article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
a substantial part of the energy carried by the solar wind can be transfered into the terrestrial magnetosphere and it is associated with the passage of southward directed interplanetary magnetic fields , bs , by the earth for sufficiently long intervals of time . discussed the energy transfer process as a conversion of the directed mechanical energy from the solar wind into magnetic energy stored in the magnetotail of earth s magnetosphere and its reconversion into thermal mechanical energy in the plasma sheet , auroral particles , ring current , and joule heating of the ionosphere .
the increase on the solar wind pressure is responsible for the energy injections and induces global effects in the magnetosphere called geomagnetic storms .
the characteristic signature of geomagnetic storms can be described as a depression on the horizontal component of the earth s magnetic field measured at low and middle latitude ground stations .
the decrease in the magnetic horizontal field component is due to an enhancement of the trapped magnetospheric particle population , consequently an enhanced ring of current .
this perturbation of the h - component could last from several hours to several days ( as described by * ? ? ?
the geomagnetic storms can consist of four phases : sudden commencement , initial phase , main phase and recovery phase .
the sudden commencement when it exists , corresponds to the moment when the initial impact of the increased solar wind pressure over the magnetopause occurs .
the initial phase at ground appears as a rapid increase on the h - component over less than 1 h almost simultaneously worldwide .
the main phase of the geomagnetic storm lasts a few hours and is characterized by a decrease in the h - component .
the recovery time corresponds to the gradual increase of the h - component value to its average level .
a detailed description of the morphology of magnetic storms is , for instance , in @xcite .
the intensity of the geomagnetic disturbance in each day is described by indices .
the indices are very useful to provide the global diagnostic of the degree of disturbance level .
there are different indices that can be used depending on the character and the latitude influences in focus . considering only the main latitudinal contributions
, the ring current dominates at low and middle latitudes and an auroral ionospheric current systems dominates at higher latitudes @xcite .
kp , ae and dst and their derivations are the most used geomagnetic indices
. the kp index is obtained from the h - component and it is divided in ten levels from 0 to 9 corresponding to the mean value of the disturbance levels within 3-h intervals observed at 13 subauroral magnetic stations ( see * ? ? ?
. however , the k index is the most difficult to be physically interpreted due to its variations be caused by any geophysical current system including magnetopause currents , field - aligned currents , and the auroral electrojets @xcite .
the minutely ae index ( sometimes @xmath2 minute interval ) is also obtained by the h - component measured from magnetic stations ( 5 to 11 in number ) located at auroral zones and widely distributed in longitude .
the ae index provides a measure of the overall horizontal auroral oval current strength .
the index most used in low and mid - latitudes is the dst index .
it represents the variations of the h - component due to changes of the ring current and is calculated every hour .
the dst index is described as a measure of the worldwide derivation of the h - component at mid - latitude ground stations from their quiet days values . at mid - latitude ,
the h - component is a function of the magnetopause currents , the ring current and tail currents .
@xcite calculated the dst index as a average of the records from @xmath3 mid - latitude magnetic stations following , @xmath4 where @xmath5 is a local time h average , @xmath6 is the h - component measured at disturbed days and @xmath7 , on quiet days .
other contributions beyond the ring current could be extracted or eliminated with the idea presented by @xcite .
those authors described the evolution of the ring current by a simple first order differential equation , @xmath8 where @xmath9 .
the contribution of the magnetopause currents to @xmath10 is proportional to the square root of the solar wind dynamic pressure ( @xmath11 ) , @xmath12 represents the injection of particles to the ring current , @xmath13 represents the loss of particles with an e - folding time @xmath14 and the constant terms @xmath15 , @xmath16 and @xmath17 are determine by the quiet days values of the magnetopause and ring currents .
the dst index is available on the kyoto world data center at http:// wdc.kugi.kyoto-u.ac.jp/dstdir/index.html .
it is traditionally calculated from four magnetic observatories : hermanus , kakioka , honolulu , and san juan .
these observatories are located at latitudes below @xmath18 which are sufficiently distant from the auroral electrojets .
the derivation of the dst index corresponds to three main steps : the removal of the secular variation , the elimination of the sq variation and the calculation of the hourly equatorial dst index ( see http://wdc.kugi.kyoto-u.ac.jp/dstdir/dst2/ondstindex.html ) .
the traditional method of calculating the baseline for the quiet day variations uses the five quietest day for each month for each magnetic observatory . in this work
, we propose a way to deal with sq variations by suggesting a method using principal component with the wavelet correlation matrix .
this method eliminates the disturbed days using a multiscale process . also , we developed an algorithm for extracting the solar quiet variations recorded in the magnetic stations time series , in order words , a way of estimation of the quiet - time baseline . to accomplish this task , we separate the solar diurnal variations using hourly data of the h - component using the technique ( described in * ? ? ?
afterward we applied the principal component wavelet analysis to identify the global patterns in the solar diurnal variations .
the rest of the paper is organized as follows : section [ the dst index calculation procedure ] is devoted to explain the main issues of the dst index calculation procedure . in section [ magnetic data ] the analyzed period and data are presented .
section [ methodology ] describes the principal component analysis and it is devoted to introduce the suggested method of principal component analysis ( pca ) using gapped wavelet transform and wavelet correlation .
it also establishes the identification of the disturbed days .
the results are discussed in section [ results and discussion ] , and finally , section [ summary ] brings the conclusions of this work .
recently , reconstructing the dst and removing the quiet - time baseline have been a motivation of several works @xcite . despite of these disagreements in the dst constrution ,
today , the dst index remains an important tool in the space weather analysis . reconstructed the dst index ( dxt ) following the original formula presented at http://wdc.kugi.kyoto-u.ac.jp /dstdir
/ dst2/ondstindex.html .
however , they encountered a few issues as : the availability and the data quality , some shifts in the baseline level of the h - component , erroneous data points and some data gaps at all / some magnetic observatories .
the dst could not be fully reproduced using the original formula because the inadequate information above effects the treatment of the related issues and therefore remains partly unscientific as described by .
a new corrected and extended version ( dcx ) of the dxt index was proposed by .
they corrected the dst index for the excessive seasonal varying quiet - time level which was unrelated to magnetic storms as previously discussed in .
they also showed that the considerable amount of quiet - time variation is included in the dxt index but none in the dcx index .
another issue related to the derivation of the dst index is that no treatment is made to normalize the different latitudinal location of the magnetic observatories .
suggested the normalization of the magnetic disturbances at the four dst stations with different latitudes by the cosine of the geomagnetic latitude of the respective station .
if no correction is made , they showed for the lowest geomagnetic station , honolulu , the largest deviations and for the highest station , hermanus , the lowest deviations of the four station .
the standard deviations reflect the annually averaged effect of the ( mainly ring current related ) disturbances at each station , ( see * ? ? ?
* for more details ) .
evaluated the effect of using more than four magnetic stations and shorter time intervals to calculate the dst index .
the obtained dst index profiles using 12 , 6 or 4 magnetic station did not show significant discrepancies and the best agreement with the standart dst was obtained using magnetic stations located at latitudes lower than @xmath19 in both hemispheres .
although , the increase of symmetrically world - wide distributed magnetic stations did not effect significantly the dst index , the longitudinal asymmetries of the ring current contributes for the average disturbances of the four dst stations be systematically different . using an extended network of 17 stations , demonstrated that the local disturbances are ordered according to the station s geographic longitude , where the westernmost station ( honolulu ) presented the largest disturbances and contributions to dst index and the easternmost ( kakioka ) the smallest
. studied the characteristics of the sq variations at a brazilian station and compared to the features from other magnetic stations to better understand the dynamics of the diurnal variations involved in the monitoring of the earth s magnetic field .
they used gapped wavelet analysis and the wavelet cross - correlation technique to verify the latitudinal and longitudinal dependence of the diurnal variations .
as previously mentioned by , also verified that magnetic stations located at lower latitudes and westernmost ( honolulu and san juan ) presented larger correlation to vassouras than the easternmost stations ( as kakioka ) .
some important aspects for the construction of the dst index as described by are : the utilization of the original data , the inspection in time and frequency domains ( removal of diurnal variation ) and the consideration of the distinction between stationary and non - stationary time series ingredients which applies to the secular variation .
also as mentioned by , some patterns of the global magnetic disturbance field are well understood and some are not which means that there is still a lot to learn about the magnetosphere , magnetic storm and earth - sun relationship .
in this paper , we use ground magnetic measurements to estimate the quiet - time baseline .
we select the four magnetic observatories used to calculate the dst index : hermanus ( her ) , kakioka ( kak ) , honolulu ( hon ) , and san juan ( sjg ) , plus other @xmath20 different magnetic observatories reasonably homogeneously distributed world wide .
one of these nine chosen stations is vassouras ( vss ) located under the south atlantic magnetic anomaly ( minimum of the geomagnetic field intensity ) .
the geomagnetic data use in this work relied on data collections provided by the intermagnet programme ( http://www.intermagnet.org ) .
the distribution of the magnetic stations , with their iaga codes , is given in fig .
[ fig : mapstations ] . the corresponding codes and locations
are given in table [ table : abbcode ] .
these selection of magnetic stations correspond to the same selection used in a previous work o@xcite for the same reasons ( exclusion of the major influence of the auroral and equatorial electrojets ) . in this work ,
we only use the data interval corresponding to the year 2007 .
we also apply the same methodology to identify geomagnetically quiet days used by .
we consider quiet days , only those days in which the kp index is not higher than 3 + . as at low latitudes the horizontal component ( h )
is mostly affected by the intensity of the ring current , we decided to use only the hourly mean value series of this component .
the magnetic stations present available data in cartesian components ( xyz system ) .
the conversion to horizontal - polar components ( hdz system ) is very simple ( see * ? ? ?
* for more details ) .
the system s conversion was performed in all the chosen magnetic stations .
.intermagnet network of geomagnetic stations used in this study .
[ cols="^,^,^,^,^ " , ] + source : http://wdc.kugi.kyoto-u.ac.jp/igrf/gggm/index.html ( 2010 ) [ table : abbcode ]
the method used in this study is based on the principal component analysis ( pca ) using gapped wavelet transform and wavelet correlation to characterize the global diurnal variation behavior . to identify periods of magnetic disturbance
, we use the discrete wavelet transform .
this technique is employed to analyze the removal of disturbed days from the magnetograms , and consequently , from the reconstructed sq signal . also in this section , a combined methodology using the pca and gapped wavelet transform is briefly described .
following , we present an identification method to distinguish the disturbed days using discrete wavelet coefficients . among the several available methods of analysis , pca is a particularly useful tool in studying large quantities of multi - variate data .
pca is used to decompose a time - series into its orthogonal component modes , the first of which can be used to describe the dominant patterns of variance in the time series @xcite .
the pca is able also to reduce the original data set of two or more observed variables by identifying the significant information from the data .
principal components ( pcs ) are derived as the eigenvectors of the correlation matrix between the variables .
their forms depend directly on the interrelationships existing within the data itself .
the first pc is a linear combination of the original variables , which when used as a linear predictor of these variables , explains the largest fraction of the total variance .
the second , third pc , etc .
, explain the largest parts of the remaining variance @xcite .
as explained by , the interpretation of the eigenvectors and the eigenvalues can be described as follow , the eigenvectors are the normalized orthogonal basis in phase space , and also , the set of vectors of the new coordinate system in the space , different from the coordinate system of the original variables ; the eigenvalues are the corresponding variance of the distribution of the projections in the new basis . in order to isolate the global contributions of each pcs mode , we applied pca using the wavelet correlation matrix computed by gapped wavelet transform .
this wavelet correlation matrix was introduced in @xcite .
we joined the properties of the pca , which are the compression of large databases and the simplification by the pcs modes , and properties of the wavelet correlation matrix , which is the correlation at a given scale , @xmath15 , in this case , the scale corresponded to the pseudo - period of @xmath1 hours .
the wavelet analysis has the following propriety : the larger amplitudes of the wavelet coefficients are associated with locally abrupt signal changes or `` details '' of higher frequency . in the work of and the following work of @xcite , a method for the detection of the transition region and the exactly location of this discontinuities due to geomagnetic storms was implemented . in these cases ,
the highest amplitudes of the wavelet coefficients indicate the singularities on the geomagnetic signal in association with the disturbed periods . on the other hand ,
when the magnetosphere is under quiet conditions for the geomagnetic signal , the wavelet coefficients show very small amplitudes . in this work
, we applied this methodology with daubechies orthogonal wavelet function of order 2 on the one minute time resolution with the pseudo - periods of the first three levels of 3 , 6 and 12 minutes .
in this section , we will present the results of reconstructed baseline for the global quiet days variation using pca technique implemented with gapped wavelet transform and wavelet correlation .
also , we will apply dwt to evaluate the day - by - day level of geomagnetic disturbance using kak magnetic station as reference . , and at bottom right , the global wavelet spectrum.,title="fig:",width=529 ] + fig .
[ fig : kakescalogram ] shows an example of the geomagnetic behavior presented at the june , 2007 magnetogram of kakioka using continuous gapped wavelet transform ( gwt ) .
the gwt can be used in the analysis of non - stationary signal to obtain information on the frequency or scale variations and to detect its structures localization in time and/or in space ( see * ? ? ? * for more details ) .
it is possible to analyze a signal in a time - scale plane , called so the wavelet scalogram . in analogy with the fourier analysis ,
the square modulus of the wavelet coefficient , @xmath21 , is used to provide the energy distribution in the time - scale plane . in the gwt analysis
, we can also explore the central frequencies of the time series through the global wavelet spectrum which is the variance average at each scale over the whole time series , to compare the spectral power at different scales .
this figure shows the h - component ( top ) , the wavelet square modulus ( bottom left ) and the global wavelet spectrum ( total energy in each scale bottom right ) . in the scalogram ,
areas of stronger wavelet power are shown in dark red on a plot of time ( horizontally ) and time scale ( vertically ) .
the areas of low wavelet power are shown in dark blue . in fig .
[ fig : kakescalogram ] , it is possible to notice peaks of wavelet power on the scalogram at the time scale corresponding to @xmath22 to @xmath23 minutes of period .
this periods are associated to pc5 pulsations during disturbed periods . also , it is possible to notice a maximum of wavelet power at the time scale corresponding to harmonic periods of the 24 hours such as 6 , 8 , 12 hours .
those periods are related to the diurnal variations .
the gwt technique is able to analyze all the informations present on the magnetograms .
it is an auxiliary tool to localize on time / space the pc1pc5 pulsations @xcite .
however , the scalogram provides a very redundancy information which difficult the analysis of each decoupling phenomenum .
for that reason , we preferred to use dwt to evaluate the day - by - day level of geomagnetic disturbance .
geomagnetically quietest days ( blue ) and @xmath24 most disturbed days ( red ) , the h - component average variation for kak obtained at june , 2007 and the square root wavelet coefficients amplitudes @xmath25 , @xmath26 and @xmath27 with the pseudo - periods of 3 , 6 and 12 minutes.,title="fig:",width=529 ] + fig .
[ fig : sqjun ] is composed of two graphs , the first one presents the reconstructed line of the sq variation where the first @xmath28 geomagnetically quietest days of each month are highlighted in blue and the @xmath24 most disturbed days in red and the second one presents the discrete wavelet analysis of the geomagnetic horizontal component obtained at kakioka station , japan .
using our criteria of removing disturbed days , we consider as gaps of 3rd , 14th , 21th and 29th day . in our case ,
the gapped wavelet technique is very helpful because it reduces two effects : the presence of gaps and the boundary effects due to the finite length of the data , for more details see .
the first graph shows the amplitude range between @xmath29 and @xmath24 nt and presents a complex pattern .
it is possible to notice that the larger amplitudes of the reconstructed sq signal correspond to the periods between the days 810 , 1315 , 2123 and 2930 .
these periods correspond to the disturbed days .
table [ table : sq10calmos ] shows these @xmath28 quietest days and @xmath24 most disturbed days of each month set by the _ geoforschungszentrum ( gfz ) potsdam _ through the analysis of the kp index that are highlighted on the second graph .
the year of @xmath30 is a representative year of minimum solar activity and it is used in this analysis due to have less disturbed periods ( see our considerations in section [ magnetic data ] ) . by analyzing these highlighted days , we expect to find out if there is a correlation between the days classified as quiet days and a small sq amplitude variation .
[ table : sq10calmos ] the second graph shows the discrete wavelet analysis applied to geomagnetic minutely signal from kak using daubechies orthogonal wavelet family 2 . from top to bottom in this graph , the h - component of the geomagnetic field and the first three levels of the square wavelet coefficients denoted by d1 , d2 and d3 .
this analysis uses the methodology developed by , and posteriorly applied by @xcite and . in order to facilitate the evaluation of the quiet periods obtained by the discrete wavelet analysis applied to geomagnetic signal from kak
, we also developed a methodology ( effectiveness wavelet coefficients ( ewc ) ) to interpret the results shown in fig .
[ fig : sumsqjun ] .
the ewc corresponds to the weighted geometric mean of the square wavelet coefficients per hour .
it is accomplished by weighting the square wavelet coefficients means in each level of decomposition as following @xmath31 where @xmath3 is equal to @xmath32 because our time series has one minute resolution .
[ c][][0.8]@xmath33 [ c][][0.8]@xmath34 + through fig .
[ fig : sumsqjun ] , it is possible to compare the global sq behavior ( top graph ) with the analysis of quiet and disturbed days obtained by one representative magnetic station of medium / low latitudes ( kak bottom graph ) in order to verify situations in which the global sq behavior presents less or more variability .
this analysis allows us to validate the quietest days , and evaluate the most disturbed days in order to establish a reliable method of global sq analysis obtained from medium / low latitude magnetic stations influenced only by the ionosphere . in fig .
[ fig : sumsqjun ] , the global sq behavior ( top graph ) shows larger amplitudes during the periods between the days 810 , 1216 , 2223 and 2930 .
most of these periods correspond to the days where the ewcs have an increase of their values as shown kak analysis ( bottom graph ) .
the increase of the ewcs values occurs during the periods between the days 14 , 810 , 1318 , 2124 and 2930 .
the ewcs can help us also to interpret the results obtained in each day , and can help us to evaluate the quietest and most disturbed days measure by the selected magnetic station of medium / low latitudes , kak , as shown in table [ table : sq10calmosdwt ] .
[ table : sq10calmosdwt ] the same methodology and analysis comparing the global sq behavior and the ewcs from kak is done for the month of march , september and december,2007 .
[ fig : sumsqmar ] shows the comparative of the global sq behavior and the ewcs from kak for the month of march , 2007 .
the amplitude range of global sq signal is between @xmath35 and @xmath36 nt , and , it also shows a complex pattern .
it is possible to notice that the larger amplitudes of the reconstructed sq signal correspond to the periods between the days 67 , 1117 and 2328 .
once more , these periods correspond to the disturbed days .
the increase of the ewcs values occurs during the periods between the days 1 , 67 , 1117 , 2328 and 3031 .
we can notice that the increase of the amplitude of the reconstructed sq signal correspond to the increase of the ewcs magnitude .
[ c][][0.8]@xmath33 [ c][][0.8]@xmath34 + the amplitude range for september , 2007 , is between @xmath35 and @xmath36 nt and the larger amplitudes of the reconstructed sq signal correspond to the periods between the days 17 , 1923 and 2630 , see fig , [ fig : sumsqset ] .
the increase of the ewcs values occurs during the periods between the days 13 , 2024 and 2730 . comparing these two analysis , the global and the ewcs
, we verify that the kak magnetic behavior represents well the increase of global sq oscillations .
[ c][][0.8]@xmath33 [ c][][0.8]@xmath34 + in fig .
[ fig : sumsqdec ] , the amplitude range is between @xmath37 and @xmath28 nt and the larger amplitudes correspond to the periods between the days 1523 .
also , the increase of the ewcs values occur during the periods between the days 1011 and 1721 .
once more , when we compare these two analysis , the global and the ewcs , we verify that the kak magnetic behavior represents well the increase of global sq oscillations .
[ c][][0.8]@xmath33 [ c][][0.8]@xmath34 + we observe in figs .
[ fig : sumsqjun ] , [ fig : sumsqmar ] , [ fig : sumsqset ] and [ fig : sumsqdec ] , in most of the cases , the major amplitude fluctuations of the reconstructed sq signal correspond to the most disturbed days and minor fluctuations , to the quietest days .
however , the diurnal global variability shows a complexity on the amplitude variation pattern even during geomagnetically quiet periods . through this study
we compare the amplitude variation of the reconstructed sq signal to effectiveness wavelet coefficients obtained at kak with the purpose of understanding the complexity of the diurnal global variability .
in this work , we suggest an alternative approach for the calculation of the sq baseline using wavelet and pca techniques .
this new approach address some issues , such as , the availability and the quality of data , abrupt changes in the level of the h - component , erroneous points in the database and the presence of gaps in almost all the magnetic observatories . to fulfill this purpose ,
we reconstruct the sq baseline using the wavelet correlation matrix with scale of @xmath1 hours ( pseudo - period ) .
the pca / wavelet method uses the global variation of first pca mode that also corresponds to phenomena with periods of @xmath1 hours .
this study shows that the largest amplitude oscillation of the reconstructed signal ( sq baseline ) corresponded to the most disturbed days and the smaller oscillations to the quietest days .
this result is consistent with the expected sq variations .
v. klausner wishes to thanks capes for the financial support of her phd ( capes grants 465/2008 ) and her postdoctoral research ( fapesp 2011/20588 - 7 ) .
this work was supported by cnpq ( grants 309017/2007 - 6 , 486165/2006 - 0 , 308680/2007 - 3 , 478707/2003 , 477819/2003 - 6 , 382465/01 - 6 ) , fapesp ( grants 2007/07723 - 7 ) and capes ( grants 86/2010 - 29 , 0880/08 - 6 , 86/2010 - 29 , 551006/2011 - 0 , 17002/2012 - 8 ) .
also , the authors would like to thank the intermagnet programme for the datasets used in this work .
gonzalez , w. d. , joselyn , j. a. , kamide , y. , kroehl , h. w. , rostoker , g. , tsurutani , b. t. , vasyliunas , v .
m. , 1994 .
what is a geomagnetic storm ?
journal of geophysical research , 99 ( a4 ) , pp .
57715792 .
kamide , y. , baumjohann , w. , daglis , i. a. , gonzalez , w. d. , grande , m. , joselyn , j. a. , mcpherron , r. l. , phillips , j. l. , reeves , e. g. d. , rostoker , g. , sharma , a. s. , singer , h. j. , tsurutani , b. t. , and vasyliunas , v. m. , 1998 .
current understanding of magnetic storms : storm - substorm relationships .
journal of geophysical research , 103 ( a8 ) , pp .
1770517713 . klausner , v. , papa , a. r. r. , mendes , o. , domingues , m. o. , frick , p. , 2011a
. characteristics of solar diurnal variations : a case study based on records from the ground magnetic observatory at vassouras , brazil . arxiv:1108.4594 [ physics.space-ph ] .
mendes , o. j. , domingues , m. o. , mendes da costa , a. , cla de gonzalez , a. l. , 2005a .
wavelet analysis applied to magnetograms : singularity detections related to geomagnetic storms .
journal of atmospheric and solar - terrestrial physics , 67 , pp .
18271836 .
mendes , o. j. , mendes da costa , a. , domingues , m. o. , 2005b .
introduction to planetary electrodynamics : a view of electric fields , currents and related magnetic fields .
advances in space research , 35 ( 5 ) , pp .
812 - 818 . ,
a. , domingues , m. o. , mendes , o. , brum , c. g. m. , 2011 .
interplanetary medium condition effects in the south atlantic magnetic anomaly : a case study .
journal of atmospheric and solar - terrestrial physics , 73 ( 1112 ) , pp .
14781491 .
murray , g. w. ; mueller , j. c. ; zwally , h. j. , 1984 matrix partitioning and eof / principal components analysis of antartic sea ice brightness temperatures .
greenbelt , md . : national aeronautics and space administration , goddard space flight center , 1 v. , nasa technical memorandum , 83916 .
nesme - ribes , e. , frick , p. , sokoloff , d. , zakharov , v. , ribes , j. c. , vigouroux , a. , laclare , f. , 1995 .
wavelet analysis of the maunder minimum as recorded in solar diameter data .
comptes rendus de lacadmie des sciences .
srie ii , mcanique , physique , chimie , astronomie 321 ( 12 ) , 525532 . | this paper describes a methodology ( or treatment ) to establish a representative signal of the global magnetic diurnal variation based on a spatial distribution in both longitude and latitude of a set of magnetic stations as well as their magnetic behavior on a time basis . for that
, we apply the principal component analysis ( pca ) technique implemented using gapped wavelet transform and wavelet correlation .
the continuous gapped wavelet and the wavelet correlation techniques were used to describe the features of the magnetic variations at vassouras ( brazil ) and other @xmath0 magnetic stations spread around the terrestrial globe . the aim of this paper is to reconstruct the original geomagnetic data series of the h - component taking into account only the diurnal variations with periods of @xmath1 hours on geomagnetically quiet days . with the developed work ,
we advance a proposal to reconstruct the baseline for the quiet day variations ( sq ) from the pca using the correlation wavelet method to determine the global variation of pca first mode .
the results showed that this goal was reached and encourage other uses of this approach to different kinds of analysis .
magnetogram data , h - component , quiet days , principal component wavelet analysis . |
sodium nitrite is a ferroelectric at room temperature . it has the orthorhombic structure , space group @xmath2 , with the dipole vector of the v - shaped nitrite anions aligned parallel to the crystallographic @xmath3 direction , as shown in fig .
[ fig : nano2 ] . crystal structure of nano@xmath0 in the ferroelectric phase . ]
the ferroelectric - paraelectric phase transition takes place at about @xmath4 k , where the high temperature phase is orthorhombic , space group @xmath5 , with the dipoles disordered with respect to the @xmath3 axis . in a narrow temperature range from @xmath6 k to @xmath4 k
, there exists an incommensurate antiferroelectric phase .
the melting temperature is @xmath7 k. distinguished from displacive ferroelectrics in which the ferroelectric transition is driven by soft phonon modes , nano@xmath0 offers a model system for research of the order - disorder structural phase transition and any associated ferroelectric instability .
@xcite extensive experimental work on nano@xmath0 has been devoted to probing the mechanism of the no@xmath8 polarization reversal that triggers the order - disorder transition .
the majority of studies support the @xmath1-axis rotation model , but there were also results favoring the @xmath9-axis rotation model.@xcite recently , refined x - ray studies over a wide temperature range reinforced the @xmath1-axis rotation model.@xcite on the theoretical side , the microscopic model calculations done by ehrhardt and michel supported the @xmath1-axis rotation mechanism,@xcite whereas mixed double rotations around the @xmath9-axis and the @xmath1-axis was suggested by kinase and takahashi.@xcite it has long been desirable to apply computer molecular dynamics ( md ) simulations to nano@xmath0 in order to achieve unambiguous understanding of the polarization reversal mechanism .
earlier md simulations with empirical born - mayer pair potentials detected the @xmath1-axis rotation in above - room - temperature nano@xmath0.@xcite unfortunately , the low - temperature structure produced by those simulations was antiferroelectric and apparently disagreed with the experimental observations .
lu and hardy pointed out that the overall phase behavior of nano@xmath0 could be simulated by using an _ a priori _ approach to construct the crystal potential surface ( pes).@xcite the lu - hardy ( lh ) approach was originally designed to deal with molecular crystals such as k@xmath0seo@xmath10 , where exists a mix of bonding types , that is , the intermolecular interactions are mostly ionic , but the constituent atoms in a molecule ( seo@xmath11 in k@xmath0seo@xmath10 ) bond covalently . in the lh approach ,
the intra - molecule interactions were treated by applying the _ ab initio _ self - consistent field method to the gas - phase molecules , while the intermolecular pair potentials were computed within the gordon - kim ( gk ) electron gas theory.@xcite the crux of their application of the gk theory is how to partition the _ ab initio _ molecular charge density between the constituent atoms .
since there is no unique way to separate the charge density of a highly covalently bonded molecule , lu and hardy suggested equal separation in a spirit similar to the mulliken population analysis ( mpa ) . by using this atomic - level method , we could successfully describe the phase transitions in fluoroperovskites,@xcite and ionic crystals with polyatomic molecules including seo@xmath11,@xcite clo@xmath12,@xcite so@xmath11,@xcite sio@xmath13,@xcite and no@xmath14.@xcite note that the mpa happens to preserve the ( zero ) dipole moment of these molecules .
however , several problems appear when we moved on to deal with nano@xmath0 where the no@xmath15 radical has nonzero dipole moment and stronger chemical bonding .
first , it is well known that the mpa , while certainly the most widely employed , is also somewhat arbitrary and the most criticized.@xcite in particular , the mpa overestimates the dipole moment of the free no@xmath8 ion by about @xmath16 .
other difficulties involved the free - ion approximation . unlike in monatomic ionic crystals , there may exist considerable _ internal _ charge - transfer effects in molecular ionic crystals .
electronic band structure calculations @xcite indicated that within a nitrite entity , the nitrogen atom and two oxygen atoms bond covalently , leading to high charge transferability between these constituent atoms
. therefore , in solid nano@xmath0 the no@xmath8 group will feel different crystal - field environments as it rotates and responds by redistributing the charge density among its three constituent atoms .
our goals in this paper are twofold .
first , we show that our atomistic level simulation methods involving pair potentials with the rigid - ion model is capable of correctly describing the phase behavior of nano@xmath0 .
second , we systematically examine the lh approach to understand the reason why it works so well in molecular ionic crystal systems by the following steps : ( i ) we develop another population analysis method that preserves the molecular dipole moment by directly fitting the _ ab initio _
charge density of a molecule ; ( ii ) we carry out _ ab initio _ hartree - fock ( hf ) calculations of the intermolecular interactions and find that the pair potentials from the rigid - ion model can correctly reproduce the _ ab initio _ results ; ( iii ) we investigate the crystal - field effects on the no@xmath8 ion by embedding the ion ( and its first shell of neighbors ) in a lattice of point charges and find a remarkable internal charge - transfer effect .
@xcite several md simulations based on these modifications of the lh approach are also performed .
the ferroelectric - paraelectric transition triggered by the @xmath1-axis rotation of the nitrite ions is observed in all versions of the lh approach .
however , the transition temperatures predicted by these simulations are lower than the experimental values .
furthermore , the transition temperatures obtained from the original version are higher than those predicted by modified versions and closer to the experimental values . after careful examination
, we notice that in the lh approach , the no@xmath17 dipole moments were generally overestimated by about 120% , which reinforces the ground state .
this implies that the crystal structure of nano@xmath0 is stabilized by the anion polarization effects .
thus , we conclude polarization effects are particularly important for the quantitative study of nano@xmath0 . this paper is organized as follows .
section ii describes the method we used to obtain the intermolecular interactions .
section iii analyzes the resulting intermolecular potentials in comparison with those obtained from _ ab initio _ calculations .
section iv presents the results of our md simulations .
the crystal - field effects on no@xmath17 are discussed in section v. concluding remarks are made in section vi .
our md simulation technique method originates from the gk model for simple ionic crystals such as alkali halides , assuming that ( molecular ) ions in a crystal environment may be described as free ions .
@xcite then it was extended to deal with molecular ionic crystals like k@xmath0seo@xmath10 , in which strong intramolecular covalency exists .
@xcite the main idea is that the molecular ion ( seo@xmath18 in k@xmath0seo@xmath10 ) is treated as a single entity , and intramolecular and intermolecular interactions are treated separately : first we perform _ ab initio _ quantum chemistry calculations for the whole molecular ion , to obtain the optimized structure , the force constants , and the whole electron density @xmath19 .
the intramolecular interactions are described by force constants within the harmonic approximation . as for the intermolecular interactions
, we have to carry out electron population analysis to separate @xmath19 onto each individual atom in the molecular ion , then use the gordon - kim electron gas model to calculate the intermolecular pair potentials .
this approach provides a parameter - free description for the crystal potential - energy surfaces , which allow structural relaxation , md simulation , and lattice dynamics calculations . in calculating the intermolecular forces
, there are three major approximations as discussed in the following : \(1 ) we assume that the geometries and electronic densities of the separate ions remain unchanged once they have been obtained under given circumstance , such as in the equilibrium state of the gas or crystal phases .
this approximation is the fundamental basis for the gk electron gas theory .
general speaking , we found that in an ionic crystal there is no strong chemical bond between ions , hence this approximation is reasonable .
\(2 ) when dealing with the intermolecular interaction , we assume that the charge density of a rigid ion can be separated into its atomic constituents .
\(3 ) we assume that the crystal potential energy is composed of the intermolecular and intramolecular interaction , where the intramolecular interaction is expressed in terms of force fields and the intermolecular interaction is a sum of interatomic pair potentials .
atomistic level simulations utilizing pair potentials and the rigid - ion model have great success in describing many ionic systems .
@xcite we showed that this scheme can correctly describe the phase transition behaviors of alkali halide fluoperovskites , @xcite and molecular crystals with tetrahedral @xcite and equilateral triangular @xcite radicals .
however , for nano@xmath0 in which no@xmath17 has only a two fold symmetrical axis , the results were less satisfactory .
@xcite note that the mean hf polarizability of no@xmath8 , 14.156 ( atomic units ) , calculated with the d95 * basis , @xcite is much higher than that of na@xmath20 , 0.343 ( atomic units ) , calculated with the 6 - 31 * basis .
therefore , in solid nano@xmath0 , the rotation of no@xmath8 in the crystal field will induce charge redistribution within the molecule .
hence this dynamic effect may invalidate the rigid ion approximation . in this paper
we shall perform hf calculations for various geometries to verify this scenario . in the gk model
, we evaluate the interaction between two molecules based on the electron density , @xcite which is approximated as the sum of component densities taken from hf calculations .
that is , if @xmath21 and @xmath22 are the component densities , then the total density is @xmath23 . whereas , interaction potential is computed as the sum of four terms : coulombic , kinetic , exchange , and correlation energies which are expressed in terms of the charge densities . therefore , suppose the @xmath24 and @xmath25 molecules are made up of @xmath26 and @xmath27 atoms , respectively , then the coulombic interaction between them is @xmath28 where @xmath29 , @xmath30 , @xmath31 and @xmath32 are the nuclear charges and coordinations of the @xmath33-th atom in the @xmath24 molecule and @xmath34-th atom in the @xmath25 molecule , respectively
this potential energy can be split into two parts : first the long - range part , @xmath35[z_{b , j}-\int \rho _ b^{}(\mathbf{r}_2)d\mathbf{r}_2]}{|\mathbf{r}_{a , i}-\mathbf{r}_{b , j}| } , \label{cl}\ ] ] and the short - range part @xmath36 eq .
( [ cl ] ) is essentially the electrostatic interaction energy when the charge densities of the molecules are distributed as point charges on the constituent atoms , which is known as the madelung potential energy .
the non - coulombic energy terms are expressed in the uniform electron gas formula , @xmath37 , \label{kxc}\ ] ] where @xmath38 is one of the energy functionals for the kinetic , exchange , and correlation interactions .
@xcite note that eq .
( [ kxc ] ) is not composed of pair potentials . in order to obtain the effective pairwise potentials , we approximate eq .
( [ kxc ] ) using @xmath39 , \label{pair}\ ] ] where @xmath40 . @xmath41 and @xmath42 are the charge densities of individual atoms in the @xmath24 and @xmath25 molecules , respectively , which are obtained by a population analysis as described in the next subsection . even though the non - coulombic forces as determined by eq .
( [ kxc ] ) are not strictly additive , the above approximation appears to be adequate except at very short distances .
as pointed out by waldman and gordon , @xcite the main reason as to why this approximation is valid is because the coulombic force , the largest contribution to the potentials , is additive . based on our calculations , we find additivity of @xmath43 holds only to within about 50% ; however , the overlap contribution to the electrostatic energy dominates @xmath43 and renders additivity to within 10% .
one final remark is in order , for the sake of simplifying the two - electron integral in eq .
( [ coul ] ) , the charge densities , @xmath41 and @xmath42 , are taken as its spherical average . as a result ,
the coulombic interaction is not exactly evaluated . nevertheless , as we shall show in figs .
[ fig : na - no2 ] and [ fig : no2-no2 ] , this error is compensated by those due to the pairwise additive approximation . to summarize this subsection , we demonstrated that it is possible to analytically express the intermolecular potentials @xmath44 using eqs .
( [ cl ] ) , ( [ cs ] ) and ( [ pair ] ) once the charge density of each individual atom is obtained by an electronic population analysis . in the next subsection
, we shall present further analysis on the charge density . in this subsection , we discuss the ways to separate the electron density @xmath19 of a molecule into its atomic constituents .
suppose the molecule consists of @xmath26 atoms , then the wave function of the molecule @xmath45 can be written as a linear superposition of atomic wave functions @xmath46 , @xmath47 , centered at each atom , @xmath48 in turn , the atomic wave functions @xmath46 can be written as a linear superposition of the basis functions @xmath49
@xmath50 where \{@xmath51 } are usually the gaussian basis functions , and the coefficients @xmath52 can be obtained from the variational method .
then the electronic density of the molecule is , @xmath53 where @xmath54 , which can be divided into two parts , namely the _ net _ ( @xmath55 ) and _ overlap _ ( @xmath56 ) populations .
the latter can not be ignored in the presence of strong intramolecular covalency .
therefore , separating @xmath19 into its atomic constituents is to split the overlap population .
however , the way to achieve that is not unique .
for example , we can introduce a set of weights @xmath57 due to different criteria such that @xmath58 then we can rewrite eq .
( [ density ] ) as following @xmath59 where @xmath60 is the atomic density of atom @xmath33 .
.[table : multipole ] electronic multipole moments of molecule ( @xmath61 ) calculated from the mulliken population analysis . the _ ab initio _ values are shown in parentheses .
all quantities are in atomic units . [ cols="<,^,^,^,^",options="header " , ] before we proceed with the molecular dynamics simulations , we perform lattice relaxation for the ferroelectric structure of nano@xmath0 both with and without the @xmath62 space group symmetry constraints .
this relaxation procedure provides the crystal structure with zero force on each atom , that is an energy extremum ; it also produces a test to the pes because the resultant structures have to agree reasonably with the experimental data for further simulations to be reliable .
we perform both static and dynamic relaxations : the static one is an application of the newton - raphson algorithm and usually results in finding a local minimum of the energy , and the dynamic one is a simulated annealing calculation for overcoming that limitation .
we start the static lattice relaxation with the experimental parameters . in table
[ table : par ] we present the lattice and basis parameters deduced from the experiments and static relaxation . in all cases , the static relaxation produced essentially the same structure that strongly resembles the experimental structure .
most of the lattice constants in the relaxed structure are shorter than the experimental values ( by 3.7% , 1.5% , and 8.5% for @xmath9 , @xmath3 , and @xmath1 , respectively , in model i , and by 0.5% , -2.4% , and 10% for @xmath9 , @xmath3 , and @xmath1 , respectively , in model ii ) .
hence the calculated volume is smaller than the experimental one by 13% for model i and 10% for model ii , a common feature for simulations using the gk model , which will be addressed in more detail in the next subsection .
next , we go on to relax the statically relaxed crystal structure to zero temperature using a simulated annealing algorithm , in which the amount of kinetic energy in the molecules slowly decreases over the course of the simulation .
we find that the ( zero temperature ) ground states in models i and ii are close to the statically relaxed structures , whereas there are substantial changes taking place in model iii . by monitoring the orientations of the nitrite ions , we find that the ground structure in model iii , still orthorhombic with @xmath63 , @xmath64 , and @xmath65 , is ferroelectric with the dipole moments of no@xmath8 aligned along the @xmath9 axis rather than the experimental @xmath3 axis .
so we conclude that the pes given by models iii did not reflect reality .
this concurs with the previous discussion ( section iii ) on the intermolecular potentials . in the following we use only models i and ii to simulate the phase transition in nano@xmath0 .
temperature variation of lattice constants @xmath9 , @xmath3 , @xmath1 ( solid , dashed , dotted lines , respectively ; left scale ) and volume of the unit cell ( open circles ; right scale ) for ( a ) model i and ( b ) model ii . ]
mean dipole moment @xmath66 and quadrupole moment @xmath67 of the whole nano@xmath0 crystal as a function of temperature for the md runs for ( a ) model i and ( b ) model ii . ]
diagonal elements of the mean - square atomic displacements @xmath68 vs temperature .
( a ) na , ( b ) n , and ( c ) o atom . ] atomic positions of nano@xmath0 viewed from the @xmath9 direction obtained from the md simulation for model i at ( a ) @xmath69 k , ( b ) @xmath70 k , ( c ) @xmath71 k , and ( d ) @xmath72 k. ] using the isoenthalpic , isobaric ensemble , our md simulation is started with a zero - temperature zero - pressure orthorhombic cell ( @xmath73 ) , consisting of 512 atoms .
periodic boundary conditions are imposed to simulate an infinite crystal .
the md calculations are carried out in the parrinello - rahman scheme @xcite which allows both the volume and the shape of the md cell to vary with time .
the calculation of the energies and forces was handled by the ewald method .
a time step of 0.002 ps was used to integrate the equations of motion . in our heating runs , we raise the temperature of the sample in stages , 20 k each time , up to 1000 k. at each stage , the first 2000 time steps were employed to equilibrate the system , then 10000 time steps were collected for subsequent statistical analysis . since our simulations employ periodic boundary conditions , we can not distinguish the incommensurate structure ( i.e. , phase ii of solid nano@xmath0 ) . in figs .
[ fig : abc ] through [ fig : ellipsoid ] , we demonstrate that as the md cell is heated , it undergoes two phase transitions : in the first one , the system retains its orthorhombic structure with a change of space group from @xmath62 to @xmath74 , in agreement with the experiments .
the critical temperature @xmath75 is around @xmath76 k for model i and @xmath77 k for model ii , respectively . in the second transition ,
the crystal structure violently changes from orthorhombic to tetragonal at temperature ( @xmath78 ) which is around @xmath79 k for model i and @xmath80 k for model ii , as shown in fig .
[ fig : abc ] .
however , we argue that the crystal has already melted before this type of transition could be observed in reality . to investigate the mechanism of the polarization reversal of no@xmath8
, we monitor the crystal polarization and display the results in fig .
[ fig : fe ] .
let the dipole moment of anion @xmath33 be @xmath81 and the quadrupole moment be @xmath82 calculated by using the point charges on the @xmath27 and @xmath83 atoms .
then the mean dipole moment per anion at temperature @xmath84 is @xmath85 where @xmath86 is the number of no@xmath8 in the md cell and the brackets denote an average over the md run .
in addition , we define the antiferroelectric polarization as @xmath87 where @xmath88 and @xmath89 is the lattice vectors associated with the @xmath33-th ion . within our statistical uncertainty
we find over all temperature range @xmath90 , while @xmath91 and @xmath92 .
this fact confirms that the transition taking place at @xmath75 is the ferroelectric - paraelectric phase transition .
furthermore , we calculated the mean quadrupole moment @xmath93 .
when the dipole vector of a no@xmath8 is aligned along the @xmath3 axis , @xmath94 , @xmath95 , @xmath96 for model i and @xmath97 , @xmath98 , @xmath99 for model ii ; thus @xmath100 .
this relation holds as the no@xmath8 ion rotates around the @xmath1 axis ; nevertheless , one would expect @xmath101 when the no@xmath8 ion rotates without directional preference .
the fact that @xmath102 for @xmath103 ( fig .
[ fig : fe ] ) reveals that the no@xmath8 anions rotate primarily about the @xmath1 axis . when @xmath104 , @xmath105 , i.e. , nano@xmath0 becomes an orientational liquid .
further , in fig . [
fig : uu ] we show the mean - square atomic displacements @xmath106 where @xmath107 denotes the displacements along the @xmath108 axes , respectively .
different models of no@xmath8 reversal are expected to satisfy the following conditions : ( 1 ) rotation around the @xmath1 axis : @xmath109 and @xmath110 ; ( 2 ) rotation around the @xmath9 axis : @xmath111 and @xmath112 .
this figure relates to recent x - ray experiments which used the same quantities to investigate the polarization reversal mechanism .
@xcite the experiments confirmed that the first condition holds for both ferroelectric and paraelectric phases .
another important feature revealed by the experiments is that @xmath113 in the ferroelectric phase , whereas @xmath114 in the paraelectric phase .
that is , @xmath115 and @xmath116 are reversed across @xmath75 .
these features are reproduced in fig .
[ fig : uu ] with exception of @xmath112 in the paraelectric phase .
this means the no@xmath8 motions in our simulations are more mobile than those in the real crystal , rendering the simulated transition temperatures lower than the experimental values of @xmath117 k and the melting temperature @xmath7 k. in other words , the barriers to no@xmath8 rotation in our models are too small .
in addition , in fig .
[ fig : ellipsoid ] , we show the average crystal structures of nano@xmath0 at different temperatures .
the ellipsoids in these pictures represent the root - mean - square deviations of the atoms from their average positions and thus indicate the thermal motions of these atoms .
the @xmath1-axis rotation mode can be clearly seen , particularly in fig .
[ fig : ellipsoid](c ) . according to the calculations described in the previous section , model ii generally gives a better description of free no@xmath8 and the intermolecular potential energies than model i. however , the simulation based on model i matches closer the experimental results than that based on model ii ,
that is , @xmath118 and @xmath78 predicted by model i are closer to experiment , this indicates that the crystal fields and polarization effects are particularly important for quantitatively studying the nano@xmath0 system , where the ferroelectric structure is considerably stabilized by anion polarization effects .
actually , the mpa employed by model i does not preserve the _ ab initio _ dipole moment of free no@xmath8 . rather , it overestimates the dipole moment by 120% , thus leading to higher no@xmath17 rotational barriers than those predicted by model ii , which in turn raises the transition temperature and provides a better simulation in comparison with the experiments .
it is worth mentioning the less desirable agreement between theoretical and experimental volumes , namely , the 13% discrepancy for model i and 10% for model ii . to address this we make one simple change : by following waldman and gordon,@xcite we increase the kinetic energy term in the gordon - kim potentials by 5% , this reduces the discrepancy to 9% for model i and 6% for model ii .
having done this we rerun the md to obtain values of @xmath119 of 360 k for model i and 303 k for model ii . while this change worsens the value for model i , the value for model ii is virtually unchanged . and in both cases the transition mechanism is unaltered .
thus the slight hardening of the short - range potentials removes most of the volume discrepancies
. however , there is no material change in the mechanism of the phase transition .
this robustness of the results with respect to minor variations in the potential demonstrates that our basic conclusion remain valid . based on the previous simulation results , the order - disorder phase transition in nano@xmath0 involves the rotation of the nitrite ions .
we devise a scheme to calculate the three rotational barriers for no@xmath8 around the crystallographic @xmath9 , @xmath3 , and @xmath1 axes with its center of mass fixed : we start from the experimental ferroelectric structure @xcite as the ground state and calculate its energy difference with one of the two nitrite ions in the unit cell being rotated about the respective axis .
the results are shown in figs .
[ fig : barrier](a ) and [ fig : barrier](b ) .
rotational barriers of one of the two nitrite ions in the unit cell around the @xmath9 , @xmath3 , and @xmath1 axes with its center of mass fixed . ( a)(b ) for the gk model , ( c)(d ) for the point charge model . left and right panels are for model i and ii , respectively . ] the bottom of each barrier , zero angle , is in the ferroelectric structure .
for both models i and ii , the rotation around the @xmath1-axis has an energy barrier 5 - 10 times smaller than those of the other rotations , which is a characteristic of nitrites .
@xcite hence , our calculations unambiguously reveal that the reorientation of no@xmath8 in the paraelectric phase occur essentially by rotations around the @xmath1 axis .
in addition , the barriers calculated in model i are higher than that in model ii , confirming that the transition temperature in model i will be higher than in model ii . in figs . [ fig : barrier](c ) and [ fig : barrier](d ) , we also plot the contribution to the rotational barriers purely from the electrostatic interaction , that is , in the point - charge model with the short range forces deleted . comparing fig . [
fig : barrier](a ) with fig .
[ fig : barrier](c ) , or fig .
[ fig : barrier](b ) with fig .
[ fig : barrier](d ) , we notice that the point - charge model gives rise to a quite different rotation barrier about the @xmath9 axis : it bottoms at about @xmath120 and is lower than the ground state energy due to the omission of short - range interactions , which comes from overlap of the charge cloud of an atom with those of its neighbors .
in this section , we investigate the crystal - field effects on the no@xmath8 ion , which include electrostatic interaction from the background lattice , overlap compression of the no@xmath8 charge cloud through interaction with its neighbors , and charge - transfer between molecules which is usually small in ionic crystals . in the studies of monatomic ions @xcite and cyanides @xcite ,
fowler _ et al .
_ showed that these effects could be successfully described by embedding the ion of interest , or a cluster consisting of the ion and its first shell of neighbors , into a lattice of point charges .
we therefore perform hf calculations based on the following two models .
@xcite first , the crystal field of ferroelectric nano@xmath0 is simulated by placing the nitrite ion at the center of a @xmath121 orthorhombic point charge lattice with spacings equal to the experimental lattice parameters .
charges in the faces of the lattice are scaled to maintain overall neutrality .
all anions except the central no@xmath8 are represented by single point charges . hence , there are 174 point charges surrounding the no@xmath8 ion
. calculations of this type are referred to as cryst .
obviously , in the cryst calculation we take into account only the crystal - field effect arising purely from electrostatic interaction . at the next level of sophistication ,
we replace the six nearest positive charges of the central no@xmath8 in the above lattice by the na cations .
calculations of this type are referred to as clust . in both cryst and clust calculations , the geometrical structure of the no@xmath17
is fixed at its experimental values .
we employ the same basis set , d95 * , for the in - crystal no@xmath17 ion as for the free no@xmath8 ion . in order to keep the clust calculations to a manageable size
, we use the minimal basis set , sto-3 g , for the na@xmath20 ions unless specified .
the cations , however , are relatively insensitive to the crystal environment and they are included here only to account for their compressing effect on the no@xmath8 wave functions .
we find that adding extra basis functions to na@xmath20 will not change the results significantly .
the in - crystal no@xmath8 initially points in the @xmath3 direction as in the ferroelectric phase of nano@xmath0 @xcite .
its dipole moment is 0.636 ( cryst ) and 0.661 ( clust ) debye , close to that in the free ion model , 0.661 debye .
thus it appears that the crystal - field effect is small for this configuration .
subsequently , we rotate the no@xmath8 about the @xmath9 , @xmath3 , and @xmath1 axes and calculate the dipole moment of the rotated no@xmath8 . the rotation center is taken to be the center of a ( na@xmath20)@xmath122 cage formed by the 6 neighboring sodium ions of the central no@xmath15,@xcite ( 0,0.279 , 0 ) in the coordinate convention of fig .
[ fig : nano2 ] . the clust and cryst results are depicted in figs .
[ fig : polarizable](a ) and [ fig : polarizable](b ) , respectively .
dipole moment of the central no@xmath8 in a @xmath121 lattice as it rotates around the @xmath108 axes through the center of its ( na@xmath20)@xmath122 cage .
( a ) clust and ( b ) cryst . ] clearly , the dipole moment of the no@xmath8 changes considerably as it rotates , indicating strong crystal field effects on the reorientation of the no@xmath8 . since the electron density of the no@xmath8 is compressed by its 6 neighboring sodium cations , the variation in magnitude of its dipole moment is smaller in the clust model than that in the cryst model .
the electron cloud of the no@xmath8 is most and least variable when it rotates around the @xmath9 and @xmath3 axes , respectively ; for the @xmath1-axis rotation , which has the lowest rotational barrier , the dipole moment goes down from 0.661 debye at @xmath123 to 0.534 debye at @xmath124 in the clust model , as shown fig .
[ fig : polarizable](a ) . in the context of population analysis , increase of the dipole moment of no@xmath15 implies that more electrons are distributed on the o atom , i.e. , electrons are flowing from the nitrogen atom to the oxygen atoms .
conversely , decrease of the dipole moment indicates a reversal in electron transfer .
therefore , we have demonstrated considerable _
intramolecular _ charge - transfer , although the intermolecular charge - transfer is usually small in ionic crystals .
this intramolecular charge - transfer could be further elucidated based on the language of molecular orbitals ( mos ) : in the minimal basis set of atomic orbitals ( aos ) on nitrogen and oxygen , each atom of the no@xmath17 molecule contributes one @xmath125-orbital perpendicular to the molecular plane .
thus their linear combinations , which are determined by the crystal field , form three different @xmath126 mos extended over the entire molecule , thus leading to the above - mentioned intramolecular charge - transfer .
note that the mean hf polarizability of free no@xmath8 , @xmath127 in the atomic units , is much higher than that of free na@xmath20 , @xmath128 , calculated by using the 6 - 31 * basis .
therefore , it is reasonable to expect that no@xmath8 in solid nano@xmath0 encountering different crystal - field environments as it rotates , redistributes its charge among the three constituent atoms to lower its energy . in fig .
[ fig : barrier444 ] _ ab initio _ barriers to rotation of the central no@xmath8 in a @xmath121 lattice around the @xmath9 , @xmath3 , @xmath1 axes through the center of its ( na@xmath20)@xmath122 cage .
( a ) clust and ( b ) cryst . ]
we show that the rotational barriers of the central no@xmath8 . on the whole ,
the clust results are similar to those in figs .
[ fig : barrier](a ) and [ fig : barrier](b ) , while the cryst results are similar to those in figs .
[ fig : barrier](c ) and [ fig : barrier](d ) .
the reason is that in the cryst calculations we consider only the crystal - field effects originating purely from electrostatic interactions with the background point charges , similar to the point charge model used to obtain figs .
[ fig : barrier](c ) and [ fig : barrier](d ) .
whereas , in the clust calculations , overlap compression is also taken into account , which is reflected in figs .
[ fig : barrier](a ) and [ fig : barrier](b ) as inclusion of short range repulsion . in fig .
[ fig : barrier444](a ) , the barrier to the @xmath1-axis rotation is the lowest , but comparable with that to the @xmath3-axis rotation .
this feature is caused by the fact that all background anions are represented by single point charges in our clust and cryst calculations ; however , we anticipate that restoring multipole moments of these background anions would increase the barrier difference among rotations about the @xmath9 , @xmath3 , and @xmath1 axes .
to further elaborate the electronic polarization effect on no@xmath15 arising from the crystal environment , we change the rotation center to the center of mass of the no@xmath15 which was assumed in the model by kremer and siems .
@xcite in this case , the dipole moment of no@xmath15 and the rotational barriers are presented in figs .
[ fig : crystal](a ) and [ fig : crystal](b ) , respectively . in the clust model ,
no@xmath8 rotates around the @xmath108 axes through its center of mass .
( a ) dipole moment of no@xmath8 , and ( b ) rotational barriers . ] the dipole moment changes in a different way from that shown in fig .
[ fig : polarizable](a ) .
in particular , the dipole moment in the ferroelectric phase ( @xmath123-rotation ) is larger than at @xmath129-rotation in fig .
[ fig : polarizable](a ) , while it is smaller in fig .
[ fig : crystal](a ) . on the other hand
, there is no qualitative discrepancy between fig .
[ fig : crystal](b ) and fig .
[ fig : barrier444](a ) ; the main differences are : the rotational barrier about the @xmath9 axis has risen by 135% while the barrier about the @xmath1 axis is depressed by 26% .
this means that the order of rotational barriers is enhanced by the change of the rotation center .
although strong crystal field effects have been revealed by these _ ab initio _ calculations , the rotational barriers obtained from the polarizable - ion models are in qualitative agreement with those from the rigid - ion models , confirming that the rigid - ion model is capable of describing the phase behavior in nano@xmath0 .
we have presented md simulations of nano@xmath0 using a hybrid _ a priori _ method consisting of _ ab initio _ calculations and gordon - kim electron gas theory to analytically calculate the crystal potential surface .
this method has been carefully examined by using different population analysis methods .
we have carried out _ ab initio _ hartree - fock calculations of the intermolecular interactions for no@xmath8:na@xmath130 and no@xmath8:no@xmath8 dimers and concluded that the pair potentials of the rigid - ion model can correctly reproduce the _ ab initio _ results .
we demonstrated that a rigid - ion model is capable of describing phase behavior in solid nano@xmath0 .
the crystal - field effects on the no@xmath8 ion are also addressed in two polarizable - ion models for which the ferroelectric phase of nano@xmath0 was found to have a larger dipole moment of no@xmath8 than the @xmath129-rotation phase .
remarkable internal charge - transfer effect is found to be stabilizing the crystal structure of nano@xmath0 . in our md simulations , two rigid - ion models using mpa and fpa , respectively , have been studied .
the one using mpa , which overestimates the dipole moment of no@xmath8 , gives rise to the more comparable results with the experiments , since such overestimation also stabilizes the crystal structure , thus mimics the anion polarization effect . to quantitatively simulate nano@xmath0 ,
a more elaborate polarizable - ion model is needed .
helpful discussions with dr .
l. l. boyer are gratefully acknowledged .
this work was supposed by the nebraska research initiative , the nebraska epscor - nsf grant eps-9720643 , and department of the army grants daag 55 - 98 - 1 - 0273 and daag 55 - 99 - 1 - 0106 .
w. n. m. is grateful for the support from the office of naval research . | we present molecular dynamics simulations of solid nano@xmath0 using pair potentials with the rigid - ion model .
the crystal potential surface is calculated by using an _ a priori _ method which integrates the _ ab initio _ calculations with the gordon - kim electron gas theory .
this approach is carefully examined by using different population analysis methods and comparing the intermolecular interactions resulting from this approach with those from the _ ab initio _ hartree - fock calculations .
our numerics shows that the ferroelectric - paraelectric phase transition in solid nano@xmath0 is triggered by rotation of the nitrite ions around the crystallographical @xmath1 axis , in agreement with recent x - ray experiments [ gohda _ et al .
_ , phys .
rev.b * 63 * , 14101 ( 2000 ) ] .
the crystal - field effects on the nitrite ion are also addressed .
remarkable internal charge - transfer effect is found . |
in the one - body approximation , the study of disordered systems amounts to the study of random schrdinger operators of the form @xmath0 where @xmath1 is a kinetic term ( _ i.e. _ a self - adjoint or essentially self - adjoint operator corresponding to some dispersion relation , typically a regularized version of @xmath2 ) and @xmath3 is a real random potential ( in the simplest case , @xmath3 is a white noise ) .
we work on a ultra - violet regular subspace of @xmath4 and we restrict ourselves to @xmath5 small so as to see @xmath6 as a kind of perturbation of the free hamiltonian . the properties of @xmath7 are usually established through the behavior of the kernel of the resolvent operator or green s function ( @xcite , @xcite , @xcite ) @xmath8 for instance , the density of states is given by @xmath9 the important point is that , in the thermodynamic limit , the system is self - averaging , _
i.e. _ mean properties are often almost sure ones .
thus the problem can be seen as a statistical field theory with respect to the random field @xmath3 . in statistical mechanics , functional integrals in the weakly coupled regime
are controled through a cluster expansion ( or polymer expansion ) with small field versus large field conditions , the problem being then to control a boltzmann weight ( @xcite , @xcite ) . in the first part of this paper ,
we derive a resolvent cluster expansion with large field versus small field conditions assuming that @xmath3 satisfies some large deviation estimates .
this would allow to prove the existence and the regularity of the mean green s function ( theorem [ thre ] ) and to get an asymptotic expansion for the density of states .
in the second part , we show that the hypothesis of theorem [ thre ] are satisfied in the case of a 2 dimensional model with a rotation invariant dispersion relation and an infra - red cut - off on the potential . from the point of view of _ renormalization group _
analysis , our results allows to control the model away from the singularity , _
i.e. _ to perform the first renormalization group steps and therefore to generate a fraction of the expected `` mass '' .
in @xmath10 we consider @xmath0 where @xmath3 is a gaussian random field with covariance @xmath11 whose smooth translation invariant kernel is rapidly decaying ( we will note the associated measure @xmath12 ) . because @xmath11 is smooth
, @xmath12 as a measure on tempered distributions is in fact supported on @xmath13 functions .
we suppose also that @xmath14 has compact support so that we do not have to deal with ultra - violet problems .
we construct the finite volume model in @xmath15 by replacing @xmath11 and @xmath1 by their `` @xmath16-periodization '' @xmath17 then we define @xmath18 where @xmath19 can be considered either as a measure on @xmath20 or as a measure on @xmath21 which is supported by the space of @xmath16-periodic functions . in the same way ,
@xmath22 will be considered as an operator either on @xmath23 or on @xmath24 .
one can note that in momentum space , because of the cut - off , the problem reduces to a finite dimensional one . because @xmath3 is almost surely regular ,
its operator norm as a multiplicative operator is equal to its @xmath25 norm ( it is easy to see that @xmath26 , equality can be obtained by taking test functions @xmath27 such that @xmath28 ) . therefore @xmath3 is bounded and self - adjoint .
then @xmath29 is almost surely an analytic operator - valued function of @xmath5 in a small domain ( depending on @xmath3 ) around the origin .
this domain can be extended to a @xmath3-dependent neighborhood of the real axis thanks to the identity ( for @xmath30 small enough ) @xmath31^{n } \right\}\ ] ] in the same way , @xmath32 is analytic in @xmath33 .
one can also check that @xmath32 has a smooth kernel and is integrable with respect to @xmath19 .
furthermore , @xmath34 will have a translation invariant kernel because @xmath19 is translation invariant .
we introduce a function @xmath35 which satisfies * @xmath35 is an odd @xmath13 function , increasing and bounded * for any @xmath36 , @xmath37 * for any @xmath38 , @xmath39 * the @xmath25 norm of its derivatives does not grow too fast then for @xmath40 , we define the operators @xmath41 , @xmath42 and @xmath43 through the fourier transform of their kernel @xmath44 - i\mu \right|^{1/2 } } \\
\hat{u}_{\lambda , \mu}^{-1}(p ) & = & \hat{d}_{\lambda , \mu}^{2}(p ) \hat{c}_{\lambda , \mu}^{-1}(p ) \end{aligned}\ ] ] given any characteristic length @xmath45 we can divide the space into cubes @xmath46 of side @xmath45 and construct an associated @xmath47 partition of unity @xmath48 where @xmath49 has support in a close neighborhood of the cube @xmath46 ( _ e.g. _ on @xmath46 and its nearest neighbors ) .
this decomposition induces an orthogonal decomposition of @xmath3 into a sum of fields @xmath50 with covariance @xmath51 for simplicity we will pretend that @xmath11 and @xmath52 have compact support , so that @xmath50 is almost surely supported on a close neighborhood of @xmath46 , moreover we will take that it is restricted to @xmath46 and its nearest neighbors .
the generalization to a fast decaying @xmath11 can be easily obtained by decomposing each @xmath50 over the various cubes and write more complicated small / large field conditions that test the size of @xmath50 in the various cubes .
this leads to lengthy expressions that we want to avoid .
finally , we note @xmath53 the distance in @xmath15 @xmath54 in the following , @xmath55 or @xmath56 will stand as generic names for constants in order to avoid keeping track of the numerous constants that will appear
. furthermore we will not always make the distinction between a function and its fourier transform but we will use @xmath36 , @xmath57 and @xmath58 as space variables and @xmath59 and @xmath60 as momentum variables .
[ thre ] + suppose that * @xmath11 is smooth and has fast decay * @xmath61^{d } } \ ! \xi^{-1}_{\lambda}(x ,
y ) \ , dx \ , dy$ ] exists * for all @xmath62 $ ] and all @xmath63 , @xmath64 , @xmath65 and @xmath66 have smooth kernels with fast decay over a length scale @xmath45 . * for all @xmath67 , we have @xmath68 such that for all @xmath16 and all triplets @xmath69 @xmath70^{n_{1 } } \left[1 + l^{-1 } d_{\lambda}(\delta_{2 } , \delta_{3})\right]^{n_{1}}}\ ] ] * there are constants @xmath71 , @xmath72 , @xmath73 and @xmath74 such that @xmath75 where @xmath76 denote the probability with respect to the measure @xmath77 ( @xmath78 ) @xmath79 then let @xmath80 , @xmath81 and @xmath82 for all @xmath83 and for all @xmath84 small enough ( in a @xmath5-dependent way ) , @xmath85 is uniformly bounded in @xmath16 and admits the following development ( in the operator norm sense ) @xmath86 where @xmath87^{d}$ ] , and @xmath88 is the characteristic function of @xmath89 .
furthermore we have the following properties * @xmath90 has a smooth , translation invariant kernel * @xmath91 and @xmath90 have high power decay @xmath92^{n_{0}}}\ ] ] and a similar relation for @xmath90 with @xmath53 being replaced by @xmath93 .
* @xmath94 is an analytic operator valued function of @xmath33 for all @xmath33 in @xmath95e_{1 } , e_{2}[$ ] with a small @xmath5-dependent radius of analyticity .
* @xmath94 is a @xmath13 operator - valued function of @xmath5 and admits an asymptotic expansion to all orders in @xmath5 . which is the formal perturbative expansion of @xmath96 ( @xmath97 denotes the scalar product , _
i.e. _ @xmath98 ) this theorem is formulated in a rather general way so as to apply with minimum transformation to various situations ( lattice or continuous models ) and in any dimension .
then we construct a concrete example with a two - dimensional model .
one can also refer to @xcite for a @xmath99 case .
we consider @xmath100 where * @xmath101 is a ultra - violet regularized inverse laplacian , _
i.e. _ there is a @xmath102 function @xmath103 equal to 1 on `` low '' momenta such that @xmath104 we will note @xmath105 instead of @xmath106 , the uv - cutoff being then implicit .
* we are interested in the mean green s function for an energy @xmath107 * @xmath108 is an infra - red cut - off which enforces @xmath109 for some large constant @xmath110 * @xmath3 has covariance @xmath11 which is a @xmath47 approximation of a @xmath111-function this corresponds to the model away from the singularity @xmath112 in a multi - scale renormalization group analysis , we will show that it generates a small fraction of the expected imaginary part which is @xmath113 .
let @xmath114 be an even integer greater than 2 , we define @xmath115 such that @xmath116 next , we construct a smooth partition of unity into cubes of side @xmath117 ( they form a lattice @xmath118 ) and we construct the fields @xmath50 s accordingly .
[ thprob2d ] + there exist constants @xmath71 and @xmath72 such that for any @xmath16 , @xmath119 and @xmath120 we have @xmath121 furthermore theorem [ thre ] applies and @xmath122 is asymptotic to its perturbative expansion so that it behaves more or less like @xmath123 it is easy to extend this result to the case of a rotation invariant dispersion relation and for energies outside the free spectrum not too close to the band edge . in this case , the cut - off is no longer needed so that the result apply to the full model .
we give here the global strategy for proving theorem [ thre ] , the main ingredient being the polymer expansion that we will detail in the following .
first we recall ( without proving them ) some quite standard properties of gaussian measures .
[ lemtrans ] complex translation + let @xmath124 be a gaussian random field with covariance @xmath55 and let @xmath125 be the associated measure .
for any regular functional @xmath126 and any function @xmath127 , we have the following identity @xmath128 [ lemipp ] integration by part + with the same notations than above we have @xmath129 those lemmas could for instance be easily proved for polynomial functionals and extended through a density argument to a wide class of functionals .
our starting point is obtained by applying lemma [ lemtrans ] with @xmath131 .
@xmath132 on one hand we earned something because now the resolvent operator in the integral is bounded in norm independently of @xmath84 ( in the following we will note @xmath58 instead of @xmath133 and show convergence for any @xmath58 such that @xmath134 , this would allow to prove analyticity in @xmath58 ) .
but on the other hand we have a huge normalization factor to pay .
however , we can remark that this normalization factor is in fact equivalent to a factor @xmath135 per @xmath45-cube .
most of the demonstration amounts to a polymer expansion of @xmath136 , _
i.e. _ we write @xmath91 as a sum over polymers of polymer activities @xmath137^{n_{0 } } } \sum_{y \in { \cal a } } \lambda^{c_{2}|y| } \gamma_{y } \ , t(y ) \right ] \chi_{\delta_{in}}\end{aligned}\ ] ] where @xmath138 and @xmath139 are small constants , @xmath140 has decay in the spatial extension of @xmath141 and @xmath142 is bounded .
furthermore , @xmath143 is given by a functional integration over fields @xmath50 s corresponding to cubes in the support of the polymer @xmath141 .
this show that @xmath91 is bounded and has a high power decay uniformly in @xmath16 .
next , when we consider @xmath144 we can divide the sum over polymers into a sum over polymers with a large spatial extension ( say @xmath145 ) and sum over `` small '' polymers .
the large polymers will have a total contribution small as @xmath146 to some large power .
for the small polymers , since we are far away from the boundaries , their contribution calculated with @xmath19 will be equal to their contribution calculated with @xmath12 up to a factor @xmath147 . in this way we can prove the development ( [ lambdadev ] ) .
smoothness of the kernel will be obtained because we will show that we can write @xmath148 the convergence for any @xmath149 allows to show analyticity ( we write @xmath58-derivatives as cauchy integrals so that we can show that they all exists and do not grow too fast ) .
then an asymptotic expansion can be generated through the repeated use of resolvent identity . finally , for the density of states , we just need to remark that @xmath150 where @xmath151 is a regularized @xmath111-function because of the presence of the ultra - violet cut - off .
thus an asymptotic expansion for @xmath152 with respect to the operator norm will yield an asymptotic expansion for the density of states .
cluster expansions in constructive field theory lay heavily on a clever application of the taylor formula with integral remainder .
writing the full taylor series would amount to completely expand the perturbation series , which most often diverges , and therefore should be avoided .
a rather instructive example of minimal convergent expansion is the brydges - kennedy forest formula : you have a function defined on a set of links between pair of cubes and you expand it not on all possible graphs but only on forests ( cf @xcite ) . for more complex objects
a way to generalize such a formula can be found in @xcite , and we refer the reader to it for a more careful treatment and for various proofs .
let us assume that we have a set of objects that we call monomers .
a sequence of monomers will be called a polymer , then we will expand a function defined on a set of monomers into a sum over allowed polymers . to be more precise ,
let @xmath153 be a set of monomers , we define the set @xmath154 of polymers on @xmath153 as the set of all finite sequences ( possibly empty ) of elements of @xmath153 .
then a monomer can be identified to a polymer of length 1 . the empty sequence or empty polymer
will be noted @xmath155 .
we define on @xmath154 * a concatenation operator : for @xmath156 and @xmath157 , we define @xmath158 * the notion of starting sequence : we say that @xmath159 is a starting sequence of @xmath141 ( equivalently that @xmath141 is a continuation of @xmath159 ) and we note @xmath160 iff there exists @xmath161 such that @xmath162 then we call allowed set ( of polymers ) any finite subset @xmath163 such that * @xmath164 * @xmath165 the first condition implies that @xmath166 whenever @xmath167 is non - empty .
finally , for @xmath141 belonging to some allowed set @xmath167 , a monomer @xmath124 is said to be admissible for @xmath141 ( according to @xmath167 ) iff @xmath168 .
[ lemclust ] + let @xmath169 be a set of @xmath170 monomers and @xmath154 the set of polymers on @xmath153 .
we assume that we have an indexation of @xmath171 by @xmath153 , _
i.e. _ a bijection from @xmath153 to @xmath172 so that an element of @xmath173 can be noted @xmath174 . for @xmath175 a regular function from @xmath173 to some banach space @xmath176 and an allowed set @xmath177
, the polymer expansion of @xmath175 according to @xmath167 is given through the following identity @xmath178\end{aligned}\ ] ] where @xmath179 is given by @xmath180 _ proof _ + the proof is made through an inductive iteration of a first order taylor formula . we start with @xmath181 and put a common interpolating parameter @xmath182 on all admissible monomers for the empty set , i.e. we make a first order taylor expansion with integral remainder of @xmath183 $ ] between 0 and 1 , with @xmath184 being the vector with entries 1 or 0 according to whether the corresponding monomer is admissible or not .
then each partial derivative acting on @xmath175 can be seen as taking down the corresponding monomer so that terms can be seen as growing polymers .
the iteration goes as follow : for a term of order @xmath185 corresponding to a given polymer @xmath141 and having @xmath185 interpolating parameters @xmath186 we put a common parameter @xmath187 interpolating between @xmath188 and @xmath189 on all monomers admissible for @xmath141 .
it is easy to check that the process is finite since @xmath190 is finite and that one obtains the desired formula .
@xmath130 in the following our monomers are sets of cubes ( that we call the support of the monomer ) and links between those cubes .
when we take down a polymer , we connect all the cubes in its support and maybe some more cubes . thus a polymer is made of several connected regions , we will say that it is connected if it has a single connected component
. the rules of admissibility will be to never take down a monomer whose support is totally contained in a connected region . in this case
, one can show that the interpolating parameters depend only of the connected component to which the corresponding monomer belongs so that one can think to `` factorize '' the connected components .
we define generalized polymers as sets of connected polymers .
then a generalized polymer @xmath191 is allowed if the polymer @xmath192 is allowed ( this does not depend of the order of the @xmath193 .
equation ( [ eqclust0 ] ) becomes @xmath194\ ] ] where the sum extends on all allowed generalized polymers , and @xmath195 is given by @xmath196 semi - perturbative expansion ( like cluster expansions ) are convergent only when the `` perturbation '' is small ( in our case the operators @xmath50 s ) .
thus it is very important to distinguish between the so called _ small field regions _ where perturbations will work and the _ large field regions _ where we must find other estimates ( they will come mostly from the exponentially small probabilistic factor attached to those regions ) .
we take a @xmath47 function @xmath84 such that * @xmath197 * @xmath198 $ ] * @xmath199 } } = 1 $ ] then for each @xmath46 we define @xmath200 where @xmath201
. then we can expand @xmath202 where @xmath203 is the large field region whose contribution will be isolated through the following lemma .
[ lemlargef ] + let @xmath203 be a large field region made of @xmath170 cubes @xmath204 , , @xmath205 and @xmath110 any operator such that @xmath206 ( @xmath207 stands for @xmath208 ) .
we have the following identity @xmath209 where @xmath210 _ proof _ + the proof relies on resolvent expansion identities @xmath211 we show by induction that for all @xmath212 we have @xmath213 the case @xmath214 is obtained by a resolvent expansion @xmath215 then we go from @xmath216 to @xmath217 with 2 steps of resolvent expansion .
we write @xmath218 finally , for @xmath219 we make a last resolvent expansion on the rest term @xmath220 by writing @xmath221
@xmath130 if we look at @xmath222 and fix @xmath223 , we can see that summing over the sequences @xmath224 and choosing a particular term for each @xmath225 amounts to construct a tree on @xmath226 .
we define an oriented link @xmath227 as a couple of cubes that we note @xmath228 , then @xmath229 is the set of oriented links . given two cubes @xmath230 and @xmath231 and a set of cubes @xmath232 we construct the set @xmath233 of oriented trees going form @xmath230 to @xmath231 through @xmath203 as the sequences @xmath234 which satisfy * @xmath235 * @xmath236 * @xmath237 * @xmath238 * @xmath239 then we have the following equivalent formulation of lemma [ lemlargef ] .
[ ldecoupling ] + let @xmath203 be a large field region made of @xmath170 cubes @xmath204 , , @xmath205 and @xmath110 any operator such that @xmath206 we have the following identity @xmath240 where @xmath241 the proof being just a rewriting of lemma [ lemlargef ] is quite immediate . @xmath130 thanks to this lemma we can factorize out the contribution of the large field region , then we need to extract spatial decay for the resolvent in the small field region
. however a kind of combes - thomas estimate ( @xcite ) would not be enough because of the normalization factor that we must pay .
for this reason , we will make a polymer expansion to determine which region really contributes to the resolvent .
for some large field region @xmath203 , we want to prove the decay of @xmath242 and get something to pay for the normalization factor .
we define the set @xmath243 of links as the set of pair of cubes , and @xmath244 as the set of links which do not connect two cubes of @xmath203 .
then for @xmath245 we define @xmath246 with the convention that @xmath247 if @xmath248 . for any fixed @xmath249 we expand @xmath250 on @xmath244 with the rule that for any growing polymer * if we have two adjacent connected components @xmath159 and @xmath161 ( such that @xmath251 ) we connect the two components * we connect @xmath252 ( resp .
@xmath253 ) to any adjacent polymer component this allows to take into account that the operators localized on a pair of cubes have their support extending to the neighboring cubes .
let us notice that if @xmath110 and @xmath254 have disjoint support , we have @xmath255 then it is easy to see that the expansion of @xmath250 involves only totally connected polymers which connect @xmath252 to @xmath253 , because the other terms necessarily contain a product of two operators with disjoint supports which gives zero .
we note @xmath256 the corresponding set of polymers which is a decreasing function of @xmath203 , _
@xmath257 then , according to ( [ eqclust ] ) , our expansion looks like @xmath258 \\ & = & \sum_{y \in { \cal a}(\omega , l_{0 } ) } \int \prod_{i } dh_{i } \left(\prod_{x \in y } \frac{\partial } { \partial u_{x}}\right ) \frac{1}{\displaystyle i + \sum_{x \in { \cal l}(\omega ) } z_{x } q_{x } + \sum_{x \in y } u_{x } q_{x } } \left[\vec{z}(y , \{h_{i}\ } ) , \vec{0}\right ] \end{aligned}\ ] ] then in the second expression , we rewrite the derivatives as cauchy integrals so that @xmath259 where @xmath260 we suppose that we fixed @xmath67 the power rate of decay in @xmath261 of @xmath262 and @xmath263 , then we have the following lemma [ lemresexp ] + for @xmath264 and @xmath5 small enough , we have @xmath265^{n_{2 } } } \gamma(y ) \mbox { with } \sum_{y \in { \cal a}^{*}(l_{0 } ) } \gamma(y ) \leqslant 1\end{aligned}\ ] ] where @xmath266 is the number of monomers in @xmath141 .
_ proof _ + since we are in the small field region @xmath267^{n_{1}-(d+1)}}\ ] ] then in ( [ eqresexp ] ) we can integrate each @xmath268 on a circle of radius @xmath269^{n_{1 } - 2 ( d+1)}\ ] ] while staying in the domain of analyticity for @xmath268 and have a resolvent bounded in norm by say @xmath270 ( if @xmath5 is small enough ) . thus @xmath271^{n_{1 } - 2(d+1 ) } } \right ) \\ & \leqslant & \frac{\lambda^{|y|/4}}{[1 + l^{-1 } d_{\lambda}(\delta_{0 } , \delta_{0'})]^{n_{2 } } } \ , \frac{\left[o(1 ) \lambda^{1/4}\right]^{|y|}}{|y| ! } \left(\prod_{{x \in y } \atop { x=\{\delta_{x } , \delta_{x}'\ } } } \frac{1}{[1+l^{-1 } d_{\lambda}(\delta_{x } , \delta_{x}')]^{d+1 } } \right ) \end{aligned}\ ] ] this demonstrates the first part of the lemma with @xmath272^{|y|}}{|y| ! } \left(\prod_{x \in y } \gamma_{x }
\right ) \quad \mbox { and } \sum_{x \ni \delta } \gamma_{x } = o(1)\ ] ] a link @xmath273 can either be a true link when @xmath274 or a tadpole when both cubes collapse .
our expansion rules insure that there is at most 1 tadpole per cube of @xmath275 the support of @xmath141 .
if we forget about proximity links ( that we connect also adjacent cubes ) then a polymer with @xmath216 true links and @xmath59 tadpoles has a support of @xmath217 cubes ( 2 of them being @xmath252 and @xmath252 ) and the true links make a tree on the support of @xmath141 . if we take into account the proximity links then two connected links in the tree on @xmath141 are adjacent instead of sharing a common cube , we will forget about this since it would induce at most a factor @xmath276 .
the links in @xmath141 are ordered , but we can take them unordered by eating up the @xmath277 we have in @xmath278 .
then the sum over @xmath141 can be decomposed as * choose @xmath279 * choose @xmath280 cubes @xmath281 * chose a tree @xmath282 on @xmath283 * choose @xmath284 * place @xmath59 tad - poles on @xmath285 we can perform the sum on tadpole configurations because for @xmath59 tadpoles , we have a factor @xmath286^{p}$ ] coming from the tadpoles and at most @xmath287 configurations .
thus @xmath288^{m+1 } \nonumber \\ & & \quad \left[o(1 ) \lambda^{1/4}\right]^{m } \left ( \prod_{x \in { \cal t } } \gamma_{x}\right)\end{aligned}\ ] ] then we fix first the form of @xmath282 then we sum over the positions of @xmath204 , , @xmath289 .
but since the cubes are now labeled we get @xmath290 the desired sum . @xmath291^{m } \frac{1}{(m-1 ) ! } \sum_{\cal t } \sum_{(\delta_{1 } , \ldots , \delta_{m-1 } ) } \left ( \prod_{x \in { \cal t } } \gamma_{x}\right)\ ] ] we choose @xmath252 as the root of our tree and suppose that the position of @xmath253 is not fixed .
then the sum over the position of the cubes is made starting from the leaves thanks to the decaying factors @xmath292 ( cf .
@xcite ) , this costs a factor @xmath293 . finally , the sum over @xmath282 , which is a sum over unordered trees , is performed using cayley s theorem which states that there are @xmath294 such trees .
@xmath291^{m } \frac{(m+1)^{m-1}}{(m-1 ) ! } \leqslant o(1 ) \lambda^{1/4 } \leqslant 1\ ] ] for @xmath5 small enough .
@xmath130 we note that we can perform the same expansion on @xmath295 we define @xmath296 we can combine equations ( [ complextrans ] ) , ( [ lsdecomp ] ) , ( [ eqldecoupling ] ) and ( [ yexpan ] ) to write @xmath297 where we pretend that the @xmath49 s are sharp otherwise we would have to deal with adjacent cubes but it s an irrelevant complication .
furthermore , for the leftmost term we made a polymer expansion of @xmath298 instead of @xmath299 so that we can write @xmath300 as @xmath301 the crucial point here is to notice that for any cube @xmath46 , each term where @xmath46 appears in @xmath203 but not in @xmath302 pairs with a corresponding term where @xmath303 and @xmath304 ( _ i.e. _ @xmath46 has been killed in every polymer expansion ) .
then the corresponding @xmath305 and @xmath306 add up back to 1 so that @xmath307 the factor @xmath308 correponds to the translation of @xmath3 by @xmath309 , this is equivalent to have translated all the @xmath50 s by @xmath310 therefore we can write it as @xmath311 then we can perform the integration on all @xmath312 so that the normalization factor reduces to @xmath313 this amounts to pay a constant per cube of @xmath314 , this is done in @xmath203 with a fraction of the probabilistic factor coming from the large field condition and in @xmath315 with a fraction of the factor @xmath316 coming from the @xmath317 .
the sums over the various @xmath318 s are controled by lemma [ lemresexp ] and we are left with a sum over a tree that we perform much in the same way we did in lemma [ lemresexp ] .
indeed one can check that spatial decay appears through factors of the form @xmath319 thus we can extract decay in @xmath320 time a bound in @xmath321 when we combine all these factors . yet we need some extra features to deal with the product of @xmath322 s , each of them being bounded in norm by @xmath323 for some @xmath324 .
the factor @xmath325 can be controled with a small fraction of the probabilistic factor attached to the cube @xmath326 if a given @xmath327 appears at a large power it has necessarily a large number of links attached to it .
because of the tree structure , the links must go further and further so that the decay of the links together with the gaussian measure allow to control the factorial coming from the accumulation of fields .
this is quite standard and the reader can refer to @xcite for instance . finally we can write @xmath300 as a sum over polymers of the form @xmath328^{n_{3 } } } \sum_{y \in { \cal a}^{*}(\delta_{in } , \delta_{out } ) } \lambda^{c_{2}|y| } \gamma_{y } t(y ) \end{aligned}\ ] ] where @xmath138 and @xmath139 are small constant , @xmath140 has decay in the spatial extension of @xmath141 and @xmath329 is bounded .
we are interested now in the particular case @xmath100 where * @xmath101 is a ultra - violet regularized inverse laplacian , that we will note @xmath331 * @xmath108 is an infra - red cut - off that forces @xmath332 * @xmath3 has covariance @xmath11 which is a @xmath47 approximation of a @xmath111-function * @xmath114 is an even integer greater than 2 , and @xmath333 is such that @xmath116 for each @xmath334 , we construct a smooth partition of unity into cubes of side @xmath335 which form a lattice @xmath336 . it follows a decomposition of @xmath3 in fields @xmath337 and we will assume for simplicity that for @xmath338 and @xmath339 @xmath340 even if it is not totally true because of irrelevant border effects .
we make a partition of unity according to the size of @xmath341 thanks to a function @xmath342 which satisfies * @xmath342 is in @xmath343 with value in @xmath344 $ ] * @xmath342 has its support inside @xmath345 $ ] and is equal to 1 on @xmath344 $ ] * the @xmath25 norm of the derivatives of @xmath342 does not grow too fast then we construct @xmath346
\\ \hat{\eta}_{j}(p)&=&\hat{\eta } \left[m^{2j}(p^{2}-e)^{2}\right ] - \hat{\eta } \left[m^{2(j+1 ) } ( p^{2}-e)^{2 } \right ] \quad \mbox{for } j > 0 \\
\end{array } \right.\ ] ] in order to shorten expressions , we assume that @xmath347 we expect that most of the physics will come from the neighborhood of the singularity @xmath348 of the free propagator . as an operator in momentum space , @xmath3 has a kernel @xmath349 .
but since @xmath59 and @xmath60 have more or less the same norm , there are only two configurations which give the sum @xmath350 ( cf @xcite ) .
we can see this in another way .
if we make perturbations and integrate on @xmath3 we will get feynman graphs with four - legged vertices where the incoming momenta have a fixed norm and must add to zero ( or almost zero ) because of ( approximate ) translation invariance .
then the four momenta approximately form a rhombus which happens to be a parallelogram .
it implies that they must be more or less opposite 2 by 2 .
thus the problem looks like a vectorial model because the angular direction of the momentum is preserved . in order to have this feature more explicit , we decompose the slice @xmath351 into @xmath352 angular sectors .
we introduce @xmath353 with * @xmath353 is an even function in @xmath354 with value in @xmath344 $ ] * @xmath353 has its support inside @xmath355 $ ] and is equal to 1 on @xmath356 $ ] * @xmath357 for @xmath358 * the @xmath25 norm of the derivatives of @xmath353 does not grow too fast then we define @xmath359 and construct sectors @xmath360 of angular width @xmath361 centered around @xmath362 ( identifying @xmath363 and @xmath364 ) , with @xmath365 .
@xmath366 afterwards , we define the operators @xmath367 s by their kernel @xmath368 they form a positive , self - adjoint partition of identity .
we will map our problem to an operator - valued matrix problem with the following lemma whose proof is quite obvious .
+ let @xmath370 be an hilbert space and suppose that we have a set of indices @xmath371 and a partition of unity @xmath372 where @xmath373 is the identity in @xmath374 and the @xmath375 s are self - adjoint positive operators . for all @xmath376 ,
we define @xmath377 then @xmath370 and @xmath374 are naturally isomorphic to @xmath378 and @xmath379 thanks to @xmath380 in our case , we define @xmath381 as the set of sectors in the slice @xmath382 and @xmath383 so that we can construct the operator - valued matrices @xmath384 s as @xmath385 for a slice @xmath386 , we define the enlarged slice @xmath387 then an angular sector @xmath388 of @xmath386 has a natural extension into an angular sector @xmath389 of @xmath390 and we have the corresponding operator @xmath391 .
let @xmath393 be defined by @xmath394 where the @xmath111 are kronecker s ones .
we can remark that @xmath395 then we have the following large deviation result [ thprobv ] + there exists constants @xmath55 and @xmath396 such that for all @xmath16 , @xmath397 , @xmath398 and @xmath399 @xmath400 where @xmath401 behaves like @xmath402 .
_ proof _ + we use the bound @xmath403^{m_{0}}\ ] ] where @xmath404 thus for any @xmath405 @xmath406^{m_{0 } } d\mu_{\xi_{\lambda } } ( v)\end{aligned}\ ] ] let us note @xmath407^{m_{0 } } \!\!\!\ !
= \mbox{tr } \left[\sum_{\alpha } ( \eta_{\alpha}^{j})^{2 } \ ,
v_{\delta } \sum_{\beta } ( \eta_{\beta}^{k})^{2 } \ , v_{\delta } \right]^{m_{0}}\ ] ]
we have the following lemma [ lemproba ] + there exists a constant @xmath55 such that for all @xmath405 we have the following bound @xmath408\ ] ] this lemma is the core of the demonstration but its proof is quite long so that we postpone it until the end of this part .
it leads to @xmath409\ ] ] we take @xmath410 and use the rough bound @xmath411 to get the desired estimate .
@xmath130 in fact , in the proof of lemma [ lemproba ]
it is easy to see that @xmath412 and @xmath413 can be replaced by @xmath414 and @xmath415 with the same result .
furthermore , thanks to the locality of @xmath3 and to the decay of the @xmath416 s , the sum of several @xmath417 s is more or less an orthogonal sum .
more precisely , for any cube @xmath252 we define @xmath418 as the set of cubes of @xmath419 which are contained in @xmath252 . then given two sets @xmath420 and @xmath421 and their smoothed characteristic functions @xmath422 and
@xmath423 we have [ lemortho ] + for any @xmath185 and @xmath55 there is a constant @xmath424 such that for any @xmath397 and @xmath425 @xmath426\left[1 + m^{-nk } d(\omega_{2},\delta_{0})^{n } \right ] } \nonumber \\ & & \quad \max \left [ \|\bar{\eta}_{j } v_{\delta_{0 } } \bar{\eta}_{k}\| , \sup_{{m < j } \atop { n < k } } \sup_{\delta \in d_{m \wedge n}(\delta_{0 } ) } \!\ !
m^{-c(j - m+k - n ) } \|\eta_{m } v_{\delta } \eta_{n}\| \right ] \end{aligned}\ ] ] where @xmath427 .
_ proof _ + we introduce @xmath428 a @xmath47 function equal to 1 on the support of @xmath429 then we write @xmath430 afterwards we introduce the sectors and the matrix formulation and we notice that when we want to compute for instance the norm of the function @xmath431 , momentum conservation tells us that we can convolve @xmath428 by a function which is restricted in momentum space to the neighborhood of @xmath432 . in this way it is quite easy to see that we can extract at the same time spatial decay and momentum conservation decay .
let @xmath433 , we call @xmath434 and @xmath435 the events @xmath436 \\ y_{c_{y } , a } & = & \left[\| d_{\lambda , \mu } \eta_{e } v_{\delta_{0 } } \eta_{e } d_{\lambda , \mu}\| \geqslant a c_{y } j_{0 } m^{j_{0}/2 } \right]\end{aligned}\ ] ] we will note @xmath437 the contrary event of @xmath438 . theorem [ thprobv ] tells us that @xmath439 one can see that thanks to lemma [ lemortho ] , @xmath440 implies @xmath441 .
thus if we call @xmath442 @xmath443 furthermore , if we work with respect to @xmath440 which is stronger than @xmath444 everything goes as if one had @xmath70^{n_{1 } } \left[1 + l^{-1 } d_{\lambda}(\delta_{2 } , \delta_{3})\right]^{n_{1}}}\ ] ] thus we will be able to apply theorem [ thre ] with an effective coupling constant @xmath445 and a length scale @xmath446 .
if we want to make perturbations it is clever to perturb around the expected green s function without cut - off , _
i.e. _ we write @xmath447 where @xmath448 is the expected contribution of the tadpole given by the self - consistent condition @xmath449 afterwards , when we compute the perturbative expansion , the tadpole with cut - off will eat up a fraction @xmath450 of the counter - term so that @xmath451 @xmath130 in fact since the tadpole has a real part , it implies that we should also renormalize the energy by a shift @xmath452 \right)\ ] ] we will note @xmath453 , @xmath454 and @xmath124 as either @xmath455 or @xmath456 .
we can perform the integration on @xmath50 so that @xmath457 appears as a sum of feynman graphs .
@xmath458 where a solid line stands for a @xmath459 , a dashed line stands for a @xmath460 and a wavy line represents the insertion of a @xmath50 . in the following
, we will prove the theorem in infinite volume with @xmath3 having a covariance @xmath111 in order to have shorter expressions .
the proof can then easily be extended to short range covariances and finite volume except for the first few slices where one must pay attention to the ultra - violet cut - off but this is irrelevant because it will cost only a factor @xmath56 .
the integration on @xmath50 consists in contracting the wavy lines together , then both ends are identified and bear an extra @xmath49 which restricts their position .
the @xmath124 s will stand as propagators and the contraction of the @xmath50 s will give birth to 4-legged vertices .
first , we notice that if we note @xmath461 the opposite sector of @xmath462 @xmath463 then we put an orientation on each propagator , so that if a @xmath464 goes from a vertex at @xmath58 to a vertex at @xmath465 it gives a @xmath466 , _
i.e. _ it is equivalent to have an incoming @xmath467 at @xmath58 and an incoming @xmath464 at @xmath465 .
now , for a given vertex with incoming propagators @xmath468 ,
@xmath469 , @xmath470 , @xmath471 , the spatial integration over its position gives a term of the form @xmath472 in momentum space , it becomes @xmath473 where we use the same notation for a function and its fourier transform . in @xmath36-space
, @xmath49 is a @xmath47 function with support inside a box of side @xmath474 , it means that in momentum space , it is a @xmath13 function with fast decay over a scale @xmath475 .
thus for all @xmath185 there exists @xmath476 such that @xmath477 we make a decomposition of @xmath49 @xmath478 where @xmath479 has its support inside the ball of radius @xmath480 , @xmath481 has its support outside the ball of radius @xmath482 and @xmath483 forces @xmath484 to be in the interval @xmath485 $ ] . in this way
, we can decompose each vertex @xmath486 into a sum of vertices @xmath487 , where a vertex @xmath487 forces momentum conservation up to @xmath488 and has a factor coming from @xmath489 we split the factor in order to have a small factor per vertex and yet retain some decay to perform the sum on @xmath490 .
a graph will present tadpoles when two neighboring @xmath50 s contract together thus yielding a @xmath491 .
suppose that we have a @xmath382-tadpole then at the corresponding vertex we will have something of the form @xmath492 between the two @xmath456 s , momentum will be preserved up to @xmath493 which in most case is much smaller than @xmath494 so that @xmath495 is very close to @xmath496 .
then we would like to forget about @xmath497 by summing over @xmath462 and see the whole thing as a kind of new @xmath498 . now if per chance the new @xmath498 makes a tadpole we will erase it , and so on recursively .
first , we define the propagators as propagators ( or links ) of order @xmath188 @xmath499 then we define links of order @xmath500 @xmath501 where we do nt write the momentum conservation indices for shortness .
we have a similar definition for @xmath502 ( obtained by erasing @xmath60 @xmath456-tadpoles of order 0 ) .
we will note @xmath503 to indicate that momentum is preserved up to @xmath504 between the leftmost and the rightmost @xmath124 s or @xmath505 if momentum conservation is worse than @xmath494 .
now , we can iterate the process in an obvious way .
yet , we must add an important restriction : we will erase a @xmath506 tadpole only if it is attached to a @xmath507 vertex . + there exist constants @xmath72 and @xmath508 ( independent of @xmath382 and @xmath509 ) such that for any tadpole obtained by erasing a total of @xmath59 @xmath455-tadpoles and @xmath60 @xmath456-tadpoles we have the following bond @xmath510 where @xmath126 is a small factor coming from the various @xmath483 that appear in the expression of @xmath124 .
thus , @xmath175 gets smaller as momentum conservation gets worse .
_ proof _ + first we will prove this result when momentum is well preserved , _
i.e. _ up to @xmath494 at worst , then we will see what has to be adapted when there is a bad momentum conservation .
the proof is by induction on the order of the tadpole .
we define @xmath72 , @xmath508 and @xmath511 such that @xmath512 it is easy to see that for level 0 tadpoles @xmath513 now , consider @xmath514 a @xmath455-tadpole of order @xmath216 and weight @xmath515 obtained by erasing @xmath185 @xmath456-tadpoles of order @xmath280 and weights @xmath516 , , @xmath517 .
we have @xmath518 the expression of @xmath519 will be of the form @xmath520 since we supposed that we have momentum conservation up to @xmath494 , the @xmath521 s will be either @xmath522 or one of its neighbors and @xmath523 will be either @xmath524 or one of its neighbors .
thus the sum on sector attribution will give a factor @xmath525 .
we have @xmath526 @xmath455 s but only @xmath185 spatial integrations because we have a tadpole .
this gives a factor @xmath527 ( we forget about the momentum conservation factor for the moment ) .
finally the @xmath528 s bring their factor so that @xmath529 which is precisely what we want .
then we can do the same for the @xmath530 s .
now we must consider the cases with bad momentum conservation .
first , let us suppose that momentum conservation is bad overall for @xmath519 but was good for the @xmath528 s , then the previous argument will work except if there are some @xmath507 vertices . in this case
we will have to pay a factor @xmath352 to find the following @xmath521 instead of a factor 3 .
but from the corresponding @xmath531 we have a small factor @xmath532 from which we can take a fraction to pay the @xmath352 and retain a small factor for @xmath126
. finally , if a @xmath528 has a bad momentum conservation it is necessarily attached to a @xmath507 vertex ( otherwise we would not erase it ) . in this case
we must pay a factor @xmath533 ( to find @xmath523 and @xmath521 ) but again we can take a fraction of the factor of @xmath531 to do so .
when tadpole elimination has been completed , we have erased @xmath534 @xmath455-tadpoles and @xmath535 @xmath456-tadpoles and we are left with @xmath536 vertices linked together by @xmath537 @xmath455 s and @xmath537 @xmath456 s ( a tadpole which has not been erased being seen as a propagator ) . for a @xmath538 ,
it is quite easy to see that to integrate on @xmath57 with fixed @xmath36 amounts more or less to the same problem for @xmath539 and that to find @xmath540 knowing @xmath462 costs a factor @xmath541 .
[ sconserv ] + _ let ( @xmath542 , , @xmath543 ) be a quadruplet of sectors of the enlarged slice @xmath390 and @xmath544 such that there are @xmath545 ,
, @xmath546 verifying @xmath547
_ then we can find @xmath548 satisfying @xmath549 where @xmath550 and @xmath551 are some constants independent of @xmath227 and @xmath552 .
_ proof _ + if we can prove the result for @xmath553 then we will be able to enlarge the result to any @xmath227 provided maybe we take some slightly bigger @xmath550 and @xmath551 .
therefore we assume that this is the case in the following .
we define @xmath554 by * @xmath555 * @xmath556 * @xmath557 * @xmath558 then , if @xmath559 we exchange @xmath560 and @xmath561 .
a sector @xmath562 is included in a tube , of center @xmath563 and whose direction is orthogonal to the direction @xmath564 , of size @xmath565 we define @xmath566 if we can prove that @xmath549 then we will be able to conclude , with @xmath567 and @xmath568 .
it is easy to check that by construction , we have * @xmath569 * @xmath570 we have a trivial bound @xmath571 therefore @xmath572 we can see that @xmath35 is very well conserved .
if @xmath573 then @xmath574 . otherwise , let us remark that @xmath575 is at a distance at most @xmath576 from a rhombus @xmath577 of center @xmath578 and of diagonals @xmath579 then , @xmath580 is at a distance at most @xmath581 from a rectangle @xmath582 of center @xmath583 and of sides @xmath584 since @xmath585 , we have @xmath586 . we define a @xmath58 axis in the direction @xmath587 .
@xmath588 has a @xmath58 coordinate @xmath589 .
this leads to the condition @xmath590 let us note that @xmath591 is an increasing function of @xmath592 and that we have @xmath593 we take @xmath594 .
@xmath595 since we are in the case @xmath596 , for @xmath597 and @xmath598 large enough we will have @xmath599 therefore we must have @xmath600 which allows us to conclude .
the previous section shows that , at each vertex , momenta come approximately by pairs of opposite sectors .
thus , for all the vertices which havent been erased by the tadpole elimination process , we can choose by a factor @xmath601 how to pair the sectors .
then we split the vertices in two half - vertices according to this pairing .
we represent graphically this as this gives @xmath602 ( split ) graphs that we will consider as our basic graphs in the following .
a graph is decomposed into a number of _ momentum cycles _ connected together by wavy lines .
we will follow those cycles to fix momentum sectors . finding the enlarged sectors ( of level @xmath382 ) will cost a factor @xmath352 per cycle times a constant per vertex
. then we will pay an extra @xmath603 for each @xmath456 propagator to find its sector .
we define @xmath604 the total number of momentum cycles that we decompose into @xmath605 tadpoles , @xmath551 bubbles ( with 2 vertices ) and @xmath227 large cycles ( with 3 or more vertices ) .
we have @xmath606 and the sector attribution costs @xmath607 notice that the constant has an exponent @xmath405 because of the tadpole elimination process .
the spatial integration of the vertices will be made with the short @xmath455 links whenever possible .
we can decompose each graph into @xmath455-cycles linked together by @xmath456 links ( because there are @xmath270 incoming @xmath455 s at each vertex ) , this allows to integrate all the vertices but one per cycle with a @xmath455 link .
the total cost is ( noticing that the last vertex is integrated in the whole cube @xmath46 ) @xmath608 where @xmath609 is the total number of short cycles that we decompose into @xmath610 short tadpoles , @xmath611 short bubbles and @xmath612 large short cycles . @xmath613
the scaling of the tadpoles and the propagators give a factor @xmath614 tadpoles that have been obtained by erasing a few vertices ( say @xmath615 for instance ) will have an extra small factor because they strongly violate momentum conservation , we can take it to be a power of @xmath616 .
tadpoles with higher weights will not have this good factor but we will see that they bring a better combinatoric .
the @xmath605 momentum tadpoles will consist in @xmath617 low weight ones and @xmath618 others while the @xmath610 short tadpoles split into @xmath619 low weight ones and @xmath620 others .
we can manage to have a factor @xmath621 if we have a short bubble we will have four incoming long propagators whose momenta must add up to zero up to @xmath622 .
if we apply lemma [ sconserv ] , we can see that knowing 3 of these momenta it cost only a factor @xmath623 to find the fourth momentum sparing us a factor @xmath603 obtained by navely fixing first the enlarged sector at slice @xmath382 .
if the bubble has a weight @xmath59 , _ i.e. _ the two short propagators have been obtained after erasing @xmath59 vertices , and a momentum conservation worse than @xmath624 then the small factor of bad momentum conservation will pay for the @xmath625 .
we will have @xmath626 such good bubbles , each of them bringing a factor @xmath627 .
in addition , we will have @xmath628 bad bubble of weight greater than @xmath629 for which we earn nothing and @xmath630 bad bubbles of weight @xmath631 bringing a factor @xmath632 .
this gives a factor @xmath633 finally we have the following bound for the contribution of a graph @xmath634 if we use equations ( [ momcycle ] ) and ( [ shortcycle ] ) we obtain @xmath635 we will take @xmath636 so that @xmath637 .
furthermore , @xmath605 and @xmath610 are at most equal to @xmath638 thus @xmath639 .
it allows us to rewrite the bound @xmath640 + let @xmath641 be the number of ways to contract @xmath642 adjacent @xmath3 s so as to make only generalized tadpoles .
we have @xmath643 _ proof _ + it is easy to see that a good contraction scheme , _
i.e _ one that gives only generalized tadpoles , corresponds to have no crossing contractions .
it means that if we label the fields @xmath644 according to their order and if @xmath645 and @xmath646 contract respectively to @xmath647 and @xmath648 then @xmath649 we have @xmath650 . for @xmath651
, we contract first @xmath652 to some @xmath645 .
@xmath653 will necessarily contract among themselves making only generalized tadpoles and so will do @xmath654 .
thus @xmath655 is necessarily odd and we have @xmath656 where by convention @xmath657 .
we introduce the generating function @xmath658 the recursion formula ( [ trecursion ] ) can be translated into an equation for @xmath605 which is @xmath659 whose resolution yields @xmath660 since the second solution is analytic around @xmath661 , we can take it as @xmath662 and the coefficients of its power expansion will give us @xmath641 .
an easy computation leads to the desired formula .
@xmath130 + the number @xmath663 of graphs with @xmath254 possible momentum bubbles obtained in the contraction of a cycle of @xmath664 @xmath3 s has the following bound @xmath665 _ proof _ + first lets us remark that bubbles come in chains ( possibly with a tadpole at one end ) of two possible types * type 1 : * type 2 : where a solid line stands here either for a @xmath455 or a @xmath456 .
we have two special cases * which can be seen as a type 1 chain * which can generate only one momentum bubble so that we can see it as a type 2 chain of length 1 . having chosen the @xmath3 s there are only two contraction schemes that yield a type 1 chain and a unique contraction scheme for a type 2 chain .
if we fix explicitly the subgraphs corresponding to @xmath254 bubbles and contract the remaining @xmath3 s in any way we will get all the desired graphs plus some extra ones so that we can bound @xmath663 .
we construct @xmath666 type 1 chains of lengths @xmath667 , , @xmath668 and @xmath669 type 2 chains of lengths @xmath670 ,
, @xmath671 .
we set @xmath672 to count the contraction schemes , first we cut the cycle of @xmath664 @xmath3 s into a sequence of @xmath3 s ( there are @xmath664 ways to do so ) . then it is easy to check that in order to build a type 1 chain we must choose two sets @xmath673 and @xmath674 of @xmath675 adjacent @xmath3 s while for a type 2 we need a set @xmath676 of @xmath677 adjacent @xmath3 s .
we distribute those @xmath678 objects in @xmath679 boxes in an ordered way , and for the @xmath680 type 1 chain the respective order of @xmath673 and @xmath674 will fix the contraction scheme . then , there remain @xmath681 @xmath3 s to contract so that we have the following number of configurations @xmath682 we can compute this @xmath683!}{[2(m - b)-2r_{1}-r_{2 } ] ! ( 2r_{1 } ) ! r_{2 } ! } \nonumber \\ & & \frac{(2 r_{1})!}{r_{1 } ! } \frac{[2(m - b - r_{1})]!}{2^{m - b - r_{1 } } ( m - b - r_{1 } ) ! } \\ & \leqslant & 2 m \!\ !
\sum_{{r_{1}\leqslant b_{1 } } \atop { r_{2}\leqslant b_{2 } } } { b_{1}-1 \choose r_{1}-1 } { b_{2}-1 \choose r_{2}-1 } 3^{2(m - b ) } 2^{2r_{1 } } r_{1 }
! 2^{m - b - r_{1 } } ( m - b - r_{1 } ) ! \\ & \leqslant & 2 m \!\ !
\sum_{b_{1}+b_{2}=b } 2^{b_{1}-1 } 2^{b_{2}-1 } 9^{m - b } 2^{m } ( m - b ) ! \\ & \leqslant & 2 m ( b+1 ) 18^{m } ( m - b ) ! \end{aligned}\ ] ] @xmath130 now we can achieve the proof of lemma [ lemproba ] in bounding @xmath684 in order to compute this sum , we fix first @xmath534 , @xmath535 and @xmath685 , where @xmath685 is the number of possible momentum bubbles and therefore is greater than @xmath551 .
then we define the set @xmath686 has the set of graphs with the corresponding @xmath534 , @xmath535 and @xmath685 and for which the erased tadpoles form @xmath185 sets of @xmath687 adjacent @xmath3 s .
we can write @xmath688 to bound @xmath689 we notice that when a graph has a bad bubble of weight @xmath631 it means that we have erased two set of generalized tadpoles @xmath690 and @xmath691 on the two propagators of the bubble with @xmath692 .
thus we have a corresponding factor @xmath693 which control the bad factor @xmath694 of the bad bubble so that @xmath695 the number of graphs in @xmath696 has the following bound @xmath697 \leqslant 2^{m_{0 } } \prod t(q_{i } ) 3^{m'_{0 } } c^{m'_{0 } } ( m'_{0}-\bar{b } ) ! \leqslant c^{m_{0 } } \prod \frac{2^{2q_{i}}}{q_{i}+1 } ( m'_{0}-\bar{b})!\ ] ] this leads to @xmath698 summing on @xmath534 and @xmath535 is equivalent to sum over @xmath537 and @xmath535 with @xmath699 . the sum over @xmath685 is roughly evaluated by taking the supremum over @xmath685 , the result depends whether @xmath537 is greater than @xmath700 or not .
@xmath701 \\ & \leqslant & c^{m_{0 } } m^{-m_{0}j } \sum_{m'_{0 } , t_{k } } m^{-m'_{0}(k - j)/2 } m^{-t_{k}(k - j ) } \left [ 1 + 1_{(m'_{0 } > m^{j/6 } ) } m^{-m'_{0}j/6 } ( m'_{0}-\bar{b } ) ! \right]\end{aligned}\ ] ] finally , the sum over @xmath535 is easy and we bound the sum over @xmath537 by finding the supremum .
one can check that it gives the announced result .
understanding the effect of perturbations on the free spectrum of hamiltonian operators is an outstanding challenge .
we think that for the two - dimensional case , this paper can help to control the model up to quite close to the scale of the expected `` mass '' ( _ i.e. _ imaginary part ) so that in studying the full model , one can focus on the thin slice of momentum @xmath702 .
then we think that a key of the problem lies in the fact that in this case the potential is very close ( in momentum space ) to a large hermitian random matrix with almost independent entries and therefore we can connect our problem to the much better understood domain of random matrices or equivalently use the vector model picture of the problem .
one can note that in dimension @xmath99 , the interpretation in term of random matrices can also be helpful ( cf .
@xcite ) , but then one has to deal with constrained matrices whose entries are no longer independent .
this paper is part of a larger program in collaboration with j. magnen and v. rivasseau to apply rigorous renormalization group methods to the study of disordered systems , the ultimate goal being to devise tools to construct extended states in weakly disordered systems in dimension @xmath703 .
they played an important part in getting these results and i thank them very much for their help .
fmrt2 m. aizenman , in _ the state of matter _ , world scientific ( 1994 ) 367 a. abdesselam and v. rivasseau `` an explicit large versus small field multiscale cluster expansion '' ( hep - th/9605094 ) to appear in _ reviews in math .
_ d.brydges , `` a short course on cluster expansion '' , in _ critical phenomena , random systems , gauge theories _
, les houches session xliii , 1984 , elsevier science publishers ( 1986 ) j.m .
combes and l. thomas `` asymptotic behaviour of eigenfunctions for multiparticle schrdinger operators '' , _ comm .
* 34 * ( 1973 ) 251 - 270 j.feldman , j. magnen , v. rivasseau and e. trubowitz , in _ the state of matter _ , world scientific ( 1994 ) 293 j.feldman , j. magnen , v. rivasseau and e. trubowitz , _ europhys .
* 24 * ( 1993 ) 521 e. fradkin , in _ dveloppement rcents en thorie des champs et physique statistique _ , les houches xxxix , north holland ( 1984 ) m.l .
mehta , _ random matrices _ , academic press , boston ( 1991 ) j. magnen , g. poirot and v. rivasseau `` the anderson model as a matrix model '' ( cond - mat/9611236 ) to appear in _ advanced quantum field theory _ , la londe les maures ( in memory of c. itzykson ) elsevier ( 1996 ) g. parisi , in _ dveloppement rcents en thorie des champs et physique statistique _ , les houches xxxix , north holland ( 1984 ) v. rivasseau , `` cluster expansions with large / small field conditions '' , minicourse in vanvouver summer school ( 1993 ) v. rivasseau , _ from perturbative to constructive renormalization _ , princeton university press ( 1991 )
thouless , _ phys . report _
* 13 * ( 1974 ) 93 | in this paper we develop a polymer expansion with large / small field conditions for the mean resolvent of a weakly disordered system . then we show that we can apply our result to a two - dimensional model , for energies outside the unperturbed spectrum or in the free spectrum provided the potential has an infra - red cut - off .
this leads to an asymptotic expansion for the density of states .
we believe this is an important first step towards a rigorous analysis of the density of states in the free spectrum of a random schrdinger operator at weak disorder .
= -0.54 truecm = -0.54 truecm = 0 truecm = 0 truecm = 0 truecm = 0 truecm = 23 truecm = 17 truecm amssym.def amssym.tex |
mathematical models are used throughout infectious disease epidemiology to understand the dynamical processes that shape patterns of disease @xcite . while early models did not include complex population structure , modeling approaches now frequently let the epidemic spread take place on a network , which enables greater realism than a model in which all individuals mix homogeneously .
it does , however , pose many technical problems for model analysis , particularly the question of how heterogeneity in the number of links each individual participates in their degree influences the epidemic @xcite . similarly , even though some of the earliest mathematical models of infectious diseases were stochastic , accounting for the chance nature of transmission of infection @xcite , much of the applied modeling that followed was deterministic and based on non - linear differential equations @xcite .
more recent applied work has , however , recognized the importance of using stochastic epidemic models @xcite and also of the development of associated methodology @xcite .
the difficulty in mathematically analyzing models which include both stochastic elements and network structure can be a reason for not including these factors , but we prefer to include them , and subsequently to systematically reduce the complexity of the resulting model .
this is the approach we adopt in this paper ; the reduction process being made possible because of the existence of a separation of time - scales : many variables `` decaying '' away rapidly , leaving a few slow variables which govern the medium- to long - term dynamics . compartmental models of epidemics typically assume that the majority of the population starts susceptible to infection , with a small number infectious , who then spread infection to others before recovering .
a key distinction is between susceptible - infectious - susceptible ( sis ) dynamics in which recovery returns an individual to the susceptible compartment and susceptible - infectious - recovered ( sir ) dynamics in which recovery removes an individual from a further role in the next epidemic @xcite , with the former being used to model e.g. sexually transmitted infections other than hiv and the latter e.g.pandemic influenza @xcite . in the theoretical analysis of epidemic models , a crucial quantity corresponds to the basic reproductive ratio @xmath0verbally described as the expected number of secondary cases infected per primary case early in the epidemic @xcite . depending on the value of @xmath0 , either @xmath1 individuals will experience infection in a population of size @xmath2the model is then said to be supercritical or @xmath3 individuals will experience infection in a population of size @xmath2the model is then subcritical thus defining an epidemic threshold . in reality , the contact network on which the disease spreads will often change over the course of the epidemic ; while approaches exist in which the dynamics of both the disease and the network are considered @xcite , it is more common to consider two limiting cases .
the first of these is a _
static _ approach ( also called ` quenched ' ) in which the network is assumed to evolve much more slowly than the disease and which is typically approached analytically through the use of pair approximation and related techniques @xcite .
the second is a _ dynamic _ limit ( also called ` annealed ' or ` discrete heterogeneous ' ) in which the network is assumed to evolve much more quickly than the epidemic , and can be described by an effective network characterized by its degree distribution , in which all individuals sharing the same degree are considered equivalent .
this case can be analyzed through use of a set of degree - indexed differential equations ( often called the ` heterogeneous mean field ' approach ) provided the maximum degree is not too large @xcite . when the distribution of degrees in the population is highly variable a situation that appears to be supported empirically @xcite it was recognized that the epidemic may not exhibit straightforward critical behavior .
this happens because as the population size @xmath2 becomes large , extremely small levels of infectiousness can lead to large epidemic sizes @xcite or more accurately speaking the critical level of infectiousness can depend very sensitively on the largest degree , @xmath4 , in the network @xcite . the behavior of highly heterogeneous network epidemics near criticality continues to generate interest in both the physics and mathematics literature @xcite . in this paper
we investigate the stochastic behavior of heterogeneous network epidemics over time .
we study a network in the dynamic limit , characterized by a power - law degree distribution such that the probability of an individual having degree @xmath5 is given by @xmath6 , with @xmath7 , although the method of analysis is applicable to other distributions .
this type of network has the property that the basic reproductive ratio , which is found to be proportional to the second moment of the degree distribution @xmath8 , diverges in the limit of infinite populations , leading to the absence of an epidemic threshold .
this is evidently not the case when the population and thus the degree cutoff is finite , but the second moment of the distribution can still be extremely large for sufficiently large @xmath2 , and heterogeneity can play an important role in the dynamics of the system . for the case of large but finite population size , we derive a two - dimensional stochastic differential equation ( sde ) approximation to full sir dynamics , which reduces to an analytically solvable one - dimensional system early in the epidemic .
we perform simulations using a power - law degree distribution with a maximum degree cutoff @xmath4 , which show that our approach provides a good approximation provided @xmath4 is not too close to the population size .
there have , to our knowledge , been no directly comparable studies of this kind . perhaps those closest have been , first the study of finite - size scaling of the sis model near the critical point to produce a phenomenological one - dimensional sde @xcite , second , the derivation of a one - dimensional sde for the sis model from first principles , but using an adiabatic approximation that would only be expected to hold near the critical point @xcite , and finally the study of a four - dimensional sde for the sir model on a static network that is much more complex than ours and less amenable to analysis @xcite .
the outline of the paper is as follows . in sec .
[ sec : formulation ] we introduce the model , and formulate it first at the level of individuals ( the microscale ) and then at the level of populations , but retaining a stochastic element ( the mesoscale ) .
we begin to explore the reduction of the mesoscopic - level sdes in sec . [
sec : det ] and complete the process in sec .
[ sec : fastvarelim ] . in sec .
[ sec : cir ] we perform a further reduction which allows us to make contact with a model which has already been discussed in the literature . in sec .
[ sec : size ] these reduced models are used to find the distribution of epidemic sizes , and we conclude with a discussion of our results in sec . [
sec : discussion ] .
there are two appendices where technical details relevant to secs .
[ sec : det ] and [ sec : fastvarelim ] have been relegated .
in this section we formulate the model first at the microscale , that is , in terms of individuals , and from this derive a mesoscopic version of the model where the variables are continuous , and represent the fractions of types of individuals who are infected or who are susceptible to infection .
the general procedure to do this is reviewed in ref .
@xcite , and has been previously applied to models of epidemics on networks @xcite , albeit in situations where the nodes of the network contained many individuals , rather than just one , as in the present context .
as discussed in the introduction , individuals are located at the nodes of a network through which they are exposed to infection from individuals at neighboring nodes .
since we work in the dynamic limit , individuals do not alter their number of contacts when acquiring the infection or recovering from it , and the network plays no role other than encoding the information about how many connections to other individuals a given individual possesses .
an individual is labeled according to whether it is ( i ) susceptible , infected or recovered , and ( ii ) the degree of the node where it is located .
thus the variables at the microscale are @xmath9 , @xmath10 and @xmath11 , @xmath12 .
we will assume that the population is closed , i.e. , that @xmath13 , with total population @xmath14 , at all times , so that we can remove one of the variables : @xmath15 .
two types of events can occur : the infection of a susceptible individual , whenever an individual of type @xmath16 comes in contact with one of type @xmath17contacts between individuals @xmath18 and @xmath19 having degrees @xmath20 and @xmath21 , respectively , are governed by a poisson process with rate @xmath22 ; and the recovery of an infectious individual with rate @xmath23 . in this paper we will be interested in the properties of an epidemic that takes place over a much shorter timeframe than demographic processes and hence will ignore the birth and death of individuals , leaving the only interactions between the individuals to be : 1 .
infection of an individual at a node of degree @xmath5 .
the transition rate for this process is given by @xmath24 where the summation is over infected individuals labeled according to the degree of the node on which they are located .
the only arguments of @xmath25 which are shown explicitly are those involving individuals on nodes of degree @xmath5 , since only these change in this process ; those on the right represent the initial state , and those on the left represent the final state .
2 . recovery of a infected individual at a node of degree @xmath5 .
this proceeds at a rate given by @xmath23 , and so the transition rate for this process is given by @xmath26 the dynamics of the process can be described by writing down an equation for the probability that the system is in a state @xmath27 at time @xmath28 , which will be denoted by @xmath29 .
this is the master equation @xcite , which equates @xmath30 to the net rate of increase of @xmath31 due to the processes given by eqs .
( [ eqn : rate_1 ] ) and ( [ eqn : rate_2 ] ) , and is given by @xmath32 \nonumber \\ & \quad - \sum^k_{k=1 } \left[t_1(s_k-1,i_k+1\vert s_k , i_k)+t_2(s_k , i_k-1\vert s_k , i_k)\right]p(s_k , i_k , t ) .
\label{eqn : master}\end{aligned}\ ] ] once again , the dependence of the probability distribution , @xmath31 , on the state variables @xmath33 has been suppressed in the sum over @xmath5 on the right - hand side of this equation .
we will refer to the discrete model defined by eqs .
as the full microscopic model , since we we shortly derive a mesoscopic model from it , and subsequently further reduce this model .
once an initial state has been specified , the model is completely determined , and in principle @xmath34 can be found for any state @xmath27 at any time @xmath28 .
this behavior can be explored numerically _ via _ monte carlo simulations ( e.g. using gillespie s method @xcite or other approaches @xcite ) however our main interest in this paper is in approximating the master equation to obtain models which are more amenable to analysis . a mesoscopic description of the process involves going over to new variables @xmath35 and @xmath36 , @xmath12 , which are assumed continuous for large @xmath2 , but retaining the stochastic nature of the system by keeping @xmath2 large , but finite .
a macroscopic description , on the other hand , would involve taking the limit @xmath37 , which renders the process deterministic , and which eliminates all effects of the original discrete nature of the system .
the construction of the mesoscopic model involves expanding the master equation , eq .
, in powers of @xmath38 and truncating this expansion at order @xmath39 .
this is discussed in many places in the literature , but here we follow ref .
@xcite , which gives explicit forms for the functions @xmath40 and @xmath41 , @xmath42 , which define the mesoscopic model .
carrying out this procedure , the master equation becomes a fokker - planck equation @xcite of the form @xmath43+\frac{1}{2n^2}\sum^k_{k=1}\sum\limits_{\mu,\nu = 1}^2\frac{\partial ^2}{\partial x^{(k)}_\mu \partial x^{(k)}_\nu}\left[b^{(k)}_{\mu \nu}(\bm{x } ) p\left(\bm{x } , t\right)\right ] , \label{eqn : fpe}\end{aligned}\ ] ] where @xmath44 ; @xmath45 , @xmath46 .
we have also introduced @xmath47 .
the explicit form of the functions @xmath40 and @xmath41 are @xcite @xmath48 where @xmath49 are the original transition rates ( scaled by @xmath2 ) given in eqs .
( [ eqn : rate_1 ] ) and ( [ eqn : rate_2 ] ) , but written in terms of the continuous state variables @xmath50 and @xmath51 .
the final stage of the construction of the mesoscopic form of the model is to adopt a coarser time scale by introducing a new time variable @xmath52 . from eq .
( [ eqn : fpe ] ) it can be see that without this change of variable , the macroscopic ( @xmath37 ) limit eliminates the right - hand side of the equation , giving a trivial macroscopic limit . with the change of variable , the first term on the right - hand side survives ( but without the factor @xmath53 ) , giving a liouville - like equation for @xmath31 , which implies a non - trivial deterministic dynamics .
a clearer way to see this , and in fact a more intuitive formulation of the mesoscopic dynamics to that of the fokker - planck equation , is to use the sde which is equivalent to eq .
( [ eqn : fpe ] ) .
we utilize the general result that a fokker - planck equation of the form given by eq .
( [ eqn : fpe ] ) is equivalent to the sde @xmath54 , where the @xmath55 are gaussian random variables with zero mean and a correlation matrix given by the matrix @xmath56 . in terms of the variables
@xmath50 and @xmath51 , the sdes take the form @xmath57 where @xmath58 here the noise is to be interpreted in the sense of it @xcite .
it is clear from eqs .
( [ eqn : sk ] ) and ( [ eqn : ik ] ) that the limit @xmath59 eliminates the noise terms leaving deterministic equations , and we recover eq .
( 2.6 ) from ref .
@xcite .
it is convenient to make the noise in the sdes explicitly multiplicative .
since the rates @xmath60 are non - negative we may introduce new functions @xmath61 , @xmath62 , and then transform to new noise variables @xmath63 it is straightforward to check that @xmath64 , and therefore that all the state dependence has been transferred from the correlator to the sdes . equations correspond to the full mesoscopic model . in order to describe the evolution of the entire population , we need to consider the sdes for all @xmath65 ; that is , we have to work with @xmath66 equations . in this paper
, we will be interested in networks where individuals can pick degrees from a truncated zipf distribution of the form @xmath67 and we let @xmath68
. since we are interested in the case when @xmath4 is large , the number of equations making up the full mesoscopic model are so large that it can make any direct treatment impractical .
therefore , in the following sections we will be concerned with reducing the number of equations , and we start by looking at the deterministic limit of the system .
this will also have a crucial role to play in the reduction of the stochastic model , described in sec .
[ sec : fastvarelim ] .
the deterministic limit which controls the dynamics of the macroscopic version of the model is found by taking the @xmath37 limit of eqs .
, and so eliminating the noise terms . for notational convenience
we reorder the entries of the state vector , so that @xmath69 , and write eqs . in the form @xmath70 . to find the fixed points of the dynamics we set the time derivatives on the left - hand side of these equations to zero ;
these fixed points will be denoted by an asterisk and obey @xmath71 .
it is immediately clear that there is a set of stable fixed points with @xmath72 and @xmath73 undetermined , for all @xmath5 , corresponding to a disease - free state . for a given initial condition ,
the fixed point is uniquely determined , however it may not be stable . to investigate the stability of the fixed point @xmath74 we perform a linear stability analysis .
the @xmath75 stability matrix @xmath76 is defined so that near @xmath77 , the dynamics are @xmath78 , and is found to have the form @xmath79 where @xmath80 is a @xmath81 matrix with entries @xmath82 and @xmath83 is the @xmath81 unit matrix
. the eigenvalues and eigenvectors of @xmath84 are straightforward to determine .
any vector which has the last @xmath4 entries equal to zero is an eigenvector with eigenvalue zero , reflecting the degeneracies of the system .
similarly a vector with the first @xmath4 entries set equal to zero and the last @xmath4 entries equal to @xmath85 , @xmath12 , is an eigenvector with eigenvalue @xmath86 if @xmath87 , that is , if @xmath88 .
these then yield another degenerate set of @xmath89 eigenvalues @xmath86 .
the final eigenvalue can be found by equating the trace of @xmath84 to the sum of its eigenvalues . in summary ,
the @xmath90 eigenvalues are @xmath91 while a suitable set of corresponding eigenvectors are listed in appendix [ sec : eve ] . from eq .
( [ eqn : eva_full ] ) we see that the condition for the fixed points to be ( marginally ) stable is @xmath92 as mentioned in the introduction , there are two possible outcomes for the deterministic dynamics depending on the parameter values : either a negligible or a significant number of individuals in the population will get infected during the course of the epidemic .
the quantity that determines which of these cases we are in is often referred to as the basic reproductive ratio ; the final size of the epidemic , in turn , can be obtained from the @xmath93 coordinates of the fixed point , @xmath94 . in order to determine these in the general heterogeneous case , we look first at the homogeneous case , where @xmath95 .
if we set an initial condition with only a few infective individuals in an almost completely susceptible population , we will have @xmath96 and @xmath97 , where @xmath98 and @xmath99 ; with this , the @xmath37 limit of eq .
( [ eqn : ik ] ) takes the form @xmath100 , early in the epidemic . from this
we obtain @xmath101 where @xmath102 corresponds to the basic reproductive ratio .
this defines an epidemic threshold , such that for @xmath103 the number of infectives grows exponentially in the early stages of the epidemic , while for @xmath104 the epidemic does not take off and its final state corresponds to @xmath105 . for the case
when @xmath103 , we can find the @xmath93 coordinate of the fixed point by dividing eq . by eq . with @xmath95
; this yields @xmath106 if we set again an initial condition @xmath107 , with @xmath108 , we can integrate the equation above to find @xmath109 where we have used that the final state of the epidemic corresponds to @xmath110 .
therefore , to a very good approximation , we can find @xmath93 coordinate of the fixed point as the smallest solution to @xmath111 note that @xmath112 is always a solution , and the only one for @xmath104 .
the total epidemic size will be given by @xmath113 , which corresponds to the fraction of recovered individuals at the end of the outbreak .
this is due to the fact that the initial number of recovered individuals is zero , and that every individual that gets infected during the course of the epidemic will eventually recover and remain permanently recovered .
now , a significant simplification can be made in the deterministic limit of the heterogeneous case , which reduces the @xmath90 dynamical equations to @xmath114 equations ( see refs .
@xcite , which consider the static case ) .
this is achieved by use of the ansatz @xmath115 ; we also define @xmath116 such that @xmath117 . with these two variables , eqs .
, in the case when @xmath118 , can be rewritten independently of @xmath5 as @xmath119 where we have introduced @xmath120 .
if we also introduce the probability generating function of the degree distribution @xmath121 , then @xmath122 . in this reduced model , there is a line of stable fixed points , uniquely determined by the initial conditions , which satisfy @xmath123 .
the stability matrix for these fixed points is the @xmath124 matrix @xmath125 which has a similar form to the stability matrix for the full system given by eq .
( [ eqn : j_2k_defn ] ) .
once again , an eigenvector exists which has the last entry equal to zero .
this corresponds to an eigenvalue zero , reflecting the existence of a line of fixed points .
the trace of @xmath126 then gives the second eigenvalue , and the two eigenvalues are thus @xmath127 with the eigenvectors given in appendix [ sec : eve ] .
the stability condition is now @xmath128 now we can find analogous expressions to those in eqs . and for the heterogeneous case .
we start by noting that , in terms of @xmath129 , the total number of susceptible individuals will be given by @xmath130 it is clear that if @xmath131 , then @xmath132 ; conversely , if @xmath133 then @xmath134 , and so @xmath135 .
this means that a completely susceptible population will correspond to @xmath129 being equal to unity .
a small initial number of infectives will , in turn , correspond to @xmath136 .
in addition from eq .
( [ eqn : lambda_det ] ) , @xmath137 , where @xmath138 , at early times .
therefore @xmath139 from eq . (
[ eqn : lambda_early ] ) it is clear that there is a critical value of @xmath140 equal to @xmath141 ; if @xmath140 is larger than @xmath141 the infection grows , and if it is below @xmath141 it does not . thus @xmath140 is the heterogeneous basic reproductive ratio , generalizing @xmath142 in the homogeneous case .
central to our investigation is the fact that this quantity does not have a deterministic threshold for networks characterized by power - law degree distributions in the case of interest , i.e. , when @xmath7 . in order to find the heterogeneous final size ,
we proceed again as in the homogeneous case ; dividing eq . by eq .
, we find @xmath143 from this , the heterogeneous analog of eq .
is given by @xmath144 where we again set an initial condition with a very small number of infectives , @xmath145 , in an almost completely susceptible population , @xmath146 , and the final state is given by @xmath123 .
the @xmath129 component of the fixed point will then correspond to the smallest solution to @xmath147\right\rbrace , \label{eqn : theta_fix}\end{aligned}\ ] ] with @xmath140 the heterogeneous basic reproductive ratio defined above .
the epidemic size , by definition , will be given by @xmath148 .
furthermore , the @xmath50 components of the fixed point correspond to @xmath149 .
our interest in this paper is in the ( finite @xmath2 ) stochastic dynamics of the model , where a complete reduction of the kind that led to eqs .
( [ eqn : theta_det ] ) and ( [ eqn : lambda_det ] ) is not possible . in order to see this , we proceed in the same manner as in the deterministic case :
we start from eq .
, this time without taking the @xmath150 limit , and use the ansatz @xmath151 ; summing over @xmath5 and dividing by @xmath152 , we arrive at @xmath153 where we have introduced @xmath154^{-1}\sum_k \sigma^{(k)}_1 \rho^{(k)}_1(\tau)$ ] . since the noises @xmath155 have zero mean and are independent of each other
, we can rewrite @xmath156 as @xmath157 where @xmath158 has zero mean and unit variance , and @xmath159^{-2}\sum_k \left[\sigma^{(k)}_1\right]^2=\left[g^\prime ( \theta)\right]^{-1}\beta \lambda \theta$ ] . with this
, we arrive at a reduced equation for @xmath129 given by @xmath160 as in the deterministic case , then , we can express the @xmath4 equations for the @xmath50 variables , described by eq .
, entirely in terms of @xmath129 and @xmath161 _ via _ eq . .
however , we will see that the same does not hold for the dynamics of the @xmath51 variables , as we have anticipated . in a similar fashion as with the @xmath50 variables , we take the equation for the @xmath51 variables , eq . ,
multiply it by @xmath5 and sum over @xmath5 , to obtain @xmath162 let us now denote the noise terms appearing in the equation for @xmath161 above as @xmath163 and @xmath164 .
these two noise terms , like @xmath156 , have zero mean ; also , it can be readily verified that the correlations between all three of them are given by @xmath165 , where now @xmath166 , with @xmath167,\qquad \bar{b}_{33}=\gamma \sum_k k^2 i_k,\end{aligned}\ ] ] and @xmath168 , where @xmath169 .
we can now , as in sec .
[ sec : formulation ] , express @xmath170 and @xmath171 in terms of three independent normally - distributed random variables , @xmath172 and @xmath173 , with zero mean and @xmath174 , @xmath166 .
the definition of @xmath158 in terms of @xmath156 has already been given by eq .
( [ eqn : xi_to_zeta ] ) ; the other two are : @xmath175 where @xmath176 and @xmath177 . substituting all this back into eq .
yields @xmath178 . \label{eqn : lambda_stoch}\end{aligned}\ ] ] the main feature we note from this is that the two - dimensional system of equations can not be closed .
this is so because @xmath179 is still dependent on @xmath51 for all values of @xmath5 , and so it is not possible to achieve a complete reduction of the stochastic system .
therefore , when we take the intrinsic noise present in the system into account , a partially reduced model of @xmath180 equations eq . for @xmath129 , with @xmath181 , and eq . for @xmath51is the best we can achieve . due to this , the exploration of the epidemic dynamics for extremely heterogeneous cases , in which @xmath4 can take very large values , is rendered impractical : even after reducing the number of equations , we will still be left with a very - high - dimensional system .
however , progress can be made by observing that , under certain conditions , the deterministic limit of the model presents a separation of time - scales , as we shall see in the following section .
we will end this section by carrying out this partial reduction in the deterministic case , as a preparation for the stochastic treatment .
this partial reduction involves applying the above reduction to the @xmath50 variables only , by writing @xmath115 , but keeping the @xmath4 @xmath51 variables .
there are now @xmath180 dynamical equations which take the form @xmath182 there is again a line of fixed points corresponding to disease - free states , @xmath183 , @xmath12 , but with @xmath129 undetermined .
the @xmath184 stability matrix is now @xmath185 where @xmath186 is a @xmath4-dimensional vector whose @xmath187-th entry is @xmath188 , @xmath80 is a @xmath81 matrix with entries @xmath189 and @xmath83 is the @xmath81 unit matrix .
the eigenvalues and eigenvectors can be found in the same way as before , the eigenvalues having the form @xmath190 with the eigenvectors given in appendix [ sec : eve ] .
the stability condition is once again given by eq .
( [ eqn : stability_cond_total ] ) .
as we have discussed , the deterministic limit of eqs . can be greatly simplified by performing a change of variables and describing the evolution of the system in terms of @xmath129 and @xmath161 .
the inclusion of demographic noise , however , hinders such a reduction of dimensionality .
it is not possible to obtain a closed two - dimensional system of equations , thus enormously complicating the analysis of the stochastic epidemic , compared to its deterministic limit . here
, we attempt to overcome this issue by exploiting the properties of the deterministic limit of the model , and by finding a way to close the equations for @xmath129 and @xmath161 in the finite-@xmath2 case , while still retaining the essential features of the full model . in order to do this , we start from eqs .
([eqn : semi_2 ] ) for the partially reduced @xmath180-dimensional deterministic system for @xmath129 and @xmath51 .
examination of the eigenvalues appearing in eq . , shows that @xmath191 .
if the ratio @xmath192 is small , then there is the potential to carry out a reduction of the system . in such case
, the deterministic dynamics would quickly evolve along the directions corresponding to the set of most negative eigenvalues towards a lower - dimensional region of the state - space , in which the system as a whole evolves in a much slower time - scale , and which contains the fixed point @xmath193 .
as illustrated in figure [ fig : l_r0 ] , showing the value of @xmath194 for different choices of @xmath4 , we would expect this time - scale separation to occur for combinations of parameters for which the final size of the epidemic is small , i.e. , for large @xmath195 . since there is one most negative eigenvalue , with a degeneracy equal to @xmath89 , after sufficient time has passed the evolution of the system
will then be constrained to a two - dimensional surface , which we will refer to as the slow subspace . with the aim of reducing the dimensionality of the problem , then , we ignore the fast initial behavior of the system and focus the analysis on the slow surface
we do so by imposing the condition that no dynamics exists along the @xmath89 fast directions , i.e. , @xmath196 where @xmath197 , and @xmath198 is the @xmath19-th left - eigenvector of the jacobian @xmath199 evaluated at the fixed point @xmath193 , with @xmath195 given by eq . .
this procedure goes under the name of adiabatic elimination , or fast - variable elimination @xcite , and it is a common practice in the simplification of deterministic non - linear dynamical systems e.g .
, systems of oscillating chemical reactions @xcite . and @xmath200 , @xmath201 , as a function of @xmath195 for different values of @xmath4 ( @xmath202 ) .
] for ease of application of the method , the left- and right - eigenvectors of the jacobian are normalized in such a way that @xmath203 these are given , respectively , by eqs . and . imposing the condition given by eq . , one finds that @xmath204 , where @xmath205 and @xmath206 is given by eq .
( [ xandy ] ) . in terms of the @xmath129 and @xmath51 variables
this reads @xmath207\,\sum^k_{l=1 } l i_l , \ \
k=2,\ldots , k , \label{eqn : slow_ik_intermediate}\end{aligned}\ ] ] and this gives the behavior of @xmath51 which limits the dynamics of the deterministic system to a slow subspace .
a more convenient form can be found by multiplying eq .
( [ eqn : slow_ik_intermediate ] ) by @xmath5 and summing from @xmath208 to @xmath209 .
this allows us to relate @xmath210 to @xmath211 to find the following @xmath89 expressions for @xmath51 : @xmath212 } k d_k \left[\gamma { \theta^*}^k-\beta \left(\phi(\theta){\theta^*}^k-\phi(\theta^*)\theta^k\right)\right ] , \ \
k=2,\ldots , k.\label{eqn : slow_ik}\end{aligned}\ ] ] it should be stressed that the @xmath89 equations ( [ eqn : slow_ik ] ) , which we may write in the form @xmath213 , @xmath205 , should be interpreted as the time - evolution that the @xmath51 need to follow , so that @xmath214 in the directions @xmath215 , @xmath205 .
this follows from the deterministic equations of motion @xmath216 and from eq .
( [ eqn : kill_fast ] ) .
thus the only temporal evolution is in the @xmath217 and @xmath218 directions .
the functional forms @xmath213 , with @xmath219 defined by eq .
( [ eqn : slow_ik ] ) , formally only hold sufficiently near the fixed point that linearization is accurate . however in fig .
[ fig : slow_km10 ] we show three - dimensional projections of the deterministic dynamics from eqs .
for an epidemic on a network characterized by a power - law degree distribution with @xmath220 and maximum degree @xmath221 , with heterogeneous basic reproductive ratio @xmath222 , which gives @xmath223 .
it appears that the solid red line , which represents the deterministic dynamics , lies in the slow subspace for most of the time , not just at late times when the system approaches the fixed point .
that this is the case can be seen from a consideration of the jacobian of the system at an arbitrary time , and not just at the fixed point : @xmath224 it is straightforward to check that @xmath225 is a right - eigenvector of the matrix in eq .
( [ eqn : j_k_plus_one_arb_time ] ) with eigenvalue @xmath86 , if @xmath226 . therefore , remarkably , the eigenvectors @xmath227 of the jacobian at the fixed point , defined in eq .
( [ eqn : eve_semi ] ) , are eigenvectors of the jacobian at all times .
the existence of this negative eigenvalue with degeneracy @xmath89 means that the system remains in the plane described by the vectors @xmath217 and @xmath218 , since any small deviations out of the plane will tend to be pushed back again . normally ,
when carrying out a fast - variable elimination , one would find that the system possesses a line of fixed points which is quickly approached at early times , dominated by the deterministic dynamics , and along which the slow , stochastic dynamics takes place .
in this case , however , the line of fixed points corresponds to an absorbing boundary of the system : the evolution of the epidemic stops altogether once it reaches the disease - free state .
therefore , it is important that the projection onto the slow variables constitutes a good approximation to the dynamics also far from the line of fixed points .
we parameterize the slow subspace using the coordinates @xmath228 and @xmath229 . using eq .
( [ eqn : leve_semi ] ) and @xmath230 , we obtain @xmath231 a linear transformation of eqs .
( [ eqn : slow_z1 ] ) and ( [ eqn : slow_z2 ] ) shows that an equivalent set of coordinates are @xmath129 and @xmath232 , previously called @xmath161 .
so , in summary , in sec .
[ sec : det ] we showed the system of @xmath90 equations leads to a closed set of equations in @xmath129 and @xmath161 . here , through a consideration of the slow modes of the deterministic equations
, we have shown that not only can we describe the system in terms of @xmath129 and @xmath161 , but that it remains in the vicinity of a subspace which can be parameterized by these variables .
( left ) and @xmath233 ( right ) for @xmath221 , @xmath220 , @xmath222 , @xmath234 .
red , dashed line : deterministic limit , eqs . ; blue , yellow , green lines : stochastic differential equations and .
the initial condition corresponds to @xmath235 , @xmath236 for @xmath205 , and @xmath237 . in both cases ,
the system is pushed back from the initial condition ( lower right in both panels ) towards the slow subspace described by eq .
, on which it seems to stay during the rest of the course of the epidemic .
note the different scales in the @xmath238 axes of both panels , so the fluctuations away from the slow subspace are small in both cases ; the fluctuations on the slow subspace can , however , be large , as seen from the departure of the stochastic trajectories from the deterministic dynamics.,title="fig : " ] ( left ) and @xmath233 ( right ) for @xmath221 , @xmath220 , @xmath222 , @xmath234 .
red , dashed line : deterministic limit , eqs . ; blue , yellow , green lines : stochastic differential equations and .
the initial condition corresponds to @xmath235 , @xmath236 for @xmath205 , and @xmath237 . in both cases ,
the system is pushed back from the initial condition ( lower right in both panels ) towards the slow subspace described by eq .
, on which it seems to stay during the rest of the course of the epidemic .
note the different scales in the @xmath238 axes of both panels , so the fluctuations away from the slow subspace are small in both cases ; the fluctuations on the slow subspace can , however , be large , as seen from the departure of the stochastic trajectories from the deterministic dynamics.,title="fig : " ] we would like to carry this reasoning over to the stochastic case , to find a reduced set of sdes , but which are closed , i.e. which can be expressed entirely in terms of the slow variables @xmath239 and @xmath240 ( or @xmath129 and @xmath161 ) .
figure [ fig : slow_km10 ] also shows three - dimensional projections of stochastic trajectories from eqs . and , for a total population of @xmath234 .
the behavior is found to be similar to that of the deterministic case : the dynamics seems to quickly reach , and then fluctuate around , the deterministic slow subspace , with the early behavior dominated by the deterministic dynamics . therefore we expect that we should be able to effectively reduce the stochastic system from @xmath180 dimensions to two dimensions , by neglecting the fast initial behavior of the epidemic .
we will do this by projecting the stochastic differential equations , eqs . and , onto the slow degrees of freedom of the deterministic limit of the model ; that is , we will neglect the fluctuations away from the slow subspace , along the @xmath89 fast directions corresponding to the most negative eigenvalue .
this can be formulated mathematically through the application of a projection operator that only picks the components of @xmath241 , @xmath242 , and @xmath243 along the slow eigenvectors of the deterministic limit , @xmath244 and @xmath245@xcite : @xmath246^\top + \bm{v}^{\{k+1\}}[\bm{u}^{\{k+1\}}]^\top .
\label{eqn : projector}\end{aligned}\ ] ] the details of the derivation are given in appendix [ sec : projection ] , where it is shown that the sdes given by eqs .
( [ eqn : theta_stoch ] ) and ( [ eqn : lambda_stoch ] ) of sec .
[ sec : det ] are recovered , except that now @xmath177 is determined in terms of @xmath129 and @xmath161 .
this closed system of stochastic differential equations for @xmath129 and @xmath161 , which we call the _ reduced mesoscopic model _
, constitutes the main result of this paper and so we reproduce it here : @xmath247 , \label{eqn : slow_lambda}\end{aligned}\ ] ] where @xmath248^{-1}\beta \lambda \theta$ ] , @xmath249 - \bar{\sigma}^2_1 $ ] and @xmath250\right ) .
\label{sigma_3_explicit}\end{aligned}\ ] ] we now compare the dynamics of the full microscopic model , defined by eqs .
, to that of the reduced mesoscopic model above .
figures [ fig : ts_full_red_3 ] and [ fig : ts_full_red_4 ] show the time series of the full microscopic model in terms of the variables @xmath129 and @xmath161 , and that of the reduced mesoscopic model , for @xmath251 and different values of @xmath4 having the same deterministic final size @xmath195 , obtained from eq . ; figure [ fig : phase_full_red_2 ] , in turn , compares the phase diagrams .
we see that , in the case of small @xmath4 , the time series of both variables in the full microscopic model are rather smooth , and are dominated by a horizontal spread due to the fact that the epidemic takes off at different moments . as heterogeneity increases ,
the dynamics of the system becomes noisier and noisier , as seen in the phase diagrams ; however , the time series of @xmath129 continues to be smooth .
this observation will be useful in the following section . from the time series of the reduced mesoscopic model we see that , except for small @xmath4 , the temporal behavior of @xmath129 and @xmath161 does not present the horizontal spread observed in the full microscopic model .
however , once the epidemic has taken off , the dynamics of the full microscopic model is correctly captured when @xmath4 is not so large ; this can be seen from the phase diagrams .
we note that the relative magnitude of the eigenvalues for this choice of parameters takes the values @xmath252 , and that the reduced mesoscopic model fares well despite @xmath194 being rather large .
we find slightly worse agreement between the full and reduced models for smaller values of @xmath195 , corresponding to larger @xmath194 .
one further detail to add is that in the case of the reduced model there are far fewer trajectories in which the epidemic does take off , compared to the full microscopic model ; that is , the reduced model results in an over - representation of very early extinctions .
note that , for the value of @xmath2 employed in these figures , @xmath253 is the maximum possible degree cutoff . in this case , we would not expect the reduced model to yield accurate results .
this is due to the fact that not even the full @xmath90-dimensional mesoscopic model ( eqs .
) provides a good approximation to the microscopic dynamics in this case , since @xmath4 is of the order of @xmath2 and this would need to be explicitly taken into account when performing the mesoscopic expansion leading to the fokker - planck equation .
we end this section with some further observations regarding the reduced stochastic system .
first , the existence of a @xmath89-fold degenerate eigenvalue equal to @xmath86 throughout the time evolution also means that stochastic trajectories , as well as deterministic ones , which venture out of the slow subspace will tend to be pushed back into the plane .
thus we would expect the reduced set of sdes to be a good approximation throughout the motion , and not just at late times .
this may be the reason why the approximation works well even if @xmath194 is not particularly small .
second , in retrospect , the elimination of the fast variables in the stochastic system is not strictly necessary , since the sdes previously obtained in sec .
[ sec : det ] are recovered , and @xmath179 could be found in terms of @xmath129 and @xmath161 using eq .
( [ eqn : slow_ik ] ) obtained from the elimination of fast variables in the deterministic system .
one can compare the value of @xmath179 in terms of the @xmath51 to the one in terms of @xmath161 in the deterministic limit and see that , indeed , the approximation is not very good for moderate to large values of @xmath194 .
this does not seem to play a huge role when considered as one of many terms in the stochastic evolution of @xmath161 .
finally , we can make a general comment about the nature of the fluctuations as the epidemic starts to grow from its initial state . in this case
the relevant eigenvalues will be those of the jacobian eq .
( [ eqn : j_k_plus_one_arb_time ] ) , but with @xmath254 and @xmath255 , and will therefore be given by those in eq .
( [ eqn : eva_semi ] ) , but with @xmath195 set equal to @xmath141 . therefore the largest eigenvalue is now @xmath256 . in this case , if @xmath257 , so that an epidemic starts to grow , the dynamics is pushed away from the line @xmath258 in its early stages , as can be seen from fig .
[ fig : slow_km10 ] where the trajectories leave the vicinity of the absorbing state after reaching the slow subspace .
+ + + + + + + + , @xmath259 and @xmath260 .
colored lines correspond realizations of the full microscopic model ( top ) and the reduced mesoscopic model ( bottom ) started with a total number of 5 infectives with degrees sampled from a truncated zipf distribution with @xmath220 and @xmath221 ( left ) , @xmath261 ( center ) , and @xmath253 ( right ) .
the black , dashed line corresponds to the deterministic solution.,title="fig:",scaledwidth=26.0% ] , @xmath259 and @xmath260 .
colored lines correspond realizations of the full microscopic model ( top ) and the reduced mesoscopic model ( bottom ) started with a total number of 5 infectives with degrees sampled from a truncated zipf distribution with @xmath220 and @xmath221 ( left ) , @xmath261 ( center ) , and @xmath253 ( right ) .
the black , dashed line corresponds to the deterministic solution.,title="fig:",scaledwidth=26.0% ] , @xmath259 and @xmath260 .
colored lines correspond realizations of the full microscopic model ( top ) and the reduced mesoscopic model ( bottom ) started with a total number of 5 infectives with degrees sampled from a truncated zipf distribution with @xmath220 and @xmath221 ( left ) , @xmath261 ( center ) , and @xmath253 ( right ) .
the black , dashed line corresponds to the deterministic solution.,title="fig:",scaledwidth=26.0% ] + + , @xmath259 and @xmath260 .
colored lines correspond realizations of the full microscopic model ( top ) and the reduced mesoscopic model ( bottom ) started with a total number of 5 infectives with degrees sampled from a truncated zipf distribution with @xmath220 and @xmath221 ( left ) , @xmath261 ( center ) , and @xmath253 ( right ) .
the black , dashed line corresponds to the deterministic solution.,title="fig:",scaledwidth=26.0% ] , @xmath259 and @xmath260 .
colored lines correspond realizations of the full microscopic model ( top ) and the reduced mesoscopic model ( bottom ) started with a total number of 5 infectives with degrees sampled from a truncated zipf distribution with @xmath220 and @xmath221 ( left ) , @xmath261 ( center ) , and @xmath253 ( right ) .
the black , dashed line corresponds to the deterministic solution.,title="fig:",scaledwidth=26.0% ] , @xmath259 and @xmath260 .
colored lines correspond realizations of the full microscopic model ( top ) and the reduced mesoscopic model ( bottom ) started with a total number of 5 infectives with degrees sampled from a truncated zipf distribution with @xmath220 and @xmath221 ( left ) , @xmath261 ( center ) , and @xmath253 ( right ) .
the black , dashed line corresponds to the deterministic solution.,title="fig:",scaledwidth=26.0% ]
in the previous section , we have derived a reduced two - dimensional system with three white - noise processes , the reduced mesoscopic model eqs . ,
which provides a good approximation to the dynamics of the full microscopic model , provided the maximum degree individuals can pick is not so large .
this is true even for relatively large values of @xmath194 , corresponding to small separation between the eigenvalues of the system .
we have also noted that the time series of @xmath129 , obtained from the full microscopic model , presents very smooth behavior over the range of parameters explored .
inspired by this observation , we attempt to further simplify the model by noting that eqs .
have the form @xmath262 + + + + taking into account the fact that , in any given realization of the process , the lowest value reached by @xmath129 will be roughly around @xmath195 , we compare the magnitudes of the noise intensities @xmath263 between @xmath264 and @xmath265 for fixed @xmath195 and different values of @xmath4 . we find that , in general , and especially during the early stages of the epidemic when @xmath266 , @xmath267 is small compared to @xmath268 and @xmath269see figure [ fig : noises ] ; the effect is more pronounced for more heterogeneous distributions with smaller deterministic final sizes , i.e. , larger @xmath195 .
therefore , the system may be described simply by a deterministic ordinary differential equation ( ode ) for @xmath129 , and eqs become @xmath270 with @xmath271 a gaussian noise with zero mean and @xmath272 , with the noise intensity , @xmath273 , given by @xmath274^{1/2}.\end{aligned}\ ] ] we call this system the _ semi - deterministic mesoscopic model _ , which is two - dimensional like the reduced mesoscopic model , but has only one white - noise process in place of the latter s three .
here we have used that @xmath275 , and @xmath276 , @xmath166 .
now , we note that eq
. corresponds to a cox - ingersoll - ross ( cir ) model @xcite @xmath277 where @xmath278 is a time - dependent parameter which is the solution of the deterministic differential equation ( [ eqn : theta_simple ] ) , @xmath279 , and @xmath280+\gamma \frac{\phi(\theta^*)+\psi(\theta^*)}{\phi(\theta^*)}}.\end{aligned}\ ] ] for fixed @xmath281 and @xmath282 , the conditional probability distribution of @xmath283 , given @xmath284 , is known in closed form , and corresponds to a non - central chi - square distribution given by @xcite @xmath285 where @xmath286 and where @xmath287 is the modified bessel function of the first kind of order @xmath288 . therefore , if we assume that @xmath129 evolves discretely , staying constant from time @xmath289 to time @xmath290 for reasonably small @xmath291 , we can use the result above to simulate eqs .
by sampling the values of @xmath161 directly from the distribution in eq . with fixed parameters @xmath292 , @xmath293 at every time - step .
the case @xmath279 corresponds to the special case of a non - central chi - square distribution with zero degrees of freedom ; thus , in order to simulate eq .
we follow the method described in refs .
@xcite and sample from a central chi - square distribution with poisson - distributed degrees of freedom , with the convention that the central chi - square distribution with zero degrees of freedom is identically zero @xcite .
figure [ fig : phase_cir ] shows the dynamics obtained in this way . while we note an additional over - representation of early extinctions , the other results are in very good agreement with the reduced mesoscopic model , eqs . . ,
@xmath259 and @xmath260 .
colored lines correspond to realizations of eqs .
, started with a total number of 5 infectives with degrees sampled from a truncated zipf distribution with @xmath220 and @xmath221 ( left ) , @xmath261 ( center ) , and @xmath253 ( right ) .
the black , dashed line corresponds to the deterministic solution.,title="fig:",scaledwidth=26.0% ] , @xmath259 and @xmath260 .
colored lines correspond to realizations of eqs .
, started with a total number of 5 infectives with degrees sampled from a truncated zipf distribution with @xmath220 and @xmath221 ( left ) , @xmath261 ( center ) , and @xmath253 ( right ) .
the black , dashed line corresponds to the deterministic solution.,title="fig:",scaledwidth=26.0% ] , @xmath259 and @xmath260 .
colored lines correspond to realizations of eqs .
, started with a total number of 5 infectives with degrees sampled from a truncated zipf distribution with @xmath220 and @xmath221 ( left ) , @xmath261 ( center ) , and @xmath253 ( right ) .
the black , dashed line corresponds to the deterministic solution.,title="fig:",scaledwidth=26.0% ]
in the previous sections , we have derived a two - dimensional reduction of the @xmath90-dimensional sir dynamics described by eqs . , by exploiting a separation of time - scales in the deterministic dynamics of the system .
we have found reasonably good agreement between the temporal behaviors of the full and reduced models .
furthermore , we have also discussed an extra simplification , consisting in neglecting the noise acting on the susceptible variable @xmath129 , and describing the system in terms of one ode and one sde instead of two sdes .
now we explore how accurately these reductions capture the distribution of epidemic sizes , given by the number of recovered individuals present in the system at the end of the epidemic , @xmath294 , which can be obtained as @xmath295 and @xmath296 in the full and reduced models , respectively , where @xmath297 and @xmath298 are the final values of @xmath50 and @xmath129not to be confused with the fixed points in the deterministic limit , @xmath73 and @xmath195 .
formal mathematical work has considered central limit theorems for models that generalize ours @xcite ; we are interested here in explicit calculation and simulation approaches .
we obtain the distribution of @xmath294 from the full model _ via _ a sellke construction @xcite , and note that it is characterized by two regions : one to the left of the interval , corresponding to early extinctions , and one to the right corresponding to the epidemic taking off .
the first thing we note when comparing this to results from the two versions of the reduced model
eqs. , and eqs .
is that , as discussed previously , there is an over - representation of very early extinctions in the case of the latter , and therefore the relative sizes of the left- and right - most regions of the distribution are not well captured .
we find that the agreement improves significantly if we condition the epidemic to take off , and only consider values of @xmath294 larger than some prescribed limit , say @xmath299that is , we only take into account realizations in which at least two individuals become infected , apart from the initial infective . with this idea in mind ,
we can go a bit further by obtaining an approximation of the height of the left - most portion of the distribution of @xmath294 .
if we denote by @xmath300 the probability that the initial infective , with degree @xmath5 , does not spread the disease to any other individuals during its _ lifetime_before recovering and , similarly , by @xmath301 the probability that it spreads the disease to one , and only one other individual with degree @xmath18 , we can write @xmath302 where we have again used the ansatz @xmath303 . the expressions are obtained by interpreting the initial behavior of the epidemic as a birth - death process , where a ` birth ' corresponds to the infection of an individual with degree @xmath18 with rate @xmath304 ; the ` death ' of the initial infective , in turn , corresponds to its recovery , which occurs with rate @xmath23 @xcite . with this , we obtain for the total probability @xmath305 , @xmath306 where we take the first infective to have degree @xmath5 with probability @xmath307 . since , initially , @xmath308 , we have @xmath309 , i.e. , the first moment of the degree distribution , which we denote by @xmath310 .
therefore , @xmath311.\label{eqn : p2}\end{aligned}\ ] ] figure [ fig : r_inf ] shows the distribution of final sizes from the reduced mesoscopic model and the cir - like model , renormalized so that the fraction of trajectories with @xmath312 is equal to @xmath313 , and compares these to the distribution from the full model .
this shows that the birth - death chain approximation we have introduced provides a simple method to adjust for the inaccuracies in early extinctions we have previously noted , yielding an accurate approximation overall
. + +
stochastic epidemics on heterogeneous networks involve two potentially large parameters .
the first of these is the size of the population , @xmath2 , which is proportional to the dimensionality of the microscopic description . for large populations , this motivates the derivation of a mesoscopic model based on a diffusion approximation whose errors are controlled by inverse powers of @xmath2 , and which we have formulated in this paper .
the second of these is the maximum degree @xmath4 , which is proportional to the dimensionality of the full mesoscopic model .
we have been able to use fast - mode elimination techniques to derive two - dimensional approximations to this scheme : the reduced mesoscopic model , which involves three white - noise processes ; and the semi - deterministic mesoscopic model , which involves only one white - noise process . our numerical results back up our theoretical understanding that these low - dimensional systems are accurate approximations to the microscopic model , except in the case where @xmath4 approaches @xmath2 , although this is to be expected since our derivation of the diffusion limit implicitly assumes that @xmath314 .
nevertheless , the reduction can be applied to cases with extremely large values of @xmath4 , as long as they are not too close to @xmath2 , thus significantly simplifying the original model .
we have observed that the elimination of fast variables in the system provides a good approximation to the full dynamics even for relatively large values of @xmath194 , corresponding to a small separation between time - scales .
we have also shown that the main inaccuracy of the reduction , relating to very early extinction probabilities of the epidemic , can be dealt with using a simple birth - death process argument . in this way
, we hope to have provided a useful tool in the study of both infectious diseases and spreading processes on heterogeneous networks .
cp - r was funded by conicyt becas chile scholarship no .
th was supported by the uk engineering and physical sciences research council ( epsrc ) .
a set of right - eigenvectors corresponding to the eigenvalues from eq . are given by @xmath315 where @xmath316 the @xmath18-th unit vector and @xmath317 a set of right - eigenvectors corresponding to the eigenvalues from eq .
are given by @xmath318 the partially reduced system with eigenvalues given by eq .
has a set of right - eigenvectors given by @xmath319 where now @xmath320 in eqs .
( [ eqn : eve_full ] ) and ( [ eqn : eve_total ] ) we did not normalize the eigenvectors .
however in this case we do need to eigenvectors to be normalized to simplify the application of the reduction technique described in sec .
[ sec : fastvarelim ] .
this requires finding the corresponding left - eigenvectors .
when finding these , we have a complication given by the fact that there is one eigenvalue which is @xmath89-fold degenerate , and the condition will not necessarily hold in general .
however , if we construct a matrix @xmath321 in such a way that its @xmath18-th column corresponds to @xmath322 , we can choose @xmath198 to be the @xmath19-th row of the inverse matrix @xmath323 , if such an inverse exists .
we find that indeed this is the case , and the left - eigenvectors in the partially reduced model are then given by @xmath324 we note that @xmath325 using this property , we can verify that the orthogonality condition is satisfied : @xmath326
in this appendix we derive the two - dimensional reduced mesoscopic model , by projecting the dynamics of the @xmath327-dimensional partially reduced mesoscopic model onto the slow degrees of freedom of its corresponding deterministic limit . using the projector defined in eq .
( [ eqn : projector ] ) on @xmath328 gives the slow variables defined by eqs .
( [ eqn : slow_z1 ] ) and ( [ eqn : slow_z2 ] ) . later ,
the inverse transformation , back to the more familiar variables @xmath129 and @xmath329 , will be required .
it is @xmath330},\label{eqn : theta_slow}\\ \lambda(\bm{z } ) & = \frac{\phi(\theta^*)}{k d_k{\theta^*}^{k}}\,z_2 , \label{eqn : lambda_slow}\end{aligned}\ ] ] however for the moment we will use the @xmath239 and @xmath240 variables , and will denote the right - hand side of eq .
( [ eqn : theta_slow ] ) as @xmath331 .
the deterministic slow dynamics , obtained by applying the projector to @xmath332 , will be given by @xmath333,\label{eqn : slow_dz1}\\ \dot{z}_2 & = z_2\left(\beta \phi ( \chi(\bm{z}))-\gamma \right),\label{eqn : slow_dz2}\end{aligned}\ ] ] where the dot represents the time derivative .
applying now the projector to the noise terms in the finite-@xmath2 case we find , for the component along @xmath244 , @xmath334\label{rhs_z2dot } \\ & = \frac{1}{\sqrt{n}}\left[\left(1-\frac{\phi(\chi(\bm z))}{\chi(\bm z)}\frac{\beta \theta^*}{-\gamma + \beta \phi(\theta^*)}\right)\bar{\sigma}_1(\bm z)\zeta_1(\tau ) + \frac{\beta \theta^*}{-\gamma + \beta \phi(\theta^*)}\left(\bar{\sigma}_2(\bm z)\zeta_2(\tau)+\bar{\sigma}_3(\bm z)\zeta_3(\tau)\right)\right ] , \label{eqn : eta_z1}\end{aligned}\ ] ] where we obtain , from eqs . , and , @xmath335^{1/2},\\ \bar{\sigma}_2(\bm{z})&= \left[\frac{\beta \phi(\theta^*)z_2}{kd_k{\theta^*}^k}\left(\phi(\chi(\bm z))+\psi(\chi(\bm z))-\frac{\phi(\chi(\bm z))^2}{\chi(\bm z)g^\prime ( \chi(\bm z))}\right)\right]^{1/2},\\ \bar{\sigma}_3(\bm z ) & = \left[\frac{z_2}{k d_{k}{\theta^*}^{k}}\left(\gamma \left[\phi(\theta^*)+\psi(\theta^*)\right]+\beta \left[\phi(\theta^*)\psi(\chi(\bm{z}))-\phi(\chi(\bm{z}))\psi(\theta^*)\right]\right)\right]^{1/2}. \label{eqn : sigma_3_bar_determined}\end{aligned}\ ] ] similarly , for the component of the fluctuations along @xmath336 : @xmath337 we now form @xmath338 and @xmath339 as linear combinations of @xmath340 and @xmath341 , using eqs .
( [ eqn : theta_slow ] ) and ( [ eqn : lambda_slow ] )
. these are equal to the appropriate linear combinations of eqs .
( [ eqn : slow_dz1 ] ) , ( [ eqn : slow_dz2 ] ) , ( [ eqn : eta_z1 ] ) and ( [ eqn : eta_z2 ] ) .
one finds that eqs .
( [ eqn : theta_stoch ] ) and ( [ eqn : lambda_stoch ] ) are exactly recovered , except that now @xmath177 is determined , given by eq .
( [ eqn : sigma_3_bar_determined ] ) . 47ifxundefined [ 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 * * , ( ) @noop * * , ( ) @noop * * ( ) , @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , , ) @noop _ _ ( , , ) @noop * * , ( ) @noop * * , ( ) , @noop _ _ ( , , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) in @noop _ _ ( ) pp . @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ , ed .
( , ) @noop * * , ( ) @noop _ _ ( , ) @noop _ _ , ed .
( , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ , vol .
( , ) `` , '' in @noop _ _ ( , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) , | networks of contacts capable of spreading infectious diseases are often observed to be highly heterogeneous , with the majority of individuals having fewer contacts than the mean , and a significant minority having relatively very many contacts .
we derive a two - dimensional diffusion model for the full temporal behavior of the stochastic susceptible - infectious - recovered ( sir ) model on such a network , by making use of a time - scale separation in the deterministic limit of the dynamics .
this low - dimensional process is an accurate approximation to the full model in the limit of large populations , even for cases when the time - scale separation is not too pronounced , provided the maximum degree is not of the order of the population size . |
within the standard model framework , the strong interaction is described by quantum chromodynamics ( qcd ) , which suggests the existence of the unconventional hadrons , such as glueballs , hybrid states and multiquark states .
the establishment of such states remains one of the main interests in experimental particle physics .
decays of the @xmath4 particle are ideal for the study of the hadron spectroscopy and the searching for the unconventional hadrons . in the decays of the @xmath4 particle ,
several observations in the mass region 1.8 gev / c@xmath7 - 1.9 gev / c@xmath7 have been presented in different experiments@xcite@xcite , such as the @xmath8@xcite@xcite , @xmath9@xcite@xmath10@xcite , @xmath11@xcite@xmath10@xcite and @xmath12@xcite .
recently , using a sample of @xmath13 @xmath4 events@xcite collected with besiii detector@xcite at bepcii@xcite , the decay of @xmath2 was analyzed@xcite , and the @xmath0 was observed in the @xmath1 mass spectrum with a statistical significance of @xmath5 . .
the dots with error bars are data ; the histogram is phase space events with an arbitrary normalization .
[ m6pi],scaledwidth=60.0% ] the @xmath1 invariant mass spectrum is shown in fig .
[ m6pi ] , where the @xmath0 can be clearly seen .
the parameters of the @xmath0 are extracted by an unbinned maximum likelihood fit . in the fit ,
the background is described by two contributions : the contribution from @xmath14 and the contribution from other sources .
the contribution from @xmath14 is determined from mc simulation and fixed in the fit ( shown by the dash - dotted line in fig .
[ m6pi_fit ] ) .
the other contribution is described by a third - order polynomial .
the signal is described by a breit - wigner function modified with the effects of the detection efficiency , the detector resolution , and the phase space factor .
the fit result is shown in fig .
[ m6pi_fit ] .
the mass and width of the @xmath0 are @xmath15 mev / c@xmath7 and @xmath16 mev , respectively ; the product branching fraction of the @xmath0 is @xmath17 . in these results ,
the first errors are statistical and the second errors are systematic . .
the dots with error bars are data ; the solid line is the fit result .
the dashed line represents all the backgrounds , including the background events from @xmath18 ( dash - dotted line , fixed in the fit ) and a third - order polynomial representing other backgrounds .
[ m6pi_fit],scaledwidth=60.0% ] figure [ comp_mw ] shows the comparisons of the @xmath0 with other observations at besiii@xcite .
the comparisons indicate that at present one can not distinguish whether the @xmath0 is a new state or the signal of a @xmath1 decay mode of an existing state . ,
title="fig:",scaledwidth=67.0% ] ( -137 , 138 ) ( -137 , 123.5 ) ( -137 , 109 ) ( -137 , 94.5 ) ( -137 , 80 )
with the same data sample , the decay of @xmath6 was searched for@xcite .
the mass spectrum of the @xmath1 is shown in fig .
[ m6pi ] , where no events are observed in the @xmath19 mass region . with the feldman - cousins frequentist approach@xcite , the upper limit of the branching fraction is set to be @xmath20 at the 90% confidence level , where the systematic uncertainty is taken into account .
with a sample of @xmath13 @xmath4 events collected at besiii , the decay of @xmath21 was analyzed@xcite .
the @xmath0 was observed in the @xmath1 invariant mass spectrum .
the mass , width and product branching fraction of the @xmath0 are @xmath15 mev / c@xmath7 , @xmath16 mev and @xmath17 , respectively .
the decay @xmath22 was searched for .
no events were observed in the @xmath19 mass region and the upper limit of the branching fraction was set to be @xmath20 at the 90% confidence level .
00 j. z. bai _ et al . _
[ bes collaboration ] , phys .
lett . * 91 * , 022001 ( 2003 ) .
j. p. alexander _ et al . _
[ cleo collaboration ] , phys .
d * 82 * , 092002 ( 2010 ) .
m. ablikim _ et al . _
[ besiii collaboration ] , phys .
lett . * 108 * , 112003 ( 2012 ) .
m. ablikim _ et al . _
[ bes collaboration ] , phys .
lett . * 95 * , 262001 ( 2005 ) .
m. ablikim _ et al . _
[ besiii collaboration ] , phys .
lett . * 106 * , 072002 ( 2011 ) .
m. ablikim _ et al . _
[ bes collaboration ] , phys .
lett . * 96 * , 162002 ( 2006 ) .
m. ablikim _ et al . _
[ besiii collaboration ] , phys .
d * 87 * , 032008 ( 2013 ) .
m. ablikim _ et al . _
[ besiii collaboration ] , phys .
lett . *
107 * , 182001 ( 2011 ) .
m. ablikim _ et al . _
[ besiii collaboration ] , chin .
c * 36 * , 915 ( 2012 ) .
m. ablikim _ et al . _
[ besiii collaboration ] , nucl .
instrum .
a * 614 * , 345 ( 2010 ) .
j. z. bai _ et al .
_ [ bes collaboration ] , nucl .
instrum .
a * 458 * , 627 ( 2001 ) .
m. ablikim _ et al . _
[ besiii collaboration ] , arxiv:1305.5333 [ hep - ex ] .
g. j. feldman and r. d. cousins , phys .
d * 57 * , 3873 ( 1998 ) . | observation of the @xmath0 in the @xmath1 invariant mass in @xmath2 at besiii is reviewed . with a sample of @xmath3 @xmath4 events collected with the besiii detector at bepcii , the @xmath0
is observed with a statistical significance of @xmath5 .
the mass , width and product branching fraction of the @xmath0 are determined .
the decay @xmath6 is searched for , and the upper limit of the branching fraction is set at the 90% confidence level .
observation of the @xmath0 at besiii |
right after the first detection of optical afterglows of gamma - ray bursts ( grbs ) and the diagnosis of grb afterglow radiation as synchrotron emission , predictions have been made for the linear and circular polarisation of grbs and their afterglows ( see for a review lazzati 2006 and references therein ) . while time resolved polarimetry of sources as faint and transient as grb afterglows is technically complicated and requires specialised instrumentation on large telescopes ,
the rewards are high : from time resolved polarimetric light curves we can determine grb parameters ( e.g. the jet structure , magnetic field configuration , viewing angle , etc . ) that can not easily be measured from light curves alone .
the first detections of polarisation of afterglows in the pre - swift era demonstrated technical feasibility , and shown that afterglows generally have low levels of polarisation ( @xmath0 ) that vary as a function of time ( see lazzati 2006 for an overview of pre - swift measurements ) .
the swift era has provided further incentive to perform detailed polarimetry : the observed richness in afterglow light curve morphology ( x - ray flares , plateaux , steep decays etc .
, see evans 2009 for statistics ) , has resulted in new models with various additional components to the standard fireball model , including for example the effects of high latitude emission , variable microphysics , energy injection mechanisms , etc .
many of these new model ingredients can be explored via the large sample of well sampled swift grb afterglow light curves and spectral energy distributions ( seds ) , but the large number of parameters and relatively low sensitivity of optical and x - ray light curves to some parameters ( e.g. @xmath1 ) make the addition of new independent constraints on the models , such as the linear or circular polarisation as a function of time , particularly useful .
one of the primary focus points of polarimetry models of grb afterglows has been the jet collimation and our viewing angle into the jet ( the angle between our sightline and the jet axis ) : simple afterglow models show that small differences in viewing angles and internal jet structure lead to strong and in principle easily identifiable differences in the behaviour of the polarisation as a function of time , in contrast with the optical and x - ray light curves , in which the differences are small and difficult to detect ( rossi 2004 and references therein ) . in the case of uniform , top - hat , jets with a unordered magnetic field , a key prediction is the existence of two bumps in the polarisation light curve , with a 90 degree change in polarisation position angle around the time of the jet break .
confirmation of the existence of such a change in position angle would give a new , light curve independent way of estimating jet opening angles , internal jet structure and viewing angle , for assumed magnetic field configurations .
however , as indicated by lazzati ( 2003 ) , the presence of polarisation caused by scattering by dust particles in the host galaxy alters both the linear polarisation and polarisation angle light curves .
this implies that to successfully use polarimetry as indicator of jet collimation , we require datasets that _ ( i ) _ span a wide time range , with data extending to far after the time of jet break ; _ ( ii ) _ measure polarisation as a function of wavelength ( e.g. through spectropolarimetry or multi - band imaging polarimetry ) , to separate the dust - induced polarisation from afterglow polarisation ; _ ( iii ) _ have well sampled multi wavelength light curves so that the presence of a light curve break can be established .
early attempts in the pre - swift era did not detect a 90 degree angle change in polarisation light curves ( see e.g. covino 2003 , greiner 2003 , rol 2003 , masetti 2003 , gorosabel 2004 ) .
the most important reason appears to be that in most cases the polarimetric light curves were too sparsely sampled ( only half a dozen sources have 3 or more data points ) , and most of these have rather uncertain jet break times .
the sources with best polarimetric coverage are 021004 and 030329 , both of these have highly irregular optical light curves , characterised by rebrightenings and bumps . the case of 030329
in particular shows some correlated behaviour between the light curve bumps and polarisation behaviour ( greiner 2003 ) , which makes interpretation in terms of simple polarimetry models difficult ( granot & knigl 2003 ) . data of grb020813 may also obey this correlation between light curve and polarisation variability : its polarisation curve is smooth ( fig 1 ; barth 2003 ; gorosabel 2004 ; lazzati 2004 ) just like the optical light curve ( laursen & stanek 2003 ) . using the very large telescope in chile , we embarked on a campaign to obtain well - sampled polarimetry light curves of swift bursts , selected solely by an initial on - board uvot identification of an afterglow , thus avoiding an observational bias towards sources that have a long - lasting shallow afterglow decay .
a first success of this campaign is the dataset presented in figure 1 ( for details see wiersema 2012 ) .
presented in this figure are the linear polarisation data points of the afterglow of grb091018 as obtained with the fors2 instrument ( in @xmath2 band , green symbols ) and a datapoint obtained with the isaac instrument ( in @xmath3 band , open square ) , gathered over 3 nights after the burst .
the optical and x - ray light curves of this afterglow show a break , with no change in the x - ray to optical spectral energy distribution , i.e. the break is achromatic .
we interpret this break as a jet break : the horizontal axis of figure 1 shows time since burst normalised by the jet break time .
immediately apparent is that data at @xmath4 have a constant polarisation angle , data after that have a higher but variable angle .
a weighted average angle of 6 degrees is found in the first interval , this is drawn as a dotted line in figure 1 .
the dotted line at @xmath5 is drawn at 96 degrees , and shows that the data is consistent with a 90 degree change in polarisation angle occurring slightly after @xmath6 .
the uniform top hat jet model with random field predicts that two bumps should be visible in the polarisation curve , and each bump has a constant polarisation angle .
the data at @xmath4 is perfectly consistent with this prediction , if the viewing angle is slightly off - axis ( @xmath7 ) .
the later data is not consistent with a simple broad bump with constant angle . highlighted in an inset in figure 1 is the behaviour of the polarisation angle around @xmath8 .
the angle shows a rapid sweep of the source through the stokes plane : angle and polarisation can jointly be explained if in addition to the expected smooth bump from the simple models there is a slowly variable , low polarisation component present with an angle nearly 90 degrees offset from the expected bump ( 96 degrees ) .
the addition of these two components can largely reproduce the observed behaviour ( wiersema in prep . ) .
we therefore consider this case the first with a polarisation - based jet break identification .
there are further features of interest in the 091018 data .
first of all , we acquired not only linear polarimetry , but also circular polarimetry in @xmath2 , again using vlt fors2 , in between the first and second datapoint in figure 1 .
these show a non - detection of circular polarisation , with a limit of @xmath9 ( wiersema 2012 ) .
there are no signs of reverse shock contribution to the afterglow of this burst , so we consider this a tight limit on the forward shock circular polarisation , and therefore on the presence of weak but ordered magnetic fields in the blast wave .
a second point of interest is the very low polarisation at early times .
this , together with the angle behaviour described above , makes structured jet models very unlikely , but also sets a strong lower limit on the size and number of any coherent patches of emission on the blast wave ( gruzinov & waxman 1999 ) .
finally , if the fast variability behaviour at @xmath10 is a common one , i.e. if it is something that may be seen in all afterglows rather than caused by something which is specific to this burst only , we may need a much larger emphasis on late - time polarimetry , in the sense that a 90 degree angle shift from early data can only be measured using several data points together .
i would like to note that data taken after the conference of another grb seem to imply that this behaviour is indeed a common one , though analysis is ongoing ( wiersema in prep . ) . on figure 1
i also plot the pre - swift grbs which have 3 or more data points and an estimate of the jet break time from light curves ( taken from zeh 2006 ) , excluding grb030329 for reasons stated above .
polarisation angles are shifted so that their early time values fall on the grb091018 value , so that it is easier to see angle changes .
this plot demonstrates that if all bursts behave like grb091018 , there are not sufficient data points beyond @xmath10 to diagnose a 90 degree angle change .
one exception is grb020813 , which has some data in this interval , but may have fallen victim to the same rapid variability behaviour as seen in grb091018 .
is the same as that of 091018 ( angle @xmath11 ) , indicated by the horizontal dashed line for @xmath12 .
the horizontal dashed line at @xmath13 is drawn at an angle @xmath14 .
the times of ( candidate ) jet breaks of the pre - swift bursts are as found in zeh ( 2006 ) . ]
as mentioned in the previous section , scattering of afterglow photons on dust particles in the host galaxy results in wavelength dependent linear polarisation . in sight lines in our own galaxy , the wavelength dependence is often describe by the empirical serkowski curve ( serkowski 1975 ) , the black dashed curve in figure 1 .
the induced polarisation peaks with polarisation value @xmath15 at wavelength @xmath16 .
if we assume this curve , or a similar parametrisation , to also be valid for extragalactic sight lines , we can express the expected polarisation in different photometric bands ( e.g. in @xmath2 and @xmath17 ) in terms of dust parameter @xmath18 ( klose 2004 ) .
as can also be seen in figure 1 , the ratio of the detected polarisation in @xmath2 and @xmath17 does not show evidence for significant dust induced polarisation .
this is also true for the other , pre - swift , cases where wavelength dependent polarimetry exists ( e.g. barth 2003 ) , reflecting the low amount of dust seen in these sightlines , a low degree of dust grain alignment , or dust grain size distributions different from galactic environments .
further study of dust induced polarisation in afterglows would be very useful , in particular because the high quality spectra and seds that can be obtained for these afterglows can be combined with polarimetry to better understand dust processing in grb environments .
figure 2 shows that even at @xmath19 the @xmath17 band is red ward of the peak of the serkowski curve , whereas the r band is blue ward for all but the very lowest redshift grbs . to exploit this fact , we are performing a small survey of afterglows using the liris instrument on the 4.2 m william herschel telescope , in imaging polarimetry mode , which has had some success already ( wiersema 2012b ) .
( red , the redshift of grb091018 is 0.97 ) and @xmath20 .
the curve is characterised by a typical wavelength @xmath16 at which maximum polarisation @xmath15 is present .
, width=264 ]
in recent years , new , dedicated , instrumentation has succeeded in robustly measuring the polarisation of gamma - rays of the grb prompt emission ( yonetoku 2011 ) . in all cases
high values of polarisation were found , in contrast with the low values found in the late time forward shocks shown in figure 1 .
the use of polarimetry instruments on robotic telescopes allows investigation of the transition of prompt to afterglow emission , and is able to probe the reverse shock ( or its absence ) , and therefore investigate the magnetisation of the grb ejecta .
the case of grb 090102 in particular showed a high polarisation likely associated with reverse shock ( steele 2009 ) .
early circular polarimetry of grb afterglows can probe the ordered field component in reverse shock emission , and in some cases even fairly late observations may be sufficient for a detection ( wiersema in prep . ) . resolving the decay of the reverse shock and rise of the forward
shock will require the ability to acquire polarimetry at short timescales ( exposure times ) , but short exposure polarimetry is also of some interest at late times : if the fast variability seen in grb 091018 after the jet break is commonplace , we need short exposures to resolve its variability timescale . secondly , the model where a large number of small patches of coherent magnetic field contribute to the received emission ( gruzinov & waxman 1999 ) can be tested through short time scale variability tests .
we use liris at the wht for this : the instrument utilises a double - wollaston , and therefore records @xmath21 simultaneously in each exposure .
we typically use 30 second exposure sets ( 3 sub exposures of 10 seconds ) , to get good sky subtraction in the @xmath3 band .
an analysis of field stars in a grb field , shown in figure 3 , shows we can do linear polarimetry with polarisation errors of 1% in 30 second exposures for @xmath22 ( vega magnitudes ; wiersema in prep . )
. band with liris on the wht .
measurements of stokes @xmath21 are done on 30 second integrations .
the values of the 4 stars are vertically displaced for clarity .
the afterglow is too faint for meaningful polarimetry on these short time scales .
, width=264 ]
it is clear from the above that polarimetry of grb afterglows is a important pursuit .
the recent measurements of @xmath23-ray polarisation of grb prompt emission , the advent of polarimeters on robotic telescopes capable of probing the very early afterglow , and the increasing capabilities for polarimetry at longer wavelengths ( e.g. alma , jvla ) highlight the importance of late - time , deep , and densely sampled polarisation curves .
the recent results on grb091018 give some long sought - after confirmation of basic predictions of blast wave models , in particular a 90 degree change in polarisation angle after the jet break .
similar campaigns on other bursts are required to probe the relation of polarisation behaviour with other burst parameters , e.g. the bulk lorentz factor , burst energetics , reverse shock properties and the viewing angle into the jet .
besides giving some support to jet break models , the dataset of grb091018 appears to show a new kind of unpredicted fast variability around or just after the jet break , illustrating that there is still plenty of discovery space left in afterglow polarisation studies .
barth , a. j. , sari , r. , cohen , m. h. 2003 , apj , 584 , 57 covino , s. , malesani , d. , ghisellini , g. 2003 , a&a , 400 , l9 evans , p. a. , beardmore , a. p. , page , k. l. 2009 , mnras , 397 , 1177 gorosabel , j. , rol , e. , covino , s. 2004 , a&a , 422 , 113 granot , j. , knigl a. , 2003 , apj , 594 , l83 granot j. , taylor g. b. , 2005 , apj , 625 , 263 greiner , j. , klose , s. , reinsch , k. 2003 , nature , 426 , 157 gruzinov , a. , & waxman , e. 1999 , apj , 511 , 852 klose s. , palazzi , e. , masetti , n. 2004 , a&a , 420 , 89 laursen , l. t. , stanek , k. z. 2003 , apj , 597 , 107 lazzati , d. , covino , s. , di serego alighieri , s. 2003 , a&a , 410 , 823 lazzati , d. , 2006 , njph , 8 , 131 masetti , n. , palazzi , e. , pian , e. 2003 , a&a , 404 , 465 rol , e. , wijers , r. a. m. j. , fynbo , j. p. u. 2003 , a&a , 405 , 23 rossi , e. m. , lazzati , d. , salmonson , j. d. 2004 , mnras , 354 , 86 serkowski k. , mathewson d.l .
, ford v.l . , 1975 ,
apj , 196 , 261 steele , i. a. , mundell , c. g. , smith , r. j. 2009 , nature , 462 , 767 wiersema , k. , curran , p. a. , krhler , t. 2012 , mnras , 426 , 2 wiersema , k. , van der horst , a. j. , levan , a. j. 2012b , mnras , 421 , 1942 yonetoku , d. , murakami , t. , gunji , s. 2011 , apj , 743 , 30 zeh , a. , klose , s. , kann , d. a. 2006 , apj , 637 , 889 | follow - up observations of large numbers of gamma - ray burst ( grb ) afterglows , facilitated by the swift satellite , have produced a large sample of spectral energy distributions and light curves , from which the basic micro- and macrophysical parameters of afterglows may be derived .
however , a number of phenomena have been observed that defy explanation by simple versions of the standard fireball model , leading to a variety of new models .
polarimetry has shown great promise as a diagnosis of afterglow physics , probing the magnetic field properties of the afterglow and geometrical effects ( e.g. jet breaks ) . unfortunately , high quality polarimetry of a significant sample of afterglows is difficult to acquire , requiring specialised instrumentation and observing modes . in this talk
i will review the recent successes in afterglow polarimetry , also showing first results of new instruments and observing campaigns .
i will particularly focus on jet breaks . |
during diffusive shock acceleration ( dsa ) cosmic rays ( cr ) gain energy by repeatedly passing back and forth between the upstream and downstream plasmas ( krymskii 1977 , axford et al 1977 , bell 1978 , blandford & ostriker 1978 ) .
cr diffuse ahead of the shock to form a precursor with an exponential scaleheight @xmath0 where @xmath1 is the shock velocity and @xmath2 is the cr diffusion coefficient upstream of the shock .
the average dwell - time spent upstream of the shock between successive shock crossings is @xmath3 for relativistic particles ( bell 2012 ) .
this , along with the corresponding downstream dwell - time , determines the rate at which cr are accelerated .
lagage & cesarsky ( 1983a , b ) showed that the time taken for cr acceleration is @xmath4 where @xmath5 is the downstream diffusion coefficient and @xmath6 is the downstream fluid velocity in the shock rest frame . since @xmath7 for a strong shock it might appear that the downstream dwell - time determines the acceleration rate , but we can expect @xmath8 partly because the magnetic field is increased by compression by the shock , partly because the compressed downstream magnetic field is more closely perpendicular to the shock normal , and partly because the downstream field is disturbed and more irregular after passing through the shock .
if the downstream and upstream dwell - times are the same @xmath9 .
a further common assumption is that bohm diffusion applies : @xmath10 or @xmath11 where @xmath12 is the cr larmor radius .
this assumes a diffusion model in which cr are scattered by irregularities in the magnetic field such that the scattering mean free path is of the order of the cr larmor radius .
there is some observational evidence for bohm diffusion ( stage et al 2006 , uchiyama et al 2007 ) .
furthermore if the mean free path were much larger than a larmor radius acceleration by snr to the knee in the galactic cr spectrum would be very difficult .
the maximum cr energy is determined by the condition that @xmath13 can not exceed the age @xmath14 of an snr ( lagage & cesarsky 1983a , b ) . assuming @xmath10 , the maximum cr energy @xmath15 in ev
is given by @xmath16pev where @xmath17 is the upstream magnetic field in microgauss , @xmath18 is the shock velocity in units of 10,000 km s@xmath19 and @xmath20 is the snr age in 1000 s of years . for a typical young snr expanding into the interstellar medium without magnetic field amplification @xmath21 , @xmath22 and @xmath23 , giving @xmath24pev which falls a factor of @xmath25 short of that required to explain the galactic cr spectrum .
snr in the sedov phase do not fare better .
their shock velocities decrease in proportion to @xmath26 and @xmath27 , so little benefit accrues from their larger age and radius .
this posed a serious problem for diffusive shock acceleration as an explanation of the galactic spectrum until it was shown that a plasma instability driven by streaming cr in the upstream precursor could amplify the magnetic field ahead of the shock and facilitate rapid acceleration to higher energies ( lucek & bell 2000 , bell 2004 , 2005 ) .
the phenomenon of magnetic field amplification provides a mechanism by which the cr energy can be raised significantly beyond 0.01pev , but there remains the question of why the fields are amplified to the observed magnitude , up to 100 s @xmath28 in the historical snr ( vink & laming 2003 , berezhko et al 2003 , vlk et al 2005 ) , and why cr are accelerated to a few pev rather than 0.1 or 10 pev .
we try to answer these questions by examining the self - consistent interaction between streaming cr , behaving kinetically , and the upstream plasma behaving magnetohydrodynamically .
it has been known for many years that the escape of cr upstream of the shock is an important part of the overall acceleration process as discussed below in the final paragraphs of section 3 .
we find that the combined cr - mhd system organises itself to allow a suitable number of cr to escape upstream . the cr drive magnetic field amplification which in turn regulates the number of escaping cr .
if a smaller number of cr escaped , the magnetic field would be insufficiently amplified to confine and accelerate the cr .
if a larger number of cr escaped , the magnetic field would grow too rapidly to allow their escape .
hence a self - regulating system is set up that determines the number and maximum energy of escaping cr .
the paper is organised as follows .
sections 2 and 3 present approximate calculations showing how a limit on the cr energy is placed by the need for cr to drive magnetic field amplification by escaping upstream .
sections 4 to 7 describe vlasov - fokker - planck ( vfp ) simulations that support the arguments of sections 2 and 3 .
sections 8 to 11 apply the results to supernova remnants and the galactic cosmic ray spectrum . readers unfamiliar with vfp simulations may wish to read sections 1 to 3 and 8 to 11 before returning to the computational validation and illustration in sections 4 to 7 .
we assume that magnetic field is generated by the non - resonant hybrid ( nrh ) instability described by bell ( 2004 ) .
this is one of a class of plasma instabilities driven by cr streaming . in its simplest form , cr have a larmor radius much greater than the wavelength of spiral perturbations in a zeroth order uniform magnetic field .
because of their large larmor radius the streaming cr , carrying an electric current density @xmath29 , are essentially undeflected by the perturbed field but the @xmath30 force acts towards the centre of the spiral . a corresponding reactive force acts on the background plasma to expand the spiral .
this stretches and increases the magnitude of the perturbed magnetic field , thereby increasing the @xmath30 force in a positive feedback loop that drives the instability .
nrh appears to be the most rapidly growing instability driven by cr streaming on a relevant scalelength .
other instabilities that have attracted significant attention are the resonant alfven instability ( kulsrud & pearce 1969 , wentzel 1974 ) that grows with spatial wavelengths spatially resonant with the cr larmor radius and the weibel instability ( weibel 1959 ) that grows quickly on the spatial scale of an electron or proton collisionless skin depth , @xmath31 km , @xmath32 .
the weibel instability is important for interactions engaging thermal or mildly suprathermal electrons and ions , and may be effective for cr acceleration to low energies , but it grows on a scale too small to scatter pev ions which have a larmor radius of @xmath33 .
the alfven instability grows on the desired spatial scale but grows less quickly than the nrh instability in the snr shock environment .
instabilities that grow on scales larger than the cr larmor radius ( bykov et al 2011 , drury & falle 1986 , drury & downes 2012 , malkov & diamond 2009 , rogachevskii et al 2012 , schure & bell 2011 ) tend to have longer growth times but turbulent amplification may increase the growth rate ( bykov et al 2011 ) . an extended discussion of instabilities driven by cosmic ray streaming can be found in schure et al ( 2012 ) . from the lagage & cesarsky ( 1983a , b ) limit
as described in section 1 it is clear that proton acceleration to a few pev is only possible if the magnetic field is strongly amplified above its characteristic interstellar value of a few @xmath34 .
the following argument places an upper limit on the maximum energy of a cr that can be strongly scattered by a magnetic field growing on the scale of a cr larmor radius .
the background thermal plasma is highly magnetised on the scale of a pev cr larmor radius .
consequently the magnetic field is ` frozen in ' to the thermal plasma .
magnetic field amplification occurs as the plasma moves and stretches field lines . assuming the perturbed magnetic field does not far exceed the initial field
, the @xmath30 force displaces a plasma element a maximum distance @xmath35 in time @xmath36 where @xmath37 is the initial seed field . for the stretched magnetic field to strongly scatter cr as required for bohm diffusion the displacement must grow to the order of a cr larmor radius , that is @xmath38 where @xmath39 is the cr energy in ev .
the cr current density @xmath29 carries an energy flux @xmath40 in the upstream plasma rest frame which can not far exceed the energy flux @xmath41 carried by the upstream plasma into the shock , where @xmath42 is the upstream mass density .
we define a cr acceleration efficiency @xmath43 such that @xmath44 .
we then have two equations : @xmath45 and @xmath46 .
when combined , they yield a cr energy @xmath47 .
this expression for @xmath39 is equivalent to @xmath48pev where @xmath18 is the shock velocity in units of @xmath49 , @xmath50 is the seed magnetic field in @xmath51 from which amplification begins , and @xmath20 is the time in 1000 s of years . characteristically for young snr , @xmath22 , @xmath23 , @xmath52 ( see appendix a ) and @xmath53 where the magnetic field of a few @xmath51 represents the seed field from which the instability grows . with these estimates , @xmath54pev , which is an order of magnitude less than the energy of the knee .
this estimate is independent of any detailed instability theory except for the assumption that magnetic field amplification takes place through plasma motions generated by the @xmath30 force acting on the scale of a cr larmor radius .
it highlights the difficulty of amplifying magnetic field by orders of magnitude on the scale of the larmor radius of a pev proton and the need for an instability that can provide non - linear growth of the magnetic field .
the nrh instability offers a way round this difficulty by initially growing very rapidly on a scale much less than the cr larmor radius .
since the small - scale magnetic field grows unstably the @xmath30 force grows exponentially in time .
the scale - size increases to the cr larmor radius during non - linear growth .
the nrh growth rate is @xmath55 where @xmath56 is the zeroth order magnetic field and @xmath57 is the wavenumber on which the instability grows .
the maximum nrh growth rate is @xmath58 which occurs at a wavenumber @xmath59 .
@xmath60 is @xmath61 times larger than the nrh growth rate for @xmath62 , thus allowing the magnetic field and the @xmath30 force to increase rapidly .
the nrh and alfven growth rates are similar at @xmath62 ( see appendix e and bell 2004 ) . using the above nomenclature , @xmath63 where @xmath64 is the alfven speed , @xmath65 and @xmath66 is the electron density in @xmath67 . for a discussion of the value of @xmath43 see appendix a and bell ( 2004 ) .
the growth time of the fastest growing mode is then @xmath68 years where @xmath69 is the energy in pev of the cr driving the instability .
years for @xmath71 , @xmath72 , @xmath22 which implies that even the nrh instability struggles to amplify the magnetic field sufficiently to accelerate cr to pev energies in the historical snr . for growth by many e - folding the nrh instability must be driven by cr with energies less than 1pev . as remarked above
, the fastest growing mode grows on a scale @xmath73 whereas efficient cr scattering requires fluctuations in the magnetic field on a scale @xmath12 .
the above analysis showed that @xmath74 .
the instability initially grows on a scale too small to effectively scatter pev cr .
two factors save the situation .
firstly , as the magnetic field grows from a few @xmath75 g to a few hundred @xmath75 g the cr larmor radius decreases by the corresponding factor of @xmath25 . secondly , the characteristic scalelength of the structured magnetic field increases during non - linear growth of the instability .
figure 1 is drawn from figures 3 and 4 of bell ( 2004 ) .
the graph of energy densities shows how the magnitude of the magnetic field grows to many times its seed value in a time @xmath76 .
the characteristic scalelength grows during this time as shown by the grey - scale images . by the time @xmath77 the scalelength has grown to the size of the computational box from its initially much smaller scale .
both @xmath57 and @xmath12 decrease by a large factor by the time @xmath78 .
the simulation reproduced in figure 1 had a fixed cr current so it was impossible to demonstrate strong cr scattering by the magnetic field after @xmath79 , but the figure provides strong evidence for rapid initial growth of the fastest growing modes into large scale magnetic structures able to scatter cr on the scale of a larmor radius .
the previous section does not provide an estimate of the magnitude of the amplified magnetic field , but it does define conditions under which strong magnetic field amplification occurs .
cr acceleration to pev energies requires strong magnetic field amplification which fixes the number of instability e - foldings to the range 5 - 10 . this in turn fixes the cr current .
from figure 1 we take the condition for strong magnetic field amplification in a particular volume of upstream plasma to be that @xmath80 in the time before the shock overtakes it .
since @xmath81 , the condition for field amplification is @xmath82 where @xmath83 is the total electric charge of cr passing through unit surface area upstream of the shock before the shock arrives .
( it makes no difference whether the cr passing through the plasma have high or low energies .
it is only the number of cr escaping upstream that counts . ) since only the highest energy cr escape , the areal charge @xmath83 is carried by the highest energy cr being accelerated .
the larmor radius of a pev proton in the interstellar medium is comparable with the radius of the historical snr and its scattering mean free path is very much greater than the snr radius .
these escaping cr pass into the interstellar medium with low probability of further encounter with the snr shock .
strongly scattered cr at lower energies are confined by the shock and are advected away downstream after acceleration .
suppose that the cr distribution follows a @xmath84 power law up to a momentum @xmath85 , then in steady state the electrical current of cr escaping in a small band of energies above the energy @xmath86 is @xmath87 where @xmath88 is the isotropic part of the cr distribution in momentum space at the shock .
equation ( 2 ) is derived from equation ( 11a ) below by integrating in space across the shock and deriving the rate at which cr reach the momentum @xmath85 .
the derivation can be found in appendix b. in comparison the cr pressure at the shock , where the cr distribution extends down to @xmath89 and @xmath90 is the proton mass is @xmath91 giving @xmath92 where we have assumed @xmath93 for a cr distribution extending from 1gev to 1pev . the condition for magnetic field amplification that @xmath94 can be rearranged to give a value for @xmath15 : @xmath95 which is equivalent to @xmath96 for characteristic values ( @xmath22 , @xmath72 , @xmath97 , @xmath98 ) @xmath99tev .
as expected from the above discussion , this falls short of the few pev required to explain galactic cr .
this will be discussed further in section 8 .
zirakashvili et al ( 2008a , b ) developed a related analytic model of the excitation of the nrh instability and cr confinement upstream of the shock .
they derived an estimate for the maximum cr energy with a similar dependency on @xmath100 in the numerator of their equation ( 19 ) but with an additional denominator that depends on the magnitude of the amplified magnetic field .
their analysis operates at a more detailed level than ours since they consider the growth in amplitude of the amplified magnetic field over a range of scales and small angle cr scattering by small scale field .
however , the dominant physics is similar in their analysis and ours , and their estimate of the cr energy ( equation ( 19 ) ) is similar to that in our equation ( 6 ) .
note that the magnetic field does not enter into the estimation of the maximum energy cr in equation ( 6 ) .
the magnetic field is assumed to grow to whatever magnitude and spatial scale required to confine cr at energies less than @xmath15 .
this assumption is justified on the basis that the magnitude of the field increases by a substantial numerical factor in time @xmath101 even in the non - linear regime and that this is accompanied by rapid growth in characteristic spatial scale .
for example , the magnetic field and its spatial scale are very much larger in figure 1 when @xmath102 than when @xmath103 even though the time is different by a factor of only @xmath104 . allowing the field to grow for just a little longer gives a much larger magnetic field and spatial scale .
consequently the primary parameter is not the magnitude of the magnetic field but the charge carried by escaping cr .
it is clear that magnetic field amplification is only possible if a population of high energy cr escape upstream of the shock . a magnetic field capable of stopping their escape is only produced after the areal charge @xmath83 has escaped .
thus cr escape upstream of the shock is an essential aspect of diffusive shock acceleration when magnetic field amplification is operational .
if a charge @xmath83 has not passed through a particular point upstream of the shock then cr can not be confined at that distance from the shock .
previous authors have also concluded that cr escape upstream is inevitable , but generally for different reasons .
one possibility is that cr escape through filamentary cavities in the magnetic field ( reville & bell 2012 , 2013 ) but usually it is assumed that cr escape at a free escape boundary in position or momentum .
free escape boundaries were introduced at an early stage in the development of shock acceleration theory ( ellison et al 1981 ) .
it was appreciated that steady state models were impossible without cr escape ( ellison & eichler , 1984 , berezhko & ellison 1999 , malkov et al 2000 ) .
the standard @xmath105 test particle energy spectrum at strong shocks diverges if integrated to infinite energy .
the divergence is even stronger if non - linear cr feedback onto the shock structure is included since the relativistic cr pressure increases the density jump at the shock resulting in a spectrum flatter than @xmath105 at high cr energy .
however , the conditions for a steady state solution do not present a compelling argument for a free escape boundary since time - dependent solutions offer an acceptable option . indeed the lagage & cesarsky limit on cr energy is based on the assumption that the maximum cr energy increases with time .
the case for cr escape was strengthened by the recognition that magnetic field amplification is essential since there will always be a distance upstream at which amplification is inoperative and cr must escape . similarly , there must always be an upper limit to the spatial scale to which field is amplified and a corresponding limit , through the larmor radius , on the momentum to which cr can be accelerated .
this led to the imposition of a free escape boundary either at an imposed distance upstream of the shock ( eg caprioli et al 2010b , ohira et al 2010 ) or at an imposed cr momentum ( eg ellison & bykov 2011 ) . whether the free escape boundary is imposed in momentum or configuration space , the momentum at which cr escape depends on the magnetic field .
some authors ( eg ptuskin et al 2010 ) assume that the field is amplified until the magnetic energy density reaches a given fraction of @xmath106 . the fractional magnetic energy density can be chosen to match observation ( ptuskin et al 2010 , vlk et al 2005 ) .
alternatively the field can be chosen to match the saturation value estimated by bell ( 2004 ) .
others ( caprioli et al 2010a , drury 2011 ) choose a mathematical form for the amplified magnetic field that allows for multiple possibilites .
ptsukin & zirakashvili ( 2003 ) and caprioli et al ( 2009a , b , 2010b ) include magnetic field generation due to the resonant instability as it develops ahead of the shock .
vladimirov et al ( 2006 ) include magnetic field generation in response to gradients in the cr pressure .
ellison et al ( 2012 ) choose an amplified magnetic field suitable for snr expansion into a circumstellar wind .
in contrast to the above , cr escape in our model is determined by the escaping cr electric charge rather than the magnetic field generated in the upstream plasma , although magnetic field amplification is implicitly required by our model .
reville et al ( 2009 ) also used the escaping flux to calculate growth rates in an effort to motivate a realistic free - escape boundary location in their steady - state non - linear model .
their discussion on self - regulation of cr precursors however did not extend to the maximum cr energy .
we now set out to test the above conclusions as far as we are able with a numerical model that includes the self - consistent interaction of cr modelled kinetically with a background plasma modelled magnetohydrodynamically .
standard mhd equations describe the background plasma except that a @xmath107 force is added to the momentum equation : @xmath108 as described in lucek & bell ( 2000 ) and bell ( 2004 ) .
the cr distribution function @xmath109 at position @xmath110 and momentum @xmath111 is defined in the local fluid rest frame and evolves according to the vlasov - fokker - planck ( vfp ) equation @xmath112 where @xmath113 is an optional collision term included to represent scattering by magnetic fluctuations on a small scale .
the electric field is zero in the local fluid rest frame .
quadratic terms in the local fluid velocity @xmath114 are neglected on the assumption that @xmath115 .
the cr current @xmath29 , required for insertion into the mhd momentum equation , is calculated by integration over @xmath116 in momentum space .
the magnetic field in the cr kinetic equation is taken from the mhd calculation . as in lucek & bell ( 2000 ) and bell ( 2004 )
three spatial dimensions are needed to represent the turbulence adequately . since our aim is to investigate the mutual interactions of magnetic field amplification , cr acceleration and cr escape upstream of the shock we need to model the complete system including the shock and the complete cr precursor . because we model the detailed interaction between the cr and the distorted magnetic field we need to resolve the cr larmor radius in configuration space and the rotation of cr trajectories in momentum space . as a consequence
the numerical model should be 6-dimensional in momentum - configuration space with spatial scales extending from the cr larmor radius of the lowest energy cr to the precursor scaleheight of the highest energy cr
. this would be impossible without extraordinary computational resources so our strategy is to design a computational model that includes all the important processes at a minimal level .
we retain the three dimensions in configuration space but limit the range of cr momentum to a factor of 10 so that we do not have to resolve the larmor radius of very low energy cr .
we choose a shock velocity @xmath117 to keep the ratio of the cr to the mhd timescale to a minimum while staying close to the range of conceivable snr expansion speeds .
our greatest approximations are made in the momentum space representation of the cr distribution function since this is the aspect of the calculation in which the number of dimensions can be reduced .
the vlasov - fokker - planck ( vfp ) equation ( equation 8) is important in the physics of laser - produced plasmas where it is solved in finite difference form to model electron transport .
the successful use of vfp simulation to model electron transport in laser - produced plasmas stretches back more than 30 years ( bell et al , 1981 ) so it is natural to apply the techniques to cr which obey the same equation .
the distribution function @xmath118 of charged particles ( cosmic rays or energetic electrons ) is usually represented in spherical co - ordinates ( @xmath119,@xmath120,@xmath121 ) in momentum space .
a common representation of the distribution function is as a sum of spherical harmonics : @xmath122 where @xmath123 is the coefficient for the ( @xmath124,@xmath90 ) spherical harmonic .
@xmath123 is function of time , position and magnitude of momentum @xmath119 .
the spherical harmonics decribe the angular structure on shells of constant magnitude of momentum .
reviews of the vfp technique , or papers containing a significant review element , are bell et al ( 2006 ) , tzoufras et al ( 2011 ) and thomas et al ( 2012 ) . for information on the application of vfp techniques
to cr acceleration the reader is referred to reville & bell ( 2013 ) which includes an appendix setting out the full vfp equations for a spherical harmonic expansion .
bell , schure & reville ( 2012 ) also apply vfp techniques to cr acceleration .
the use of an expansion in tensors ( used here ) as an alternative to spherical harmonics is discussed by schure & bell ( 2012 ) .
vfp simulation was used by bell et al ( 2011 ) for the calculation of cr acceleration by oblique shocks .
they found that an expansion to 15th harmonic could be needed for oblique shocks because of the abrupt change in magnetic field direction at the shock , but that only a few harmonics are needed for quasi - parallel shocks .
here we reduce the computational size of the problem by modelling parallel shocks in which the zeroth order magnetic field is parallel to the shock normal .
there are good reasons to suppose that the first few terms in the expansion capture the essential physics .
firstly , as shown by bell ( 2004 ) the nrh instability is driven by the cr current density @xmath29 .
higher order anisotropies do not directly contribute to the instability .
secondly , the cr precursor scaleheight is @xmath125 times the cr scattering mean free path in standard dsa theory .
dsa theory is based on the diffusive approximation in which only the first order anisotropy is needed and only the first two terms ( isotropy and drift anisotropy ) in the harmonic expansion are retained . in diffusion theory
the higher order anisotropies are damped by scattering .
hence it might be supposed that only the first two terms are needed and an adequate representation of the cr distribution function might be @xmath126 .
this ` @xmath127 ' expansion allows free cr propagation along magnetic field lines but restricts transport across the magnetic field because the direction of the anisotropic drift term is rotated by the field .
however , this expansion omits an essential feature of cr transport . it does not allow cr to gyrate as they travel along a magnetic field line .
the @xmath127 expansion allows cr to propagate along field lines and separately it allows cr to gyrate around field lines , but it does not allow cr to do both at the same time .
spiral trajectories require the inclusion of the off - diagonal components of the stress tensor @xmath128 . without the stress tensor ,
cr can not resonantly interact in space with magnetic perturbations on the scale of a larmor radius .
clearly this would rule out the alfven instability , and more surprising it also rules out the nrh instability .
the role of the stress tensor is discussed by schure et al ( 2011 ) in which the linear nrh dispersion relation is derived with the perturbed cr distribution expressed as a tensor expansion .
we proceed on the basis that the cr distribution function may be adequately represented by an isotropic part ( 0th order in @xmath129 ) plus a drift component ( 1st order in @xmath129 ) plus a term representing the stress tensor ( 2nd order in @xmath129 ) .
this second order expansion is more easily represented in the equivalent tensor notation instead of spherical harmonics : @xmath130 where the trace of @xmath131 is zero because it is already accounted for in the isotropic term @xmath88 . following johnston ( 1960 ) the reduced vlasov equation for the cr distribution function is then : @xmath132 @xmath133 @xmath134 @xmath135 where
quadratic order terms in the velocity @xmath114 have been neglected and we have omitted terms involving @xmath136 times a gradient of @xmath114 apart from the term in equation ( 11a ) for the evolution of @xmath88 .
this amounts to the neglect of second order fermi acceleration and acceleration resulting from shear motions in the background hydrodynamics . in our problem , these processes are small in comparison with acceleration at the shock .
we have chosen to terminate the set of equations at equation ( 11c ) by setting @xmath137 to zero . additionally , we soften the termination of the tensor expansion by replacing the magnetic rotation in equation ( 11c ) by a damping term with a damping rate equal to the magnetic gyration frequency .
the logic of this approximation is that random rotation of the stress tensor anisotropy leads to its damping .
a similar assumption underlies bohm diffusion which replaces rotation of @xmath138 by a damping rate , thereby terminating the tensor expansion at @xmath138 whereas we terminate it at @xmath131 .
our approximation makes intuitive sense , and we support it in appendices d and e by examining its effect on propagating modes and on the nrh instability .
the notation is simplified by introducing the vector @xmath139 related to shear and vorticity ( @xmath140 ) in cr motion to represent the off - diagonal components of the stress tensor .
the on - diagonal components of @xmath131 ( @xmath141 ) are approximately accounted for by multiplying the @xmath142 term in equation ( 11b ) by 9/5 to allow for the stress tensor contribution to compressional waves and to allow freely streaming cr to propagate at @xmath143 instead of @xmath144 , as shown in appendix d. the equations to be solved ( their derivation is given in appendix c ) are then @xmath145 @xmath146 @xmath147 equivalently , expressed in vector notation , @xmath148 @xmath149 @xmath150 where @xmath151 is the vector cr larmor frequency and we take @xmath152 .
the presence of the curls of @xmath153 and @xmath154 facilitates the propagation of transverse modes in @xmath153 and @xmath154 as needed for helical motion along magnetic field lines or cr propagation at an angle to the wavevector @xmath155 .
the termination of the harmonic expansion at the stress tensor makes the computation tractable with available resources .
appendices d and e show that the truncated expansion provides an adequate representation of the essential physics of cr propagation and cr - driven instability .
a full 3d simulation with realistic parameters is not possible because of the large ratio of the largest distances ( the cr precursor scaleheight and the cr free escape distance ) to the smallest distance ( the shortest wavelength on which the nrh instability grows ) .
the correspondingly large ratio of the largest timescale ( the snr expansion time ) to the shortest timescale ( the shortest nrh growth time ) similarly makes substantial demands on computer resources .
the computational constraints are discussed in appendix f. we artificially increase the magnetic field and stretch other parameters in a favourable direction , choosing : @xmath156 @xmath157 where @xmath158 is the energy at which cr are injected at the shock .
the zeroth order magnetic field is aligned along the shock normal .
the instability is seeded by imposing random fluctuations on the magnetic field with a mean magnitude of @xmath159 .
cr are injected at the shock into the lowest momentum bin ( width @xmath160 ) according to the rule @xmath161 where @xmath162 is the density upstream of the shock and @xmath163 is the local background fluid velocity relative to the initially stationary upstream plasma .
@xmath164 is the local thermal energy density .
this prescription is designed to inject a suitable energy density of cr at the shock , dependent on the choice of the constant , whilst avoiding a negative value of @xmath164 due to excessive transfer of energy to cr from cold plasma at the foot of the shock .
the resulting cr current density ahead of the shock is displayed in panel ( e ) of figures 2 and 3 .
the peak cr current density in the population freely escaping ahead of the shock in figure 2 is @xmath165 this current density corresponds to @xmath166 , giving @xmath167 @xmath168 @xmath169 we use a spatial grid with @xmath170 .
ten cells in momentum cover an energy range from 100tev to 1pev with logarithmic spacing .
there are 32 cells in each of @xmath171 and @xmath172 with periodic boundary conditions . in contrast
5676 cells are used in @xmath173 with reflective boundary conditions for the mhd part of the code and for cr at the right hand boundary .
cr reaching the left hand boundary are disposed of on the assumption that they escape freely .
correspondingly , the computational box extends @xmath174 m by @xmath175 m by @xmath175 m .
the box - size in @xmath171 and @xmath172 is comparable with the initial cr larmor radius , but the larmor radius contracts significantly as the magnetic field is amplified .
the cell - size is @xmath176 times the wavelength of the fastest growing mode , which is barely sufficient to represent the initial growth of the instability , but the characteristic scalelength of the instability increases rapidly as its amplitude grows .
these parameters are only marginally sufficient to represent the physics , but for example a halving of the computational cell - size would increase the computational cost by a factor of 16 .
the simulation in figure 3 was run for @xmath177 using 128 processors for 75 hours on the scarf - lexicon cluster at the uk rutherford appleton laboratory .
the mhd part of the code is the same as that used by lucek & bell ( 2000 ) and bell ( 2004 ) .
the vfp part of the code that models the cr uses a second order runge - kutta scheme in configuration space , a donor - cell scheme in magnitude of momentum , and the boris algorithm for rotation in momentum due to magnetic field as used in particle - in - cell codes ( birdsall & langdon 1985 ) .
the vfp equations are formulated in tensor notation rather than in spherical harmonics , but the two formulations are similar for the low order expansion used here , and their numerical solution is informed by experience with the kalos spherical harmonic code ( bell et al , 2006 ) where runge - kutta advection is found to be robust , sufficiently accurate , and a good fit to the form of the equations .
numerical diffusion due to limited spatial resolution is ameliorated by designing the simulation to initialise the upstream background plasma at rest relative to the computational grid .
the more usual approach of initiating the simulation by setting the background plasma in motion towards a reflective boundary would exacerbate the effect of numerical diffusion on small structures during advection . instead , a dense piston is initialised moving leftward from the right boundary , pushing a shock before it into the stationary background plasma .
results from the 3d simulation are presented in figures 2 and 3 at @xmath178 and @xmath179 respectively .
note that the horizontal spatial scales , in the direction of the shock normal , are artificially compressed by a factor of 24 .
the actual aspect ratio of the computational box is 177:1 .
the plot of the background plasma density in panel ( a ) shows the position of the dense plasma piston propagating leftwards and pushing the shock ahead of it .
the high pressure region in panel ( b ) is due to plasma heating at the shock .
the cr pressure is plotted in panel ( c ) .
the maximum cr pressure occurs at the shock .
panel ( c ) of figure 2 shows the cr population dividing into a population freely propagating ahead of the shock and a population confined by magnetic field ( panel ( d ) ) near the shock .
cr confined near the shock continue to be accelerated .
the banded structure in the cr pressure ahead of the shock in panel ( c ) is exaggerated by the colour coding .
it represents only a variation of a few percent in the cr pressure and probably arises from a small oscillation in the injection rate at the shock .
also , the horizontal spatial compression of figure 2 distorts the aspect ratio of the bands .
the escaping cr are an essential aspect of the process of cr confinement by the self - generated magnetic field .
they excite the instability and carry the escaping areal charge @xmath180 identified in equation ( 1 ) as necessary for substantial field amplification . in figure 2
the escaping cr are only just reaching the left hand of the grid where the integral @xmath181 is small and field amplification is negligible .
in contrast , immediately in front of the shock in figure 2 @xmath181 has reached the critical value needed for strong field amplification causing the onset of cr confinement . by the time of figure 3 most of the escaping population has passed through the free escape boundary at the left hand end of the grid . by this time a large magnetic field in the upstream plasma has switched off cr escape creating an expanse of low cr density between the confined and escaping cr populations .
panel ( d ) shows that the magnetic field is amplified by an order of magnitude by the escaping cr over a large distance ahead of the shock .
the separation of cr into escaping and confined populations matches expectations from the argument in section 3 .
the cr charge per unit shock area in the escaping population is @xmath182 coulomb m@xmath183 in figure 2 in good agreement with the estimate of @xmath184 coulomb m@xmath183 from equation ( 1 ) .
panels ( e ) in figures 2 and 3 plot the cr current density ahead of the shock .
figure 4 follows the calculations of figures 2 and 3 to later times with the same parameters but in 2d instead of 3d to reduce the demand on computer resources .
the top row of plots in figure 4 is the 2d equivalent of the 3d results in figure 2 at @xmath178 .
the dimensions of the spatial grid are the same in figure 2 and the top row of figure 4 . in subsequent rows of figure 4
the spatial dimensions normal to the shock are expanded by factors of 2 , 3 and 4 such that cr travel the length of the grid in the respective times ( @xmath185 , @xmath186 & @xmath187 ) . in 3d at @xmath178
the dip in the cr density separating the escaping and confined cr has only recently formed as seen in figure 2(c ) . at the same time in 2d ( figure 4 top row ) the separation between the confined and escaping populations is relatively more developed .
otherwise the results are similar in 2d and 3d .
the reason for the slightly earlier development of cr confinement in figure 4 is that magnetic structures can expand more freely in 2d . in 2d ,
magnetic structures expand in the ignored dimension without running into stationary or counterpropagating plasma .
nonlinear development in 2d is therefore less inhibited and the magnetic field grows more rapidly .
the magnetic field in 2d reaches a steady value similar to that found in the 3d calculation .
the eventual limit on the magnitude of the magnetic field may be set by magnetic tension which operates equally in 2d and 3d .
comparison of the cr distributions at different times in figure 4 shows that at early times ( @xmath178 ) only the low energy cr are confined . at @xmath188 cr
are confined at 100 and 167tev with confinement beginning to occur at 278tev and no evidence of the development of two separate populations in the few cr reaching 464tev . by @xmath189 a larger number of cr have reached 464tev with a local minimum in cr density @xmath190 evident at all energies up to 464tev in front of the shock . by @xmath189 cr at 100 and 167
tev are strongly confined with a short precursor scaleheight ahead of the shock , which is unsurprising since the larmor radius of a 100tev proton in a @xmath191 magnetic field is @xmath192 . by @xmath193 cr
are confined at all energies up to 464tev . at all four times in figure 4
the cr spectrum is much steeper than the steady - state test particle spectrum ( @xmath194 ) because of cr loss into the downstream plasma .
magnetic field amplification takes place ahead of the shock and the downstream field only becomes large when the shock overtakes the field amplified in the upstream .
hence at early times cr escape downstream .
this causes a reduction in the number of cr accelerated to high energy at the shock , thereby steepening the spectrum .
downstream confinement at the shock improves at later times as shown by the downstream gradients in @xmath88 at @xmath189 and @xmath193 .
however , unphysical numerical relaxation of the spatially compressed downstream magnetic field due to limited spatial resolution may be playing a part in allowing cr to escape downstream .
there is slight evidence of pulsed acceleration as seen in the three separate peaks in the 278tev cr density at @xmath189 and @xmath193 but overall the cr profiles develop in an orderly manner .
our simulations support the model developed in section 3 . according to our model cr
are confined and accelerated if the electrical charge of cr escaping upstream of the shock reaches @xmath195 .
we now apply this to spherical snr shocks .
the cr current density at a radius @xmath196 is @xmath197 due to cr accelerated to energy @xmath198 when the shock radius was @xmath199 . since only the highest energy cr escape upstream we assume that the cr reaching the radius @xmath196 are monoenergetic with energy @xmath200 .
@xmath39 evolves as the shock expands . when the snr shock reaches the radius @xmath196 , cr are confined if @xmath201 differentiating this equation with respect to @xmath196 and assuming a power law dependence of density on radius , @xmath202 , @xmath203 defining @xmath204 , @xmath205 , @xmath206 , @xmath207 ( such that @xmath66 is approximately the electron density in @xmath208 ) , and taking @xmath209 for expansion into a uniform medium , @xmath210 a snr with @xmath211 , @xmath72 and @xmath212 , representative of cas a , would then accelerate cr to @xmath213140tev
. a snr with @xmath22 , @xmath214 and @xmath215 , representative of sn1006 , would accelerate cr to @xmath213180tev .
these energies are a factor of 10 lower than the energy of the knee .
abbasi et al ( 2012 ) place the knee at 4 - 5pev , although data from other experiments indicate a lower energy and the turnover in the spectrum is not well defined as shown in their figure 15 .
our model places a considerable question mark over the ability of the well - known historical snr to accelerate cr to the knee .
acceleration by snr such as cas a and sn1006 fails to reach the knee in our analysis because their expansion is already significantly decelerated .
zirakashvili & ptuskin ( 2008a ) also reach the conclusion ( their table 2 ) that the historical snr do not accelerate cr beyond 100 - 200 tev , but note that their parameter @xmath216 differs from our parameter @xmath43 by a factor of two ( @xmath217 )
. initial expansion velocities of very young snr can reach @xmath218 ( manchester et al 2002 ) or possibly higher for some types of sn ( chevalier & fransson 2006 ) . according to equation
( 21 ) expansion into a density of @xmath219 at @xmath218 for 16 years to a shock radius of 0.5 parsec accelerates cr to @xmath220 .
moreover , the energy processed through the shock is comparable to that for expansion to 1.5 parsec at @xmath221 , thereby contributing a comparable cr energy content to the galactic energy budget .
a complementary perspective on the same problem is obtained by expressing the maximum cr energy in terms of the mass @xmath222 swept up by the shock and the characteristic energy of the blast wave @xmath223 . if @xmath224 is the mass in units of a solar mass and @xmath225 is the energy in units of @xmath226j , the maximum cr energy is @xmath227 which indicates that a typical snr should be able to accelerate cr to @xmath228 and that the maximum cr energy is greater for the same energy @xmath229 given to a smaller mass @xmath230 .
the maximum cr energy is nearly independent of density @xmath66 because a shock expands to a larger radius in a low density medium before deceleration ( hillas 2006 ) . in the self - similar sedov phase
, @xmath229 is constant , @xmath230 is proportion to @xmath231 and hence @xmath232 . the maximum cr energy decreases with radius during the sedov phase until magnetic field amplification ceases , at which point our analysis in terms of the charge carried by escaping cr is inapplicable
. equation ( 20 ) with @xmath233 indicates that the highest cr energies might be achieved by snr shocks expanding into a dense circumstellar medium previously ejected as a wind from the sn progenitor .
@xmath234 is independent of radius in a steady wind , in which case the cr energy depends only on @xmath235 and @xmath43 according to equation ( 20 ) thus favouring cr acceleration in the very early stages of rapid expansion as previously suggested by vlk & biermann ( 1988 ) and bell & lucek ( 2001 ) .
the maximum cr energy for a snr expanding into a wind carrying @xmath236 solar masses per year and shed with a velocity of @xmath237 is @xmath238 indicating that pev energies are attainable . cr may be accelerated to energies beyond the knee if the initial shock velocity is @xmath239 and the shock expands into a particularly dense wind or otherwise dense circumstellar medium .
the estimates made in this section are subject to considerable numerical uncertainty .
for example our estimates for the efficiency @xmath240 or the escaping charge @xmath241 could be uncertain by a factor of 2 or 3 .
however , the accumulated error in our estimates would need to be a factor of 10 for the well - known historical snr to account for acceleration to the knee .
our arguments are not completely watertight but we tentatively conclude that acceleration to the knee takes place in younger relatively undecelerated snr .
in this section we discuss the energy spectrum of escaping cr integrated over the lifetime of the snr .
the energy @xmath39 of escaping cr changes as the snr evolves in radius and expansion velocity .
the integrated energy spectrum of cr escaping into the interstellar medium ( ism ) need not be the same as the spectra of cr at the shock during acceleration or that of cr carried downstream into the interior of the snr .
related analyses based on different models of cr escape can be found in caprioli et al ( 2010a ) , drury ( 2011 ) , ohira et al ( 2010 ) , ptuskin & zirakashvili ( 2003 ) and ptuskin et al ( 2010 ) .
we assume a power - law density gradient , @xmath242 where @xmath209 for a uniform circumstellar medium and @xmath233 for a steady pre - supernova wind .
we further assume that the shock velocity can be approximated as a power law @xmath243 over a sufficiently large part of the snr s evolution . in this approximation @xmath244 and @xmath245 . in the sedov phase @xmath246 . from equation ( 20 ) @xmath247
. in a uniform circumstellar medium ( @xmath209 ) the energy @xmath39 of escaping cr decreases during expansion if @xmath248 , which is equivalent to @xmath1 decreasing more rapidly than @xmath249 .
if a snr expands into a steady pre - supernova wind ( @xmath233 ) the maximum cr energy always decreases with time provided @xmath1 decreases with time ( @xmath250 ) as expected .
let @xmath251 be the total energy given to cr above an energy @xmath39 . by definition of @xmath43
the cr energy flux escaping upstream is @xmath252 , so @xmath253 and @xmath254 where cr with energy @xmath39 escape when the the shock radius is @xmath196 . from equation ( 20 )
@xmath255 and @xmath162 , @xmath256 and @xmath257 are the values of @xmath42 , @xmath1 and @xmath39 at a reference radius @xmath258 .
manipulating these equations gives @xmath259 @xmath260 and @xmath261 is the cr differential spectrum in energy .
for snr expansion into a uniform medium the cr spectral index is @xmath262 , and for expansion into a wind the index is @xmath263 . during the sedov phase @xmath209 and @xmath246 ,
giving @xmath264 although the analysis only applies while the magnetic field is being amplified by the nrh instability .
a slightly less rapid decrease in shock velocity , @xmath265 ( @xmath266 ) , would reproduce the cr spectral index ( @xmath267 ) inferred for cr at their source at energies less than 1pev ( gaisser et al 1998 , hillas 2005 ) .
the spectral index of cr escaping in the sedov phase , @xmath268 , is the same as that for test particle acceleration at a strong shock .
there is no obvious reason why this should be so .
we emphasise that this discussion and the derived spectral index @xmath269 or @xmath270 applies only to cr escaping upstream from the shock during snr expansion
. a further population of lower energy cr are carried into the centre of the snr where they reside until they are released into the interstellar medium when the snr slows , disintegrates and dissolves into its surroundings .
these lower energy particles lose energy adiabatically as the snr expands but they can be expected to contribute most of the galactic population of low energy cr .
in section 8 we suggested that acceleration beyond the knee might result from high velocity expansion into a dense circumstellar medium .
the cr population beyond 1pev is an uncertain mix of protons and heavy ions .
the overall spectral index of cr released into the galaxy by snr beyond 1pev can be approximated as @xmath271 with further spectral steepening occurring during propagation from the source to the earth .
this spectral index at source is predicted by our analysis if @xmath272 ( @xmath273 ) for expansion into a steady wind .
it would correspond to a reduction in the shock velocity by a quite reasonable factor of 2.1 between @xmath274 and @xmath275 .
expansion into a uniform medium would require @xmath276 ( @xmath277 ) equivalent to a reduction in shock velocity by a less reasonable , but not impossible , factor of 5.9 between @xmath274 and @xmath275 .
this supports the contention that acceleration to @xmath278 may occur at a snr shock expanding into a circumstellar wind .
the hillas parameter @xmath279 ( shock velocity times magnetic field times spatial scale ) provides an estimate of the energy @xmath15 to which cr can be accelerated under various circumstances ( hillas 1984 ) .
@xmath280 is a numerical factor of order unity , probably lying in the range between 1/8 and 3/8 depending upon the cr diffusion coefficient ( lagage & cesarsky 1983a , b , bell 2012 ) .
@xmath281 where @xmath282 ( @xmath283 for @xmath284 ) . combining this with equation ( 21 )
gives an estimate of the pre - shock magnetic field required to accelerate cr to @xmath15 : @xmath285 the post - shock field is @xmath286 larger due to compression at the shock . for our simulation parameters (
@xmath214 , @xmath287 ) the estimated post - shock maximum magnetic field is @xmath288 which is consistent with the field of @xmath289 seen in panel ( d ) of figures 2 and 3 .
this confirms that the magnetic field in the simulation is amplified sufficiently to confine and accelerate cr and that diffusion in the cr precursor is approximately bohm , @xmath290 .
the above estimate of the magnetic field is derived on the basis that the magnetic field must be sufficient to contain and accelerate cr to the energy estimated in equation ( 21 )
. continued growth in the magnetic field would inhibit cr escape and remove the cr current that drives the nrh instability .
part of the energy generated by the nrh instability is stored in the kinetic energy of plasma motions .
these motions might continue to stretch magnetic field lines and further increase the magnetic field after the cr current becomes inhibited .
further release of cr into the upstream plasma would then be heavily restricted until the magnetic field relaxes to a lower level .
this might result in oscillation about a marginal state defined by a balance between magnetic field amplification and cr escape upstream .
weak evidence for periodic releases of cr into the upstream plasma can be found in the plot of the 278tev cr density at @xmath291 in figure 4 but on the whole the system appears to evolve without oscillation . in planar geometry escaping cr are in principle capable of generating magnetic field at an unlimited distance ahead of the shock . in the spherical geometry of an expanding snr
the cr current decreases with distance ahead of the shock , @xmath292 , so continuous cr escape is needed to amplify the magnetic field at a general radius @xmath196 before the shock reaches that point .
hence the marginal balance between cr escape and magnetic field generation is more likely in spherical than planar geometry .
the above discussion assumes that magnetic field growth and cr acceleration is determined by the growth rate of the nrh instability .
however , it is possible that the instability might saturate and stop growing before it reaches that given by equation ( 28 ) .
bell ( 2004 , 2009 ) argues that tension in the field lines limits amplification when @xmath293 for magnetic field structured on the scale of a cr larmor radius .
this implies a saturation magnetic energy density @xmath294 and predicts a saturated upstream magnetic field @xmath295 with a further @xmath296 increase at the shock .
the ratio of the magnetic field given by equation ( 28 ) to the saturated magnetic field is @xmath297 which implies that tension in the magnetic field does not stop the field growing to that given in equation ( 28 ) .
however , it would not require magnetic field growth to overshoot that indicated by equation ( 28 ) by a large factor before the magnetic tension intervenes to halt growth .
according to equation ( 28 ) the magnetic energy density is proportional to @xmath106 , whereas the magnetic energy density determined by saturation is proportional to @xmath298 .
observations of snr slightly favour a dependence on @xmath298 , but the difference is too close to be called ( vink 2006 ) . from equation ( 20 )
( @xmath299 for @xmath209 ) and the modified hillas condition ( @xmath300 ) the required post - shock magnetic energy density is @xmath301 for magnetic field structured on the scale of the larmor radius of the highest energy cr we should assume @xmath240 as the fraction of @xmath106 given to the highest energy cr in which case the the post - shock magnetic energy density would be @xmath302 of @xmath106 , allowing for compression at the shock .
vlk et al ( 2005 ) find observationally that the post - shock magnetic energy density is typically @xmath303 of @xmath106 in the historical snr .
however , magnetic field will also be amplified on the scale of the larmor radii of low energy as well as high energy cr .
the difference between @xmath302 and @xmath303 may be explained by integration over the magnetic structures on scales varying by six orders of magnitude corresponding to the difference between the larmor radii of gev and pev protons .
if this is the case , most of the magnetic energy at the shock resides at scalelengths too short to accelerate cr to pev energies . the magnetic field inferred from x - ray synchrotron observations of a snr shock
should not be inserted without adjustment into the hillas parameter to estimate the maximum cr energy .
our analysis of cr escape leads to the result that the energy density of the magnetic field confining the highest energy cr is proportional to @xmath106 .
previous analyses of cr escape often start from the assumption that the magnetic energy density is proportional to @xmath106 and consequently they produce similar results for the spectrum of escaping cr .
comparable results for the cr spectrum produced in the sedov phase can be found in caprioli et al ( 2010a ) , drury ( 2011 ) , ohira et al ( 2010 ) , ptuskin & zirakashvili ( 2003 ) and ptuskin et al ( 2010 ) .
for example a @xmath105 energy spectrum for escaping cr in the sedov phase has previously been derived by berezhko & krymskii ( 1988 ) , ptuskin & zirakashvilii ( 2005 ) , caprioli et al ( 2010a ) and drury ( 2011 ) .
in our model only the highest energy cr escape upstream of the shock . at any given time more cr energy
is carried away downstream into the snr than escapes into the galaxy .
efficient production of galactic cr might therefore seem impossible .
however , the cr energy carried into the interior of the remnant is subsequently recycled to drive the snr expansion and is available to accelerate further cr at a later stage .
the overall efficiency of the production of galactic cr is demonstrated by integrating over the lifetime of the snr . assuming a @xmath105 energy spectrum ( @xmath304 , @xmath246 ) equation ( 26 ) can be integrated to deduce a total cr energy input to the galaxy : @xmath305 between energies @xmath306 and @xmath307 ( @xmath308 and @xmath228 respectively ) . a small value of @xmath43 ( @xmath309 ) is balanced by the factor @xmath310 which results from recycling cr energy carried into the interior of the snr .
it is occasionally remarked that cr acceleration by very young snr during the first few years can not inject sufficient energy into the galaxy to account for cr at energies beyond the knee because the snr shock is small .
however , if the cr spectrum connects smoothly across the knee and the spectrum beyond the knee of escaping cr matches observation as shown to be possible in section 9 then it follows that the cr energy input is sufficient to match the galactic energy budget .
the central message of this paper is that cr of a given energy escape freely ahead of a shock until magnetic field amplification takes place to inhibit propagation .
the condition for propagation inhibition is that a sufficient number of cr must escape upstream for the nrh instability to grow through @xmath311 e - foldings at the growth rate of the fastest growing mode .
since the instability is driven by the cr current , the condition is that a cr electric charge @xmath312 per unit area must escape through a spherical surface surrounding the snr to amplify the magnetic field and inhibit cr escape through that surface . since high energy cr
carry less charge than low energy cr for a fixed cr energy flux the condition on @xmath83 determines the energy of escaping cr .
we find that the energy @xmath200 of escaping cr is proportional to @xmath313 as given by equation ( 21 ) .
the energy @xmath200 varies during the evolution of the snr and determines the energy spectrum of cr injected into the interstellar medium by snr . in our estimation , the historical snr ( cas a , tycho , kepler , sn1006 ) are currently accelerating cr to @xmath314 tev .
acceleration to the knee at a few pev takes place in snr at an earlier stage of evolution when the shock velocity is @xmath315 or greater .
this is an unsurprising conclusion since if the historical snr were to accelerate cr to the knee we would be asking why even higher energy cr were not being produced by younger snr .
observations by the planned cherenkov telescope array ( cta ) should be crucial in testing our conclusions ( hinton & hofman 2010 , aharonian 2012 ) .
acceleration beyond the knee may take place in very young snr expanding at @xmath316 into a dense circumstellar pre - supernova wind .
the spectral index of escaping cr is consistent with the measured galactic cr spectrum at energies less than 1pev . beyond the knee the proton spectral index is uncertain both theoretically and observationally .
the theoretical prediction depends on the rate at which the snr shock decelerates during its early expansion .
the magnetic field can be estimated from the hillas parameter as the field needed to accelerate cr to the escape energy .
the field is close to , but slightly less than , the saturation field determined by tension in the magnetic field .
the predicted magnetic fields are consistent with those observed in snr if allowance is made for the large range of scalelengths , corresponding to the range of cr larmor radii , in the structure of the magnetic field .
the model is tested against numerical simulation .
the mhd / vfp code is three - dimensional in space and one dimensional in cr momentum with anisotropy included to second order .
the computational parameters are pushed to their limit to allow solution of this multi - scale multi - dimensional problem but the results support the analytic model . in the simulation cr
are seen to escape upstream with the electrical charge predicted by theory and magnetic field is strongly amplified to the predicted level .
the research leading to these results has received funding from the european research council under the european community s seventh framework programme ( fp7/2007- 2013 ) / erc grant agreement no .
247039 and from grant number st / h001948/1 made by the uk science technology and facilities council .
we thank the stfc s e - science facility for access to the scarf computing cluster .
we thank an anonymous referee for comments that considerably improved the presentation of this paper .
abbasi r. et al , 2012 , submitted to astropart phys , arxiv.1207.3455 aharonian f.a . , 2012 , astropart phys , available on - line : 10.1016/j.astropartphys.2012.08.007 axford w.i .
, leer e. & skadron g. , 1977 , proc 15th int .
cosmic ray conf . , 11 , 132 bell a.r . , 1978 , mnras , 182 , 147 bell a.r . , 2004 , mnras , 353 , 550 bell a.r . , 2005 , mnras , 358 , 181 bell a.r . , 2009 , plasma phys cont fusion , 51,124004 bell a.r . , 2012 , astropart phys , available on - line : 10.1016/j.astropartphys.2012.05.022 bell a.r .
, evans r.g . & nicholas d.j . , 1981 , phys rev lett 46 , 243 bell a.r . &
lucek s.g .
, 2001 mnras 321 433 bell a.r . ,
robinson a.p.l .
, sherlock m. kingham r.j . & rozmus w. , 2006 , plasma phys cont fusion , 48 , r37 bell a.r . ,
schure k.m . , reville b. , 2011 , mnras 418 , 1208 berezhko e.g. & ellison d.c . , 1999 ,
apjl 526 , 385 berezhko e.g. & krymskii g.f . , 1988 , sov phys usp 31 , 27 berezhko e.g. , ksenofontov l.t .
& vlk h.j . , 2003 ,
a&a 412 , l11 birdsall c.k .
& langdon a.b . , 1985 ,
_ plasma physics via computer simulation _ , mcgraw hill , singapore .
blandford r.d . &
ostriker j.p . , 1978 , apj , 221 , l29 bykov a.m. , osipov s.m . &
ellison d.c . , 2011 ,
mnras 410 , 39 caprioli d. , blasi p. & amato e. , 2009a , mnras 396 , 2065 caprioli d. , blasi p. , amato e. & vietri m. , 2009b , mnras 395 , 895 caprioli d. , blasi p. & amato e. , 2010a , astropart phys 33 , 160 caprioli d. , blasi p. & amato e. , 2010b , astropart phys 33 , 307 chevalier r.a . & fransson c. , 2006 , apj , 651 , 381 drury l.oc .
, 2011 , mnras 415 , 1807 drury l.oc . & downes t.p . ,
2012 , mnras in press , arxiv:1205.6823 drury l.oc . & falle s.a.e.g . , 1986 , mnras 222 , 353 .
ellison d.c . &
bykov a.m. , 2011 , apj 731 , 87 ellison d.c .
& eichler d. , 1984 , apj 286 , 691 ellison d.c .
, jones f.c . &
eichler d. , 1981 , j geophys 50 , 110 ellison d.c . , slane p. , patnaude d.j .
& bykov a.m. , 2012 , apj 744 , 39 gaisser t.k .
protheroe r.j .
& stanev t. , 1998 , apj 492 , 219 hillas a.m. , 1984 , ara&a 22 , 425 hillas a.m. , 2005 ,
j phys g 31 , r95 hillas a.m. , 2006 , j phys conf ser 47 , 168 hinton j.a .
& hofmann w. , 2010 , ann rev astron astrophys , 47 , 523 johnston t.w .
, 1960 , phys rev 120 , 1103 krymskii g.f .
, 1977 , sov phys dokl , 23 , 327 kulsrud r. & pearce w.p . , 1969 , apj 156 445 lagage o. & cesarsky c.j .
, 1983a , a&a 118 223 lagage o. & cesarsky c.j .
, 1983b , apj 125 249 lucek s.g . &
, 2000 , mnras 314 65 malkov m.a .
& diamond p.h . , 2009 ,
apj 692 , 1571 malkov m.a .
, diamond p.h . &
, 2000 , apj 533 , l171 manchester r.n . ,
gaensler b.m . ,
wheaton v.c . , staveley - smith l. , tzioumis a.k .
, bizunok n.s . , kesteven m.j .
& reynolds j.e . , 2002 , pub .
austr . , 19 , 207 ohira y. , mirase k. & yamazaki r. , 2010 , a&a 513 a17 ptuskin v.s . &
zirakashvili v.n . , 2003 , a&a 403 , 1 ptuskin v.s .
& zirakashvili v.n . , 2005 ,
a&a 429 , 755 ptuskin v.s .
, zirakashvili v.n . & seo e .- s . , 2010 , apj 718 , 32 reville b. & bell a.r . , 2012 ,
mnras 419 , 2433 reville b. & bell a.r .
, 2013 , mnras in press , arxiv:1301.3173 reville b. , kirk j.g . &
duffy p. , 2009
, apj 694 , 951 rogachevskii i. , kleeorin n. , brandenburg a. & eichler d. , 2012 , apj 753:6 schure k.m . & bell a.r .
, 2011 , mnras 418 , 782 schure k.m .
, bell a.r .
, drury l.oc . &
bykov a.m. , 2012 , space science reviews 173 , 491 stage m.d .
, allen g.e .
, houck j.c . & davis j.e . , 2006 ,
nature physics 2 , 614 thomas a.g.r .
, tzoufras m. , robinson a.p.l . , kingham r.j . , ridgers c.p . , sherlock m. , & bell a.r . , 2012 , j comp phys 231 , 1051 tzoufras m. , bell a.r . ,
norreys p.a .
& tsung f.s . , 2011 , j comp phys 230 , 6475 uchiyama y. et al , 2007 , nature 449 , 576 vink j. 2006 , proc of the the x - ray universe 2005 ( esa sp-604 ) p319
a. wilson vink j. & laming j.m . , 2003 , apj , 584 , 758 vladimirov a. , ellison d.c .
& bykov a. , 2006 , apj 652 , 1246 vlk h.j . ,
berezhko e.g. & ksenofontov l.t . , 2005 ,
a&a , 433 , 229 vlk h.j . & biermann p. , 1988
apjl 333 , l65 weibel e.s . , 1959
, phys rev lett 2 , 83 wentzel d.g .
, 1974 , ara&a 12 , 71 zirakashvili v.n . & ptuskin v.s . , 2008a , apj 678 , 939 zirakashvili v.n . , ptuskin v.s
& vlk h.j .
, 2008b , apj 678 , 255 the cr electric current @xmath29 drives the amplification of magnetic field through the nrh instability . throughout this paper
we express @xmath317 as a fraction of the cr current needed to carry the characteristic energy flux @xmath298 : @xmath318 where @xmath39 is the characteristic cr energy in ev . in this appendix
we briefly explain why we choose @xmath319 as our best estimate ( see also bell 2004 ) . in the absence of cr acceleration
the thermal energy density downstream of a strong shock is @xmath320 from the rankine - hugoniot relations . assuming that a third of this energy is given to cr as required for efficient cr production by snr , the downstream cr energy density is @xmath321 . from continuity across the shock the cr energy density immediately ahead of the shock is also @xmath321 and the cr energy flux relative to the upstream plasma is @xmath322 .
however , only the highest energy cr escape upstream .
lower energy cr do not penetrate far upstream and they amplify magnetic field on too small a scale to engage the escaping cr . hence the analysis in this paper depends on the current carried only by high energy cr . for a @xmath105 cr energy spectrum extending from @xmath323 to @xmath324 the energy
is spread equally across each decade in energy with a fraction @xmath325 associated with any energy @xmath39 .
the energy flux carried by cr with energy @xmath39 is then @xmath326 .
the energy flux carried by cr streaming at velocity @xmath327 with number density @xmath328 and energy @xmath200 is @xmath329 .
consequently we assume that @xmath318 where @xmath240 .
equation 2 for the electric current @xmath317 carried by escaping cr assumes ( i ) that cr escape upstream in a small range of momenta just above @xmath85 ( ii ) that any cr reaching a momentum @xmath85 or higher freely stream ahead of the shock at a velocity much greater than the shock velocity @xmath1 ( iii ) that escape is predominantly upstream .
the electric current density is @xmath330 where @xmath331 is the anisotropic part of the distribution fucntion representing cr drift in the @xmath199 direction . in a steady state on an acceleration timescale @xmath332 can be neglected from equation 11a .
the second term @xmath333 can also be neglected since the cr are assumed to be freely streaming at energies above @xmath334 and advection at the fluid velocity is small .
in one spatial dimension @xmath199 equation 11a then reduces to @xmath335 for energies above @xmath334 .
integrating in space across the shock and in momentum from @xmath85 upwards gives @xmath336 where @xmath337 is the change in velocity across the shock .
since @xmath338 for a strong shock @xmath339 as in equation 2 . here
we show how equation 12 and its vector equivalent can be derived from equations 11 . for simplicity we omit terms representing advection at the fluid velocity and the effect of the magnetic field .
these can easily be inserted at the end of the derivation .
the difficult part is the replacement of the stress tensor @xmath131 by the vector @xmath139 .
the most transparent way of presenting the derivation is to write it out in terms of individual components in @xmath171 , @xmath172 and @xmath173 directions . in the following
we present the derivation of the equation for @xmath340 .
the derivations of the equations for @xmath341 and @xmath342 follow the same pattern . in component form , the relevant equations 11 are @xmath343 @xmath344 @xmath345 @xmath346 @xmath347 eliminating the components of the stress tensor between these equations gives @xmath348 @xmath349 which is the @xmath171 component of the vector equation @xmath350 and the same derivation holds for the @xmath172 and @xmath173 components of the equation .
this equation is second order in time differential .
it can be separated into two first order equations : @xmath351 @xmath352 these equations become equations 12 and 13 with the addition of fluid advection , rotation of @xmath153 by the magnetic field , and damping of the stress tensor at a rate @xmath353 as discussed above equation 12 in section 4 . here
we demonstrate that the equations set out in section 4 describe the essential cr propagation modes for monenergetic cr .
for simplicity we neglect the damping term ( @xmath358 ) . propagating cr solutions in a stationary background plasma
can be found by setting @xmath359 , @xmath360 and @xmath361 in equation ( 13 ) : @xmath362 @xmath363 @xmath364 the resulting dispersion relation is @xmath365 @xmath366 where @xmath120 is the angle between @xmath155 and @xmath357 .
there are two independent modes in the absence of magnetic field ( @xmath367 ) .
the mode propagating at @xmath368 represents the motion of freely propagating cr with @xmath153 parallel to @xmath155 .
the mode propagating at @xmath369 is the transverse mode representing the motions of cr with @xmath153 perpendicular to @xmath155 .
the transverse mode propagates more slowly because cr velocities are aligned preferentially away from the direction of propagation . in the presence of a magnetic field
a longitudinal mode still propagates parallel to @xmath357 ( @xmath370 ) at @xmath368 representing free propagation along field lines unaffected by the field .
the transverse mode is relatively unaffected by the magnetic field at wavelengths shorter than a larmor radius @xmath371 . at wavelengths
longer than the larmor radius the transverse mode propagates more slowly as the transverse cr current rotates rapidly in the magnetic field .
when mode propagation is directed across the magnetic field ( @xmath372 ) the wave frequency is given by @xmath373 where @xmath12 is the cr larmor radius . in the limit of wavelengths smaller than the larmor radius
, the frequency converges to those derived for the longitudinal and transverse waves in zero magnetic field as expected . at long wavelengths ( @xmath374 ) the frequency converges to the larmor frequency , representing cr rotation in the magnetic field , without significant propagation .
we now investigate whether the approximate treatment of cr kinetics in section 4 is adequate to model the nrh instability .
we derive the dispersion relation for the nrh instability in a simple case with the following assumptions .
the cr are mono - energetic with momentum @xmath119 .
the zeroth order cr current @xmath375 , the wavenumber @xmath155 and the zeroth order magnetic field @xmath376 are all parallel .
the background plasma is at rest to zeroth order , @xmath377 .
the first order perturbations to the magnetic field @xmath378 , cr current @xmath379 , and plasma velocity @xmath380 are all perpendicular to the zeroth order magnetic field and the wavevector @xmath155 . since the modes are transverse the plasma density is unperturbed to first order : @xmath381 .
the coupled linearised forms of equations ( 7 ) , ( 13 ) and the maxwell equation for @xmath382 are then : @xmath383 @xmath384 @xmath385 @xmath386 where @xmath387 . for circular polarisation
any first order perturbation @xmath388 ( @xmath389 , @xmath390 , @xmath391 or @xmath392 ) satisfies @xmath393 where @xmath394 is a unit vector in the direction of @xmath376 , giving @xmath395 @xmath396 @xmath397 @xmath398 as discussed in section 4 we set the scattering frequency @xmath353 equal to the cr larmor frequency . for snr conditions the nrh growth rate ( @xmath399 ) is much less than the cr larmor frequency , in which case the dispersion relation simplifies to @xmath400 where @xmath401 is the alfven speed , @xmath12 is the cr larmor radius in the magnetic field @xmath56 and @xmath402 is the nrh growth rate where it dominates in the range @xmath403 .
@xmath404 according to the sense of the circular polarisation as determined by the sign of @xmath57 . for wavelengths much shorter than the larmor radius ( @xmath405 )
@xmath406 which represents a purely growing instability for wavenumbers less than @xmath407 in the appropriate polarisation @xmath408 .
tension in the magnetic field correctly damps waves with wavenumbers greater than @xmath407 ( bell 2004 ) .
the truncated tensor analysis is correct on scales smaller than the larmor radius because cr trajectories are then relatively unaffected by perturbations in the magnetic field .
this is the important regime in which the rapidly growing nrh instability amplifies the magnetic field . at wavelengths
longer than the larmor radius ( @xmath409 ) the approximate tensor expansion gives @xmath410 in comparison the correct dispersion relation in this limit is @xmath411 for the resonant instability and @xmath412 for the non - resonant instability which in fact is stable for mono - energetic cr in this limit .
the alfven term @xmath413 is negligible at long wavelengths so the tensor expansion gives a growing mode with @xmath414 in both resonant and non - resonant polarisations .
figure 5 compares the approximate tensor dispersion relation ( top row ) with the correct dispersion relations of bell ( 2004 ) ( bottom row ) for both the resonant instability ( @xmath415 ) that dominates for @xmath416 and the rapidly growing non - resonant nrh instability ( @xmath415 ) that dominates for @xmath417 .
crucially the tensor expansion accurately calculates the nrh growth rate in the range @xmath418 . outside this range
the growth rate is too small for the instability to be effective .
the tensor expansion reproduces the growth of the resonant instability for @xmath416 although it incorrectly produces weak growth of the non - resonant instability in this regime . the crucial point for the simulation
is that the tensor expansion gives an accurate account of instability where the growth rate is large .
the simulation makes heavy demands on computational resources because it models spatial scales encompassing the wavelength of the fastest growing mode , the cr larmor radius , and the free propagation of cr ahead of the shock .
it is three - dimensional in configuration space and models the cr distribution in momentum .
the spatial cell - size @xmath419 must be small enough to allow the nrh instability to grow from its initially small scale ( @xmath420 ) and the timestep @xmath421 must be short enough to resolve cr crossing one computational cell ( @xmath422 ) .
the simulation must be run for a time @xmath423 at least 10 times the growth time of the fastest growing mode @xmath424 and the length of the computational grid @xmath425 in the direction parallel to the shock normal must be large enough to allow the cr to escape upstream : @xmath426 . provided the boundary conditions are periodic in the directions perpendicular to the shock normal they can be much smaller than @xmath425 but they must be able to accommodate a cr larmor radius : @xmath427 .
the number of computational operations is proportional to @xmath428 , which from definitions and equations presented above , is of order @xmath429 as discussed above the nrh instability grows strongly in a wavenumber range between @xmath430 and @xmath431 .
saturation due to magnetic tension occurs on scalelengths comparable to the cr larmor radius @xmath432 when @xmath433 as discussed in section 10 .
the ratio of the saturated field @xmath434 to the initial field @xmath56 is @xmath435 where @xmath431 and @xmath12 are defined in the initial field @xmath56 .
the equation for @xmath431 can be found in section 2 , giving @xmath436 where the acceleration efficiency @xmath43 is also defined in section 2 .
the alfven mach number @xmath437 must be large to allow significant amplification of the magnetic field ( @xmath438 ) . from equation ( c1 )
large @xmath437 imposes a heavy demand on computational resources , but from equation ( c2 ) the cost can be minimised by making the shock velocity @xmath1 a large fraction of the speed of light .
we initiate the simulation with an unrealistically large magnetic field of @xmath439 to reduce the alfven mach number from its typical value of @xmath440 for shocks in young snr . | galactic cosmic ray ( cr ) acceleration to the knee in the spectrum at a few pev is only possible if the magnetic field ahead of a supernova remnant ( snr ) shock is strongly amplified by cr escaping the snr . a model formulated in terms of the electric charge carried by escaping cr predicts the maximum cr energy and the energy spectrum of cr released into the surrounding medium .
we find that historical snr such as cas a , tycho and kepler may be expanding too slowly to accelerate cr to the knee at the present time .
[ firstpage ] cosmic rays , acceleration of particles , shock waves , magnetic field , ism : supernova remnants |
physically speaking , the goal of this paper is to study the exotic statistics of loop - like defects in a 4-dimensional topological field theory called @xmath0 theory .
we call these entities ` closed strings ' for short , though they behave differently from the closed strings familiar in string theory : the relevant lagrangian is different . in fact
, we postpone the study of their dynamics to another paper @xcite .
the considerations of this paper are purely topological , and accessible we hope to mathematicians with only a passing interest in physics . mathematically speaking ,
the point of this paper is to study some representations of a higher - dimensional analogue of the braid group : the ` loop braid group ' .
just as the braid group describes the topology of points moving in the plane , the loop braid group describes the topology of circles moving in @xmath2 . in the body of this paper , we describe this group and certain representations of it coming from the moduli space of flat bundles on @xmath2 with these circles removed .
but since everything we do has a more familiar analogue one dimension down , let us start by recalling that .
the behavior of a collection of identical particles when they are exchanged goes by the name of ` statistics ' .
traditionally , statistics was described using representations of the symmetric group .
however , it is well known that in 3d spacetime , ` exotic ' statistics are possible , in which the process of exchanging identical particles is described by a representation of the braid group . for example , exchanging two ` abelian anyons ' multiplies their wavefunction by a phase , which need not be @xmath3 as it is for bosons , nor @xmath4 as for fermions .
this possibility has been investigated in experiments on the fractional quantum hall effect @xcite .
now researchers have begun the search for ` nonabelian anyons ' , whose statistics are described by more complicated representations of the braid group @xcite .
plans are already afoot to use these in quantum computers @xcite .
exotic statistics also arise naturally in the context of 3d quantum gravity . as we ` turn on gravity ' , letting newton s gravitational constant @xmath5 become nonzero , ordinary quantum field theory on 3d minkowski spacetime deforms into a theory where the poincar group goes over to a quantum group called the @xmath5-poincar group .
moreover , if we begin with a field theory of bosons , their statistics become exotic as we turn on gravity . for a thorough treatment of these fascinating phenomena , see the papers by freidel and collaborators @xcite , the paper by krasnov @xcite , and the many references therein .
in fact , the reason for exotic statistics in 3d quantum gravity is very simple . in 3d
spacetime , einstein s equations say that spacetime is _ flat _ except in regions where matter is present .
a point particle at rest bends the nearby space into a cone .
this cone is flat everywhere except at its tip , where there is a deficit angle proportional to the particle s mass .
if we parallel transport a vector around the particle , it gets rotated by this angle @xmath6 : @xmath7 more generally , if we have @xmath8 particles , space will be flat except for conical singularities at @xmath8 points . if we exchange these particles by moving them around the plane , they trace out a loop in the space of @xmath8-point subsets of the plane .
their energy - momenta will change in a way that depends on this loop but only on the _ homotopy class _ of this loop , because they are being parallel transported with respect to a flat connection .
a homotopy class of such loops is just an @xmath8-strand braid : @xmath9 so , the group @xmath10 of @xmath8-strand braids acts on the hilbert space of states for @xmath8 identical particles .
in fact , this result holds classically as well : we get an action of @xmath10 on the configuration space for @xmath8 identical particles .
the above argument uses the fact that 3-dimensional gravity ( with vanishing cosmological constant ) can be described by @xmath0 theory with the lorentz group @xmath11 as gauge group . to understand this paper
, the reader only needs to know one thing about @xmath0 theory : _ it involves a flat connection on space_. for completeness , however , we recall that @xmath0 theory in @xmath8-dimensional spacetime with gauge group @xmath1 involves two fields : a connection @xmath12 and a @xmath13-valued @xmath14-form @xmath15 . in the absence of matter ,
the lagrangian is simply @xmath16 here @xmath5 plays the role of newton s constant in the case of 3d gravity , and @xmath17 is the curvature of @xmath12 .
the resulting equations of motion : @xmath18 = 0,\ ] ] imply that the connection @xmath12 is flat . in 3d @xmath0 theory , point particles can be included by considering spacetimes with curves removed : we think of these as the particles worldlines . away from these worldlines
the above equations still hold , while along the worldlines @xmath12 becomes singular .
the holonomy around a loop circling a worldline gives an element of the group @xmath1 .
a collection of @xmath8 particles in the plane thus gives rise to an @xmath8-tuple of elements of @xmath1 .
for simplicity , consider the case @xmath19 . as we exchange two particles by rotating them around each other counterclockwise ,
they trace out this braid : @xmath20 as we recall in section [ 3d ] , this operation acts as the following map on @xmath21 : @xmath22 applying this map twice does _ not _ give the identity , so we do not obtain an action of the symmetric group on @xmath21 , but only an action of the braid group .
in other words , the particles have exotic statistics ! in the case of 3d gravity ,
the singularity of the connection along a particle s worldline reflects the fact that the particle s mass creates a conical singularity in the metric .
the holonomy around the worldline , an element of @xmath23 , describes the particle s _ energy - momentum_. this may seem odd , since we are used to thinking of energy - momentum as a vector in minkowski spacetime .
however , in 3 dimensions minkowski spacetime is naturally isomorphic to the lie algebra @xmath24 , and we can reinterpret lie algebra elements as group elements via the map : @xmath25 so , we can encode the energy - momentum @xmath26 of a particle in the holonomy @xmath27 resulting from parallel transport around this particle s worldline . thanks to the factor of @xmath5 here , the group @xmath11 effectively ` flattens out ' to @xmath24 in the @xmath28 limit .
for example , multiplication in the group reduces to addition in the lie algebra plus small corrections : @xmath29 + \cdots ) \ ] ] this implies that in terms of @xmath24-valued energy - momenta , the braiding in equation ( [ braiding ] ) is given by @xmath30 + \cdots \ , , \ , p_1 ) \ ] ] so , the exotic statistics reduce to ordinary bosonic statistics in the limit where newton s constant goes to zero .
they also reduce to bosonic statistics in the limit where the particles are at rest relative to each other , since then @xmath31 and @xmath32 become proportional and their commutator vanishes .
the corrections to the usual law for addition of energy - momenta implicit in equation ( [ doubly_special ] ) are interesting in themselves . like the exotic statistics , these corrections become negligible in the limit @xmath28 . under the name of
` doubly special relativity ' , modified laws for adding energy - momentum have already been studied by many authors .
the paper by freidel , kowalski - glikman and smolin @xcite gives a good account of doubly special relativity in the context of 3d quantum gravity ; their paper also explains more of the history of this subject .
1em besides exotic statistics and corrections to the usual rule for adding energy - momenta , there is yet another surprising consequence of the switch from vector - valued to group - valued energy - momentum as we turn on gravity in 3d physics
. the classification of elementary particles changes ! in ordinary quantum field theory on minkowski spacetime , the lorentz group acts on the space of possible energy - momenta , and the orbits of this action correspond to different types of spin - zero particles . when spacetime is 3-dimensional , the space of energy - momenta is @xmath24 , and the orbits look like this : @xmath33 if we write the energy - momentum as @xmath34 and let @xmath35 , we have six families of orbits , corresponding to six types of spin - zero particles : 1
. positive - energy tardyons of mass @xmath36 : @xmath37 , 2 .
negative - energy tardyons of mass @xmath36 : @xmath38 , 3 .
positive - energy luxons : @xmath39 , 4 .
negative - energy luxons : @xmath40 , 5 .
tachyons of mass @xmath41 for @xmath36 : @xmath42 , 6 .
particles of vanishing energy - momentum : @xmath43 . given any orbit @xmath44 ,
the hilbert space for a single particle of type @xmath45 is just @xmath46 .
the same philosophy applies when we turn on gravity , but now the space of energy - momenta is not the lie algebra @xmath24 but the lorentz group itself .
this acts on itself by conjugation , and the orbits are conjugacy classes .
types of spin - zero particles now correspond to conjugacy classes in the lorentz group . near the identity
these conjugacy classes look just like orbits in the lie algebra , so the classification of particles reduces to the above one in the limit of small energy - momenta .
however , there are important differences , which show up for large energy - momenta .
most notably , under the map @xmath47 the lie algebra element @xmath48 is mapped to a rotation by the angle @xmath49 in the @xmath50 plane .
so , the holonomy around a stationary particle of energy @xmath15 is a rotation by the angle @xmath49 .
this rotation does not change when we add @xmath51 to the particle s energy . up to factors of order unity ,
this quantity @xmath51 is just the _ planck energy_. if we call it the planck energy , then masses in 3d quantum gravity are defined only modulo the planck mass .
this ` periodicity of mass ' affects the classification of tardyons that is , the most familiar sort of particles , those with timelike energy - momentum . instead of positive - energy tardyons of arbitrary mass @xmath36 and negative - energy tardyons of arbitrary mass @xmath36
, we just have tardyons of arbitrary mass @xmath52 . more generally , for any lie group @xmath1
, the various allowed types of spin - zero particles in 3d @xmath0 theory with gauge group @xmath1 correspond to conjugacy classes @xmath53 .
any conjugacy class is closed under the operations @xmath54 and these operations satisfy equations making @xmath45 into an algebraic structure called a ` quandle ' @xcite , whose definition we recall in section [ quandle ] .
the hilbert space for a single particle of type @xmath45 is just @xmath46 , defined using a measure on @xmath45 that is invariant under these operations . in an easy generalization of 3d @xmath0 theory , we can study the exotic statistics of ` particles of type @xmath45 ' for any quandle @xmath45 equipped with an invariant measure .
this takes advantage of the well - known relation between quandles and the braid group @xcite
. it would be wonderful to generalize all the above results to 4d gravity , but for now all we can handle is a simpler theory : 4d @xmath0 theory .
this may eventually be relevant to gravity , since one can describe general relativity in 4 dimensions either as the result of _ constraining _ 4d @xmath0 theory with a certain gauge group , or _ perturbing around _
4d @xmath0 theory with some other gauge group .
the first approach goes back to plebanski @xcite , and it underlies a great deal of work on spin foam models of quantum gravity @xcite , especially the barrett crane model .
the second approach goes back to macdowell and mansouri @xcite , and has recently been explored by freidel and starodubtsev @xcite .
however , we do not dwell on these possible applications here .
they only focus our attention towards certain choices of gauge group : @xmath55 our idea is simply to increase the dimension of everything in the previous section by 1 .
thus , we consider @xmath0 theory on a 4-dimensional spacetime with the worldsheets of several ` closed strings ' removed .
we focus on the case where the manifold representing space is @xmath56 , where @xmath57 is an ` @xmath8-component unlink ' : a collection of @xmath8 unknotted unlinked circles .
a flat connection on @xmath58 gives us a group element for each circle , namely the holonomy of some standard loop going around this circle : @xmath59
so , just as before , we obtain @xmath8-tuples of elements of @xmath1 .
moreover , any way to exchange the circles in @xmath57 gives a map from @xmath60 to itself .
it is often said that exotic statistics are only possible when space has dimension 2 or less .
however , this folklore only applies to point particles . as pointed out by balanchandran and others
@xcite , exotic statistics are possible for closed strings in 3-dimensional space , since there are topologically nontrivial ways to exchange unknotted unlinked circles in @xmath2 .
the statistics of such theories are governed not by the braid group @xmath10 , but by a larger group : the ` loop braid group ' @xmath61 .
using recent work of lin @xcite , we show that this group is isomorphic to the ` braid permutation group ' of fenn , rimnyi and rourke @xcite .
this is an apt name , because @xmath61 has a presentation with generators @xmath62 that describe two strings trading places without passing through each other , just as if they were point particles : @xmath63 but also generators @xmath64 that describe one string passing through another : @xmath65 so , this group is a kind of ` hybrid ' of the symmetric group and the braid group .
indeed , the elements @xmath62 generate a copy of the symmetric group @xmath66 in @xmath61 , while the elements @xmath64 generate a copy of the braid group @xmath10 . in a one - dimensional unitary representation of the loop braid group , the permutation generators @xmath62 all act as @xmath67 , while the braid generators @xmath64 all act as an arbitrary phase @xmath68 .
we could call particles that transform in this way ` abelian bose - anyons ' and ` abelian fermi - anyons ' , respectively .
they act like bosons or fermions when we switch them using the generators @xmath62 , but like abelian anyons when we switch them using the generators @xmath64 .
@xmath0 theory gives us more interesting unitary representations of the loop braid group : whenever the group @xmath1 is unimodular , we obtain a unitary representation of @xmath61 on @xmath69 .
all the groups listed above are unimodular , so we get an interesting variety of exotic statistics for closed strings in 4d @xmath0 theory
. we can also restrict attention to a specific conjugacy class @xmath53 and get a unitary representation of the loop braid group on @xmath70 , as long as @xmath45 is equipped with a measure invariant under conjugation .
as already mentioned , in the case of 3d gravity a choice of conjugacy class in @xmath23 essentially amounts to choosing a specific _ mass _ for our point particles , which is a very natural thing to do . in the case of 4d @xmath0 theory with @xmath71 , choosing a conjugacy class essentially amounts to choosing a specific _ mass density _ for our closed strings .
the loop braid group @xmath61 consists of all ways a collection of oriented , unknotted , unlinked circles can move around in @xmath2 and come back to their original positions , perhaps trading places .
more precisely , it consists of ` isotopy classes ' of such motions .
this group thus plays the same role in describing the interchange of closed strings in @xmath2 that the symmetric group @xmath66 plays for point particles in @xmath2 , and the braid group plays for point particles in @xmath72 . in this section
we use the work of lin @xcite to obtain two presentations of the loop braid group .
first , however , we explain the sense in which the loop braid group , the symmetric group and the braid group are all examples of ` motion groups ' .
the general idea of a ` motion group ' goes back at least to dahm s 1962 thesis @xcite , which unfortunately was never published . in the 1970 s and 80 s ,
some papers by wattenberg @xcite and goldsmith @xcite clarified and expanded on dahm s work .
more recently , mccool @xcite and rubinsztein @xcite have studied the motion group for unknotted and unlinked circles in @xmath2 .
surya has also given a description of the loop braid group as an iterated semidirect product @xcite .
much of this work considers the motion of unoriented circles .
since we use oriented circles , we obtain a smaller motion group , which lacks the ` circle - flipping ' operations that reverse orientations . quite generally ,
suppose that @xmath73 is a smooth oriented manifold and @xmath74 is a smooth oriented submanifold .
let @xmath75 be the group of orientation - preserving diffeomorphisms of @xmath73 .
let @xmath76 be the subgroup of @xmath75 maps that restrict to give orientation - preserving diffeomorphisms of @xmath57 .
we define a * motion * of @xmath57 in @xmath73 to be a smooth map @xmath77 \times s \to s$ ] , which we write as @xmath78 ( @xmath79 $ ] ) , with the following properties : * for all @xmath80 , @xmath81 lies in @xmath75 ; * for all @xmath80 sufficiently close to @xmath82 , @xmath81 is the identity ; * for all @xmath80 sufficiently close to @xmath3 , @xmath81 is independent of @xmath80 and lies in @xmath76 .
intuitively , a motion is a way of moving @xmath57 through @xmath73 so that it comes back to itself not pointwise , but as a set at @xmath83 .
this suggests that one can ` multiply ' motions by doing one after the other , and indeed this is true .
given motions @xmath84 and @xmath85 , one can define a motion @xmath86 called their * product * as follows : @xmath87 given a motion @xmath84 we can also define a motion called its * reverse * , denoted @xmath88 , by : @xmath89 we say two motions @xmath84 and @xmath85 are * equivalent * if @xmath90 is smoothly homotopic , as a path in @xmath75 with fixed endpoints , to a path that lies entirely in @xmath76 .
one can check that this is indeed an equivalence relation and that the operations of product and reverse make equivalence classes of motions into a group .
this is called the * motion group * @xmath91 .
next we turn to examples : * when @xmath92 is a collection of @xmath8 points and @xmath93 , @xmath94 is the symmetric group @xmath66 . * when @xmath95 is a collection of @xmath8 points , @xmath96 is the braid group @xmath10 .
* when @xmath97 is a collection of @xmath8 unknotted and unlinked oriented circles , we call @xmath98 the * loop braid group * @xmath61 .
we shall use the work of lin @xcite to give two presentations of @xmath61 .
first note that there is a homomorphism @xmath99 which simply forgets the details of the braiding , remembering only how the circles get permuted in the process .
the image of @xmath26 is all of @xmath66 .
we call the kernel of @xmath26 the * pure loop braid group * @xmath100 .
suppose , just to be specific , that @xmath101 where @xmath102 are disjoint unit circles in the @xmath50 plane , lined up from left to right with their centers on the @xmath103 axis .
lin proves that @xmath100 has a presentation with generators @xmath104 for @xmath105 with @xmath106 .
the generator @xmath104 describes a motion in which the @xmath107th circle floats up and over the @xmath108th circle , shrinks slightly and passes down through the @xmath108th circle , expands to its original size , and then moves straight back to its starting position . we draw this as follows :
@xmath109 where for purely artistic reasons we let the @xmath108th circle move a bit to the left in the process .
here we are using a drawing style adapted from carter and saito s work on surfaces in 4 dimensions @xcite .
crossings in a braid or knot are usually drawn with an artificial ` break ' in one of the strands to indicate that it lies under the other : @xmath110 similarly , carter and saito draw 3d projections of knotted surfaces in 4 dimensions , indicating by a broken surface which one passes ` under ' the other in the suppressed fourth dimension . in our context , we take this suppressed dimension to be one of the spatial dimensions , in order to make room for _ time _ , which we decree to flow downward in all our diagrams .
the broken surfaces in @xmath104 indicate whether one circle is above or below the other in the suppressed spatial dimension , so that the following diagram and ` movie ' illustrate the same process : @xmath111{ltrmovie.ps } } ; \endxy\ ] ] the inverse of @xmath104 is of course obtained by running the movie backwards , which in diagrammatic notation becomes : @xmath112 one advantage of this drawing style is that it immediately suggests reidemeister like moves for loop braids , such as this : @xmath113 we shall study the loop braid group algebraically , relying on such diagrams for our intuition . given lin s presentation of @xmath100 , we can obtain a presentation of @xmath61 using the short exact sequence @xmath114 first , note that there is a homomorphism @xmath115 which takes a given permutation to what lin calls a ` permutation path ' in the motion group : a loop braid in which circles trade places without any circle passing through another in a topologically nontrivial way .
for example , we can have them trade places while remaining on the @xmath50 plane .
this map @xmath108 is well - defined since all such permutation paths are homotopic .
moreover , the composite @xmath116 is the identity homomorphism on @xmath66 , so @xmath108 is a splitting of the short exact sequence above .
since @xmath108 is one - to - one , we may identify elements of @xmath66 with their images in @xmath61 . since @xmath100 is a normal subgroup , elements of @xmath66 act on @xmath100 via conjugation .
this allows us to define the semidirect product @xmath117 , and thanks to our split exact sequence , we get an isomorphism @xmath118 with inverse @xmath119 writing the loop braid group as a semidirect product in this way , we easily obtain a presentation for it : [ presentation1 ] the loop braid group @xmath61 has a presentation with generators @xmath62 for @xmath120 and @xmath104 for @xmath121 with @xmath106 , together with the following relations : the relations for the standard generators @xmath62 of @xmath66 : @xmath122 lin s relations for the generators @xmath104 of @xmath100 : @xmath123 relations expressing the action of @xmath66 on @xmath100 : @xmath124 * proof . * since the presentation ( @xmath125 ) of @xmath66 is well - known , and lin @xcite proved that @xmath100 has the presentation ( @xmath126 ) , to present their semidirect product @xmath61 it suffices to add relations that express the result of conjugating any of lin s generators @xmath104 by the symmetric group generators @xmath127 .
for @xmath120 we have : @xmath128 for @xmath129 and @xmath130 all distinct , we have : @xmath131 for @xmath132 and @xmath133 distinct , we have : @xmath134 and using a similar picture we see that for @xmath135 and @xmath108 distinct , @xmath136 .
the reader may notice that we have not included all possible conjugations of generators of @xmath100 by generators of @xmath66we would naively expect two additional such classes , yielding two more relations : @xmath137 but these follow , respectively , from ( [ rel9 ] ) and ( [ rel11 ] ) combined with ( [ rel3 ] ) .
so , we have precisely the relations in part @xmath138 , as desired . from this presentation of the loop braid group
we now derive a presentation with fewer generators .
we keep all the generators @xmath62 , but replace the @xmath104 with new generators defined as follows : @xmath139 for @xmath120 .
we can draw these as follows : @xmath140 where we twist the picture a bit in the second step . to see that the generators @xmath62 and @xmath64 indeed give a new presentation , note that we can express the old generators @xmath104 in terms of these new ones as follows .
first , repeatedly applying ( [ rel9 ] ) we obtain : @xmath141 if instead of ( [ rel9 ] ) we use its equivalent form ( [ rel9prime ] ) , we obtain : @xmath142 rewriting these in terms of the new generators @xmath64 , and in the second case using relation ( [ rel7 ] ) , we obtain a way to write @xmath104 in terms of the new generators : @xmath143 sometimes it is more convenient to use an alternate formula , obtained by applying ( [ rel11 ] ) , its equivalent form ( [ rel11prime ] ) , and ( [ rel7 ] ) again : @xmath144 what these formulas say is that when @xmath145 we can construct the loop braid @xmath104 by permuting either the @xmath107th circle or the @xmath108th until they are adjacent , braiding one through the other , and then permuting the circles back to where they started .
the nice thing about using @xmath62 and @xmath64 as generators of the loop braid group is that @xmath62 describes how two neighboring circles can trade places by going around each other : @xmath146 while @xmath64 describes how two neighboring circles can trade places with the right one passing over and then down through the left one : @xmath147 as a result , the generators @xmath62 generate a subgroup of @xmath61 isomorphic to the symmetric group @xmath66 , while the @xmath64 generate a subgroup isomorphic to the braid group @xmath10 .
there are also ` mixed relations ' involving generators of both kinds : [ presentation2 ] the loop braid group @xmath61 has a presentation with generators @xmath62 and @xmath64 for @xmath120 together with the following relations : * relations for the standard generators @xmath62 of @xmath66 : @xmath148 * relations for the standard generators @xmath64 of @xmath10 : @xmath149 * the following * mixed relations * : @xmath150 * proof . * the proof is somewhat lengthy , so we defer it to the appendix .
it is , however , simple to convince oneself using pictures that the given relations express topologically allowed moves for loop braids .
perhaps the least obvious of these is ( [ rel9 ] ) , for which we supply a visual proof below : @xmath151 if we omit relations ( [ rel9 ] ) we obtain the ` virtual braid group ' @xmath152 of vershinin @xcite .
this plays a role in virtual knot theory analogous to that of the usual braid group in ordinary knot theory . if we include these relations , which say : @xmath153 then we obtain precisely the ` braid permutation group ' @xmath154 of fenn , rimnyi and rourke @xcite .
so , the loop braid group is isomorphic to the braid permutation group .
the isomorphism @xmath155 yields a simplified diagrammatic way of working with loop braids , which is in fact the method used by fenn , rimnyi and rourke in their original paper on @xmath154 . in the theory of ` welded braids ' , the generators @xmath64 in @xmath154 correspond to the kind of crossings found in ordinary braids : @xmath156 , while the @xmath62 describe ` welded crossings ' , drawn like this : @xmath157 .
these crossings are called ` welded ' because one imagines that the two strands have been ` welded down ' at the crossing .
the point is that elements of the abstract group presented in theorem [ presentation2 ] can be represented either as loop braid diagrams or as welded braid diagrams , as follows : @xmath158 for the _ pure _ loop braid group @xmath100 , the above correspondence implies the following welded braid pictures of the generators @xmath159 and their inverses : @xmath160 the other generators @xmath104 can be obtained from these by conjugation , using ( [ generatorcorrespondence1 ] ) or ( [ generatorcorrespondence2 ] ) .
for example : @xmath161 diagrammatic calcuations with welded braids and hence with loop braids can be carried out by using the usual reidemeister moves for real crossings , along with ` welded reidemeister moves ' : @xmath162 which are of course simply graphical restatements of the relations in @xmath163 and @xmath164 .
the nonexistence of the following move : @xmath165 is the rationale for the term ` welded braid'we are not allowed to pass a strand under the weld .
it is easy from the presentation in theorem [ presentation2 ] to work out the 1-dimensional unitary representations of the loop braid group .
if @xmath166 is such a representation , we must have @xmath167 and @xmath168 for all @xmath169 , where @xmath68 is a fixed phase .
we call the representations with @xmath170 * bose - anyons * , and the representations with @xmath171 * fermi - anyons*. these have been studied in physics at least since the work of balachandran @xcite , and recently niemi has shown how they arise in the dynamics of vortices in a quantum fluid @xcite . in section [ 4d ]
we describe more interesting unitary representations of the loop braid group , using some technology which we now develop . in related
work , szabo @xcite has obtained a different class of representations using @xmath0 theory with abelian gauge group .
surya @xcite has also studied representations of the loop braid group .
in this section we recall dahm s @xcite action of the motion group @xmath91 on the fundamental group of @xmath172 and describe how this gives a unitary representation of the motion group on a certain hilbert space of states for @xmath0 theory on @xmath172 .
we consider @xmath0 theory in @xmath8-dimensional spacetime .
so , we take ` space ' to be of the form @xmath173 , where @xmath73 is an oriented manifold of dimension @xmath174 , and @xmath175 is an oriented submanifold .
we let @xmath1 be a lie group and let @xmath176 be a principal @xmath1-bundle .
the ` naive configuration space ' of @xmath0 theory is @xmath177 , where @xmath178 is the space of flat connections on @xmath179 and @xmath180 is the group of gauge transformations . by ` naive ' we mean that we are ignoring boundary conditions ; there are no boundary conditions to worry about when @xmath181 is compact , but we shall mainly be interested in two examples where it is not : 1 .
@xmath181 is @xmath72 with a finite set of points removed ( describing _ point particles _ ) : @xmath182 2 .
@xmath181 is @xmath2 with a finite set of unlinked unknotted circles removed ( describing what one might call _ closed strings _ ) : @xmath183 a rigorous study of @xmath0 theory may require that we impose boundary conditions at @xmath57 .
we ignore this issue now , leaving it for future research .
the space @xmath177 is a bit difficult to handle .
it is often more convenient to start by fixing a basepoint @xmath184 and working with @xmath185 , where @xmath186 the group @xmath187 acts on @xmath185 in a natural way .
this lets us form @xmath177 as the quotient of the bigger space @xmath185 by this action of @xmath1 .
the advantage of the space @xmath185 is that any point @xmath188 $ ] in this space gives a homomorphism @xmath189 ) { \colon}\pi_1(x ) \to g\ ] ] which sends any homotopy class of loops @xmath190 $ ] to the holonomy of @xmath12 around @xmath191 .
this gives a map @xmath192 which is known to be one - to - one .
note that @xmath1 acts on @xmath193 by conjugation : @xmath194 where @xmath195 is any homomorphism .
moreover , the map @xmath196 is compatible with this group action : @xmath197 ) = g \ , \hol([a ] ) .\ ] ] so far we have fixed a principal @xmath1-bundle @xmath179 .
but , in gauge theory it is often better to treat this bundle as variable
part of the physical field along with the connection @xmath12 .
for example , path integrals in quantum chromodynamics involve a sum over bundles , which represent instantons .
the mathematical advantage of treating @xmath179 as variable is that all points of @xmath193 are in the image of @xmath196 if we allow ourselves to vary @xmath179 @xcite .
a point in this space represents a ` @xmath1-bundle with flat connection over @xmath181 , mod gauge transformations that equal the identity at the basepoint ' .
modding out by the rest of the gauge transformations we get a space known as the ` moduli space of flat bundles ' , @xmath198 .
this is the naive configuration space for @xmath0 theory where we treat the bundle @xmath179 as variable . applying schrdinger quantization to this configuration space
, we obtain the ( naive ) hilbert space for @xmath0 theory : @xmath199 of course , defining this @xmath200 space requires that we choose a measure on the moduli space of flat bundles .
alternatively , we can try to form a hilbert space @xmath201 on which @xmath1 acts as follows : @xmath202 again , this requires choosing a measure on @xmath193 .
moreover , @xmath1 will only have a unitary representation on @xmath203 if this measure is @xmath1-invariant .
in sections [ 3d ] and [ 4d ] we will show that for the two examples above , there is a ` natural ' choice of @xmath1-invariant measure on @xmath193 . in both these examples the motion group @xmath91 acts on @xmath204 and thus on @xmath193 . by saying a measure on @xmath193 is ` natural ' , we simply mean that it is preserved by this action .
using such a natural measure to define the hilbert space @xmath205 , we obtain a unitary representation of the motion group on this hilbert space .
this representation describes the statistics of point particles or closed strings in @xmath0 theory .
as we have seen , in the first example the motion group is the braid group @xmath10 , while in the 4d case it is the loop braid group @xmath61 .
so , we obtain ` exotic statistics ' in both cases . this fact is somewhat familiar in 3 dimensions , but less so in 4 dimensions .
so , in the following sections we first review the 3d case , and then move on to the 4d case after a brief digression on ` quandle field theory ' . before doing this , however , let us see how the motion group acts on @xmath204 .
the idea goes back to dahm s original work on the motion group @xcite , and it has been nicely explained by goldsmith @xcite . the idea is simple : elements of the motion group @xmath91 give equivalence classes of diffeomorphisms of @xmath173 , and these act on homotopy classes of loops in @xmath181 .
the only problem is that the fundamental group is defined using _ based _ loops , and the diffeomorphisms used in the definition of the motion group need not preserve the basepoint in @xmath181 .
luckily , wattenberg @xcite has shown that we can use compactly supported diffeomorphisms in the definition of the motion group without changing this group . in the examples
above , we can assume without loss of generality that these diffeomorphisms are supported in a fixed large ball containing @xmath57 .
so , if we choose a basepoint @xmath206 that is sufficiently far from @xmath57 , we can assume this basepoint is preserved by all the diffeomorphisms in the definition of the motion group .
this makes it easy to check that @xmath91 acts as automorphisms of @xmath204 .
[ 3d ] now let us apply the general ideas of the previous section to the case of a plane with @xmath8 punctures : @xmath207 if we interpret these punctures as ` particles ' , we shall see that a state of 3d @xmath0 theory on this space describes a collection of identical point particles with exotic statistics governed by the braid group .
the fundamental group of @xmath181 is the free group on @xmath8 generators , so we have @xmath208 the @xmath8 group elements here are nothing but the holonomies of a flat connection around based loops going clockwise around the particles : @xmath209 having described particles as punctures in this theory , let us now consider what sort of statistics such particles obey .
the previous section shows that the interchange of identical particles is described by an action of the @xmath8-strand braid group @xmath10 on @xmath60 , but we would like to work it out explicitly . for simplicity , consider the case @xmath19 and consider what happens when the two particles switch places . as remarked earlier , there are infinitely many topologically distinct ways for the particles to move around each other , but they are all powers of the braid group generator @xmath210 : @xmath211 if the holonomies around the two particles are @xmath212 and @xmath213 : @xmath214 switching them via @xmath210 induces a diffeomorphism of the plane which deforms the loops around which the holonomies are taken : @xmath215 to see how the system changes in this process , compare the final frame in this ` movie ' to the first frame . given that @xmath216 describes the holonomies initially , a slight deformation of the loops in the final frame : @xmath217 makes it clear that the corresponding holonomies around _
these _ loops in the final configuration : @xmath218 is @xmath219 . thus the effect of switching the two particles via @xmath210 is to send @xmath220 to @xmath221 . we can work out the action of @xmath222 in the same way , or simply derive it algebraically from the fact that it must undo the effect of @xmath210 .
the easiest way to remember the results is with this picture : @xmath223 more generally , we have a right action of the braid group @xmath10 on @xmath60 given as follows : @xmath224 as mentioned in the previous section , we also have a left action of @xmath1 on @xmath60 via gauge transformations at the basepoint @xmath225 .
this works as follows : @xmath226 we would like a measure on @xmath60 that is invariant under both these group actions , so that the braid group and gauge transformations act as unitary operators on @xmath69 .
such a measure exists whenever @xmath1 is * unimodular * , meaning that its left - invariant haar measure is also right - invariant .
a lie group is automatically unimodular if it is compact , or abelian , or semisimple .
in particular , the groups @xmath227 are all unimodular .
since these groups act on minkowski spacetime in a way that preserves its lebesgue measure , the poincar groups @xmath228 are also unimodular . also , the identity component of a unimodular group is unimodular , as is any covering space of a unimodular group . from this
we see that the 3d lorentz group @xmath11 is unimodular , as are its identity component @xmath229 and the double cover of its identity component , namely @xmath230 .
all these are reasonable choices of gauge group when treating 3-dimensional or more properly , ( 2 + 1)-dimensional lorentzian gravity as a @xmath0 theory . given a unimodular lie group , haar measure is typically not the only measure invariant under conjugation : we can multiply haar measure by any function that only depends on the conjugacy class . as an extreme example , we can even try to multiply haar measure by a ` delta function ' supported on one conjugacy class .
more precisely , we can look for a conjugation - invariant measure supported on a single conjugacy class of @xmath1 . in this case
we might as well be working not with @xmath1 but with just the conjugacy class .
it turns out that in the case of 3d quantum gravity , this amounts to studying identical particles _ of a specified mass_. this leads us to our next subject : quandle field theory .
in the previous section we considered @xmath0 theory in 3 dimensions , and were led to a natural action of the braid group @xmath10 on the space @xmath60 for any group @xmath1 .
notice that we did not actually need the multiplication in @xmath1 to define this action ; we only needed the operation of _
conjugation_. this suggests that we can work more generally , replacing the group @xmath1 by some algebraic structure that captures the properties of conjugation .
such a thing is called a ` quandle ' .
more precisely , a * quandle * is a nonempty set
@xmath45 equipped with two binary operations @xmath231 and @xmath232 called * left * and * right conjugation * , which satisfy : ( i ) : : left idempotence : @xmath233 ( i@xmath234 ) : : right idempotence : @xmath235 ( ii ) : : left inverse law : @xmath236 ( ii@xmath234 ) : : right inverse law : @xmath237 ( iii ) : : left distributive law : @xmath238 ( iii@xmath234 ) : : right distributive law : @xmath239 for all @xmath240 . in general , the operations of left and right conjugation in a quandle are neither associative nor commutative .
quandles were first introduced as a source of knot invariants by david joyce @xcite in @xmath241 .
many examples of quandles can be found in the work of fenn and rourke @xcite and other authors @xcite . for us ,
the most important examples come from taking a group @xmath1 , letting @xmath45 be any union of conjugacy classes of @xmath1 , and making @xmath45 into a quandle with @xmath242 we are especially interested in the case where @xmath45 is either the whole group @xmath1 or a single conjugacy class .
we can do some of the same things with quandles as with groups . for example
, we can define a * topological quandle * to be a topological space that is also a quandle in such a way that the quandle operations @xmath243 and @xmath244 are continuous @xcite .
if @xmath1 is a lie group and @xmath53 is a conjugacy class , @xmath45 becomes a topological quandle with the induced topology .
given a topological quandle @xmath45 , we define an * invariant measure * on @xmath45 to be a borel measure that is invariant under left conjugation by any element of @xmath45or equivalently , invariant under right conjugation by any element of @xmath45 .
this implies that @xmath245 for any @xmath246 and any integrable function @xmath84 on @xmath45 .
as noted earlier , invariant measures on quandles are far from unique in general .
in particular , we may multiply an invariant measure on a lie group by any class function and obtain a new invariant measure . in the previous section
, we saw that the @xmath8-strand braid group @xmath10 acts on @xmath60 for any group @xmath1 .
but , since our argument relied only on properties of conjugation , it works just as well for a quandle .
the idea is that we can braid two elements of a quandle past each other using left conjugation : @xmath247 the inverse braiding uses right conjugation : @xmath248 it is well known that with these rules , the braid group relations follow from the quandle axioms .
so , generalizing our result from the previous section , we easily obtain : [ 3dtheorem ] suppose @xmath45 is a topological quandle equipped with an invariant measure . then there is a unitary representation @xmath249 of the braid group @xmath10 on @xmath70 given by @xmath250 for all @xmath251 , where @xmath10 has a right action on @xmath252 given by : @xmath253 there is also a unitary operator @xmath254 on @xmath70 for each element @xmath255 , given by @xmath256 the operators @xmath254 represent gauge transformations when @xmath45 is a group , so we can think of them as representing some sort of ` gauge transformation ' even when @xmath45 is a quandle .
of course , if @xmath45 is a conjugacy class in a group @xmath1 , there will be gauge transformations even for elements of @xmath1 that do not lie in @xmath45 .
it is instructive to work out the details in the case of ( 2 + 1)-dimensional quantum gravity .
this theory can be viewed as a @xmath0 theory with @xmath1 being the connected lorentz group @xmath229 , or perhaps better , its double cover @xmath230 . in either case
we shall see that different conjugacy classes @xmath45 describe different types of spinless particles .
the hilbert space for @xmath8 particles of this type is @xmath70 , and theorem [ 3dtheorem ] describes the exotic statistics and gauge invariance of this @xmath8-particle system . in quantum field theory without gravity on 3d minkowski spacetime , we can describe the energy - momentum of a particle by an element @xmath257 : @xmath258 note that @xmath259 the adjoint action of @xmath230 on its lie algebra : @xmath260 preserves the determinant of @xmath26 .
so , the adjoint action gives an action of @xmath230 as lorentz transformations on the space of energy - momenta .
as explained in the introduction , an orbit of this action is just a type of spin - zero particle .
when we turn on gravity , we must describe energy - momenta not by elements of the lie algebra @xmath261 but by elements of the group @xmath230 .
particle types are then described not by adjoint orbits but by conjugacy classes @xmath262 . however ,
this new description is compatible with the old one , at least for energy - momenta that are small compared to the planck energy @xmath51 .
the reason is that we can identify group elements near the identity with lie algebra elements via the map @xmath263 this maps any adjoint orbit of @xmath261 into a conjugacy class of @xmath230 .
indeed , it gives a one - to - one correspondence between the set of adjoint orbits close to @xmath264 and the set of conjugacy classes close to @xmath265 .
but , as mentioned in the introduction , important differences show up for large energy - momenta .
to understand the conjugacy classes in @xmath230 , it is handy to use the representation @xmath266 which says @xmath230 is geometrically a ` unit hyperboloid ' in a space of signature @xmath267 . since conjugate matrices have the same eigenvalues , the _ trace _ and thus the number @xmath125 is an invariant of conjugacy classes .
it is not a complete invariant , but it is except for matrices with @xmath268 .
every matrix in @xmath230 is conjugate to one of these five kinds : @xmath269 some explanation of this table is in order .
every ` rotation ' maps to a rotation in the connected lorentz group @xmath229 : in other words , a transformation that preserves a timelike vector in 3d minkowski spacetime .
similarly , every ` boost ' maps to a transformation that preserves a spacelike vector , and every ` shear ' maps to a transformation that preserves a lightlike vector .
since the two - to - one map from @xmath230 to @xmath229 maps the matrix @xmath4 to the identity , ` antiboosts ' get mapped to the same elements as boosts , and ` antishears ' get mapped to the same elements as shears .
( an ` antirotation ' would be just another rotation . )
the above chart counts certain conjugacy classes more than once .
first of all , there is an overlap at @xmath270 , since the identity rotation is also the identity shear and identity boost .
similarly , there is an overlap at @xmath271 , since a rotation by @xmath272 is also an antishear and an antiboost . finally , all shears ( resp . antishears ) with @xmath273 are conjugate to each other , and all shears ( resp .
antishears ) with @xmath274 are conjugate to each other . these are all the redundancies . knowing this
, we can list all the conjugacy classes in @xmath230 without any redundancies .
however , it is less tiresome to list the conjugacy classes in @xmath229 , since the elements @xmath275 get identified in @xmath229 , so we do not need to worry about ` antiboosts ' and ` antishears ' . here
are all the conjugacy classes in @xmath229 , and the corresponding five types of spin - zero particles : 1 . for any @xmath276
there is a conjugacy class containing the image of @xmath277 this corresponds to a tardyon of mass @xmath278 .
2 . for any @xmath279
there is a conjugacy class containing the image of @xmath280 this corresponds to a tachyon of mass @xmath41 .
3 . there is a conjugacy class containing the image of @xmath281 this corresponds to a positive - energy luxon .
4 . there is a conjugacy class containing the image of @xmath282 this corresponds to a negative - energy luxon .
there is a conjugacy class containing the image of @xmath283 this corresponds to a particle of vanishing energy - momentum .
the factors of @xmath284 here arise from the double cover @xmath285 .
as explained in the introduction , masses of tardyons really take values in the circle @xmath286 .
each conjugacy class @xmath287 admits an invariant measure which is unique up to an overall scale .
so , theorem [ 3dtheorem ] applies : we can form a hilbert space @xmath46 for particles of type @xmath45 , and more generally an @xmath8-particle hilbert space @xmath70 , on which the braid group and @xmath229 gauge transformations act as unitary transformations .
[ 4d ] all the work in the previous two sections generalizes nicely from 3 to 4 dimensions , using the loop braid group as a substitute for the braid group .
let space be @xmath2 with @xmath8 unknotted and unlinked circles removed : @xmath288 the fundamental group of @xmath181 is the free group on @xmath8 generators , so for any lie group @xmath1 we have @xmath289 as explained in section [ motion ] , a point in this space represents a @xmath1-bundle with flat connection over @xmath181 , mod gauge transformations that equal the identity at a chosen basepoint .
the @xmath8 elements of @xmath1 describing this point are just the holonomies around the circles @xmath102 .
physically , we think of these circles as string - like ` topological defects ' where the flat connection on space becomes singular . we explained quite generally in section [ motion ] how the motion group @xmath91 acts on @xmath193 . in the present case
the motion group is just the loop braid group @xmath61 , and its generators act on @xmath290 as follows : @xmath291 @xmath224 this is easy to see using pictures . for example , the generator @xmath210 has the following effect : @xmath292{lassortl.ps } } ; ( .5,1.2)*{g_1 } ; ( 2,1.3)*{g_2 } ; ( .5,-1.3)*{g_2 } ; ( 2.5,-1.2)*{g_1 } ; \endxy\ ] ] by an argument like the one we made in section [ 3d ] for the ordinary braid group action in 3d @xmath0 theory , it follows that @xmath210 acts on the holonomies @xmath293 by switching them while left conjugating @xmath213 by @xmath212 : @xmath294 similarly , the inverse of @xmath210 acts to switch the group elements while right conjugating @xmath212 by @xmath213 : @xmath295 the generator @xmath296 simply switches the holonomies @xmath212 and @xmath213 : @xmath297 it is easy to see that if @xmath1 is unimodular , this action of the loop braid group on @xmath60 gives rise to a unitary representation of the loop braid group on @xmath69 . and , just as in 3 dimensions , we can generalize this result to the case of a quandle : .1em [ 4dtheorem ] suppose @xmath45 is a topological quandle equipped with an invariant measure . then there is a unitary representation @xmath249 of the loop braid group @xmath61 on @xmath70 given by @xmath250 for all @xmath298 , where @xmath61 has a right action on @xmath252 given by : @xmath299 there is also a unitary operator @xmath254 on @xmath70 for each element @xmath255 , given by @xmath256 * proof . *
while the proof is straightforward , it is worth comparing theorem 5.1 of fenn , rimnyi and rourke @xcite .
this says that the braid permutation group @xmath154 is the group of automorphisms of the free quandle on @xmath8 generators .
since @xmath154 is isomorphic to the loop braid group @xmath61 , it follows that @xmath61 acts on @xmath252 for any quandle @xmath45 .
the action is precisely as above .
let us illustrate these ideas in the case where the gauge group is the connected lorentz group @xmath300 or its double cover @xmath301 . with either of these gauge groups , @xmath0 theory in 4 dimensions
is sometimes called ` topological gravity ' . in section [ quandle ]
we recalled the classification of conjugacy classes in @xmath229 and its double cover @xmath230 .
the classification for @xmath300 and its double cover @xmath301 is very similar , but simpler , because every complex number has a square root .
it is also more familiar , since any element of @xmath302 gives a fractional linear transformation @xmath303 such transformations are precisely the conformal transformations of the riemann sphere .
note that both @xmath3 and @xmath4 in @xmath301 map to the identity fractional linear transformation , so the conformal group of the riemann sphere is @xmath304 indeed , lorentz transformations can be thought of as conformal transformations of the ` celestial sphere ' : the set of light rays through an observer at the origin @xcite . a list of conjugacy classes in @xmath300 can thus be read off from the well - known classification of conformal transformations of the riemann sphere @xcite .
but in fact , it is easy enough to construct this list from first principles .
every element of @xmath300 is either conjugate to the image of the shear @xmath305 or conjugate to the image of @xmath306 for some @xmath307 .
the conjugacy class of the latter element is unchanged if we make the replacement @xmath308 , and its image in @xmath300 is unchanged if we make the replacement @xmath309 .
these replacements ( and their composite ) are the only ways we can change @xmath310 without changing the conjugacy class of the corresponding element of @xmath300 . using this
, we can see there there are five types of conjugacy classes in @xmath300 : 1 . for any real @xmath278 with @xmath311
there is a conjugacy class containing the image of @xmath312 an element conjugate to one of this form is called * elliptic*. 2 . for any purely imaginary @xmath278 with @xmath313
there is a conjugacy class containing the image of @xmath314 an element conjugate to one of this form is called * hyperbolic*. 3 . for any @xmath315 with @xmath316 and @xmath313
there is a conjugacy class containing the image of @xmath314 an element conjugate to one of this form is called * loxodromic*. 4 .
there is a conjugacy class containing the image of @xmath317 an element conjugate to one of this form is called * parabolic*. 5 .
there is a conjugacy class containing the image of @xmath318 this class contains only the identity element .
now let us return to @xmath0 theory with gauge group @xmath300 , taking space to be @xmath2 with a collection of unknotted unlinked circles @xmath102 removed . for brevity
let us call these circles ` closed strings ' .
a flat connection on space will have some holonomy @xmath319 around the @xmath107th string .
the above list of conjugacy classes lets us list possible ` types ' of strings , just as we used conjugacy classes in @xmath229 to list types of point particles in 3d gravity : 1 .
if @xmath320 is elliptic , it acts on minkowski spacetime as a spatial rotation in some reference frame . in this reference frame ,
parallel transport around the string @xmath321 is a spatial rotation by some angle @xmath322 about some axis .
( a rotation by an angle @xmath323 is a rotation by @xmath324 about the opposite axis . )
this angle @xmath6 is proportional to the real number @xmath278 which appears in item 1 of the above list , as follows : @xmath325 by analogy to 3d gravity , we could call the string a * tardyon * in this case , and call the number @xmath278 its ` mass density ' . the number
@xmath278 is real and takes values @xmath311 .
if @xmath320 is hyperbolic , it acts on minkowski spacetime as a boost in some reference frame . in this reference frame ,
parallel transport around the string @xmath321 is a boost with rapidity @xmath326 along some axis .
the rapidity @xmath327 is proportional to the imaginary number @xmath278 which appears in item 2 of the above list , as follows : @xmath328 by analogy to 3d gravity , we could call the string a * tachyon * in this case , and call the number @xmath278 its ` mass density ' .
the number @xmath278 is purely imaginary and takes values in the upper half of the imaginary axis : @xmath313 .
if @xmath320 is loxodromic , it acts on minkowski spacetime as a combined rotation and boost about the same axis in some reference frame . in this reference frame ,
parallel transport around the string @xmath321 is a combination of a rotation by an angle @xmath329 and a boost with rapidity @xmath326 about the same axis , where @xmath330 this case has no analogue in 3d gravity .
we can still think of @xmath278 as some sort of mass density , but it is complex , with @xmath316 and @xmath313 .
if @xmath320 is parabolic , it acts on minkowski spacetime as a lorentz transformation fixing a single null vector . by analogy to 3d gravity , we could call the string a * luxon * in this case , and say @xmath331 .
if @xmath320 is the identity , we can say the string carries no energy - momentum , and again say @xmath331 .
each of these conjugacy classes @xmath332 is a quandle .
the question then arises which of these quandles admits an invariant measure , and whether this measure is unique up to scale .
one can work this out on a case - by - case basis .
one important case is when @xmath45 is the conjugacy class containing all rotations by some fixed angle @xmath333 .
this conjugacy class corresponds to a ` tardyonic ' closed string with a given mass density @xmath311 .
it is easy to see that this conjugacy class @xmath45 indeed admits an invariant measure . to see this , note
that to specify a rotation by the angle @xmath6 one must first pick a future - pointing unit timelike vector @xmath334 , to split minkowski spacetime into space and time , and then pick a unit spacelike vector @xmath335 orthogonal to @xmath336 , to serve as the axis of rotation .
the allowed choices of @xmath336 lie in the hyperboloid @xmath337 this hyperboloid @xmath338 is a riemannian submanifold of @xmath339 .
an allowed choice of @xmath336 together with @xmath335 amounts to a point in @xmath340 , the unit sphere bundle of @xmath338 .
so , we have @xmath341 .
since the unit sphere bundle of a riemannian manifold is itself a riemannian manifold in a natural way , we get a well - defined lebesgue measure on @xmath340 and thus @xmath45 , which is invariant under @xmath300 , since our construction respected the lorentz group symmetry . given an invariant measure on @xmath45
, we obtain a hilbert space @xmath70 for @xmath8 strings of type @xmath45 .
note that we do not try to ` symmetrize ' the states in this hilbert space .
instead we describe the statistics using a representation of the loop braid group , following theorem [ 4dtheorem ] . of course , one should work out the details explicitly , but we leave this for future research .
much more needs to be done to ferret out the physical significance of the theory we have been considering here . first , there are some nice projects for the mathematician .
one should determine for various lie groups @xmath1 which conjugacy classes @xmath53 admit invariant measures , and when these invariant measures are unique up to an overall scale .
we have only done this for @xmath342 , but for applications to 4d physics other groups are more relevant especially the lorentz , poincar , desitter and anti - desitter groups . then , given a conjugacy class @xmath53 with an invariant measure , one should work out explicitly the representation of the loop braid group @xmath61 on the hilbert space @xmath70 , if possible decomposing this representation into irreducibles , so as to understand in detail the workings of the exotic statistics .
it would also be interesting to study how , in the @xmath28 limit , the exotic statistics approach ordinary bosonic statistics . for the physicist , one interesting project would be to study the _ dynamics _ and _ interactions _ of the ` closed strings ' discussed at the purely kinematical level here . in a paper with perez @xcite we describe a lagrangian whereby these objects can couple to the fields in @xmath0 theory .
we work out the equations of motion and propose a strategy for quantizing the resulting theory , analogous to the known quantization of point particles coupled to 3d gravity @xcite
. a more ambitious project would be to generalize all our results from collections of unlinked unknotted circles to arbitrary embedded graphs .
finally , a still more ambitious project would be to use these ideas as part of a perturbative expansion of macdowell mansouri gravity about 4d @xmath0 theory , as proposed by freidel and starodubtsev @xcite .
here we present a proof of theorem [ presentation2 ] on p. .
* proof . *
we begin by demonstrating that the relations in the statement of theorem [ presentation2 ] follow from those given in theorem [ presentation1 ] .
it clearly suffices to show that the relations in @xmath343 and @xmath344 follow from the relations in @xmath345 and @xmath138 . in what follows ,
we make frequent use of the correspondence between generators @xmath104 of @xmath100 and generators @xmath64 of @xmath61 as given in ( [ generatorcorrespondence1 ] ) and ( [ generatorcorrespondence2 ] ) .
in fact , since these follow from different relations in the presentation of theorem [ presentation1 ] , it suffices for our purposes to take one expression from each of these , say @xmath346 these representations of @xmath104 follow directly from the definition of @xmath64 along with the relations ( [ rel7 ] ) , ( [ rel9 ] ) , and ( [ rel11 ] ) . + @xmath347 relation ( [ rel7 ] ) : we wish to show that @xmath348 for @xmath349 to check this , we begin with relation ( [ rel8 ] ) in the form : @xmath350 where @xmath351 using ( [ generatorcorrespondence ] ) above , this becomes : @xmath352 applying relation ( [ rel1 ] ) of to the left - hand side and then cancelling @xmath62 from each side gives @xmath353 when @xmath354 which is ( [ rel7 ] ) .
+ @xmath347 relation ( [ rel8 ] ) : we wish to show that @xmath355 for @xmath356 beginning with relation ( [ rel9 ] ) with @xmath357 , we obtain : @xmath358 by ( [ generatorcorrespondence ] ) this gives : @xmath359 multiplying on the right by @xmath360 and on the left by @xmath361 , we have : @xmath362 this can be rewritten as @xmath363 which is ( [ rel8 ] )
. + @xmath347 relation ( [ rel9 ] ) : we wish to show that @xmath364 for @xmath356 to verify this we use relation ( [ rel5 ] ) with @xmath365 and @xmath366 , which gives : @xmath367 multiplying by @xmath370 on the left produces : @xmath371 which is ( [ rel9 ] ) .
+ @xmath347 relation ( [ rel4 ] ) : we wish to show that @xmath372 for @xmath373 . to do so
, we use relation ( [ rel4 ] ) with @xmath374 , which are clearly all distinct for @xmath375 .
we therefore have : @xmath376 which , by ( [ generatorcorrespondence ] ) , becomes : @xmath377 applying ( [ rel7 ] ) to both sides of this equation , followed by relation ( [ rel1 ] ) , we obtain : @xmath378 with @xmath379 which is ( [ rel4 ] )
. + @xmath347 relation ( [ rel5 ] ) : we wish to show that @xmath380 for @xmath381 . to check this
we start with relation ( [ rel6 ] ) with @xmath382 and @xmath366 , which are clearly all distinct .
thus , we have : @xmath383 using the correspondence given in ( [ generatorcorrespondence ] ) and cancelling @xmath62 from both sides , we obtain : @xmath384 cancelling @xmath385 on the left and multiplying by @xmath385 on the right produces : @xmath386 where in the last step we used ( [ rel5 ] ) in the form @xmath387 .
this is ( [ rel5 ] ) .
+ the loop braid group thus has generators that satisfy all of the relations of the braid permutation group .
it remains to show that these relations are sufficient , which we do by demonstrating that the relations in the statement of theorem [ presentation1 ] follow from those given in theorem [ presentation2 ] . in this direction of the proof it is convenient to use both of the equivalent expressions ( [ generatorcorrespondence1 ] ) and ( [ generatorcorrespondence2 ] ) as the correspondence between generators @xmath64 and @xmath104 .
+ @xmath347 relation ( [ rel7 ] ) : this relation simply says @xmath388 , which is immediate from ( [ generatorcorrespondence ] ) since both sides are equal to @xmath64 .
+ @xmath347 relation ( [ rel8 ] ) : we wish to show @xmath389 , whenever @xmath390 are distinct .
when either @xmath391 or @xmath392 , @xmath127 commutes with each of the factors in the expansion @xmath393 by ( [ rel1 ] ) and ( [ rel7 ] ) .
similarly , when @xmath394 or @xmath395 , @xmath127 commutes with each factor in @xmath396 when @xmath397 we also need two applications of ( [ rel2 ] ) : @xmath398 the only remaining case is @xmath399 , which is handled similarly
. + @xmath347 relation ( [ rel9 ] ) : we wish to show that @xmath400 whenever @xmath401 .
when @xmath402 we have : @xmath403 and the case @xmath404 is similar .
+ @xmath347 relation ( [ rel11 ] ) : the proof that @xmath405 is essentially the same as the proof of ( [ rel9 ] ) above .
+ @xmath347 relation ( [ rel4 ] ) : we wish to show @xmath406 , whenever @xmath407 and @xmath408 are distinct .
naively there are @xmath409 orderings of @xmath410 to consider , but symmetry of the relation implies only 8 are independent .
all cases are proved similarly ; we demonstrate only the case @xmath411 : @xmath412 @xmath347 relation ( [ rel5 ] ) : we wish to show that @xmath413 when @xmath414 are distinct .
we have three independent cases : @xmath415 , @xmath416 , and @xmath417 . in the case
@xmath392 , we first note that if @xmath145 , then by ( [ rel8 ] ) and ( [ rel9 ] ) we have : @xmath418 by repeated application of these facts , it suffices to consider the subcase where @xmath419 .
similarly , if @xmath420 , we can use ( [ rel8 ] ) and ( [ rel11 ] ) to reduce to the case where @xmath421 .
thus it suffices to consider only the cases where @xmath129 are consecutive : @xmath422 this proves the case @xmath392 .
the remaining two cases are similar
. + @xmath347 relation ( [ rel6 ] ) : we wish to show that @xmath423 when @xmath129 are distinct . in light of ( [ rel5 ] )
this equation is symmetric under the interchange of @xmath107 and @xmath424 , and this symmetry reduces the number of independent cases to 3 : @xmath392 , @xmath425 , and @xmath395 . in the case
@xmath392 , we first note that if @xmath145 , then by ( [ rel8 ] ) and ( [ rel9 ] ) we have @xmath426 by repeated application of these facts , it suffices to consider the subcase where @xmath419 .
similarly , if @xmath420 , we can use ( [ rel8 ] ) and ( [ rel11 ] ) to reduce to the case where @xmath421 .
thus it suffices to consider only the cases where @xmath129 are consecutive : @xmath427 this proves the case of @xmath392 .
the other two independent cases are similar . as pointed out by blake winter
, one can also prove theorem [ presentation2 ] as follows .
fenn , rimnyi , and rourke @xcite show that the braid permutation group @xmath154 is isomorphic to the subgroup , of @xmath428 generated by all permutations of basis elements , together with all operations of conjugating one basis element by another .
let @xmath181 be @xmath2 with unlinked unknotted circles @xmath102 removed .
as we have seen , @xmath429 , the free group on @xmath8 generators , so by the work of dahm , the loop braid group acts as automorphisms of @xmath430 .
let @xmath431 be the resulting homomorphism .
goldsmith @xcite shows that the image of @xmath432 is precisely the above subgroup of @xmath428 and that , moreover , @xmath432 is one - to - one .
it follows that @xmath61 and @xmath154 are isomorphic .
since fenn , rimnyi and rourke prove that @xmath154 has the presentation given in theorem [ presentation2 ] , it follows that @xmath61 also has this presentation .
most of all we would like to thank xiao - song lin for allowing us to read a draft of his paper on the loop braid group .
this inspired us to write the present paper .
we would also like to thank a. p. balachandran , james dolan , sam nelson , carlo rovelli , sumati surya , and blake winter .
j. c. baez , an introduction to spin foam models of bf theory and quantum gravity , in _ geometry and quantum physics _
, eds.h .
gausterer and h. grosse , lecture notes in physics * 543 * , springer , berlin , 2000 , pp .
25 - 93 . also available as gr - qc/9905087 .
a. p. balachandran , statistics , strings and gravity , in _ mathematical physics : proceedings , 3rd regional conference , islamabad , pakistan , february 1724 , 1989 _ , eds .
f. hussain and a. qadir , world scientific , singapore , 1990 , pp .
31 - 45 .
f. e. camino , w. zhou and v. j. goldman , realization of a laughlin quasiparticle interferometer : observation of fractional statistics , * phys .
b * * 72 * ( 2005 ) , 075342 . also available as cond - mat/0502406 .
l. freidel and d. louapre ,
ponzano - regge model revisited i : gauge fixing , observables and interacting spinning particles , _ class .
* 21 * ( 2004 ) , 5685 - 5726 . also available as hep - th/0401076 | after a review of exotic statistics for point particles in 3d @xmath0 theory , and especially 3d quantum gravity , we show that string - like defects in 4d @xmath0 theory obey exotic statistics governed by the ` loop braid group ' .
this group has a set of generators that switch two strings just as one would normally switch point particles , but also a set of generators that switch two strings by passing one through the other .
the first set generates a copy of the symmetric group , while the second generates a copy of the braid group .
thanks to recent work of xiao - song lin , we can give a presentation of the whole loop braid group , which turns out to be isomorphic to the ` braid permutation group ' of fenn , rimnyi and rourke .
in the context 4d @xmath0 theory this group naturally acts on the moduli space of flat @xmath1-bundles on the complement of a collection of unlinked unknotted circles in @xmath2 .
when @xmath1 is unimodular , this gives a unitary representation of the loop braid group .
we also discuss ` quandle field theory ' , in which the gauge group @xmath1 is replaced by a quandle . |
planets that transit their host stars play a special role in our understanding of the characteristics of exoplanets : their transit allows us to accurately determine the radius and the orbital inclination of the planet from the photometric light curve so that an actual mass can be derived from a spectroscopic orbit of the host star .
the mass and radius enable us to infer a bulk composition of the planet , and although there are degeneracies associated with the bulk composition , it allows us to put constraints on models of planetary structure and formation theories .
the incredible diversity of the over 60 discovered transiting planets , ranging from dense planets with a higher mean density than that of copper to strongly irradiated puffed - up planets with a mean density comparable to that of corkwood , have baffled the exoplanet community , and no unified theory has been established to explain all the systems consistently . transiting extrasolar planet ( tep ) discoveries are primarily the result of dedicated ground based searches , such as superwasp @xcite , hatnet @xcite , tres @xcite and xo @xcite , and space - borne searches , such as corot @xcite and kepler @xcite .
since its commissioning in 2003 , the hungarian - made automated telescope network ( hatnet ; * ? ? ?
* ) survey has been one of the major contributors to the discoveries of teps .
hatnet has discovered over a dozen teps since 2006 by surveying bright stars ( @xmath17mag ) in the northern hemisphere and has now covered approximately 11% of the northern sky .
hatnet consists of six wide field automated telescopes ; four of these are located at the fred lawrence whipple observatory ( flwo ) in arizona , and two on the roof of the submillimeter array hangar ( sma ) of the smithsonian astrophysical observatory ( sao ) in hawaii . in this paper we report a new tep discovery of hatnet , called hat - p-16b .
the structure of the paper is as follows . in
we summarize the observations , including the photometric detection , and follow - up observations .
in we describe the analysis of the data , such as the stellar parameter determination ( ) , ruling out blend scenarios ( ) , and global modeling of the data ( ) .
we discuss our findings in .
the transits of hat - p-16b were detected with the hat-6 and hat-7 telescopes in arizona and the hat-8 and hat-9 telescopes in hawaii .
the star gsc 2792 - 01700 lies in the intersection of 3 separate hatnet fields internally labeled as 123 , 163 and 164 .
field 123 was observed with an @xmath18 filter and a 2k@xmath192k ccd , while fields 163 and 164 were observed through @xmath20 filters with 4k@xmath194k ccds .
all three fields were observed with a 5 minute exposure time and at a 5.5 minute cadence in the period between july 2007 and september 2008 during which over 12,000 exposures were gathered for the three fields .
each image contains between 27,000 and 76,000 stars down to a magnitude of @xmath21 for field 123 and @xmath22 for fields 163 and 164 , yielding a photometric precision for the brightest stars in the field of @xmath23mmag for field 123 and @xmath24mmag for fields 163 and 164 .
frame calibration , astrometry and aperture photometry were done in an identical way to recent hatnet tep discoveries , as described in @xcite and @xcite .
the resulting light curves were decorrelated ( cleaned of trends ) using the external parameter decorrelation technique in `` constant '' mode ( epd ; see * ? ? ?
* ) and the trend filtering algorithm ( tfa ; see * ? ? ?
the light curves were searched for periodic box - like signals using the box least - squares method ( bls ; see * ? ? ? * ) .
a significant signal was detected in the light curve of gsc 2792 - 01700 ( also known as 2mass 00381756 + 4227470 ; @xmath25 , @xmath26 ; j2000 ; @xmath27 ; @xcite ) , with a depth of @xmath28mmag , and a period of @xmath29days .
the dip had a relative duration ( first to last contact ) of @xmath30 , corresponding to a total duration of @xmath31 hr ( see figure [ fig : hatnet ] ) .
transiting planet candidates found by ground - based , wide - field photometric surveys must undergo a rigorous vetting process to eliminate the many astrophysical systems mimicking transiting planets ( called false positives ) , the rate of which has proved to be much higher than the occurrence of true planets ( 10 to 20 times higher ) . low signal to noise ratio ( s / n ) high - resolution reconnaissance spectra are used to extract stellar parameters such as effective temperature , gravity , metallicity and rotational and radial velocities to rule out these false positives .
examples of false positives which are discarded are eclipsing binaries and triple systems .
the latter can be either hierarchical or chance alignment systems where the light of the eclipsing pair of stars is diluted by the light of a third brighter star .
rapidly rotating and/or hot host stars whose spectrum is unsuitable for high precision velocity work are also discarded .
we acquired @xmath32 reconnaissance spectra with the cfa digital speedometer ( ds ; * ? ? ?
* ) mounted on the flwo 1.5 m tillinghast reflector between december 2008 and january 2009 .
the extracted modest precision radial velocities gave a mean absolute rv=@xmath33@xmath34 with an rms of @xmath35 , which is consistent with no detectable rv variation .
the stellar parameters derived from the spectra @xcite , including the effective temperature @xmath36 , surface gravity @xmath37 ( log cgs ) and projected rotational velocity @xmath38 , correspond to a f8 dwarf
. the high significance of the transit detection by hatnet , together with the stellar spectral type and small rv variations encouraged us to gather high resolution , high s / n spectra to determine the orbit of the system .
we have taken 21 spectra between august and october 2009 using the fibre fed chelle spectrograph ( fies ) at the 2.5 m nordic optical telescope ( not ) at la palma , spain @xcite .
we used the medium and the high resolution fibers ( @xmath39 projected diameter ) with resolving powers of @xmath40 and @xmath41 , respectively , giving a wavelength coverage of @xmath42 .
recently , fies has also been used to obtain a spectroscopic orbit for one of the first kepler planets , namely kepler-7b @xcite .
we also used the hires instrument @xcite on the keck i telescope located on mauna kea , hawaii .
we obtained 6 exposures with an iodine cell , plus 1 iodine - free template with keck - i / hires .
the observations were made on 6 nights between 2009 july 3 and 2009 october 29 .
the width of the spectrometer slit used on hires was @xmath43 , resulting in a resolving power of @xmath44 , with a wavelength coverage of @xmath45 .
the iodine gas absorption cell was used to superimpose a dense forest of @xmath46 lines on the stellar spectrum and establish an accurate wavelength fiducial ( see * ? ? ?
relative rvs in the solar system barycentric frame were derived as described by @xcite , incorporating full modeling of the spatial and temporal variations of the instrumental profile .
the final rv data and their errors ( for both instruments ) are listed in table [ tab : rvs ] .
the folded data are plotted in figure [ fig : rvbis ] .
the systemic gamma velocities ( that are the result of the global modeling , as laid out in ) have been subtracted to ensure a common zero point .
the best orbital fit ( see ) is superimposed in the figure .
[ ht ] since this is the first time the not has been used to determine the orbit of a hatnet planet , we describe briefly the extraction of the radial velocities .
we have built a custom pipeline to rectify and cross correlate spectra from chelle spectrographs with the goal of providing very precise radial velocities .
the first step in the extraction process is to remove the bias level and crop the raw images . to effectively remove cosmic rays ,
the observation is divided into three separate exposures , which enables us to combine the raw images using median filtering , removing virtually all cosmic rays .
we use a flat field to trace the chelle orders and we rectify the spectra by using the `` optimal extraction '' algorithm @xcite . the blaze function is determined by fitting cubic splines to a combined high s / n flat field exposure and is saved separately in order to preserve the original flux in the stellar exposure . by dividing the normalized blaze function into the rectified flat field spectrum , we can determine the pixel to pixel variations and correct for these . the scattered light in the two dimensional raw image
is determined and removed by masking out the signal in the chelle orders and fitting the inter
order scattered light flux with a two dimensional polynomial . thorium argon ( thar ) calibration images are obtained through the science fiber before and after the stellar observation .
the two calibration images are combined to form the basis for the fiducial wavelength calibration .
we have determined that the best wavelength solution is achieved by choosing an exposure time that saturates the strong argon lines but enhances the forest of weaker thorium lines .
fies is not a vacuum spectrograph and is only temperature controlled to @xmath47 .
consequently , the radial velocity errors are dominated by our limited ability to correct for shifts due to pressure , humidity and temperature variations .
in order to successfully remove these large variations ( @xmath48 ) , it is critical that the thar light travels through the same light path as the stellar light and thus acts as an effective proxy to remove these variations .
we have therefore chosen to interleave the stellar observation between two thar exposures instead of using the simultaneous thar technique , which may not exactly describe the induced shifts in the science fiber .
the centers of the thar lines are found by fitting a gaussian function to the line profiles and a two dimensional fifth order legendre polynomial is used to describe the wavelength solution .
once the spectra have been extracted , a multi order cross correlation is performed order by order .
first , the spectra are interpolated to a common oversampled log wavelength scale with the same monotonic wavelength increment in all orders .
a high and low pass filter is applied in the fourier domain and the ends of the spectra are apodized with a cosine bell function .
the orders are cross correlated using a fast fourier transform ( fft ) and the cross correlation function ( ccf ) for each order is co added .
this automatically weights each order by its flux , giving more weight to the orders with more photons .
this ccf peak is fitted with a gaussian function to determine its center .
the internal precision is estimated by finding the radial velocity for each order and the rv error is thus @xmath49 , where @xmath50 is the rv of the individual orders and @xmath51 is the number of orders . for the final radial velocities ,
a template spectrum is constructed by shifting and co adding all the observed spectra , and the individual spectra are cross correlated against this co
added template spectrum to minimize the contribution of noise in the template . to confirm the transit signal and obtain high - precision light curves for modeling the system
, we conducted photometric follow - up observations with the keplercam ccd on the telescope .
we observed 4 transit events of hat - p-16 on the nights of 2009 september 10 , september 21 , october 19 and october 30 ( figure [ fig : lc ] ) . on the four nights
, @xmath52 and @xmath53 frames were acquired with a cadence of @xmath54 and @xmath55 seconds ( @xmath56 and @xmath57 seconds of exposure time ) in the sloan @xmath58 band , respectively .
the reduction of the images , including frame calibration , astrometry , and photometry were performed as described in @xcite .
we also performed epd and tfa to remove trends simultaneously with the light curve modeling ( for more details , see ) .
the final light curves are shown in the upper plots of figure [ fig : lc ] , superimposed with our best - fit transit light curve models ( see also ) ; the photometry is provided in table [ tab : phfu ] .
lrrrr @xmath59 & @xmath60 & @xmath61 & @xmath62 & @xmath63 + @xmath64 & @xmath65 & @xmath61 & @xmath66 & @xmath63 + @xmath67 & @xmath68 & @xmath61 & @xmath69 & @xmath63 + @xmath70 & @xmath71 & @xmath61 & @xmath72 & @xmath63 + @xmath73 & @xmath74 & @xmath61 & @xmath75 & @xmath63 +
the stellar atmospheric parameters were derived from the template spectrum obtained with the keck / hires instrument .
we used the spectroscopy made easy ( sme ) analysis package of @xcite , along with the atomic - line database of @xcite .
this yielded the following _ initial _ values and uncertainties ( which we have conservatively increased to include our estimates of the systematic errors ) : effective temperature @xmath76k , stellar surface gravity @xmath77(cgs ) , metallicity @xmath8}}={\ensuremath{0.15\pm0.06}}$]dex , and projected rotational velocity @xmath78 .
we could now use the effective temperature and the surface gravity found by sme to determine other stellar parameters , such as the mass , @xmath79 , and radius , @xmath80 , using model isochrones .
however , there may be strong correlations between effective temperature , gravity and metallicity in the spectroscopic determination of these parameters . also , the effect of @xmath81 on the spectral line shapes is typically subtle , and as a result @xmath81 is generally a rather poor luminosity indicator .
instead , we used the @xmath82 , the normalized semimajor axis ( related to @xmath83 , the mean stellar density ) , which can be derived directly from the light curves @xcite .
the stellar parameters were determined simultaneously with the modeling of the light curves and radial velocities , as described next .
we began by adopting the values of @xmath84 , @xmath85 $ ] , and @xmath81 from the sme analysis to fix the quadratic limb - darkening coefficients from the tabulations by @xcite , which are needed to model the light curves(@xmath86 , @xmath87 ) .
this modeling yields the probability distribution of @xmath82 via a monte carlo approach , which is described fully in .
we then used the @xmath82 distribution together with gaussian distributions for @xmath84 and @xmath85 $ ] , with 1@xmath88 uncertainties as reported previously , to estimate @xmath79 and @xmath80by comparison with the stellar evolution models of @xcite .
this was done for each of 15,000 simulations .
parameter combinations corresponding to unphysical locations in the h - r diagram ( 4% of the trials ) were ignored , and replaced with another randomly drawn parameter set .
the resulting stellar parameters and their uncertainties were determined from the _ a posteriori _ distributions obtained in this way .
in particular , the resulting surface gravity of @xmath89 is somewhat different from that derived in the initial sme analysis , which is not surprising in view of the possible strong correlations among @xmath84 , @xmath85 $ ] , and the surface gravity .
therefore , in a second iteration we adopted this value of @xmath81 and held it fixed in a new sme analysis , adjusting only @xmath84 , @xmath85 $ ] , and @xmath90 .
this gave @xmath91k , @xmath8}}={\ensuremath{+0.17\pm0.08}}$ ] , and @xmath92@xmath34 , which we adopt as final atmospheric values for the star .
finally , we repeated the calculation of the stellar mass and radius with these values and the stellar evolution models , and obtained @xmath79=@xmath93@xmath5 , @xmath80=@xmath94@xmath7 and @xmath95=@xmath96@xmath97 , along with other parameters summarized in table [ tab : stellar ] .
the model isochrones from @xcite for metallicity @xmath85$]= are plotted in figure [ fig : iso ] , with the final choice of effective temperature @xmath98 and @xmath82 marked , and encircled by the 1@xmath88 and 2@xmath88 confidence ellipsoids .
the stellar evolution modeling provides color indices that may be compared against the measured values as a sanity check .
the best available measurements are the near - infrared magnitudes from the 2mass catalogue @xcite , @xmath99 , @xmath100 and @xmath101 ; which we have converted to the photometric system of the models ( eso ) using the transformations by @xcite .
the resulting measured color index is @xmath102 .
this is within 1@xmath88 of the predicted value from the isochrones of @xmath103 .
the distance to the object may be computed from the absolute @xmath104 magnitude from the models ( @xmath105 ) and the 2mass @xmath106 magnitude , which has the advantage of being less affected by extinction than optical magnitudes .
the result is @xmath107pc , where the uncertainty excludes possible systematics in the model isochrones that are difficult to quantify .
lll @xmath98 ( k ) & & sme + @xmath8}}$ ] ( dex ) & & sme + @xmath108 ( @xmath34 ) & & sme + @xmath109 ( @xmath34 ) & & sme + @xmath110 ( @xmath34 ) & & sme + @xmath111 ( @xmath34 ) & @xmath112 & ds + @xmath86 & @xmath113 & sme+claret + @xmath87 & @xmath114 & sme+claret + @xmath115 ( @xmath116 ) & @xmath93 & yy+@xmath82+sme + @xmath117 ( @xmath118 ) & @xmath94 & yy+@xmath82+sme + @xmath119 ( cgs ) & @xmath120 & yy+@xmath82+sme + @xmath121 ( @xmath122 ) & @xmath96 & yy+@xmath82+sme + @xmath123 ( mag ) & 10.812 & tass + @xmath124 ( mag ) & @xmath125 & yy+@xmath82+sme + @xmath104 ( mag , eso ) & @xmath126 & 2mass+carpenter + @xmath127 ( mag , eso ) & @xmath128 & yy+@xmath82+sme + age ( gyr ) & @xmath129 & yy+@xmath82+sme + distance ( pc ) & @xmath130 & yy+@xmath82+sme + we performed a line bisector analysis of the observed spectra following @xcite to explore the possibility that the main cause of the radial velocity variations is actually distortions in the spectral line profiles due to stellar activity or a nearby unresolved eclipsing binary .
the bisector analysis was performed on the keck spectra as described in 5 of @xcite ; the spectra from the not were analyzed in a similar manner .
the resulting bisector span variations are plotted in figure [ fig : rvbis ] and have a low amplitude compared to the orbital semi amplitude and show no correlation with the radial velocities .
we therefore conclude that the velocity variations are real , and that the star is transited by a close - in giant planet . our model for the follow - up light curves used analytic formulae based on @xcite for the eclipse of a star by a planet , where the stellar flux is described by
quadratic limb - darkening .
the limb darkening coefficients are based on the sme results ( ) and the tables provided by @xcite for the @xmath58 band .
the transit shape was parametrized by the normalized planetary radius @xmath131 , the square of the impact parameter @xmath132 , and the reciprocal of the half mid - ingress to mid - egress duration of the transit @xmath133 .
we chose these parameters because of their simple geometric meanings and the fact that they show negligible correlations ( see * ? ? ? * ) .
our model for the hatnet data was the simplified `` p1p3 '' version of the @xcite analytic functions ( an expansion by legendre polynomials ) , for the reasons described in @xcite . following the formalism presented by @xcite ,
the rv curve was modeled as an eccentric keplerian orbit with semiamplitude @xmath104 , and lagrangian orbital elements @xmath134 we assumed that there is a strict periodicity in the individual transit times . in practice , we assigned the transit number @xmath135 to the first high quality follow - up light curve gathered on 2009 september 10
. the adjusted parameters in the fit were the first transit center observed with hatnet , @xmath136 , and the last high quality transit center observed with the telescope , @xmath137 .
the transit center times for the intermediate transits were interpolated using these two epochs and the @xmath138 transit number of the actual event @xcite .
the model for the rv data contains the ephemeris information through the @xmath136 and @xmath137 variables .
altogether , the 8 parameters describing the physical model were @xmath136 , @xmath137 , @xmath131 , @xmath132 , @xmath133 , @xmath104 , @xmath139 , @xmath140 .
nine additional parameters were related to the instrumental configuration .
these are the three instrumental blend factors @xmath141 of hatnet ( one for each of the three fields ) , which account for possible dilution of the transit in the hatnet light curve due to the wide ( 20 wide fwhm ) psf and possible crowding , the three hatnet out - of - transit magnitudes , @xmath142 , and three relative rv zero points @xmath143 ( one each for the keck , high - resolution fies , and medium - resolution fies observations ) .
we extended our physical model with an instrumental model that describes the systematic variations of the data .
this was done in a similar fashion to the analysis presented in @xcite .
the hatnet photometry has already been epd- and tfa - corrected before the global modeling , so we only considered systematic correction of the follow - up light curves .
we chose the `` eltg '' method , i.e. epd was performed in `` local '' mode with epd coefficients defined for each night , and tfa was performed in `` global '' mode using the same set of stars and tfa coefficients for all nights .
the underlying physical model was based on the @xcite analytic formulae , as described earlier .
the five epd parameters were the hour angle ( characterizing a monotonic trend that linearly changes over time ) , the square of the hour angle , and the stellar profile parameters ( equivalent to fwhm , elongation , position angle ) .
the functional form of the above parameters contained six coefficients , including the auxiliary out - of - transit magnitude of the individual events .
the epd parameters were independent for all four nights , implying 24 additional coefficients in the global fit .
for the global tfa analysis we chose 20 template stars that had good quality measurements for all nights and on all frames , implying an additional 20 parameters in the fit .
thus , the total number of fitted parameters was 17 ( physical model ) + 24 ( local epd ) + 20 ( global tfa ) = 61 .
the joint fit was performed as described in @xcite .
we minimized @xmath144 in the parameter space by using a hybrid algorithm , combining the downhill simplex method ( amoeba ; see * ? ? ?
* ) with the classical linear least squares algorithm .
uncertainties for the parameters were derived using the markov chain monte - carlo method ( mcmc , see * ? ? ?
* ; * ? ? ?
the eccentricity of the system appeared as significantly non - zero ( @xmath145 , @xmath146 ) .
the best - fit results for the relevant physical parameters are summarized in table [ tab : planetparam ] .
we also list the rv `` jitter , '' which is a component of assumed astrophysical noise intrinsic to the star that we add in quadrature to the rv measurement uncertainties in order to have @xmath147 from the rv data for the global fit .
in addition , some auxiliary parameters ( not listed in the table ) are : @xmath148 ( bjd ) , @xmath149 ( bjd ) , @xmath150@xmath151 , @xmath152@xmath151(fies , high resolution ) , @xmath153@xmath151(fies , medium resolution ) , @xmath154 , @xmath155 , @xmath156 .
note that these gamma velocities do _ not _ correspond to the true center of mass rv of the system , but are only relative offsets .
the true systemic velocity of the system is rv=@xmath112@xmath34found by cross correlating the observed spectrum against a library template spectrum .
the planetary parameters and their uncertainties can be derived by the direct combination of the _ a posteriori _ distributions of the light curve , rv and stellar parameters .
we found that the planet is fairly massive with @xmath157 , and compact with radius of @xmath158 , yielding a mean density of @xmath159@xmath14 .
the final planetary parameters are summarized at the bottom of table [ tab : planetparam ] .
lr @xmath160 ( days ) & @xmath161 + @xmath162 ( @xmath163 ) & @xmath164 + @xmath165 ( days ) & @xmath166 + @xmath167 ( days ) & @xmath168 + @xmath169 & @xmath170 + @xmath133 & @xmath171 + @xmath131 & @xmath172 + @xmath132 & @xmath173 + @xmath174 & @xmath175 + @xmath58 ( deg ) & @xmath176 + @xmath104 ( @xmath151 ) & @xmath177 + @xmath178 & @xmath179 + @xmath180 & @xmath181 + @xmath182 & @xmath183 + @xmath184 & @xmath185 + rv jitter ( @xmath151 ) & @xmath186 + rv rms from fit ( @xmath151 ) & @xmath187 + @xmath188 ( bjd ) & @xmath189 + @xmath190 & @xmath191 + @xmath192 & @xmath193 + @xmath194 ( @xmath195 ) & @xmath196 + @xmath197 ( @xmath198 ) & @xmath199 + @xmath200 & @xmath201 + @xmath202 ( @xmath14 ) & @xmath203 + @xmath204 ( au ) & @xmath205 + @xmath206 ( cgs ) & @xmath207 + @xmath208 ( k ) & @xmath209 + @xmath210 & @xmath211 + @xmath212 ( @xmath213@xmath214 ) & @xmath215 +
we present the discovery of hat - p-16b with a period of @xmath216 , a mass of @xmath217 and a radius of @xmath218 . currently , there are only a handful of planets residing near hat - p-16b in the mass radius diagram .
corot-2b @xcite and corot-6b @xcite have similar masses and periods , but both have eccentricities consistent with zero .
hd80606 @xcite and hd17156 @xcite are also similar in mass , but both planets have long periods and very eccentric orbits .
wasp-10b ( @xmath219d , @xmath220 , see @xcite ) is probably the transiting planet which most resembles hat - p-16b with a similar period and eccentricity , albeit slightly lower mass and radius . at @xmath221 ,
hat - p-16 is one of the brightest stars known to host a @xmath222 planet .
we have compared hat - p-16b to the theoretical models from @xcite by interpolating the models to the solar - equivalent semi major axis of @xmath223 , the result of which can be seen overplotted in figure [ fig : exomr ] .
we find that the mass and radius of hat - p-16b are quite consistent with the 1 gyr model and conclude that hat - p-16b is most likely a h / he - dominated planet .
the radial velocity dataset used to characterize hat - p-16b consist of three different velocity sets from two different telescopes ( fies medium and high resolution spectra from the not and hires spectra from keck - i ) , requiring additional free parameters in the global fit to account for the arbitrary velocity offset between the measurements .
nevertheless , the different observations yield very similar semi - amplitude and we find an rms from the best fit model of only @xmath224@xmath151 .
the rich dataset consisting of 27 rv observations allows us to accurately determine the small non - zero eccentricity of the orbit as @xmath15 with a high significance of @xmath16 , with @xmath225 at @xmath16 and @xmath226 at @xmath227 .
the eccentricity is consistent and clearly evident even when fitting separate orbits to the three different velocity datasets .
most short period teps have eccentricities consistent with zero , as is expected because of circularization driven by tidal dissipation within the planet with a typical timescale substantially less than 1 gyr ( see @xcite ) .
in fact , the estimated circularization timescale for hat - p-16b is @xmath228 ( when the tidal dissipation parameter is assumed to be @xmath229 ) , which is much less than the estimated @xmath230 age of the system . on the other hand ,
@xcite argue that circularization timescales could be significantly higher and the planets with eccentric orbits may simply be in the process of being circularized .
@xcite argue that for multiple systems with a hot jupiter as the inner planet , secular excitation of the eccentricity by companion planets could explain the non
zero eccentricity of these systems .
thus the origin of the eccentricity of hat - p-16b remains unclear . with a stellar rotation of @xmath231 @xmath34and an impact parameter of @xmath232 ,
hat - p-16 is a favorable target for measuring the rossiter
mclaughlin ( rm ) effect .
this effect can be used to determine the alignment of the planet s orbital angular momentum vector with the stellar spin axis @xcite .
the estimated amplitude of the rm effect is @xmath233 for hat - p-16b .
hatnet operations have been funded by nasa grants nng04gn74 g , nnx08af23 g and sao ir&d grants .
work of g..b . and j. johnson were supported by the postdoctoral fellowship of the nsf astronomy and astrophysics program ( ast-0702843 and ast-0702821 , respectively ) .
gt acknowledges partial support from nasa grant nnx09af59 g .
we acknowledge partial support also from the kepler mission under nasa cooperative agreement ncc2 - 1390 ( d.w.l .
thanks the hungarian scientific research foundation ( otka ) for support through grant k-81373 . based in part 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 astrofisica de canarias .
this research has made use of keck telescope time granted through nasa ( n018hr ) .
fischer , d. a. , vogt , s. s. , marcy , g. w. , butler , r. p. , sato , b. , henry , g. w. , robinson , s. , laughlin , g. , ida , s. , toyota , e. , omiya , m. , driscoll , p. , takeda , g. , wright , j. t. , & johnson , j. a. 2007 , , 669 , 1336 latham , d. w. borucki , w. j. , koch , d. g. , brown , t. m. , buchhave , l. a. , basri , g. , batalha , n. m.,caldwell , d. a. , cochran , w. d. , dunham , e. w. , furesz , g , gautier iii , t. n. , geary , j. c. , gilliland , r. l. , howell , s. b. , et al .
2010 , arxiv:1001.0190 | we report the discovery of hat - p-16b , a transiting extrasolar planet orbiting the @xmath0 f8 dwarf gsc 2792 - 01700 , with a period @xmath1 , transit epoch @xmath2 ( bjd ) , and transit duration @xmath3d .
the host star has a mass of @xmath4@xmath5 , radius of @xmath6@xmath7 , effective temperature k , and metallicity @xmath8}}= { \ifthenelse{\equal{{ii}}{i}}{{\ensuremath{0.15\pm0.06}}}{{\ensuremath{+0.17\pm0.08}}}}$ ] .
the planetary companion has a mass of @xmath9@xmath10 , and radius of @xmath11@xmath12 yielding a mean density of @xmath13@xmath14 . comparing these observed characteristics with recent theoretical models , we find that hat - p-16b is consistent with a 1 gyr h / he - dominated gas giant planet .
hat - p-16b resides in a sparsely populated region of the mass radius diagram and has a non - zero eccentricity of @xmath15 with a significance of @xmath16 . |
the layered transition metal dichalcogenides ( ltmds ) mx@xmath7 of the group v metals ( m = v , nb , ta ; x = s , se ) and their intercalation compounds have been the subject of numerous studies on the inter - relationship between superconductivity and charge density waves ( cdw ) , both of which arise from the strong electron - phonon ( @xmath6-@xmath10 ) coupling within the layers @xcite .
parameters such as stoichiometry , polymorphism , disorder and intercalation have been extensively used to study the physical properties of these low dimensional compounds .
there is however , no satisfactory explanation for their effect on the above transitions . amongst the binary compounds , all di - selenides ( v , nb , ta ) and all polymorphs of tas@xmath7 show cdw transitions , while the 2h and 4h nb and ta compounds are also superconducting .
hnbs2 and 1t - vs@xmath7 are unusual in the former , any cdw is suppressed below the superconducting @xmath11 6.2 k due to @xmath6-@xmath6 interactions in the narrow unhybridised nb @xmath12 bands .
1t - vs@xmath7 is structurally metastable , supposedly due to reduced covalency which destabilizes the layered structure in its stoichiometric composition .
the effects of @xmath6-@xmath6 interactions are also significant in 1t - vse@xmath7 in which the normal to incommensurate cdw transition temperature increases from 110 k under pressure induced broadening of the narrow v @xmath12 bands .
it also shows a small curie - like contribution in @xmath13 from a small excess of v between the layers .
2h - nbse@xmath7 ( @xmath14 33 k ) , 2h - tas(se)@xmath7 ( @xmath14 75(122 ) k ) and 4h@xmath15-tas(se)@xmath7 ( @xmath14 22(75 ) k ) also exhibit @xmath11 7.2 k , @xmath16 .6 k and @xmath17 1 k respectively .
high pressure lowers and raises towards its ` undistorted ' values ( @xcite and references therein ) .
lattice - dynamics calculations suggest that in these materials , @xmath6-@xmath10 renormalisation effects on phonon frequencies are crucial in causing lattice instability , as well as raising @xcite .
the above transitions are absent in di - tellurides in which strong intermetallic bonding and metal atom clustering leads to more stable but distorted structures .
stoichiometry , intercalation and disorder significantly affect the cdw and superconductivity in these compounds . in 1t - tas@xmath7 a mere 50 ppm of isoelectronic nb impurities were observed to destroy the long range phase coherence of the commensurate cdw and completely suppress the associated metal - insulator ( mi ) transition @xcite .
on intercalation , the in ta compounds increases from @xmath18 1 k upto 5 k while it is lowered for nb @xcite@xcite . in self - intercalated 2h - nb@xmath19se@xmath7
the was suppressed below 2.2 k @xcite , while 2h - nbse@xmath7(eda)@xmath20 showed no superconductivity but a resistance minimum at @xmath17 25 k @xcite .
post - transition metal intercalated hnbs2 also did not show superconductivity @xcite . `
kondolike ' resistance minima at 20 k along with a cdw were observed for 2h - fe@xmath21nb(ta)se@xmath7 @xcite .
li@xmath9nbs@xmath7 ( @xmath22 ) has a complex dependence of on @xmath2 due to interplay of polymorphic changes ( 2h3r ) and electron transfer effects upon intercalation @xcite .
in fact , comparable disorder induced resistance minima at low temperatures are also seen in structurally similar graphite intercalation compounds , and explained using weak localisation and @xmath6-@xmath6 interactions @xcite .
a less studied polymorph of 2 is the 3r phase ( space group @xmath23 ) . while the inherently stoichiometric hnbs2 forms at high temperatures ( @xmath24 ) under high s vapour pressure , any excess nb ( at low s vapour pressure ) results in the metal - rich 3r phase @xcite .
the stoichiometry limit for single phase rnbs2 depends on the preparation temperature , and a minimum of @xmath25 has been reported at 650 @xcite .
samples prepared at high temperature and reported closer to stoichiometry are essentially a 2h3r mixture with properties correspondingly in between .
the 3r phase can thus be considered a self - intercalated phase with the excess nb in octahedral interlayer vacancy sites @xcite .
while no superconductivity has been observed in rnbs2 down to 1.7 k , there is also no evidence of a cdw which could depress the same .
conduction in ltmds is essentially intralayer .
hence the role of even small amounts of interlayer metal in suppressing superconductivity between different polymorphs needs to be investigated to understand better the effects of polymorphism , stoichiometry , disorder and dimensionality on the properties of these compounds .
we have prepared hnbs2 , rnbs2 ( @xmath26 ) and 3r-2 ( @xmath27 ) .
hnbs2 has @xmath17 6.2 k , as expected .
the metal - rich rnbs2 shows no superconductivity ; rather , a resistance minimum @xmath17 20 k. non - magnetic ga - intercalated 3r phases exhibit similar low temperature resistance minima ( 20 k @xmath28 60 k )
. thermopower also shows anomalies around .
we observe a clear correspondence of and extent of localisation to the cation - to - anion ratio and the preparation temperature , to the extent of structural distortion from the ideal layered host .
the compounds were prepared by heating stoichiometric mixtures of the pure elements ( ga 99.999% , nb 99.7% , s 99.999% pure , cerac ) in evacuated quartz ampoules .
hnbs2 was obtained using 10% excess sulfur .
chemical analysis by icp - aes and edx techniques gave fe @xmath18 240 ppm , mn @xmath18 15 ppm and cr @xmath18 10 ppm .
pressed polycrystalline pellets were sintered at different temperatures
950 for the binary compound , and 850 ( lt phase ) & 1100 ( ht phase ) for ga - intercalated compounds and quenched to room temperature .
single crystal flakes were obtained by vapour transport ( 950 900 ) for hnbs2 and 3r - ga@xmath29nbs@xmath7 .
some sulfur rejection was observed at high temperature in all cases , indicating a metal rich composition .
the compounds were characterised by room temperature powder xrd .
we compare the xrd patterns of hnbs2 , rnbs2 and 3r - ga@xmath292(lt ) in fig .
[ fig1 ] . the 2h phase ( @xmath30 = 3.32 , @xmath31 = 11.97 ) had broad ( @xmath32 ) lines indicating well known disorder in the stacking of the weakly coupled antiparallel layers @xcite .
the metal rich 3r phases on the other hand had sharp lines , the ht phases more ordered than the lt ones .
the excess metal ( nb / ga ) occupying the octahedral interlayer vacancies @xcite pins the layers together and prevents stacking faults .
rnbs2 ( @xmath30 = 3.32 , @xmath31 = 17.88 ) and 3r-2 ( @xmath2 = .1 , .25 , .33(lt phase ) ) ( @xmath30 = 3.33@xmath33 , @xmath31 = 17.90@xmath33 ) have similar lattice parameters . however , the ga - intercalated phases exhibit superlattice reflections proportional in number and intensity to ga content and the preparation temperature , indicating progressive distortion .
the .33 ga ( ht ) phase ( @xmath30 = 7.19 , @xmath31 = 17.30 ) is structurally very different , with about 10% reduction in volume and a doubling of the @xmath30-axis .
the superlattice lines order into a doubled @xmath30-axis and show much lower @xmath34 .
thus with increasing cation concentration and preparation temperature , the trigonal prismatic close - packed layers transform towards a distorted octahedral coordination of nb by staggered sulfur layers .
progressive metal clustering and vacancy formation within the nb layers finally leads to a well ordered phase with a different structure .
this is similar to cation - rich nb@xmath7se@xmath35 , nb@xmath35s@xmath36 , cu@xmath372 , which have strong nb - nb bonds giving rise to zig - zag chains / clusters .
these changes are also reflected in the electronic properties .
the exact nature of the nb atom clustering in our compounds would require a more detailed structural study .
the electronic properties were studied by 4-probe d.c .- resistance ( 4.2300 k ) , thermopower ( @xmath38 ) ( 14300 k ) and magnetic susceptibility ( @xmath39)(80300 k ) measurements .
the d.c .- resistance of pellets as well as single crystal flakes was measured in van der pauw geometry @xcite using ag - paste contacts .
the @xmath38 of pellets pressed between cu stubs was measured in the differential mode .
absolute @xmath40 was determined by calibrating with pb @xcite , and correcting for the cu leads .
the @xmath39 was measured on compacted powders under a 9.7 koe field in a vibrating sample magnetometer ( vsm ) @xcite .
the results are summarised in table [ tbl1 ] .
the samples showing resistance minimum obviously do not follow mattheissen s rule for scattering from dilute impurities under the quasi - free electron approximation .
therefore , we consider the @xmath41 as characteristic of the residual resistance , and in subsequent discussions the residual resistance ratio ( rrr ) is defined by @xmath42 ; while , for the superconducting samples the rrr = @xmath43 , where @xmath44 is the value of the resistivity at .
the resistivity results on single crystal flakes and polycrystalline pellets of both 2h and 3r polymorphs are shown in fig .
[ fig2 ] and [ fig3 ] .
grain boundary scattering in the pellets and the anisotropy of conduction in flakes would prevent their inter - comparison .
figure [ fig3 ] therefore shows the normalised resistance behaviour after subtracting low temperature residual resistances .
the 2h phase @xmath45 k and its large rrr of @xmath46 for in - plane conduction in the flakes compare well with earlier studies @xcite .
thus inspite of a considerable degree of stacking disorder , the essentially in - plane conduction and also the superconductivity is not affected in 2h polymorphs .
the 3r phases on the other hand have no stacking disorder as inferred from their sharp xrd pattern , but gave a resistance minimum around and above 20 k. the temperature and extent of the minimum increases with the amount of intercalate atoms .
the overall conduction behaviour of the 3r phases showing a resistance minimum is also significantly different .
firstly , we notice a large increase in residual resistance upon intercalation ( table [ tbl1 ] ) .
for example , the .1 ga crystal flake measured along the plane has a rrr of only 1.65 , although its resistivity is actually much lower than superconducting hnbs2 ( rrr @xmath17 69 ) .
the scattering of electrons at higher temperatures is also significantly different .
we observe in the 3r phases a comparatively faster drop in resistance of flakes as well as pellets and also a larger contribution of t@xmath47 - term above their resistance minimum ( fig .
[ fig3 ] ) .
the reduction in saturation effects in the high temperature conduction of the polycrystalline rnbs2 phase probably indicates an increase in its isotropy of conduction . a proper explanation of this difference would require a detailed investigation of the effects of non - stoichiometry and polymorphic changes on the band structure and @xmath6-@xmath10 interaction effects .
the observed differences , however , can not be simply related to doping of carriers since extra nb or ga atoms have similar effects .
we presently conclude that in the intrinsically non - stoichiometric rnbs2 phases the intralayer scattering potentials are significantly increased , leading to large residual resistance , increased @xmath6-@xmath6 scattering effects , and consequently low temperature localisation of carriers .
the @xmath38 and @xmath39 results on polycrystalline 2h and 3r phases are shown in fig .
the @xmath38 of 3r phases of self intercalated and .1 ga intercalated phase are very similar but differ markedly from that of hnbs2 .
for the former a large , fairly constant , negative @xmath38 at high temperature changing rapidly towards positive below 100 k can be seen .
for @xmath39 we have subtracted a small saturated contribution which was observed in the @xmath48 behaviour at low fields ( @xmath49 koe ) .
our data differ from previous studies , as we observe a continuous reduction in @xmath39 instead of a slight increase on cooling @xcite .
these are the typical dependences observed in compounds showing cdw instabilities .
we have earlier mentioned that group v tmd compounds of other than the present studies , layered 1t - vse@xmath7 , 2h - nbse@xmath7 and various polymorphs of tas(se)@xmath7 , show varying degrees of cdw formation .
the associated anomalies in their transport and magnetic properties are much weaker than in 1d cdw structures such as nbs@xmath35 , tas@xmath35 etc .
for example , except for 1t - tas@xmath7 which has a cdw lead mi transition , the usual increase in resistance at , is barely observable and only a steep increase in its slope is observed below .
recent optical studies on 2h - tase@xmath7 confirm the absence of any abrupt formation of a charge - excitation gap at @xcite .
the steep increase in the resistance slope below is found to be a consequence of freezing - out of scattering channels since the drude scattering peak in @xmath50 ( @xmath51 0 ) becomes narrower below . in the light of the above observations , we can not rule out cdw - correlations in our 3r phases , more so since superstructural distortions are clearly observed in intercalated phases .
the thermopower and magnetic susceptibility variation also suggest some non - magnetic electronic correlations developing on cooling . a careful structural study of the 3r polymorphs of 2 at low temperature is therefore required to ascertain the relation of the superlattice with the observed transport and magnetic properties . in fig . [ fig5 ]
we show a scaled plot of resistance behaviour with temperature for the compounds of our study between @xmath4 .
there is seen to be a close similarity for the crystal flakes and pellets of compounds having different intercalate ( ga / nb ) concentration and showing vastly different resistivity values .
it clearly indicates the role of defects in these compounds since @xmath41 and show a systematic increase with the intercalate concentration .
the resistance minimum may be sought to be explained by kondo scattering from dilute magnetic impurities such as fe or mn @xcite . however , as stated earlier , the maximum concentration of such impurities in our compounds ( @xmath52 ppm ) is too low for significant spin - flip scattering contribution .
neither do we observe the expected logarithmic dependence of resistivity upon temperature for @xmath53 .
the temperature of the kondo minimum plotted as @xmath54 vs @xmath55 is in fact only weakly dependent upon the extrinsic impurity concentration .
instead , we find a strong dependence of on intercalate concentration which scales as shown in fig .
moreover , our @xmath38 is smoothly varying in the temperature range of resistance minimum instead of the broad , large maxima expected in kondo systems .
the pauli like @xmath39 of our compounds at high temperature is also in contrast to the curie - weiss like behaviour of kondo systems for @xmath56 .
thus kondo scattering can not be invoked to explain the observed behaviour of our compounds .
on the other hand , disorder and coulomb interaction effects on the scattering of electrons in a narrow anisotropic band of these compounds are expected to be quite important . for @xmath57 , ( @xmath58 fermi momentum , @xmath59 elastic mean free path ) ,
the quasi - classical treatment of elastic scattering leading to mattheissen s rule in dilute alloys breaks down .
the rise in resistance on cooling can be understood in terms of either increasing interference corrections for the elastically scattered waves from the static disorder or subtle changes in the excitation spectrum at the fermi level caused by @xmath6-@xmath6 interaction effects in the presence of ionic disorder , as first suggested by altshuler and aronov @xcite .
the relative importance of these quantum correlations and their temperature dependence depends on dimensionality and the details of band structure of the conduction electrons .
the plots of @xmath60 at low temperature are shown in fig .
[ fig6 ] , where @xmath61 is the value of conductivity at 0 k obtained by extrapolating below 4.2 k. our preliminary results show a nearly @xmath62 dependence of the conductivity far below its maximum , the coefficient of the @xmath62 term depending upon the stoichiometry and structural details .
the observed behaviour suggests quantum corrections to the dos due to long range coulomb interactions between conduction electrons @xcite .
the metal rich .33 ga ( ht ) phase also exhibits similar behaviour , though with a much larger slope . here the localisation effects are dramatically increased due to the clustered nature of the metallic lattice .
we mention here an interesting possibility which may increase the long range coulomb interaction effects between charge carriers .
the theory of altshuler and aronov gives a depression in dos ( ) @xcite .
the present compounds are prone to cdw correlations which also reduce dos ( ) . however , for cdw the @xmath6-@xmath6 interactions are very special since they require @xmath6-@xmath10 coupling across a nesting wave vector .
therefore , the role of incipient cdw fluctuations in the presence of disorder to give the observed resistance minimum behaviour in these compounds should be seriously explored .
to summarise , we have observed low temperature resistance minima in non - superconducting metal - rich 2 derivatives
non - stoichiometric rnbs2 ( @xmath17 20 k ) as well as 3r-2 ( 20 k @xmath28 60 k ) .
a common physical origin of these minima is evident in the scaling of resistance as @xmath3 between @xmath4 for different phases whose resistivity differs by an order of magnitude .
the low temperature behaviour of these non - magnetic compounds can not be explained by kondo scattering effects .
instead , we find that the conductivity varies as @xmath62 below its maximum ( below ( @xmath41 ) ) .
we , therefore , propose that a possible cause for the observed behaviour in these narrow band anisotropic systems is a correction to the dos ( ) due to @xmath6-@xmath6 interaction effects in the presence of ionic disorder .
* acknowledgements : * we are grateful to prof .
deepak kumar for illuminating discussions .
an thanks csir , new delhi for financial support during his doctoral work .
99 wilson j.a .
, di salvo f.j . and mahajan s. , _ adv .
_ , * 24 * ( 1975 ) 117 .
friend r.h .
and yoffe a.d .
_ , * 36 * ( 1987 ) 1 .
nishio y. , , _ j. phys .
* 63 * ( 1994 ) 156 ; nishio y. , _ j. phys
* 63 * ( 1994 ) 223 .
zwick f. , , _ phys .
_ , * 81 * ( 1998 ) 1058 .
revelinski e. , spiering g.a . and beernstein d.j .
_ j. phys .
chem . solids _ , * 26 * ( 1965 ) 1029 .
di salvo f.j .
, , _ j. chem .
_ , * 59 * ( 1973 ) 1922 .
meyer s.f . ,
, _ j. chem .
_ , * 62 * ( 1975 ) 4411 .
karnezos n. , welsh l.b . and
shafer m.w .
b _ , * 11 * ( 1974 ) 1808 . gamble f.r . and
geballe t.h . , _ treatise on solid state chemistry , inclusion compounds _ , * 3 * plenum , new york ( 1976 ) ch .
3 . dai z. , , _ phys .
b _ , * 48 * ( 1993 ) 14543 . mcewen c.s . and sienko m.j . ,
_ rev . de chim . min .
_ , * 19 * ( 1982 ) 309 .
piraux l. , , _ solid state commun .
_ , * 82 * ( 1992 ) 371 .
fisher w.g . and
sienko m.j . , _ inorg . chem .
_ , * 19 * ( 1980 ) 39 .
tatsuki k. , wakihara m. and taniguchi m. , _ j. less common metals _ , * 68 * ( 1979 ) 183 .
powell d.r . and jacobson r.a . , _
j. solid state chem .
_ , * 37 * ( 1981 ) 140 .
jellinek f. , _ arkiv kemi _ ,
* 20 * 1963 447 .
van der pauw l.j . , _
phillips res . repts .
_ , * 13 * ( 1958 ) 1 .
roberts r.b . , _ phil . mag .
_ , * 36 * ( 1977 ) 91 .
niazi a. , poddar p. and rastogi a.k . ,
_ current science _ , * 79 * ( 2000 ) 99 .
naito m. and tanaka s. , _ j. phys .
* 51 * ( 1982 ) 219 , 228 .
hamaue y. and aoki r. , _ j. phys
* 55 * ( 1986 ) 1327 .
bowmeester h.j.m .
, , _ phys .
b _ , * 43 * ( 1991 ) 9431 .
vescoli v. , , _ phys .
_ , * 81 * ( 1998 ) 453 .
kondo j. , _ solid state physics _ , * 23 * , ed .
f. seitz , d. turnbull and h. ehrenresch , academic press , new york ( 1969 ) 183 .
heeger a.j . ,
_ solid state physics _ ,
* 23 * , ed .
f. seitz , d. turnbull and h. ehrenresch , academic press , new york ( 1969 ) 283 .
maple m.b . , delong l.e . and sales b.c . , _
handbook on the physics and chemistry of rare earths _ ed .
gschneider , jr . and l. eyring , north holland publishing co. ( 1978 ) 797 .
altshuler b.l .
and aronov a.g . ,
_ electron - electron interactions in disordered systems _ , ed
efros and m. pollak , elsevier science publishers b.v . , amsterdam ( 1985 ) 1 . & & & + & @xmath63 & @xmath64 & t@xmath65 & r.r.r .
& ( @xmath66v / k ) & ( @xmath67 + & & ( k ) & ( @xmath68 ) & & emu / mole ) + 2h - nbs@xmath7 @xmath69 & 0.241 & 0.0035 @xmath70 & & 68.83 & & + 2h - nbs@xmath7 & 2.02 & 0.80 @xmath70 & & 2.52 & @xmath712.092 & @xmath72 200 + 3r - nb@xmath8s@xmath7 & 3.23 & 2.296 & 20 & 1.408 & @xmath714.114 & @xmath72 100 + ga@xmath29nbs@xmath7 @xmath69 & 0.178 & 0.1075 & 20 & 1.656 & & + ga@xmath29nbs@xmath7 ( lt ) & 1.96 & 1.380 & 20 & 1.420 & @xmath714.501 & @xmath72 100 + ga@xmath73nbs@xmath7 ( lt ) & 2.29 & 1.874 & 28 & 1.222 & @xmath710.199 & @xmath72 30 + ga@xmath73nbs@xmath7 ( ht ) & 4.32 & 3.575 & 32 & 1.207 & 1.305 & @xmath72 35 + ga@xmath37nbs@xmath7 ( lt ) & 4.74 & 3.950 & 41 & 1.200 & 12.667 & @xmath72 30 + ga@xmath37nbs@xmath7 ( ht ) & 12.00 & 10.60 & 58 & 1.128 & 3.012 &
@xmath72 140 + 2 ( lt ) .
the ( 002 ) and ( 003 ) peaks were very strong due to preferred orientation and are truncated to magnify the rest .
the broad peaks of the 2h phase , ( a ) , are in contrast to the sharp lines of the metal - rich 3r phases , ( b ) and ( c ) , as discussed in the text .
emerging superlattice lines ( @xmath74 ) can be seen in ( c).,width=566 ] and ( b ) .
@xmath39 , in hnbs2 , rnbs2 and 3r-2 ( lt phases ) , @xmath75 . in ( a ) data from @xcite , ( @xmath76 ) , compares well with our rnbs2 and .1 ga phases . in ( b )
the inset shows data from @xcite , ( 1 ) hnbs2 and ( 2 ) 3r - nb@xmath77s@xmath7 .
the diamagnetic contribution of core - electrons is expected to be @xmath78 emu / mole.,width=566 ] for the above phases .
the figure legend is same as fig .
the solid lines are guides to the eye .
all curves scale well as @xmath62 below @xmath79 .
the slope of the curves depends on the stoichiometry details discussed in the text.,width=566 ] | we report the structural and electron transport properties of rnbs2 ( @xmath0 ) and 3r-2 ( @xmath1 ) prepared as polycrystalline pellets as well as single crystals grown by vapour transport .
we observe a resistance minimum in these compounds between 2060 k , with the proportional to @xmath2 .
the resistance scales as @xmath3 between @xmath4 for different phases with @xmath5 whose resistivity differs by an order of magnitude .
powder x - ray diffraction ( xrd ) also shows progressively increasing intensity of superlattice lines with cation concentration .
the thermopower changes sign around the resistance minimum .
the explanation of the resistance minimum and the simultaneous rapid suppression of superconductivity is sought in @xmath6-@xmath6 scattering effects in the presence of cation disorder in these narrow band anisotropic materials .
2nbs@xmath7 hnbs22h - nbs@xmath7 rnbs23r - nb@xmath8s@xmath7 2ga@xmath9nbs@xmath7 * pacs numbers : * + 72.80.ga transition metal compounds .
+ 71.45.lr charge density wave systems .
+ 74.62.bf effects of material synthesis , crystal structure , and chemical composition . + * running head : * low temperature resistance minimum , |
the colour dipole model has proven to be very successful in describing a wide variety of small-@xmath4 inclusive and diffractive processes at hera .
this talk @xcite is based on work done on this subject in collaboration with h. kowalski and l. motyka @xcite .
r0.5 [ fig : diagram ] the amplitude for an exclusive diffractive process , @xmath5 , shown in fig .
[ fig : diagram ] , such as vector meson production , @xmath6 , or deeply virtual compton scattering ( dvcs ) , @xmath7 , can be expressed as @xmath8\cdot\boldsymbol{\delta}}\;\frac{\mathrm{d}\sigma_{q\bar q}}{\mathrm{d}^2\boldsymbol{b}},\end{gathered}\ ] ] up to corrections from the real part of the amplitude and from skewedness ( @xmath9 ) . here
, @xmath10 is the fraction of the photon s light - cone momentum carried by the quark , @xmath11 is the transverse size of the @xmath12 dipole , while @xmath13 is the impact parameter , that is , @xmath14 is the transverse distance from the centre of the proton to the centre - of - mass of the @xmath12 dipole ; see fig .
[ fig : diagram ] .
the transverse momentum lost by the outgoing proton , @xmath15 , is the fourier conjugate variable to the impact parameter @xmath13 , and @xmath16 .
the forward overlap function between the initial - state photon wave function and the final - state vector meson or photon wave function in is denoted @xmath17 , while the factor @xmath18 $ ] in originates from the non - forward wave functions .
taking the imaginary part of the forward scattering amplitude immediately gives the formula for the total @xmath19 cross section : @xmath20 the dipole picture therefore provides a unified description of both exclusive diffractive processes and inclusive dis at small @xmath4 .
the unknown quantity common to and is the @xmath14-dependent dipole
proton cross section , @xmath21 where @xmath22 is the imaginary part of the dipole
proton scattering amplitude , which can vary between zero and one , where @xmath23 is the unitarity ( `` black disc '' ) limit .
the scattering amplitude @xmath22 is generally parameterised according to some theoretically motivated functional form , with the parameters fitted to data .
we will consider two such parameterisations .
the first parameterisation is based on lo dglap evolution of an initial gluon density , @xmath24 , with a gaussian @xmath14 dependence , @xmath25 .
we refer to this parameterisation as the `` b - sat '' model @xcite : @xmath26 where the scale @xmath27 , @xmath28 gev@xmath29 is fixed from the @xmath30-slope of exclusive @xmath1 photoproduction , and the other three parameters ( @xmath31 , @xmath32 , @xmath33 ) are fitted to zeus @xmath34 data with @xmath35 and @xmath36 $ ] gev@xmath37 .
the second parameterisation is a modified version of the colour glass condensate ( cgc ) dipole model of iancu , itakura and munier : @xmath38 where a gaussian impact parameter dependence is introduced into the saturation scale : @xmath39^{\frac{1}{2\gamma_s}}.\ ] ] we refer to the parameterisation given by and as the `` b - cgc '' model @xcite .
the parameter @xmath40 gev@xmath29 is fixed from the @xmath30-slope of exclusive @xmath1 photoproduction , while the other four parameters ( @xmath41 , @xmath42 , @xmath43 , @xmath44 ) are fitted to zeus @xmath34 data with @xmath35 and @xmath45 $ ] gev@xmath37 .
the optimum fitted value of the anomalous dimension at the saturation scale , @xmath46 , is close to the value of @xmath47 determined from numerical solution of the balitsky
kovchegov equation . however , the value of @xmath48 obtained from the fit is lower than the perturbatively calculated value of @xmath49 , and suggests that the saturation scale comprises significant non - perturbative dynamics .
r0.4 it is customary to define a saturation scale @xmath50 , that is , the momentum scale at which the dipole proton scattering amplitude @xmath22 becomes sizable such that non - linear effects start to become important .
there is no unique definition of @xmath50 and various choices are used in the literature .
we define the saturation scale @xmath51 , where the saturation radius @xmath52 is the dipole size where the scattering amplitude @xmath53 note that we use lower - case @xmath54 and upper - case @xmath55 to distinguish between the two scales defined by and respectively .
the model - independent saturation scale @xmath50 is shown in fig .
[ fig : q2sx ] : it is generally less than @xmath56 gev@xmath37 in the hera kinematic regime for the most relevant impact parameters @xmath57@xmath58 gev@xmath59 .
l0.4 [ fig : apom ] a wealth of hera data exists on exclusive diffractive vector meson ( @xmath2 , @xmath1 , @xmath60 , @xmath61 ) production and dvcs .
it is therefore a significant challenge for an essentially parameter - free model to describe all aspects of the @xmath62 , @xmath63 and @xmath30 dependence .
extensive comparison of the b - sat and b - cgc model predictions with hera data on exclusive processes has been made in refs .
@xcite . in general , both models describe almost all features of the available data . here
we focus on only two aspects of the data which differentiate the models . in fig .
[ fig : apom ] we show the effective pomeron trajectory @xmath64 vs. @xmath65 , where @xmath64 is determined by fitting @xmath66}$ ] .
the b - cgc model gives a better description of @xmath67 , where @xmath68 .
this suggests that the b - cgc dipole cross section better models the interplay between the @xmath4 and @xmath14 dependence . in fig .
[ fig : crossw ] we show the @xmath63 dependence of the total cross sections for exclusive @xmath1 , @xmath60 and @xmath61 meson production .
the @xmath63 dependence of @xmath1 photoproduction is much better described by the b - sat model than by the b - cgc model .
this difference can be traced to relatively small differences in the form of the dipole cross sections . vs. @xmath63 for exclusive @xmath1 , @xmath60 and @xmath61 meson production
compared to predictions from the b - sat and b - cgc models @xcite.,title="fig:",scaledwidth=33.0% ] vs. @xmath63 for exclusive @xmath1 , @xmath60 and @xmath61 meson production compared to predictions from the b - sat and b - cgc models @xcite.,title="fig:",scaledwidth=33.0% ] vs. @xmath63 for exclusive @xmath1 , @xmath60 and @xmath61 meson production compared to predictions from the b - sat and b - cgc models @xcite.,title="fig:",scaledwidth=33.0% ] [ fig : crossw ]
r0.45 @xmath69 ( nb ) & @xmath70 ( nb ) + tevatron & @xmath71 & @xmath72 + lhc & @xmath73 & @xmath74 + + @xmath75 & @xmath69 ( pb ) & @xmath70 ( pb ) + tevatron & @xmath76 & @xmath77 + lhc & @xmath78 & @xmath79 + + @xmath3 & @xmath69 ( fb ) & @xmath70 ( fb ) + tevatron & @xmath80 & @xmath81 + lhc & @xmath82 & @xmath83 + the equivalent - photon approximation allows predictions to be made for the rapidity distributions of exclusive photoproduced @xmath1 and @xmath2 mesons , and @xmath3 bosons , expected at the tevatron and lhc .
a given rapidity @xmath84 corresponds to a photon energy @xmath85 .
the hadron
hadron cross sections are obtained from the photon
hadron cross sections by multiplying by the flux @xmath86 of quasi - real photons .
we neglect the possible interference between photon pomeron and pomeron photon fusion .
we also neglect absorptive corrections from soft rescattering .
both these effects should be largely washed out for cross sections integrated over final state momenta , in which case the rapidity gap survival factor is expected to be @xmath87@xmath88 . in table
[ tab : results ] we present @xmath89 at @xmath90 and the total cross sections integrated over rapidity .
the @xmath1 and @xmath2 predictions have been scaled by factors 1.08 and 2.96 , respectively , in order to give the best agreement with the existing hera data . h. kowalski , l. motyka and g. watt , phys . rev .
d * 74 * ( 2006 ) 074016 [ arxiv : hep - ph/0606272 ] . g. watt and h. kowalski , phys .
d * 78 * ( 2008 ) 014016 [ arxiv:0712.2670 [ hep - ph ] ] .
l. motyka and g. watt , phys .
d * 78 * ( 2008 ) 014023 [ arxiv:0805.2113 [ hep - ph ] ] . | we discuss two different models for the impact parameter dependent dipole cross section : one based on dglap evolution and the other inspired by the balitsky kovchegov equation .
the parameters are determined from fits to data on the total @xmath0 cross section measured at hera .
the impact parameter dependent saturation scale is extracted .
predictions are then confronted with hera data on exclusive diffractive vector meson production and deeply virtual compton scattering .
finally , predictions are given for the cross sections of exclusive photoproduced @xmath1 and @xmath2 mesons , and @xmath3 bosons , expected at the tevatron and lhc . |
during the last few years , agile and fermi observations of the galactic plane have indicated the existence of a possible population of unidentified transient mev - gev sources characterized by flares lasting no more than a few days at most e.g. @xcite , @xcite .
notably , no blazar - like counterparts are known within their error boxes so they could represent a completely new class of galactic high energy transients .
the task of identifying their counterparts at lower energies is very challenging , mainly because of their fast transient nature and large positional uncertainty ( e.g radii typically from 10 arcmin to 0.5 degrees ) .
however these difficulties are compensate by the fact that it is most probably within this group of gamma - ray sources that peculiar objects ( or even a new class of objects ) could emerge , leading to novel and unexpected discoveries .
the ibis instrument onboard the integral satellite is particularly suited to search for reliable best candidate counterparts in the energy range 20100 kev thanks to i ) a large field of view ( 30@xmath1@xmath230@xmath1 ) which ensure a total coverage of the gamma - ray error box ii ) a good angular resolution ( 12 arcminutes ) which is mandatory to disentangle the hard x - ray emission of different sources in crowded fields such as those on the galactic plane iii ) a good sensitivity above 20 kev ( @xmath3 10 mcrab during a typical ibis observation lasting @xmath3 2,000 seconds ) . in particular ,
recent integral / ibis results @xcite , @xcite , @xcite provided intriguing hints that best candidate counterparts could be found among the members of the supergiant fast x - ray transients population ( sfxts ) .
sfxts are a new subclass of high mass x - ray binaries ( hmxbs ) unveiled in the last few years mainly thanks to integral observations of the galactic plane @xcite , @xcite , @xcite .
they host a massive ob supergiant star as identified by optical spectroscopy , the compact object is generally assumed to be a neutron star because of the broad band x - ray spectral shape ( 0.2100 kev ) strongly resembling those of accreting x - ray pulsars in classical hmxbs .
as support to this assumption , x - ray pulsations have been firmly detected in several sfxts e.g. @xcite . as for their x - ray behavior , sfxts
are intriguingly characterized by remarkable fast x - ray flares lasting from few hours to no longer than a few days at most and reaching typical peak x - ray luminosities of l@xmath4 @xmath3 10@xmath5 erg s@xmath6 .
the duty cycles of activity are very low ( 0.1%3% ) , conversely sfxts spend most of their time in a low level x - ray activity with typical l@xmath7 @xmath3 10@xmath810@xmath9 erg s@xmath6 , rarely they are also observed in x - ray quiescence at much lower x - ray luminosities ( l@xmath7 @xmath10 10@xmath11 erg s@xmath6 ) .
the typical dynamic range of classical sfxt spans three to five order of magnitude , however some systems show a lower value of the order of @xmath310@xmath12 and so they have been named as intermediate sfxts @xcite , @xcite .
the extreme case of fast variability and high dynamical range characterizing sfxts is at odds with the behavior of their historical parent population of classical wind - fed supergiant hxmbs which , since more than 40 years of x - ray astronomy , are known to be bright and persistent x - ray sources always detectable around l@xmath4 @xmath3 10@xmath5 erg s@xmath6 .
although sfxt hunting is not an easy task , in a very few years @xmath3 10 firm systems have been reported in the literature ( see list in @xcite ) plus a similar number of candidates : sfxts could represent a major population of transient hmxbs hidden on the galactic plane of our galaxy . the physical mechanism driving their peculiar fast x - ray transient behaviour is unclear and still highly debated .
several models have been proposed in the literature ( see @xcite for a review ) .
it is worth noting that in the last few years observations performed by fermi , agile and cherenkov telescopes have provided secure evidences that particles can be efficiently accelerated to very high energies in some hmxbs .
in fact , a number of classical hmxbs have been firmly detected from mev to tev energies as persistent and variable sources e.g. @xcite .
in addition , the two microquasar hmxb cygnus x-1 and cygnus x-3 have been detected as transient mev - gev emitters whose flaring activity lasted typically 1 - 2 days @xcite , @xcite , @xcite . similarly to this general and emerging class of gamma - ray hmxbs , in principle sfxts could be able to produce photons up to mev - tev energies since they have the same `` ingredients '' in term of a neutron star compact object as well as a bright and massive ob star which could act as a source of seed photons ( for the inverse compton emission ) and target nuclei ( for hadronic interactions ) .
in particular , the high energy emission from sfxts should be in the form of fast flares ( from few hours to few days duration ) and the duty cycle of activity should be very small , this would make their high energy detection very challenging . despite this drawback
, some observational evidences have been recently reported in the literature on sfxts as best candidate counterparts of unidentified transient mev - gev sources located on the galactic plane @xcite , @xcite , @xcite .
these evidences are merely based on intriguing hints such as a spatial correlation and a common transient behaviour on similar , though as yet not simultaneous , short time scales .
this scenario is also supported from an energetic standpoint by a theoretical model based in the microquasar accretion / jet framework @xcite .
the so far proposed associations represent an important first step towards obtaining reliable candidates on which to concentrate further efforts in order to obtain quantitative proofs for a real physical association . in this respect , so far , the best test case is represented by the proposed association between the two sources igr j17354@xmath03255 and agl j1734@xmath03310 .
each data point represents the average flux during one pointing ( scw ) lasting 2,000 seconds . ]
igr j17354@xmath03255 is a hard x - ray transient discovered by integral in 2006 during an outburst having average flux of @xmath3 2.1@xmath210@xmath13 erg @xmath14 s@xmath6 ( 2060 kev ) and unconstrained duration ( @xcite , @xcite .
subsequent studies with the swift satellite showed that its broad band x - ray spectrum ( 0.2100 kev ) is characterized by a spectral shape very similar to that of accreting x - ray pulsars in hmxbs , in addition a periodicity of @xmath3 8.4 days was unveiled in the swift / bat data @xcite . in the soft x - ray domain ( 0.210 kev ) ,
the source has been observed by swift / xrt @xcite , chandra @xcite and xmm @xcite .
specifically , during two observations ( march 2008 with xrt and march 2011 with xmm ) the source was not detected leading to an inferred 3@xmath15 upper limit of 7@xmath210@xmath16 erg @xmath14 s@xmath6 , conversely during two other observations ( february 2009 with chandra and april 2009 with xrt ) the source was indeed detected and it was strongly variable as well , being the average flux and the peak flux equal to @xmath3 1.3@xmath210@xmath17 erg @xmath14 s@xmath6 and @xmath3 9@xmath210@xmath17 erg @xmath14 s@xmath6 , respectively .
in addition , chandra provided a very accurate positioning which allowed to pinpoint a single bright 2mass infrared source as counterpart @xcite .
subsequent infrared spectroscopic performed by @xcite unveiled its nature of supergiant star with spectral type o9.5iab .
the temporal x - ray behavior of igr j17354@xmath03255 above 20 kev , which is crucial to allow a proper and firm classification as classical persistent supergiant hmxb or alternatively as a sfxt , is largely unknown .
we performed a detailed temporal study with integral in the energy band 1860 kev @xcite .
as result , we found that igr j17354@xmath03255 is a weak persistent hard x - ray source spending a major fraction of the time in a out - of - outburst state with an average 1860 kev flux of @xmath3 1.4@xmath210@xmath17 erg @xmath14 s@xmath6 .
this is occasionally interspersed with fast hard x - ray flares having duration in the range 0.560 hours , a total of 16 hard x - ray flares have been detected by integral / ibis over a total exposure of @xmath3 115 days though not in sequence .
1 shows an example of a fast flare lasting only half an hour and detected with a significance of about 6 @xmath15 ( 1860 kev ) .
the dynamic range of the source in the hard x - ray band ( 1860 kev ) is as high as 200 , it is even higher ( @xmath181,250 ) in the softer x - ray band ( 0.210 kev ) .
based on the above findings , igr j17354@xmath03255 can be classified as a member of the sfxt population . in the framework of the clumpy wind scenario ,
the fast x - ray flares could be explained as due to accretion onto the compact object of dense clumps of material in the highly structured wind of the supergiant companion donor star .
conversely , during the weak persistent out - of - outburst x - ray state , accretion is likely still at work although at much lower rate , e.g. the compact object is likely accreting from the much less dense background wind .
this is supported by the fact that the 1860 kev spectral shape , as measured by integral / ibis , is identical during both the flaring and out - of - outburst x - ray states , i.e. a power law shape with @[email protected] .
3255 within a set window of 3 days . ]
our investigation of the integral / ibis long - term light curve using the lomb - scargle periodogram method strongly confirmed the 8.4 days orbital period .
2 shows the corresponding phase folded light curve where it is evident a smooth orbital modulation of the flux ; it peaks during the periastron passage and becomes consistent with zero around apastron .
the red crosses indicate the 16 outbursts detected by integral / ibis within these orbital ephemeris . as clearly evident
their occurrence is consistent with the region of orbital phase around periastron .
we note that the shape of the orbital profile is rather smooth and appears to be dominated by lower level x - ray emission rather than by the bright x - ray outbursts . to test this assumption
we employed the recurrence analysis technique which searches for periastron detections by summing the x - ray emission during each periastron passage within a set window of 3 days ( periastron @xmath20 1.5 days ) .
the source could then be detected even though a significant detection is not achieved in the individual ibis pointing lasting 2,000 seconds .
the same process was performed for each apastron passage and the distributions compared .
3 shows the recurrence analysis results , there is a clear excess in detections above 3@xmath15 during the periastron passages , corresponding to detectable emission on about 26% of periastron passages covered by the data set .
this value is taken as a lower limit as there may still be emission that is occurring during other periastron passages that is below the sensitivity of ibis .
on the contrary no detections above 3@xmath15 are recorded during apastron passages .
our recurrence analysis suggests that the 16 individual outbursts detected by ibis / isgri can not explain the smooth shape seen in the phase folded light curve . in fact , assuming a source distance of 8.5 kpc ( the distance is still unknown and the source is located in the direction of the galactic center ) these outbursts all have x - ray luminosities of the order of 10@xmath5 erg s@xmath6 and so represent the most luminous x - ray outburst events .
hence we would not expect these events to define the orbital emission profile over the extent of these long baseline observations .
instead we attribute the shape to lower level x - ray emission that is below the instrumental sensitivity of ibis in an individual scw ( i.e. @xmath3 10 mcrab ) .
however when the whole data set , covering about 300 orbital cycles of 8.4 days , is folded this emission sums to a significant detection and reveals the smooth profile shown in fig .
2 . this emission could be the superposition of many lower intensity x - ray flares at luminosity values of @xmath3 10@xmath810@xmath9 erg s@xmath6 .
interestingly , the sfxt igr j17354@xmath03255 is located in the sky region of two unidentified gamma - ray sources : agl j1734@xmath03310 and 3eg j1734@xmath03232 ( see fig .
4 ) : * agl j1734@xmath03310 was discovered by the agile gamma ray satellite on 2009 april 14 during a flare lasting only 1 day and detected with a significance of 4.8@xmath15 at e@xmath18100 mev @xcite .
after its discovery , extensive searches for further flaring gamma - ray emission have been carried out by the agile team @xcite . as a result
, several additional gamma - ray flares have been discovered in the agile data archive 20072009 .
they have a similar duration ( about 1 day ) and significance detection in the range ( 35)@xmath15 .
this clearly shows that agl j1734@xmath03310 is a recurrent transient gamma - ray source .
the significance of the sum of all the flares detected by agile is 7.3@xmath15 with a 95% statistical and systematic positional error radius of 0.46 degrees .
4 shows the 1860 kev integral / ibis significance mosaic map ( @xmath3 10 ms exposure ) of the sky region surrounding igr j17354@xmath03255 with superimposed the positional uncertainty of agl j1734@xmath03310 ( green circle ) . clearly , igr j17354@xmath03255 is the only hard x - ray source detected inside the agile error circle , the same holds in the softer x - ray band ( 310 kev ) from an integral / jem x deep mosaic ( @xmath3 700 ks exposure ) .
we took into account the possibility of a chance coincidence and to this aim we calculated the probability of finding a supergiant hmxb , such as igr j17354@xmath03255 , inside the agile error circle by chance .
given the number of supergiant hmxbs detected by ibis within the galactic plane @xcite , defined here as restricted to a latitude range of @xmath205@xmath1 , we estimated a probability of @xmath31@xmath21 . * 3eg j1734@xmath03232 is a still unidentified gamma - ray source listed in the third egret catalog with average flux of ( [email protected])@xmath210@xmath22 photons @xmath14 s@xmath6 ( e @xmath18100 mev ) and significance detection of 6.2@xmath15 @xcite .
it is also designated as confused source , which means that it may have significant uncertainties due to the overlapping psfs .
it is in fact possible that 3eg j1734@xmath03232 is the blend of more than one gamma - ray source , this is particularly evident thanks to recent observations performed by agile and fermi whose excellent angular resolution allowed to pinpoint the two gamma - ray sources likely responsible for the entire emission from 3eg j1734@xmath03232 ( see fig .
4 ) : agl j1734@xmath03310 ( discussed above ) and 2fgl j1732.5@xmath03131 . the latter is a firmly identified gamma - ray pulsar @xcite with average flux @xmath3 20@xmath210@xmath22 photons @xmath14 s@xmath6 ( 100 mev100 gev ) .
such hypothesis is strongly supported by i ) the blending of 3eg j1734@xmath03232 elongated towards the direction of both agl j1735@xmath03258 and 2fgl j1732.5@xmath0313 ii ) the average egret gamma - ray flux compatible with the sum of the gamma - ray fluxes from the agile and fermi sources ( i.e. @xmath345@xmath210@xmath22 photons @xmath14 s@xmath6 at e @xmath18100 mev ) .
in addition , 3eg j1734@xmath03232 is likely variable as suggested by the value of its variability index i @xcite .
the spatial match , the possible variability , the gamma - ray flux values , all strongly suggest that at least a fraction of the entire gamma - ray emission from 3eg j1734@xmath03232 might well be associated with agl j1735@xmath03258 .
the remaining part is very likely coming from 2fgl j1732.5@xmath03131 . for the sake of completeness , we note that in the surroundings of agl j1734@xmath03310 and 3eg j1734@xmath03232 there are two other persistent gamma - ray sources @xcite as detected by fermi ( see fig .
4 ) : i ) 2fgl j1737.2@xmath03213 is an unidentified gamma - ray source with average flux @xmath3 11@xmath210@xmath22 photons @xmath14 s@xmath6 ( 300 mev100 gev ) , it is not variable and it is likely associated with a supernova remnant or pulsar wind nebula ii ) 2fgl j1731.6@xmath03234c is still unidentified with average flux of @xmath3 6@xmath210@xmath22 photons @xmath14 s@xmath6 ( 300 mev100 gev ) , however it is found in a region with possibly incorrected diffuse emission . as such
, its position and even existence may not be reliable , i.e. it could be a fake source potentially confused with interstellar emission .
we note that none of the above three fermi gamma - ray sources is spatially associated with agl j1734@xmath03310 and/or igr j17354@xmath03255 . moreover , despite 2fgl j1737.2@xmath03213 and 2fgl j1731.6@xmath03234c are in the surroundings of 3eg j1734@xmath03232 , they likely give no significant contribution to the flux measured by egret since both are very weak gamma - ray sources .
based on spatial correlation as well as on a flaring nature on similar short timescales ( although not simultaneous yet ) , we propose the sfxt igr j17354@xmath03255 as the best candidate counterpart of agl j1734@xmath03310 , to date .
although such proposed association is merely based on intriguing hints , it represents an important first step towards obtaining a reliable test case on which to concentrate further efforts to obtain quantitative proofs for a real physical association . in this respect ,
further agile , fermi and integral studies of igr j17354@xmath03255/agl j1734@xmath03310 are under way . if fully confirmed , the implications of sfxts producing gamma - ray emission are huge , both theoretically and observationally since i ) it would open the study to an unexplored energy window , ii ) it would allow a deep inspection of the extreme physical mechanisms able to accelerate particles up to relativistic energies in hmxbs , iii ) it would add a further extreme characteristic to the already extreme class of sfxts .
v. sguera is very grateful to andrea bulgarelli and the agile team for sharing the results on agl j1734@xmath03310 before publication | among the different types of sources shining in the high energy sky , gamma - ray binaries are rapidly becoming the subject of major interest .
in fact , in the last few years a number of high mass x - ray binaries ( hmxbs ) have been firmly detected from mev to tev energies , providing secure evidences that particles can be efficiently accelerated up to very high energies in such galactic systems .
similarly to this general and emerging class of gamma - ray binaries , in principle supergiant fast x - ray transients ( sfxts ) have all the `` ingredients '' to be transient high energy emitters . in this context , the sfxt igr j17354@xmath03255 is a good bench test and we present intriguing hints likely suggesting that it is a transient gamma - ray source flaring on short timescales .
if fully confirmed by further studies , the implications stemming are huge , both theoretically and observationally , and would add a further extreme characteristic to the already extreme class of sfxts . |
the invention of the scanning tunneling microscope@xcite in 1981 and a consequent development in the beginning of the nineties of the remarkably simple experimental technique known as mechanically controllable break junction ( mcbj)@xcite led to the possibility of fabrication of metallic point contacts approaching the atomic scale .
the recent review article ( ref . ) summarizes the numerous achievements in this field . in the experiments the conductance measured as a function of the elongation of the nanocontacts decreases in a stepwise fashion @xcite with steps of order of the conductance quantum @xmath0
. such behavior of the conductance is attributed to atomic rearrangements that entails a discrete variation of the contact diameter . @xcite the electron transport in metallic nanocontacts is purely ballistic and phase - coherent because their size is much smaller than all scattering lengths of the system .
according to landauer,@xcite conductance is understood as transport through nonmixing channels , @xmath1 where @xmath2 s are transmission probabilities .
they are defined as eigenvalues of the transmission matrix @xmath3 . here
the matrix element @xmath4 gives the probability amplitude for an incoming electron wave in the transverse mode ( channel ) @xmath5 on the left from the contact to be transmitted to the outgoing wave in the mode @xmath6 on the right .
consequently , the eigenvectors of @xmath3 are usually called eigenchannels .
it was shown in the pioneering work by scheer _
et al._@xcite that a study of the current - voltage relation for the superconducting atomic - sized contacts allowed to obtain transmission probabilities @xmath2 s for particular atomic configurations realized in mcbj experiments .
the @xmath2 s are found by fitting theoretical and experimental @xmath7 curve which has a peculiar nonlinear behavior for superconducting contacts at voltages @xmath8 smaller than the energy gap @xmath9 of a superconductor@xcite .
the origin of such effect is explained in terms of multiple andreev reflections.@xcite the analysis of mcbj experiments within the tight - binding ( tb ) model suggested by cuevas _
et al._@xcite gave a strong evidence to the relation between the number of conducting modes and the number of valence orbitals of a contact atom .
to describe the electronic and transport properties of nanocontacts , quite a big number of different methods which supplemented each other were developed during the last 15 years .
early models employed a free - electron - like approximation.@xcite further approaches based on density functional theory ( dft ) used psuedopotentials to describe atomic chains suspended between jellium electrodes.@xcite the tb models were applied to the problem of the conduction eigenchannels@xcite and to the study of the breaking processes of nanowires.@xcite the up - to - date fully self - consistent _ ab initio _
methods@xcite allowed to treat both the leads and the constriction region on the same footing and to evaluate the non - equilibrium transport properties as well@xcite .
the scattering waves , underlying a concept of eigenchannels introduced by landauer and bttiker,@xcite do not form an appropriate basis for the most of _ ab initio _ methods .
instead , one considers conduction channels as eigenvectors of some hermitian transmission matrix written in terms of local , atom centered basis set.@xcite one of the goals of the present paper is to establish a missing link between these approaches .
below we introduce a formalism for the evaluation of conduction eigenchannels , which combines an _ ab initio _ korringa - kohn - rostoker ( kkr ) green s function method@xcite for the electronic structure calculations and the baranger and stone formulation of the ballistic transport@xcite . in recent publications,@xcite
we have successfully applied this method to the study of the electron transport through atomic contacts contaminated by impurities . in the present paper ,
mathematical aspects of the problem are considered , followed by some applications . in particular , we analyze the symmetry of channels and relate our approach to the orbital classification of eigenmodes introduced by cuevas _
et al._@xcite the paper is organized as follows .
a short description of the kkr method is given in sec .
we proceed in sec .
iii with a formal definition of eigenchannels for the case of realistic crystalline leads attached to atomic constriction .
iv supplemented by appendices a and b contains mathematical formulation of the method . briefly , using the equivalence of the kubo and landauer approaches for the conductance , @xcite we build the transmission matrix @xmath10 in the scattering wave representation .
the angular momentum expansion of the scattering bloch states within each cell is used further to find an equivalent , kkr representation of the transmission operator for which the eigenvalue problem can be solved .
applications of the method are presented in sec .
v. in particular , we focus on transition metal contacts ( such as ni , co and pd ) , since experimental@xcite and theoretical studies @xcite of their transport properties have been attracting much attention during the last years .
experiments @xcite regarding ballistic magnetoresistance ( bmr ) effect in ferromagnetic contacts are commented .
a summary of our results is given in sec .
the systems under consideration consist of two semi - infinite crystalline leads , left ( l ) and right ( r ) , coupled through a cluster of atoms which models an atomic constriction . in fig . 1 a typical configuration used in the calculations is shown the two fcc ( 001 ) pyramids attached to the electrodes are joined via the vertex atoms .
we employed the _ ab initio _ screened kkr green s function method to calculate the electronic structure of the systems . since details of the approach can be found elsewhere , @xcite only a brief description is given below . in the kkr formalism one
divides the whole space into non - overlapping , space - filling cells , with the atoms ( and empty spheres ) positioned at the sites @xmath11 , so that the crystal potential @xmath12 is expressed in each cell as @xmath13 .
the one - electron retarded green s function is expressed in terms of local functions centered at sites @xmath11 : @xmath14 where @xmath15 , @xmath16 are restricted to the cells @xmath5 and @xmath17 ; @xmath18 , @xmath19 denote one of the two vectors @xmath15 or @xmath16 with the smaller or the larger absolute value , and local functions @xmath20 and @xmath21 are the regular and irregular solutions of the schrdinger equation for the single potential @xmath22 . here the index @xmath23 stands for the angular momentum quantum numbers and atomic units are used : @xmath24 , @xmath25 , @xmath26 . the structural green s function @xmath27 ( structure constants ) in eq .
( [ grr ] ) is related to the known structure constants of the appropriately chosen reference system by the algebraic dyson equation which includes the difference @xmath28 between local @xmath29-matrices of the physical and a reference system . in the screened kkr method@xcite
we use a lattice of strongly repulsive , constant muffin - tin potentials ( typically , @xmath30ry height ) as reference system that leads to structure constants which decay exponentially in real space . within the screened kkr method both a constriction region and
the leads are treated on the same footing .
this is achieved by using the hierarchy of green s functions connected by a dyson equation , so that we perform the self - consistent electronic structure calculations of complicated systems in a step - like manner .
first , using the concept of principal layers together with the decimation technique,@xcite we calculate the structural green s function of the auxiliary system consisting of semi - infinite leads separated by a vacuum barrier . at the second step ,
the self - consistent solution of the impurity problem is found by embedding a cluster with perturbed potentials caused by the atomic contact into the auxiliary system . due to
effective screening of perturbation , the algebraic dyson equation for the structure constants is solved in real space.@xcite
the concept of eigenchannels is introduced in the landauer approach to ballistic transport , where the problem of the conductance evaluation is considered from viewpoint of scattering theory . following landauer,@xcite we look at the system shown in fig . 1 as consisting of two semi - infinite leads ( electrodes ) attached to a scattering region ( atomic - sized constriction ) .
far away from the scattering region the propagating states are the unperturbed bloch waves @xmath31 of the left ( l , @xmath32 ) and right ( r , @xmath33 ) leads , where @xmath34 belongs to the isoenergetic surface @xmath35 ( fermi surface , @xmath36 , in case of conductance ) and a common notation @xmath37 is used to denote bloch vector @xmath38 and band index @xmath39 . for
the eigenchannel problem one considers in - coming and out - going states in the l and r leads normalized to a unit flux .
the in - states in l and out - states in r are @xmath40 with positive velocity @xmath41 along @xmath42-axis .
the conjugated states @xmath43 are the out - states in l and in - states in r with negative velocity @xmath44 . here
@xmath45 is a @xmath42-component of the group velocity @xmath46 ; a proportionality factor between @xmath47 and @xmath48 related to a particular choice of normalization of the bloch waves is introduced further in sec.iv.b .
the potential @xmath49 describing the constriction introduces a perturbation to the perfect conductor .
let @xmath50 be a perturbed state which is a solution of the lippmann - schwinger equation for an in - coming state in l : @xmath51 where the integral goes over all space , and @xmath52 is the retarded green s function of the perfect conductor .
asymptotic behavior of @xmath50 is @xmath53 where @xmath54 and @xmath55 are transmission and reflection amplitudes , assuming elastic scattering ( @xmath56 ) . according to the landauer - bttiker formula,@xcite conductance
is given by @xmath57 , where trace goes over in - coming states ( @xmath58 ) in the left electrode and @xmath36 .
an equivalent formulation with respect to in - coming states ( @xmath59 ) from the right electrode reads as @xmath60 .
eigenchannels appear from a unitary transformation of in- and out - states .
let @xmath61 be a unitary transform of in - states in l : @xmath62 .
the corresponding solution @xmath63 of eq .
( [ ls ] ) for an arbitrary * r * is @xmath64 the unitary transform @xmath61 is defined such way that the transmission matrix @xmath65 is diagonal in the basis @xmath66 : @xmath67 and the conductance reads as @xmath68 , where the @xmath69 s are transmission probabilities of eigenchannels .
the matrix @xmath70 , however , is not diagonal in basis @xmath66 .
following ref .
one can introduce a unitary matrix @xmath71 which satisfies the equation : @xmath72 where all quantities are energy dependent .
the solution is @xmath73 .
the following properties of @xmath71 can be checked : @xmath74,@xmath75 , thus @xmath71 is indeed the unitary matrix .
it diagonalizes @xmath76 : @xmath77 matrix @xmath71 performs a unitary transform of out - states in r , so that the linear combination @xmath66 of in - coming states in l scatters into the linear combination @xmath66 of the out - states in r , @xmath78 with the transmission amplitude @xmath79 , namely : @xmath80
. one can show@xcite that for the bloch states at the same energy ( @xmath81 ) in the ideal leads the following relations hold for the current matrix elements : @xmath82 \ = \ \frac{v_{\boldsymbol{k}}}{2\pi } \ , \delta_{\boldsymbol{kk'}},\ ] ] where the bloch waves are either left ( @xmath83 ) or right - travelling ( @xmath84 ) , the operator @xmath85 is defined as @xmath86 , and the integral goes over infinite plane @xmath87 ( cross - section of the lead ) which is perpendicular to the current direction @xmath42 . in case of the perturbed system the orthogonality relation holds for current matrix elements in the basis of eigenchannels . using eqs .
( [ phi_k_limit ] ) , ( [ omega ] ) and ( [ curr_bloch1 ] ) we can compute it in the asymptotic region of the right ( r ) lead:@xcite @xmath88_{z\to + \infty } = \frac{t_{\nu}(e)}{2\pi}\,\delta_{\nu\mu}.\ ] ] we note , that eq .
( [ curr_right ] ) holds for _ any _ position @xmath42 of the plane @xmath87 . because the wave functions of channels are solutions of the schrdinger equation with a real potential corresponding to the same energy , the flux through arbitrary plane @xmath87 is conserved ( a proof is similar to that of appendix a of ref . ) .
to calculate the ballistic conductance , we employ the kubo linear response theory as formulated by baranger and stone : @xcite @xmath89 where @xmath90 and @xmath91 are retarded and advanced green s functions , respectively , and @xmath92 .
the integration is performed over left ( @xmath93 ) and right ( @xmath94 ) planes which connect the leads with the scattering region ( fig . 1 ) .
the current flows in @xmath42 direction . the implementation of eq .
( [ cond_g ] ) within the kkr method , related convergence tests and further details were discussed in recent publication.@xcite in this paper we present a further extension of ref . , a method for the evaluation of conduction eigenchannels .
for that , we will follow closely the analysis of refs . where the equivalence of the kubo and landauer approaches to conductance problem was shown .
we proceed in four steps : ( i ) we remind the kkr representation of the bloch functions ; ( ii ) we build up the asymptotic expansion of the green s function in terms of unperturbed bloch states of the leads ; ( iii ) we construct the transmission matrix @xmath95 in @xmath58-space ; ( iv ) finally , we find an equivalent representation of the transmission matrix within the local kkr basis
. further solution of the eigenvalue problem leads us to conduction channels .
we mention here one aspect of the problem : for realistic calculations the planes @xmath93 and @xmath94 ( see fig
. 2 for details ) are usually placed in finite distance from the atomic constriction . nevertheless , we first focus on the asymptotic limit before we discuss the realistic situation which is considered in sec .
iv.e and appendix a. let @xmath96 be arbitrary point in the asymptotic region of the lead ( fig . 2 ) . in the kkr method
the local , energy dependent basis of atomic orbitals is defined at each unit cell @xmath5 : @xmath97 where @xmath98 s are real regular solutions @xcite of the schrdinger equation for the potential @xmath99 at cell @xmath5 , and @xmath100function is 1 inside cell @xmath5 and is 0 outside it .
the unperturbed bloch function is given by expansion over atomic orbitals at all sites @xmath5 in the born - von krmn supercell : @xmath101 with @xmath102 . here the common notation @xmath37 for the bloch vector @xmath38 of the 1st brillouin zone ( bz ) and the band index @xmath39 is used .
the @xmath103 are solutions of the kkr band structure equations @xcite with energy @xmath104 . considering a waveguide geometry
, we will assume the bloch functions to be normalized per cross section of the born - von krmn supercell ( see fig .
2 ) with open boundary conditions along @xmath42-axis . thus , @xmath105 with @xmath106 being number of atoms per cross - section , and orthogonality condition for bloch waves takes a form @xcite : @xmath107 here @xmath108 is a volume of the supercell , @xmath109 is an area of the @xmath110-unit cell in the electrode , @xmath111 are signs of @xmath112 and @xmath113 , relation @xmath114 has been used , and velocity @xmath47 along current flow is defined as @xmath115 . since the bloch waves form a complete set , a back transform of eq .
( [ psik_nl ] ) exists : @xmath116 where @xmath117 sum runs over all @xmath118-points in the 1st bz and over all bands @xmath119 .
the @xmath120 symbol means hermitian conjugate .
the expression for @xmath121 can be obtained @xcite from known matrix @xmath122 .
one can prove further , that @xmath122 and @xmath121 obey the following orthogonality relations : @xmath123 where in the second equation a sum over @xmath58 is restricted to states with @xmath124 .
starting from the site angular momentum representation ( [ grr ] ) of the retarded green s function within the kkr method and using eqs .
( [ psik_nl ] ) and ( [ phi_nl ] ) we obtain the asymptotic expansion for @xmath125 over the unperturbed bloch waves : @xcite @xmath126 with @xmath127 , @xmath128 ( see fig . 2 ) , and @xmath129 or in a matrix form : @xmath130 . formally ,
the @xmath58-sums in eq .
( [ g_knl ] ) are performed over all @xmath38-states in the 1st bz and over all bands @xmath39 . however , since the green s function for @xmath131 is a solution of the schrdinger equation without a source term , only states @xmath132 at the isoenergetic surface of energy @xmath104 contribute to the sum in the asymptotic expansion.@xcite therefore , @xmath133 where @xmath134 . here
@xmath135 is number of atom sites in born - von krmn supercell along @xmath42 axis , @xmath136 is volume of the 1st
bz , @xmath137 is the area of the isoenergetic surface corresponding to band @xmath39 , @xmath138 . for the discrete @xmath58-points , the function @xmath139 equals 1 if @xmath140 , and is 0 otherwise .
in addition , boundary conditions for the green s function@xcite constrain matrix elements @xmath141 to be non - zeros only if @xmath58 and @xmath117-states are right - travelling waves ( with positive velocity along @xmath42-axis ) that corresponds to the in - states @xmath58 in the left lead and out - states @xmath117 in the right one .
we proceed further , and use the asymptotic representation ( [ g_knl ] ) of the green s function to evaluate conductance according to eq .
( [ cond_g ] ) . assuming the integration planes to be placed within the leads infinitely far from the scattering region , we obtain : @xmath142,\ ] ] where the diagonal operators of velocities @xmath143 and
@xmath144 ( related to the left and right planes ) acting in the @xmath58-space were introduced : @xmath145_{\boldsymbol{kk ' } } = \left({v_{\boldsymbol{k}}}/{2\pi}\right ) \delta_{\boldsymbol{kk'}}$ ] .
formally , the trace ( tr ) in eq .
( [ g_tr_vava ] ) goes over all @xmath58-states and the fermi surface is taken into account by means of eq .
( [ a0_kk ] ) where @xmath146 .
the velocity operators can be decomposed into sum of two operators related to the bloch states with positive and negative velocities along @xmath42 : @xmath147 where @xmath148 is nonzero for right - travelling waves only , while @xmath149 is nonzero for left - travelling ones . in the asymptotic limit , only in - coming and
out - going @xmath58-states with positive velocities contribute to the sums in eq.([g_tr_vava ] ) . using the relation between expansion coefficients @xmath150 and the transmission amplitudes @xmath151 derived in refs .
, @xmath152 we obtain : @xmath153 \\ \nonumber { } & = & g_0 \mathrm{tr}_{(\boldsymbol{k})}\,[t(e_f)],\end{aligned}\ ] ] where a representation of @xmath65 in @xmath58-space is given by @xmath154 with a positive definite operator under square - roots .
the @xmath58-representation is formal but not suitable for implementation . to solve the eigenvalue problem for @xmath65 the mapping on the site - angular momentum @xmath155-space of the kkr method
should be presented .
such mapping is realized through the expansion ( [ psik_nl ] ) of the bloch functions over atomic orbitals , so that velocity operators in @xmath58-space take a form : @xmath156_{\boldsymbol{kk ' } } = \sum_{nn'\in s^0_{\mathrm{l } } } \sum_{ll ' } c_{\boldsymbol{k},nl}\ , \bigl[d_{\mathrm{l(r)}}\bigr]^{nn'}_{ll'}\ , c^{\dagger}_{n'l',\boldsymbol{k'}},\ ] ] where site - diagonal operators @xmath157 and @xmath158 defined on atomic orbitals are the kkr - analogue of velocities : @xcite @xmath159^{nn'}_{ll'}\ = } \\ \nonumber\displaystyle & & \pm\ , \delta_{nn'}\ , \int_{s^n_{\mathrm{l(r ) } } } ds \left [ r^n_l({\bf r},e)\ , i\stackrel{\leftrightarrow}{\partial_z } r^{n}_{l'}({\bf r},e ) \right]\end{aligned}\ ] ] here integral is restricted to the cross - section of the unit cell around site @xmath5 .
now we can evaluate conductance according to eq.([g_tr_vava ] ) .
taking into account that @xmath160 [ eq.([bc ] ) ] , we obtain : @xmath161 here ( and further ) all matrices are assumed to be taken at the fermi energy , @xmath162 stands for matrix notation of the structural green s function introduced in eq .
( [ grr ] ) , and trace ( tr ) involves sites and orbitals related to the atomic plane @xmath163 ( fig . 2 ) .
the operators @xmath158 , @xmath157 are anti - symmetric and hermitian , thus their spectrum consists of pairs of positive and negative eigenvalues : @xmath164 .
let @xmath165 be the unitary transform of matrix @xmath166 ( either @xmath158 or @xmath157 ) to a diagonal form : @xmath167 decomposition of operator @xmath168 into two terms , @xmath169 , related to the right- and left - travelling waves is naturally given in the basis of eigenvectors : @xmath170 with @xmath171 where @xmath172 , @xmath173 are positive ( non - negative ) diagonal matrices .
the analogue of eq .
( [ g_vp_a_vp_a ] ) reads as @xmath174 now we are ready to build the @xmath155-representation for transmission matrix @xmath65 . for that
, one should extract a square - root from the positive definite operator @xmath175 defined on @xmath58-space with help of @xmath155-space .
namely , because the @xmath176 is positive definite , it can be represented in the following form : @xcite @xmath177 .
here an operator @xmath178 maps the @xmath58-space on the @xmath155-space , @xmath179 . the solution for @xmath180 is @xmath181 where @xmath182 is an arbitrary unitary matrix ( @xmath183 ) and @xmath184^{1/2 } u^{\dagger}_l $ ] with @xmath185^{1/2}$ ] being square - root of the positive definite diagonal matrix [ eq.([d_ud0u ] ) ] . to find @xmath65 , we start from eq .
( [ g_vp_a_vp_a ] ) and proceed as follows : @xmath186\end{aligned}\ ] ] where eq.([a_kk ] ) was used .
because of @xmath187 [ eq .
( [ bc ] ) ] , we obtain : @xmath188 where , in the asymptotic limit , the transmission matrix @xmath65 in @xmath155-representation is given by @xmath189 the trace in eq.([g_tr_nl ] ) goes over all sites @xmath5 and orbitals @xmath190 of the atomic plane @xmath163 in the left lead ( fig . 2 ) .
equivalent formula can be written for the right lead .
to conclude , one can prove that spectrum of the obtained matrix coincides with spectrum of transmission matrix @xmath191 [ eq .
( [ tke ] ) ] defined in @xmath58-space ( see appendix b for details ) .
therefore , solution of the eigenvalue problem for @xmath192 gives us required transmission probabilities .
in practical calculations of conductance with the use of the kubo formula , the left ( l ) and right ( r ) planes are positioned _ somewhere _ in the leads ( fig . 2 ) .
expression ( [ g_tracedgdg ] ) is valid in general case and result is exactly the same as in ref . .
operator @xmath158 in eq.([g_tracedgdg ] ) is sum of two terms : @xmath193 .
therefore , we can write down @xmath194 where @xmath195 and @xmath196 denote two contributions . in a formal theory , when the atomic plane @xmath93 is placed in the asymptotic region of the left lead far from the atomic constriction the second term @xmath196 in eq.([2_terms ] )
is equal to zero . in practice ,
the real space summation of current contributions includes only a finite number of sites at both atomic planes , because the current flow along @xmath42 direction is localized in the vicinity of the contact . due to numerical effort
we are forced to take integration planes closer to the constriction in order to obtain convergent value for the conductance with respect to number of atoms included in summation .
in addition , even better convergence for matrix elements is required to solve the eigenvalue problem .
a compromise can be usually achieved but positions of the atomic planes @xmath93 and @xmath94 do not meet the asymptotic limit criterion .
however , since the electron current through the structure is conserved , any position of the planes is suitable for the calculation of conductance .
if @xmath93 is placed somewhere in the scattering region we have to sum up all multiple scattering contributions .
we show in appendix a , that all multiple scattering contributions in direction of the current cause @xmath195 , whereas all scattering contributions in opposite direction give rise to @xmath196 .
thus , the first term , @xmath195 , in eq .
( [ 2_terms ] ) is always positive , while the second one , @xmath196 , is always negative . to make this statement clear , an illustration of scattering events
is shown in fig .
3 assuming a simple free - electron model . in the region of the lead where the potential is a small perturbation with respect to the bulk potential
the contribution to the conductance due to @xmath196 is one order of magnitude smaller than @xmath195 . to find transmission probabilities of eigenchannels
, one has to apply the procedure introduced in the previous section independently for both terms , @xmath195 and @xmath196 .
we refer to appendix a for a mathematical justification .
expression for conductance takes a form : @xmath197 with @xmath198 we show in appendix a that all eigenvalues of @xmath199 are either positive or zeros whereas all eigenvalues of @xmath200 are negative or zeros . to identify transmission probabilities @xmath2 of channels the spectra of operators @xmath199 and @xmath200 have to be arranged in a proper way .
then transmission of the @xmath5-th channel is given by @xmath201 , where @xmath202 are positive and negative eigenvalues of the operators @xmath199 and @xmath200 , respectively .
the @xmath2 does not depend on the positions of the left ( @xmath93 ) and right ( @xmath94 ) planes , while @xmath203 are @xmath42-dependent . in the asymptotic limit @xmath204 and @xmath205 , so that the landauer picture is restored . in general case
the direct way to find the pairs of eigenvalues is not evident without a back transform to the @xmath58-space .
however , from the point of view of applications to the extremely small symmetric atomic contacts , as the ones we are studying in this work , the problem is easy to handle .
since the number of contributing eigenmodes is limited , the pairs of eigenvalues can be found by symmetry analysis of the eigenvectors of @xmath199 and @xmath200 .
namely , using the symmetry properties of the structural green s function @xmath206 and current matrix elements @xmath207 one can show that channel s transmission @xmath2 bounded between 0 and 1 is defined by eigenvalues @xmath208 and @xmath209 which belong to the same irreducible representation of the symmetry point group .
in recent papers @xcite we have verified the method described here by studying systematic changes in the conductance of metallic constrictions in the presence of defect atoms .
illustrative examples presented below focus on single - atom contacts made of pure metals such as cu , ni , co and pd . copper serves mainly for test purposes .
it is a representative of the noble metals and has electronic properties similar to ag and au , for which a lot of experimental results @xcite as well as dft based calculations @xcite are available .
in particular , a large number of experiments for alkali metals ( li , na , k ) @xcite and noble metals ( au , ag , cu ) , @xcite employing different techniques under room an liquid - he temperatures , show that conductance histograms have a dominant peak very close to one conductance quantum @xmath0 , and smaller peaks close to integer values .
however , for transition metal contacts ( examples of which are ni , co and pd )
the situation differs significantly.@xcite only one broad maximum centered somewhere between @xmath210 and @xmath211 is usually observed in conductance histograms.@xcite that is a signature of the nontrivial decomposition of conductance consisting of more than one perfectly transmitting channel , @xcite since for transition metal atoms @xmath212 states of different symmetry are available at the fermi level .
the question on half - integer conductance quantization has been addressed.@xcite however , recent experiments @xcite do not confirm this hypothesis thus pointing to the conclusion that the electron transport through ferromagnetic contacts can never be fully spin - polarized .
another issue is a large magnetoresistance ( mr ) effect @xcite observed for metallic point contacts made of different magnetic materials .
this field is known to be full of controversy .
there is still a continuing discussion on whether or not the enormous mr values found experimentally could be of electronic origin @xcite or the effect is just due to atomic rearrangements in the neck region of a contact as a response to the applied magnetic field.@xcite extensive discussion on this topic can be found in the recent review paper by marrows .
@xcite we will comment further on the above issues .
an atomic configuration of a constriction used in our calculations is presented in fig . 1 .
the single atom contact was modeled by a small cluster attached to the semi - infinite fcc ( 001 ) leads .
the cluster consists of two pyramids joined via the vertex atoms separated by a distance @xmath213 . here
@xmath214 is the experimental lattice constant of fcc metals : 6.83 a.u . for cu , 6.66 a.u . for ni , 6.70 a.u . for co , and 7.35 a.u
. for pd .
the metals under consideration do not have a tendency to form chains.@xcite an atomic bridge is most likely to be broken just after the single atom limit is achieved .
thus , a configuration shown in fig .
1 resembles one of limiting configurations of point contacts which could appear in the mcbj experiments .
our calculations are based on dft within the local density approximation .
the parametrization of vosko , wilk , and nusair @xcite for the exchange and correlation energy was used .
the potentials were assumed to be spherically symmetric around each atom ( atomic sphere approximation , asa ) .
however , the full charge density , rather than its spherically symmetric part , was taken into account . to achieve well converged results the angular momentum cut - off for the wave functions and the green s function
was chosen to be @xmath215 that imposed a natural cut - off @xmath216 for the charge density expansion . in case of heavy element
pd the scalar relativistic approximation@xcite was employed . for the conductance calculation the surface green s function was computed using a small imaginary part @xmath217 and about 250.000 were taken in the 2d brillouin zone . instead of integration over planes ,
current matrix elements ( [ d_lr ] ) were averaged over atomic layers as described in detail in ref . .
a typical error in the calculation of conductance was @xmath218 . to understand the relation between the electronic structure and the transport properties of atomic contacts we consider the energy - dependent total transmission , @xmath219 , and its decomposition to the conduction eigenchannels , @xmath220 .
results are shown in figs .
4,6,8,9 for the case of cu , ni , co and pd point contacts , respectively . the investigated structure ( fig . 1 )
has a @xmath221 symmetry .
further we denote individual channels by the indices of irreducible representations of this group using notations of ref . ,
common in band theory .
in addition , each channel can be classified according to the angular momentum contributions when the channel wave function is projected on the contact atom of the constriction .
this is very helpful since the channel transmission can be related to the states of the contact atom.@xcite for example , the identity representation @xmath222 of the @xmath221 group is compatible with the @xmath223 , @xmath224 and @xmath225 orbitals ( here the @xmath42 is perpendicular to the surface and passes through the contact atom ) , while the two - dimensional representation @xmath226 is compatible with the @xmath227 , @xmath228 , @xmath229 , @xmath230 orbitals .
the basis functions of @xmath231 and @xmath232 are @xmath233 and @xmath234 harmonics , respectively . the energy - dependent transmission of cu atomic contact ( shown in fig .
1 ) is presented in fig.4 together with the eigenchannel decomposition . at the fermi energy
the calculated conductance value is @xmath235 .
it mainly consists of one open channel of @xmath222 symmetry which arises locally from @xmath223 , @xmath224 and @xmath225 orbitals when the wave function is projected on the contact atom .
this result is in a good agreement with a lot of experiments@xcite mentioned previously as well as with other calculations involving different approaches.@xcite the additional twofold degenerate channel has @xmath236 symmetry .
transmission of this channel increases at energies above the fermi level ( @xmath237 ) together with an increase of the @xmath238 contribution to the local density of states ( ldos ) at the contact atom .
however , at @xmath239 ev below the @xmath237 the @xmath236 channel is built mainly from the @xmath240 orbitals of the cu atom .
we would like to point out that in case of noble metals the conductance of single - atom contact is not necessarily restricted to one channel .
an example of a configuration which has more channels ( but still has only one cu atom at the central position ) is presented on the top of fig .
5 . because of the larger opening angle for incoming waves
as compared with the preceding case , conductance of such system is @xmath241 with major contribution from four channels ( fig . 5 and its caption ) .
the value @xmath242 correlates with a position of the third peak in the conductance histogram of cu , which is shifted from @xmath243 to smaller values.@xcite as it is seen from the presented example , conductance quantization does not occur for the metallic atomic - sized contacts . in general , even for noble metals , conduction channels are only partially open@xcite in contrary to the case of quantum point contacts realized in the two - dimensional electron gas where a clear conductance quantization was observed.@xcite for illustration , we present in fig .
5 probability amplitudes of the eigenchannles @xmath244 in real space resolved with respect to atoms of the 2nd plane ( s-2 ) below the surface .
we see that the wave functions of the 1st and the 4th channel with the highest symmetry ( @xmath222 ) obey all eight symmetry transformations of the @xmath221 group , while two different wave functions of the double degenerate channel ( @xmath236 ) are transformed to each other after some symmetry operations .
we turn to transition metals , and consider the ferromagnetic ni assuming a uniform magnetization of the sample .
transmission @xmath219 split per spin of a ni contact is shown in fig .
a shift about @xmath245 ev along the energy axis between transmission curves is seen that is in agreement with exchange - splitting of the ni @xmath212 states .
similar computational results regarding transmission of ni constrictions were reported by solanki _
et al_.,@xcite rocha _
et al._@xcite and smogunov _
et al._.@xcite exchange splitting estimated from their works varies from 0.8 ev till 1.0 ev , but fine details differ because of different atomic configurations and employed methodologies . in this
regard , exchange splitting about 2.0 ev observed in transport calculations of jacob _
et al._@xcite in case of ni contact seems to be overestimated .
the shift in energy due to different spins is observed as well for the transmissions of individual channels ( fig .
we see from fig . 7 and table i that at the fermi energy the spin - up ( majority ) conductance of ni contact is mainly determined by one open @xmath222 channel ( similar to the case of cu ) , while three partially open channels , of @xmath222 and @xmath236 symmetry , contribute to the spin - down ( minority ) conductance .
the minority @xmath236 channel arises locally from @xmath229 and @xmath230 states , rather than from @xmath246 states whose contribution to the spin - down ldos at the fermi energy is much smaller ( fig .
the calculated conductance , @xmath247 , correlates with a position of the wide peak in the conductance histogram of ni centered between @xmath210 and @xmath248.@xcite [ cols=">,^,^,^,^,^,^ " , ] coming to the experimental situation on ballistic magnetoresistance ( bmr ) effect in ferromagnetic contacts , we point out that large mr values @xcite were usually measured for much thicker constrictions ( as compared with atomic - sized contacts ) with resistance in the range of hundreds of ohms .
it is believed,@xcite that such experiments suffer from many unavoidable artifacts induced by magnetomechanical effects that mimics the real mr signal which would come from the spin - polarized transport alone .
however , recent studies by sullivan _ et al_.@xcite and chopra
_ et al._@xcite on ni and co atomic - sized contacts report bmr values in the range of @xmath249 with discussion on the electronic origin of the effect due to domain wall scattering .
nevertheless attempts to minimize magnetostrictive effects were undertaken , we just can repeat@xcite that a natural explanation of these@xcite and similar experiments @xcite is that , due to magnetization reversal processes , unstable in time atomic constriction changes its contact area when magnetic field is applied .
characteristic steps and jumps in the measured field - dependent conductance ( fig .
4 of ref . ) or resistance ( fig .
3a of ref . ,
fig . 3 of ref . ) are distinct evidence of atomic reconstructions and fractional changes of the contact cross section .
for example , just eliminating one contact atom from the configuration shown in fig .
1 changes conductance of a ni constriction from @xmath250 ( chain of two atoms , see table i ) up to @xmath251 ( one contact atom only , see ref . ) , thus producing @xmath252% mr .
further increase of a contact area can give arbitrary high mr values , that supports hypothesis on mechanical nature of the effect .
to summarize , we have presented a formalism for the evaluation of conduction eigenchannels of metallic atomic - sized contacts from first - principles .
we have combined the _ ab initio _ kkr green s function approach with the kubo linear response theory .
starting from the scattering wave formulation of the conductance problem , we have built a special representation of the transmission matrix in terms of local , energy and angular momentum dependent basis inherent to the kkr method .
we have proven that solutions of the eigenvalue problem for the obtained matrix are identical to conduction eigenchannels introduced by landauer and bttiker .
applications of the method have been presented by studying ballistic electron transport through cu , pd , ni and co single - atom contacts .
the symmetry analysis of eigenchannels and its connection to the orbital classification known from the tight - binding approach were discussed in detail .
experiments on the electron transport through magnetic contacts were commented .
we acknowledge a financial support through the deutsche forschungsgemeinschaft ( dfg ) , priority programme 1165 : `` nanowires and nanotubes '' .
we consider a practical implementation of the kubo formula ( [ cond_g ] ) : positions @xmath253 and @xmath254 of the @xmath255 and @xmath256 planes are chosen within the semi - infinite electrodes , where @xmath257-matrix describing a scattering region is sufficiently small ( for details , see fig . 2 ) .
a sketch as how to find the conduction eigenchannels in this case has been given in sec .
below we complete that discussion by presenting a mathematical justification of the method .
we proceed in three steps : ( i ) since a position of @xmath255 and @xmath256 planes in eq .
( [ cond_g ] ) does not meet an asymptotic limit criterion , we expand the green s function in terms of scattering bloch states in contrary to expansion ( [ g_knl ] ) involving unperturbed states ; ( ii ) we use the kubo formula ( [ cond_g ] ) and express conductance as a trace of the appropriately defined current operator in @xmath58-space ; the obtained operator is identified with transmission matrix @xmath76 ; ( iii ) finally , we build up the equivalent site - angular momentum representation of the transmission matrix , that leads us to the required formulae [ eqs .
( [ g_2terms ] ) , ( [ tt ] ) of sec .
iv.e ] . to begin the proof
, we note that the asymptotic representation ( [ g_knl ] ) of the green s function can be rewritten as @xmath258_{\boldsymbol{kk'}}}\ \phi_{\boldsymbol{k'}}(\mathbf{r'},e)\end{aligned}\ ] ] where an expansion is performed over perturbed bloch waves @xmath259 and @xmath260 which were introduced in sec .
we remind , that function @xmath261 is the solution of the lippmann - schwinger equation ( [ ls ] ) associated with the in - coming unperturbed bloch wave @xmath262 in the left lead propagating towards atomic constriction ( @xmath263 ) , while the second function , @xmath264 , is the solution of the equation associated with the in - coming unperturbed bloch wave @xmath265 in the right lead ( @xmath266 ) .
consider now a general case , when @xmath253 and @xmath254 points related to planes @xmath255 and @xmath256 ( see fig .
2 ) are not taken infinitely far from atomic constriction .
we note , that for the conductance evaluation one needs only the back - scattering term @xmath267 in eq .
( [ grr ] ) , which is the solution of the schrdinger equation without a source term .
therefore , we can expand @xmath268 over the eigenfunctions of the whole system corresponding to energy @xmath104 .
these are the propagating perturbed bloch states , @xmath269 and the evanescent states , @xmath270 where , in turn , both type of functions are expanded over atomic orbitals , and @xmath271 and @xmath272 are expansion coefficients .
in particular , matrix @xmath273 is related to the matrix @xmath274 corresponding to the unperturbed bloch waves ( for details , see ref . ):
@xmath275^{nn'}_{ll ' } { c}_{\boldsymbol{k},n'l'}(e)\ ] ] where the @xmath257-matrix describes the whole scattering region ( a vacuum barrier plus an atomic constriction ) and @xmath276 is the structural green s function of the three - dimensional periodic bulk crystal .
both equations ( [ psik_pert ] ) and ( [ x_a ] ) can be joined to one matrix equation : @xmath277 \phi(\mathbf{r},e),\ ] ] where ( ) stands for a column - vector , while [ ] denotes a matrix .
we also expand the conjugated states as @xmath278 \phi(\mathbf{r},e).\ ] ] we expand the back - scattering term @xmath267 over eigenfunctions @xmath279 and @xmath280 with energy @xmath104 : @xmath281 \left ( \begin{array}{c } \psi(\mathbf{r'},e ) \\
{ \cal{x}}(\mathbf{r'},e ) \end{array } \right ) \\ \nonumber & = & \phantom{+\ } \sum_{\boldsymbol{k k ' } } \overline\psi_{\boldsymbol{k}}(\mathbf{r},e)\ , f^{00}_{\boldsymbol{kk'}}(e)\ , \psi_{\boldsymbol{k'}}(\mathbf{r'},e ) \\ \nonumber & & + \
\sum_{\boldsymbol{k}\beta } \overline\psi_{\boldsymbol{k}}(\mathbf{r},e)\ , f^{01}_{\boldsymbol{k}\beta}(e)\ , { \cal{x}}_{\beta}(\mathbf{r'},e ) \\
\nonumber & & + \ \sum_{\alpha\boldsymbol{k ' } } { \cal{x}}^{*}_{\alpha}(\mathbf{r},e)\ , f^{10}_{\alpha\boldsymbol{k'}}(e)\ , \psi_{\boldsymbol{k'}}(\mathbf{r'},e ) \\ \nonumber & & + \ \sum_{\alpha\beta } { \cal{x}}^{*}_{\alpha}(\mathbf{r},e)\ , f^{11}_{\alpha\beta}(e)\ , { \cal{x}}_{\beta}(\mathbf{r},e ) \\ \label{g_rr_e_fexp } & = & \delta g_{+}^{(1)}(\mathbf{r},\mathbf{r'},e ) + \delta g_{+}^{(2)}(\mathbf{r},\mathbf{r'},e).\end{aligned}\ ] ] only the first term @xmath282 exists in the asymptotic case , while other three terms , denoted as @xmath283 , contain contributions of the evanescent states which decay within the leads .
the unknown matrix @xmath284 in eq .
( [ g_rr_e_fexp ] ) satisfies the equation ( below we will skip for simplicity the obvious energy dependence ) : @xmath285 \times \left [ \begin{array}{cc } f^{00 } & f^{01 } \\
f^{10 } & f^{11 } \end{array } \right ] \times \left [ \begin{array}{c } { \mathscr{c } } \\
\gamma \end{array } \right ] = g_{\mathrm{lr}},\ ] ] where symbol @xmath191 denotes transpose , and matrix @xmath286 is introduced which contains selected matrix elements of the structural green s function @xmath287 with @xmath288 and @xmath289 . to find matrices @xmath290
, we introduce the block - matrix @xmath291 $ ] which is a solution of the following equation ( here symbol @xmath120 denotes hermitian conjugate ) : @xmath292 \times \left [ { \cal{b}}^{\dagger}\ \delta^{\dagger } \right ] = \left [ \begin{array}{cc } { \mathscr{c } } { \cal{b}}^{\dagger } & { \mathscr{c}}\delta^{\dagger } \\
\gamma { \cal{b}}^{\dagger } & \gamma \delta^{\dagger } \end{array } \right ] = \left [ \begin{array}{cc } 1 & 0 \\ 0 & 1 \end{array } \right].\ ] ] this means , that matrix @xmath293 $ ] is the inverse or ( in general case ) the pseudoinverse @xcite matrix to the matrix with @xmath294- and @xmath295-blocks.@xcite .
the equation similar to eq .
( [ eq_cg_bd ] ) exists for other matrices : @xmath296 \times \left [ \widetilde{\mathscr{c}}^t\ \widetilde\gamma^t \right ] = \left [ \begin{array}{cc } \widetilde{\cal{b}}^{*}\ , \widetilde{\mathscr{c}}^t & \widetilde{\cal{b}}^{*}\ , \widetilde{\gamma}^t \\
\widetilde{\delta}^{*}\ , \widetilde{\mathscr{c}}^t & \widetilde{\delta}^{*}\ , \widetilde{\gamma}^t \end{array } \right ] = \left [ \begin{array}{cc } 1 & 0 \\ 0 & 1 \end{array } \right],\ ] ] where symbol @xmath297 denotes complex conjugate of the matrix elements only . with the use of eqs .
( [ eq_cg_bd ] ) , ( [ eq_cg_bdx ] ) we find solution for @xmath284 : @xmath298 = \left [ \begin{array}{cc } \widetilde{\cal{b}}^ { * } g_{\mathrm{lr } } { \cal{b}}^{\dagger } & \widetilde{\cal{b}}^ { * } g_{\mathrm{lr } } { \delta}^{\dagger } \\ \widetilde{\delta}^ { * } g_{\mathrm{lr } } { \cal{b}}^{\dagger } & \widetilde{\delta}^ { * } g_{\mathrm{lr } } { \delta}^{\dagger } \end{array } \right],\ ] ] then expansion coefficients of the @xmath299 term read as @xmath300^{nn'}_{ll ' } { \cal{b}}^{\dagger}_{n'l',\boldsymbol{k'}}(e).\end{aligned}\ ] ] with help of matrix @xmath301 , @xmath302_{\boldsymbol{kk ' } } = -\sqrt{v_{\boldsymbol{\kappa}}}\ \frac{f^{00}_{\boldsymbol{kk'}}(e)}{2\pi i } \
\sqrt{v_{\boldsymbol{\kappa'}}}\ ] ] ( with positive velocities under square - roots ) we write down eq .
( [ g1_f ] ) as @xmath303_{\boldsymbol{kk ' } } \phi_{\boldsymbol{k'}}(\mathbf{r'},e).\end{aligned}\ ] ] comparing obtained equation with one for the asymptotic limit [ eq . ( [ g_pert_wf ] ) ] and taking into account that in fact @xmath301 is independent on the positions of the l and r planes , we obtain : @xmath304_{\boldsymbol{kk ' } } = \left[\tau^{-1}(e)\right]_{\boldsymbol{kk'}}.\ ] ] when conductance is evaluated , only @xmath305 term in expansion ( [ g_rr_e_fexp ] ) contributes . other term , @xmath306 , contains the evanescent states which decay within the leads and , therefore , have zero velocities along the current flow . when eq .
( [ g1_tau ] ) is inserted into the kubo formula ( [ cond_g ] ) we obtain : @xmath307_{\boldsymbol { k k ' } } { \cal{j}}^{\mathrm{r}}_{\boldsymbol{k'k'_1 } } \left[\frac{1}{\tau^{\dagger}_{\mathrm{rl } } } \right]_{\boldsymbol{k'_1 k_1}},\ ] ] where all matrices are evaluated at the fermi energy ( @xmath308 ) .
here we have introduced the current operators related to the left ( l ) lead , @xmath309 \ = \ \bigl [ \tau^{\dagger}\tau \bigr]_{\boldsymbol{kk'}},\end{aligned}\ ] ] and to the right ( r ) one : @xmath310 \ = \ \bigl [ \tau\tau^{\dagger}\bigr]_{\boldsymbol{kk'}},\end{aligned}\ ] ] where basis @xmath66 of eigenchannels ( see sec .
iii ) was used to evaluate the matrix elements .
with help of ( [ jl_def ] ) and ( [ jr_def ] ) , equation ( [ g_jtjt ] ) for conductance @xmath311 takes a form : @xmath312 where @xmath313 is the unitary operator in the @xmath58-space : @xmath314 and eqs .
( [ tau_lr_tau ] ) and ( [ jr_def ] ) were used .
when the perturbed bloch waves are expanded over atomic orbitals , @xmath315 current operators @xmath316 and @xmath317 take a form ( energy dependence is skipped ) : @xmath318^{nn'}_{ll ' } \widetilde{x}^{*\dagger}_{n'l',\boldsymbol{k ' } } , \\ \nonumber { \cal{j}}^{\mathrm{r}}_{\boldsymbol{kk ' } } & = & \sum_{nn'\in s_r}\sum_{ll ' } { x}_{\boldsymbol{k},nl } \left[d_{\mathrm{r } } \right]^{nn'}_{ll ' } { x}^{\dagger}_{n'l',\boldsymbol{k'}},\end{aligned}\ ] ] where operators @xmath319 and @xmath320 are defined as @xmath321 and operators @xmath158 and @xmath157 were introduced in eq .
( [ d_lr ] ) . according to eq .
( [ g_trjl ] ) conductance is determined by operator @xmath322 .
the formal representation ( [ jlr_kk ] ) of the current operator @xmath323 in @xmath58-space is not suitable for practical implementation .
it can be used , however , to find an equivalent representation in @xmath155-space . using ( [ jlr_kk ] ) together with representation of operator @xmath158 [ eqs .
( 23 ) , ( 24 ) ] we can decompose @xmath323 in two terms : @xmath324 where @xmath325^{\dagger}\ ] ] has only positive ( non - negative ) eigenvalues , while @xmath326^{\dagger}\ ] ] has only negative ( or zero- ) eigenvalues , and @xmath327 are diagonal matrices with nonnegative elements [ eqs . (
23 ) , ( 24 ) ] .
the advantage of operators @xmath328 over operator @xmath329 is that matrices @xmath330 are positive definite , while anti - symmetric matrix @xmath158 is not .
as we show below in appendix b , that makes possible to find for operators @xmath331 an equivalent , site - angular momentum @xmath155 representation , while it is not the case for full operator @xmath332 .
we mention also , that in the asymptotic limit the contribution due to @xmath333 vanishes .
however , in general case , spectra of both operators should be found . to clarify these ideas ,
let us elucidate a physical meaning of eq .
( [ jl_2tems ] ) .
consider new basis functions in @xmath155-space which are constructed with help of unitary matrix @xmath334 , @xmath335_{nl,\alpha},\ ] ] so that the velocity operator @xmath336 in @xmath155-space introduced in eq .
( 21 ) is diagonal in the new basis : @xmath337_{\alpha\beta } } \\ \nonumber & = & \int_{s_{\mathrm{l } } } ds \left [ \chi^{*}_{\alpha}(\mathbf{r},e)\ , i\!\stackrel{\leftrightarrow}{\partial_{z } } \chi_{\beta}(\mathbf{r},e)\right ] \ = \
\delta^{\circ}_{\alpha}(e ) \ , \delta_{\alpha\beta}.\end{aligned}\ ] ] let us examine further the wave function @xmath264 which defines the current operator @xmath316 [ eq .
( [ jl_def ] ) ] .
this function is the solution of the lippmann - schwinger equation associated with the in - coming unperturbed bloch wave @xmath338 ( initial channel ) propagating from the right ( r ) lead towards the nanocontact and being scattered on it . within the left ( l ) lead ( @xmath339 ) we have : @xmath340 where @xmath341 here the ( @xmath342)-sums run over the half of the indices of the basis functions @xmath343 , corresponding either to the `` positive '' or to the `` negative '' window of the spectrum
thus , the wave function @xmath264 in the l lead is a linear combination of two functions .
the first one , @xmath344 carries the flux in the initial direction of the in - coming wave with momentum @xmath345 , from the l electrode to the r one .
the second function , @xmath346 , carries the flux in the direction being opposite to the propagation direction of the in - coming wave .
an example illustrating this idea for free electrons is shown in fig .
one can check , that operator @xmath347 is defined on functions @xmath344 only , while @xmath348 is defined on @xmath349 : @xmath350_{\boldsymbol{kk ' } } = 2\pi \int\limits_{s_l } ds \left [ \overline\phi^{(\pm)\,*}_{\boldsymbol{k}}(\mathbf{r},e)\ , i\!\stackrel{\leftrightarrow}{\partial_{z } } \overline\phi^{(\pm)}_{\boldsymbol{k'}}(\mathbf{r},e)\right],\ ] ] because , according to eqs .
( [ dl_delta ] ) and ( [ phi_k_pm ] ) , the cross - terms involving both functions , @xmath351 and @xmath352 , vanish .
obviously , that if the @xmath93 plane is chosen way behind the scattering region , the @xmath346 turns to zero and the contribution due to @xmath353 vanishes . in general case
, both operators , @xmath354 and @xmath355 , should be considered .
in this section we show that transmission probability of the conduction eigenchannel @xmath66 is given by @xmath356 , where @xmath357 and @xmath358 are positive and negative eigenvalues of operators @xmath359 and @xmath333 , respectively .
assume , that we know matrix @xmath71 ( introduced in sec.iii ) which is the unitary transform to the basis of eigenchannels .
according to ref .
we have : @xmath360 thus , we have decomposed the wave function of each eigenchannel in two terms : @xmath361 .
consider further the transmission matrix @xmath191 in the basis of eigenchannels : @xmath362_{\nu\mu } \\ & = & 2\pi \int_{s_{\mathrm{l } } } ds\ \left[\overline\phi_{\nu}^{\,*}(\mathbf{r},e)\ i\!\stackrel{\leftrightarrow}{\partial_{z } } \overline\phi_{\mu}(\mathbf{r},e ) \right ] \\
& = & t^{(+)}_{\nu\mu } + t^{(-)}_{\nu\mu},\end{aligned}\ ] ] with @xmath363.\ ] ] operators @xmath364 are related to current operators @xmath331 via the unitary transform @xmath71 : @xmath365 .
the unitary transform @xmath71 diagonalizes the full operator @xmath366 . using the representation ( [ jlr_kk ] ) of operators @xmath331 we can check , that @xmath359 and @xmath333 do not commutate . thus , the @xmath364 operators do not have a diagonal form in the basis of eigenchannels .
therefore , we represent each of these operators as sum of diagonal and off - diagonal terms : @xmath367 here @xmath368 , and off - diagonal contributions are of different signs because the sum of two matrices is the diagonal matrix @xmath369 .
the diagonal terms are @xmath370 where @xmath371.\ ] ] and transmission probability of the eigenchannel is given by @xmath372 . here
the first term arises due to all multiple scattering contributions in the direction of the current , while the second term is due to scattering contributions in the opposite direction . in spite of the fact that operators @xmath373 and @xmath374 can not be diagonalized within the unique unitary transform
, one can show that off - diagonal terms @xmath375 are small .
they are determined by functions @xmath376 with negative velocities with respect to the initial velocity of the in - coming channel @xmath66 incident from the right ( r ) lead .
the function @xmath376 in the left ( l ) lead is small as compared with @xmath377 .
when the @xmath93 plane is placed far enough from the surface , the @xmath376 collects all multiple scattering events in the opposite direction to the current flow .
however , such scattering processes are possible only due to small inhomogeneities of the potential which is not exactly the bulk one around @xmath378 ( see fig . 3 ) .
thus , we can write down : @xmath379 where parameter @xmath380 has a meaning of reflection amplitude ( see caption of fig . 3 ) .
the @xmath381 is of the order of reflection probability , which ( in realistic calculations ) is @xmath382 .
thus , off - diagonal terms , @xmath383 , contain small parameter @xmath384 .
since @xmath385 , according to the perturbation theory @xcite the difference between eigenvalues of operators @xmath386 [ eq.([tpm_full ] ) ] and @xmath387 [ eq.([t0_diag ] ) ] _ appears only in the second order _ , which is @xmath388 .
therefore , the operators @xmath331 provide a feasible way to find @xmath389 with good enough precision . to find the spectrum of positive definite operators
@xmath390 we represent them in the form @xmath391 , where @xmath392 maps the site - orbital space on the @xmath58-space
. we can prove ( see appendix b ) that spectrum of @xmath393 is the same as for operator @xmath394 which is defined in @xmath155-space and reads as @xmath395^{\dagger } { \cal{o}}^{\,\mathrm{lr}}\widetilde{x}^ { * } \left(\pm d^{\pm}_{\mathrm{l}}\right)^{1/2}.\ ] ] here we have used the unit operator in @xmath58-space @xmath396 introduced in eq .
( [ o_kk ] ) . by using equations ( [ tau1_kk ] ) , ( [ o_kk ] ) , ( [ jlr_kk ] ) , ( [ xc_1 ] ) , we obtain expression for operator @xmath397 : @xmath398^{\dagger } \left [ \tau^{-1}_{\mathrm{lr}}\ , { \cal{j}}_{\mathrm{r } } \bigl(\tau^{\dagger}_{\mathrm{rl}}\bigr)^{-1 } \right ] \widetilde{x}^ { * } \left(\pm d^{\pm}_{\mathrm{l}}\right)^{1/2 } \\ & = & \pm \left(\pm d^{+}_{\mathrm{l}}\right)^{1/2 } \bigl[\widetilde{\mathscr{c}}^{*}\bigr]^{\dagger } \left(\widetilde{\cal{b}}^{*}\ , g_{\mathrm{lr}}\ , { \cal{b}}^{\dagger } \right ) \mathscr{c}\ , d_{\mathrm{r}}\ , { \mathscr{c}}^{\dagger } \\ & & \times \left({\cal{b}}\ , g^{\dagger}_{\mathrm{rl}}\ , [ \widetilde{\cal{b}}^{*}]^{\dagger } \right ) \widetilde{\mathscr{c}}^ { * } \left(\pm d^{\pm}_{\mathrm{l } } \right)^{1/2 } \\ & = & \pm \left(\pm d^{\pm}_{\mathrm{l}}\right)^{1/2}\ , g_{\mathrm{lr}}\ , d_{\mathrm{r } } g^{\dagger}_{\mathrm{rl } } \left(\pm d^{\pm}_{\mathrm{l}}\right)^{1/2},\end{aligned}\ ] ] where orthogonality relation@xcite @xmath399^{\dagger } \widetilde{\cal{c}}^ { * } = 1 $ ] was used .
thus , we come to eq .
( 30 ) of sec .
let us consider operator @xmath400 acting in @xmath58-space of a following kind : @xmath401 where @xmath402 is a positive definite hermitian operator acting in the site - orbital @xmath155-space .
let @xmath403 be a unitary transform which diagonalizes operator @xmath404 , @xmath405 our goal is to find a spectrum of @xmath404 with help of @xmath155-space .
following sec.iii , we represent @xmath404 in the form : @xmath406 with @xmath407 and @xmath408 .
a square - root from the positive definite hermitian operator is defined as @xmath409 where the unitary matrix @xmath165 transforms @xmath402 to the diagonal form @xmath410 .
let us show that a spectrum of operator @xmath411 acting in the @xmath58-space , is the same as a spectrum of operator @xmath412 acting in the @xmath155-space .
let @xmath413 be such matrix that @xmath414 solution for @xmath413 is @xmath415 .
matrix @xmath413 diagonalizes @xmath416 : @xmath417 to finish the proof , one has to check that @xmath413 is indeed the unitary transform , i.e. the following properties should hold : ( i ) @xmath418 , and ( ii ) @xmath419 . first , we check property ( i ) : @xmath420 next , we check property ( ii ) : @xmath421^{-1 } \sigma^{\dagger } = 1,\end{aligned}\ ] ] so the proof is complete . ya .
m. blanter and m. bttiker , physics reports * 336 * , 1 ( 2000 ) ; m. bttiker , phys . rev .
b * 46 * , 12485 ( 1992 ) ; m. bttiker , phys .
lett . * 57 * , 1761 ( 1986 ) ; m. bttiker , y. imry , r. landauer , and s. pinhas , phys .
b * 31 * , 6207 ( 1985 ) .
j. a. torres , j. i. pascual , and j. j. senz , phys .
b * 49 * , 16581 ( 1994 ) ; j. a. torres , and j. j. senz , phys .
lett . * 77 * , 2245 ( 1996 ) ; a. m. bratkovsky and s. n. rashkeev , phys .
b * 53 * , 13074 ( 1996 ) .
n. kobayashi , m. brandbyge and m. tsukada , jpn .
. phys . * 38 * , 336 ( 1999 ) ; n. kobayashi , m. brandbyge and m. tsukada , phys .
b * 62 * , 8430 ( 2000 ) ; n. kobayashi , m. aono , and m. tsukada , phys .
b * 64 * , 121402(r ) ( 2001 ) ; k. hirose , n. kobayashi , and m. tsukada , phys .
b * 69 * , 245412 ( 2004 ) .
a. nakamura , m. brandbyge , l. b. hansen , and k. w. jacobsen , phys .
* 82 * , 1538 ( 1999 ) ; k. s. thygesen , m. v. bollinger , and k. w. jacobsen , phys . rev .
b * 67 * , 115404 ( 2003 ) ; p. jelnek , r. prez , j. ortega , and f. flores , phys .
b * 68 * , 085403 ( 2003 ) ; y. fujimoto and k. hirose , phys .
b * 67 * , 195315 ( 2003 ) ; k. palots , b. lazarovits , l. szunyogh , and p. weinberger , phys .
b * 70 * , 134421 ( 2004 ) ; p. a. khomyakov and g. brocks , phys .
b * 70 * , 195402 ( 2004 ) .
m. brandbyge , j .- l .
mozos , p. ordejn , j. taylor , and k. stokbro , phys .
b * 65 * , 165401 ( 2002 ) ; j. taylor , h. guo , and j. wang , phys .
b * 63 * , 245407 ( 2001 ) ; h. mehrez , a. wlasenko , b. larade , j. taylor , and p. grtter , h. guo , phys . rev .
b * 65 * , 195419 ( 2002 ) .
a. pecchia and a. di carlo , rep .
phys . * 67 * , 1497 ( 2004 ) ; g. c. solomon , a. gagliardi , a. pecchia , th .
frauenheim , a. di carlo , j. r. reimersa , and n. s. hush , j. chem .
* 125 * , 184702 ( 2006 ) ; s. kurth , g. stefanucci , c .- o .
almbladh , a. rubio , and e. k. u. gross , phys .
b * 72 * , 035308 ( 2005 ) ; c. verdozzi , g. stefanucci , and c .- o .
almbladh , phys .
lett . * 97 * , 046603 ( 2006 ) .
r. zeller , p. h. dederichs , b. ujfalussy , l. szunyogh and p. weinberger , phys .
b * 52 * , 8807 ( 1995 ) ; r. zeller , phys . rev .
b * 55 * , 9400 ( 1997 ) .
n. papanikolaou , r. zeller , and p. h. dederichs , j. phys . : condens .
matter * 14 * , 2799 ( 2002 ) .
we note , that a similar result holds for the eigenfunctions @xmath422 of the system defined as superposition of scattered states coming from the right lead : @xmath423 , where the perturbed bloch state @xmath424 is the solution of eq .
( [ ls ] ) corresponding to the initial in - coming state @xmath425 in r ( @xmath33 ) .
the notation with bar , @xmath422 , is used to distinguish from @xmath426 .
taking into account the property of microscopic reversibility for the transmission amplitudes , @xmath427 , and eqs .
( [ theta ] ) and ( [ curr_bloch1 ] ) , we obtain : @xmath428_{z\to -\infty } = -\ , \frac{t_{\nu}(e)}{2\pi}\ , \delta_{\nu\mu}.\ ] ] since cell potential @xmath99 is real , both regular @xmath429 and irregular @xmath430 solutions of the radial schrdinger equation can be found as real valued functions .
however , in practical implementation of the kkr method , we find @xmath431 as solution of the lippmann - schwinger equation for the incoming spherical bessel function @xmath432 , where @xmath433 is a real spherical harmonic .
for example , in case of spherical potentials ( asa ) employed in this work , such defined functions @xmath431 carry multipliers @xmath434 $ ] , where @xmath435 is a scattering phase shift . without loss of generality
, these phase factors can be ascribed to structure constants @xmath436 [ see eq .
( [ grr ] ) ] and solutions @xmath429 can be considered as real valued functions .
this expression reads @xmath437 @xmath438 $ ] , where a matrix @xmath404 is introduced : @xmath439 $ ] , here @xmath440 is a volume of a unit cell , whereas @xmath441 is number of atoms in a supercell . because of orthogonality of the basis functions , @xmath442 \sim \delta_{ll'}\ , \delta(e - e')$ ] , the major contribution to eq .
( [ phi_nl ] ) comes from @xmath443-states with @xmath444 .
this property is used to prove the second relation in eq .
( [ bc ] ) .
w. f. egelhoff , jr . ,
l. gan , h. ettedgui , y. kadmon , c. j. powell , p. j. chen , a. j. shapiro , r. d. mcmichael , j. j. mallett , t. p. moffat , m. d. stiles , and e. b. svedberg , j. appl .
phys . * 95 * , 7554 ( 2004 ) .
recent _ ab intio _
calculations by burton _
( see above ref . )
predict 340% domain wall mr in case of one - dimensional , fcc @xmath445 , ni wire .
this result follows from the quantized conductance being @xmath446 in case of uniformly magnetized wire reduced down to @xmath447 in the presence of domain wall .
however , one has to take into account that transmission probabilities of many wire channels even without domain wall will be significantly reduced ( especially those which are built from @xmath212 electrons ) when realistic geometry of a contact is considered . that can diminish mr to quite moderate values comparable to ones estimated in the present work . without a loss of generality , the number @xmath449 of @xmath58-points on the isoenergetic surface @xmath450 plus number of ( possibly existing ) evanescent states ( with energy @xmath104 )
@xmath451 can be chosen to be @xmath452 where @xmath453 is number of atoms in the l ( or r ) atomic planes , so that the inverse matrix operation is well defined .
even if @xmath454 is not exactly equal to @xmath455 the rectangular matrix @xmath291 $ ] can be found as the pseudoinverse matrix ( ref . ) to the matrix with @xmath294- and @xmath295-blocks . from eq .
( [ eq_cg_bd ] ) we obtain : @xmath456 , and @xmath457 . in order the @xmath5-sum
be converged , matrix elements of @xmath294 , @xmath458 , @xmath295 and @xmath459 should carry coefficients @xmath460 ( where @xmath461 is number of atoms per cross - section of the born - von krmn supercell ) . on the other hand ,
since matrix @xmath462 $ ] is the inverse to the one with @xmath294 and @xmath295 blocks , the following relation holds : @xmath463 .
the second term in this equation is zero , because the number of evanescent ( surface ) states @xmath464 at given energy @xmath104 is much smaller than the total number of surface states @xmath465 , so that @xmath466 .
finally , we obtain : @xmath467 . | we develop a formalism for the evaluation of conduction eigenchannels of atomic - sized contacts from first - principles .
the multiple scattering korringa - kohn - rostoker ( kkr ) green s function method is combined with the kubo linear response theory .
solutions of the eigenvalue problem for the transmission matrix are proven to be identical to eigenchannels introduced by landauer and bttiker .
applications of the method are presented by studying ballistic electron transport through cu , pd , ni and co single - atom contacts .
we show in detail how the eigenchannels are classified in terms of irreducible representations of the symmetry group of the system as well as by orbital contributions when the channels wave functions are projected on the contact atom . |
there might be two kinds of dibaryon@xcite .
one is the loosely bound type consisting of two octet baryons ; the deuteron is a typical example . the others are tightly bound ;
the @xmath4 particle had been predicted to be such a six quark state although later calculations cast doubt on it@xcite . instead , a non - strange @xmath5 @xmath6 and a strangeness -6 @xmath7 di-@xmath2 have been predicted to be tightly bound six quark states , which are formed from decuplet baryons@xcite . the strangeness -3 @xmath1 n@xmath2 has also been predicted to be of the tightly bound type@xcite .
the tensor interaction due to @xmath8 exchange plays a vital role in the formation of loosely bound deuteron . in the @xmath6 case
the tensor interaction contribution to its mass is minor but is critical for its d - wave decay to the nn final state@xcite .
there are other near threshold and deeply bound dibaryon candidates found in two systematic quark model calculations@xcite .
this naturally raises the question as to whether or not the tensor interaction adds sufficient strength to bind these other near threshold states to become strong interaction stable , as in the deuteron case ?
conversely , is the tensor interaction weak enough to leave the high spin , deeply bound states as narrow dibaryon resonances , as was shown in the @xmath6 case@xcite ?
the present calculation is aimed at answering these two questions for the dibaryon candidates in the u , d , s three flavor world within the extended quark delocalization , color screening model ( qdcsm ) .
our results show that both the effective s - d wave transition interactions due to gluon and @xmath8 exchanges decrease rapidly with increasing strangeness , and that the tensor contributions of k and @xmath0 mesons are negligible after a short range truncation . altogether , the tensor contributions are not strong enough to bind other near threshold six quark states , such as the @xmath9 , to become strong interaction stable with the sole exception of the deuteron .
the d - wave decay widths of high spin , six quark states , such as the @xmath10 and the @xmath11 , are in the range of tens of mev to tens of kev and so these states might be narrow dibaryon resonances .
the extended qdcsm is briefly introduced in section ii . in section iii , we present our results . the discussion and conclusion are given in section iv .
the qdcsm was put forward in the early 90 s .
details can be found in refs.@xcite .
although the short range repulsion and the intermediate range attraction of the @xmath12 interaction are reproduced by the combination of quark delocalization and color screening , the effect of the long - range pion tail is missing in the qdcsm .
recently , the extended qdcsm was developed@xcite , which incorporates this long - range tail by adding @xmath8-exchange but with a short - range cutoff to avoid double counting because the short and intermediate range interactions have been accounted for by the quark delocalization and color screening mechanism@xcite .
the exchange of k and @xmath0 mesons has been shown to be negligible in this model approach@xcite .
nevertheless , their effect , especially the tensor part , has been included in this calculation to check further whether they are negligible in our model approach .
the extended qdcsm not only reproduces the properties of the deuteron well , but also improves agreement with nn scattering data as compared to previous work@xcite .
the hamiltonian of the extended qdcsm , wave functions and the necessary equations used in the current calculation are given below .
the tensor interactions due to effective one gluon and the octet goldstone boson exchanges are included .
the details of the resonating - group method ( rgm ) have been presented in refs.@xcite .
the hamiltonian for the 3-quark system is the same as the well known quark potential model , the isgur model .
for the six - quark system , we assume @xmath13 , \nonumber \\ v_g(r_{ij } ) & = & \alpha_s \frac{\vec{\lambda}_i \cdot \vec{\lambda}_j } { 4 } \left [ \frac{1}{r_{ij}}-\frac{\pi}{2 } \delta ( \vec{r_{ij } } ) \left ( \frac{1}{m^2_i}+\frac{1}{m^2_j}+\frac{4\vec{\sigma}_i \cdot \vec{\sigma}_j}{3m_im_j } \right)+\frac{1}{4m_im_jr_{ij}^3 } s_{ij } \right ] , \nonumber \\ v_{\pi}(r_{ij } ) & = & \theta ( r - r_0 ) \frac{g_8
^ 2}{4\pi}\frac{m_{\pi}^2 } { 4m_q^2 } \frac{1}{r_{ij } } e^{-m_{\pi } { r_{ij}}}\left[\frac{1}{3 } \vec{\sigma}_i \cdot \vec{\sigma}_j + z(r_{ij})s_{ij}\right]\vec{\tau}_i\cdot \vec{\tau}_j , \label{hamiltonian } \\
s_{ij } & = & 3\frac{\vec{\sigma}_i \cdot \vec{r}_{ij } \vec{\sigma}_j \cdot \vec{r}_{ij}}{r_{ij}^2}-\vec{\sigma}_i \cdot \vec{\sigma}_j , \nonumber \\
z(r ) & = & \frac{1}{3}+\frac{1}{m_{\pi}r}+\frac{1}{(m_{\pi}r)^2 } , \nonumber \\ v_{conf}(r_{ij } ) & = & -a_c \vec{\lambda}_i \cdot \vec{\lambda}_j \left\ { \begin{array}{ll } r_{ij}^2 & \qquad \mbox{if } i , j\mbox { occur in the same baryon orbit } , \\ \frac{1 - e^{-\mu r_{ij}^2 } } { \mu } & \qquad \mbox{if } i , j\mbox { occur in different baryon orbits } , \end{array } \right . \nonumber
\\ \theta ( { r_{ij}}-r_0 ) & = & \left\ { \begin{array}{ll } 0 & \qquad r_{ij } < r_0 , \\ 1 & \qquad \mbox{otherwise } , \end{array } \right .
\nonumber\end{aligned}\ ] ] where @xmath14 is the short range cutoff for pion exchange between quarks .
@xmath15 is the @xmath8 quark coupling constant .
@xmath16 is the measured @xmath8 mass .
the k and @xmath0 meson exchange interactions , which have not been shown explicitly in the above equation but have been included in this calculation , have a form very similar to that for the @xmath8 .
the color screening constant , @xmath17 , is to be determined by fitting the deuteron mass in this model .
all other symbols have their usual meanings , and the confinement potential @xmath18 has been discussed in refs.@xcite .
the pion exchange interaction , @xmath19 , affects only the @xmath20 and @xmath21 quarks . we take these to have a common mass , @xmath22 , i.e. , ignoring isospin breaking effects .
the quark wave function in a given nucleon ( orbit ) relative to a reference center ( defined by @xmath23 ) is taken to have a gaussian form characterized by a size parameter , @xmath24 , @xmath25 the light quark mass , @xmath26 , is chosen to be @xmath27 of the nucleon mass .
the strange quark mass , @xmath28 , baryon size parameter , @xmath24 , effective quark - gluon coupling constant , @xmath29 , and the strength of confinement , @xmath30 , are all determined by reproducing the nucleon mass , the @xmath31 mass difference , an overall fit to other ground state baryon masses and by requiring the nucleon mass to be variational stable with respect to its size parameter , b. the quark - pion coupling constant @xmath32 is obtained from the nucleon - pion coupling constant by a slight ( @xmath33% ) correction to the classic symmetry relation , viz .
, @xmath34 where @xmath35 is the nucleon mass and the last factor provides the correction due to the extent of the quark wavefunction in the nucleon .
the k and @xmath0 are assumed to have a flavor su(3 ) symmetric quark - meson coupling constant and the same short range cutoff , @xmath14 , as the @xmath8 s .
the color screening parameter , @xmath17 , has been determined by matching our calculation to the mass of the deuteron .
all of the model parameters are listed in table i. table i : model parameters [ cols="^,^,^,^,^,^,^",options="header " , ] from table iv we see that the width of @xmath6 decay to the nn d - wave is 6.57 mev .
comparison with our previous results@xcite , confirms that the width is not sensitive to the value of cutoff @xmath14 .
the width of @xmath36 decays to @xmath3 d - wave is about tens of kev , about three orders of magnitude smaller .
the result is not changed significantly for decay channels with different spin .
for example , the n@xmath37 ( spin=0 ) d - wave decay width is calculated to be 12.0 kev , and the n@xmath38 ( spin=1 ) d - wave decay width is 21.9 kev .
these results confirm our expectation that the n@xmath2 is a narrow dibaryon resonance .
( the width of n@xmath37 is smaller than that of the @xmath39 nn decay mainly due to the reduction of the tensor interaction , but also due to the fact that the n and @xmath40 have the same flavor content while n , @xmath2 and @xmath41 , @xmath42 have differing flavor content in each baryon . )
the effects of the tensor interactions of gluon and goldstone boson exchanges on the dibaryon mass and decay width have been studied in the extended qdcsm . only in the deuteron channel
is the tensor interaction of @xmath8 exchange strong enough to bind the two nucleons into a loosely bound state .
no other near threshold six - quark state studied so far in the u , d , s three flavor world can be bound together by the additional attraction induced by these tensor interactions . the s - d wave effective transition interactions due to @xmath8 and gluon tensor forces both decrease rapidly with increasing strangeness . in the @xmath43 and @xmath44 channels
the effective transition interactions due to the gluon tensor term become repulsive and cancel the attractive @xmath8 contribution .
the tensor contribution of explicit k and @xmath0 meson is confirmed as negligible due to the same short range truncation as for the @xmath8 , in our model approach@xcite the mass shift of the @xmath1 n@xmath2 state induced by the tensor interaction is small ( not more than 1 mev ) and the d - wave partial decay widths to @xmath3 with spin-0 and spin-1 are only 12.0 and 21.9 kev , respectively .
hence , the @xmath1 n@xmath2 state appears to be a good candidate for a narrow dibaryon resonance .
altogether there are only two promising narrow dibaryon resonances in the light flavor world in our model approach : the @xmath5 @xmath6 and the @xmath1 n@xmath2 .
the @xmath4 particle and di-@xmath2 are marginally strong interaction stable in our model .
however the theoretical binding energies of both are small ( only few mev@xcite ) .
table ii shows that the calculated ground octet and decuplet baryon masses deviate from the measured ones about 18 mev on average
. a reasonable estimate of the model uncertainty for the dibaryon mass would be at least that large .
therefore , in our model , it is unjustified to assert that the @xmath4 particle and di-@xmath2 might be strong interaction stable dibaryon candidates .
this estimation is consistent with the latest di-@xmath41 hypernuclear findings@xcite .
there are various broad resonances with widths @xmath45 mev around the @xmath6 mass ( @xmath46 2180 mev ) in our model which makes the analysis of the nn scattering more difficult in the energy region 2.1 @xmath47 2.4 mev where a broad bump has been found in the pp and np total cross sections .
we will report on those results later .
the @xmath48 n@xmath2 state is quite convincingly lower in mass than the n@xmath2 threshold , and quite possibly lower than the @xmath49 threshold , as well .
we have shown the decay width to be as small as tens of kev .
such a narrow dibaryon resonance might be detected by reconstructing the invariant mass of its two body decay products , @xmath41 and @xmath42 , in high @xmath2 production reactions using rhic at brookhaven and compass at cern and the future ones at jhf in japan and fair in germany .
this model , the extended qdcsm , which proposes a new mechanism to describe the nn intermediate range attraction instead of the @xmath50 meson , well describes , with the fewest parameters , the properties of the deuteron and the existing nn , n@xmath41 and n@xmath51 scattering data . up to now
, it is the only model which gives an explanation of the long standing fact that the nuclear and molecular forces are similar in character despite the obvious length and energy scale differences and that nuclei are well described as collection of a nucleons rather than 3a quarks . in view of the fact that the @xmath4 particle has not been observed experimentally , the bb interaction in the @xmath52 channel@xcite predicted by this model may be a good approximation of the real world . based on these facts
we suppose the predictions about dibaryon states of this model might also be approximately correct . of course
, the qdcsm is only a model of qcd .
the high spin , high strangeness dibaryon resonance , @xmath1 n@xmath2 , may be a good venue to search for new hadronic matter and to test whether or not the qdcsm mechanism for the intermediate range attraction is realistic .
this work is supported by nsfc contracts 90103018 and 10375030 . and by the u.s .
department of energy under contract w-7405-eng-36 .
f. wang would like to thank the itp for their support through the visiting program .
99 f. wang , j.l .
ping and t. goldman , _ in _ aip conference proceedings 338 , ed .
s.j.seestrom , aip press , woodbury , new york , 1995 , p.538
. f. wang _
et al_. , phys .
* c51 * , 3411 ( 1995 ) .
jaffe , phys .
lett . * 38 * , 195 ( 1977 ) .
jaffe and f. wilczek , phys .
* 91 * , 232003 ( 2003 ) .
hep - ph/0307341 .
_ et al_. , nucl - th/0306043 .
t. goldman _
et al_. , phys . rev .
* c39 * , 1889 ( 1989 ) . t. goldman , _ et al_. , mod
. phys . lett .
* a13 * , 59 ( 1998 ) .
ping , h.r .
pang , f. wang and t. goldman , phys .
rev * c65 * , 044003 ( 2002 ) .
q.b . li _ et al_. , nucl
a683 * , 487 ( 2001 ) .
pang , j.l .
ping , f. wang and t. goldman , commun .
phys . * 38 * , 424 ( 2002 ) ; phys . rev . *
c66 * , 025201 ( 2002 ) .
t. goldman _
et al_. , phys .
lett . * 59 * , 627 ( 1987 ) .
et al_. , commun .
. phys . * 41 * , 67 ( 2004 ) .
j.l . ping , f. wang and t. goldman , nucl . phys . *
a688 * , 871 ( 2001 ) .
f. wang , g.h .
wu , l.j . teng and t. goldman , phys .
69 * , 2901 ( 1992 ) .
pang , j.l .
ping , f. wang and t. goldman , phys .
* c65 * , 014003 ( 2001 ) .
ping and f. wang , chin .
* 20 * , 42 ( 2003 ) .
buchmann , y. yamauchi and a. faessler , nucl . phys .
a496 * , 621 ( 1989 ) . f. wang , j. l. ping and t. goldman , phys . rev . * c51 * , 1648 ( 1995 ) .
j.l . ping , f. wang and t. goldman , phys . rev . * c62 * , 054007 ( 2000 ) . h. takahashi _
et al_. , phys
lett . * 87 * , 212502 ( 2002 ) . | the influence of gluon and goldstone boson induced tensor interactions on the dibaryon masses and d - wave decay widths has been studied in the quark delocalization , color screening model .
the effective s - d wave transition interactions induced by gluon and goldstone boson exchanges decrease rapidly with increasing strangeness of the channel .
the tensor contribution of k and @xmath0 mesons is negligible in this model .
there is no six - quark state in the light flavor world studied so far that can become bound by means of these tensor interactions besides the deuteron .
the partial d - wave decay widths of the @xmath1 n@xmath2 state to spin 0 and 1 @xmath3 final states are 12.0 kev and 21.9 kev respectively .
this is a very narrow dibaryon resonance that might be detectable in relativistic heavy ion reactions by existing rhic detectors through the reconstruction of the vertex mass of the decay product @xmath3 and by the compas detector at cern or at jhf in japan and the fair project in germany in the future . |
quantitative studies of galaxy demographics that is , of the multivariate distribution functions that connect global galaxy properties provide the empirical bedrock on which theoretical models of galaxy formation and evolution are founded .
the quality of a cosmological model of galaxy formation ( _ e.g. _ * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) is judged by its ability to reproduce the most basic demographics of real galaxy samples .
this includes univariate distributions like the mass or luminosity functions , and/or bivariate distributions like the size
mass , colour mass or mass density relations .
the field of galaxy formation and evolution is thus largely data - driven , and is likely to remain so for the foreseeable future .
it has long since been established that there exist a number of empirical ` laws ' that describe the remarkably tight correlations between most , if not all , of the global parameters used to describe galaxies : _
e.g. _ , luminosity , mass , colour , mean stellar age , star formation rate , size , shape , structure , morphology , dynamics , etc .
( _ e.g. _ * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
since 2000 , an explosion in the volume and quality of galaxy survey data at low- and high - redshifts has helped to consolidate and make concrete these insights , at least for high mass galaxies .
one of the most important and influential insights has been the idea that these scaling relations can be best understood as a unidimensional sequence in stellar mass ( _ e.g. _ , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) but see also , _
e.g. _ , @xcite , who argue that stellar surface density or central velocity dispersion may be the more fundamental parameter . in this picture ,
once mass is controlled for , local environment potentially plays an important but secondary role ( _ e.g. _ * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
there is also the longstanding idea that the galaxy population can be naturally subdivided into two ( and only two ) broad classes . even before galaxies came to be called
` galaxies ' , @xcite recognised that the ` extragalactic nebulae ' could be qualitatively separated into two distinct phenomenological classes , based on their morphologies .
broadly speaking , at fixed mass , the elliptical ` early type ' galaxies are older , redder , less likely to be star forming , and smaller than the ` late type ' spirals ( _ e.g. _ * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
in this way , there appear to be two ( and only two ) distinct sets of relations that describe how galaxies properties scale with mass ; one set for each of the early- and late - type populations . further ,
early types are , on average , more massive and more likely to be found in denser environments ( _ e.g. _ , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the idea has thus been that these two populations correspond to two ( and only two ) distinct evolutionary states .
one aspect of this ` bimodality'or , better , the dichotomy between the ` developing ' and ` developed ' galaxy populations has been particularly influential , inasmuch as it has received a huge amount of attention from observers and modellers alike . in order to reproduce the distributions of galaxy _ colours _ ( _ e.g. _ , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , and in particular the evolving mass functions ( mfs ) of red / blue galaxies ( _ e.g. _ , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , cosmological models have had to introduce an _ ad hoc _ ` quenching ' mechanism ( or mechanisms ) to disrupt or prevent star formation . within the models ,
these inclusions act on more massive galaxies and/or galaxies in denser environments , either by the removal / expulsion of the existing gas reservoir , or by preventing the accretion of new material .
the physical nature of the quenching process remains controversial .
the most popular candidates are energetic ` feedback ' from an agn ( _ e.g. _ * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , or a distinction between ` hot- ' and ` cold - mode ' accretion ( _ e.g. _ * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) resulting from the presence or absence of persistent shock - fronts in infalling gas . the quenching mechanism is usually taken to be linked to halo mass , and may or may not have an environmental component ( _ e.g. _ * ? ? ?
. with the above as background , our immediate goal in this paper is to derive a quantitative , phenomenological description of the bivariate colour mass distribution function for field galaxies in the local universe , with particular emphasis on the colour mass relations ( cmrs ) and mass functions ( mfs ) for the redder and bluer of the galaxy subpopulations .
in essence , our goals are similar to those of @xcite , who set out to quantitatively model the two apparently distinct but overlapping ` red ' and ` blue ' populations seen in the colour magnitude plane . the colour magnitude diagram is astronomy s most basic diagnostic plot . for galaxies , as a measure of the integrated starlight ,
magnitude is typically taken as a proxy for total stellar content ; _ i.e. _ , stellar mass .
colour is a simple observable diagnostic parameter that characterises galaxies stellar populations . in particular , modulo dust
, colour acts as a proxy for the luminosity - weighted mean stellar age , which can also be taken as an average specific star formation rate ( sfr ) over long ( @xmath5 gyr ) timescales .
our analysis improves on that of @xcite in two ways .
first , we use the results of stellar population synthesis ( sps ) modelling of broadband spectral energy distributions ( seds ) , rather than simple restframe luminosities and colours .
specifically , we use sps - derived stellar mass estimates as our proxy for total stellar content , and we use dust - corrected intrinsic stellar colour as ( at least in principle ) a more direct tracer of galaxies stellar populations . second , we extend the @xcite analysis by developing and applying a statistically rigorous mixture modelling formalism to derive a quantitative , phenomenological description of the bimodality in galaxies stellar populations .
the crux of the problem is that the ( optical ) colour distributions of the apparently distinct red and blue populations are seen to overlap . in the first instance
, this presents an operational problem : how best to disentangle these two populations . in the second instance , the fact of overlap makes it difficult to interpret the terms ` red ' and ` blue ' in concrete , astrophysical terms .
given the role that these kinds of observations have in guiding theories of galaxy formation and evolution , a secondary goal of this work is to elucidate some of the important conceptual subtleties and difficulties inherent to this kind of analysis , which are too often glossed over if not ignored altogether .
our discussion proceeds in four parts , as follows .
the gama data and our basic analysis of them are laid out in and .
we discuss our ability to meaningfully constrain dust obscurations and intrinsic stellar colours in and fig .
[ fig : colours ] . the ( limited ) role of incompleteness and selection effects in our results are discussed in , as well as . in the second part ,
we motivate and describe our approach to the problem . in
, we show how and why previous studies have found qualitatively and quantitatively different results for the red / blue mfs : namely , the different and almost always arbitrary
ways that the ` red ' and ` blue ' galaxy samples have been selected or defined .
the extent to which these results provide meaningful constraints on the process of galaxy formation and evolution is therefore limited by the extent to which the terms ` red ' and ` blue ' can be shown to be astrophysically meaningful .
this is why we have set out to derive objective and phenomenological , operational definitions for the terms ` red ' and ` blue ' .
in addition to the description of our modelling procedure given in , we provide a more pedagogical discussion of our approach in appendix [ app : themodel ] , in which we develop our analysis starting from a simple @xmath6-minimisation fit . this material is intended to help fast - track researchers intending to apply a similar mixture - modelling analysis to their own data .
the reader that is concerned about how we have decided to parameterise our fits to the bivariate colour mass distributions should focus on .
in the third part , we present the results of our descriptive modelling .
the quality of our fits is illustrated in fig .
[ fig : rfdist ] and fig .
[ fig : stardist ] , and discussed in .
the specific question of how to interpret our results at very low masses is discussed in . in , we present and discuss our characterisations of the cmrs for the two galaxy populations ( ) , the objective classification scheme that we derive from our modelling of the cmds ( ) , and the mfs for the two populations ( ) .
our most important astrophysical results and conclusions can be found in , in which we describe the essential characteristics of the bimodal ( or , better , two - population ) distribution of galaxies stellar populations .
the fourth and final part comprises a discussion of our results and methods ( ) .
we revisit the results of earlier studies in light of our analysis and results in , including illustrating how our objective , phenomenological red / blue classifications compare to those used previously ; this effectively closes the loop opened in . in
, we show how our objective classification scheme maps onto two commonly used diagnostic diagrams .
[ fig : nircols ] and fig .
[ fig : hdd4 ] provide important validation and illustration of how our objective classifications discriminate between galaxies with qualitatively different stellar populations .
finally , in we discuss potential objections to our analysis and results .
this paper is long .
given the increasing awareness of the need for more detailed and rigorous statistical analysis of large galaxy catalogues , our hope is that this paper will serve as a useful pedagogical resource for researchers working on similar problems in the future .
therefore , some of the more technical description and discussion of our statistical formalism may not be of interest to some readers ; or , for those researchers familiar with bayesian mcmc fitting techniques , for example , they may seem _ overly _ detailed . in recognition of this
, we have made efforts to make the structure of the paper as modular as possible , so that the reader can choose which sections to read closely , and which to skip altogether .
for the casual or first time reader , we make the following recommendations .
start with fig .
[ fig : colours ] and its caption .
then , read and the opening of for the motivation for our analysis , and an outline of the basic assumptions that underpin our approach . and are particularly important , in that they provide our rationale for favouring the more neutral designations ` b ' and ` r ' , in place of the more laden terms ` blue ' and ` red ' .
next , move to and , which offer an intuitive way of understanding how a mixture modelling approach can be used to characterise the two populations without ever specifying which galaxies belong to which population . after reviewing fig.s [ fig : rfcmr][fig : starmf ] and their captions , move to , in which we discuss our main results and conclusions in astrophysical terms .
, in which we show that our r - type galaxies really do have different and much older stellar populations than b - type galaxies , is very important . , and
especially , in which we compare our results to those of @xcite , is also important for readers interested in the problem of quenching .
readers that remain concerned about the validity of our methods and results as well as those of previous studies should read carefully . a summary of our analysis , results , and conclusions is given in .
[ fig : mcmc ] serves as a table of results for the various parameters that we have fit for .
machine - readable tables of the results shown in fig.s [ fig : rfcmr][fig : starmf ] are made available as additional online content .
we are happy to provide the source code for our modelling on request . throughout
, we adopt the concordance cosmology : @xmath7 = @xmath8 .
all stellar mass estimates have been derived assuming , or have been approximately scaled to match , a @xcite stellar initial mass function ( imf ) .
all magnitudes are expressed in the ab system .
finally , a note on notation : in the more technical sections of this work , which describe the formal basis and justification for our modelling , we will represent vectors as @xmath9 , matrices as @xmath10 , and sets as @xmath11 , as distinct from scalar quantities like @xmath12 , @xmath13 , @xmath14 , @xmath15 , @xmath16 , or @xmath17 .
as an optical spectroscopic survey , the galaxy and mass assembly ( gama ; * ? ? ?
* ; * ? ? ?
* ) survey has now completed its observations of three separate equatorial fields of 60 @xmath18 each .
the spectroscopic target selection is described by @xcite .
targets have been selected on the basis of dust - corrected petro @xmath19-band magnitudes from the sdss dr7 @xcite . for gama - ii , all three fields
have been surveyed to a depth of @xmath20 mag . in gama - ii nomenclature , these define the ` survey_class ` @xmath21 sample selection limits .
the gama survey strategy @xcite has been optimised for uniform and near total spectroscopic completeness ( @xmath22% ) , even in regions with high target density .
targeting completeness is better than 99.9% , with only 160/189059 main survey targets not having been observed . as a function of the sdss @xmath19-band ` fiber ` magnitudes ,
redshift success is 99 and 95% for @xmath23 and 20.5 respectively , where success is defined as @xmath24% confidence that the given redshift is correct . whereas previously , the gama spectroscopic redshifts were based on by - eye determinations done by observers at the telescope ,
the spectroscopic redshifts given in the gama - ii catalogues have been derived using an automated pipeline , as described by @xcite .
this has reduced the standard redshift error from @xmath25 to @xmath26 km / s , and reduced the redshift blunder rate for high confidence redshifts from @xmath27 to @xmath28% , as determined through comparisons between repeat observations of gama targets , and through comparisons between gama and sdss observations of common targets . for the purposes of calculating luminosity and comoving distances ,
these heliocentric redshifts have been corrected for local bulk flows using the model of @xcite for the very lowest redshifts ( @xmath29 ) , and then tapering to a cosmic microwave background ( cmb)-centric frame for @xmath30 .
the details of this conversion are given by @xcite .
it is these flow - corrected redshifts that we will use as the basis of our analysis , including sample selection .
the photometric backbone of the gama - ii dataset comprises optical _ ugriz _ imaging from sdss ( dr7 ) and near infrared _ zyjhk _ imaging from the viking survey .
the sdss data have been extensively described ( see , _
e.g. _ , * ? ? ?
* ; * ? ? ?
* ) , and have been obtained from the sdss data archive server .
the viking data reduction has been done by the cambridge astronomical data unit ( casu ) pipeline for vista , and have been obtained from the vista science archive ( vsa ; * ? ? ?
the gama - ii photometric catalogue is based on an independent reanalysis of these imaging data ( see * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . for the purpose of constructing multi - band spectral energy distributions ( seds ) , a set of psf - matched mosaics ( @xmath31 fwhm ) have been made .
these have been fed to sextractor @xcite , which has been run in dual - image mode , using the @xmath19-band mosaics as the detection images , to yield seeing- and aperture - matched _ ugrizyjhk _ seds .
comparisons between earlier versions of this photometry and the sdss model and petro photometry are presented by @xcite and @xcite . for this work ,
we have used the latest gama - ii photometric catalogue ( internal designation ` apmatchedcatv05 ` ) , which will be described by driver et al .
( in prep . ) . in comparison to the earlier ` apmatchedcatv01 ` catalogue presented by @xcite , the most significant change is the supersession of the ukidss las nir data with _
_ data from vst viking .
as is well known , the finite auto aperture is prone to miss a significant amount of flux for galaxies that are faint and/or have significant low - surface brightness wings . to account for this
, we characterise the total , observers frame @xmath19-band flux by fitting a srsic profile to the observed 2d light distributions for each galaxy . as described by @xcite ,
this has been done using galfit3 @xcite , incorporating a galaxy - specific model for the psf , and taking care to isolate and deblend the target from any and all nearby galaxies . in the fits
, the srsic profile has been truncated at 10 @xmath32 , which typically corresponds to a surface brightness limit of @xmath33 mag / @xmath34 . for this work
, we have used these srsic - fit estimates of total @xmath19-band flux , taken from the ` sersiccatv09 ` catalogue , to normalise the auto seds described above . at fixed srsic index , the rms in the values of these corrections - to - total
is of order 0.05 mag or less , even for the faintest apparent magnitudes .
the redshifts and multiband photometry described above have been combined to estimate stellar population parameters including masses , restframe photometry , luminosity - weighted mean stellar ages , dust obscurations , metalliticies , specific star formation rates , etc .
the basic procedure is the same for the gama - i masses presented by @xcite , but with one significant improvement . for gama - ii , each band is weighted such that the sps fits are done to a fixed _ restframe _ wavelength range of 300011000 , which corresponds roughly to restframe @xmath35@xmath36
. between this change , and the change from the ukidss to the viking nir data , the large systematic errors in the sps fits discussed at length in @xcite have been reduced significantly in the @xmath37 , @xmath36 , and @xmath38 bands .
this suggests that at least part of these problems were due to calibration errors in the ukidss data .
the issues with the @xmath39 and @xmath40 band data persist , but at a lower level , which is why we have not pushed further into the restframe nir for the sps fits . following standard practice , in the course of these fits , an arbitrary error floor is imposed on each photometric point by adding an additional uncertainty of 0.05 mag in quadrature to the catalogued photometric uncertainties .
this decision is typically justified as providing protection against both errors and uncertainties in the relative or cross calibration of the photometry in different bands , as well as against template mismatch and/or aliasing errors in the sps fitting .
given that these uncertainties are treated as being both random and independent , however , neither of these justifications are really well - founded . with the exception of the @xmath35-band photometry , the catalogued errors are almost always comparable to , or even less than , 0.05 mag .
this imposed error floor is thus the limiting factor in setting the formal uncertainties on our sps - derived results including , in order of increasing importance , @xmath41 , @xmath42 , @xmath43 , and @xmath44 .
in fact , as we will argue in below , this decision leads to drastic overestimates of the ` true ' errors on the intrinsic ( _ i.e. _ , dust corrected ) stellar colours @xmath44 .
the sed fitting process involves comparing the observed photometry to a library of synthetic stellar population spectra .
this stellar population library ( spl ) was constructed using the @xcite stellar evolution models for a @xcite imf , and making the following common simplifying assumptions : 1 . ) exponentially declining star formation histories , 2.)uniform , single screen dust attenuation , and 3 . )
uniform stellar metallicities .
the estimates for both the values of and uncertainties in the sp parameters for individual galaxies have been made in a bayesian way .
the rms difference between our values for @xmath45 and those from the mpa / jhu catalogues for sdss dr7 is @xmath46 dex , with no appreciable systematic differences as a function of colour , structure , mass , or apparent magnitude .
we note that the quantitative values of the mass estimates can be recovered to high precision ( @xmath47 error of @xmath48 dex ) using the following simple , empirical relation @xcite : @xmath49 = 1.15 + 0.70 \ , ( g - i ) - 0.4 ~ m_i ~ .\ ] ] in this way , the @xmath41@xmath50 colour stellar mass diagram can be transparently viewed as simple , linear shearing of the @xmath41@xmath51 colour magnitude diagram .
further , for the reader wishing to compare their data to our results , this relation offers a simple and transparent basis for comparison between our ( and by extension , the sdss ) stellar mass estimates .
galaxies in the colour stellar mass plane using restframe @xmath41 colour , and using the intrinsic ( _ i.e. _ , corrected for internal dust extinction ) stellar colour @xmath44 , respectively .
the middle panel shows the dust extinction , @xmath43 , as a function of mass , and can be thought of as linking the other two . within each panel ,
the blue and red points distinguish galaxies that do or do not show clear h@xmath52 line emission ; optically identified agn are highlighted yellow .
these spectral classifications are entirely independent of the sed - fit results .
note that the agn seem to be almost exclusively associated with the blue sequence in the @xmath44 cmd .
as expected , there is a strong correspondence between galaxies with old , red stellar populations , galaxies with little to no dust , and galaxies with little or no h@xmath52 emission .
that is , there is good consistency between the wholly independent photometric and spectroscopic classifications of blue / red , young / old , and active / passive galaxies .
( but this does not imply that all ` blue ' galaxies are star - forming , or vice versa . )
this should give some confidence in the sps - derived modelling used to derive all of @xmath50 , @xmath41 , @xmath43 , and @xmath44 .
[ fig : colours],title="fig:",width=328 ] galaxies in the colour stellar mass plane using restframe @xmath41 colour , and using the intrinsic ( _ i.e. _ , corrected for internal dust extinction ) stellar colour @xmath44 , respectively .
the middle panel shows the dust extinction , @xmath43 , as a function of mass , and can be thought of as linking the other two . within each panel ,
the blue and red points distinguish galaxies that do or do not show clear h@xmath52 line emission ; optically identified agn are highlighted yellow .
these spectral classifications are entirely independent of the sed - fit results .
note that the agn seem to be almost exclusively associated with the blue sequence in the @xmath44 cmd .
as expected , there is a strong correspondence between galaxies with old , red stellar populations , galaxies with little to no dust , and galaxies with little or no h@xmath52 emission .
that is , there is good consistency between the wholly independent photometric and spectroscopic classifications of blue / red , young / old , and active / passive galaxies .
( but this does not imply that all ` blue ' galaxies are star - forming , or vice versa . )
this should give some confidence in the sps - derived modelling used to derive all of @xmath50 , @xmath41 , @xmath43 , and @xmath44 .
[ fig : colours],title="fig:",width=328 ] galaxies in the colour stellar mass plane using restframe @xmath41 colour , and using the intrinsic ( _ i.e. _ , corrected for internal dust extinction ) stellar colour @xmath44 , respectively .
the middle panel shows the dust extinction , @xmath43 , as a function of mass , and can be thought of as linking the other two . within each panel ,
the blue and red points distinguish galaxies that do or do not show clear h@xmath52 line emission ; optically identified agn are highlighted yellow .
these spectral classifications are entirely independent of the sed - fit results .
note that the agn seem to be almost exclusively associated with the blue sequence in the @xmath44 cmd .
as expected , there is a strong correspondence between galaxies with old , red stellar populations , galaxies with little to no dust , and galaxies with little or no h@xmath52 emission .
that is , there is good consistency between the wholly independent photometric and spectroscopic classifications of blue / red , young / old , and active / passive galaxies .
( but this does not imply that all ` blue ' galaxies are star - forming , or vice versa . )
this should give some confidence in the sps - derived modelling used to derive all of @xmath50 , @xmath41 , @xmath43 , and @xmath44 .
[ fig : colours],title="fig:",width=328 ] restframe luminosities and colours are derived for each galaxy in the course of the sps fitting process , in the same way as for @xmath50 or @xmath45 .
naturally , these values reflect the galaxies constituent stellar population , modulated by interstellar dust within each galaxy . in order to more directly trace the stellar populations
, we will therefore also consider dust corrected or intrinsic stellar colours . to our knowledge , this approach was first pursued by @xcite .
it has also been described by , _
e.g. _ , @xcite and @xcite , on the basis of 32-band photometric redshifts from the musyc narrow - medium band survey ( nmbs ) . here
, we demonstrate the feasibility and applicability of this kind of analysis for broadband seds , given spectroscopic redshifts .
we will focus on the intrinsic stellar @xmath41 colour , which we will represent as @xmath44 .
this parameter is a very good proxy for luminosity weighted mean stellar age for these models .
however , the correlation between the sps - inferred and h@xmath52-derived ssfr for real galaxies is weak at best .
it seems to us that sps fits are good at constraining @xmath53 , which is very closely related to @xmath54 , but much less so at constraining the instantaneous ssfr .
thus , we consider @xmath44 to be a better proxy for @xmath53 than for ssfr . ] , @xmath53 .
quantitatively , at fixed @xmath44 , the range in @xmath53s is everywhere @xmath28 dex , and @xmath55 dex for @xmath56 .
the values of @xmath44 have also been derived in the course of the sps fits , but can be derived to within @xmath57 mag from the values of @xmath41 and @xmath43 directly : @xmath58 the coefficient of 6.0 in this relation reflects @xcite dust extinction applied to the average ( sed - fit ) galaxy spectrum within our sample .
the first order effect of adopting a different dust obscuration law would be a scaling of this coefficient .
the choice of dust obscuration law is thus a potential source of unaccounted - for random and systematic error in our characterisations of the @xmath44 cmrs ( but much less so for the mfs ) . the formal uncertainties in the derived values of @xmath43 ranges from @xmath59 mag for very blue galaxies to @xmath60 for very red ones .
again , formally , the dominant factor in these uncertainties is the imposed 0.05 mag error floor on each point in the sed ( rather than the catalogued photometric uncertainties ) , and thus they are not strongly magnitude dependent .
given that the parameter @xmath43 is bounded
the amount of dust can not be negative
these random uncertainties will lead to a systematic overestimate of @xmath43 in the case that the true value of @xmath43 is close to zero .
this may in turn induce a differential systematic bias in the inferred @xmath44 colours of galaxies with little to no dust ( compared to those with some dust ) , such that the value of @xmath44 may be too blue by @xmath61 mag , but not more than 0.2 mag .
this limits our ability to accurately determine the locus of the cmr for red sequence galaxies , at least where such galaxies have little to no dust . without diminishing this point
, what is more important for our analysis and especially when it comes to determining the mfs for the red and blue populations is that we are able to make the _ qualitative _ distinction between an old stellar population and a younger one with some dust . with this in mind ,
[ fig : colours ] * shows a simple sanity check on these values . in this figure
, we distinguish between galaxies with and without strong h@xmath52 line emission .
specifically , those galaxies with an equivalent width of 1 or greater are plotted as blue ; agn - host galaxies ( see ) are plotted as yellow crosses ; the remainder of the population with h@xmath52 seen in absorption are plotted as red .
the top panel of fig .
[ fig : colours ] shows the effective , restframe @xmath41 cmd for our @xmath0 sample .
as expected , galaxies with h@xmath52 seen in absorption can be seen to form a tight red sequence in @xmath41 colours . however , there are also many galaxies with strong h@xmath52 emission that lie embedded in or very near to this red sequence . in general terms , these are the dusty star - formers . the central panel of fig .
[ fig : colours ] shows the sed - fit values of @xmath43 as a function of @xmath42 , using the same plotting scheme to distinguish ` active star - formers ' from ` quiescent ' galaxies .
the emission- and absorption - line galaxies can be seen to follow different @xmath42@xmath43 relations : in general , the galaxies without h@xmath52 emission have low ( but non - zero ; @xmath620.35 ) dust extinctions .
while this is as expected , it is crucial to realise that the spectral classifications are independent of the sed - fit values for @xmath42 and @xmath43 .
these results thus demonstrate that our sps fits are indeed able to reliably distinguish between old sps with little or no dust , and dusty star - forming galaxies .
( see also the discussions based on optical
nir colours or stellar spectral diagnostics described in , below . ) for the galaxies with h@xmath52 emission , there is a trend towards higher values of @xmath43 with increasing @xmath50 .
the obvious implication is that the star - forming population will be observed to become redder in @xmath41 towards higher @xmath50 by virtue of their higher dust content , independently of any variation in their stellar populations .
this complicates any attempt to disentangle the young / star forming and old / passive populations based on @xmath41 alone . as can be seen in the lower panel of fig .
[ fig : colours ] , however , the active and quiescent populations are much better separated using the dust - corrected , intrinsic stellar colour , @xmath44 . again , we stress that the determination of @xmath44 is independent of the spectral classification the fact that galaxies that show h@xmath52 in absorption are almost all red in @xmath44 thus demonstrates that we are in fact able to distinguish between ` red and dead ' galaxies from dusty star - forming galaxies on the basis of their broad - band seds .
( see for further discussion of this point . )
note that our immediate goal in this paper is _ not _ to distinguish between galaxies based on their instantaneous sfrs .
( we will do this in another paper , using the h@xmath52 measurements . ) here , our goal is to characterise galaxies stellar populations , using the intrinsic stellar colour , @xmath44 , which is a close proxy for luminosity weighted mean stellar age , @xmath53 .
this relies on our ability to meaningfully constrain the dust obscuration , which is what fig .
[ fig : colours ] is intended to show .
taken together , the three panels of fig .
[ fig : colours ] should thus give some confidence in the reliability of our estimates of all of @xmath50 , @xmath41 , @xmath43 , and @xmath44 .
when we come to fitting the galaxy distributions in colour mass space in , we will want to account for the fact that the measurement errors / uncertainties in @xmath50 and @xmath41 are correlated .
the strength of this correlation is characterised by the ( pearson ) correlation coefficient , @xmath63 here , @xmath12 and @xmath64 can be taken to be @xmath42 and either @xmath41 or @xmath44 ; @xmath65 and @xmath66 are the uncertainties in these values ; and @xmath67 represents the expectation value for a generic quantity @xmath13 . in the parlance of @xcite , @xmath67 is the bayesian ` most likely ' value , which is computed as the probability - weighted integral over the posterior distribution function ( pdf ) for that quantity ( see eq . 5 of * ? ? ?
* ) . by definition
, the value of @xmath68 is constrained to be @xmath69 , with the cases @xmath70 corresponding to total anti - correlation , total independence , and total correlation , respectively .
the values of @xmath71 have been computed with the formal uncertainties @xmath65 and @xmath66 on a per galaxy basis in the course of the sps fitting process . for the galaxies in our sample ,
the covariance between @xmath42 and @xmath41 is typically @xmath72 ; the @xmath73@xmath44 covariance is typically in the range @xmath74 . with this definition , the error / uncertainty ellipse for any individual galaxy
can then be expressed in the usual way for a bivariate gaussian distribution : @xmath75 , \ ] ] where the vector @xmath76 represents the observed data point and the associated error / uncertainty matrix , @xmath77 , is : @xmath78 note that if @xmath79 , then the matrix @xmath80 is diagonal with entries @xmath81 and @xmath82 , and reduces to the familiar form for a 2d gaussian with @xmath83 $ ] .
as mentioned at the beginning of this section , the sps fits to the seds includes an error floor of 0.05 mag , and it is this decision that largely determines the formal uncertainties in @xmath44 .
the median formal uncertainty in @xmath44 within our sample is 0.18 mag ; 99% of our sample have uncertainties greater than 0.10 mag . by comparison ,
the _ observed _ width of the blue and red sequences in the @xmath44 cmd are on the order of 0.10 mag ( see fig .
[ fig : stardist ] , below ) ; _ i.e. _ , significantly smaller than the formal uncertainties .
this indicates that the formal ( random ) errors in @xmath44 are badly overestimated .
for this reason , when we model the cmd , we rescale the formal error estimates using a multiplicative factor @xmath84 .
the value of this scaling factor is fit for as a nuisance parameter along with the rest of the model . from our modelling of @xmath44 cmd
, we find @xmath85 ; in effect , we are ultimately using nearly uniform uncertainties in @xmath44 of @xmath86 0.05 mag .
note that we do not rescale the formal uncertainties for @xmath42 , nor do we adjust the correlation coefficients @xmath71 . for comparison , fitting to the @xmath41 cmd , the inferred value is @xmath87 ; _ i.e. _ , we see no signs that the formal uncertainties on @xmath41 ought to be rescaled . our analysis is based on a subset of the full gama database .
specifically , we limit our analysis to those gama galaxies with @xmath1 ( _ i.e. _ , @xmath88 m@xmath89 ) and @xmath0 .
these mass and redshift limits are motivated and justified in , below . to ensure the reliability and robustness of the spectroscopic redshift measurements
, we will only consider those galaxies with ` nq ` @xmath90 .
we only consider the @xmath19-band selected sample ; that is , we ignore 12 h - atlas selected galaxies , and 355 filler targets with @xmath91 . with these selections , we have a sample of 26368 galaxies .
98.5% of our sample has effective surface brightness @xmath92 .
based on the completeness curves shown in @xcite , we expect there to be no significant surface brightness selection effects inherited from the ( sdss ) photometric parent catalogues , at least for @xmath93 . we have explored the impact of surface brightness - dependent redshift failure rates , by applying completeness corrections as a function of the sdss @xmath94 magnitude . the effect on the mfs is negligible : only 1% for @xmath95 , and still just 3% for @xmath96 .
we do not explicitly exclude agn from our analysis . in fig .
[ fig : colours ] , we highlight the 1522 galaxies that are identified as agn hosts , based on their position in the bpt @xcite diagram , coupled with an h@xmath52 equivalent width @xmath97 6 selection . this is similar in spirit to the whan selection described by @xcite , and was chosen to approximately reproduce the by - eye spectral classifications by @xcite .
we note that the vast majority of these agn hosts are inferred to have ` normal ' b - type @xmath44 colours .
we have verified that none of our main results or conclusions ( including the shape of the b and r mfs ) change if we choose to exclude these galaxies .
the upper panels of fig . [ fig : colours ] show the basic data for our analysis ; namely the @xmath41 and @xmath44 cmds . in both cases , the relative number of red sequence galaxies in both the @xmath41 and the @xmath44@xmath50 diagrams peaks somewhere around @xmath98 .
there is a drop - off in the fraction of red galaxies below this mass , such that there is little to no clear evidence for a continuation of the @xmath44 red sequence below @xmath99 .
the principal difficulty in interpreting this result is the extent to which our @xmath0 sample is sensitive to truly ` red and dead ' galaxies at these relatively low masses of @xmath100 .
we explore this issue in two complementary ways in this section .
[ ch : vmax ] we use the standard @xmath102 technique @xcite to account and correct for incompleteness as a function of both stellar mass and stellar population .
the essential idea behind @xmath102 corrections is to estimate that maximal volume , @xmath103 , over which any given galaxy would satisfy our ( @xmath19-band ) selection criteria .
if we can estimate or predict the apparent @xmath19-band magnitude for a given galaxy if it were to be placed at some generic redshift @xmath104 as @xmath105 , then @xmath103 can be derived by integrating over the survey volume in which @xmath105 is brighter than our selection limits .
this has been done in the course of the sps sed fits , using the single best - fit spl template , as described in @xcite .
there are two things that need to be accounted for when estimating the values of @xmath103 for galaxies in our sample .
first , there is the fact that the gama target selection has been done on the basis of sdss petro magnitudes .
we can account for this by calculating @xmath105 as @xmath106 , where @xmath107 is the ( foreground extinction corrected ) sdss petro magnitude , and @xmath108 can be thought of as the @xmath104-dependent @xmath109-correction implied by the sps fit .
second , there is the difference between the cosmological redshift , which maps directly to comoving distance , @xmath110 , as @xmath111 , and the heliocentric redshift , which includes doppler shifting from peculiar motions due to local bulk flows .
this is done by recognising that @xmath112 , where @xmath113 is the peculiar velocity arising from local bulk flows ( see ) .
the value of @xmath103 is then defined via the maximum ( flow - corrected ) redshift , @xmath104 , for which both the @xmath114 and @xmath115 selection criteria are satisfied .
we have experimented with using a density corrected @xmath102 weighting to account for large scale structure at the lowest redshifts .
@xcite have shown first that large scale structure in the @xmath116 gama volume can have a significant impact on the recovered mass functions , and second that these effects can be largely mitigated by using a density defining population .
our case is rather different , however : even for @xmath117 , most of our galaxies lie at @xmath118 .
( plus , the gama - ii survey area is 25% larger , as well as 0.4 mag deeper in two of three fields . ) using the @xcite scheme , the corrections to the mfs are at the level of @xmath119% for @xmath120 .
the problem is that using different ddps yield different corrections .
the difference in the recovered @xmath121 mf when defining the ddp to be @xmath122 or @xmath123 galaxies is on the order of @xmath124% ; that is , comparable to the size of the corrections themselves . for this reason , we do not apply these negligible corrections .
note that to protect against catastrophic errors in the @xmath103 estimates , we limit the maximum relative weighting of any individual galaxy to be @xmath125 . in effect
this means that we will be under - correcting for any galaxies that have @xmath126 . as can be seen in fig .
[ fig : cmds ] , this decision affects only 16 galaxies in our @xmath1 sample .
limiting our sample to being @xmath127 excludes all of these ` problem ' objects . with this @xmath127 limit ,
the inferred mass functions are depressed by @xmath28 dex for @xmath128 ( due to incompleteness ) , but none of our qualitative results or conclusions change .
we have done the usual consistency tests @xcite to check the reasonableness of these incompleteness corrections .
we have verified that where the values of @xmath103 imply that we are properly volume limited ( _ i.e. _ , mass complete ) , the median @xmath129 is approximately equal to the volumetric centre of the @xmath0 survey window .
for the bluest galaxies ( @xmath130 ) , this is true for @xmath131 ; for the reddest galaxies , this is true for @xmath132 .
we have also verified that the median value of @xmath133 ; even after binning by colour , this is true for all masses @xmath134 .
the problem is that there are too few low mass red galaxies in our sample for us to look at our completeness for @xmath135 galaxies with @xmath136 in this way .
we have only 12 galaxies with @xmath137 and @xmath138 in our sample , all of which are at @xmath139 .
leaving aside the question of field - to - field variance , the concern is whether the apparent dearth of such red , low - mass galaxies in the gama catalogues is a fair characterisation of the gama @xmath139 survey volume , or if instead we have over - estimated our sensitivity to these very faint galaxies . and @xmath44 cmds , respectively , for all @xmath0 galaxies in the gama catalogue . in both panels ,
the points show individual galaxies , with the sizes of each point chosen to reflect the implied @xmath102 incompleteness corrections ( see ) .
the line contours show the observed data density without corrections for incompleteness ; the filled coloured contours show the @xmath102-weighted results .
the white lines show empirical estimates of how the @xmath20 selection limit maps onto the cmd at different redshifts ( see ) .
the @xmath140 curve shows that we are complete ( volume limited ) for all galaxy colours for @xmath141 .
our decision to limit our @xmath142 weightings to @xmath143 effectively means that we will be under - correcting for incompleteness for @xmath144 ; those galaxies with relative weightings @xmath145 , or @xmath146 , are plotted in black .
the extent to which the empirical completeness curve for @xmath147 approximately bounds the black @xmath148 points thus shows the consistency between these two complementary means of estimating our selection limits as a function of colour , mass , and redshift . our analysis is thus _ conservatively
_ limited to @xmath1 ( _ i.e. _ , @xmath149 m@xmath89 ) ; this limit is shown as the vertical dotted line .
[ fig : cmds ] , title="fig:",width=321 ] and @xmath44 cmds , respectively , for all @xmath0 galaxies in the gama catalogue . in both panels ,
the points show individual galaxies , with the sizes of each point chosen to reflect the implied @xmath102 incompleteness corrections ( see ) .
the line contours show the observed data density without corrections for incompleteness ; the filled coloured contours show the @xmath102-weighted results .
the white lines show empirical estimates of how the @xmath20 selection limit maps onto the cmd at different redshifts ( see ) .
the @xmath140 curve shows that we are complete ( volume limited ) for all galaxy colours for @xmath141 .
our decision to limit our @xmath142 weightings to @xmath143 effectively means that we will be under - correcting for incompleteness for @xmath144 ; those galaxies with relative weightings @xmath145 , or @xmath146 , are plotted in black .
the extent to which the empirical completeness curve for @xmath147 approximately bounds the black @xmath148 points thus shows the consistency between these two complementary means of estimating our selection limits as a function of colour , mass , and redshift .
our analysis is thus _ conservatively _ limited to @xmath1 ( _ i.e. _ , @xmath149 m@xmath89 ) ; this limit is shown as the vertical dotted line .
[ fig : cmds ] , title="fig:",width=321 ] in * fig .
[ fig : cmds ] * , we again show the distribution of @xmath0 galaxies in the @xmath41 and @xmath44 colour mass diagrams . in both panels ,
the filled , coloured contours show the inferred bivariate colour mass distribution function after applying our @xmath102 weightings to account for incompleteness . these contours should be compared to the black and white line - contours , which show the raw , observed data - density in the colour mass diagrams ; _ i.e. _ , without incompleteness corrections .
we have plotted the individual galaxies in our sample as the gray points ; the size of each point directly reflects the magnitude of the @xmath102 factor used to account for incompleteness .
the black points in these panels highlight those few galaxies with relative weightings @xmath150 .
since we have chosen to limit our weightings to be @xmath151 , these are the galaxies for which ( formally ) we would be under - correcting for incompleteness .
it is entirely possible , however , that these points reflect somehow catastrophic errors in our @xmath103 and/or @xmath50 estimations : there are many more galaxies with similar masses and colours for which the implied values of @xmath152 are considerably smaller . in order to investigate our sensitivity to low mass , red sequence galaxies further ,
we have therefore sought to quantify our sample completeness limits in a way that is independent of our @xmath103 calculations .
we have done so by taking all observed galaxies and simply scaling their total luminosities / masses down to match the @xmath153 selection limit .
then , by dividing our sample in narrow redshift intervals of width @xmath154 and centred on @xmath155 , 0.02 , ... , 0.12 , we take the median value of this limiting mass in narrow bins of restframe or intrinsic colour .
this analysis thus provides an empirical description of our 50% mass completeness limits , as a function of redshift and colour , but in a way that is independent of the sps fits that have been used to derive @xmath50 , @xmath41 , and @xmath44 ; the results are shown as the heavy white - and - black lines in fig .
[ fig : cmds ] . the @xmath156 , 0.04 , 0.06 , 0.08 , and 0.10 curves can be taken as corresponding to relative volume completenesses of @xmath157 0.02 , 0.04 , 0.13 , 0.30 , and 0.59 , respectively .
the first point to make is that this independent , empirical characterisation of our mass - completeness limits agrees very well with the results of our @xmath102 calculations .
the fact that the distribution of the black points in each panel of fig .
[ fig : cmds ] is approximately bounded by the empirical completeness limit for @xmath158 should thus give some confidence in our @xmath103 estimates .
in the same way , the @xmath159 curve can be taken as indicative of where we are truly volume limited .
taken together , these two curves thus bound the region of the colour - mass diagrams in which our incompleteness corrections are important and reasonable .
since the black and colour - filled contours in these panels show the data density without and with @xmath102 corrections , where these contours coincide shows where incompleteness corrections are unnecessary .
again , the fact that the @xmath160 curve very accurately bounds the regions in both the @xmath41@xmath50 and @xmath44@xmath50 diagrams over which this is true should give confidence in our @xmath103 estimates .
given all of the above , are we ( or are we even capable of ) seeing the low - mass end of the red sequence
? we can be all but certain that there are very few red galaxies with @xmath161 galaxies in the @xmath116 gama survey volume ( @xmath162 gpc@xmath163 ) . addressing this question any further
is made problematic , however , by the effects of large scale structure , and particularly by the degree to which low - mass and red galaxies are biased towards rich - group and cluster environments .
@xcite have looked at the fraction of low mass galaxies in sdss that are ( spectroscopically ) identified as having _ both _ old stellar populations _ and _ no ongoing star formation , and found that all such galaxies with @xmath1649 are satellites within 1.5 mpc ( comoving , projected ) of a @xmath165 ` host ' galaxy . for our sample ,
the number density of @xmath166 galaxies at @xmath116 is actually 10% higher than for @xmath167 : that is , we may in fact be biased _ towards _
old , low mass galaxies .
based on the group catalogue described by @xcite , our sample contains 580 @xmath166 and @xmath168 galaxies , in 232 separate groups , 22 of which have multiplicities of 10 or more . included in these groups
are 194/2895 ( 6.7% ) of the @xmath169 and @xmath116 galaxies in our sample , all of which are within 0.5 mpc ( _ cf._the @xcite limit of 1.5 gpc ) of their hosts .
these numbers give some sense of the environments we are probing_i.e .
_ , from isolation up to low- and - moderate sized groups . at the same time , we point out that based on the derived values of @xmath103 , the results shown in fig . [
fig : cmds ] suggest that our completeness may still be @xmath170% even for @xmath171 .
further , we stress that our relative volume completeness is greater than 50% ( @xmath172 ) for even the reddest galaxies with @xmath173 ; we consider it highly unlikely that our results above this mass scale are strongly affected by incompleteness .
we have also verified that none of our results change significantly if we limit our analysis to @xmath116 , or to @xmath173 . in light of all this , and with the above caveats , we continue our analysis with a nominal mass limit of @xmath1 .
there are a number of ways of discriminating between ` developed ' and ` developing ' galaxies , based on , _
e.g. _ , restframe colour , spectral classification , hubble type ( _ i.e. _ , morphology ) , or srsic index ( _ i.e. _ , structure ) .
there is considerable , but by no means total , overlap between these different kinds of selections ( see , _ e.g. _ , * ? ? ?
. however , as we will show in future papers in this series , inappropriate conflation of the terms red / blue , early-/late - type , and quiescent / active has the potential to be dangerously misleading .
our overarching goal in this work is to look at the bimodality as seen in the optical cmd in other words , we are specifically interested in the bimodality that exists in galaxies _ stellar populations_. with this in mind , our specific goal is to derive a quantitative description of phenomenology of the joint colour mass distribution of galaxies , in terms of both the cmrs and the mfs for the apparently distinct ` red ' and ` blue ' populations . as a motivating introduction to our method for attacking this problem ,
consider * fig .
[ fig : others]*. in this figure , we show our best attempts at reproducing the sdss - based analyses of @xcite , @xcite , and @xcite using our @xmath1 and @xmath174 gama sample . in the upper panels of this figure
, we show the different ways that each set of authors have separated the red and blue galaxy populations , based on either a colour - magnitude , or a colour mass diagram . in rough terms , the @xcite cut can be seen as a relatively conservative means of selecting ` red ' galaxies : the selection line appears to hug the lower limits of the red sequence .
by contrast , the @xcite cut is rather aggressive : it falls closer to the upper edge of the blue cloud .
the cut advocated by @xcite is in a sense intermediate : it can be seen to be aggressive at lower luminosities , and conservative at higher luminosities . in the lower panels ,
the filled squares show the inferred red / blue galaxy mass functions , when applying each of the different selections to our gama dataset . in general
, the agreement between each set of sdss- and gama - derived results is very good .
further to our discussion of incompleteness in the previous section , we also highlight the fact that the gama mfs including the red mfs are continuous for @xmath175 .
this is despite the distracting and unfortunate downtick in the number of galaxies with @xmath176 ( our mass selection limit ) .
we are not obviously incomplete for @xmath177 .
there are some obvious systematic differences in the inferred number densities for @xmath100 . as a result ,
the integrated number density of galaxies with @xmath1 from gama is 7% , 12% , and 13% lower than that from @xcite , @xcite , and @xcite , respectively .
( not surprisingly , however , we agree almost exactly with * ? ? ?
* not shown ) .
these differences come down to the different means of estimating stellar masses . the role of various kinds of systematic errors / uncertainties in determining the net mf ( the black curves and points in fig .
[ fig : others ] ) has been explored by @xcite ; that is not our main purpose here . for our purposes , it is sufficient to note that having controlled for everything we can ( _ e.g. _ , taking sdss model fluxes as total ; matching imfs and cosmologies ) we match the inferred integrated stellar mass density for @xmath173 to within 34% in each case . instead , we are specifically concerned with sources of systematic error or uncertainty on the mfs for the ` red ' or for the ` blue ' galaxy populations .
that is , we are particularly interested in the red and blue lines / points shown in fig .
[ fig : others ] . in this regard , the most noticeable discrepancy is our failure to reproduce the apparent upturn in the red mass function seen by @xcite for @xmath121 .
we suggest that at least part of this discrepancy is due to differences in how we have derived our @xmath178 colours . comparing our @xmath178 colours , derived in the course of the sps fits , to those from @xmath179 @xcite
, we find that there is considerable scatter ( at the level of 0.15 mag ) , even when analysing the same sdss petro photometry .
we have tried simply perturbing our @xmath178 colours by 0.15 mag .
this has the net effect of scattering a small fraction of ` blue ' galaxies into the ` red ' sample , which leads to a significant increase in the inferred numbers of red galaxies , as illustrated by the thin vertical lines in the lower - right panel of fig .
[ fig : others ] .
note that the other mfs are more robust to photometric scatter at this level , as is shown .
we also note that , in the middle panel , the agreement between our gama - derived results and the @xcite fits is imperfect , particularly for the red mf
. we will defer detailed discussion of this discrepancy to .
for now , we simply note that the @xcite cut is based on their fits to the mf , rather than the other way around ; we therefore expect some small quantitative differences between the @xcite fits and the mfs derived using the @xcite cut . at this stage ,
the important point is that we see the same qualitative results . in order to facilitate easy comparison between the results of these different analyses , in each of the lower panels of fig .
[ fig : others ] , there are two sets of small open squares , which re - plot the results of the other two analyses .
the range spanned by these points thus reflects the systematic uncertainty on the blue and red mass functions , arising from the different ways that the ` blue ' and ` red ' galaxy samples have been selected / defined .
note in particular the size of these uncertainties at and around the knee of the mass function , as well as at low masses , for both the red and blue mass functions .
based on fig .
[ fig : others ] , we make the following four observations : * comparing the different authors fit mfs _ to one another _ , there are important discrepancies in the shapes of both the red and the blue mass functions for @xmath180 .
this is unfortunate , because this leads to large uncertainties ( @xmath181 dex ) in the mass - scale at which galaxies transition from one population to the other .
* there are also large discrepancies in the values of the low - mass slope of the red mass function : it might be slowly declining ( @xmath182 ) , or nearly constant ( @xmath183 ) , or has a sharp upturn ( @xmath184 ) .
this is unfortunate , as it leaves the behaviour of the low mass red population largely unconstrained .
* further , there is not even consensus as to how the two mfs ought to be described and understood _
qualitatively_. the @xcite mfs are each well described by a single schechter function ; @xcite finds a need for a second schechter component to describe the blue mf ; @xcite find instead that it is the red mf that needs a second schechter component .
* that we can reproduce each set of results using our dataset shows that these discrepancies come from differences in how the data are analysed , rather than differences in the data themselves .
in other words , _ current understanding of the mfs for the red and blue galaxy populations is limited by systematic errors_. as mentioned in the introduction , measurements of the mfs for the red and blue galaxy populations have played a pivotal role in informing our understanding of galaxy formation and evolution .
it is therefore critically important to understand how and why there can be such large discrepancies between the results of these different analyses .
only then will we be able to formulate an analytical approach that will allow us to robustly measure these quantities .
given that there appears to be a better separation between the ` red ' and ` blue ' populations in the @xmath44 cmd shown in fig.s [ fig : colours ] and [ fig : cmds ] , the natural question is whether the discrepancies described in the previous section can be alleviated or removed by focusing on intrinsic stellar colours / magnitudes .
what happens if we try modifying these analyses to account for dust obscuration / extinction ?
we address this question in * fig .
[ fig : otherdust]*. the main difference between this figure and fig .
[ fig : others ] is that we have now shifted to intrinsic ( _ i.e. _ , dust corrected ) stellar colours and luminosities , so as to more directly probe galaxies stellar populations . as discussed in and shown in fig .
[ fig : colours ] , the lowest values for the sed - fit dust extinctions is @xmath185 , even for galaxies with no h@xmath52 emission . in light of this fact
, we need to also rescale each of the selection lines shown in the upper panels of fig .
[ fig : others ] .
what we have done is to shift each selection line by the equivalent of @xmath186 mag .
this ` correction ' is much larger than the expected dust obscuration for a canonically ` red and dead ' galaxy .
it should be thought of as a conservative way to exclude the dustiest galaxies , while retaining those galaxies with genuinely ` red ' stellar populations .
the bottom panels of fig .
[ fig : otherdust ] shows the mfs for ` red ' and ` blue ' galaxies , selected in this way ; the selections themselves are shown in the upper panels .
it is clear that adopting these selections would lead to a rather different picture of the makeup of the field galaxy population .
comparing the range of values for the red mf that come from these different selections , the differences are at the level of a factor of @xmath187 .
this is considerably larger than the differences seen in fig .
[ fig : others ] .
that is , modifying these selections to account for dust _ exacerbates _ the tension between these different authors results , rather than alleviating it .
( this implies that , even when looking at dust - corrected , intrinsic stellar colours , there is still substantial overlap between the ` blue ' and ` red ' populations .
we will show in fig .
[ fig : stardist ] , below , that this is indeed the case .
) no less worrying is the size of discrepancy between the @xmath100 mfs for the @xmath44-selected ` red ' galaxies ( fig .
[ fig : otherdust ] ) and the @xmath41selected ones ( fig .
[ fig : others ] ) .
perhaps not surprisingly , the size of this discrepancy is very sensitive to how much one chooses to shift each selection line that is , how much dust to allow for in otherwise ` red sequence ' galaxies .
so which of the six analyses we have now trialled is right ? are the results shown in fig .
[ fig : otherdust ] any more or less reliable or meaningful than those shown in fig .
[ fig : others ] ?
the crux of the problem is that there are no clear theoretical grounds for preferring any one of these ` red'/`blue ' cuts over any other . in the absence of a solid , astrophysically meaningful argument for such a cut
, this is necessarily true without further information , we have no compelling way to answer this question .
while we might offer some empirical or phenomenological argument in support of our specific cut , the point is that this decision will always be arguable ; that is , _ arbitrary_. this is a point that we will return to in and .
for now , we can say with some confidence that these different results can be taken to approximately bracket the range of allowed values for the red / blue mfs that come from reasonable choices for hard - cut red / blue selections .
the truth probably lies somewhere between the different results shown in fig .
[ fig : others ] and fig . [ fig : otherdust ] .
this is not very satisfactory , however , as it means that the shapes of the red and blue mfs are not even well constrained _
qualitatively _ , much less quantitatively . in short :
the quantitative _ and qualitative _ discrepancies between the results shown in the lower panels of fig .
[ fig : others ] and fig .
[ fig : otherdust ] are entirely due to the different ways that each set of ` red ' and ` blue ' samples has been selected or , said another way , to the different operational definitions of the terms ` red ' and ` blue ' .
for example , the fact that @xcite see an upturn to the red mass function at low masses where @xcite and @xcite do not , despite their using essentially similar datasets , and even pushing to lower masses is a direct consequence of the fact that the @xcite selection line is relatively bluer than earlier authors .
it is therefore worth reflecting on the two implicit assumptions that underpin the use of a hard cut to separate ` red ' from ` blue ' galaxies , and thus the results shown in fig .
[ fig : others ] and in fig .
[ fig : otherdust ] .
first , it is presupposed that ` blueness ' and ` redness ' are physically meaningful designations , inasmuch as they encapsulate some fundamental distinction between the origins or natures of two distinct kinds of galaxies .
the second , and more problematic , assumption is that there is something special something astrophysically meaningful about the particular boundary used to separate the two galaxy classes .
the methodological appeal of such an approach is that it is well - defined , inasmuch as the hard cut can be written _
explicitly _ and _ exactly _ , which makes such analyses easily reproducible . in the early days of the sdss , the astronomical motivation was also clear .
@xcite , @xcite , and many others had shown that there is substantial ( but not total ) overlap between a ` red sequence ' sample and an ` early type ' sample selected on the basis of srsic index . in this way ,
` redness ' and ` blueness ' were thought of as indirect proxies for structure , and thus for morphology . as was common at the time , @xcite , @xcite and others presented their mass function determinations for ` red ' and ` blue ' galaxies in terms of the hubble early- and late - type classifications .
@xcite , on the other hand , have phrased their results in terms of ` star - forming ' and ` quiescent ' galaxies .
( this is also the explicit goal of , for example , some colour colour selections , which are discussed further in . ) here again , we caution against this conflation of terminology when interpreting these results . while ` early type ' samples selected on the basis of colour , spectral type , morphology , and structure are often treated as if they are interchangeable , it is now becoming clear that _ they are not_. this point , and its importance , have most recently been forcefully made by @xcite , who consider the cmds for morphologically classified ` early- ' and ` late - types ' .
further , the use of a hard cut overlooks the empirical fact of scatter around each of the distinct cmrs for the ` red ' and ` blue ' populations , however they are defined .
any number of authors have shown that , at fixed magnitude or mass , the distribution of galaxies ( optical ) colours can be well described as the sum of two gaussians , and that the separation and widths of these two gaussian distributions are such that there is considerable overlap between the two ( see , _ e.g. _ , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
considering these two distributions as arising from two distinct populations , the implication is that the use of a hard red / blue cut will yield samples that are both incomplete , and contaminated ( see also , _ e.g. _ , * ? ? ?
we will return to this issue towards the end of this paper , in . in light of the above
, we will not take quite so simple a view .
we will assume that there _ is _ some meaningful astrophysical distinction to be made between the two populations : that there is some unknown astrophysical process that acts to determine whether any given galaxy is a member of either the ` blue ' or the ` red ' population .
that is , _ we will assume that there are two distinct cmrs . _ but we will also allow that some ` hidden ' parameter ( or parameters ) mean that , at fixed mass , there are a range of colours among the members of each of the two populations , to the extent that these two distinct populations are observed to overlap in the cmd . that is , _ we will assume that there is some intrinsic scatter around each of the two cmrs .
_ adopting this ( non - controversial ) view of two overlapping populations , the conceptual difficulty that arises is that some members of the ` blue ' population will have quantitatively redder @xmath41 or @xmath44 colours than some members of the ` red ' population .
further , two galaxies might have identical values of @xmath50 and @xmath41 , but one might ` really ' belong to the ` blue ' population , and the other to the ` red ' one . without further information , it would be impossible to unambiguously determine which is which .
this means that any ` red'/`blue ' classification of individual galaxies can only be done probabilistically , in terms of the odds that that galaxy has been drawn from either the ` red ' or the ` blue ' population .
while our approach brings these conceptual quandaries into sharp focus , we stress that similar criticisms can be levelled at the simple , binary ` blue'/`red ' distinction used above : the inferred scatter around the cmrs derived for the hard - cut ` blue ' and ` red ' populations leads to precisely the same conundrum . without solid astrophysical justification ,
the terms ` blue ' and ` red ' must be understood to be defined _ operationally _ , and as such are useful only as _ qualitative _ descriptors . in acknowledgement of this point
, we will from now on abandon the terms ` blue ' and ` red ' as classifiers , and instead use the more generic idea of a b- and an r - population .
note that these descriptors do not properly apply to individual _ galaxies _ , but instead to distinct _
populations of galaxies_. obviously , the designations b and r have been chosen with a nod towards one being for the bluer population , and the other for the redder one .
but we want to be absolutely clear that these designations are based on _ phenomenological _ descriptions of the joint colour - magnitude distributions and should not be taken to be rigorously grounded in astrophysical theory .
astrophysical _ interpretation of our descriptive b- and r- population modelling , including those offered in , must be done with care .
let us stress in particular that we are _ not explicitly _ trying to select galaxies that are quiescent , quenched , early - type , etc .
our explicit goal here is _ only _ to distinguish between the generic ` developed ' and the ` developing ' galaxy populations on the basis of their stellar populations . in this sense , all that the b and r designations are intended to encapsulate are the distributions of luminosity weighted mean stellar ages , as probed by either @xmath41 or @xmath44 .
of course , at least for moderate- to high - masses ( @xmath188 ) , it turns out that the galaxies that comprise the r population largely conform to the prevalent notion of ` red and dead ' or ` quenched ' ( see , as well as fig .
[ fig : colours ] ) .
this being the case , our results _ can _ be used to gain insight on the process of quenching , but _ only insofar as our operational definition of ` red'- or ` r'-ness can be taken to mean ` quenched'_. while our approach brings this issue into sharp focus , the same degree of caution is merited when interpreting the results of past studies of ` blue ' and ` red ' galaxies : bearing in mind the qualitative and quantitative discrepancies between the results shown in fig . [
fig : others ] or fig .
[ fig : otherdust ] , which of the selections shown in these figures can be said to best represent the idea of ` quenched ' ?
adopting the working hypothesis of two distinct but overlapping b- and r - populations in the cmd , the question becomes technical : how best to distinguish and characterise the two populations on the basis of the observed cmd .
while the designations ` b ' and ` r ' must be understood to be qualitative , inasmuch as they are phenomenological , we want to be able to classify galaxies _ quantitatively_. this can be done probabilistically , according to the chances that they are members of either the b or r population .
further , we want these classifications to be _
objective_. in order to achieve these goals , it is necessary to describe or otherwise account for the actual underlying colour distributions , including the degree of overlap , and as a function of mass .
this is therefore the task that we have undertaken .
this section is devoted to describing and validating our descriptive modelling of the bivariate or joint @xmath41 and @xmath44 colour mass distributions for field galaxies at @xmath0 .
the most general form of our model is laid out in , including definitions and descriptions of the 40 parameters that define the model in its most general form . in
, we describe the numerical methods that have been used to fit for the free parameters . in
, we describe the process by which we have selected the best and simplest description of the data from within the more general family of models that we have considered .
( for the more motivated reader , we present a pedagogical development of the model in appendix [ app : themodel ] , in which we build up our formalism as successive generalisations of the conventional weighted-@xmath6 approach to fitting a single line . ) the conceptual basis of our descriptive modelling is this : that the observed data are a sampling of some ` true ' , astrophysical , bivariate colour mass distribution .
this being the case , our data can be seen to have been drawn from generated by some 2d probability distribution function , @xmath189 , where @xmath190 denotes some generic location in our 2d data plane .
( in this section , we will thus use @xmath12 and @xmath64 to further abbreviate the quantities @xmath42 and either @xmath41 or @xmath44 . )
we can not absolutely know the `
true ' form of the distribution function @xmath189 .
so instead , we aim to construct a parametric description for what @xmath189 might be , and use the data to constrain the possible and even likely form of @xmath189 .
let us denote the full set of parameters used or required to describe @xmath189 as @xmath191 . assume for a moment that we know or can guess the correct form of @xmath192 .
naturally , one does not observe this distribution directly . instead ,
observational errors mean that the observed distribution in the @xmath193 plane will be a smeared out version of the true distribution .
let us also assume that that the observational errors / uncertainties for a given data point , @xmath194 , are gaussian , and so can be described by the covariance matrix @xmath77 ( see equations [ eq : rho][eq : covar ] ) .
using @xmath195 as shorthand for a bivariate gaussian , the likelihood of observing a particular datum @xmath196 is then given by the convolution of the ` true ' , underlying distribution and the bivariate gaussian that describes that measurement , and its associated uncertainty ; _ i.e. _ , @xmath197 note that to satisfy the requirement that a point actually be observed , we impose to the normalisation conditions that the integral over ( @xmath198 ) space for @xmath199 and @xmath200 , and hence @xmath201 , be equal to one .
it is crucial to recognise that the value of @xmath199 , and thus the value of @xmath201 , can only be computed indeed , are only defined given an assumed or trial set of values for each and every of the parameters in @xmath191 . in recognition of this fact
, these quantities have been written in and all that follows as @xmath202 and @xmath203 .
the crux of the problem is then to construct an appropriate parametric description of @xmath192 .
that is our task in this section . at this stage ,
the casual or credulous reader whose interest lies only in our results may wish to skip these technical sections , and move directly to , in which we demonstrate the quality of our fits to the observed bivariate @xmath41 and @xmath44@xmath50 distributions . in order to accommodate the apparent bimodality in the @xmath41 and @xmath44 cmds , we split the model for the ` true ' , astrophysical bivariate colour
mass distribution_i.e .
_ , the scalar function @xmath204into two distinct b and r components , which are denoted as @xmath205 and @xmath206 .
each component has its unique parameter set , denoted as @xmath207 and @xmath208 .
because , in general , these two populations will be observed to overlap , the probability density at any point @xmath209 is given by the sum of these two distributions ; _ i.e. _ , @xmath210 for now , the ` good ' subscript can be ignored ; its significance will become clear in a moment .
note that , in line with the probabilistic nature of this generative model , all of @xmath205 , @xmath206 , and @xmath211 should be understood to be integral normalised to one .
the parameter @xmath212 thus sets the relative normalisation of the b and r components , in terms of the relative number of r - population galaxies among the global population ( given our sample selection limits ) . at fixed mass
, we treat the colour distributions of each of the r and b populations as being gaussian , and so characterised by three numbers : 1 . ) a centre , 2 . ) a width , and 3 . ) a normalisation .
each of these three quantities is allowed to vary parametrically , and independently , as a function of mass , so that we can constrain : 1 . )
the locii of the b and r cmrs , @xmath213 and @xmath214 , 2 . ) the intrinsic scatters around these cmrs , @xmath215 and @xmath216 , and 3 . ) the mass functions for each population , @xmath217 and @xmath218 . using @xmath219 as short hand for a (
properly normalised ) 1d gaussian with center @xmath220 and width @xmath66 , our model for the bivariate colour mass distribution for the r - population can be written as : @xmath221 \bigg ) ~ , \nonumber\end{aligned}\ ] ] with an analogous expression for @xmath222 .
each of these aspects of the model are described in turn below .
as discussed in detail in , we do not actually model the mass distributions of red and blue galaxies as being continuous . instead
, we model the mass function using the sum of many kronecker delta functions whose
amplitudes are modulated by the continuous dual - schechter mass function , @xmath15 , defined in . in ,
@xmath15 is thus accompanied by the kronecker delta function , @xmath223 , and the underlying models @xmath189 can be seen to be constructed as the sum of many discrete components evaluated at @xmath224 .
the reason for this decision is to allow the convolutions in eq.s ( [ eq : elli ] ) , ( [ eq : model ] ) , and ( [ eq : bad ] ) to be done analytically .
our method can be seen as evaluating an approximate model , which has a discretised mass function , in an exact , analytical way .
we define the @xmath225s as @xmath226 , 1 , ... , 65 ; that is , as a uniformly is ( very slightly ) approximate . ]
spaced grid in @xmath12 with a spacing of 0.05 dex , with grid edges running from our nominal mass limit of 8.7 up to 12 . with this grid spacing ,
the typical galaxy with a mass uncertainty @xmath227 dex has 5 or 6 @xmath225s within its fwhm .
the normalised mass functions for the b and r components , @xmath228 and @xmath229 , respectively , are described using the sum of two @xcite functions : @xmath230 here , the parameter @xmath231 can be understood to govern the relative normalisations of the @xmath232s by describing the relative number of galaxies the make up the second of the two schechter functions , and the shapes of the two schechter functions , @xmath233 , are described by a low - mass power law with logarithmic slope , @xmath52 , and a characteristic mass , @xmath234 , which describes the ` knee ' of the mass function .
thus we have up to 5 parameters for each of the b and r populations , plus the dimensionless parameter @xmath212 defined above , for a total of 11 parameters to describe the full mass distribution of galaxies , down to our selection limit .
again , each of @xmath15 , @xmath235 , and @xmath236 should be understood to be integral normalised to unity ( given our @xmath237 and @xmath0 selections ) .
we must therefore estimate the global normalisation of the mass function independently of the modelling described in this section .
this has been done after the modelling on the basis of the integrated mass density among galaxies in our sample ; _ i.e. _ , @xmath238 mpc@xmath239 .
this step introduces a @xmath240% systematic uncertainty into all of our fit mass functions ( but not the observed ones ) . with
this value fixed , we can compute the values of the usual characteristic densities , @xmath241 , in units of mpc@xmath239 dex@xmath242 , based on the values of all 11 of the mf - defining parameters . when we give the values of the fit parameters in fig .
[ fig : mcmc ] , we quote the @xmath241 values in place of the @xmath243s . next , the cmrs , @xmath244 and @xmath245 .
we allow the slope of the b and r cmrs to vary as a function of mass by describing them in the following way : @xmath246 recalling that @xmath247 , @xmath248 , and @xmath249 , this definition can be transparently viewed as the combination of two linear relations .
there is a smooth transition from a low - mass regime , in which the cmr goes like @xmath250 , to a high - mass regime where the cmr goes like @xmath251 .
we highlight two special cases : first , if @xmath252 , then this parameterisation is equivalent to the line - plus - tanh parameterisation used by @xcite ; second , if @xmath253 , then we have a smooth transition around the point of intersection between two lines .
the parameter @xmath254 defines precisely where the transition takes place , and the parameter @xmath255 governs how sharp / smooth this transition is .
thus we have six parameters to describe each of the red and blue cmrs , bringing our running total of fit parameters to 23 .
finally , there is the scatter around the cmrs , @xmath216 and @xmath215 . in the most general form of the model
, we adopt the same parametric form for the @xmath14s as for the @xmath16s ; _ viz . _ : @xmath256 this adds another six parameters to describe the scatters around each of the r and b cmrs , which brings the running total number of parameters in @xmath191 to 35 . in order to protect against biasing of our results from outliers , catastrophic errors or otherwise un- or under - modelled aspects of the observed distribution in the @xmath257 plane ,
our generative model includes a parametric description for ` bad ' data . to this end , we split the model into two components ; one for each of the ` good ' and ` bad ' data distributions : @xmath258 here , the parameter @xmath259 describes the fraction of datapoints encompassed within the ` bad ' distribution .
this is wholly analogous to the use of @xmath212 to parameterise the relative normalisations of the b and r components of the model . in the model ,
these ` bad ' data are described by an additional ( large ) error in the measured values of @xmath12 and @xmath64 .
in other words , the ` bad ' component of the model is simply generated by convolving the ` good ' component with an additional 2d gaussian : @xmath260 the significance of the ` good ' subscript in should thus now be clear . the defining covariance matrix for this gaussian , @xmath261 has diagonal entries @xmath262 and @xmath263 ; the off - diagonal entries are zero .
( in fact , as we describe below , the fit values of @xmath262 are @xmath264 , and we are able to exclude this parameter without compromising the quality of the fits . ) the ` bad ' parameters @xmath259 and @xmath265 deserve some further comment .
first , what exactly is meant by ` bad ' ? before , we have distinguished the b and r components as having different cmrs as well as different mass function .
by contrast , the ` bad ' distribution can thus be seen to be just a ` poor copy ' of the ` good ' , r plus b distribution , having been ` smeared ' with a large gaussian , and with a much lower relative normalisation .
essentially , we are using these ` bad ' quantities to parameterise our ignorance of any and all features in the observed cmds that are not easily explained by the ` good ' model .
this includes catastrophic errors in the measurements of either @xmath12 or @xmath64 , but also includes at least in principle any additional components in the true , astrophysical , joint colour mass distribution .
given this , what justification is there for treating the distribution of ` bad ' data as gaussian ? in short , there is none .
that said , we stress that our characterisation of the ` bad ' data is simply in terms of the rms of ` bad ' data points around the ` true ' cmrs .
it is true that the link between the value of @xmath265 and the true shape of the distribution of ` bad ' data in @xmath257 space does implicitly assume gaussianity .
but we have no interest in accurately modelling the shape of this distribution ; for the purposes of objectively identifying and censoring such ` bad ' data , simply knowing ( or , better , modelling ) the rms scatter is sufficient .
further , we are not even really that interested in the precise values of the parameters @xmath259 and @xmath265 : what we are interested in using these parameters to limit the influence of outliers on the fit values of all the other , more astrophysically meaningful parameters . that being the case , when we come to reporting our results we will marginalise over the values of both of these nuisance parameters ( see below ) , leaving us only with the parameters of genuine interest and importance .
readers that remain concerned about the role of these parameters in our calculation are referred to our more detailed discussion in , and also to the excellent primer on data fitting by @xcite . in summary ,
eq.s ( [ eq : elli][eq : zeta ] ) define a model for the distribution of galaxies in the cmd , which is fully described by up to 40 parameters .
( again , in we will describe the model selection process by which we have ensured that we reduced this parameter set to ensure that we have the best and simplest description of the data possible .
) there are five parameters describing each of the red and blue mass functions , so that @xmath266 , with an analogous five parameters for @xmath267 . added to these , there is the parameter @xmath268 , which describes the relative number of red galaxies in our @xmath1 and @xmath0 sample .
there are also six parameters to describe each of the red and blue cmrs , so that @xmath269 , and similarly for @xmath270 . and
finally there are six parameters to describe the scatter around each of the red and blue cmrs , so that @xmath271 , and similarly for @xmath272 . then , we have three parameters to describe outliers or otherwise ` bad ' data , @xmath273 .
to these should be added the two parameters , @xmath84 and @xmath274 , which are used to rescale the formal uncertainties in @xmath44 , as discussed in . note that each of these different subsets should be understood to be formally independent . the mass scale and softening describing the transition between the high- and low - mass regimes for @xmath16 and @xmath14
are not assumed to be related , nor are these transitions in any way formally connected to the shapes of the mass functions , nor do we place any restrictions on the relations between parameters for the b and r populations . with all of the above definitions , and
given a set of trial values for the parameters in @xmath191 , we now have the means to compute the value of @xmath275 , as defined in .
armed with this information , it is then straightforward to compute the global likelihood , @xmath17 of observing the full dataset @xmath276 , given the associated uncertainties @xmath277 , as the product of all the individual @xmath201s .
in practice , it is more convenient to work in terms of @xmath278 , so that : @xmath279 here , @xmath280 is the set of 1/@xmath103 weighting factors that we use to account for incompleteness due to the gama apparent magnitude selections , as defined and discussed in . given the particular parametric form of our model , and given that we have observed our specific dataset , what we want to do is to use the data to constrain the possible values of the parameters in @xmath191 . in other words ,
we want to construct the posterior probability density function ( pdf ) for each of the individual parameters , @xmath281 , so that we can evaluate the probability that the parameter @xmath282 has the value @xmath283 ; that is , @xmath284 this is done using bayes theorem , which can be written as : @xmath285 here , @xmath286 is just the scalar likelihood function as defined in , which can only be computed given a full set of values for the model parameters .
in contrast , @xmath287 is the full , high - dimensional pdf for the values of the parameters @xmath191 , which is what we are interested in deriving .
bayes theorem links these two quantities via the _ prior _ distribution function , @xmath288 , which is an assumed , _ a priori _ statement of our expectations for the probability of different parameter combinations with respect to one another . by invoking bayes theorem ,
we are therefore required to explicitly state our priors on the relative probabilities , @xmath289 , of different values for each of the individual parameters @xmath281 .
the decision of what priors to adopt is by no means trivial , but it is also inescapable . _
all fitting algorithms include priors_. one can not compare the relative likelihoods of two different trial parameter values without an implicit or explicit prior , even if that prior is that the two values are , _ a priori _ , equally likely . in the absence of any clearly better alternatives , we adopt uniform ( or uninformative ) priors on each of the parameters in @xmath191 .
this includes uniform priors for , for example , the @xmath290s , which is equivalent to logarithmic priors for the @xmath234s . the exception to this rule is for the slope parameters for the linear relations that go into both @xmath291 and @xmath292 . here
we take uniform on the _ angle _ of the relation ; that is , our priors are uniform in , for example , @xmath293 and @xmath294 .
it is worth noting that with this choice of uniform priors , the prior function @xmath295 is constant , and so @xmath296 is directly proportional to @xmath297 . with this decision ,
the bayesian formalism thus all but reverts to that of traditional , frequentist statistics .
in other words , at least in our case , the only difference between the bayesian and the more familiar frequentist approach is that , as bayesians , our priors are made _
explicit_. formally , the pdf for the single parameter @xmath282 is derived by marginalising over all of the other parameters in @xmath191 ; ie : @xmath298 where @xmath299 is constant for uniform priors , and the integral should be understood to be evaluated for all @xmath300 .
( here , the symbol ` @xmath301 ' means the set complement . ) in words , this expression is best understood as a probability weighted integral over all possible combinations of parameter values , with the condition that the specific parameter of immediate interest , @xmath282 , takes the particular value @xmath283 .
note that this formalism also works for any quantity @xmath302 that can be deterministically computed from the defining parameters of the model .
this includes , for example , the values of the characteristic densities , @xmath303 , for each of the schechter functions .
this also includes the values of the individual @xmath304s ; _ i.e. _ , the values of the mfs at any of the discrete @xmath225s used to define the model . here , since @xmath304 is not a member of the defining parameter set @xmath191 , the ` complementary ' parameter set of @xmath305s in is the full set @xmath191 . in this way
, we can derive formal statistical uncertainties on @xmath304 , @xmath306 , or @xmath307 that fully account for any and all covariances between the 40 parameters in @xmath191 .
we will discuss this point in more detail in .
the fitting of the model thus entails mapping out the scalar likelihood function @xmath308 over the 40-dimensional parameter space defined by @xmath191 .
this is done using the technique of monte - carlo markov chain ( mcmc ) sampling .
in essence , mcmc is just a random walk through the high - dimensional parameter space .
the key to mcmc techniques is that possible steps are considered randomly , but are accepted or rejected probabilistically .
more specifically , the chances of a step being accepted are defined by the ratios of the pdf that is , by the prior - weighted likelihood function at the present and potential future locations in @xmath191-space . in the first instance
, this makes mcmc a very robust means of exploring the parameter space with a view to finding the global maximum of the pdf . in the second instance
once the algorithm has found itself near to the maximally likely solution mcmc sampling represents an extremely convenient means of sampling the high - dimensional pdf . in this phase of the fitting process
, the key to the utility of mcmc sampling is that it is _ ergodic _ ; that is , the chances of a point in @xmath191-space being sampled is directly proportional to the value of the pdf at that point . as a consequence
, the _ distribution _ of mcmc - sampled points converges to a faithful mapping of the _ value _ of the pdf in @xmath191 space .
this means that the marginalisation integral in can be very easily computed to a high level of accuracy by simply taking a histogram of sampled values of any of the quantity .
similarly , the joint pdf for any two ( or more ) quantities can be computed by taking the two- ( or more-)dimensional histogram over those parameters .
further , the marginalisation integral for the ` most likely ' value of any quantity @xmath302properly speaking , the expectation value for @xmath13can be trivially computed by taking the mean of all mcmc sampled values for that quantity ( _ cf . _ ) ; _ i.e. _ : @xmath309 , \end{split}\ ] ] where @xmath310 represents the individual ( post - convergence ) mcmc sampled sets of trial values for the parameters @xmath191 .
similarly , the uncertainty on the value of a single parameter can be simply computed as the rms of mcmc sampled values ( _ i.e. _ , @xmath311 ) , and the joint , covariant uncertainties on multiple parameters can be computed via the pearson correlation coefficient ( _ i.e. _ , as in eq .
[ eq : rho ] ) .
we have used the publicly available python package emcee @xcite to actually perform the mcmc fits presented in this work .
compared to the standard metropolis - hastings mcmc sampling algorithm , the most important feature of emcee stems from its use of multiple mcmc ` walkers ' when sampling the parameter space .
the step size for individual walkers are based on the distribution of the ensemble of all walkers , using an affine - invariant ` stretch move ' algorithm , which leads to very efficient sampling , even in the case of strongly anisotropic pdfs . for this work ,
a key practical advantage of using emcee is that it is trivially parallelisable .
the results of this mcmc fitting process are illustrated in * fig .
[ fig : mcmc]*. this figure shows the individual sets of trial values for each of the 40 parameters in @xmath191 for unique mcmc samplings .
these samplings are , in a sense , our results
they represent the high dimensional pdf for the values of the parameters in @xmath191 . for clarity , the tracks for individual parameters in fig .
[ fig : mcmc ] have been scaled according to the mean and rms values of the thinned and post - burn mcmc samplings ; that is , according to the bayesian estimator for the most likely value , and the uncertainty in that value .
these values are given in each panel of fig .
[ fig : mcmc ] ; this figure thus also serves as a table of the results of our fits . the fact that each of these lines is horizontal shows that the fits have in fact converged .
to the extent that the pdf for any given parameter value is truly gaussian , these values can be used to fully describe the pdfs .
again , the distribution of mcmc samplings converges to a faithful mapping of the pdf , with no embedded assumptions of gaussianity ; these distributions are shown in fig .
[ fig : mcmc ] as the grey histograms . in all cases ,
the pdfs can be seen to be well - sampled .
further , in most but not all cases , the pdfs can be seen to be well described by a simple gaussian .
again , we stress that these distributions naturally and fully account for covariances among the values of ( many ) different parameters
. how can we be satisfied that , for example , the b population really is ( or is not ) better described by a double- rather than a single - schechter mass function , or whether or not the blue cmr can be adequately described using just a simple linear relation ? to explore these kinds of issues , we have made many fits to our dataset , in which we have eliminated one or more of the 40 parameters that go into our most general model .
our tests have been systematic , but by no means exhaustive .
as described below , we have used these tests to ensure that we are not grossly overfitting the data .
there are a number of bayesian approaches to the problem of model selection which we have explored : the bayes factor , @xmath40 , the closely related akaike and bayesian information criteria ( aic and bic , respectively ) , and the deviance information criterion ( dic ) .
while these different approaches are each based on slightly different assumptions , and are thus strictly valid in slightly different circumstances , they can all be thought of as being similar in spirit to a traditional frequentist log - likelihood - ratio test . in asymptotic limits , aic @xmath86 bic @xmath86 dic @xmath312 .
a difference of 2.5 ( or 10 ) in the ic of two different models implies a likelihood ratio of @xmath86 3.5 ( or 150 ) , where the model with the lower value for the ic is the preferred one .
however , unlike a simple likelihood ratio test , each of these quantities includes an explicit or implicit penalty for larger numbers of parameters , so as to protect against over - fitting of the data .
we have focussed primarily on the aic and bic , which are defined with reference to the maximum of the likelihood function , @xmath313 .
is scaled by a factor @xmath314 .
a lower @xmath315 makes steps to lower values of @xmath316 harder than they would otherwise , effectively corralling the mcmc walkers where @xmath316 is high .
it is thus possible to robustly determine the value of @xmath313 by successively reducing the value of @xmath315 . ] since extra parameters can only increase the value of @xmath313 , the question is whether or not this improvement is sufficient to merit the inclusion of an extra parameter .
the penalty terms for additional parameters for the aic and bic are @xmath317 and @xmath318 , respectively , where @xmath109 is the number of free parameters in the model , and @xmath319 is the number of data points .
thus it can be seen that , all else being equal , the bic penalises additional parameters more strongly than the aic for @xmath320 .
( in our case , @xmath321 , so the bic penalty is roughly 5 times larger . ) in other words , the bic prefers models with fewer parameters .
thus , where the bic disfavours a simpler model , this model is definitely too simple , and should not be used .
conversely , the aic prefers models with more parameters .
where the aic disfavours a model with more parameters , then that the use of that model is definitely overfitting the data . in order to ensure that we are not abusing our data
, we have used these two information criteria to explore the consequences of omitting individual parameters from the model described above .
given the number of parameters that go into our general model in its most general form , it is impractical to do this in a properly exhaustive way . instead ,
starting from the most general form of our model , we have considered omitting individual parameters one at a time to see whether or how our model might be simplified .
first , we have tried successively omitting the parameters @xmath322 , @xmath323 , @xmath19 , and @xmath324 ( _ i.e. _ , the parameters that describe the step or bend in the locus of or scatter around the cmrs ) . when fitting to the @xmath41 cmd , it is possible that a simpler description of the b cmr is possible : omitting the parameters @xmath325 and @xmath326 improves both information criteria by 8 and by 10 , respectively . while this constitutes positive statistical evidence against the need for one of these parameters , it is ambiguous which one should be excluded .
when fitting to the @xmath44 cmd , the results are similar . based on the bic , it is preferable to omit either @xmath327 or @xmath328 ( but not both ) , and also possibly and @xmath326 and @xmath329 ( @xmath330bic = 8 for both ) .
the ambiguity in these results make perfect sense looking at the table of results given in fig .
[ fig : mcmc ] . these parameters which may or may not be necessary are those whose fit values are statistically consistent with being zero .
their inclusion or exclusion thus makes little if any difference to the fits , and the decision as to whether or not to include these parameters has no practical consequences .
we therefore elect to use the most general description possible for the locii of and scatters around the cmrs , by fitting for all of these parameters . in this sense , our results
can be thought of as hypothesis testing the need for each of the @xmath322s , @xmath323s , @xmath19s and @xmath324s .
where the fit values for any of these parameters are consistent with zero , then that parameter may be unnecessary for a good description of the data .
these results nonetheless encapsulate positive information about the forms of locii of and scatters around the b and r cmrs .
perhaps more interesting is what happens when we trial alternate descriptions of the mfs . as well as the general double schechter parameterisation
, we have trialled a coupled twin - schechter description , where the two schechter functions that describe either the r or b population have the same characteristic mass ( _ i.e. _ , @xmath331 ) .
we have also trialled using only a single schechter function ( _ i.e. _ , @xmath332 . when fitting to the @xmath41 cmd , the bic definitely disfavours a single schechter function description for the b mf ( @xmath330bic @xmath333 ) , indicating that such a model definitely underfits the data .
the bic also definitely disfavours the coupled , twin schechter function description for the b mf ; the data definitely prefer a double schechter function description of the b mf , with @xmath334 . at the other extreme
, the aic disfavours the most general , double schechter model ( @xmath330aic @xmath335 ; odds @xmath86 20:1 ) for the r mf described .
that said , it is worth noting that both the aic and bic ( weakly ) prefer a single schechter description for the r mf .
this may not be surprising , given how weakly constrained the values of @xmath336 , and @xmath337 are .
when fitting to the @xmath44 cmd , the bic disfavours a single schechter mf for the r population ( @xmath330bic @xmath338 ) .
the bic does not obviously prefer the more complicated , independent double schechter descriptions of either the b or r mfs ; the coupled , twin schechter functions are just as good .
the variation of the model that best balances between the two criteria is the one that uses coupled , twin schechter functions to describe both the b and the r mf . in light of all of the above
, we will continue our analysis using the most general form for our model , which is fully defined by 40 parameters .
this is despite the fact that both the aic and bic prefer a single schechter description for the r mf when analysing the @xmath41 cmd .
( but of course , this being the case , the fits do not make use of the additional freedom that the second schechter component provides , precisely because it is not necessary for a good description of the data . ) in this sense , our fit results can be taken as limiting the deviations from schechter - ness in the observed mf for r type galaxies . for all of the rest
, we can say that we are not definitely overfitting the data , nor are we definitely underfitting the data . beyond this point
, however , all we can say is that we have done the best that we know how ( and monopolised 64 cores for more than 2 months ) to ensure that we are using the best and simplest model that we can to describe the existing data . in some of the above
, there is some ambiguity , inasmuch as the data do not provide strong evidence for or against the inclusion of some parameters .
this is particularly true for whether or how the b and r mfs should be described with a combination of schechter functions . to the extent that there is ambiguity , our decisions about whether or not to include these parameters are subjective , hence _
arbitrary_. to the extent that these points are ambiguous , however , they are also _ unimportant_. this is true in the sense that these decisions have no strong impact on the statistical quality of the fits to the data .
this is also true in the sense that the best fit models based on the less - general parameterisations provide essentially identical results .
what we have done in the above is to explore special cases or restricted classes of the general form of our model .
the question that naturally arises is whether there are some alternative parameterisation might provide a better description of the data .
for instance , it could be that a lorentzian or a student s @xmath339 or a skewed gaussian description of the scatter around the cmrs yields a better statistical description of the data .
of course , a properly exhaustive search of all conceivable models is impossible .. remarkably though , using a fifth order ( _ i.e. _ , six parameter ) polynomial description for @xmath215 , we obtain virtually identical results : the differences in the fit values of @xmath340 are nowhere more than 0.01 mag , and the ic is only slightly worse . ]
this brings us to the final and most important caveat on our results . while the formal statistical uncertainties on our cmr and mf fits are impressively small , the values themselves _ can not but _ be determined by the decisions made in constructing the parameterised , descriptive model that has been used . on the other hand
, the same criticism can be levelled at _ any _ model or modeller even a fully physically - minded explanatory model is forced to presuppose the validity of the theoretical framework on which it is based . at this level ,
some degree of subjectivity is inescapable .
once we have set the parametric form of our descriptive model , however , it is then left to the data to decide on the particular parameter values that provide the best description , including all the characteristics of the b- and r- populations .
that is , given our choice of parameterisation , it is the data themselves that define the bimodal distributions .
the results of this modelling thus provide objective b / r classifications _ insofar as objectivity is possible_. these results can thus be said to provide an accurate , reliable , and robust phenomenological description of the observed data and this is all that we have set out to obtain .
in order to illustrate our ability to perform such a detailed descriptive modelling of the data , as well as the quality of the resultant fits , consider * fig . [
fig : rfdist ] * and * fig .
[ fig : stardist]*. our task in this section is to describe and discuss these figures .
building on the discussion begun in , we will also flag the major issues with , and caveats on , the interpretation of our results , which should be kept in mind in all that follows .
[ fig : rfdist ] and fig .
[ fig : stardist ] show the @xmath41 and @xmath44 cmds , respectively . in each panel , we have split our sample up into bins of @xmath73 , each with width 0.2 dex .
the solid black histograms then show the incompleteness - corrected colour distributions within each of these mass bins : the lowest histogram is for the @xmath341 bin , the next is for the @xmath342 bin , and so on .
the normalisation of each of these histograms is arbitrary , but is the same for all bins in both figures .
illustrative statistical uncertainties in the observed colour distributions are shown by the black error bars .
these uncertainties have been estimated using bootstrap resampling .
( we have not attempted to account for field to field variance due to large scale structure . ) for @xmath343 , these statistical uncertainties are negligible .
further , even for the lowest @xmath341 bin , the statistics are reasonably good . particularly for @xmath344 , then , there should be no question as to whether or not the data are good enough to allow an independent double - gaussian ( or some other parametric ) fit to any or all of the individual , mass - binned colour distributions .
the smooth curves in fig . [ fig : rfdist ] and fig . [ fig : stardist ] show the results of our descriptive modelling of the bivariate @xmath41 and @xmath44 colour - mass distributions .
the red and blue curves show the distinct r and b components of the model ; the black curve is the total , r plus b distribution .
note that for these illustrative purposes , we have convolved the modelled colour - distributions with typical @xmath41 or @xmath44 uncertainties for each mass bin , so that these curves are directly comparable to histograms for the observed data . before moving on , let us stress that we have _ not _ binned the data in the course of fitting it : the binning in mass and colour in these figures is for illustrative purposes only .
further , nowhere in the modelling is it ever specified or even relevant whether any particular galaxy belongs to either the r or the b populations ; there is no binning in this sense either .
if , for example , we were to just fit double gaussian distributions to the observed colour distributions for the distinct mass bins that are shown in fig .
[ fig : stardist ] , then the inferred values for the centres , scatters , and normalisations of the r and b distributions in each bin would be completely independent . given the relatively poor sampling for @xmath345 and @xmath346 galaxies , we would have no means of robustly constraining the properties of either of the two gaussian components of the @xmath44 colour distributions for these very low masses or even whether or not two gaussian components should be used .
but this is not what we have done . in effect , we have assumed that the centres , scatters , and normalisations of the ( gaussian ) b and r colour distributions vary smoothly as a function of mass . in this sense ,
the derivation of each and every of the individual modelled b and r colour distributions shown in fig .
[ fig : stardist ] is based on each and every datapoint that we have .
ultimately , all the model considers is the relative probability of finding a galaxy irrespective of type at a particular point in the colour mass plane . in other words
, we are using a mixture model of two distinct but overlapping populations to describe the joint colour mass distribution function of _ all _ galaxies we are fitting for the 2d scalar function that is represented in fig . [
fig : rfdist ] and fig .
[ fig : stardist ] as the black curve . in this way
, we are able to characterise the cmrs and mfs of the two populations _ without ever explicitly considering which galaxies belong to which population_. @xmath50 distribution. the histograms in this figure show the observed @xmath102-weighted distribution of restframe @xmath41 colours of @xmath0 galaxies , computed in bins of @xmath42 centred on @xmath42 = 8.8 , 9.0 , ... , 11.4 . the errorbars show the statistical uncertainties on each these distributions , derived by bootstrap resampling .
the smooth curves show the results of our modelling : the blue and red curves show the fit distributions for the b- and r - populations ; the black curves shows the net b+r distributions .
underlaid beneath all this , the grey points show the data themselves .
the size of each point is proportional to @xmath102 .
data objectively classified as being ` bad ' are marked with a cross .
note that we have not binned the data in the course of the fits ; the binning in this figure is for illustrative purposes only .
it is clear that the fit model provides a good description of the observed data .
[ fig : rfdist ] , height=714 ] @xmath50 distribution. the histograms and smooth curves in this figure show the observed and fit distributions of intrinsic @xmath44 colours in bins of @xmath42 ; all symbols in this figure and their meanings are analogous to fig .
[ fig : rfdist ] .
in contrast to fig . [
fig : rfdist ] , the general galaxy population is better separated into distinct r and b populations on the basis of @xmath44 than of @xmath41 , in the sense that there is less overlap between these two distributions in this figure than in fig .
[ fig : rfdist ] .
further , as in fig .
[ fig : rfdist ] , we see no compelling evidence for the need to include a third , ` green ' population . in comparison to fig .
[ fig : rfdist ] , the need for a multiple - gaussian description of the _ observed _ colour distributions for @xmath345 is less clear ; this point is discussed in detail in . in any case , as in fig .
[ fig : rfdist ] , the fit model can be seen to provide an excellent description of the observed data .
[ fig : stardist ] , height=714 ] while our parametric model does provide a very good description of the data , _ a good _ description is not necessarily the same as _ the right _
one ( see related discussions in and ) .
other parameterisations of the colour distributions ( at fixed mass ) are possible , and may lead to different results but this is _ always _ true . the fact that the answer we get depends on how we have devised
our analysis is inescapable .
if our assumptions , which are clearly stated and motivated in and , are shown to be invalid , then our results go with them .
indeed , we have gone to great lengths to make this point in . without denying these inescapable truisms
, the fact remains that our model does provide a very good description of the data , and so offers one potential avenue for understanding the data . at least for @xmath188
, the model does perform its intended function : decomposing the observed data into a mixture of two populations , which are distinguished and defined by their own distributions of colours , which we take to be a tracer of the constituent stellar populations .
the skeptical reader s eye may be drawn to the fits at @xmath128 , however , where the suitability of a double gaussian fit becomes increasingly problematic . certainly , in both fig .
[ fig : rfdist ] and fig .
[ fig : stardist ] , below @xmath3 , it becomes difficult for us to claim that we have robustly separated the general galaxy population into two distinct r- and b - populations .
how then should one interpret our results at these low masses ? considering this problem from the modelling perspective , we should ask what aspects of the data drive the fits most strongly . clearly , it is the shape and normalisation of the b colour distributions that are the best constrained at these low masses .
also , from the nature of the fits , it should be clear that the fits to the b colour distributions are decided primarily by the @xmath347 or @xmath348 data .
bluewards of the peak of the observed colour distributions , the data are well described by a gaussian .
then , since the colour distributions for each population are assumed to be symmetric , it is left to the r population to accommodate whatever asymmetries there are in the observed colour distributions . from a more astrophysical perspective , then , what matters is the extent to which the colour distribution for the b population is expected to be symmetric . however , this will certainly not be the case for @xmath41 , where the asymmetric effect of dust is expected to skew the b colour distribution to the red . in this case
, one would expect the inferred number of r galaxies to absorb some of the reddest b - type galaxies .
indeed , looking at the lowest two or three bins of fig .
[ fig : rfdist ] , one can see immediately how such a description might work . rather than trying to fit for an asymmetric or skewed @xmath41 distribution ( which would be considerably computationally more complicated )
, we can look at the @xmath44 distribution , where we have tried to remove dust as a complicating factor . but looking at the lower three bins in fig .
[ fig : stardist ] , there is no way to decide whether the slight asymmetry of the observed colour distribution ought to be interpreted as an indication of a separate population , or instead as nothing more than a slight asymmetry in the @xmath44 colour distribution of the b population .
our conclusion is therefore that at these low masses , the ( field ) red sequence dissolves into obscurity we no longer see clear evidence of two distinct populations in the @xmath41 or @xmath44 colour distributions for @xmath349 .
( as we have argued in , we do not believe that the apparent dearth of low mass red galaxies is due to incompleteness . ) instead , we present our inferred mfs for r - type galaxies as _ an upper limit _ on the number densities of galaxies that have moved ( or are moving ) away from the colour distribution that describes most low - mass galaxies ; _ i.e. _ , the b population . again , the fact that the inferred b- and r- populations have substantial overlap serves to underline the subtleties involved in interpreting our results in concrete , astrophysical terms .
in particular , even at intermediate masses , it would be unwise to blithely equate the r population with ` quenched'but recognise how much more problematic it is to apply the term ` quenched ' to the hard - cut ` red ' samples discussed in . to sum up : using the parametric model described in , we have derived a very good description of the observed bivariate distributions between both @xmath41 and @xmath44 and stellar mass .
this analysis is intended to provide a phenomenological description of the essential characteristics of the bivariate colour mass distributions .
such a description clearly requires ( at least ) two populations with their own distinct cmrs and mfs .
our approach enables us to simultaneously and self - consistently describe the bivariate colour - mass distribution functions of the two populations ; indeed , this is how these populations are defined .
again , we stress that the designations b and r refer primarily to the two _ populations _ , rather than to individual galaxies . that is , instead of characterising the demographics of galaxy samples that are pre - selected to be ` blue ' or ` red ' , what we have done is decompose the full population into a mixture of two distinct , but overlapping , subpopulations ; we dub these two populations ` b ' and ` r ' . in this way , we can _ derive _ an operational definition for the terms ` r - type ' and ` b - type ' . once this is done , however , it is possible to use these fits to quantitatively if probabilistically
classify individual galaxies according to the chances that the galaxy in question has been drawn from either the b or r population ; _
i.e. _ , the relative contributions of the b- and r - populations to the data density at any given point in the cmd .
( see , below . ) but it is important to understand that these classifications _ follow from _ the fits : during the fitting process itself , the b - ness or r - ness of any particular galaxy is irrelevant .
we make no pretensions , however , about providing an explanatory model for the observations .
that is , we justify our splitting of the general population into b and r components on the grounds that _ these are distinctions that galaxies somehow care about_. in this way , we derive a phenomenological description of the bimodal
better , the two population character of the galaxy population , in terms of galaxies stellar populations .
the underlying physical differences in the origins and natures of the b- and r - populations , as well as those responsible for the observed ranges of @xmath44 colours within each population , remain to be determined ( but see ) , and will be the focus of future works in this series .
cmd , highlighting the cmrs for the b and r populations. this figure is discussed at length in . in this figure , the points show the data . as in fig .
[ fig : rfdist ] the size of each point reflects the value of the @xmath102 weighting for incompleteness , and the crosses show those datapoints objectively identified and censored as ` bad ' .
the small arrows show where galaxies fall outside the plotted range .
the filled greyscale contours show the incompleteness - corrected data density , in steps of powers of 2 .
these should be compared to the smooth line contours , which show the bivariate distribution function from our fits . for illustration purposes ,
these have been convolved with the typical observational errors , shown at the bottom right . here , the dotted ellipse shows the inferred errors for ` bad ' galaxies .
the main feature of this plot are the smooth curves , which show our fits for the locii of , and scatter around , the @xmath41 cmrs for the distinct r- and b - populations .
the width of these lines show the 68% confidence intervals on the values of these functions , as a function of mass , and including all convariances between model parameters .
this figure should be contrasted with fig .
[ fig : starcmr ] .
[ fig : rfcmr],width=325 ] in fig .
[ fig : rfdist ] and fig .
[ fig : stardist ] , we have illustrated the quality of our fits to the bivariate @xmath42@xmath41 and @xmath42@xmath44 distributions for our sample of @xmath1 and @xmath0 galaxies .
our task in this section is to lay out the actual fit results_i.e .
_ , the cmrs and mfs for r- and b - type galaxies which describe the two populations .
fit results are given in a machine readable table as additional online material .
the results of our fits for the @xmath41 and @xmath44 cmrs are shown in fig.s [ fig : rfcmr ] and [ fig : starcmr ] , respectively . in both of these figures ,
the grey points show the data themselves . as in figures [ fig : cmds ] , [ fig : rfdist ] , and [ fig : stardist ] , the size of each point has been chosen to reflect the value of the @xmath102 weighting factor .
further , as in figures [ fig : rfdist ] and [ fig : stardist ] , each individual datapoint is marked with a black cross , the size of which has been chosen to reflect the probability that that datapoint has been drawn from the ` bad ' distribution .
the marked points have thus been objectively identified as outliers ; they make little to no contribution to the fit cmrs shown .
this objective censoring can be seen to be very effective .
the semi - transparent filled contours in fig . [
fig : rfcmr ] and fig .
[ fig : starcmr ] show the observed , @xmath102-weighted bivariate data density ; these contours have the same log@xmath350 scaling as those in fig.s [ fig : colours][fig : otherdust ] .
these should be compared to the smooth line contours , which are interlaid between the data points and the fit cmrs , and which show the logarithmic probability density contours from the model . as for fig .
[ fig : rfdist ] and fig .
[ fig : stardist ] , to generate these contours , we have convolved the model fits with typical uncertainties , so that these contours are directly comparable to those for the data themselves .
these contours thus reflect the combination of the mass functions and the cmrs for these populations .
these contours are included mostly for illustration ; the mass functions themselves are shown in fig.s [ fig : rfmf ] and [ fig : starmf ] , and are described separately below .
the models provide very good descriptions of the observed bivariate @xmath41 and @xmath44 colour - mass distributions . in each figure ,
the smooth lines show the fit locii for the two cmrs , as well as the fits for the rms scatter around each cmr .
the width of each of these lines shows , at fixed mass , the 68% confidence intervals for each quantity . these uncertainties
can be seen to behave reasonably : they are very small where the data concentration is high ( _ e.g. _ , the centres of the red sequences ) , and become large where the data concentration is low ( _ e.g. _ , the very high mass ends of both the red and blue sequences ) .
in connection with our discussion in , note how the uncertainties in the locations of the @xmath351 point of the r @xmath44 distributions are less than that of the locus of the cmr , which are in turn smaller than those of the @xmath352 point .
this again shows how the descriptions of each population are constrained principally by the outer edges of the colour distributions , as well as illustrating how our bayesian approach can yield meaningful uncertainties on secondary aspects of the model ( _ e.g. _ , the cmrs ) that nicely and naturally propagate all the relevant uncertainties in the defining parameters of the model .
cmd highlighting the cmrs for the b and r populations. this figure is discussed at length in .
all symbols and their meanings are analogous to fig .
[ fig : rfcmr ] . in comparison to the @xmath41 cmd
, we make the following qualitative observations . the ` blue sequence ' in this @xmath44 cmd is both tighter and more linear than the @xmath41 ` blue cloud ' .
this implies that the upturn in the ` blue ' cmr seen in fig .
[ fig : rfcmr ] is caused by a change in the dust properties of blue galaxies with @xmath353 , rather than a change in the stellar populations of these galaxies .
note how the @xmath354 upturn to the b cmr in this figure coincides with the convergence between the @xmath41 cmr for the two populations in fig .
[ fig : rfcmr ] .
the rather steeper slope of the r - type cmr the shows how this population is less homogenous than the b - population : lower mass r - type galaxies have rather different stellar populations to their higher mass cousins . at the same time , the relatively tight and smoothly sloping cmr show how across the r - type population , mass is a relatively good predictor of stellar population , suggesting a common evolutionary pathway for these galaxies .
these points are discussed further in .
[ fig : starcmr ] , width=325 ] in fig .
[ fig : rfcmr ] , there is an upturn to the b cmr that begins around @xmath355 . as expected from fig .
[ fig : colours ] , the behaviour seen in fig .
[ fig : starcmr ] is rather different : what is seen in the @xmath41 cmd as the blue cloud is seen in the @xmath44 cmd as a considerably tighter and more linear blue sequence .
this implies that the slope of the @xmath41 cmr in fig .
[ fig : rfcmr ] for the b population is more a product of increasing dust obscuration in higher mass galaxies , rather than differences in the colours of the underlying stellar populations . at the very end of this @xmath44 blue sequence
, there is the hint of an upturn to redder @xmath44 colours for @xmath356 , but this is where the uncertainties become large .
intriguingly , looking at fig .
[ fig : rfcmr ] , the b cmr becomes indistinguishable from the r one in the @xmath41 cmd at this mass range : the locii of the two cmrs converge , and the scatter in the b cmr becomes small .
that is , the b population becomes indistinguishable in the apparent @xmath41 cmd for @xmath356 .
further , those @xmath354 b - type galaxies identified in the intrinsic @xmath44 cmd have rather different stellar populations to the rest of the blue sequence . turning to the r population , the most remarkable aspect of the @xmath41 cmd ( fig .
[ fig : rfcmr ] ) is how the @xmath357 r cmr flattens and tapers to have essentially no intrinsic scatter .
these are the galaxies that one might expect to be truly ` red and dead ' . below this mass ,
the @xmath41 cmr for the r population bends to bluer colours for lower masses .
the @xmath44 cmr shows slightly different behaviour : here , the stellar populations of r - type galaxies become very gradually redder across the range @xmath358 .
taking the results shown in fig .
[ fig : starcmr ] at face value , the simplest interpretation would be that the r - type galaxies are moving towards the ` dead sequence ' only slowly , and in such a way that creates or preserves a relatively tight relation between a galaxy s mass and its stellar population . in this scenario
, the higher mass galaxies would appear to have progressed further in this long migration .
we will expand further on the simple observations above in .
colour mass diagram. this figure is discussed in detail in .
there are two main features to this figure .
first , the smooth line contours show the bivariate distributions for the b and r components of the models , using the same logarithmic scale as in earlier plots .
this information can be used to classify individual galaxies according to the probability that they have been drawn from one or the other population .
individual points in this figure are colour - coded according to these classifications , @xmath359 .
given the empirical fact of scatter in both the blue cloud and the red sequence ( see fig . [
fig : rfdist ] ) , and thus the overlap between the two populations , the individual classifications can be ambiguous , particularly at high masses .
for example , a galaxy observed right on the locus of the red sequence still has a @xmath36025% chance of having come from the bluer b - population .
[ fig : rffrac],width=325 ] colour mass diagram. all symbols and their meanings are analogous to fig .
[ fig : rffrac ] .
note how , particularly at higher masses , the r - population is more unambiguously distinguished in the @xmath44 cmd than the @xmath41 cmd shown in fig .
[ fig : rffrac ] ; only a small fraction of galaxies have ambiguous classifications ( _ i.e. _ , intermediate values of @xmath359 ) .
as discussed in and , this means that our empirical determinations of the mfs for b / r classified galaxies are not strongly dependent on the quality of our fits . at lower masses , and particularly around @xmath361 , the populations are inferred to overlap to the extent that up to @xmath360% of galaxies in the blue sequence are members of the r population .
the overlap between the two populations in this figure , as well as fig .
[ fig : rffrac ] demonstrate the importance of accounting for scatter around the cmr when characterising the two populations .
[ fig : starfrac ] , width=325 ] in fig .
[ fig : rffrac ] and fig .
[ fig : starfrac ] , we re - present the results of our fits to the @xmath41 and @xmath44 cmds in a different way , in order to illustrate how these fits can be used to develop an objective , quantitative b / r classification scheme for individual galaxies .
in essence , the idea is to give each galaxy a score , which encapsulates the relative probability , based on our fits , that that galaxy has been drawn from the r- , b- , or even the ` bad ' population .
the r - type score , @xmath362 is given by the relative number of r - type galaxies expected to be found at the location @xmath363 , after convolving the fit models with the observational uncertainties encapsulated within the covariance matrix @xmath77 .
formally , and using the notation and definitions introduced in , these values are computed as : @xmath364 here , @xmath365 is the set of fit - for values for the parameters @xmath191 , and @xmath366 .
and of course @xmath367 and @xmath368 can be defined / computed in an analogous way , so that @xmath369 .
we illustrate how these classifications work in fig .
[ fig : rffrac ] and fig .
[ fig : starfrac ] . in these figures ,
the data have been colour coded according to their particular values of @xmath359 .
note that , because these classifications depend on the measurement uncertainties as well as the measured values themselves , there can be some variation in the @xmath359s for galaxies with very similar colours and masses .
note how , particularly at higher masses ( @xmath132 ) , there is considerably less ambiguity in the classifications based on the @xmath44 cmd , in comparison to those based on the @xmath41 cmd . as can be seen in fig .
[ fig : rffrac ] , the b and r populations overlap in the @xmath41 cmd to the extent that , even along the locus of the r cmr , @xmath370 - 15% of galaxies come from the b population ( see also fig . [
fig : rfdist ] ) .
note , too , how the situation is reversed for @xmath100 .
the point to be made here is that , where there is substantial overlap between the two populations , it is not possible to unambiguously determine whether a particular galaxy is a member of the r or the b population without some additional information . it is a legitimate question to ask what additional information could or should be used to refine these classifications . while it may be tempting to incorporate morphological or structural information srsic index , @xmath319 , for example into the ` red ' selections , we note that this would result in a sample of ` red and high-@xmath319 ' galaxies , rather than a sample of ` red ' galaxies .
( see * ? ? ? * for a detailed discussion of this problem . ) in this sense , there is the very real danger that inclusion of additional parameters makes things more confusing , rather than less .
we will explore this issue further in future papers in this series . in this context , we note that these objective , quantitative classifications can also be cast as _ weights _ , which can be useful in studying the properties of r- and b - type galaxy samples .
[ fig : rffrac ] and fig .
[ fig : starfrac ] can thus be alternatively understood as illustrating a ` soft ' red / blue selection scheme , which is a smooth function of mass and colour .
this scheme naturally accounts for the scatter around the r and b cmrs , as well as the r / b fraction , as a function of mass .
it also fully accounts for photometric scatter , due to error / uncertainties in the measurements themselves .
this figure thus helps show how our ` b- ' and ` r - type ' designations are qualitatively different to simple ` blue ' and ` red ' colour selections . as discussed in detail in , below , this is the crucial point that explains why our determinations of the mfs for the b and r populations differ strongly from those for ` blue ' and ` red ' galaxies given by previous authors .
cmd. the smooth blue and red curves show the fit mass functions for the b and r populations ; the black curve shows the net b plus r mass function .
the inset panel shows the two separate schechter components that go into each of the b and r mfs .
the solid lines in this inset show ` the ' fit mfs ; the shaded regions show the 68 and 99% confidence intervals for each component .
the uncertainties on the total b and r mfs are shown in the same way in the main panel .
although the individual schechter components are partially degenerate , the overall mfs are very well constrained . except for the highest masses ,
the uncertainties are comparable to the width of the lines used to show the fits , and so are difficult to see . for comparison to the fit results ,
the blue and red histograms show the empirical mfs , where individual galaxies have been weighted by their values of @xmath359 or @xmath367 .
_ the curves are not fit to the histograms _ ; instead , the histograms are derived using the quantitative classifications that come from the fits .
[ fig : rfmf],width=325 ] cmd. in contrast to fig .
[ fig : rfmf ] , this figure shows the mass functions for the r and b populations as inferred from the @xmath44 , rather than the @xmath41 , cmd ; otherwise , all symbols and their meanings are the same as in fig . [
fig : rfmf ] . note that we do not consider the apparent low mass upturn in the r mf to be robust , for the reasons given in , and ( see also fig .
[ fig : stardist ] ) .
further , there is not conclusive statistical evidence that a second schechter component to the r mf is required .
the quantitative discrepancies between the results shown in this figure and in fig .
[ fig : rfmf ] for @xmath371 can be understood looking at the distributions shown in fig .
[ fig : rfdist ] and fig .
[ fig : stardist ] .
the implication from this figure and fig .
[ fig : rfmf ] is that systematic uncertainties in the low - mass values of the r mf are at least a factor of 2 .
it is worth stressing , however , that the two sets of results agree at the @xmath372 level for all @xmath132 .
compared to previous results , we find that the b - population_i.e .
_ , galaxies with young stellar populations extends to much higher masses , and we find considerably fewer ` old ' galaxies at lower masses .
( see also fig .
[ fig : punchline]).[fig : starmf ] , width=325 ] the inferred mfs for the r and b populations , based on the @xmath41 and @xmath44 cmds , are shown in fig .
[ fig : rfmf ] and fig .
[ fig : starmf ] , respectively . in the main panels ,
the solid lines show our fits for the mass functions when fitting the model as described in .
the heavier and lighter shaded regions around each line show the 68% and 99% confidence intervals for each mass function .
note that these are only visible at the highest masses , and at the low - mass end of the r mf .
elsewhere , the formal statistical uncertainties in the fits are smaller than the width of the lines . in both cases ,
the results shown in these figures involve implicit marginalisation / averaging over all possible permutations of counting each individual galaxy as being either b- or r - type .
it is true that , in general , we can not necessarily say with absolute confidence whether any particular galaxy is r - type or b - type . as discussed in , however , _
this has virtually no impact on the statistical accuracy of our determination of the mass functions_. we can quantify this last statement by marginalising over the uncertainties in the b / r classifications _ explicitly_. we have done this by recomputing the values shown by the histograms ( discussed further below ) for different mcmc - samplings of the model parameter space , and so over the range of statistically allowed classifications for each independent galaxy .
the 1@xmath373 error on the r mf is typically 25% , and is everywhere @xmath374% ; for the b mf , the error is at the 12% level for @xmath375 , rising to @xmath376% by @xmath377 . for comparison , we can also quantify the statistical uncertainties on the mf determinations shown in figures [ fig : rfmf ] and [ fig : starmf ] , via bootstrap resampling .
even using relatively wide mass bins of 0.10 dex , the statistical uncertainties from sampling are everywhere at least twice as large as those from uncertainties in the classifications , and are more typically larger by a factor of @xmath378 .
recall from that , within the modelling , each of the b and r mfs is parametrically described as the sum of two schechter functions .
this is illustrated in the insets to each of figures [ fig : rfmf ] and [ fig : starmf ] , which show the individual schechter components of the b and r mfs .
note that while the values of these individual components are generally not well constrained , the uncertainties or degeneracies in the separate components are largely irrelevant to the results . the uncertainties in the primary and secondary schechter functions are so strongly covariant that they leave the overall mf essentially unchanged . in this sense ,
the parameters that describe the individual schechter components can be treated as ` nuisance parameters ' to be marginalised over in order to determine the values of the mfs for the two populations at any given mass ( see also footnote 10 . )
while the secondary schechter component for the r mf is constrained to be negligibly small for all @xmath379 , both fits show an excess above a simple schechter function for @xmath128 .
this is more pronounced in fig .
[ fig : starmf ] than in fig .
[ fig : rfmf ] .
we do not consider this ( mildly pathological ) behaviour robust , for a number of reasons .
first , as discussed in , there is not a clear statistical need for a second schechter function to describe the r mf .
second , as discussed in , based on the binned colour distributions in fig .
[ fig : stardist ] , it is not clear whether or not our two - population model really provides a good description of the data at these low masses .
third , we note that this behaviour is not clearly seen in the red histograms , which are relatively flat across the range @xmath380 .
this last point requires some explanation .
the blue and red histograms shown have been derived from the data , by weighting the contribution of each datum by both @xmath102 , and also by the relative likelihood of each galaxy having been drawn from the b or r population that is , by either @xmath367 or @xmath359 , defined as per . in this sense , rather than
the curves being fit to the histograms , the histograms are in fact _ derived based on the fits_. in general , there is very good agreement between the fit mfs and these weighted mass functions for the data .
but it is important to recognise that this agreement is not strictly by construction .
the difference between the two sets of the results is subtle , but important . as a simple example to illustrate this point , imagine if we had only used a single schechter function to describe both the b and the r populations .
given that the b mf is demonstrably non - schechter ( the two components are not completely degenerate ) , we would not be using ` the right ' description of the b mf .
but given how slight the deviations are from a pure schechter function , we would not see very large differences in the values of the @xmath359s or @xmath367s . in this case
, the histograms in these figures would change hardly at all : they would still show the same non - schechter features as are seen in fig .
[ fig : rfmf ] or fig .
[ fig : starmf ] .
( in fact , when we do this experiment , the values of the histograms change by @xmath27% . ) in this way , the close agreement between the curves and the histograms thus provides a useful _ consistency check _ , inasmuch as it shows that the fit cmrs and mfs do provide a reasonable _ and self - consistent _ description of the data .
in fact , for this reason , we favour these empirical mf measurements over the fits ; it is these histograms that should be taken as ` the ' results of our analysis .
( the exception is at the highest masses , where the histograms are prone to the eddington bias in the normal way . )
the fits assume that we have used _ the right _
parameterisation for our model . instead , the weightings that go into the histograms rely only our having chosen _ a good _ parameterisation or at least a parameterisation that is _ good enough _ to derive reasonable classifications .
that is , the histograms have a weaker dependence on the precise parameterisation used to construct the model .
first and foremost , let us stress what we do _ not _ observe .
virtually nowhere are the colour distributions of galaxies _ at fixed mass _ observed to be ` bimodal ' in the strict definition of the word . almost none of the distributions in either fig .
[ fig : rfdist ] or fig .
[ fig : stardist ] have two distinct peaks . with the exception of the limited range @xmath381 , and even then only in the @xmath44 cmd , there is substantial overlap between the two galaxy populations .
while our r population fits do extract a distinct and relatively narrow ` red sequence ' , particularly over the mass range @xmath382 , we do not see this continuing down to lower masses in the either the @xmath41 or the @xmath44 cmd .
there are essentially no galaxies with @xmath41 or @xmath44 colours that are consistent with ` red and dead ' with stellar masses @xmath338 .
instead , the @xmath44 cmr for the r population becomes progressively bluer towards lower masses .
indeed , it becomes highly problematic to distinguish two separate populations , in either the @xmath41 or the @xmath44 cmd , for masses below @xmath3 ( see ) .
in other words , the ` red sequence ' dissolves into obscurity for @xmath383 .
further , we see no clear evidence for a low - mass upturn to the mf for the r population of the kind discussed by @xcite ( see in particular fig . [
fig : rfdist ] ) .
we do not believe that these results can be easily explained by mass incompleteness , for the reasons given in and . nor
do we believe that these results can be easily explained by outliers or otherwise ` bad ' data , for the reasons given in and .
that said , at least over the range @xmath384 , the r and b populations can be seen in both fig .
[ fig : stardist ] and fig .
[ fig : starcmr ] to be remarkably well separated in the @xmath44 cmd .
for @xmath385 , the fits can be seen to provide an excellent description of what has been dubbed the ` blue cloud ' in the effective , restframe cmd , and what we see as a more uniform ` blue sequence ' in the dust - corrected , intrinsic stellar cmd . as a corollary to this observation
, we note that , based on either fig . [
fig : rfdist ] or fig .
[ fig : stardist ] , there is no obvious need for the inclusion of a distinct ` green ' , intermediate or transition population .
the data are extremely well described by the double gaussian model . given all of the above , as we describe the basic properties of the bivariate @xmath41 and @xmath44@xmath50 distributions in this section , we will relax but not completely abandon our self - imposed prohibition against using the terms ` blue ' and ` red ' in connection with our b- and r - population fits .
however , we will limit ourselves to using the terms ` blue ' and ` red ' to those regimes where the b and r population fits can be directly related to the empirically and astrophysically sensible blue and red sequences described above ( with all the appropriate caveats ) . with this as introduction
, we make the following qualitative observations about the bivariate colour mass distributions from our fits to the r and b populations . 1 . )
the mf for the r - type population is relatively constant ( @xmath386@xmath387 mpc@xmath239 dex@xmath242 ) for @xmath388 .
there is the possibility of a slight upturn to the r mf ( albeit at a much lower level than that reported by * ? ? ?
* see ) , but we do not consider this result to be robust . the very smooth decline in the _ relative _ numbers of r - type galaxies_i.e._the r - type fraction towards lower masses suggests that mass is not _ the _ critical parameter for determining which population a galaxy is a member of : mass is not a good predictor of b / r - ness .
that is , even though more massive galaxies are more likely to have ` old ' stellar populations , quenching can not be ( uniquely ) associated with mass . 2 .
) with regards to the non - schechter features in the overall galaxy mf , there are very slight but statistically significant deviations from a simple schechter function in the b mf .
specifically , the fits suggest a slight deficit of galaxies with @xmath38910.3 , which coincides with the apparent ` dip ' in the overall mf . below this mass ,
the upturn in the total mf is clearly associated with b - type , rather than r - type galaxies . 3 . ) at the very highest masses , the @xmath41 and @xmath44 colours of the r population are consistent with ` red and dead ' stellar populations . for @xmath390 , however , the slope of the @xmath44 cmr implies that r - type galaxies at these intermediate - to - low masses have relatively younger luminosity - weighted mean stellar ages and higher dust contents than their higher mass cousins .
taken at face value , this would imply that the @xmath391 r population has not yet evolved into fully fledged ` red and dead ' galaxies .
certainly we can say that , even within the r population , there are very few @xmath100 galaxies with genuinely ` red and dead ' stellar populations ( see also ) . 4 . ) at high masses ( @xmath392 ) , the scatter around the cmr for the high mass r population tapers down to become small : @xmath393 mag in either @xmath41 or @xmath44 .
this behaviour can be understood in terms of mergers among ` red and dead ' galaxies ( see * ? ? ?
* ) : by the central limit theorem , the mixing of stellar populations in the individual merger products leads to convergence towards the mean colour for the population as a whole .
it is interesting that the flattening and focussing of the r cmr apparently begin at around @xmath394 , whereas major mergers are thought to be most prevalent at slightly higher masses ( @xmath395 ) ; see _ e.g. _ * ? ? ?
5 . ) at intermediate - to - low masses ( @xmath396 ) , the scatter in the @xmath44 cmr for the r - population is relatively small and relatively uniform .
( note that this qualitative statement is at worst weakly dependent on how well we understand the observational errors on the values of @xmath44 for individual galaxies . )
coupled with the apparent dearth of genuinely ` red and dead ' galaxies with @xmath396 , this implies that the ongoing evolution of r - type galaxies must proceed in such a way as to create or preserve the relation between stellar mass and stellar population .
this would seem to go against the idea that ( at least at these masses and redshifts , and outside of the richest clusters ) galaxies move on to the ` dead sequence ' rapidly and stochastically . 6 . ) in the @xmath44 cmd , the cmr for the b population ( _ i.e. _ , the blue sequence ) is relatively shallow and very nearly linear for @xmath371 .
this implies that the slope seen in the @xmath41 cmd reflects greater dust attenuation in higher mass galaxies , as expected from fig .
[ fig : colours ] .
further , the relatively shallow slope of the @xmath44 cmr implies an approximate self - similarity in the stellar populations of b - type galaxies with @xmath397 .
the relatively small and constant scatter around the relation in this mass range strengthens this idea .
together , these results suggest , albeit weakly , that the process of star formation or at least stellar assembly proceeds in a roughly self - similar fashion among the moderate - to - low mass b population . 7 . )
when we fit to the @xmath41 cmd , the cmrs for the b and the r population appear to converge for @xmath356 .
this mass range coincides with where there may be a slight upturn to the b cmr in the @xmath44 cmd , and with the knee in the overall field galaxy mass function .
while it must be stressed that this is where the statistics for b - type galaxies becomes poor , this hints that these very massive b - type galaxies may be qualitatively different from their lower mass cousins , in that they have rather redder stellar populations , while still having significant amounts of dust .
one possible interpretation is that these b - type galaxies with @xmath398 are well on their way to joining the r population .
an alternative is that some recent event ( _ e.g. _ , a merger event ) has briefly rejuvenated the stellar populations and dust content of these massive galaxies , perturbing them out of the main r population .
either way , the implication is that the apparent self - similarity among b - type galaxies breaks down at the highest masses .
we have now derived phenomenological descriptions of the b and r populations (; fig.s [ fig : rfcmr][fig : starmf ] ) , and have used this information to develop an objective , quantitative b / r classification scheme (; fig .
[ fig : rffrac ] and fig .
[ fig : starfrac ] ) .
our final task is to compare our results to the existing results introduced in .
our discussion is based on * fig .
[ fig : punchline]*. the basic point that we are trying to elucidate with this figure is how our b / r - type classifications compare to the hard blue / red selection cuts employed or advocated by @xcite , @xcite , and @xcite , and how these different operational definitions for ` blue ' and ` red ' lead to very different quantitative results and qualitative conclusions .
let us begin with a general description of what is shown in the different panels .
as in fig .
[ fig : others ] , the top panels reproduce the colour - magnitude and colour - mass diagrams used by these authors to define their blue and red samples . in these panels ,
the filled coloured contours show how the fraction of r - type galaxies ( as inferred from our fits to the @xmath44 cmd ) varies across the different colour - magnitude and colour - mass diagrams .
these values have been obtained by 1/@xmath103-weighted averaging of the values of @xmath399that is , the objective b / r - type classifier defined and discussed in in bins of colour and magnitude / mass .
in essence , if we were to observe many galaxies with the same or similar colour and magnitude / mass , the contours thus show what fraction of these galaxies we would expect to be members of the r population .
each of these cuts falls slightly blueward of the point where the r - type fraction is 50% , which is shown by the heavy green line .
unsurprisingly , there is a rather broad range of colours where there is substantial overlap between the b- and r - populations ( @xmath400 ) in each of these diagrams .
as we alluded in , this means that the different ` blue ' and ` red ' samples selected / defined by these different hard cuts will comprise a mix of both r- and b - type galaxies , in different proportions . as a means of selecting r- and b - type galaxies , these red and blue samples
will all be both incomplete and contaminated .
further , the relative proportions of r and b galaxies in each sample_i.e . _ , the degree of completeness / contamination will be a strong function of mass , and will be sensitive to the precise cut used . this is quantified in the middle row of fig . [
fig : punchline ] . we can define a kind of quasi - completeness , @xmath401 , for each of the red selections by looking at the relative numbers of r - type galaxies that satisfy each of the different red selections .
this is simply derived as @xmath402 , where the sum is over all galaxies satisfying the red selection , and similarly for @xmath403 .
( note that here , we are using the subscripts ` blue ' and ` red ' to denote the hard - cut samples , as distinct from our overlapping ` b ' and ` r ' populations . )
similarly , we can define a quasi - reliability , @xmath404 , for each red selection as @xmath405 , and similarly for @xmath406 .
the completeness of each of the three different red and blue selections , so defined , are shown as a function of mass by the solid lines in the middle panels .
we also show the degree of contamination , ( @xmath407 ) , as the dashed lines .
this same information is re - presented in a different form in the lower panels of fig .
[ fig : others ] , which also serves to illustrate and elucidate how these effects lead to qualitatively and quantitatively different determinations of the red and blue , or r- and b - type , mfs . in each panel , the histograms with errors show our determinations of the r- and b - type mfs , @xmath218 and @xmath217 , reproduced from fig .
[ fig : starmf ] ; these are the same in each panel .
then , the solid lines show the fit mfs , @xmath408 and @xmath409 , from each of @xcite , @xcite , and @xcite .
note that for the purposes of this figure , we have renormalised the literature mfs to match the integrated number density for the gama sample , in order to focus on differences in the _ shapes _ of the b / r and the blue / red mfs . in each panel of fig .
[ fig : others ] , the points show the absolute numbers of b- or r - type galaxies that are ` correctly ' selected as being blue or red ; in other words , these are the galaxies that we can all agree on
. these values are equivalent to @xmath410 and @xmath411 .
the difference between the points and the histograms thus reflects the effect of ` incompleteness ' in the blue / red samples selected using a hard cut .
similarly , the difference between the points and the curves show the impact of ` contamination ' in the hard - cut samples .
having now described the content of fig .
[ fig : punchline ] in general terms , let us now turn to discussing the results of each of these three works in the context of our more sophisticated analysis
. looking at the left panels of fig .
[ fig : others ] , it can be seen that our b mf agrees well with the blue mf from @xcite for @xmath412 . but
this agreement is at least partly coincidental . at these masses , compared to our b - classifications , the @xcite blue cuts are 8090% complete , and 8090% reliable .
the red cuts , on the other hand , are only 50 - 60% complete , and only 4060% reliable .
it turns out that these two effects offset one another : the extra 10 - 20% of b - type galaxies that are counted as red almost exactly balances the 50% of r - type galaxies that are missed . in the context of our results , @xcite thus get the ` right ' answer for the @xmath413 red / blue mfs , but not necessarily for the ` right ' reasons : even where our mfs are similar , they are counting very different galaxies . at higher masses , there are large discrepancies between our mfs and those of @xcite .
many of what we call ` b - type ' galaxies are counted by @xcite as being ` red ' : the quasi - completeness of the @xcite selection drops rather smoothly from @xmath5 75% at @xmath414 to @xmath415% for @xmath416 .
this means that , even though the red selection is @xmath417% reliable , it is also contaminated by b - type galaxies at the 2030% level .
the obvious culprit here is dust , and specifically the dust - induced upturn in the @xmath41 cmr for the b population , as can be seen in fig .
[ fig : rfcmr ] and fig .
[ fig : rffrac ] .
the net effect of these effects is to inflate the red mf by a factor of @xmath418 , and to depress the blue mf by a factor of @xmath419 . in light of all this , the fact that the @xcite red / blue mfs can each be well described by a single schechter function is somewhat coincidental .
the @xcite determinations of the red and blue mfs are based on similar assumptions to the ones that we have made .
these authors have used an _ ad hoc _ iterative procedure to fit simultaneously for the centres of and scatters around the blue / red cmrs , as well as the mfs . compared to our analysis ,
the major differences are : 1 . )
that their fits are to the @xmath420@xmath421 colour magnitude diagram , rather than the @xmath44@xmath73 cmd ; 2 . )
that in their analysis , they bin the data first by magnitude , and then by colour , and fit to these binned distributions ; and 3 . )
that they then rescale their values for each magnitude bin to stellar mass , using a simple relation between @xmath420 and @xmath422 , whereas we are explicitly working with @xmath50 estimates from sps modelling of optical seds .
given the general similarities in , and important differences between , the two approaches , it is extremely encouraging that there is such good agreement between the b and r mfs that we derive , and those from @xcite at least at the high - mass end .
there are still rather large differences between the red mf from @xcite and the r mf that we derive from the @xmath44 cmd , but we note that the results based on the @xmath41 cmd are in rather better agreement . the cut shown in fig .
[ fig : punchline ] has been derived by @xcite on the basis of their fits .
this cut is designed to maximise the product ( @xmath423 ) in @xmath420@xmath421 space . in this sense
, it is designed to be an optimal hard - cut blue / red selection line ( given the specific tanh parameterisation used ) . from fig .
[ fig : punchline ] , it can be seen that this optimal selection is nevertheless imperfect . using the @xcite selection line , @xmath401 drops from @xmath424% completeness at @xmath425 to @xmath426% for @xmath356 ; by the same token , @xmath406 is @xmath427% for all masses .
( these numbers are entirely consistent with the caveats given by * ? ? ?
this again serves to unambiguously demonstrate the difficultly of using a hard cut in an optical colour magnitude or colour mass diagram to meaningfully select red / blue galaxy samples .
finally , let us turn our attention to the comparison between our results and those of @xcite .
this comparison is particularly interesting and important , given the elegant ` semi - empirical ' model for the quenching of star formation within galaxies that these authors have advanced based on their results .
the @xcite model predicts a single schechter function for star - forming galaxies , and a two - component schechter function for quenched / passive galaxies .
the model also makes the specific prediction that the secondary component of the mf for quenched galaxies should have the same shape as , but a lower normalisation than , the mf for star - forming galaxies .
they show that their observed blue / red mfs , based on data from sdss and zcosmos , and selected using the cut in the @xmath178-@xmath428 colour - magnitude diagram , can be understood in this way . from fig .
[ fig : punchline ] , it is clear that the @xcite cut is rather bluer than others , such that their blue sample is only @xmath424% complete for b - type galaxies , even at the lowest masses .
in other words , fully @xmath429% of what we classify as ` b - type ' galaxies would be counted by @xcite as ` red ' . by the same token ,
their red sample is heavily contaminated : more than half of those galaxies selected as red / quenched by @xcite are classified by us as being b - type .
this is absolutely crucial in the context of the @xcite model , which makes the specific prediction that the secondary component of the mf for quenched galaxies should have the same shape as , but a lower normalisation than , the mf for star - forming galaxies .
it should now be clear how exactly this kind of behaviour can be produced by using too - blue a cut : once the red - selected sample becomes highly contaminated , it becomes simply a shadow of the blue mf .
this effect does not rely on , but will be amplified by , the effects of photometric scatter discussed in .
we thus highlight the possibility that the low mass upturn in the red mf seen by @xcite and others simply reflects a high degree of ` contamination ' of the ` red ' sample by members of the ` blue ' population .
this happens where the relative numbers of red galaxies is low , and the hard red / blue selection limit enters the red wings of the blue distribution .
this interpretation also explains why the red upturn has a similar shape to the blue mf : because the galaxies responsible for this upturn _ are _ actually members of the blue population . taken at face value , our results would therefore appear to be in conflict with the quenching model advanced by @xcite .
however , this statement is only true to the extent that our r - type galaxies can be interpreted , in astrophysical terms , as being genuinely ` quenched ' .
but still , looking at fig . [ fig : punchline ] and back to fig .
[ fig : stardist ] , the same criticism could be levelled at @xcite . in light of the above ,
the strongest point that we are prepared to advocate at this stage is only that the relative shapes and absolute values of the b and r ( or blue and red ) mfs depends very sensitively on how the different populations are defined / selected / classified ( see also , _
e.g. _ * ? ? ?
* ; * ? ? ?
* among others ) .
further , we stress the dangers of over - interpretting the physical significance of any hard red / blue selection , given the basic fact of overlap between the two r and b populations .
this is precisely why we have set out to derive objective and phenomenological b / r classifications , which are empirically derived from our statistical description of the observed data .
the basic problem that we have sought to overcome in our analysis is that the _ optical _ colour distributions of the b and r populations overlap . we are certainly not the first to apprehend the difficulties in using optical colours to distinguish ` developed ' from ` developing ' galaxies .
our solution to this problem has been to devise a mixture modelling approach to account for this fact .
most authors , however , have sought to circumvent these problems , rather than attempt to confront them head on .
many others have argued that an optical nir colour colour diagram ( see _ e.g. _ , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) can be used to isolate ` quiescent ' galaxies .
the idea is that the dust and age / metallicity vectors are no longer parallel in a such a diagram , and so ` red and dead ' galaxies and ` dusty star - formers ' are more easily separated .
this kind of colour selection thus acts as a very simple ( two - colour ) sed - based classification .
taking this idea to its logical conclusion , other authors have explicitly used sps modelling of seds to develop a young / old classification scheme . for example , @xcite have selected ` passive ' galaxies based on the best - fit sps template .
after the fact , they then show that these ` passive ' galaxies are indeed concentrated in the expected region of the nuv@xmath19@xmath38 colour colour diagram . by considering the intrinsic stellar @xmath44 color ,
we have used the information encoded in the multicolour seds to , insofar as is possible , break the degeneracy between dust and age / metallicity . in this way , we are using @xmath44 as a continuous and quantitative diagnostic of a galaxy s stellar population .
then , we have used a phenomenological , descriptive model of the bivariate @xmath50@xmath44 distribution to derive a quantitative , probabilistic classification scheme .
next , similar to @xcite , we can elucidate the nature of and validate our interpretation of
these classifications by showing how our classifications map onto other common diagnostic plots .
we explicitly show how our b / r - classification scheme maps onto the _ uvj _ diagram in * fig .
[ fig : nircols ] * , which shows the galaxies in our sample , separated into broad mass bins , and colour - coded according to their objective b / r - type classification , @xmath359 .
these plots can be compared to the restframe _ uvj _ colour colour diagram used by , for example , @xcite to and isolate the ` dead sequence ' of truly passive galaxies at high redshifts . for this figure ,
the restframe _ uvj _ photometry have been derived using a heavily refactored version of the interrest algorithm for interpolating restframe photometry @xcite .
it is worth stressing that the derivation of these restframe colours is completely independent from the sed fits used to derive the values of @xmath50 , @xmath41 , @xmath44 , etc .
it can be seen that , in broad terms , our @xmath44-based separation between b- and r - type galaxies behaves as expected in the _ uvj _ colour colour diagram : the r - type galaxies are tightly clustered , and in a location that is above and/or to the left of the ` blue sequence ' .
this demonstrates that our optical sed fits can distinguish between the seds of ` red and dead ' galaxies and ` dusty star formers ' , and thus that our values of @xmath44 do provide a meaningful characterisation of galaxies stellar populations .
the different panels fig .
[ fig : nircols ] show how our b / r - type classifications is projected onto the _ uvj _ diagram at different mass ranges .
looking at the classifications , the point of transition from mostly b - type ( blue points ) to mostly r - type ( red points ) can be seen to correspond qualitative changes in the 2d data density that , in simple terms , look like different populations . it can be seen in this figure that there is some mass - dependent ` creep ' to how our b / r classification maps onto the _ uvj _ diagram . from a phenomenological point of view
, however , the argument could be made that this is desirable .
certainly , it is clear that the locus of ` red ' galaxies , however defined , shifts to redder colours at higher masses .
this is just the colour mass relation in another guise .
indeed , in the lower mass bins our b / r - type classifications might provide a better characterisation of the two - population nature of the data than the standard hard selection box shown .
it is worth explicitly noting that the _ uvj _ selection box shown would capture predominately b - type galaxies at the lowest masses : the ` reliability ' of a _ uvj _ selected sample of r - type galaxies would be only @xmath430 % . in comparison to our determination of the r mf , such a _ uvj_-selected ` red ' mf would be inflated .
moreover , this would act in such a way that the inferred ` red ' mf would have approximately the same low mass slope as the ` blue ' one .
this is highly significant in the context of the results of @xcite . *
[ fig : hdd4 ] * shows a similar but different test , this time using the @xmath431 and h@xmath432 spectral diagnostics .
these measurements have been made using the available gama or literature spectra for each galaxy in our sample .
for the purposes of this plot , we have imposed a signal - to - noise cut across the relevant wavelength range to ensure reasonable data quality .
this effectively introduces a bias against the reddest galaxies , particularly at low masses .
representative error bars for each bin are shown . in this diagram ,
stars with ages @xmath433 gyr will have both @xmath434 and h@xmath435 ( see , _ e.g. _ * ? ? ?
* among many others ) . because these measurements are made over a narrow wavelength range , they are very weakly sensitive to dust .
this diagnostic diagram shows that the optical spectra of r type galaxies are completely dominated by old stars , with little to no contribution from stars less than @xmath436 gyr old . in other words , and modulo the caveat in the previous paragraph , these results suggest that those galaxies that we have classified as r - type have seen no significant star formation activity in the past 1 gyr or more .
the results presented in this section are intended to demonstrate two important facts .
first , the results shown in each of these two figures illustrate that and how our b - r classifications faithfully and meaningfully encapsulate differences in galaxies sed shapes , and hence stellar populations . in particular , these figures show how , particularly for @xmath132 , the r - type population is analogous to commonly used ` red and dead ' , ` passive ' , or ` quenched ' selection criteria .
second , and more perhaps more importantly , we have presented these diagnostic diagrams as a means to evaluate the meaning of our _ phenomenological and empirically derived _
b / r classification scheme . earlier , in
, we have said that our results can be used to gain insight into the process of quenching , but only to the extent that our operational definitions of ` red ' or ` r'-ness can be taken to mean ` quenched ' . in this regard , we can now offer a more astrophysical characterisation of what distinguishes r- and b - type galaxies : namely , that the light from r - type galaxies is completely dominated by stars with ages @xmath433 gyr . in this work ,
we have used the intrinsic , dust - corrected @xmath44 colour as a diagnostic parameter for a galaxy s stellar population . as mentioned in ,
this quantity is a very good proxy for luminosity - weighted mean stellar age .
we have then used our parametric , descriptive mixture - modelling of the @xmath44 cmd to construct an objective b / r - type classification scheme , and so derive operational definitions for these designations . in this section ,
we have used complementary diagnostics to show that the stellar populations of b- and r - type galaxies do indeed differ , and in particular that r - type galaxies really do host older stars .
this demonstrates that the @xmath44 parameter is indeed a useful diagnostic for distinguishing galaxies based on their stellar populations . the line of argument that we have now developed is similar to the one in @xcite .
in particular , we have now explicitly demonstrated , in a mass dependent way , that galaxies classified as r - type or b - type have different spectral and sed shapes , and thus that the two populations have qualitatively different stellar populations .
( we also show how the kind of _ uvj _ cuts commonly used can give very misleading results at low masses . )
the crucial point of difference between our analysis and those mentioned above is this : we have not explicitly set out to distinguish ` young ' and ` old ' ( or ` star forming ' and ` quenched ' ) stellar populations .
instead , we have started from a phenomenological description of the optical colour mass diagram , and used this to disentangle the two apparently distinct populations . identifying the distinguishing features of galaxies in these two populations is thus , for us , a secondary problem . in this sense , the _ implication _ from fig . [ fig : nircols ] and fig .
[ fig : hdd4 ] is that massive r - type galaxies have relatively old stellar populations , and little to no dust .
this point is non - trivial , even if it may seem so at first blush . in
, we have described the various tests that we have done in an attempt to devise the best and simplest description of the data possible .
this is not to say , however , that our parameterisation is right . instead , our claim is only that we do provide a good description of the data , inasmuch as our model encapsulates all the qualitative features of the bivariate colour mass distributions , as illustrated in fig.s [ fig : rfdist][fig : starmf ] .
there are some indications that some parameters could be eliminated from the model without significantly degrading the statistical quality of the fit : most notably , the data do not clearly demand a two - schechter description of the r mf . while there is some ambiguity in
whether or not certain parameters ought be excluded or included , we can say with confidence that we are not grossly overfitting the data .
there are some parameters for which the inferred values are consistent with zero ( see fig . [
fig : mcmc ] ) . to the extent
that the model does not make use of the additional freedoms that these parameters allow , they would appear unnecessary for the best , simplest possible description of the data . at the same time , and for the same reason , excluding these parameters would make no difference to the fits , or to our results . to the extent that the decision as to whether or not to include these parameters is arbitrary , it is also unimportant .
as discussed in , this same criticism could be levelled at any model or modeller .
the decision as to how to model one s results is an inescapable part of any modelling .
furthermore , in the specific context of modelling the bimodality in the cmd , this concern is misplaced . as discussed in and , the use of a hard cut to select ` red sequence ' and
` blue cloud ' galaxies directly determines the inferred shapes of the mass functions . and
as discussed in , in the absence of some solid theoretical justification for the specific cuts used , these cuts must be viewed as being to some extent arbitrary .
this criticism is thus much more pertinent for the vast majority of existing bimodality studies ( the notable exception being * ? ? ?
indeed , as laid out in , the motivation for our analysis is precisely to address this concern , insofar as is possible .
our primary motivation for our mixture model of the bivariate colour - mass distributions is to develop an objective classification scheme , which can be used to disentangle the apparently distinct populations .
but this is not to deny the truth that the form of the model influences the outcome . indeed , as discussed in , this idea leads to an important caveat on our mf determinations for @xmath345 , where the application of our model to the observed colour distributions becomes problematic ( see fig . [
fig : rfdist ] and fig .
[ fig : stardist ] ) . by using a bayesian mcmc sampling scheme to constrain the values of the parameters in our model_i.e .
_ , to do the actual fitting we fully determine and account for the covariances between the model parameters .
further , as discussed in , we are not specifically interested in the actual values of most of the parameters that define the model .
for example , the values of the parameters @xmath438 and @xmath439 are important only insofar as they describe the shape of the mf for the b population . as shown in fig .
[ fig : rfmf ] and fig .
[ fig : starmf ] , we do not need to be able to uniquely determine the values of the @xmath52s or @xmath241s in order to obtain very precise determinations of the mf . as discussed in , ` the results ' of our analysis should be taken to be the histograms shown in fig . [
fig : rfmf ] and fig .
[ fig : starmf ] . in deriving these results ,
the role of our parameterised model is only to provide objective and quantitative b- and r - type classifications , as discussed in , which account for the overlap between the two populations in the cmds .
as discussed in , uncertainties in the values of the different model parameters , propagated through to uncertainties in the classifications , produce uncertainties in our mfs that are at most @xmath370 % , and are more typically 15% .
for any given mass in either fig .
[ fig : rfmf ] or fig .
[ fig : starmf ] , the contribution to the total error budget associated with the construction of the model is always @xmath440% , and more typically @xmath338% . as discussed in ,
our analysis is predicated on two assumptions , neither of which are controversial , first , we are assuming that there are two populations , which are characterised / distinguished by their own distinct cmrs .
second , we are assuming that there is some scatter around these relations , to the point where the two populations are observed to overlap .
taken together , these two assumptions lead to the situation where , in principle and in practice , the bluest r - type galaxies may have bluer stellar colours than the reddest b - type galaxies .
this is precisely the reason why we choose to refer to the populations using the more neutral designations ` b ' and ` r ' : to try to avoid some of the confusion that comes from the strong connotations that have come to be associated with these words .
some care is therefore required in interpreting our results in astrophysical terms ( see the caveats given in ) .
furthermore , we stress that previous results that have used a hard cut overlook the empirical fact of scatter in the colour distributions for the ` blue ' and ` red ' populations .
some care is also required in interpreting the results of past bimodality studies in astrophysical terms .
we would therefore invert this criticism to argue that past studies have selected the red tail of the b population galaxies that have young , blue stellar populations and called these galaxies ` quenched ' ( see fig .
[ fig : punchline ] and in particular fig .
[ fig : nircols ] ) .
again , we make no pretence that our particular parameterisation is in any way physically meaningful ; only that it yields a good description of the phenomenology of the cmd .
our results thus offer one potential means of understanding the data , which is the most that any modeller can do . included within this
is a phenomenological description of how the general population can be decomposed into two distinct but overlapping populations .
our hope here is that our empirical results can be used to guide and inform the future development of genuine semi - analytical and sph models of galaxy formation in a cosmological context .
what we have done in this work is to develop a means for objectively and quantitatively classifying galaxies according to their stellar populations . in the broader context of this series of papers ,
our ultimate goal is to use these classifications to identify the physical differences between b- and r - type galaxies , with a view to deriving empirical constraints on the physical processes that act to determine whether any particular galaxy is b- or r - type .
we therefore defer more astrophysically - minded observational studies of the galaxy bimodality ( or bimodalities ? ) to future works . in the first instance , our goal has been to separate the general galaxy population into two subpopulations , on the basis of their constituent stellar populations , and in a mass dependent way . for our analysis of the effective @xmath41 cmd ,
dust is a confounding factor .
for this reason , we have repeated our analysis looking at intrinsic , dust corrected @xmath44 colours .
it is true that degeneracies between dust , metallicity , and sfh in such fits mean that the uncertainties in the values of @xmath441 , @xmath37 , or @xmath442 can be large , but we maintain that these data are sufficient to make the qualitative distinction between dusty and old sed shapes . in support of this claim ,
we have shown in that those galaxies that we classify as being either b- or r - type occupy different regions of the h@xmath432-d@xmath443 spectral line diagnostic diagram ( fig . [
fig : hdd4 ] ) , as well as the restframe _ uvj _ colour - colour diagram ( fig .
[ fig : nircols ] ) .
in other words , the b-/r - type classifications really do encapsulate meaningful information about galaxies stellar populations .
further , these classifications closely correspond to the two - population distributions seen in both diagnostic plots .
conversely , we have also argued the kinds of ` red ' selections that are commonly used are a poor proxy for ` quenched ' .
we have thus accomplished our primary goal of disentangling and characterising the two apparently distinct populations seen in the cmd . _
some _ of our r - type galaxies conform to the usual picture of ` red and dead ' galaxies , but we have shown that these galaxies are just the high mass tip of a more continuous r population . with our primary goal accomplished , our phenomenological characterisations of these two populations can then shed light on their astrophysical nature .
this includes the mf for each population , which now represents a target for cosmological simulations to aim at .
this also includes our characterisations of the @xmath44 and @xmath41 cmrs for each population , which provide qualitative constraints on the processes of star formation and of star formation quenching within galaxies .
the process of galaxy formation is complicated , and there are myriad well - known and studied galaxy types and classes that we have not considered in our analysis : _ e.g. _ , radio loud agn , starbursts , e+a or post - starburst galaxies , satellites / centrals , interacting and merging galaxies , etc . some or all of these processes may play an important role in determining whether a particular galaxy is r - type or b - type , or conversely , some of these processes may act exclusively on or within r- or b - type galaxies . in this way , these processes are presumably also responsible for producing the observed intrinsic scatter around each of the two cmrs . by the same token , galaxies are in general multi - component systems . as a simple example , most massive galaxies have both a disk and a bulge component , with separate stellar populations and formation mechanisms .
for example , @xcite have argued that galaxies _ global _ colours are driven by the mixture of ( blue ) disk stars and ( red ) bulge stars , and that the colour bimodality is thus best understood in terms of the ( bimodal ) distribution of bulge - to - disk mass ratios .
we have deliberately avoided these kinds of questions in this work ; our intention is only to derive an empirical , phenomenological description of the apparent dichotomy in the stellar populations of field galaxies that is , the nature of the r- and b galaxy populations . in doing so
, we have derived an objective , operational definition for these designations .
this is a necessary prerequisite for future studies of the different astrophysical natures of these two galaxy populations , which we will pursue in future papers . in light of the above ,
it is remarkable that this game can be played at all . here
again , we stress there are no clear signs of an intermediate or transition ` green ' population .
that the bivariate colour mass distribution for the field galaxy population is extremely well described by a simple two - population model implies that our phenomenological distinction between b- and r - type galaxies does indeed encapsulate some important astrophysical differences in the formation histories or evolutionary states of these two populations even if we can not yet articulate what these differences are , or what the driver for these differences may be .
our particular interest lies in characterising the mass functions ( mfs ) and colour - mass relations ( cmrs ) for the apparently distinct populations seen in the optical colour - mass diagram ( cmd ) , where the two populations are distinguished on the basis of galaxies stellar populations .
our analysis is based on a sample of @xmath1 and @xmath0 field galaxies from the galaxy and mass assembly ( gama ) survey .
this sample is properly mass - complete ( volume limited ) for @xmath132 ; for lower masses , we have used the standard @xmath102 formalism to account for incompleteness ( see ) .
note that none of our results or conclusions change if we limit our analysis to @xmath173 , or to @xmath116 . the immediate motivation for reconsidering this well - studied problem is that , as discussed in , there are quantitative and qualitative disagreements between the mf determinations that exist in the literature .
first , we have shown that if we analyse our gama - sample in the same way as each of @xcite , @xcite , and @xcite , we are able to reproduce each of these authors sdss - based results
. then , we argue that the discrepancies between the results and conclusions of these studies are due entirely to the different ( and most often arbitrary ) ways that the ` blue ' and ` red ' galaxy samples have been selected / defined .
our first and most important conclusion is therefore a qualitative one : that the largest uncertainty in previous characterisations of the mass functions for ` red ' and ` blue ' galaxies is tied to how these terms are defined .
put bluntly , the main reason why , say , @xcite see an upturn at the low end of the ` red ' mf where , say , @xcite do not is simply because the @xcite definition of ` red ' is considerably bluer than the @xcite one . as a direct consequence , a significant fraction of the @xcite ` red ' sample
are star forming galaxies with young stellar populations ( see fig.s [ fig : otherdust ] , [ fig : punchline ] , and [ fig : nircols ] ) .
the direct implication is that the power of these results to provide useful constraints on the process of galaxy evolution is directly limited by the extent to which the specific ` red ' and ` blue ' selections used can be shown to be astrophysically meaningful . in the absence of convincing arguments in favour of any one of these selections over the others , important questions like the mass range over which the galaxy population transitions from mostly blue to mostly red , or the low mass slopes of the blue and red mass functions , are left largely unconstrained . in order to address these questions ,
what is needed is a well - motivated operational definition for the technical terms ` red ' and ` blue ' . to redress this ,
we have developed a descriptive model for the distribution of galaxies in cmd , with the specific goal of distinguishing between ` developed ' and ` developing ' galaxies on the basis of their stellar populations . in our modelling , we treat the observed data distribution as being the sum of two distinct but overlapping populations .
the model also includes a ` bad ' component to allow for outliers , catastrophic errors , or otherwise un- or under - modelled aspects of the observed distributions ( see and ) .
the formalism for our modelling , which is developed pedagogically in appendix [ app : themodel ] , is based on the method of ( gaussian ) mixture modelling ( see , _
e.g. _ , * ? ? ?
as outlined in , in its most general form , our descriptive model is fully defined by 40 parameters . as discussed in ,
not all of these parameters are strictly necessary for a ` good ' description of the data .
in particular , it is not clear that the data strictly demand a second schechter component to well describe the mf for the r population . as discussed in , however , where there is ambiguity about whether or not a parameter is necessary , it will have little to no impact on the final results . beyond that , the most that we can say is that we have made every attempt to ensure that we are not grossly overfitting our data . in effect
, there are two assumptions that underpin our approach .
first , we are assuming that some physical process(es ) or hidden parameter(s ) act to determine whether a given galaxy is a member of one or the other population ; that is , _ we assume that there are two populations , which follow distinct cmrs_. then , some secondary process(es ) or parameter(s ) determines where that galaxy falls with respect to the main cmr for that population ; that is , _ we assume that there is some ( gaussian ) scatter around each of the two cmrs _ , to the point that the two populations are observed to overlap in an optical cmd ( or , indeed , in a _ uvj _
colour colour diagram ) .
neither of these assumptions ought to be controversial . allowing that , at fixed mass
, there is some overlap between the ` blue ' and ` red ' colour distributions inescapably implies that some ` red ' galaxies will have bluer colours than some ` blue ' galaxies . in acknowledgement of this semantic trap , and to avoid some of the confusing connotations associated with the terms ` blue ' and ` red ' , we have adopted the more generic designations of ` b ' and ` r ' to describe the two populations seen in the cmd ( see and ) .
that is , rather than considering blue and red _ galaxies _ , we focus on two galaxy _ populations _ , which we dub b and r. it is not necessarily true that a particular b- galaxy will be bluer than some other r - galaxy .
it is also not necessarily true that an r - type galaxy can be thought of as being ` red and dead ' , ` early type ' , ` quiescent ' , etc .
it nevertheless remains true that the distribution of stellar population properties are different for galaxies in each of the two populations in particular , at fixed mass , and on average , galaxies in the b - population have bluer ( and so younger ) stellar populations than those in the r - population .
in other words , we are assuming that the phenomenological separation of the general galaxy population into b- and r - type subpopulations somehow reflects a qualitative binarity in the formation histories or evolutionary states of galaxies but it remains to describe and explain precisely how and why this is the case . by modelling these populations
, we can actually _ derive from the data _ phenomenological working definitions for the terms ` b- ' and ` r - type ' ( see , fig .
[ fig : rffrac ] , and fig .
[ fig : starfrac ] ) .
what is more , the classifications are _ objective _ , insofar as objectivity is possible ( see ) .
bearing in mind the caveats given above , we go on to describe the basic characteristics of the b- and r - population in terms of both the effective @xmath41 cmd ( fig . [ fig : rfcmr ] ) , as well as the intrinsic , dust - corrected colour , @xmath44 cmd ( fig .
[ fig : starcmr ] ) . that the members of these two populations comprise
genuinely different stellar populations has been demonstrated in , where we show that r- and b - type galaxies occupy distinct regions of the _ uvj _ and @xmath431h@xmath432 diagnostic diagrams ( fig .
[ fig : nircols ] and fig .
[ fig : hdd4 ] ) .
we find that the intrinsic @xmath44 cmr for the b population is both considerably tighter and more linear than the effective @xmath41 cmr .
this implies that the both the upturn in the @xmath41 cmr at @xmath361 and the relatively large scatter around the @xmath41 cmr at all masses is driven by the distribution of dust properties , rather than differences in the stellar populations of these galaxies . at least for @xmath444 ,
the relative flatness of the @xmath44 cmr for b - type galaxies also implies that , in terms of stellar colours , the b population is relatively homogenous
. this behaviour changes , however , for @xmath356 . at this mass range , the b population appears to converge with the red sequence in the @xmath41 cmd .
this is also where the mf for b galaxies drops off very rapidly . in other words , there are few if any galaxies with @xmath356 with young ( b - type ) stellar populations ; this represents the top end of the blue sequence . for the r population ,
if nothing else , we can say with confidence that there are essentially no @xmath128 field galaxies with stellar populations that are the same as or similar to those of the genuinely ` red and dead ' galaxies seen at the highest masses . instead , at least for @xmath445 , we see a gradual trend whereby less massive galaxies have progressively bluer stellar colours ( _ i.e. _ , younger luminosity - weighted mean stellar ages and/or lower mean stellar metallicities ) than their more massive cousins .
the relatively small dispersion in @xmath44 cmr for the r - population suggests that , at fixed mass , r - type galaxies have rather similar stellar populations , and hence similar star formation / stellar assembly histories . at the same time
, the slope of the @xmath44 cmr for r - type galaxies shows that there are differences in the evolutionary histories of different members of the r populations with different present day masses .
said another way , the evolution of individual r - type galaxies in the field proceeds in such a way as to create or preserve the correlation between the stellar mass on the one hand , and stellar population on the other .
these observations the relative homogeneity of the stellar colours of b - type galaxies , and the correlation between stellar mass and stellar population for r - type galaxies run counter to prevalent notions of the ` red sequence ' and the ` blue cloud ' .
they also provide meaningful targets for theoretical models of galaxy formation and evolution to aim for . for @xmath121 ,
it becomes increasingly difficult to meaningfully and robustly distinguish two separate b and r populations from the joint @xmath41 and @xmath44 colour - mass distributions .
indeed , looking _ only _ at the observed colour distributions in bins of mass ( fig .
[ fig : rfdist ] and fig .
[ fig : stardist ] ) , there is little to no clear evidence for a distinct r population at @xmath128 .
our claim is therefore that , below @xmath446 , we are no longer able to unambiguously identify a distinct r - component to the general field population : at least for field galaxies , below @xmath3 , the r population dissolves into obscurity . at these low masses ,
our characterisation of the r population may be better interpreted as describing the degree of asymmetry in the observed colour distributions , and our mf for the r population may be taken as an _ upper limit _ on the number of galaxies that lie outside the main colour distribution for ` normal ' b - type galaxies ( see ) .
we note in passing that the data do not clearly demand an intermediate ` green ' population : the data are very well described by a two - population model ( see fig .
[ fig : rfdist ] and fig .
[ fig : stardist ] ) . taking this remarkable observation at face value
, this might imply that movement between populations is quick ( see , _ e.g. _ , * ? ? ?
alternatively , it could imply that movement _ within _ a population ( for example , by satellite accretion / minor mergers ) is much faster than movement _ between _ populations : a galaxy may be ` passed ' from one population to the other , and then have its colour rapidly re - randomised according to the ` normal ' colour distribution for its new population .
we also note that optically - identified agn reside exclusively within the blue sequence defined by the b population ( fig .
[ fig : colours ] ) .
that is , in terms of their stellar populations , agn hosts do not clearly differ from ` ordinary ' star forming galaxies .
certainly they do not represent a transition population that is intermediate between the b and r populations .
the mfs that we derive for the r and b populations are rather different to those for ` red ' and ` blue ' galaxies presented by other authors ( see fig . [
fig : punchline ] ) .
the reasons for these differences are discussed first in , and then again in close detail in .
in particular , we find considerably more ` b ' galaxies with @xmath141 , and a much more abrupt drop - off in the b mf at @xmath356 . whereas others put the crossover mass , where the two mfs intersect , at @xmath44710.3 , we find it to be closer to 10.5 .
further , our mf for r - type galaxies is also considerably lower at low masses than those for ` red ' galaxies from the literature . _
if _ these red or r - type mfs can be used to probe the process of quenching , our results would imply that quenching is less prevalent or equivalently , that massive galaxies continue forming new stars for longer than has been previously thought .
but this leads to the final and most important caveat : as discussed at length in and , there is real danger in inappropriately reifying the terms ` b- ' and ` r - type'or , equally , the terms ` blue ' or ` red ' .
these terms must be understood as qualitative , phenomenological designations for galaxy _ populations_. in particular , one should not conflate the terms ` r ' or ` red ' with the term ` quenched ' .
this is especially true at low masses , where the stellar populations of ` red ' or r - type galaxies are rather different to those found at higher masses . as discussed throughout this paper , it is necessarily and inescapably true that our results depend on the choices made in the construction of our model we can only answer the question that we have asked . in particular , the decision to use gaussians to describe the colour distributions at fixed mass is primarily one of convenience .
our justification for these decisions is ultimately empirical
the proof is in the pudding .
we have thus shown that our descriptive modelling provides one potential way of understanding the data .
the fact nevertheless remains that using a different operational definition for the terms ` red ' and ` blue ' will lead to quantitatively and perhaps qualitatively different results .
we do not pretend to be immune to these difficulties . instead , our goal is to highlight the importance of these issues , and to make it clear that the same criticisms can and should be levelled at any study of the ` red ' and ` blue ' galaxy populations .
but then again , we have also validated our results by showing how the phenomenological b / r classifications that come from our model do indeed select galaxies with qualitatively different stellar populations .
this point is particularly significant in connection with the elegant semi - empirical model for quenching presented by @xcite , which hinges on the presence or absence of an upturn to the mf for quenched galaxies at @xmath121 .
looking at the distributions shown in fig .
[ fig : rfdist ] or fig .
[ fig : stardist ] , ( or even the _ uvj _ diagrams in fig .
[ fig : nircols ] ) it should be immediately obvious how problematic it is to meaningfully distinguish a separate ` red ' or ` quenched ' or even ` r'-type population at these low masses ( see ) .
further , it must be recognised how a sample of ` old ' or ` quenched ' galaxies selected using a hard cut in any of these diagrams including the _ uvj _ diagram will be dominated at low masses by spillover from the red tail of the distribution of otherwise normal , young , star - forming , b - type galaxies .
the critical issue is therefore whether or in what sense these galaxies can be thought of as a distinct , coherent population of quenched galaxies .
this paper is the first in a series in which we explore the different aspects or manifestations of the bimodality ( or bimodalities ) in galaxy properties . in papers
ii and iii , we will perform a similar analysis to explore bimodalities in terms of line emission properties , and in terms of srsic - fit structural parameters .
taken together , these three papers will provide a basis for robustly and objectively classifying galaxies according to the stellar populations , present day star formation and/or agn activity , and structure . in future works
, we will go on to use these results to explore the natures of , and interrelations between , these bimodalities .
it is our genuine pleasure to thank the referee , eric bell , for his thorough and thoughtful feedback , which greatly improved the clarity and strength of the arguments presented .
part of this work was made possible through use of the edward hpc cluster , maintained by information technology services research at the university of melbourne .
pn acknowledges the support of the royal society through the award of a university research fellowship and the european research council , through receipt of a starting grant ( degas-259586 ) .
mg is supported by a european research council starting grant ( degas-259586 ) .
gama is a joint european - australasian project based around a spectroscopic campaign using the anglo - australian telescope .
the gama input catalogue is based on data taken from the sloan digital sky survey and the ukirt infrared deep sky survey .
complementary imaging of the gama regions is being obtained by a number of independent survey programs including galex mis , vst kids , vista viking , wise , herschel - atlas , gmrt and askap providing uv to radio coverage .
gama is funded by the stfc ( uk ) , the arc ( australia ) , the aao , and the participating institutions .
the gama website is ` http://www.gama-survey.org/ ` .
our overarching goal is to derive an empirical , phenomenological description of the ` bimodality ' in the galaxy population , as seen in the @xmath41 and @xmath44 cmds .
this apparently simple project is made problematic by the fact that the apparently distinct ` red ' and ` blue ' populations are seen to overlap in both of these cmds .
accordingly , we want to avoid imposing some arbitrary ` hard ' cut to distinguish ` red ' from ` blue ' ; instead , we want to develop the means to simultaneously and flexibly describing these two distinct but overlapping populations .
the solution to this modelling problem the method of mixture modelling can alternatively be viewed as the development of an _ empirical _ and _ objective _ ` red'/`blue ' classification scheme , based on the likelihood that a given data point has been drawn from one or the other population . to reduce the problem to the simplest possible terms ,
what we want to do is to construct a parametric model that describes the distribution of data points in the cmds .
this model will need to have two components one for each of the ` red ' and ` blue ' populations each of the form @xmath448 , where @xmath449 and @xmath450 or @xmath44 .
the model itself will be fully described by a set of parameters , @xmath191 .
the observed data are then considered as having been randomly drawn from generated by this modelled density distribution , and observed subject to the appropriate measurement errors or uncertainties .
the generative model is thus explicitly intended to describe the ( scalar ) likelihood , @xmath201 , of observing any given ( vector ) datapoint , @xmath76 , and the associated uncertainties , which are assumed to be gaussian , and are for now generically represented as @xmath451 .
it is crucial to recognise that the generative model is only calculable indeed , is only defined given or assuming a particular set of trial values for each and every of the parameters in @xmath191 .
that is , @xmath201 describes the likelihood of observing the datapoint @xmath363 with formal observational uncertainties @xmath451 , given or assuming a specific set of values for @xmath191 . to reflect this fact ,
the likelihood function is represented as @xmath452 .
the primary goal is thus to use the full observed dataset , @xmath453 , along with the associated set of observational uncertainties , @xmath454 , to constrain the ` true ' values of the parameters in @xmath191that is , we aim to generate the posterior probability distribution function ( pdf ) for the values of the parameters @xmath191 , given that we have observed our data , @xmath455 .
let us begin our discussion by considering only one population , and taking the simplest possible relation between @xmath12 and @xmath64 : a perfect line , @xmath456 . in this case
, the model parameter set @xmath191 just comprises the slope and normalisation of the line ; _ i.e. _ , @xmath457 . the traditional but overly simple and somewhat nave approach to this kind of fitting problem is to use the method of weighted @xmath6 minimisation , which is described immediately below , to fit independently fit for a linear relation between @xmath64 and @xmath12 . among extragalactic astronomers ,
this approach is also frequently referred to as ` maximum likelihood ' .
however , as is well known and accepted in many other disciplines ( including particle physics and cosmology ) , and as we shall endeavour to make clear in the next few sections , the weighted @xmath6 fitting formalism is only one instance of the more general class of maximum likelihood fits that are possible . moreover , the very restrictive set of assumptions that underpin the weighted @xmath6 formalism means that it is only appropriate to use in very specific and , in astronomy , very rare situations . in any case , following the weighted @xmath6 minimisation formalism , including the assumption of gaussian errors , the values of the @xmath458s are computed as : @xmath459 ~ , \end{split}\ ] ] where we have abbreviated @xmath460 ) as @xmath461 , and @xmath462 is the observational uncertainty associated with the measurement of @xmath463 .
we have also used this equation to introduce @xmath464 as our shorthand for a 1d gaussian distribution , centred on @xmath465 and with variance @xmath466 , and evaluated at the location @xmath467 . note that in what follows , @xmath468 should always be understood to be integral normalised to unity . the global likelihood , @xmath17 , of observing the full dataset
is then given by the product of all the individual @xmath201s .
in practice , it is more convenient to work in terms of logarithms , so that : @xmath469 within the traditional weighted @xmath6 formalism , @xmath470 is written as @xmath471 , where @xmath472 and @xmath473 .
note that a ` least squares ' fit corresponds to the case where all the @xmath474s , and hence all of the @xmath462s , have the same , constant value . with these definitions
, @xmath316 can be seen to be equal to @xmath475 minus a constant .
the parameter values that minimise the sum of @xmath476 can thus be seen to also maximise @xmath17 .
this is the statistical justification for the otherwise _ purely geometric _ rationale that underpins the @xmath6 formalism .
a weighted-@xmath6 minimisation is just a special case of a maximum likelihood fit .
the ` miracle ' @xcite of the weighted @xmath6 formalism is that maximisation condition @xmath477 can be solved analytically .
since we have defined @xmath478 , each of the individual @xmath278s , and thus the summed @xmath316 , can be seen to be quadratic in @xmath479 . scaling the @xmath474s so that @xmath480 , the result is : @xmath481 note how similar this expression is to cov@xmath482var@xmath483 . for our purposes fitting the red and blue cmrs this procedure would be done separately and independently for the red and blue subpopulations , using some prior distinction to separate the two .
there are a number of important and implicit assumptions involved in writing and , each of which make the traditional @xmath6 approach unsuitable for our purposes . 1 . )
any and all uncertainties in measured values of the @xmath484s are ignored@xmath485 does not appear in .
there are two facets to the embedded assumption 2 . ) that the distribution of datapoints is well described by a perfect line .
first , 2a.)all values of @xmath12 are considered equally likely ; no attempt is made to account for the underlying distribution function for @xmath12 ( _ i.e. _ , mass ) .
second , 2b.)the relation between @xmath12 and @xmath64 is assumed to be infinitely narrow ; no allowance is made for there being an intrinsic ( astrophysical ) scatter in the relation between @xmath12 and @xmath64 .
no allowance is made for outliers , catastrophic measurement errors , or otherwise ` bad ' data ; the results of the fit are , in both principle and practice , sensitive to data that contain little or no useful information . and finally , 4 . )
any two or more populations must be fit independently using an _ a priori _ distinction ; this approach can not deal with multiple overlapping populations .
our task in this appendix is therefore to develop a better descriptive model for the data density distribution in the cmd that can overcome the considerable limitations of the traditional approach , and which can be used to objectively distinguish between red and blue galaxies .
our treatment and discussion of the problem , including that immediately above , is heavily influenced by the excellent pedagogic work of @xcite .
in general , and certainly in our case , there are significant observational uncertainties on the values of the @xmath484s as well as the @xmath463s .
further , the measurement errors in @xmath12 and @xmath64 are , in general , correlated , in the sense that if a galaxy s colour is overestimated , then so too will its mass - to - light ratio , and thus its total stellar mass . in this case , and sticking with the assumption of gaussian errors , what we have generically referred to as @xmath451 is most simply represented by a covariance matrix , @xmath77 , as defined in . consider observing an instance drawn from our generative model that ` really ' lies at the position @xmath486 , which lies somewhere along the line @xmath487 .
so long as the measurement errors can be treated as being gaussian , then the probability of observing this point at the position @xmath363 is given by : @xmath488 ~ , \nonumber\end{aligned}\ ] ] where we have now also introduced @xmath489 as a shorthand for the ( normalised ) 2d gaussian centred on the point @xmath490 , and with covariance given by the matrix @xmath491 .
this expression can thus be understood as describing the contribution to the expected , observed data density at the location @xmath363 in the observed cmd , owing to the ` true ' , underlying distribution at the location @xmath209 , when observed with measurement errors described by @xmath77 . in order to derive an expectation for the net observed data density at the location @xmath363or , in other words , the overall likelihood of observing the datapoint @xmath363it is therefore necessary to integrate over all possible values of @xmath209 .
if the underlying distribution is truly a perfect , uniformly populated and infinitely thin line , then this line integral takes the following form : @xmath492 ~ { \mathcal{g}}_2({\underline{x } } { { } _ i } - { \underline{x } } ' , { \underline{\underline{\mathrm{s } } } } { { } _ i } ) ~ , \end{aligned}\ ] ] in the second line , we have re - written the original line integral to highlight the fact that it can alternatively be understood as a convolution between the underlying data distribution and the measurement error ellipse ; _ i.e. _ , @xmath493 \otimes { \mathcal{g}}_2 ( { \underline{x } } { { } _ i } , { \underline{\underline{\mathrm{s } } } } { { } _ i } ) $ ] .
here , we have used the kronecker delta function , @xmath494 $ ] , to enforce the condition that @xmath487 for points on the line ; _ i.e. _ , @xmath495 $ ] is taken to represent the intrinsic distribution in @xmath496 space .
we have also used to re - express @xmath497 as @xmath498 . and @xmath499 .
the difference between these two quantities boils down to the distinction between _ error _ and _ uncertainty_. when a measurement is made , there is ( almost ) always some _ error _ ; that is , a difference between the true and measured values .
this quantity is , by definition , unknowable .
at the same time , whenever a measurement is made , it will ( or should ) always come with an associated _ uncertainty _ , which reflects the allowed range of ` true ' values that are consistent with the measured values .
this is something that can ( and should ) be estimated . in this sense , the uncertainties represent our priors on the error distribution , to be marginalised over .
the assumption underpinning the sleight of hand by which we have made @xmath500 is therefore that the probability distribution for the measurement _ errors _ is faithfully described by our formal measurement _ uncertainties_. in particular , the assumption is that the errors are random , not systematic , and that the error distribution function is gaussian in form . ] sticking with the assumption that the underlying distribution is a perfect line , this line or convolution integral can be done analytically .
the result is simply a 1d gaussian : @xmath501 here , @xmath502 and @xmath503 are the ( scalar ) projections of the vector @xmath363 and the error ellipse described by @xmath77 , respectively , onto the normal vector for the line , @xmath504 . in other words , if the slope of the line is @xmath479 , then @xmath504 is a unit vector in the direction @xmath505 , and @xmath506 and @xmath507 are defined as @xmath508 and ( @xmath509 ) , respectively . can just as easily be thought of in terms of the angle of the line @xmath510 , in which case @xmath504 becomes @xmath511 . ]
while this result may at first appear to be surprisingly simple , with a moment s reflection it becomes immediately intuitive . by a symmetry argument
, the probability density around the line _ must _ depend only on the perpendicular distance from it .
( think of the electrical field above a charged wire or plate . ) in the simple case of a distribution with @xmath512 , the scatter in the @xmath12 direction is immaterial for each point from the
true distribution that is scattered to the right , another will be scattered from the left to take its place .
any covariance between @xmath513 and @xmath66 , which just represents a shearing of @xmath65 along the @xmath64 axis , is similarly immaterial .
at least in the case that the errors in @xmath12 and @xmath64 are uncorrelated , the solution of @xmath514 is still analytic .
( this is sometimes referred to as deming regression , particularly among chemists and in the medical sciences . )
it is also possible to derive an analytic solution in the more general case of correlated errors .
however , we would argue that , in this day and age , it is just as easy to solve the problem numerically , which can be almost trivially done using any number of established optimisation algorithms .
this is doubly true if one wants to quantify the uncertainties in the fit parameters ( as one should ) .
furthermore , once one adopts a bayesian perspective ( as we strongly advocate ) , the ` maximum likelihood ' solution becomes all but meaningless .
instead , the bayesian strives to constrain the values of the model parameters _ given the observed data _ ; that is , to derive the posterior probability distribution function ( pdf ) for the full range of allowed parameter values .
if the above discussion has been somewhat long and laboured , it is to make the following point very clear : even within the framework of traditional maximum likelihood or @xmath6 minimisation fitting , _ it is very easy to account for completely general gaussian uncertainties . _
the only change to the formalism that is required is to shift from the offset and scatter in the @xmath64 direction to those perpendicular to the linear relation ; that is , to use the definition of @xmath201 given in in place of that in . almost without exception ,
if a more general description of the observational errors or uncertainties is available or can be assumed , there is no good reason not to use this information .
the next assumption we intend to relax is that the underlying data distribution comes from an infinitely thin line .
instead , we will allow that there is some intrinsic , astrophysical scatter around the blue and red cmrs , which we intend to fit for . in the absence of any other better motivated alternatives
, we will make the simplest and most convenient assumption that this intrinsic scatter can be treated as being gaussian .
this requires the introduction of a new parameter into the set @xmath191 , which we will denote as @xmath515 , and which should be understood as being the intrinsic variance around the ` true ' cmr .
intrinsic scatter can be accommodated by considering it as an additional kind of ` error ' on points drawn from a perfect linear relationship .
a linear relationship with gaussian scatter can be represented as the convolution of perfect line with a gaussian ; _ i.e. _ , @xmath516 \otimes { \mathcal{g}}_1(y { { } _ i } , \zeta^2)$ ] . by the associativity of convolutions
the procedures @xmath517 and @xmath518 are equivalent ; by the commutativity of convolutions , it does not matter which of the two gaussians in this schema is represents the intrinsic scatter , and which represents the measurement errors .
then , since the convolution of two gaussians with variances @xmath515 and @xmath466 is itself a gaussian with variance @xmath519 , it is clear that we are justified in treating @xmath14 as being akin to ` just another source of error '
. thus we have : @xmath520 \otimes { \mathcal{g}}_1 ( y { { } _ i } , \zeta^2 ) \otimes { \mathcal{g}}_2 ( { \underline{x } } { { } _ i } , { \underline{\underline{\mathrm{s } } } } { { } _ i } ) ~ .\ ] ] there are then two options for how to include @xmath515 . if @xmath515 is taken to be the intrinsic variance _
perpendicular to the relation _ , then it can be folded into by simply replacing @xmath521 with @xmath522 .
the alternative is to define @xmath515 as being the variance _ in the y direction_. since this is the more physically sensible way to envisage astrophysical scatter in the cmrs , this is how we choose to treat @xmath515 ; it should thus be thought of as representing the intrinsic variance in colours at fixed mass .
once we have defined @xmath14 in this way , using the arguments presented in the previous section , we can now define @xmath523 , where @xmath524 is the angle of the linear relation .
then , becomes : @xmath525 again , we stress that it is not all that hard to incorporate intrinsic scatter into the traditional maximum likelihood fitting framework . at this stage , however , ( we believe that ) the maximum likelihood problem can no longer be solved analytically ; it requires the use of a computer .
it is true that , in most cases , the inferred values of the linear fit parameters will not be strongly covariant with the inferred scatter .
however , any ` goodness of fit ' statistic and thus uncertainties on these parameters_will _ depend on the true , underlying variance .
furthermore , in many cases including ours the scatter is itself a quantity of astrophysical interest , and it makes little if any sense to neglect it .
as @xcite stress , traditional weighted @xmath6 minimisation fitting methods are generically sensitive to outliers , or ` bad data ' .
however , these authors also outline a solution to this problem the method of ( gaussian ) mixture modelling . the basic idea here
is to add a secondary component to our generative model , so that it is able to generate both ` bad ' and ` good ' data . constructing a generative model for such ` bad ' data obviously requires that we make some working assumptions about the ` true ' distribution of ` bad ' data in the observed @xmath257 plane .
developing a full and an accurate description of such ` bad ' data is not something that is easily done ; further , it is not something that we are particularly interested in doing .
that said , we echo @xcite in saying that `` the power of this [ method ] comes not from making an _ accurate _ model of the outliers , it comes simply from _ modelling _ them '' ( emphasis in the original ) . and
in the end , as described in , we will marginalise over these ` nuisance parameters ' that describe the ` bad ' data distribution , leaving us only with the parameters that we genuinely care about .
since we have no concrete knowledge of the ` bad ' distribution , we make the simplest possible assumption and treat this distribution as gaussian .
we also consider all data points as having an equal probability of being or becoming ` bad ' ; our desire to objectively identify ` bad ' data points requires that we make no _ a priori _ assumptions as to the ` badness ' or otherwise of individual data points .
we thus choose to model the ` badness ' within our data as an additional source of error in the measured values of @xmath12 and @xmath64.@xmath64 space , characterised by five parameters : the mean and variance in both @xmath12 and @xmath64 , plus an overall normalisation factor .
our solution is thus less general : we have fixed the mean values of @xmath12 and @xmath64 for the ` bad ' component to be the same as for the ` good ' one . ] in other words , the model for the bad data is constructed by convolving the ` good ' distribution with an additional ( large ) gaussian error distribution ; this error can then be treated in the same way as in the previous two sections .
assuming for the moment that a given data point _ is _ bad , then its likelihood can be written : @xmath526 in writing this expression , we have defined @xmath527 , where @xmath261 would be the error matrix that describes the ` badness ' , in direct analogy to the @xmath528 introduced in .
again , @xmath529 should be understood as the likelihood of finding a datapoint at the location @xmath530 given or assuming that it is ` bad ' .
if some fraction @xmath259 of all datapoints are bad , then the net likelihood of observing the datapoint @xmath363 becomes : @xmath531 where @xmath532 is the likelihood for ` good ' data , defined , to now , as in .
further , the ` bad ' parameter set @xmath533 should now be understood to be included as a subset of the larger parameter set @xmath191 .
the above represents our first real departure from traditional frequentist statistical analysis , inasmuch as we have now introduced the nuisance parameters @xmath534 .
while these parameters are certainly important in the calculation , we are not interested in their values _ per se_. it is through the bayesian process of marginalisation that we can push these parameters into the background , and thus focus on the quantities of genuine interest .
the utility of these parameters is that they allow us to objectively identify and ` mask ' those datapoints that can not reasonably be considered to have been drawn from the ` true ' underlying distribution , so as to limit the influence of any and all ` bad ' points on the values of the ` good ' parameters .
taking this idea just a little bit further : what we have done is created a mechanism within the fitting process that operates to accommodate outlying data within a secondary , ` bad ' component to the error / uncertainty distribution function .
further , given set of trial values for the parameters in the full set @xmath191 , the ` badness ' of any individual datapoint can be evaluated be considering the relative likelihood that that point has been drawn from either the ` bad ' or the ` good ' populations .
specifically , the probability that a point is ` bad ' is given by @xmath535 .
this mechanism can equally well be understood in two ways .
first , it can be viewed as using a two - component gaussian to describe the uncertainties on each point , with the understanding that we are now _ fitting _ for ( part of ) the error / uncertainty distribution function .
alternatively , it can be viewed as modelling the observed data distribution as being the sum of two components : a ` bad ' one , and a ` good ' one .
the ` bad ' one has both a larger scatter , as given by @xmath261 , and a lower normalisation , given by @xmath259 . by virtue of the fact that the values of these parameters are determined in the course of the fit , their inclusion has the effect of flexibly and objectively determining , on the basis of their ` badness ' , which points ought be downweighted when fitting for the ` good ' parameters . in this way
, points are classified according to their ` badness ' in a way that is both objective , and empirical ; _ i.e. _ , based on the observed dataset , in its entirety .
until now , our discussion in this section has only considered the case of fitting a single line to an observed dataset .
but as we have seen , the general galaxy population can be decomposed into two distinct but overlapping populations in the cmd .
our generative model therefore needs to simultaneously describe these two separate populations .
the conceptual basis for how we can go about doing this has already been laid out in the previous section . in the same way as we have split the observed @xmath257 distribution into ` good ' and ` bad ' components , we now split the ` good ' distribution into distinct ` blue ' and ` red ' components .
we can then consider separately the likelihoods of a particular galaxy as having been drawn from or as being a member of either the red or the blue subpopulation .
this can be done by redefining @xmath458 as : @xmath536 here , we have defined two independent subsets of @xmath191 that comprise those parameters pertaining exclusively to each population ; _
e.g. _ , @xmath537 and similarly for @xmath538 .
then , @xmath539 and @xmath540 are defined analogously to as the likelihood of drawing the datapoint @xmath363 from either the red or blue distributions . finally , in a similar way to @xmath259 above , the parameters @xmath541 and @xmath542 define the relative amplitudes of the blue and red probability distribution functions at the point @xmath363 in the ` true ' , astrophysical cmd .
further , it should now be clear the ways in which mixture modelling and objective classification are two sides of the same coin .
in contrast to @xmath259 , we have deliberately written @xmath541 and @xmath542 with a subscript @xmath196 . to understand the motivation for this decision ,
consider the following two examples .
if we were to somehow have perfect _ a priori _ knowledge of which galaxies were blue / red ( if such a thing is even possible ) , then we could set each individual @xmath541/@xmath542 to either 1 or 0 . this case would be equivalent to simultaneously but still independently fitting for the relations within the two subpopulations .
the next level of complexity would be to fit for the relative fraction of red galaxies by treating @xmath541 and @xmath542 in the same way as we have @xmath259that is , to use a global parameter @xmath212 to modulate the relative amplitudes of the gaussians used to define @xmath539 and @xmath540 , so that @xmath543 and @xmath544 .
of course , neither of these two simple cases are suitable for our purposes . instead , what we want to do is to account for the relative numbers of red and blue galaxies _ as a function of mass _ ; that is , we want to allow the values of @xmath541 and @xmath542 to vary explicitly with @xmath12 .
our method for doing so is the subject of the next section .
our next task is to find a way to incorporate a more general , non - uniform distribution in @xmath12 values_i.e .
_ , the mass function into our modelling / fitting algorithm . as the starting point for this section ,
let us restate in the following form : @xmath545 \otimes { \mathcal{g}}_1 ( y { { } _ i } , \zeta_\mathrm{blue / red}^2 ) \otimes { \mathcal{g}}_2 ( { \underline{x } } { { } _ i } , { \underline{\underline{\mathrm{s } } } } { { } _ i } ) ~ , \end{split}\ ] ] where , again , @xmath546 is the defining function for a linear relation .
let also restate equation as : @xmath547 now , realise that the principal advantage of assuming gaussian distribution functions to characterise all the different aspects of our generative model is that by doing so , each of these convolutions can be done analytically .
this is why the method of mixture modelling is typically phrased in terms of gaussian distributions .
so , how are we to proceed ?
one way of allowing for @xmath541 and @xmath542 to vary with @xmath12 would be to treat the red and blue populations as 2d gaussian distributions in the cmd with finite widths along the direction of the cmr line , by replacing the @xmath548 \otimes { \mathcal{g}}_1 ( y { { } _ i } , \zeta^2 ) $ ] with something like @xmath549
. this would be the well established method of gaussian mixture modelling , and would have the advantage of keeping the calculation of the @xmath201s analytic . of course , this is not what we want to do , because it is not astrophysically sensible .
the @xcite function : @xmath550 has been found to provide a very good description for the mass distribution function of field galaxies as well as that of many important subpopulations .
ultimately , this is the distribution function that we want to use as the basis for our characterisation of the blue / red mass functions . for now , though , let us continue our discussion in generalised terms , taking @xmath551 as a generic , parametric , functional description of the distribution of @xmath12 values .
as the first step towards folding in a more astrophysical and more general ( _ i.e. _ , non - uniform and non - gaussian ) mass distribution , let us define @xmath552 \otimes { \mathcal{g}}_1 ( y ' , \zeta_\mathrm{blue / red}^2 ) $ ] to represent the linear parts of our generative model ; _ i.e. _ , the convolution between the ( infinitely thin ) locii for the blue and red cmrs , as given by @xmath553 , and some scatter in the @xmath64 direction , described by the variance @xmath515 .
the values of these @xmath554s are analytic , as are the convolutions @xmath555 , which are necessary to describe the effect of observational errors / uncertainties .
now , the idea is to use the @xmath232s to modulate the relative amplitudes of these gaussians both the relative amplitudes of the ` blue ' and ` red ' gaussians _ at fixed mass _ , and the relative amplitudes of these distributions _ as a function of mass_. in this way , the model for the data distribution in the ( @xmath556 ) plane becomes : @xmath557 it is important at this point to remember the integral normalisation condition on @xmath199 , and thus also on the two @xmath232s .
as written , all of @xmath199 , @xmath558 , and @xmath559 should be understood to be integral normalised to unity over the ( @xmath560 ) domain ; in our case , this is @xmath561 .
this means that we can not fit for the absolute overall normalisation of @xmath562 or @xmath563 .
that said , we can fit the relative normalisations of @xmath562 and @xmath563 , using the global parameter @xmath212 , which we have now effectively defined via , and which from now on should be considered to be an element of @xmath191 .
now , as per , the value of @xmath201 comes from the convolution between @xmath192 , and the ( gaussian ) measurement error ellipse : @xmath564 but this presents a problem : the convolution @xmath565 is no longer analytic . at this point
we are thus forced to make our first formal approximation .
our solution is to break the continuous mass function @xmath232 into the sum of many discrete @xmath432 functions : @xmath566 with the appropriate scaling to satisfy the integral constraint .
note that we have not had to appeal to any special properties of the schechter function to make this approximation ; we are free to use any form of @xmath232 .
( indeed , we could even include the many @xmath567 as a suite of independent parameters to be fit for ; see , _
e.g. _ , * ? ? ?
* ) let us now turn from the general case , and define the specific parametric form for the mass functions that we actually use .
we have chosen to model both the blue and red mass functions as the sum of two independent schechter functions , _
i.e. _ : @xmath568 ~ , \end{aligned}\ ] ] with an analogous expression to define @xmath217 . here
, the two parameters @xmath569 govern the relative amplitudes of the two mass functions @xmath570 and @xmath571 in the same way that @xmath212 does for @xmath217 and @xmath218 . the approximation given in can be thought of in two complementary but , at least at this stage , equivalent ways . for any value of @xmath109
, @xmath567 is just a scalar normalisation factor for @xmath572 , and thus for the series of convolutions ; _ i.e. _ , @xmath573 , which is analytic . with this way of thinking ,
we are exactly and analytically solving an approximate model with a discrete , stepped mass function .
alternatively , the analytic convolutions @xmath574 could be thought of as being computed first . in this way of thinking ,
the shift to a discretised @xmath232 can be seen simply as computing the convolution @xmath575 numerically , using euler s method for numerical integration . in the former way of thinking , we have made a formal approximation in the _ construction _ of the model for @xmath189 ; in the latter way of thinking we have made a numerical approximation in the _ computation _ of the values of @xmath189 . for reasons that will become clear in the following two sections , we would advocate the former interpretation over the latter . with this in mind ,
our last task for this section is to explicitly define the values of @xmath225 that we adopt for the fits .
let us take the @xmath225s to be evenly spaced in the @xmath12 dimension , with a spacing given by @xmath576 , ranging from the lower limit for our sample up to some high value . in a sense , this can be thought of as using a histogram with bin centres @xmath577 in place of the continuous @xmath15 .
however , because the mass function is not generally linear in @xmath12 , the mean value of @xmath578 for this bin will be slightly different from the geometric centre of the bin ; similarly the population of the bin will be slightly different to the value of @xmath304 .
this is akin to the so - called eddington bias , and is most true in the high mass , exponential tail . to explicitly account for this ,
then , we compute : @xmath579 using trapezoidal numerical integration .
this is perhaps an unnecessary flourish : this choice has very little impact on our results .
for this reason , we have glossed over this aspect of our calculation in , where we describe the full model _ in toto_. given that we have to perform this numerical integration anyway to enforce the normalisation condition on the @xmath15s , however , making this choice has a negligible cost in terms of computational runtime .
we have therefore elected to retain this very minor ` correction ' , if only because we can .
[ [ allowing - for - more - general - relations - between - x - and - y ] ] allowing for more general relations between @xmath12 and @xmath64 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ with the approximation made in , we were able to relax the assumption of a uniformly populated line , and in so doing , accommodate a wholly general ( if parametric ) description of the @xmath12 distribution function ; _ i.e. _ , the mass function .
until now , the significance of assuming a uniform distribution function for @xmath12 has been that it made the line / convolution integral in analytic . freed from the assumption of a uniform distribution function for @xmath12 ,
the approximation in also allows us to relax the assumption of both a purely linear relation between @xmath12 and @xmath64 , as well as the assumption of a uniform scatter around the cmrs .
instead , what appears in is nothing more than the centre and width of a gaussian distribution at each and every of the discrete @xmath225s ; _ i.e. _ , @xmath580 , where @xmath581 \otimes { \mathcal{g}}(y { { } _ i } ; \zeta^2)$ ] . in the same way as we have chosen to model the ` red ' and ` blue ' mass functions as the sum of two schechter functions
, we must now state our specific parametric description of the locus of , and scatter around the ` red ' and ` blue ' cmrs ; we do this in the next two sections . in order to allow the slope of each of the cmrs to vary with @xmath12 , we elect to describe them as a combination of two linear relations .
the exact definitions for the descriptive model for @xmath582 are given in eq.s [ eq : cmrs ] and [ eq : zeta ] ; the exact functional form @xmath582 is immaterial in this pedagogical discussion . here , let us simply note that @xmath583 is described by five new parameters @xmath584 , with analogous parameters used to describe @xmath213 .
now , using the arguments made in the previous section , we can relax the assumption of a perfectly linear relation between @xmath12 and @xmath64 simply by substituting this two - line definition for @xmath16 in place of the linear @xmath16 we have assumed thus far . using the same formalism as before , @xmath585 remains defined as @xmath586 , with no further adjustments required in our formalism .
we note , however , that this is only true because we have independently characterised the distribution of points in the @xmath12 dimension as in the previous section .
if @xmath587 were uniformly populated _ along the line _ , the equivalent of the line integral given in would have to include a factor of @xmath588 to account for the fact that there would be more points ( per unit @xmath12 ) where the slope of @xmath16 is steeper .
this would mean that the line / convolution integral would no longer be analytic . in this sense
, the transformation mapping the @xmath12-axis ( _ i.e. _ , the line @xmath589 ) to the relation described by @xmath16 should be understood in terms of shearing and shifting , rather than rotation , since the @xmath12 distribution function is left unchanged .
[ [ the - next - level - of - sophistication - in - the - intrinsic - scatter - in - y ] ] the next level of sophistication in the intrinsic scatter in @xmath64 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ the final aspect of the model that remains to be developed is the allowance for the scatter in the cmrs to vary with @xmath12 . using the same arguments advanced in the previous two sections
, it should be clear that we can now simply redefine the ( until now ) constant @xmath590 that goes into the definitions of @xmath559 with some functional description @xmath591 .
the exact description that we have adopted comes from @xcite , and is given in .
the final formal step that remains to be taken , then , is to fold this change into our existing definitions for @xmath585 , @xmath189 , and @xmath592 . combining the results of the last three sections
, we have : @xmath593 that is , as previously described , the model for the true @xmath496 distribution of @xmath594 and @xmath595 galaxies comprises a discrete distribution of @xmath12 values ; at each of these discrete masses , the distribution of @xmath64 values ( _ i.e. _ colours ) is taken to be a ` bimodal ' distribution of two ( 1-d ) gaussians , the normalisations , centres , and widths of which are all allowed to vary parametrically as a function of @xmath12 ; _ i.e. _ , mass . for completeness , then , we have the model probability density distribution for ` good ' data as given by the sum of distinct ` blue ' and ` red ' components : @xmath596 with an additional ` bad ' component , which is just a badly smeared version of the ` good ' distribution : @xmath597 the generative model for the net , observed ( @xmath556 ) distribution is a then mixture of these ` good ' and ` bad ' components : @xmath598 and the likelihood of observing any given datapoint is determined by convolving this model with the observational uncertainties / errors : @xmath599 the last remaining aspect of the observational dataset that remains to be accounted for is incompleteness arising from the fact that our sample is ultimately selected on the basis of apparent magnitudes ; specifically , @xmath19-band petro magnitudes , as reported in the sdss dr6 catalogues .
this has been accounted for using the standard technique of @xmath102 weighting as motivated , described , and validated in .
the basic idea is as follows .
take the specific example of a galaxy with mass and colour such that it would only satisfy the gama spectroscopic target selection criteria over , say , 1/3 of the total gama survey volume this side of our analysis redshift interval of @xmath595 . in this case
, there would be ( modulo the effects of large - scale clustering ) 3 times as many galaxies in the real @xmath595 universe as are found in the gama catalogues . to account for this , we should count this single putative galaxy 3 times over .
that is , each galaxy should be given a weighting @xmath600 . in this way
, we arrive at last at our final expression for the overall likelihood of observing our dataset , assuming some fiducial trial values for each and every parameter in the set @xmath191 : @xmath601 or , equivalently : @xmath602 [ [ ch : modelsumm ] ] summary : a generative model for the observed distribution of galaxies in the colour mass plane ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ we have now fully developed our model to predict or describe the observed distribution of galaxies in the cmd , accounting for distinct blue and red populations with different mass functions , and allowing both the slope of and scatter around the blue and red cmrs to vary with mass .
we have also accounted for fully generalised ( if gaussian ) covariant errors in the measured values of @xmath12 and @xmath64 , as well as allowing for un - modelled or under - modelled aspects of the ( @xmath556 ) distribution as ` bad ' data .
the origins of this ` bad ' data may be be astrophysical , in the sense of some additional component in the @xmath257 plane not included in the model , or observational , in the sense of catastrophic errors in the measurements .
finally , we have also accounted for incompleteness arising from the magnitude - limited nature of the gama sample . in our pedagogical development of the model ,
we have attempted to make it clear that the conceptual framework that underpins our modelling is just an extension of the traditional ( frequentist ) weighted @xmath6 or ` maximum likelihood ' formalism for fitting a perfectly straight , perfectly narrow , and uniformly populated line .
indeed , this simple and highly idealised situation can be seen as a special case of our ( much ) more general model . in this sense , there is only one ` trick ' that we have introduced here the method of ( gaussian ) mixture modelling .
this is what allows us to flexibly describe , and thus objectively identify and quantify , the distinct ` blue ' , ` red ' , and ` bad ' components . as stated in ,
this method is intrinsically bayesian , inasmuch as the formal justification involves implicit marginalisation over all possible binomial combinations of blue-/red-/bad - ness for each individual point , with the implicit prior that all points are treated as having an equal _ a priori _ probability of being blue , or red , or ` bad ' . with this caveat ,
there is nothing preventing us from using this model to perform a frequentist ` maximum likelihood ' fit this would simply involve identifying the set of parameter values @xmath191 that maximises the scalar likelihood function @xmath603 .
however , at least with the formalism as we have laid it out , this would be dishonest . to see this , recognise that @xmath17 is only defined for a given or assumed set of values for @xmath191 .
any comparison between the values of @xmath17 for different @xmath191s thus inescapably , if implicitly , assumes that the different values of @xmath191s are equally likely , or not .
_ all fitting algorithms include priors _ ; the difference between bayesian and a frequentist statistics is only that these priors are made explicit in a bayesian setting .
this is especially important when it comes to deriving formal uncertainties on the values of the fit parameters by considering only @xmath603 , the frequentist does not have a good formal basis for making such a calculation , since they can not simultaneously assume two distinct sets of values for @xmath191 . in this sense , a traditional , frequentist ` maximum likelihood ' fit is just a bayesian maximum _ a posteriori _ ( map ) determination_i.e .
_ , the identification of the global maximum for the posterior pdf with ( unstated ) uniform priors .
we have chosen our specific priors with this in mind . in the absence of any clearly better alternatives , we assume uniform priors for just about all of the parameters in @xmath191 .
this includes uniform priors on the _ logarithm _ of the characteristic masses for the schechter functions , @xmath604 , rather than uniform priors on @xmath234 _ per se _ ) , as well as on the values of the @xmath243s ( _ cf .
_ @xmath605 ) that are used in place of the @xmath606s to modulate the relative amplitudes of the different schechter function components . the only exception to this rule is for the slopes of any and all relations , which are assumed to be linear in @xmath510 , rather than linear in @xmath479 .
this ensures that steeper slopes are not ` artificially ' down - weighted in preference of flatter ones . in all cases , we have checked to ensure that the allowed range for the priors is reasonable , and in particular that these ranges are broad enough to ensure that they do not ` artificially ' cut off the pdfs .
the only exceptions to this rule are the sensible and obvious ones ; _
e.g. _ , we require that all of the @xmath243s be in the range @xmath607 $ ] , aand that @xmath14s be positive . | we measure the mass functions for generically red and blue galaxies , using a @xmath0 sample of @xmath1 field galaxies from the galaxy and mass assembly ( gama ) survey .
our motivation is that , as we show , the dominant uncertainty in existing measurements stems from how ` red ' and ` blue ' galaxies have been selected / defined .
accordingly , we model our data as two naturally overlapping populations , each with their own mass function and colour mass relation , which enables us characterise the two populations without having to specify _ a priori _ which galaxies are ` red ' and ` blue '
. our results then provide the means to _ derive _ objective operational definitions for the terms ` red ' and ` blue ' , which are based on the phenomenology of the colour mass diagrams
. informed by this descriptive modelling , we show that : 1 . ) after accounting for dust , the stellar colours of ` blue ' galaxies do not depend strongly on mass ; 2 . )
the tight , flat ` dead sequence ' does not extend much below @xmath2 ; instead , 3 . )
the stellar colours of ` red ' galaxies vary rather strongly with mass , such that lower mass ` red ' galaxies have bluer stellar populations ; 4 . )
below @xmath3 , the ` red ' population dissolves into obscurity , and it becomes problematic to talk about two distinct populations ; as a consequence , 5 . ) it is hard to meaningfully constrain the shape , including the existence of an upturn , of the ` red ' galaxy mass function below @xmath4 .
points 14 provide meaningful targets for models of galaxy formation and evolution to aim for .
galaxies : formation and evolution galaxies : mass functions galaxies : statistics galaxies : stellar content galaxies : fundamental parameters [ [ section ] ] _ `` all nature is perverse & will not do as i wish it '' _ + charles darwin ( 1855 ) |
three models which study the stochastic behaviour of the prices of commodities that take into account several aspects of possible influences on the prices were proposed by e schwartz @xcite in the late nineties . in the simplest model ( the so - called one - factor model )
schwartz assumed that the logarithm of the spot price followed a mean - reversion process of ornstein uhlenbeck type .
the one - factor model is expressed by the following @xmath0 evolution equation@xmath1where @xmath2 measures the degree of mean reversion to the long - run mean log price , @xmath3 is the market price of risk , @xmath4 is the standard deviation of the return on the stock , @xmath5 is the stock price , @xmath6 is the drift rate of @xmath5 and @xmath7 is the time .
@xmath8 is the current value of the futures contract which depends upon the parameters @xmath9 , _
i.e. _ , @xmath10 .
generally @xmath11 , @xmath3 , @xmath4 and @xmath6 are assumed to be constants .
in such a case the closed - form solution of equation ( [ 1fm.01 ] ) which satisfies the initial condition@xmath12was given in @xcite .
it is @xmath13with @xmath14 .
it has been shown that the closed - form solution ( [ 1fm.02 ] ) follows from the application of lie point symmetries .
in particular it has been shown that equation ( [ 1fm.01 ] ) is of maximal symmetry , which means that it is invariant under the same group of invariance transformations ( of dimension @xmath15 ) as that of the black - scholes and the heat conduction equation @xcite . the detailed analysis for the lie symmetries of the three models , which were proposed by schwartz , and the generalisation to the @xmath16-factor model
can be found in @xcite .
other financial models which have been studied with the use of group invariants can be found in leach05a , leach06a , naicker , sinkala08a , sinkala08b , wafo , consta , lescot , dimas2 and references therein .
solution ( [ 1fm.02 ] ) is that which arises from the application of the invariant functions of the lie symmetry vector @xmath17and also leaves the initial condition invariant . in a realistic world parameters are not constants , but vary in time and depend upon the stock price , that is , the parameters have time and space dependence @xcite , where as space we mean the stock price parameters as an analogue to physics . in this work
we are interested in the case for which the parameters @xmath11 , @xmath3 , @xmath4 and @xmath6 are space dependent , _ ie _ , are functions of @xmath5 .
we study the lie point symmetries of the space - dependent equation ( 1fm.01 ) . as we see in that case ,
when @xmath18 , there does not exist any lie point symmetry which satisfies the initial condition ( 1fm.01a ) .
the lie symmetry analysis of the time - dependent black - scholes - merton equations was carried out recently in @xcite , it has been shown that the autonomous , and the nonautonomous black - scholes - merton equation are invariant under the same group of invariant transformations , and they are maximal symmetric .
the plan of the paper is as follows .
the lie point symmetries of differential equations are presented in section [ preliminaries ] .
in addition we prove a theorem which relates the lie point symmetries of space - dependent linear evolution equations with the homothetic algebra of the underlying space which defines the laplace operator . in section [ space1 ]
we use these results in order to study the existence of lie symmetries of for the space - dependent one - factor model ( 1fm.01 ) and we show that the space - dependent problem is not necessarily maximally symmetric . the generic symmetry vector and the constraint conditions are given and we prove a corollary in with the space - dependent linear evolution equation is always maximally symmetric when we demand that there exist at least one symmetry of the form ( [ 1fm.03 ] ) which satisfies the schwartz condition ( [ 1fm.01a ] ) . furthermore in section [ proof2 ] we consider the time - dependence problem and we show that the model is always maximally symmetric . finally in section
[ con ] we discuss our results and we draw our conclusions .
appendix[proof1 ] completes our analysis .
below we give the basic definitions and properties of lie point symmetries for differential equations and also two theorems for linear evolution equations . by definition a lie point symmetry , @xmath19 of a differential equation @xmath20 where the @xmath21 are the independent variables , @xmath22 is the dependent variable and @xmath23 is the generator of a one - parameter point transformation under which the differential equation @xmath24 is invariant .
let @xmath25 be a one - parameter point transformation of the independent and dependent variables with the generator of infinitesimal transformations being @xmath26 the differential equation @xmath24 can be seen as a geometric object on the jet space @xmath27 .
therefore we say that @xmath24 is invariant under the one - parameter point transformation with generator , @xmath28 , if @xcite @xmath29 } } \theta = 0 .
\label{go.11}\]]or equivalently@xmath29 } } \theta = \lambda \theta ~,~{mod}\theta = 0 , \label{go.12}\]]where @xmath30 } $ ] is the second prolongation of @xmath28 in the space @xmath31 .
it is given by the formula @xmath32 } = x+\eta _ { i}\partial _ { u_{,i}}+\eta _ { ij}\partial _ { u_{,ij } } , \label{go.13}\]]where @xmath33 , @xmath34 and @xmath35 is the operator of total differentiation , _ ie _ , @xmath36@xcite .
moreover , if condition ( [ go.11 ] ) is satisfied ( equivalently condition ( [ go.12 ] ) ) , the vector field @xmath28 is called a lie point symmetry of the differential equation @xmath24 . a geometric method which relates the lie and the noether point symmetries of a class of second - order differential equations
has been proposed in jgp , ijgmmp .
specifically , the point symmetries of second - order partial differential equations are related with the elements of the conformal algebra of the underlying space which defines the laplace operator . similarly , for the lie symmetries of the second - order partial differential
equation,@xmath37where @xmath38 is the laplace operator , @xmath39 is a nondegenerate tensor ( we call it a metric tensor ) and @xmath40 , the following theorem arises . ] .
[ theom1]the lie point symmetries of ( [ 1fm.04 ] ) are generated by the homothetic group of the metric tensor @xmath41 which defines the laplace operator @xmath42 .
the general form of the lie symmetry vector is @xmath43where @xmath44 is the homothetic factor of @xmath45 , @xmath46 for the killing vector ( kv , @xmath47 for homothetic vector ( hv ) , @xmath48 and @xmath49 are solutions of ( [ 1fm.05 ] ) , @xmath50 is a kv / hv of @xmath51 and the following condition holds , namely,@xmath52note that @xmath53 . another important result for the @xmath0 linear evolution equation of the form of ( [ 1fm.04 ] ) is the following theorem which gives the dimension of the possible admitted algebra .
[ theom2 ] the one - dimensional linear evolution equation can admits 0 , 1 , 3 and 5 lie point symmetries plus the homogenous and the infinity symmetries @xcite .
however , as equation ( [ 1fm.04 ] ) is time independent , it admits always the autonomous symmetry @xmath54 . in the following we apply theorems [ theom1 ] and [ theom2 ] in order to study the lie symmetries of the space - dependent one - factor model
the space - dependent one - factor model of commodity pricing is defined by the equation@xmath55 the parameters , @xmath56 , @xmath57 , @xmath58 and @xmath3 , depend upon the stock price , @xmath5 . in order to simplify equation ( [ 1fm.06 ] )
we perform the coordinate transformation @xmath59 , that is , equation ( 1fm.06 ) becomes@xmath60or @xmath61where @xmath42 is the laplace operator in the one - dimensional space with fundamental line element@xmath62and admits a two - dimensional homothetic algebra .
the gradient kv is @xmath63 and the gradient hv is @xmath64with homothetic factor@xmath65 equation ( [ 1fm.08 ] ) is of the form of ( [ 1fm.04 ] ) where now @xmath66and @xmath67without performing any symmetry analysis we observe that , when @xmath68 , ( [ 1fm.08 ] ) is in the form of the heat conduction equation and it is maximally symmetric , _ ie _ , it admits @xmath15 symmetries . in the case
for which @xmath69 from ( [ 1fm.10 ] ) we have that @xmath70where @xmath71 .
however , this is only a particular case whereas new cases can arise from the symmetry analysis .
let @xmath72 be the two hvs of the space ( [ 1fm.09 ] ) with homothetic factors @xmath44
. as ( [ 1fm.08 ] ) is autonomous and linear , it admits the lie symmetries @xmath73 , where @xmath74 is a solution of ( 1fm.08 ) , therefore from theorem [ theom1 ] we have that the possible additional lie symmetry vector is @xmath75for which the following conditions hold@xmath76@xmath77 we study two cases : a ) @xmath78 and b ) @xmath79 let @xmath78
. then ( [ 1fm.s05 ] ) is satisfied .
hence from ( [ 1fm.s04 ] ) we have the system @xmath80where @xmath81 , _ ie _ , @xmath82 .
this means that from any vector field @xmath83 we have only one symmetry .
hence from theorem [ theom2 ] condition ( [ 1fm.s06 ] ) should hold for @xmath84 and @xmath85 in this case the space - dependent one - factor model admits @xmath86 lie point symmetries .
consider that @xmath87 . from ( [ 1fm.s04 ] )
we have that@xmath88and then ( [ 1fm.s05 ] ) gives@xmath89 consider the case for which @xmath90 .
recall that for the space ( [ 1fm.09 ] ) , @xmath91 that is , from ( [ 1fm.s07 ] ) we have the conditions@xmath92@xmath93where @xmath94 .
we continue with the subcases : [ [ subcase - b1 ] ] subcase b1 + + + + + + + + + + let @xmath95 that is , @xmath96 . in this case
the symmetry conditions are : @xmath97hence we have the following system @xmath98 if system ( [ 1fm.s11])-([1fm.s13 ] ) holds for @xmath84 or @xmath99 , then equation ( [ 1fm.08 ] ) admits @xmath86 @xmath100 lie symmetries and in the case for which conditions ( [ 1fm.s11])-([1fm.s13 ] ) hold , _ ie _ , admits @xmath15 lie symmetries which is the maximum for a @xmath0 evolution equation .
[ [ subcase - b2 ] ] subcase b2 + + + + + + + + + + in the second subcase we consider that @xmath101 .
hence , if b2.a ) @xmath102 , then from ( [ 1fm.s08 ] ) and ( [ 1fm.s09 ] ) it follows that@xmath103@xmath104@xmath105where from theorem [ theom2 ] these conditions must hold for @xmath84 and @xmath106 and equation ( [ 1fm.08 ] ) is maximally symmetric .
b2.b ) let @xmath107 .
then it follows that@xmath108@xmath109@xmath110these conditions hold for @xmath84 or @xmath99 .
if these conditions hold for both @xmath84 and @xmath111 , then equation ( [ 1fm.08 ] ) is maximally symmetric .
we collect the results in the following theorem .
[ theom3]the autonomous @xmath0 linear equation ( [ 1fm.08 ] ) , apart from the symmetry of autonomy , the linear symmetry and the infinity symmetry , can admit : \a ) the two lie symmetries @xmath112 , where @xmath83 is a hv of the one - dimensional flat space with @xmath113 if and only if condition ( [ 1fm.s06 ] ) holds for @xmath84 and @xmath99 .
b1 ) the two or four lie symmetries @xmath114 if conditions ( [ 1fm.s11])-([1fm.s13 ] ) hold for @xmath84 or @xmath115 and @xmath116 and @xmath111 , respectively , where @xmath117 and @xmath118 b2.a ) the four lie symmetries ( [ 1fm.s20a ] ) if conditions ( [ 1fm.s14])([1fm.s16 ] ) hold for @xmath84 and @xmath111 , where @xmath119 and @xmath120 is given by ( [ 1fm.s20 ] ) .
b2.b ) the two or four lie symmetries ( [ 1fm.s20a ] ) if and only if conditions ( [ 1fm.s17])([1fm.s19 ] ) hold for @xmath84 or @xmath115 and @xmath84 and @xmath111 , respectively , where @xmath121 furthermore , we comment that theorem ( [ theom3 ] ) holds for all linear autonomous equations of the form of ( [ 1fm.04 ] ) . here
we discuss the relation among the lie symmetries and the initial condition ( [ 1fm.01a ] ) . in the case of constant parameters ,
_ ie _ , in equation ( [ 1fm.01 ] ) the lie symmetry vector ( [ 1fm.03 ] ) is the linear combination among the linear symmetry @xmath122 and the symmetry which is generated by the kv of the underlying space , which is @xmath123for @xmath124 .
however , for a general function , @xmath125 , in order for the symmetry which is generated by the kv @xmath126 to satisfy the initial condition @xmath127 or the initial condition has to change .
consider now that @xmath128 and satisfies the conditions @xmath129then from theorem [ theom3 ] , b2.a , we have that @xmath130 is given by ( [ 1fm.11 ] ) and at the same time @xmath131 generates two lie point symmetries for equation ( [ 1fm.08 ] ) .
the lie point symmetries are @xmath132@xmath133@xmath134plus the autonomous and trivial symmetries .
the symmetry vector field @xmath135 is the kv of the one - dimensional space .
therefore , if we wish the field @xmath136 to satisfy an initial condition such as @xmath137 , then it should be @xmath138 which gives @xmath139 .
from this we can see that , when @xmath140 , we have the initial condition ( [ 1fm.01a ] ) .
let @xmath11 , @xmath3 and @xmath6 be constants .
hence from ( [ 1fm.11 ] ) we have that @xmath141where for @xmath142 we have @xmath143and the solution for position @xmath144@xmath145 let now @xmath146 and consider that the kv @xmath147 generates a lie point symmetry of equation ( 1fm.08 ) from case a of theorem [ theom3 ]
. then from condition ( 1fm.s06 ) we have that@xmath148that is , @xmath149 however , in that case , equation ( [ 1fm.08 ] ) is maximally symmetric and admits @xmath150lie point symmetries .
consider reduction with the lie symmetry @xmath151 which keeps invariant the initial condition@xmath152 the application of @xmath153 in ( [ 1fm.08 ] ) gives @xmath154@xmath155 as another application of theorem [ theom3 ] we select @xmath156 .
then the kv @xmath126 is @xmath157 .
let this generate a lie point symmetry for equation ( [ 1fm.08 ] ) from the case a of theorem [ theom3 ] , that is , conditions ( [ 1fm.s06 ] ) give@xmath158where now we can see that equation ( [ 1fm.08 ] ) is maximally symmetric and admits @xmath15 point symmetries . consider the lie symmetry @xmath159 , which leaves invariant the modified initial condition @xmath160 .
the invariant solution which follows is @xmath161 we observe that , when @xmath126 generates a lie point symmetry for equation ( [ 1fm.08 ] ) , the functional form of @xmath162 , which includes @xmath163 and @xmath164 has a specific form , such that equation ( [ 1fm.08 ] ) is maximally symmetric and equivalent with the black - scholes and the heat equations . in general , for unknown function @xmath130 , from theorem [ theom1 ]
we have the following corollary .
[ cor]when the kv of the underlying space which defines the laplace operator in equation ( [ 1fm.08 ] ) generates a lie point symmetry , the functional form of @xmath162 is @xmath165and equation ( [ 1fm.08 ] ) is maximally symmetric .
the symmetry vectors , among the autonomous , the homogeneous and the infinity symmetries , are:@xmath166@xmath167@xmath168for @xmath169 , @xmath170 , @xmath171@xmath172@xmath173for @xmath169 , @xmath174 and@xmath175@xmath176@xmath177for @xmath178 , @xmath179 , where @xmath180 and @xmath181 are the elements of the homothetic algebra of the underlying space .
we note that corollary [ cor ] holds for all autonomous linear 1 + 1 evolution equations . in the following section we discuss the group invariants of the time - dependent problem .
when the parameters @xmath182 of equation ( [ 1fm.01 ] ) depend upon time , the one - factor model can be written as @xmath183where@xmath184 without loss of generality we can select @xmath185 . by analysing the determining equations as provided by the sym package dimas05a , dimas06a
, dimas08a we find that the general form of the lie symmetry vector is @xmath186 \partial _ { x } \nonumber \label{a3 } \\[1pt ] & & + \left [ f(t)+\frac{1}{4}\left ( 4xbq+x(1 - 2p)a^{\prime 2}qa^{\prime } -4x(b^{\prime } + ap^{\prime } ) \right .
\right .
\nonumber \\[1pt ] & & \left .
+ a^{\prime \prime } ) \right ) \right ] f\partial _ { f},\end{aligned}\]]where functions @xmath187 are given by the system of ordinary differential equations , @xmath188 & & + b^{\prime } -2pb^{\prime } -2f^{\prime } + ap^{\prime } -2app^{\prime } + aq^{\prime } -\frac{a^{\prime \prime } } { 2},\end{aligned}\]]@xmath189 & & + 3a^{\prime } p^{\prime } -aq^{\prime } -2bq^{\prime } + 2apq^{\prime } + 2b^{\prime \prime } + 2ap^{\prime \prime } \end{aligned}\]]and @xmath190 in addition to the infinite number of solution symmetries .
consequently the algebra is @xmath191 so that it is related to the classical heat equation by means of a point transformation . in the following we discuss our results .
in the models of financial mathematics the parameters of the models are assumed to be constants .
however , in real problems these parameters can depend upon the stock prices and upon time . in this work we considered the one - factor model of schwartz
and we studied the lie symmetries in the case for which the parameters of the problem are space - dependent . in terms of lie symmetries , the one - factor model it is maximally symmetric and it is equivalent with the heat equation , but in the case where the parameters are space dependent , that is not necessary true , and we show that the model can admit 1 , 3 or 5 lie point symmetries ( except the trivial ones ) . to perform this analysis we studied the lie symmetries of the autonomous linear evolution equation and we found that there exist a unique relation among the lie symmetries and the collineations of the underlying geometry , where as geometry we define the space of the second derivatives .
however , for a specific relation among the parameters of the model the system is always maximally symmetric .
in particular , that holds when @xmath130 is an arbitrary function and @xmath192where @xmath193 are constants . in that case , the correspoding symmetry ( 1fm.03 ) becomes @xmath194 .
consider that @xmath195 , and ( con.01 ) holds .
then the application of the lie symmetry @xmath196 in ( 1fm.08 ) gives the solution@xmath197where in the limit @xmath198 , solution ( [ con.02 ] ) becomes@xmath199which can compared with solution ( [ 1fm.02 ] ) .
consider now that @xmath130 is periodic around the line @xmath200 .
let @xmath201 that and ( [ con.01 ] ) holds .
hence the solution of the space - dependent one - factor model ( [ 1fm.08 ] ) which follows from the lie symmetry @xmath196 is @xmath202which is a periodic function of the stock price @xmath203 . for @xmath204 the taylor expansion of the static solution ( [ con.04 ] ) around the point @xmath205 , is@xmath206 in figure [ fig3 ]
we give the static evolution of the solutions , ( con.03 ) and ( [ con.04 ] ) , for various values of the constant @xmath207 . ) ( left figures ) and solution ( [ con.04 ] ) ( right .
figure ) for various values of the constant @xmath208 .
solid line is for @xmath209 , dash dash line is for @xmath209 , and the dash dot line is for @xmath210.,height=264 ] on the other hand , in section [ proof2 ] we studied the case for which the parameters of the one - factor model are time - dependent and we showed that the model is always maximally symmetric and equivalent with the heat equation , that is , the time - depedence does not change the admitted group invariants of the one - factor model ( [ 1fm.01 ] ) . a more general consideration will be to extend this analysis to the two - factor and three - factor models and also to study the cases for which the parameters are dependent upon the stock price and upon the time , _ ie _ , the parameters are space and time dependent .
this work is in progress .
finally we remark how useful are the methods which are applied in physics and especially in general relativity for the study of space - dependent problems in financial mathematics .
the reason for this is that from the second derivatives a ( pseudo)riemannian manifold can be defined .
this makes the use of the methods of general relativity and differential geometry essential .
the research of ap was supported by fondecyt postdoctoral grant no .
rmm thanks the national research foundation of the republic of south africa for the granting of a postdoctoral fellowship with grant number 93183 while this work was being undertaken .
in @xcite it has been shown that for a second - order pde of the form , @xmath211the lie symmetries are generated by the conformal algebra of the tensor @xmath212 .
specifically the lie symmetry conditions for equation ( [ eq.02 ] ) are @xmath213@xmath214@xmath215where@xmath216 by comparison of equations ( [ 1fm.04 ] ) and ( [ eq.02 ] ) we have that @xmath217 , _ ie _ , @xmath218 and@xmath219where @xmath220 . therefore the symmetry vector @xmath28 for equation ( [ 1fm.04 ] ) has the form @xmath221we continue with the solution of the symmetry conditions .
when we replace @xmath222 in ( [ eq.05 ] ) , it follows that @xmath223 which means that @xmath224 , where @xmath225 is a ckv of the metric , @xmath51 , with conformal factor @xmath44,@xmath226 _ ie_,@xmath227 and @xmath228 furthermore , from ( eq.04 ) the following system follows ( recall that @xmath229 and @xmath230 ) @xmath231@xmath232moreover we observe that @xmath233 , where @xmath234 is a constant ; that is , @xmath50 is a kv / hv of @xmath51 . finally for the function , @xmath235 it holds that @xmath236 case i : let @xmath247 .
then from ( [ eq.18 ] ) @xmath248 which means that @xmath249 .
however , from ( eq.19 ) we have that @xmath250 which gives the linear symmetry @xmath251 . in that case from the form of @xmath252 the autonomous symmetry @xmath253 arises .
sophocleuous c , leach pgl and andriopoulos k ( 2008 ) algebraic properties of evolution partial differential equations modelling prices of commodities _ mathematical methods in the applied sciences _ * 31 * 679 - 694 leach pgl , ohara jg & sinkala w ( 2006 ) symmetry - based solution of a model for a combination of a risky investment and a riskless investment _ journal of mathematical analysis and application _ * 334 * 368 - 381 sinkala w , leach pgl & ohara jg ( 2008 ) optimal system and group - invariant solutions of the cox - ingersoll - ross pricing equation _ mathematical methods in the applied sciences _ * 31 * 679 - 694 ( doi : 10.1002/maa.935 ) ibragimov nh & soh cw ( 1997 ) solution of the cauchy problem for the black - scholes equation using its symmetries , _ proceedings of the international conference on modern group analysis
. mars , nordfjordeid , norway_. cimpoiasu r & constantinescu r , ( 2012 ) new symmetries and particular solutions for 2d black - scholes model , _ proceedings of the 7th mathematical physics meeting : summer school and conference on modern mathematical physics , belgrade , serbia _ tamizhmani km , krishnakumar k & leach pgl ( 2014 ) algebraic resolution of equations of the black scholes type with arbitrary time - dependent parameters , _ applied mathematics and computations _ * 247 * 115 - 124 paliathanasis a & tsamparlis m ( 2014 ) the geometric origin of lie point symmetries of the schrodinger and the klein - gordon equations , _ international journal of geometric methods in modern physics _ * 11 * 1450037 dimas s & tsoubelis d ( 2005 ) sym : a new symmetry - finding package for mathematica _ group analysis of differential equations _
ibragimov nh , sophocleous c & damianou pa edd ( university of cyprus , nicosia ) 64 - 70 | we consider the one - factor model of commodities for which the parameters of the model depend upon the stock price or on the time .
for that model we study the existence of group - invariant transformations . when the parameters are constant , the one - factor model is maximally symmetric .
that also holds for the time - dependent problem .
however , in the case for which the parameters depend upon the stock price ( space ) the one - factor model looses the group invariants . for specific functional forms of the parameters the model admits other possible lie algebras . in each case
we determine the conditions which the parameters should satisfy in order for the equation to admit lie point symmetries .
some applications are given and we show which should be the precise relation amongst the parameters of the model in order for the equation to be maximally symmetric . finally we discuss some modifications of the initial conditions in the case of the space - dependent model .
we do that by using geometric techniques .
* keywords : * lie point symmetries ; one - factor model ; prices of commodities * msc 2010 : * 22e60 ; 35q91 |
linear - optical architectures belong to the most prominent platforms for realizing protocols of quantum information processing @xcite .
they are experimentally feasible and they work directly with photons without the necessity to transfer the quantum state of a photonic qubit into another quantum system like an ion etc .
the latter feature is quite convenient because photons are good carriers of information for communication purposes .
linear - optical quantum gates achieve the non - linearity necessary for the interaction between qubits by means of the non - linearity of quantum measurement . unfortunately , quantum measurement is not only non - linear but also probabilistic
. therefore linear - optical implementations of quantum gates are mostly probabilistic too their operation sometimes fails .
partly , this is a fundamental limitation .
but in many cases when data qubits appear in an improper state after the measurement on an ancillary system they can still be corrected by applying a proper unitary transformation which depends only on the measurement result . in these situations ,
implementation of the feed forward can increase the probability of success of the gate @xcite . in the present paper
, we apply this approach to a linear - optical programmable quantum gate .
the most of conventional computers use fixed hardware and different tasks are performed using different software .
this concept can be , in principle , applied to quantum computers as well : the executed unitary operation can be determined by some kind of a program . however , in 1997 nielsen and chuang @xcite showed that an @xmath1-qubit quantum register can perfectly encode at most @xmath2 distinct quantum operations .
although this bound rules out perfect universally - programmable quantum gates ( even unitary transformations on only one qubit form a group with uncountably many elements ) , it is still possible to construct approximate or probabilistic programmable quantum gates and optimize their performance for a given size of the program register .
such gates can either operate deterministically , but with some noise added to the output state @xcite , or they can operate probabilistically , but error free @xcite
. combination of these regimes is also possible .
a probabilistic programmable phase gate was proposed by vidal , masanes , and cirac @xcite .
it carries out rotation of a single - qubit state along the @xmath0-axis of the bloch sphere .
the angle of rotation ( or the phase shift ) is programmed into a state of a single - qubit program register .
it is worth noting that an exact specification of an angle of rotation would require infinitely many classical bits .
but here the information is encoded into a single qubit only .
the price to pay is that the success probability of such a gate is limited by 50% @xcite .
the programmable phase gate was experimentally implemented for the first time in 2008 @xcite .
however , the success probability of that linear - optical implementation reached only 25% .
in the present paper we will show how to increase the success probability of this scheme to its quantum mechanical limit of 50% by means of electronic feed forward .
the programmable phase gate works with a data and program qubit .
the program qubit is supposed to contain information about the phase shift @xmath3 encoded in the following way : @xmath4 the gate performs a unitary evolution of the data qubit which depends on the state of the program qubit : @xmath5 without loss of generality we can consider only pure input states of the data qubit : @xmath6
so the output state of the data qubit reads : @xmath7 experimentally the programmable phase gate can be implemented by an optical setup shown in fig .
[ fig - scheme ] .
each qubit is represented by a single photon which may propagate in two optical fibers .
the basis states @xmath8 and @xmath9 correspond to the presence of the photon in the first or second fiber , respectively .
when restricted only to the cases where a single photon emerges in each output port , the conditional two - photon output state reads ( the normalization reflects the fact that the probability of this situation is @xmath10 ) : @xmath11 , \ ] ] where @xmath12 . if we make a measurement on the program qubit in the basis @xmath13 then also the output state of the data qubit collapses into one of the two following states according to the result of the measurement : @xmath14 . if the measurement outcome is @xmath15 then the unitary transformation @xmath16 has been applied to the data qubit .
if the outcome is @xmath17 then the state acquires an extra @xmath18 phase shift , i.e. , @xmath19 has been executed .
this is compensated by a fast electro - optical modulator which applies a corrective phase shift @xmath20 ( in practice we apply phase shift @xmath18 what is equivalent ) .
the scheme of the setup is shown in fig .
[ fig - scheme ] .
pairs of photons are created by type - ii collinear frequency - degenerate spontaneous parametric down conversion ( spdc ) in a two - millimeter long bbo crystal pumped by a diode laser ( coherent cube ) at 405 nm .
the photons are separated by a polarizing beam splitter and coupled into single - mode fibers . using fiber polarization controllers the same polarizations
are set on the both photons . by means of fiber couplers and
electro - optical phase modulators the required input states of the program and data qubits are prepared . to prepare state ( [ eq - prog_qubit ] ) of the program qubit a fiber coupler ( fc ) with fixed splitting ratio 50:50 is used .
an arbitrary state of the data qubit ( [ eq - dat_qubit_in ] ) is prepared using electronically controlled variable ratio coupler ( vrc ) .
all employed phase modulators ( eo space ) are based on the linear electro - optic effect in lithium niobate .
their half - wave voltages are about 1.5v .
these phase modulators ( pm ) exhibit relatively high dispersion .
therefore one pm is placed in each interferometer arm in order to compensate dispersion effects . because the overall phase of a quantum state is irrelevant it is equivalent to apply either a phase shift @xmath21 to @xmath9 or @xmath22 to @xmath8 .
the gate itself consists of the exchange of two rails of input qubits and of the measurement on the data qubit ( see fig .
[ fig - scheme ] ) .
the measurement in basis @xmath23 is accomplished by a fiber coupler with fixed splitting ratio 50:50 and two single photon detectors . in this experiment
we use single photon counting modules ( perkin - elmer ) based on actively quenched silicon avalanche photodiodes .
detectors d@xmath24 , d@xmath25 , and d@xmath26 belongs to a quad module spcm - aq4c ( total efficiencies 5060% , dark counts 370440s@xmath27 , response time 3340ns ) . as detector d@xmath28 , serving for the feed forward , a single module spcm aqr-14fc is used because of its faster response ( total efficiency about 50% , dark counts 180s@xmath27 , response time 17ns ) .
the output of the detector is a 31ns long ttl ( 5v ) pulse . to implement the feed forward the signal from detector d@xmath28
is led to a passive voltage divider in order to adapt the 5v voltage level to about 1.5v ( necessary for the phase shift of @xmath18 ) and then it is led directly to the phase modulator .
the coaxial jumpers are as short as possible . the total delay including the time response of the detector is 20ns . to compensate this delay , photon wave - packets representing data qubits are retarded by fiber delay lines ( one coil of fiber of the length circa 8 m in each interferometer arm ) .
timing of the feed - forward pulse and the photon arrival was precisely tuned .
coherence time of photons created by our spdc source is only several hundreds of femtoseconds .
the right - most block in fig . [ fig - scheme ] enables us to measure the data qubit at the output of the gate in an arbitrary basis .
these measurements are necessary to evaluate performance of the gate .
the whole experimental setup is formed by two mach - zehnder interferometers ( mzi ) .
the length of the arms of the shorter mzi is about 10.5 m ( the upper interferometer in fig .
[ fig - scheme ] ) .
the length of the arms of the longer one is about 21.5 m ( the lower interferometer one in fig .
[ fig - scheme ] ) . to balance the arm lengths we use motorized air gaps with adjustable lengths . inside the air gaps , polarizers and wave plates
are also mounted .
they serve for accurate setting of polarizations of the photons ( to obtain high visibilities the polarizations in the both arms of each mzi must be the same ) . to reduce the effect of the phase drift caused by fluctuations of temperature and temperature gradients we apply both passive and active stabilization .
the experimental setup is covered by a shield minimizing air flux around the components and the both delay fiber loops are winded on an aluminium cylinder which is thermally isolated . besides , after each three seconds of measurement an active stabilization is performed .
it measures intensities for phase shifts 0 and @xmath29 and if necessary it calculates phase compensation and applies corresponding additional corrective voltage to the phase modulator .
this results in the precision of the phase setting during the measurement period better than @xmath30 . for the stabilization purposes we use a laser diode at 810 nm . to ensure the same spectral range ,
both the laser beam and spdc generated photons pass through the same band - pass interference filter ( spectral fwhm 2 nm , andover ) . during the active stabilization
the source is automatically switched from spdc to a laser diode .
any quantum operation can be fully described by a completely positive ( cp ) map . according to the jamiolkowski - choi isomorphism
any cp map can be represented by a positive - semidefinite operator @xmath31 on the tensor product of input and output hilbert spaces @xmath32 and @xmath33 @xcite .
the input state @xmath34 transforms according to @xmath35.\ ] ] combinations of different input states with measurements on the output quantum system represent effective measurements performed on @xmath36 .
a proper selection of the input states and final measurements allows us to reconstruct matrix @xmath31 from measured data using maximum likelihood ( ml ) estimation technique @xcite . for each phase shift ,
i.e , for a fixed state of the program qubit , we used six different input states of the data qubit , namely @xmath37 and @xmath38 . for each of these input states the state of the data qubit at the output of the gate
was measured in three different measurement basis , @xmath39 and @xmath40 .
each time we simultaneously measured two - photon coincidence counts between detectors d@xmath24 & d@xmath25 , d@xmath24 & d@xmath26 , d@xmath28 & d@xmath25 , d@xmath28 & d@xmath26 in 12 three - second intervals
. the unequal detector efficiencies were compensated by proper rescaling of the measured coincidence counts . from these data
we have reconstructed choi matrices describing the functioning of the gate for several different phase shifts . in figs .
[ fig - choi - halfpi ] and [ fig - choi - pi ] there are examples of the choi matrices of the gate for @xmath41 and @xmath42 , respectively . reconstructed : ideal : + + reconstructed : ideal : + + to quantify the quality of gate operation we have calculated the process fidelity . if @xmath43 is a one - dimensional projector then the common definition of process fidelity is @xmath44 / ( \mathrm{tr}[\chi]\mathrm{tr}[\chi_{\mathrm{id}}]).\ ] ] here @xmath43 represents the ideal transformation corresponding to our gate . in particular , @xmath45 where @xmath46 stays for the unitary operation ( [ eq - uphi ] ) applied by the gate .
we have also reconstructed density matrices of the output states of the data qubit corresponding to all input states and calculated fidelities and purities of the output states . the fidelity of output state @xmath47 is defined as @xmath48 with @xmath49 being the ( pure ) input state . the purity of the output state is defined as @xmath50 $ ] .
if the input state is pure the output state is expected to be pure as well .
table [ tab - withff ] shows process fidelities for seven different phase shifts .
it also shows the average and minimal values of output state fidelities and average and minimal purities of output states .
fidelities and purities are averaged over six output states corresponding to six input states described above .
also the minimum values are related to these sets of states .
.process fidelities ( @xmath51 ) , average ( @xmath52 ) and minimal ( @xmath53 ) output - state fidelities , average ( @xmath54 ) and minimal ( @xmath55 ) output - state purities for different phases ( @xmath3 ) _ with _ feed forward ( @xmath56 ) . [ cols="^,^,^,^,^,^",options="header " , ]
we have implemented a reliable and relatively simple electronic feed - forward system which is fast and which does not require high voltage .
we employed this technique to double the success probability of a programmable linear - optical quantum phase gate .
we showed that the application of the feed forward does not affect substantially either the process fidelity or the output - state fidelities . beside the improvement of efficiency of linear - optical quantum gates , this feed forward technique can be used for other tasks , such as quantum teleportation experiments or minimal disturbance measurement .
the authors thank lucie elechovsk for her advice and for her help in the preparatory phase of the experiment .
this work was supported by the czech science foundation ( 202/09/0747 ) , palacky university ( prf-2011 - 015 ) , and the czech ministry of education ( msm6198959213 , lc06007 ) .
increasing the number of program bits in semi - classical deterministic case one increases the precision , increasing the number of program qubits in quantum probabilistic case one increases the probability of success ( see also ref . | we have successfully used a fast electronic feed forward to increase the success probability of a linear optical implementation of a programmable phase gate from 25% to its theoretical limit of 50% .
the feed forward applies a conditional unitary operation which changes the incorrect output states of the data qubit to the correct ones .
the gate itself rotates an arbitrary quantum state of the data qubit around the @xmath0-axis of the bloch sphere with the angle of rotation being fully determined by the state of the program qubit .
the gate implementation is based on fiber optics components .
qubits are encoded into spatial modes of single photons .
the signal from the feed - forward detector is led directly to a phase modulator using only a passive voltage divider .
we have verified the increase of the success probability and characterized the gate operation by means of quantum process tomography .
we have demonstrated that the use of the feed forward does not affect either the process fidelity or the output - state fidelities . |
lifetime distribution represents an attempt to describe , mathematically , the length of the life of a system or a device .
lifetime distributions are most frequently used in the fields like medicine , engineering etc .
many parametric models such as exponential , gamma , weibull have been frequently used in statistical literature to analyze lifetime data .
but there is no clear motivation for the gamma and weibull distributions .
they only have more general mathematical closed form than the exponential distribution with one additional parameter .
+ recently , one parameter lindley distribution has attracted the researchers for its use in modelling lifetime data , and it has been observed in several papers that this distribution has performed excellently .
the lindley distribution was originally proposed by lindley @xcite in the context of bayesian statistics , as a counter example of fudicial statistics which can be seen that as a mixture of exp(@xmath0 ) and gamma(2 , @xmath0 ) .
more details on the lindley distribution can be found in ghitany et al .
+ a random variable x is said to have lindley distribution with parameter @xmath0 if its probability density function is defined as : + @xmath1 with cumulative distribution function @xmath2 some of the advances in the literature of lindley distribution are given by ghitany et al .
@xcite who has introduced a two - parameter weighted lindley distribution and has pointed that lindley distribution is particularly useful in modelling biological data from mortality studies .
mahmoudi et .
@xcite have proposed generalized poisson lindley distribution .
bakouch et al .
@xcite have come up with extended lindley ( el ) distribution , adamidis and loukas @xcite have introduced exponential geometric ( eg ) distribution .
shanker et .
@xcite have introduced a two - parameter lindley distribution .
zakerzadeh et al.@xcite have proposed a new two parameter lifetime distribution : model and properties .
hassan @xcite has introduced convolution of lindley distribution .
ghitany et al.@xcite worked on the estimation of the reliability of a stress - strength system from power lindley distribution .
elbatal et al.@xcite has proposed a new generalized lindley distribution .
+ risti @xcite has introduced a new family of distributions with survival function given by @xmath3 in this paper we introduce a new family of distribution generated by a random variable @xmath4 which follows one parameter lindley distribution .
the survival function of this new family is given as : @xmath5 where @xmath6 and @xmath7 is a cumulative distribution function(cdf ) which we use to generate a new distribution .
the cdf @xmath8 is referred to as a transformer and the corresponding probability density function ( pdf ) is given by @xmath9 we consider the transformer to follow exponential distribution with cdf @xmath10 .
hence the survival function of the new distribution is given by @xmath11 with corresponding density given by @xmath12 we refer the random variable with survival function ( 4 ) as lindley - exponential(l - e ) distribution with parameters @xmath0 and @xmath13 which we denote by l - e(@xmath14 ) .
the aim of this paper is to study the mathematical properties of the l - e distribution and to illustrate its applicability .
the contents are organized as follows .
the analytical shapes of the pdfin equations ( 5 ) are established in section 2 . the quantile function presented in section 3 .
the expressions for the moment generating function and moments corresponding to equation ( 5 ) are given in section 4 .
limiting distribution of sample statistics like maximum and minimum has been shown in section 5 . in section 6 ,
entropy of l - e distribution is presented .
the maximum likelihood estimation procedure is considered in section 7 .
the performance of the maximum likelihood estimators for small samples is assessed by simulation in section 8 .
section 9 gives estimation of stress - strength parameter r by using maximum likelihood estimation method .
finally we conclude the paper by showing applicability of the model to the real data sets .
here , the shape of pdf ( 5 ) follows from theorem 1 . + * theorem 1 : * the probability density function of the l - e distribution is decreasing for @xmath15 and unimodel for @xmath16 . in the latter case , mode is a root of the following equation : @xmath17 _ proof : _ the first order derivative of @xmath18 is @xmath19 where , @xmath20 . for @xmath21 ,
the function @xmath22 is negative .
so @xmath23 for all @xmath24 .
this implies that @xmath25 is decreasing for @xmath26 .
also note that , @xmath27 and @xmath28 .
this implies that for @xmath16 , @xmath29 has a unique mode at @xmath30 such that @xmath31 for @xmath32 and @xmath33 for @xmath34 .
so , @xmath35 is unimodal function with mode at @xmath36 .
the pdf for various values of @xmath13 and @xmath0 are shown in figure 1 .
@xmath37 and @xmath0.,scaledwidth=90.0% ] we , now , consider the hazard rate function ( hrf ) of the l - e distribution , which is given by @xmath38 + * proposition 1 * : for @xmath39 the hazard rate function follows relation @xmath40 .
+ * proof : * the proof is straight forward and is omitted .
+ in figure 2 , hazard function for different values of parameters @xmath41 and @xmath42 .
@xmath37 and @xmath0.,scaledwidth=90.0% ]
the cdf , @xmath43 , can be obtained by using eq.([4 ] ) .
further , it can be noted that @xmath44 is continuous and strictly increasing so the quantile function of @xmath45 is @xmath46 , @xmath47 . in the following theorem
, we give an explicit expression for @xmath48 in terms of the lambert @xmath49 function . for more details on lambert @xmath50 function
we refer the reader to jodr @xcite .
* theorem 2 : * for any @xmath51 , the quantile function of the l - e distribution @xmath45 is @xmath52 where @xmath53 denotes the negative branch of the lambert w function .
+ + _ proof : _ by assuming @xmath54 , the cdf can be written as @xmath55 for fixed @xmath56 and @xmath57 , the @xmath58 quantile function is obtained by solving @xmath59 .
by re - arranging the above , we obtain @xmath60 taking exponential and multiplying @xmath61 on both sides , we get + @xmath62 by using definition of lambert - w function ( @xmath63 , where @xmath64 is a complex number ) , we see that @xmath65 is the lambert @xmath49 function of the real argument @xmath66 .
thus , we have @xmath67 moreover , for any @xmath51 it is immediate that @xmath68 , and it can also be checked that @xmath69 since @xmath57 . therefore , by taking into account the properties of the negative branch of the lambert w function , we have @xmath70 also by substituting @xmath54 in cdf and solving it for @xmath71 , we get @xmath72 @xmath73 further the first three quantiles we obtained by substituting @xmath74 in equation ( 11 ) . @xmath75
the moment generating function of the random variable @xmath76 follow l - e distribution is given as + @xmath77 where , @xmath78 and @xmath79 known as digamma function . +
hence the first and second raw moments can be obtained by @xmath80 and @xmath81 respectively .
+ @xmath82 where @xmath83 is eulergamma constant = 0.577216 .
+ table 1 displays the mode , mean and median for l - e distribution for different choices of parameter @xmath13 and @xmath0 .
it can be observed from the table that all the three measures of central tendency decrease with increase in @xmath13 and increase with an increase in @xmath0 .
also for any choice of @xmath13 and @xmath0 it is observed that mean @xmath84 median @xmath84 mode , which is an indication of positive skewness . | in this paper , we introduce a new distribution generated by lindley random variable which offers a more flexible model for modelling lifetime data . various statistical properties like distribution function , survival function ,
moments , entropy , and limiting distribution of extreme order statistics are established .
inference for a random sample from the proposed distribution is investigated and maximum likelihood estimation method is used for estimating parameters of this distribution .
the applicability of the proposed distribution is shown through real data sets .
* keyword : * lindley distribution , entropy , stress - strength reliability model , maximum likelihood estimator . +
* ams 2001 subject classification : * 60e05 |
a few enumerative problems of combinatorial theory lead to generating functions that are expressed as infinite products and admit the unit circle as a natural boundary . functions with a fast growth near the unit circle are usually amenable to the saddle point method , a famous example being the integer partition generating function .
we consider here functions of moderate growth , which are outside the scope of the saddle point method .
we do so in the case where neither singularity analysis nor darboux s method is directly applicable , but the function to be analysed can be factored into the product of an elementary function with isolated singularities and a sufficiently smooth factor on the unit circle .
such decompositions are often attached to infinite products exhibiting a regular enough structure and are easily obtained by the introduction of suitable convergence factors . under such conditions ,
we prove that coefficients admit full asymptotic expansions involving powers of logarithms and descending powers of the index @xmath0 , as well as periodically varying coefficients .
applications are given to the following combinatorial - probabilistic problems : the enumeration of permutations with distinct cycle lengths , the probability that two permutations have the same cycle - length profile , the number of permutations admitting an @xmath1th root , the probability that a polynomial over a finite field has factors of distinct degrees , and the number of forests composed of trees of different sizes .
we start by recalling in section [ darbsa - sec ] the principles of two classical methods dedicated to coefficient extraction in combinatorial generating functions , namely darboux s method and singularity analysis , which are central to our subsequent developments .
the hybrid method _
per se _ forms the subject of section [ hybrid - sec ] , where our main result , theorem [ hybrid2-thm ] , is established .
section [ discyc - sec ] treats the asymptotic enumeration of permutations having distinct cycle sizes : this serves to illustrate in detail the hybrid method at work .
section [ hodgepodge - sec ] discusses more succinctly further combinatorial problems leading to generating functions with a natural boundary
these are relative to permutations , forests , and polynomials over finite fields .
a brief perspective is offered in our concluding section , section [ conc - sec ] .
in this section , we gather some previously known facts about darboux s method , singularity analysis , and basic properties of analytic functions that are central to our subsequent analyses . throughout this study
, we consider analytic functions whose expansion at the origin has a finite radius of convergence , that is , functions with singularities at a finite distance from the origin . by a simple scaling of the independent variable
, we may restrict attention to function that are analytic in the open unit disc @xmath2 but not in the closed unit disc @xmath3 .
what our analysis a priori excludes are thus : @xmath4 entire functions ; @xmath5 purely divergent series .
( for such excluded cases , either the saddle point method or ad hoc manipulations of divergent series are often instrumental in gaining access to coefficients @xcite . )
furthermore we restrict attention to functions that remain of moderate growth near the unit circle in the following sense .
[ order - def ] a function @xmath6 analytic in the open unit disc @xmath2 is said to be of _ global order _
@xmath7 if @xmath8 that is , there exists an absolute constant @xmath9 such that @xmath10 for all @xmath11 satisfying @xmath12 .
this definition typically excludes the partition generating function @xmath13 which is of infinite order and to which the saddle point method ( as well as a good deal more ) is applicable @xcite .
in contrast , a function like @xmath14 is of global order @xmath15 , while @xmath16 are of global order @xmath17 .
we observe , though we do not make use of the fact , that a function @xmath6 of global order @xmath7 has coefficients that satisfy @xmath18f(z)=o(n^{-a})$ ] .
the proof results from trivial bounds applied to cauchy s integral form @xmath19f(z)=\frac{1}{2i\pi}\int_c f(z)\ , \frac{dz}{z^{n+1}},\ ] ] upon integrating along the contour @xmath20 : @xmath21 .
( in @xcite , braaksma and stark present an interesting discussion leading to refined estimates of the @xmath22 bound . )
what we address here is the asymptotic analysis of functions whose local behaviour at designated points involves a combination of logarithms and powers ( of possibly fractional exponent ) . for the sake of notational simplicity ,
we write @xmath23 simplifying the theory to what is needed here , we set : [ logpow - def ] a _ log - power function at @xmath24 _ is a finite sum of the form @xmath25 where @xmath26 and each @xmath27 is a polynomial .
a _ log - power function at a finite set of points _
@xmath28 , is a finite sum @xmath29 where each @xmath30 is a log - power function at 1 . in what follows
, we shall only need to consider the case where the @xmath31 lie on the unit disc : @xmath32 .
it has been known for a long time ( see , e.g. , jungen s 1931 paper , ref .
@xcite , and @xcite for wide extensions ) that the coefficient of index @xmath0 in a log - power function admits a full asymptotic expansion in descending powers of @xmath0 . [ basic - lem ] the expansion of the coefficient of a log - power function is computable by the two rules : @xmath33(1-z)^\alpha & \sim & \ds \frac{n^{-\alpha-1}}{\gamma(-\alpha ) } + \frac{\alpha(\alpha+1 ) n^{-\alpha-2}}{\gamma(-\alpha ) } + \cdots \\ \ds [ z^n](1-z)^\alpha \l(z)^k & = & \ds(-1)^k \frac{\partial^k}{\partial \alpha^k } \left([z^n](1-z)^\alpha\right ) \\ & \sim & \ds ( -1)^k \frac{\partial^k}{\partial \alpha^k } \left[\frac{n^{-\alpha-1}}{\gamma(-\alpha ) } + \frac{\alpha(\alpha+1 ) n^{-\alpha-2}}{\gamma(-\alpha ) } + \cdots\right ] .
\end{array}\ ] ] the general shape of the expansion is thus @xmath34(1-z)^{\alpha}l(z)^k & \ds\mathop{\sim}_{n\to+\infty } & \ds \frac{1}{\gamma(-\alpha)}n^{-\alpha-1}(\log n)^k & ( \alpha\not\in{{\bbb z}}_{\ge0 } ) \\ { } [ z^n](1-z)^{r}l(z)^k & \ds\mathop{\sim}_{n\to+\infty } & \ds ( -1)^r k(r ! ) n^{-r-1}(\log n)^{k-1 } & ( r\in{{\bbb z}}_{\ge0 } , k\in{{\bbb z}}_{\ge1 } ) .
\end{array}\ ] ] in the last case , the term involving @xmath35 disappears as its coefficient is @xmath36 .
in essence , smaller functions at a singularity have asymptotically smaller coefficients and logarithmic factors in a function are reflected by logarithmic terms in the coefficients expansion ; for instance ,
@xmath37\frac{\l(z)}{\sqrt{1-z } } & \sim & \ds \frac{\log n + \gamma+2\log 2}{\sqrt{\pi n}}- \frac{\log n + \gamma+2\log 2}{8\sqrt{\pi n^3}}+\cdots\\ \ds[z^n](1-z)\l(z)^2 & \sim & \ds -\frac{2}{n^2}(\log n+\gamma-1)-\frac{1}{n^3 } ( 2\log n+2\gamma-5)+\cdots\ , . \end{array}\ ] ] when supplemented by the rule @xmath38\sigma\left(\frac{z}{\zeta}\right)=\zeta^{-n}[z^n]\sigma(z),\ ] ] lemma [ basic - lem ] makes it effectively possible to determine the asymptotic behaviour of coefficients of all log - power functions .
once the coefficients of functions in some basic scale are known , there remains to translate error terms .
precisely , we consider in this article functions of the form @xmath39 and need conditions that enable us to estimate the coefficients of the error term @xmath40 .
two conditions are classically available : one based on smoothness ( i.e. , differentiability ) is summarized here , following classical authors ( e.g. , @xcite ) ; the other based on growth conditions _ and _ analytic continuation is discussed in the next subsection .
[ smooth - def ] let @xmath41 be analytic in @xmath12 and @xmath42 be a nonnegative integer .
the function @xmath41 is said to be _
@xmath43smooth is said to be _
weakly @xmath44smooth _ if it admits a continuous extension to the closed unit disc @xmath45 and the function @xmath46 is @xmath42 times continuously differentiable .
this weaker notion suffices for lemmas [ darboux - lem ] , [ dar - lem ] < and theorem [ basic - thm ] . ] on the unit disc _ ( or of class @xmath47 ) if , for all @xmath48 , its @xmath49th derivative @xmath50 defined for @xmath12 admits a continuous extension on @xmath45 .
for instance , a function of the form @xmath51 for some @xmath52 and @xmath53 , is @xmath47-smooth ( both in the standard sense and in the strong sense ) .
conversely , the fact that smoother functions have asymptotically smaller coefficients lies at the heart of darboux s method .
[ darboux - lem ] if @xmath41 is @xmath43smooth , then @xmath38 h(z)=o(n^{-s}).\ ] ] one has , by cauchy s coefficient formula and continuity of @xmath41 : @xmath38h(z)=\frac{1}{2\pi}\int_{-\pi}^\pi h(e^{i\theta } ) e^{-ni\theta } \ , d\theta.\ ] ] when @xmath54 , the statement results directly from the riemann - lebesgue theorem @xcite . when @xmath55 , the estimate results from @xmath42 successive integrations by parts followed by the riemann - lebesgue argument .
see olver s book @xcite for a neat discussion .
[ smoothsa - def ] a function @xmath56 analytic in the open unit disc @xmath2 is said to admit a _ log - power expansion of class @xmath57 _ if there exist a finite set of points @xmath28 on the unit circle @xmath58 and a log - power function @xmath59 at the set of points @xmath60 such that @xmath61 is @xmath57smooth on the unit circle .
[ dar - lem ] if @xmath56 admits a log - power expansion of class @xmath62 with @xmath59 an associated log - power function , its coefficients satisfy @xmath38q(z)=[z^n]\sigma(z)+o\left(n^{-t}\right).\ ] ] one has @xmath63 , with @xmath64 being @xmath62 smooth .
the coefficients of @xmath64 are estimated by lemma [ darboux - lem ] .
consider for instance @xmath65 both are of global order @xmath66 in the sense of definition [ order - def ] . by making use of the analytic expansion of @xmath67 at 1 ,
one finds @xmath68 where @xmath69 , which is of the order of @xmath70 as @xmath71 , is @xmath72-smooth .
the sum of the first two terms ( in parentheses ) constitutes @xmath59 , in this case with @xmath73 .
similarly , for @xmath74 , by making use of expansions at the elements of @xmath75 , one finds @xmath76 where @xmath77 is also @xmath72smooth .
accordingly , we find : @xmath78q_1(z)=e\frac{1}{\sqrt{\pi n}}+o\left(\frac{1}{n}\right ) , \qquad [ z^n]q_2(z)=\frac{e\sqrt{2}}{\sqrt{\pi n}}+o\left(\frac{1}{n}\right).\ ] ] the next term in the asymptotic expansion of @xmath18q_2 $ ] involves a linear combination of @xmath79 and @xmath80 , where the latter term reflects the singularity at @xmath81 .
such calculations are typical of what we shall encounter later .
what we refer to as singularity analysis is a technology developed by flajolet and odlyzko @xcite , with further additions to be found in @xcite .
it applies to a function with a finite number of singularities on the boundary of its disc of convergence .
our description closely follows chapter vi of the latest edition of _ analytic combinatorics _ @xcite .
singularity analysis theory adds to lemma [ basic - lem ] the theorem that , under conditions of analytic continuation , @xmath82- and @xmath83-error terms can be similarly transferred to coefficients .
@xmath84-domain _ associated to two parameters @xmath85 ( the radius ) and @xmath86 ( the angle ) by @xmath87 where @xmath88 denotes the argument of @xmath89 taken here in the interval @xmath90 . by definition a @xmath84-domain properly contains the unit disc , since @xmath91 .
( details of the values of @xmath92 are immaterial as long as @xmath85 and @xmath93 . )
the following definition is in a way the counterpart of smoothness ( definition [ smoothsa - def ] ) for singularity analysis of functions with isolated singularities .
[ ordersa - def ]
let @xmath41 be analytic in @xmath12 and have isolated singularities on the unit circle at @xmath28 .
let @xmath94 be a real number .
the function @xmath41 is said to admit a _ log - power expansion of type @xmath95 _ ( relative to @xmath60 ) if the following two conditions are satisfied : * the function @xmath41 is analytically continuable to an indented domain @xmath96 , with @xmath84 some @xmath84-domain .
* there exists a log - power function @xmath97 such that , for each @xmath98 , one has @xmath99 as @xmath100 in @xmath101 .
observe that @xmath59 is _ a priori _ uniquely determined only up to @xmath102 terms .
the minimal function ( with respect to the number of monomials ) satisfying is called the _ singular part _ of @xmath41 ( up to @xmath95 terms ) .
a basic result of singularity analysis theory enables us to extract coefficients of functions that admit of such expansions .
[ sa - lem ] let @xmath28 be a finite set of points on the unit circle , and let @xmath103 be a function that admits a log - power expansion of type @xmath104 relative to @xmath60 , with singular part @xmath59 .
then , the coefficients of @xmath105 satisfy @xmath106p(z)=[z^n]\sigma(z)+o\left(n^{-t-1}\right).\ ] ] the proof of lemma [ sa - lem ] starts from cauchy s integral formula and makes use of the contour @xmath20 that lies at distance @xmath107 of the boundary of the analyticity domain , @xmath108 .
see @xcite for details . for future reference
( see especially section [ discyc - sec ] ) , we gather here facts relative to the _ polylogarithm _ function @xmath109 , which is defined for any @xmath110 by @xmath111 one has in particular @xmath112 in the most basic applications , one encounters polylogarithms of integer index , but in this paper ( see the example of dissimilar forests in section [ hodgepodge - sec ] ) , the more general case of a real index @xmath113 is also needed .
for any index @xmath110 , the polylogarithm @xmath109 is analytically continuable to the slit plane @xmath114 . if @xmath115 , the singular expansion of @xmath116 near the singularity @xmath117 is given by @xmath118 for @xmath113 not an integer , the singular expansion of @xmath109 is @xmath119 the representations are given as a _ composition _ of two explicit series .
the expansions involve both the harmonic number @xmath120 and the riemann zeta function @xmath121 defined by @xmath122 ( @xmath121 , originally defined in the half - plan @xmath123 , is analytically continuable to @xmath124 by virtue of its classical functional equation ) .
first in the case of an integer index @xmath125 , since @xmath116 is an iterated integral of @xmath126 , it is analytically continuable to the complex plane slit along the ray @xmath127 . by this device
, its expansion at the singularity @xmath117 can be determined , resulting in .
( the representation in is in fact exact and not merely asymptotic .
it has been obtained by zagier and cohen in @xcite , and is known to the symbolic manipulation system maple . ) for @xmath113 not an integer , analytic continuation derives from a lindelf integral representation discussed by ford in @xcite .
the singular expansion , valid in the slit plane , was established in @xcite to which we refer for details . in the sequel
, we also make use of smoothness properties of polylogarithms . clearly , @xmath128 is @xmath129smooth in the sense of definition [ smooth - def ] .
a simple computation of coefficients shows that any sum @xmath130\ ] ] with @xmath131 polynomially bounded in @xmath132 , is @xmath129smooth .
many similar sums are encountered later , starting with those in equations and .
the heart of the matter is the treatment of functions analytic in the open unit disc that can , at least partially , be `` de - singularized '' by means of log - power functions . our first theorem , which essentially relies on the darboux technology , serves as a stepping stone towards the proof of our main statement , theorem [ hybrid2-thm ] below .
[ basic - thm ] let @xmath6 be analytic in the open unit disc @xmath2 , of global order @xmath133 , and such that it admits a factorization @xmath134 , with @xmath135 analytic in @xmath2 .
assume the following conditions on @xmath136 and @xmath137 , relative to a finite set of points @xmath28 on the unit circle : * the `` darboux factor '' @xmath56 is @xmath43smooth on the unit circle ( @xmath138 ) .
* the `` singular factor '' @xmath103 admits , for some nonnegative integer @xmath94 , a log - power expansion relative to @xmath60 , @xmath139 ( with @xmath140 the log - power function and @xmath64 the smooth term ) , that is of class @xmath57 .
assume also the inequality ( with @xmath141 the integer part function ) : * @xmath142 , where @xmath143 let @xmath144 . if @xmath145 denotes the hermite interpolation polynomial such that all its derivatives of order @xmath146 coincide with those of @xmath137 at each of the points @xmath147 , one has @xmath148f(z)=[z^n]\left(\tilde p(z)\cdot h(z)\right)+o(n^{-u_0}).\ ] ] since @xmath149 is itself a log - power function , the asymptotic form of its coefficients is explicitly provided by lemma [ basic - lem ] .
let @xmath150 be a positive integer whose precise value will be adjusted at the end of the proof .
first , we decompose @xmath137 as @xmath151 where @xmath152 is the polynomial of minimal degree such that all its derivatives of order @xmath153 at each of the points @xmath154 coincide with those of @xmath137 : @xmath155 the classical process of hermite interpolation @xcite produces such a polynomial , whose degree is at most @xmath156 .
since @xmath152 is @xmath157smooth , the quantity @xmath158 is @xmath159smooth .
this function @xmath160 is also `` flat '' , in the sense that it has a contact of high order with 0 at each of the points @xmath31 .
we now operate with the decomposition @xmath161 and proceed to examine the coefficient of @xmath162 in each term .
_ the product @xmath163 .
_ since @xmath140 is a log - power function and @xmath164 a polynomial , the coefficient of @xmath162 in the product admits , by lemma [ basic - lem ] , a complete descending expansion with terms in the scale @xmath165 , which we write concisely as @xmath166\tilde p \cdot \bar q\in \left\ { n^{-\beta}(\log n)^k~\bigm|~k\in{{\bbb z}}_{\ge0},~\beta\in { { \bbb r}}\right\}.\ ] ] _ the product @xmath167 .
_ this is where the hermite interpolation polynomial @xmath152 plays its part . from the construction of @xmath152 , there results that @xmath158 has all its derivatives of order @xmath153 vanishing at each of the points @xmath154 .
this guarantees the existence of a factorization @xmath168 where @xmath169 is now @xmath170smooth ( division decreases the degree of smoothness ) . then , in the factorization @xmath171 the quantity @xmath140 is , near a @xmath31 , of order at most @xmath172 ( with @xmath173 the global order of @xmath174 ) .
thus , @xmath175 is at least @xmath176smooth , with @xmath177 . since @xmath178 , darboux s method ( lemma [ dar - lem ] ) yields @xmath179\widetilde p \cdot s = o\left(n^{-u(c)}\right ) , \qquad u(c):=\min(\lfloor c+a\rfloor , s - c).\ ] ] _ the product @xmath180 .
_ this quantity is of class @xmath181 by elementary product rules .
hence , by darboux s method , @xmath182 r\cdot \bar q = o\left(n^{-t}\right).\ ] ] _ the product @xmath183 .
_ this product is of class @xmath184 and , by darboux s method again , @xmath185 r\cdot s = o\left(n^{-\min(s , t)}\right).\ ] ] it now only remains to collect the effect of the various error terms of , , and in the decomposition : @xmath38f = \left([z^n ] \tilde p \cdot \bar q \right ) + o(n^{-u(c)})+o(n^{-t})+o(n^{-\min(s , t)}).\ ] ] given the condition @xmath142 in @xmath186 , the last two terms are @xmath187 . a choice , which maximizes @xmath188 ( as defined in ) and suffices for our purposes , is @xmath189 the statement then results from the choice of @xmath190 , as well as @xmath191 and @xmath192 , the corresponding hermite interpolation polynomial .
theorem [ basic - thm ] is largely to be regarded as an existence result : due to the factorization and the presence of a hermite interpolation polynomial , it is not well suited for effectively deriving asymptotic expansions . in this subsection
, we develop the hybrid method _ per se _ , which makes it possible to operate directly with a small number of radial expansions of the function whose coefficients are to be estimated .
let @xmath6 be analytic in the open unit disc .
for @xmath193 a point on the unit circle , we define the _ radial expansion _ of @xmath174 at @xmath193 with order @xmath194 as the smallest ( in terms of the number of monomials ) log - power function @xmath195 at @xmath193 , provided it exists , such that @xmath196 when @xmath197 and @xmath132 tends to @xmath198 .
the quantity @xmath195 is written @xmath199 the interest of radial expansions is to a large extent a computational one , as these are often accessible via common methods of asymptotic analysis while various series rearrangements from within the unit circle are granted by analyticity .
in contrast , the task of estimating directly a function @xmath6 as @xmath200 _ on the unit circle _ may be technically more demanding .
our main theorem is accordingly expressed in terms of such radial expansions and , after the necessary conditions on the generating function have been verified , it provides an _ algorithm _ ( equation ) for the determination of the asymptotic form of coefficients .
[ hybrid2-thm ] let @xmath6 be analytic in the open unit disc @xmath2 , of finite global order @xmath201 , and such that it admits a factorization @xmath202 , with @xmath135 analytic in @xmath2 .
assume the following conditions on @xmath136 and @xmath137 , relative to a finite set of points @xmath28 on the unit circle : * the `` darboux factor '' @xmath56 is @xmath203smooth on the unit circle ( @xmath138 ) .
* the `` singular factor '' @xmath103 is analytically continuable to an indented domain of the form @xmath204 .
for some non - negative real number @xmath205 , it admits , at any @xmath98 , an asymptotic expansion of the form @xmath206 where @xmath207 is a log - power function at @xmath24 .
assume also the inequality : * @xmath208 .
then @xmath174 admits a radial expansion at any @xmath209 with order @xmath210 .
the coefficients of @xmath6 satisfy : @xmath211f(z ) \quad = \quad \ds
[ z^n]{\bf a}(z)+o\left(n^{-u_0}\right ) , \\
\hbox{where } \qquad \ds { \bf a}(z):=\sum_{j=1}^m \asym(f(z),\zeta_j , u_0 ) .
\end{array}\ ] ] let us denote by @xmath212 the sum of the singular parts of @xmath136 at the points of @xmath31 .
the difference @xmath213 is @xmath214-smooth for any integer
@xmath94 satisfying @xmath215 ( in particular , we can choose @xmath216 , this by assumption @xmath217 .
the singular factor @xmath136 has thus been re - expressed as the sum of a singular part @xmath218 and a smooth part @xmath64 .
the conditions of theorem [ basic - thm ] are then precisely satisfied by the product @xmath219 , the inequality @xmath220 implying condition @xmath221 , so that one has by @xmath222f(z)=[z^n]\sigma(z)h(z)+o(n^{-u_0}),\ ] ] where @xmath145 is the hermite polynomial associated with @xmath137 that is described in the proof of theorem [ basic - thm ] and @xmath223 is given by . in order to complete the proof , there remains to verify that , in the coefficient extraction process of above , the quantity @xmath224 can be replaced by @xmath225 .
we have @xmath226\sigma(z)h(z ) = \sum_j [ z^n]\sigma_j(z/\zeta_j)h(z).\ ] ] now , near each @xmath31 , we have ( with @xmath144 according to ) @xmath227 respectively by assumption @xmath228 , by the high order contact of @xmath145 with @xmath137 due to the hermite interpolation construction , and by the global order property of @xmath6 .
there results from equation , condition @xmath229 , and the value of @xmath230 in that @xmath231 the proof , given and , is now complete .
thanks to theorem [ hybrid2-thm ] , in order to analyse the coefficients of a function @xmath174 , the following two steps are sufficient .
* establish the _ existence _ of a proper factorization @xmath134 . usually , a crude analysis is sufficient for this purpose .
* analyse _ separately _ the asymptotic character of @xmath6 as @xmath11 tends radially to a few distinguished points , those of @xmath60 .
as asserted by theorem [ hybrid2-thm ] , it then becomes possible to proceed with the analysis of the coefficients @xmath18f(z)$ ] _ as though _ the function @xmath174 satisfied the conditions of singularity analysis ( whereas in general @xmath6 admits the unit circle as a natural boundary ) .
manstaviius @xcite develops an alternative approach that requires conditions on generating functions in the disc of convergence , but only weak smoothness on the circumference .
his results are however not clearly adapted to deriving symptotic expansions beyond the main terms .
the function @xmath232 has been studied by greene and knuth @xcite , in relation to a problem relative to factorization of polynomials over finite fields that we treat later . as is readily recognized from first principles of combinatorial analysis @xcite , the coefficient @xmath18f(z)$ ]
represents _ the probability that , in a random permutation of size @xmath0 , all cycle lengths are distinct .
_ one has @xmath233 and the coefficients constitute the sequence * eis : a007838*. in @xcite , the authors devote some seven pages ( pp .
5258 ) to the derivation of the estimate ( @xmath234 is euler s constant ) @xmath235f(z ) = e^{-\gamma}+\frac{e^{-\gamma}}{n}+o\left(\frac{\log n}{n^2}\right),\ ] ] starting with a tauberian argument and repeatedly using bootstrapping .
in our treatment below , we recycle some of their calculations , though our asymptotic technology is fundamentally different . _ global order . _
the first task in our perspective is to determine the global order of @xmath6 .
the following chain of calculations , @xmath236 shows @xmath6 to be of global order @xmath237 .
it is based on the usual introduction of convergence factors , the exp log transformation ( @xmath238 ) , and finally the logarithmic expansion .
note that this preliminary determination of global order only gives the useless bound @xmath239 .
actually , from the infinite product expression of the gamma function @xcite ( or from a direct calculation , as in @xcite ) , there results that @xmath240 hence , from the second line of ( [ chain ] ) , @xmath241 which is compatible with , but far from sufficient to imply it . _
the hybrid method .
_ given the last line of , which is re - expressed in terms of polylogarithms as @xmath242 the right factorization of @xmath6 is obtained transparently .
define @xmath243 so that @xmath244 clearly @xmath245 is @xmath43smooth , and so is @xmath246 given usual rules of differentiation .
thus @xmath247 is our darboux factor .
the first factor @xmath248 satisfies the condition of theorem [ hybrid2-thm ] : it is the singular factor and it can be expanded to any order @xmath94 of smallness .
consequently , the hybrid method is applicable and can provide an asymptotic expansion of @xmath18f(z)$ ] to any predetermined degree of accuracy . _
the nature of the full expansion .
_ given the existence of factorizations of type with an arbitrary degree of smoothness ( for @xmath249 ) and smallness ( for @xmath250 ) , it is possible to organize the calculations as follows : take the primitive roots of unity in sequence , for orders @xmath251 .
given such a root @xmath252 of order @xmath253 , each radial restriction admits a full asymptotic expansion in descending powers of @xmath254 tempered by polynomials in @xmath255 .
such an expansion can be translated formally into a full expansion in powers of @xmath256 tempered by polynomials in @xmath257 and multiplied by @xmath258 .
all the terms collected in this way are bound to occur in the asymptotic expansion of @xmath259f(z)$ ] . for the sequel
, it proves convenient first to adjust the expansion , by taking out the @xmath260 factor .
we find , by the same techniques as in @xmath261\right),\ ] ] and start with the expansion as @xmath262 , then consider in turn @xmath81 , @xmath263 ( with @xmath264 ) , and finally @xmath265 , a primitive @xmath253th root of unity . _ the expansion at @xmath117 .
_ calculations simplify a bit if we set @xmath266 as in . by summing the singular expansions of polylogarithms ,
one arrives at an asymptotic expansion as @xmath267 of the form : @xmath268 there , the first two terms in the exponential correspond to summing the special terms in the singular expansions of polylogarithms : @xmath269 the last three terms inside the exponential of ( [ main1 ] ) arise from summation over values of @xmath270 of the regular part of , namely , @xmath271 upon distinguishing between the three cases : @xmath272 ( giving rise to @xmath273 ) , @xmath274 ( giving @xmath275 ) , and @xmath276 ( giving @xmath277 ) , and exchanging the order of summations .
the calculation of @xmath275 and @xmath273 is immediate .
first @xmath278 and since @xmath193 values at negative integers are rational numbers , the expansion of @xmath273 involves only rational coefficients .
next , one finds @xmath279 a variant of the lambert and cayley functions .
finally , the function @xmath277 has coefficients _ a priori _ given by sums like in , but with the summation extending to @xmath280 : @xmath281 each infinite sum in the expansion of @xmath282 is expressible in finite form : it suffices to start from the known expansion of @xmath283 at @xmath54 , which gives ( @xmath284 is the logarithmic derivative of the gamma function ) @xmath285 and differentiate an arbitrary number of times with respect to @xmath42 , then finally set @xmath286 .
one finds for instance , in this way , @xmath287 equation provides a complete algorithm for expanding @xmath6 as @xmath288 .
the first few terms found are @xmath289 then , an application of theorem [ hybrid2-thm ] yields the terms in the asymptotic expansion of @xmath290 arising from the singularity @xmath117 : @xmath291}=e^{-\gamma}+\frac{e^{-\gamma}}{n}+\frac{e^{-\gamma}}{n^2 } ( -\log n+c_{2,0})+\frac{e^{-\gamma}}{n^3}(\log^2 n+c_{3,1}\log n+c_{3,0})+\cdots\,\ ] ] where @xmath292 from preceding considerations , the coefficients all lie in the ring generated by @xmath293 , @xmath234 , @xmath294 , and @xmath295 .
the terms given in provide quite a good approximation .
figure [ error - fig ] displays a comparison between @xmath290 and its asymptotic approximation @xmath296}$ ] , up to terms of order @xmath297 .
we find @xmath298}\left(1+\frac{r_n}{n^3}\right),\qquad bound to remain valid for all @xmath299 .
_ expansions at @xmath81 and at @xmath263 .
_ we shall content ourselves with brief indications on the shape of the corresponding singular expansions .
note that figure [ error - fig ] clearly indicates the presence of a term of the form @xmath300 in the full asymptotic expansion of @xmath290 , where @xmath301 is a function with period 6 .
this motivates an investigation of the behaviour of @xmath6 near primitive square and cube roots of unity .
start with @xmath302 .
the definition of @xmath6 implies that @xmath303 ( the infinite product telescopes ) .
set @xmath304 and restrict attention to the type of the expansions at @xmath81 .
only half of the polylogarithms in are singular , so that the expansion at @xmath81 is of the _ type _
@xmath305 there , we have replaced all unspecified coefficients by the constant 1 for readability .
this singular form results in a contribution to the asymptotic form of @xmath290 : @xmath306}=(-1)^n\left(\frac{d_3}{n^3}+\frac{d_{4,1}\log n+d_{4,0}}{n^4}+\cdots\right).\ ] ] ( compared to roots of unity of higher order , the case @xmath81 is special , because of the factor @xmath260 explicitly present in the definition of @xmath6 . )
a simple calculation shows that @xmath307 , so that @xmath306}\sim 2\frac{(-1)^n}{n^3}.\ ] ] next , let @xmath308 and set @xmath309 .
the type of the expansion at @xmath310 is @xmath311 since now every third polylogarithm is singular at @xmath310 .
this induces a contribution of the form @xmath312}=\omega^{-n}\left(\frac{e_3}{n^3}+\frac{e_{4,1}\log n+e_{4,0}}{n^4}+\cdots\right),\ ] ] arising from @xmath310 , and similarly , for a conjugate contribution arising from @xmath313 .
another simple calculation shows that @xmath314 and leaves us with the task of estimating @xmath315 .
the use of the formula , @xmath316 a mere avatar of the product formula for the gamma function , yields then easily @xmath317 the fluctuations of period 6 evidenced by figure [ error - fig ] are thus fully explained : one has @xmath306}+\hat
f_n^{[\omega]}+\hat f_n^{[\omega^2]}=\frac{\omega(n)}{n^3}+ o\left(\frac{\log n}{n^4 } \right),\ ] ] where the periodic function @xmath318 is @xmath319 designates a real part ) @xmath320 _ expansions at @xmath265 , a primitive @xmath253th root of unity .
_ let @xmath321 and @xmath322 .
the expansion of @xmath6 is now of the type @xmath323 where @xmath324 corresponds to the singular term in @xmath325 .
consequently , fluctuations start appearing at the level of terms of order @xmath326 in the asymptotic expansion of @xmath290 as @xmath327 .
the value of @xmath328 is expressible in terms of gamma values at algebraic points , as we have seen when determining @xmath315 .
the coefficients in the expansion also involve values of the @xmath329-function ( @xmath330 ) and its derivatives at rational points , which include @xmath193 values as particular cases .
[ grkn - prop ] the probability that a permutation is made of cycles of distinct lengths admits a full asymptotic expansion of the form @xmath331 \\ & & \ds { } + \sum_{r\ge 4}\frac{p_r(n)}{n^r } , \end{array}\ ] ] with @xmath332 and @xmath333 as given by .
there , @xmath334 is a polynomial of exact degree @xmath335 in @xmath336 with coefficients that are periodic functions of @xmath0 with period @xmath337 .
figure [ error4-fig ] displays the error of the approximation obtained by incorporating all terms till order @xmath297 included in the asymptotic expansion of @xmath290 .
fluctuations of period 12 ( due to the additional presence of @xmath338 ) start making an appearance .
we now examine several combinatorial problems related to permutations , polynomials over finite fields , and trees that are amenable to the hybrid method .
the detailed treatment of permutations with distinct cycle lengths can serve as a beacon for the analysis of similar infinite product generating functions , and accordingly our presentation of each example will be quite succinct . in the examples that follow ,
the function @xmath174 whose coefficients are to be analysed is such that there is an increasing family of sets @xmath339 ( ordered by inclusion , and with elements being roots of unity ) , attached to a collection of asymptotic expansions having smaller and smaller error terms . in that case , a full asymptotic expansion is available for the coefficients of @xmath174 .
the general asymptotic shape of @xmath18f$ ] involves standard terms of the form @xmath340 modulated by complex exponentials , since @xmath18\sigma(z/\zeta)=\zeta^{-n}\sigma_n$ ] .
we formalize this notion by a definition .
a sequence @xmath341 is said to admit a _ full asymptotic expansion with oscillating coefficients _ if it is of the form @xmath342 where the exponents @xmath343 increase to @xmath344 and each @xmath334 is a polynomial in @xmath336 whose coefficients are periodic functions of @xmath0 . in a small way
, proposition [ grkn - prop ] and the forthcoming statements , propositions [ pi2-prop][for - prop ] , can be regarded as analogues , in the realm of functions of slow growth near the unit circle , of the hardy - ramanujan - rademacher analysis @xcite of partition generating functions , the latter exhibiting a very fast growth ( being of infinite order ) as @xmath345 .
the problem of determining the number of permutations that are squares or equivalently `` have a square root '' is a classical one of combinatorial analysis : see wilf s vivid account in @xcite .
the problem admits an obvious generalization .
we shall let @xmath346 be the exponential generating function of permutations that are @xmath1th powers or , if one prefers , admit an @xmath1th root . for the generating function @xmath347 , we follow wilf s account . upon squaring a permutation @xmath348 ,
each cycle of even length of @xmath348 falls apart into two cycles of half the length , while an odd cycle gives rise to a cycle of the same length .
hence , if @xmath349 has a square root , then the number of cycles it has of each even length must be even . by general principles of combinatorial analysis ,
the exponential generating function of the number @xmath350 of permutations of @xmath0 elements that have square roots satisfies @xmath351 the series starts as @xmath352 its coefficients being * eis : a003483 * , with the quantity @xmath18\pi_2(z)$ ] representing the probability that a permutation is a square .
the gf of is given by bender @xcite who attributes it to blum @xcite .
it is interesting to note that bender mentions the following estimate from @xcite , @xmath353\pi_{2}(z ) \sim \frac{2}{\sqrt{\pi n } } e^g,\ ] ] and derives it by an application of the tauberian theorem is monotonically nonincreasing in @xmath0 .
] of hardy , littlewood , and karamata . accordingly
, no error terms are available , given the nonconstructive character of classical tauberian theory .
we state : [ pi2-prop ] the probability that a random permutation of size @xmath0 has a square - root admits a full asymptotic expansion with oscillating coefficients .
in particular , it satisfies @xmath354\pi_{2}(z ) \sim \sqrt{\frac{2}{\pi n } } e^g \left[1-\frac{\log n}{n}+\frac{c_3+(-1)^n}{4n}\right ] -\frac{2e^g(-1)^{\lfloor n/2\rfloor}}{n^2 } + o\left(\frac{\log n}{n^{5/2}}\right),\ ] ] where @xmath355 first a rough analysis suffices to see that @xmath356 in order to refine these expansions , introduce the normalized tangent numbers by @xmath357 the usual exp - log reorganization of the @xmath347 series yields @xmath358 in passing , this provides for @xmath359 , the fast convergent series @xmath360 on which the numerical estimate of is based .
next , take out the @xmath361 factor , leading to @xmath362\right).\ ] ] at @xmath117 , the largest singular term in the exponential arises from @xmath363 , the contributions from the other polylogarithms being of smaller order @xmath364 where , here and later , @xmath365 designates a computable constant that we leave unspecified for the purpose of readability .
similarly , at @xmath81 , we find @xmath366 at @xmath367 ( hence at @xmath368 , by conjugacy ) , we have @xmath369 with @xmath370 an unspecified polynomial that does not leave a trace in the coefficients expansion . the singular contribution to @xmath347 arising from any root of unity @xmath193 of order @xmath371
is at most @xmath372 , which translates to an @xmath373 term .
the proof of is then completed upon making @xmath365 in the expansion at 1 explicit , which introduces the new constants @xmath374 .
existence of the full expansions finally follows from the usual analysis of polylogarithms of powers , taken at roots of unity . like the previous one
, this problem is briefly mentioned in bender s survey @xcite .
we follow again wilf s exposition @xcite . for a pair @xmath375 of positive integers
, we define @xmath376 to be @xmath377 ( thus , @xmath376 gathers from the prime decomposition of @xmath1 all the factors that involve a prime divisor of @xmath253 . )
the characterization of permutations that are @xmath1th powers then generalizes @xcite : a permutation has an @xmath1th root if and only if , for each @xmath378 , it is true that the number of cycles of length @xmath253 is a multiple of @xmath376 .
this observation leads to an expression for the corresponding generating function . indeed , define the `` sectioned exponential '' , @xmath379 so that @xmath380 and @xmath381 .
the exponential generating function of permutations that are @xmath1th powers is then @xmath382 the generating function of has been investigated by pouyanne @xcite , whose paper provides the first order asymptotic estimate of @xmath18\pi_m$ ] .
there is a fundamental factorization , @xmath383 where @xmath384 gathers from the product the numbers @xmath253 that are relatively prime to @xmath1 and @xmath385 gathers the rest .
the factor @xmath384 is found by series rearrangements to be an algebraic function expressible by radicals , @xmath386 with @xmath387 the mbius function . for @xmath388 ( square permutations ) , this is the ubiquitous prefactor @xmath389 . for @xmath390 , for instance , the prefactor becomes @xmath391 the factor @xmath392 is a transcendental function that admits the unit circle as a natural boundary @xcite .
it is expressible as an infinite product of sectioned exponentials : @xmath393 for @xmath388 , this is the infinite product of hyperbolic cosines . for @xmath390 , one has @xmath394 these singular factors can be analysed just like in the case of @xmath347 by an exp - log transformation .
one first observes that the limit value @xmath395 is well defined , since the infinite product converges at least as fast as @xmath396 .
it is seen next that singularities are at roots of unity , and the radial expansions can be computed in the usual way from the polylogarithmic expansion .
we can now state a ( somewhat minor ) improvement over @xcite : the probability that a random permutation of size @xmath0 has an @xmath1th root admits a full asymptotic expansion with oscillating coefficients .
to first asymptotic order , it satisfies @xmath38\pi_m(z)\sim \frac{\varpi_m}{n^{1-\varphi(m)/m } } , \qquad \varpi_m:= \frac{b_m(1)}{\gamma(\varphi(m)/m)}\prod_{k~|~m } k^{-\mu(k)/k},\ ] ] where @xmath397 is the euler totient function and @xmath398 in particular , when @xmath399 is a prime number , one has @xmath400\pi_p(z)\sim \frac{\varpi_p}{n^{1/p } } , \qquad \varpi_p:=\frac{p^{1/p}}{\gamma(1 - 1/p)}\prod_{\ell=1}^\infty \exp_p\left(\frac{1}{\ell p}\right).\ ] ] for instance , we find : @xmath38\pi_6(z)=b_6(1)\left [ \frac{\root6\of{12}}{\gamma(\frac13)}n^{-2/3 } + \frac{2\root12\of{12}}{\gamma(\frac16 ) } \re\left({e^{-i\frac{\pi}{3}(n+\frac14)}}\right)n^{-5/6}+ o\left(\frac1n\right)\right],\ ] ] and improves upon an early estimate of turn ( * ? ? ?
iv ) . given a permutation @xmath349 , its _ cycle type _ is the ( unordered ) multiset formed with the lengths of the cycles entering its decomposition into cycles . for a permutation of size @xmath0 , this type can be equivalently represented by a partition of the integer @xmath0 : for instance @xmath401 represents the profile of any permutation of size 25 that has three cycles of length 2 , one cycle of length 5 , and two cycles of length 7 .
the probability @xmath402 that a random permutation of size @xmath0 has profile @xmath403 is , by virtue of a well - known formula @xcite , @xmath404 corresponding to the generating function in infinitely many variables @xmath405 which is such that @xmath406\phi$ ] . in this subsection , we estimate _ the probability that two permutations of size @xmath0 taken uniformly and independently at random have the same cycle type .
_ ( each pair of permutations is taken with probability @xmath407 . ) the quantity to be estimated is thus @xmath408 where the summation ranges over all partitions @xmath409 of @xmath0 .
this problem was suggested to the authors by the reading of a short note of wilf @xcite , who estimated the probability that two permutations have the same number of cycles ( the answer to the latter question turns out to be asymptotic to @xmath410 ) .
given , it is not hard to find the generating function of the sequence @xmath411 : @xmath412 the reason being that @xmath413 the function written @xmath414 , which is obviously entire , is a variant of the bessel function @xmath415 ( see , e.g. , @xcite ) : @xmath416 . also ,
from , the expansion of @xmath417 is readily computed : one has @xmath418 where the coefficients are * eis : a087132 * ( `` sum of the squares of the sizes of conjugacy classes in the symmetric group@xmath419 '' ) .
the probability @xmath420 that two permutations of size @xmath0 have the same cycle type satisfies @xmath421 furthermore , this probability admits a full asymptotic expansion with oscillating coefficients .
we shall only sketch the analysis of the dominant asymptotic term , the rest being by now routine . from the expression of @xmath417 ,
the exp - log transformation yields @xmath422 let @xmath423h(z)$ ] , that is , @xmath424 ( the sequence @xmath425 is * eis : a002190 * , which occurs in the enumeration of certain pairs of permutations by carlitz ) .
then @xmath426 this expression ensures that the hybrid method can be applied at any order , with @xmath417 being of global order @xmath17 .
the first term of the asymptotic estimate is provided by the factorization relative to @xmath427 , namely @xmath428 , with @xmath429 since @xmath56 is clearly @xmath430 on the closed unit disk , the hybrid method applies : with the notations of theorem [ hybrid2-thm ] , one can take @xmath431 , to the effect that @xmath432 . using the algorithmic scheme of section [ discyc - sec ] ,
we find @xmath433 \\
w(z ) & \ds \mathop{=}_{z\to -1^+ } & \ds w(-1)\left[1+o((1+z)^2)\right ] , \end{array}\ ] ] with @xmath434 .
theorem [ hybrid2-thm ] then directly yields @xmath435 , and further asymptotic terms can easily be extracted .
regarding other statistics on _ pairs _ of permutations , it is well worth mentioning dixon s recent study of the probability that two randomly chosen permutations generate a transitive group @xcite . for the symmetric group @xmath436 ,
this probability is found to be asymptotic to @xmath437 and , up to exponentially smaller order terms , this expansion also gives the probability that two random permutations generate the whole symmetric group . in that case , the analytic engine is bender s theory of coefficient extraction in divergent series @xcite .
factoring polynomials in @xmath438 $ ] is a problem of interest in symbolic computation , as it has implications in the determination of partial fraction expansions and symbolic integration , for instance .
most of the existing algorithms proceed by a reduction to a few factorizations of polynomials with coefficients in some finite field @xmath439 ( with @xmath440 a prime power ) : see the books by berlekamp , knuth , or von zur gathen and gerhard @xcite .
several of the algorithms for factoring polynomials over finite fields involve what is known as the _ distinct degree factorization_. to wit , let @xmath441 $ ] be completely factored as @xmath442 where the @xmath443 are distinct irreducible polynomials .
the distinct degree factorization produces the decomposition of the square - free part of @xmath444 under the form @xmath445 we refer to the paper @xcite by three of us for details
. the distinct degree factorization of @xmath444 as defined above coincides with the complete factorization of @xmath444 if and only @xmath444 has all its irreducible factors of different degrees . for the purpose of analysis of algorithms , it is then of interest to quantify the probability of that event .
this problem has been considered independently by the authors of @xcite and by knopfmacher & warlimont in @xcite .
first , without loss of generality , we restrict attention to monic polynomials . if @xmath440 is the cardinality of the base field @xmath446 , the number of monic polynomials of degree @xmath0 is @xmath447 , with ordinary generating function
@xmath448 the unique factorization property implies that the class @xmath449 of all ( monic ) polynomials is isomorphic to the class obtained by applying the multiset construction to the class @xmath450 of ( monic ) irreducible polynomials .
this , by well known principles of combinatorial analysis @xcite translates to an equation binding the generating function @xmath414 of irreducible polynomials , namely , @xmath451 where @xmath452 is the number of monic irreducible polynomials of degree @xmath0 .
the solution is obtained by taking a logarithm of the second form and applying mbius inversion : @xmath453 and , as a consequence , the number @xmath452 admits the explicit formula ( already known to gau , see references in @xcite ) : @xmath454 in passing this approach also shows that @xmath455 where @xmath40 is analytic in @xmath456 , so that @xmath414 has a logarithmic singularity at @xmath457 .
let @xmath458 be the number of ( monic ) polynomials composed of irreducible factors of distinct degrees .
the corresponding generating function , @xmath459 satisfies @xmath460 and the probability of the event over polynomials of degree @xmath0 is @xmath461\ , d\left(zq^{-1}\right).\ ] ] it is in fact the simplification of this problem for fields of large cardinalities that originally led greene and knuth to consider the problem of enumerating permutations with distinct cycle lengths ( see @xcite for context ) .
we have : [ poly - prop ] the probability that a random polynomial of degree @xmath0 over @xmath439 has all its irreducible factors of distinct degrees satisfies the asymptotic estimate @xmath462 where the constant @xmath463 is given by @xmath464 in addition , this probability admits a full asymptotic expansion with oscillating coefficients .
we start with a rough analysis based on comparing @xmath459 and @xmath103 .
the second form of implies @xmath465 at @xmath117 , the general factor in the product satisfies , for large @xmath0 , @xmath466 given the known form of @xmath452 .
this is enough to ensure the convergence of the infinite product defining @xmath463 and proves that @xmath467 which is evidently compatible with the asserted estimate .
also , the estimate points to the fact that convergence of the infinite product giving @xmath463 is very slow , but also suggests the modified scheme ( compare with ) , @xmath468 which now exhibits geometric convergence . in order to complete the derivation of
, we apply the exp - log transformation to : @xmath469 also , we have @xmath470 . taking out the factor @xmath471 gives @xmath472 where @xmath473 is analytic in @xmath474 .
next , we observe from that @xmath475 where @xmath476 is analytic in @xmath474 .
the existence of a factorization fulfilling the conditions of theorem [ hybrid2-thm ] now follows from the usual split of the sum in the exponential of , last line : @xmath477 the first split sum is singular at roots of unity of order @xmath478 and admits a log - power expansion of type @xmath104 for any @xmath94 ; the second split sum has its @xmath0th coefficient that is of order @xmath479 , and is accordingly @xmath480 . by theorem [ hybrid2-thm ] , these considerations imply that @xmath18d(z / q)$ ] admits a full asymptotic expansion with oscillating coefficients .
it is not hard then to track the order of the term that corrects the dominant asymptotic regime : in essence , it arises from the factor @xmath481 whose coefficients are @xmath482 . a variant of this problem is also considered in @xcite : _ determine the probability that a random polynomial of degree @xmath0 is such that its square - free factorization consists of polynomials of distinct degrees .
_ the generating function is then @xmath483 by devices entirely similar to those employed in the proof of proposition [ poly - prop ] , one finds that the corresponding probability satisfies @xmath484 where the constant @xmath485 is @xmath486 the treatment of @xcite is based on a crude application of the hybrid method , that of @xcite relies on elementary bounds and coefficient manipulations , but it is not clearly applicable to derive asymptotic expansions beyond the first term .
fix a class @xmath487 of rooted trees .
an unordered forest is a multiset of trees in @xmath487 , and we let @xmath488 be the class of all forests .
a forest is said to be _ dissimilar _ if all the trees that it contains are of different sizes .
we let @xmath489 denote the class of dissimilar forests .
for instance , the collection @xmath487 of all plane trees has generating function @xmath490 a well - known fact @xcite .
the generating function of unordered forests is given by a construction analogous to : @xmath491 ( the coefficients constitute * eis : a052854 * ) .
the generating function of dissimilar forests , which is our actual object of study , is @xmath492 ( currently unlisted in @xcite ) .
the coefficients @xmath18t(z)$ ] are the famous catalan numbers of combinatorial theory , @xmath493 the asymptotic estimate being in agreement with the square - root singularity that is visible in @xmath494 .
an easy analysis of @xmath495 then shows that , as @xmath496 , @xmath497 resulting in the estimate @xmath498 [ for - prop ] the probability that a random forest of rooted plane trees having total size @xmath0 is dissimilar is asymptotic to the constant @xmath499 this probability admits a complete asymptotic expansion with oscillating coefficients .
we limit the calculation to first order asymptotics . as usual , start from the exp - log trick , which yields @xmath500 where the functions @xmath501 that play the rle of special polylogarithms are defined by @xmath502 observe that the functions @xmath503 are @xmath84-analytic , by virtue of a theorem asserting the closure of functions of singularity analysis class under hadamard products @xcite .
( the function @xmath503 is roughly comparable to @xmath504 . )
splitting the sum in according to @xmath505 and @xmath506 , for an arbitrarily large @xmath9 , yields a factorization of @xmath507 that fulfills the conditions of theorem [ hybrid2-thm ] .
hence the existence of a full expansion for coefficients is granted . for first order asymptotics
, it suffices to analyse what happens as @xmath508 .
we find , from and with the notations of @xmath509 which implies the stated estimate .
the analysis obviously extends to any simple family of trees , in the sense of meir and moon @xcite , including the case of cayley trees that are enumerated by @xmath510 .
as demonstrated by our foregoing examples , several infinite - product generating functions occurring in combinatorics are amenable to the hybrid method , despite the fact that they admit the unit circle as a natural boundary . for comparison purposes ,
figure [ comp - fig ] displays a few typical infinite - product generating functions and the general asymptotic shapes ' to represents asymptotic proportionality , that is , asymptotic equivalence up to an unspecified multiplicative constant .
] of their coefficients .
perhaps the most well - known generating function in this range is @xmath511 whose coefficients enumerate partitions into distinct summands .
in such a case , the analysis of coefficients , first performed by hardy and ramanujan , is carried out by mean of the saddle point method , in accordance with the fact that the function @xmath512 gets exponentially large at its `` main '' singularity @xmath117 .
this and similar cases are best treated by means of meinardus method @xcite .
the function @xmath513 is also fast growing near @xmath117 and is best transformed by means of convergence factors : @xmath514 this expression shows a growth similar to that of the function @xmath515 , to which the saddle point method is applicable .
all computations done , we find @xmath516{2\pi n}\right ) \qquad\hbox{with}\quad g=\prod_{n\ge1}\left(1+\frac{1}{\sqrt{n}}\right)e^{-1/\sqrt{n}+1/(2n)}.\ ] ] the next three examples of figure [ comp - fig ] are relative to functions of the form @xmath517 with @xmath518 .
these are typical cases where the hybrid method is applicable .
the function @xmath519 is exactly the one encountered when enumerating permutations with distinct cycle lengths .
the function @xmath520 provides a simplified analytic model of dissimilar forests , while @xmath521 behaves like the generating function of permutations having the same cycle type .
the general pattern is , for @xmath522 : @xmath523 at the other end of the spectrum , we find the function @xmath524 which is the exponential generating function of set partitions , all of whose blocks are of distinct sizes . clearly , this is an entire function to which the hybrid method is not applicable .
coefficients have been precisely studied by the seven authors of @xcite , with some ( but not all ) of the regimes being accessible to the saddle point method .
finally , gourdon has developed in @xcite methods leading to asymptotic expansions of a shape similar to the ones of the present paper in relation to a refinement of golomb s problem posed by knuth in ( * ? ? ?
1.3.3.23 ) .
he showed that the expected length of the longest cycle in a permutation of size @xmath0 admits an asymptotic representation which starts as @xmath525 ) @xmath526 with higher order roots of unity appearing at higher asymptotic orders .
the proof has analogies to the techniques of the present paper , with the additional need of precise expansions that describe the behaviour of the truncated logarithmic series on the unit disc .
philippe flajolet , xavier gourdon , and daniel panario , _ random polynomials and polynomial factorization _ , automata , languages , and programming ( f. meyer auf der heide and b. monien , eds . ) , lecture notes in computer science , no . 1099 , 1996 , proceedings of the 23rd icalp conference , paderborn , july 1996 . , pp .
232243 .
philippe flajolet and robert sedgewick , _ analytic combinatorics _ ,
april 2006 , chapters i ix of a book to be published by cambridge university press , 717p.+x , available electronically from p. flajolet s home page .
w. b. ford , _ studies on divergent series and summability and the asymptotic developments of functions defined by maclaurin series _
, 3rd ed . , chelsea publishing company , new york , 1960 , ( from two books originally published in 1916 and 1936 . ) .
g. h. hardy , _ ramanujan : twelve lectures on subjects suggested by his life and work _
, third ed .
, chelsea publishing company , new - york , 1978 , reprinted and corrected from the first edition , cambridge , 1940 .
a. knopfmacher , a. m. odlyzko , b. pittel , l. b. richmond , d. stark , g. szekeres , and n. c. wormald , _ the asymptotic number of set partitions with unequal block sizes _ , electronic journal of combinatorics * 6 * ( 1999 ) , no . 1 ,
r2:137 .
p. turn , _ on some connections between combinatorics and group theory _ ,
combinatorial theory and its applications , iii ( proc .
, balatonfred , 1969 ) , north - holland , amsterdam , 1970 , pp . 10551082 . | a `` hybrid method '' , dedicated to asymptotic coefficient extraction in combinatorial generating functions , is presented , which combines darboux s method and singularity analysis theory .
this hybrid method applies to functions that remain of moderate growth near the unit circle and satisfy suitable smoothness assumptions this , even in the case when the unit circle is a natural boundary .
a prime application is to coefficients of several types of infinite product generating functions , for which full asymptotic expansions ( involving periodic fluctuations at higher orders ) can be derived .
examples relative to permutations , trees , and polynomials over finite fields are treated in this way . |
as it is known , in modulation - doped ( or selectively - doped ) structures free electrons , spatially separated from ionized donors , are confined to a narrow potential well forming the so - called two - dimensional electron gas ( 2deg)@xcite . @xmath0 the @xmath1 axis , i. e. perpendicular to the 2deg plane , the confinement potential has a triangular shape , and the spectrum of the 2deg is composed of a sequence of quantum subbands.@xcite the shape of the confining potential has influence only to the carrier transitions between different subbands ; meanwhile for excitations within each subband so - called intraband excitations this effect is very small .
@xmath2 to the @xmath1 axis , i. e. in the 2deg , or @xmath3 plane , the carriers display very high electron mobilities exceeding values of 2@xmath4/(v@xmath5s ) at liquid nitrogen and 1@xmath6/(v@xmath5s ) at liquid helium temperatures.@xcite this fundamental property of these structures allowed to invent hemts
high electron mobility transistors.@xcite the carrier transport in 2deg layers at low electric fields is described by the ohm s law together with the einstein relation for the diffusion coefficient , i.e. @xmath7 , \end{aligned}\ ] ] where @xmath8 is drift velocity , @xmath9 stands for electron mobility , @xmath10 is the electric field and @xmath11 is the free carrier concentration ; @xmath12 is the electron charge , @xmath13 and @xmath14 denote boltzmann constant and lattice temperature , respectively .
this so - called `` drift - diffusion equation '' describes the quasi - classical situation when the quantization of the electronic states - induced effects are not important . on the other hand
, it does not take into account the hot - electrons phenomena when the applied in - plane electric field is strong enough to heat the electrons up to the energy exceeding their equilibrium value.@xcite the electric field in modelling , as a rule , is assumed to be uniform .
although this assumption is very convienent for calculations , however , in many cases , in order to obtain the real picture of physical processes , it is necessary to include spatial variation of the carrier concentration or / and the electric field along the channel . in particular , when it becomes comparable with the range of a mean free path . under these circumtances ,
effects such as velocity overshoot begin to predominate defining thus the performance of the device .
such an approach is very important , for instance , in understanding the behavior of devices containing asymmetric channel profile produced by a relevant doping .
more specifically , in a quarter - micron @xmath11-type silicon metal - oxide semiconductor field effect transistors ( mosfet ) with asymmetric channel profile , formed by the tilt - angle ion - implantation after gate electron formation , allows one to achieve high current drivability and hot - electron reliability.@xcite for example , in 0.1 @xmath9 m gate length asymmetric @xmath11-mosfet structures this technological innovation allows to attain higher electron velocity in comparison with conventional devices of such type.@xcite in this article , we report on experimental and theoretical investigation of the _ electric field gradient - induced effects _ due to asymmetrically in - plane shaped gaas / algaas modulation - doped structures .
we show that current voltage characteristics of such structures at low , 480 k , temperatures exhibit pronounced asymmetry .
the physics behind is attributed to a two - dimensional bigradient effect which is equivalent to the phenomenon observed earlier in a bulk asymmetricaly - shaped semiconductors@xcite .
we demonstrate that depending on the values of the in - plane electric fields and their gradients , the effect can reveal itself as a result of different distribution of accumulating two - dimensional electrons due to the asymmetrical shape of the structure , and/or the exhibition of nonlocal drift velocity which becomes pronounced in a different manner due to the presence of two different gradients of the in - plane electric field .
the paper is organized as follows . in sec .
ii we present the design of the gaas / algaas modulation - doped structures , their electrical parameters and geometry features ; we also describe briefly used measurement techniques .
section iii reports on experimental results obtained in gaas / algaas modulation - doped structures of various designs and at different lattice temperatures .
section iv is devoted to theoretical models and illustrates the concept of the electric field gradients - induced phenomenon the bigradient effect with special emphasis on manifestation of the electron drift velocity . in sec .
v possible applications of the effect for the sensing of electromagnetic radiation within ghz thz frequencies are disccused , while in sec .
vi features of three - dimensional vs. two - dimentional effect are compared . finally , conclusions are given in sec .
two types of modulation - doped structures of gaas / al@xmath15ga@xmath16as ( structure 2deg - a ) and gaas / al@xmath17ga@xmath18as ( structure 2deg - b ) were grown by molecular beam epitaxy technique .
their design parameters are given in the caption of fig . 1
. = 500 @xmath9 m ; @xmath19=250 @xmath9 m , @xmath20=50 @xmath9 m , @xmath21=100 @xmath9 m .
white color denotes active part containing 2deg which is shown schematically as a black sheet ; grey colored parts depict ohmic contact areas of length @xmath22=100 @xmath9 m . layer sequence of gaas / algaas modulation - doped structures ( from the top ) : structure 2deg - a and structure 2deg - b , respectively : 20 nm @xmath23-gaas cap layer ; 80 nm si - doped , 1@xmath24 , layer of al@xmath15ga@xmath16as and 60 nm si - doped , 2@xmath25 of al@xmath17ga@xmath18as ; undoped spacers , 45 nm al@xmath15ga@xmath16as and 10 nm al@xmath17ga@xmath18as ; 1000 nm and 600 nm of @xmath23-gaas ; smoothing superlattice twenty and six periods of 9 nm algaas/1.5 nm - gaas layers ; 0.5 @xmath9 m and 0.6 @xmath9 m layer of @xmath23-gaas ; semi - insulating substrate.,width=264,height=188 ] electron sheet density , @xmath26 , and low - field mobility , @xmath9 , at different , 300 k , 77 k and 4.2 k , temperatures for structure 2deg - a are the following : 5.5@xmath27 and @xmath28/(v@xmath5s ) , 1.9@xmath29 and @xmath30/(v@xmath5s ) , 1.9@xmath31 and 2@xmath32/(v@xmath5s ) , respectively ; while for structure 2deg - b : 5.6@xmath29 and @xmath33/(v@xmath5s ) , 4.3@xmath27 and @xmath34/(v@xmath5s ) , 4.3@xmath27 and 1@xmath32/(v@xmath5s ) , respectively .
the wafers were then processed into asymmetrically - shaped samples of different geometry which was changed varying the width @xmath35 of the neck .
we have used the optical lithography for the samples with @xmath35 = 12 @xmath9 m .
the meza with height , @xmath36 , of 2 @xmath9 m in these samples was fabricated by wet etching . in order to fabricate samples with a narrower neck ,
respectively 5 @xmath9 m , 2 @xmath9 m , and 1 @xmath9 m width , we have employed electron - beam lithography and shallow wet etching ; the meza height in this case was 300 nm .
the ohmic contacts were produced by a rapid annealing procedure of evaporated au / ge / ni compound .
the wired samples were then mounted into a two - stage closed - cycle helium cryostat , and @xmath37 characteristics were measured within 4300 k temperature range . in microwave experiments performed at 10 ghz frequency , we have used magnetrons delivering pulses of 1.55 @xmath9s duration with the repetition rate of 3540 hz ; the samples were then placed into rectangular waveguides , and the microwave radiation - induced voltage arising over the ends of the sample was measured by an oscilloscope ; the experiments were performed at room and liquid nitrogen temperatures . in the terahertz ( thz ) range , 0.5842.52 thz , the source of thz radiation was an optically - pumped molecular laser operating in the continuous wave regime ; the illumination into the sample , located in the closed - cycle helium cryostat behind the thz transparent high - pressure polythene window , was focused by spherical mirrors .
as in the microwave case , the external radiation - induced voltage arising over the ends of the sample was recorded by lock - in amplifier at a chopper frequency of 187 hz .
figure 2 shows the experimental results of @xmath37 characteristics measured in structure 2deg - a in the samples with different neck widths recorded at liquid helium temperature .
( the experimental data of the structure 2deg - b are equivalent and therefore are not presented here ) .
characteristics of the asymmetrically shaped samples with different neck width at liquid helium temperatures .
the inset depicts the measurement scheme : the contact b was always grounded , positive ( filled symbols ) and negative ( empty symbols ) voltage was applied to the contact a.,height=260 ] as one can see , the salient feature of all the measured @xmath37 characteristics is an asymmetry becoming well - procounced above 0.1 v. moreover , the character of the characteristics strongly distingtive to the reduction of the neck width : for the sample with neck width of 12 @xmath9 m two parts in the curve can be distinguished the initial rising part ( below 0.3 v ) which bends with voltage and above 0.7 v transits to the second , saturated one . with the decrease of the width @xmath35
the saturated part transforms to a boosting region with highly expressed asymmetry .
a complex behaviour is observed in the 5 @xmath9 m width sample : careful look can resolve initial rising part below 0.15 v , rudiments of saturated part within 0.150.6 v and already emerging boosting region above 0.6 v with spread influence on all the curve .
it should be pointed out that the asymmetry is strongly sensitive to the temperature variation .
this is illustrated by the experimental data for 2 @xmath9 m neck sample given in fig .
it is seen that at low temperatures the asymmetry is rather large
if the coefficient of asymmetry of the @xmath37 characteristics is defined as @xmath38 at liquid helium temperature under 0.8 v it is 0.506 , at 80 k its value is about 0.369 , meanwhile at room temperature the asymmetry is nearly invisible . to resolve the effect ,
the experimental data should be multiplied by a factor of one hundred .
characteristics of the asymmetrically shaped sample vs. temperature . the neck width is 2 @xmath9 m .
inset shows the change of asymmetry varying the lattice temperature .
measurement scheme is the same as in the previous plot.,height=260 ]
to understand the physics behind the observed phenomenon , we have followed the same ideology as in the bulk case of asymmetrically - shaped structures : we explore an asymmetrically shaped two - dimensional structure by a solving a coupled system of one - dimensional poisson and current - flow equations neglecting quantum effects@xcite : @xmath39=- \frac{|e|}{\epsilon}\left(n(x)-n_{d}\right)},\\ i=\{|e|n(x)\mu(e)e(x)+|e|d(e(x))\frac{d}{dx}[n(x)]\}s(x),\end{aligned}\ ] ] where @xmath40 denotes the position along the symmetry axis of the asymmetrically - shaped sample , @xmath41 is the cross - section area of the sample , @xmath42 is the electron concentration , @xmath43 is the donor concentration ( used here as a free fit parameter ) , @xmath44 denotes the electric field strength , @xmath45 is the dielectric constant of the sample , @xmath46 denotes the diffusion coefficient , and @xmath47 is the electron mobility .
the diffusion coefficient may be set to a constant value as its dependence on the electric field has no appreciable influence on the results .
as it is known , mobility @xmath47 changes with the electric field:@xcite @xmath48 $ ] , where @xmath49 , and @xmath50 is the cut - off velocity equal to @xmath51 cm / s . for a given value of the current @xmath52 , one can employ an iterative scheme to solve self - consistently the system and calculate the voltage .
results of the electric field and relative carrier density calculations are given in fig .
the model calculations do give two different gradients of the electric field induced by the asymmetrical shape of the structure depicted in fig .
one must note further that the distribution depends on the current direction , and this difference is appreciable only within several microns from the neck : the electric field peaks are sharper in the case of the reverse current , when the a - contact of the sample is under a positive bias .
m neck width under @xmath52 = 0.03 ma at liquid helium temperature .
, height=257 ] hence , the in - plane geometrical asymmetry induces two different gradients of the electric field in the sample . as a consequence , different spatial accumulation of two - dimensional electrons in the vicinity of the neck
can be a reason for asymmetric @xmath37 curves .
this prediction is confirmed by the theoretically calculated @xmath37 characteristics and their comparison with the experimental data presented in fig .
5 . as one can see , the measured and calculated @xmath37 curves in 12 @xmath9 m neck width samples behave similarly : in low electric fields , where ohm s law is still valid , the current values are the same and do not depend on the polarity of the applied voltage .
however , with the increase of voltage when sample is tuned to a strong field regime the electrons become hot , they accumulate differently in the vicinity of the neck due the asymmetrical shape of the structure . as a result , the asymmetry in the @xmath37 characteristics is observed .
since this phenomenon relies on hot - carriers , it is reasonable that is temperature - sensitive : the decrease of lattice temperature will make the effect much more pronounced since the carrier heating at low temperatures is stronger .
this is clearly seen in fig . 3 : the difference between the forward and the reverse - bias curves , for instance , at room temperature@xcite is approximately by three orders of magnitude smaller than that at 4.2 k temperature in the given voltage range .
the temperature effect is also confirmed by theoretical calculations ( data are not given here ) .
therefore , the presence of two different electric field gradients in the hot two dimensional electron regime , allows one to explain the asymmetry of @xmath37 characteristics at low temperatures .
the phenomenon can be called thus the two - dimensional bigradient effect .
it is reasonable that the reduction of the neck should cause stronger asymmetry in @xmath37 curves .
experimentally it is clearly illustrated in fig .
2 . however ,
if one compares the behavior of the curves , for instance , for the 5 @xmath9 m neck sample with a corresponding theoretical estimates ( fig . 5 ) , evident differences can be indentified : ( bottom plot ) characteristics in the asymmetrically - shaped samples with different 12 @xmath9 m and 5 @xmath9 m neck widths at liquid helium temperature.,height=343 ] the theory does not follow the experiment anymore neither in curve shape nor in asymmetry value . therefore , in order to understand the origin of the behavior , a more complete theoretical model must be applied .
althought the equation ( 1 ) is the backbone of electronic device simulations , in the current situation one needs to be extended taking into account the electric field gradient - induced term.@xcite the electron average velocity can be then expressed using the extended / augmented drift - diffusion model@xcite @xmath53 where @xmath54 is the drift velocity in a homogeneous electric field , @xmath19 denotes diffusion coefficient , @xmath55 depics the length coefficient .
the latter parameter is in the order of electron mean free path and determines the severity of the velocity overshoot effect which is pronounced when the variation range of the electric field is equal to or shorter than @xmath55 .
also , it is worth noting that in the given equation we have assumed that nonlocality in diffusion coefficient is negligible and has no observable influence on the effect .
a point of departure for analyzis is an estimate of a mean free path ( @xmath56 ) of two dimensional electrons at liquid helium temperatures . as electron mobility in the studied structures
is of about 1@xmath32/(v@xmath5s ) , this results to the value of @xmath57 where @xmath58=0.068@xmath59 , the latter is the free electron mass .
meanings of other symbols are the same as in previous formulae .
the second step in considering the data is an estimate of the distribution of the electric field in the vicinity of the neck .
voltage value of 0.25 v was chosen as a reference point , since experimentally observed asymmetry of the @xmath37 characteristic starts to be strongly expressed .
theoretical calculations using the set of equations ( 2 3 ) are given in fig .
as one can see , the shape of the electric field distribution strongly varies with the reduction of the neck width : it becomes sharper and the asymmetry becomes more pronounced .
futhermore , distribution of the electric field is squeezed to the narrower range . if for 12 @xmath9 m width sample the electric field extends over 1.7 @xmath9 m , in the 5 @xmath9 m and 2 @xmath9 m necks it is of about 1.2 @xmath9 m and 0.9 @xmath9 m , respectively .
one can assume therefore that in 5 @xmath9 m neck samples the nonlocality can start to predominate since the change of the electric field along the structure is comparable with the electron balistic length .
calculations illustrating the effect of nonlocality in electron drift velocity on @xmath37 characteristics are given -characteristics of 5 @xmath9 m neck sample using different length coefficient @xmath55 .
note the increase of the asymmetry with @xmath55 in the dependences .
bottom plot : experimental results in 5 @xmath9 m neck sample given for comparison.,width=241,height=317 ] in fig . 7 .
one can see that increase of @xmath55 has obvious effect in 5 @xmath9 m neck sample even small increase in length coefficient ( in the range of 0.10.2 @xmath9 m ) stimulates the raise in the asymmetry .
it is evident that calculations for the 5 @xmath9 m are very similar to the given experimental results .
it is deserve to remark that in 12 @xmath9 m neck samples such the influence is not observed ( these theoretical data are omitted here ) .
the inherent feature of the nonlocal effects is a manifestation of the velocity overshoot . to evidence the latter we have performed the monte carlo calculations of the drift velocity in a `` frozen '' electric field taken from a solution of the coupled set of poisson and current - flow equations ( eq . 2 - 3 ) as described previously .
we have simulated the drift of monte carlo particles along the x - direction in the given electric field distribution assuming that the particles enter the modelling region having the boltzmann distribution at the lattice temperature .
the standard three - valley gaas model is applied in the simulation@xcite with an impurity scattering rate corresponding to a @xmath60 donor concentration .
figure 8 presents simulated electron velocity profiles in forward and reverse directions at the electric field profiles for 5 @xmath9 m neck samples .
for the sake of evidence , the electric field profiles taken from fig .
6 are given , too .
as one can see , the velocity overshoot / undershoot and strong nonlocality effects are clearly visible .
m neck sample .
note oscilations in the drift velocity caused by optical phonon emission .
the vertical dashed line indicates the neck position .
background is decorated schematically as the geometrical shape of the structure.,width=287,height=249 ] m neck sample .
note opposite phases in the velocity and the scattering rate .
background schematically shows the geometrical shape of the structure.,width=268,height=306 ] the sharp peaks at the top of the velocity profiles are related to the locality in optical phonon emission rate @xmath61 . note that the velocity profile is asymmetric due to the presence of the different electric field gradients and exhibits nearly constant value in the forward direction ranging from 0.75 up to 3 @xmath9 m .
the monte carlo simulation for electric field profiles from fig .
8 gives the coefficient of current asymmetry of 0.48 .
the factor strongly depends on impurity scattering rate : it decreases as the scatering rate increases .
the simulation with donor concentrations 2@xmath62 , 1@xmath63 , and 5@xmath64 gives the coefficient values of 0.62 , 0.48 , and 0.25 , respectively . to get close - up
view in the physics behind , we have depicted the velocity and @xmath61 profiles in both , forward and reverse , directions ( fig .
also , stationary velocity is added to illustrate the overshoot / undershoot areas .
it is obvious that the electron velocity minima coincide with the @xmath61 maxima forming thus `` comb '' in electron distribution .
the effect is equivalent to the the free - carrier grating formation phenomenon caused by locality of optical phonon emission as already demonstrated in @xmath65 inn structures.@xcite
since the two - dimensional bigradient effect induces asymmetry in @xmath37-characteristics , it is reasonable to consider its possible applications in detection of the high - frequency electromagnetic radiation . to test this idea ,
we have placed samples in a microwave field of 10 ghz frequency and recorded microwave - induced voltage arising over the ends of the sample .
the results for two samples of different neck width recorded at room and liquid nitrogen temperatures are displayed in fig .
10(a ) .
m and 2 @xmath9 m at room and liquid nitrogen temperatures .
( b ) comparison of the two - dimensional bigradient effect - based samples with bow - ties diodes containing 2deg layer .
geometrical dimensions and are the same for both type of samples .
data recorded in 12 @xmath9 m neck size samples at room and liquid nitrogen temperatures .
insets show measurement schemes and design differences .
note much higher sensitivity of the 2deg bow - ties diodes.,width=291,height=222 ] one can see that with the increase of incident power the signal increases linearly in all the range of studied power .
thus , due to the asymmetry in @xmath37 curves caused by the two - dimensional bigradient effect the microwave radiation can be detected @xcite . however , as estimates show , the sensitivity at room temperature is rather low .
for instance , it is only about 12 mv / w and 40 mv / w for the samples with neck width of 12 @xmath9 m and 2 @xmath9 m , respectively .
since the effect is related to the electron heating , the decease in temperature should induce the increase in the detected signal .
the temperature effect , however , is not as large as might be expected from hot - electron physics where change of the signal is proportional to the carrier mobility . as one can see , at 77 k , the detected voltage increases reaching the values of 40 mv / w ( 12 @xmath9 m neck sample ) and 150 mv / w ( 2 @xmath9 m neck sample ) . in fact , the rise in only by a factor of about 4 , while the mobility increases about 14 times .
we associate the outcome with a presence of ballistic electrons which cross the effective length of the samples without scattering.@xcite it is evident that for implementation of the effect into practical devices , such a sensitivity values are not sufficient . following the ideology of the bigradient effect
, one needs to produce much higher non - uniformity of the electric field in the sample s neck vicinity .
this can be achieved easily by metallizing one of the asymmetric parts of the structure .
this is evidenced by calculations of the distribution of the electric field given in fig .
11 . characteristics .
room temperature value of the electron mobility @xmath9= @xmath28/(v@xmath5s ) is used .
the neck width is 5 @xmath9 m .
the filled squares denote the forward current @xmath66 , the empty squares - the reverse current @xmath67 . the coefficient of asymmetry is defined as @xmath38.,height=253 ] it is seen that the non - uniformity of the electric field with one part of the sample metallized is much stronger than in the conventional asymmetrically - shaped structure .
it is therefore reasonable to expect significantly higher values of the sensitivity .
the experimental illustration is given in fig .
10(b ) where the bigradient structures are compared with the so - called 2deg bow - ties diodes@xcite structures where one of the parts of the sample is covered by metal .
one can see that these devices are nearly in two orders of magnitude more sensitive compared to the structures of bigradient - design
. of particular interest for the direct applications are the values of the voltage sensitivity at room temperature . in order to increase the sensitivity of the diode , we have chosen the design 2deg - b ( due to the higher value of the sensitivity at room temperature ) and reduced the size of the apex down to 2 @xmath9 m keeping the other geometrical dimensions the same ( fig .
12 ) . as one can see
, in this case the detected signals are significantly larger than for the 12 @xmath9 m neck diodes at 300 k , and the sensitivity amounts to 2.5 v / w , which is nearly one order of magnitude higher compared to the diodes of 12 @xmath9 m neck width , where it is about 0.32 v / w .
the decrease in temperature down to 77 k allows one to increase the sensitivity up to nearly 40 v / w , i.e. close to a factor of 14.5 which corresponds to the change of 2deg mobility . the latter effect via thickness of spacer is also nicely expressed in fig .
12 were the detected signal in both devices , fabricated from 2deg - a- and 2deg - b - type structures is compared .
additionally , it deserves mentioning that the further decrease of the neck size down to 1 @xmath9 m 800 nm , gives , however , no increase in the sensitivity ( the data are not presented here ) .
we attribute it to weaker coupling of the microwave field into the structure with the reduction of the neck width .
m and 2 @xmath9 m neck size samples based on structures 2deg - a and 2deg - b at room and liquid nitrogen temperatures .
linear function is also given as a guide for an eye.,height=264 ] as to dynamical range , the devices can be used to detect rather different power levels . as already indicated above , the bigradient modulation - doped structures at 300 k in microwaves exhibit sensitivity of about 12 mv / w and 40 mv / w for the samples with neck width of 12 @xmath9 m and 2 @xmath9 m , respectively , however , the sensing range is below 1 w in a pulse mode , i.e. the resistance to overloading is small .
bulk silicon - based 12 @xmath9 m neck width structures are also not sensitive only about 0.1 v / w but they can withstand up to 10 kw maximum applied power at room temprature without any negative outcome.@xcite asymmetrically - shaped bulk gaas structures containing @xmath68-junction were found to be suitable for broadband sensing within ghz thz frequencies at 300 k.@xcite finally , it is worth to remark on the operation of the bigradient structures within the thz range .
one can note that at 300 k the sensitivity at 0.763 thz in the sample with neck width of 12 @xmath9 m it is about 13 mv / w , i.e. close to the value in 10 ghz range .
narrowing the neck width down to 2 @xmath9 m , the sensitivity at 0.763 thz can increase up to 17 mv / w . within frequencies
1.41 thz , 1.63 thz and 2.52 thz the detection observed is rather poor .
we explain it by a weak coupling of the incident thz field into the structure . at helium temperatures ,
the sensitivity reaches the value of 60 mv / w at 0.763 thz for the 12 @xmath9 m neck width sample.@xcite
the effect was originally discovered by measuring @xmath37 characteristics in bulk semiconductors , @xmath11-ge and @xmath11-si , at room and liquid nitrogen temperatures.@xcite in what follows we compare the two - dimensional bigradient effect observed in asymmetrically in - plane shaped gaas / algaas modulation - doped structures with its bulk equivalent desribed in details.@xcite m neck size sample at liquid helium and room temperatures .
insert depicts the same plot for comparison in @xmath69doped asymetrically - shaped silicon samples with specific resistance of 2.5 @xmath70 cm at 300 k. the length of the sample is 23 mm , angles of the asymmetry are @xmath71=@xmath72 and @xmath73=@xmath74 .
more details can be found in ref .
19.,height=230 ] the inherent fearure of both the two - dimentional and bulk effects is asymmetry in @xmath37 characteristics .
different material parameters and geometrical dimensions of the samples determine rather distinctive voltage scale and temperature conditions for experimental observation .
the effect of the dimensionallity we associate with unusual manifestation of the nonlocality in drift velocity .
figure 13 presents a change of the relative current in 5 @xmath9 m neck size sample at 4.2 k and 300 k. for comparison of the features , inset shows equivalent bulk data measured in @xmath11-type silicon according to ref .
it is seen that in the bulk semiconductor the nonlocality is associated with the sign change in the asymmetry clearly observed within 1 - 10 v voltage scale at 77 k ( see inset ) .
in contrast , in modulation - doped structure , no such peculiarity is observed .
it is worth noting , however , that above 0.3 v the change of the relative current is nearly independent of voltage as was indicated earlier ( fig . 7 , above 0.25 v range , where asymmetry of @xmath37 characteristics strongly increases )
we attribute it to the apperance of the nonlocal drift velocity .
asymmetrically in - plane shaped gaas / algaas modulation - doped structures of various design are studied within 4300 k temperature range .
we have shown that current voltage characteristics of such structures at low , 480 k , temperatures exhibits pronounced asymmetry which we explain by a two - dimensional bigradient effect .
this phenomenon can be decomposed into two constituents carrier accumulation and nonlocality in the electron drift velocity each of which manifests itself depending on the values of the in - plane electric fields and their gradients .
varying these parameters via change of the sample neck we have inferred that the effect is induced by different accumulation of two - dimensional electrons due to asymmetrical shape of the structure when the neck size down is 12 @xmath9 m ; in 5 @xmath9 m neck size structures in electric fields close to 3 kv / cm , the nonlocality of the electron drift velocity becomes predominating causing thus further increase of the asymmetry in @xmath37 curves .
the theoretical models suggested conform well with the obtained experimental results .
monte carlo simulation has shown strong nonlocal effects in the electron drift velocity and presence of overshoot / undershoot regions depending on the electric field and its gradient values .
the features of the two - dimensional bigradient effect and its bulk equivalent are compared and possible applications of this phenomenon to detect electromagnetic radiation of ghz and thz frequencies are discussed as well .
we sincerely grateful to vincas tamoinas , jurek lusakowski and kirill alekseev for valuable and comprehensive discussions ; our highest possible appreciations go to adolfas dargys for his illuminating remarks and delicate suggestions in the manuscript preparation , and antanas reklaitis for useful notes on simulation matters .
the work was supported , in part , by the nato sfp project 978030 `` broadband detectors '' .
the research at the semiconductor physics institute at vilnius was performed under the topic `` study of semiconductor nanostructures for terahertz technologies '' ( no .
144.3 ) .
d. seliuta , e. irmulis , v. tamoinas , s. amontas , a. suiedlis , j. gradauskas , g. valuis , p. d. steenson , w .- h .
chow , p. harrison , a. lisauskas , h. g. roskos , and k. khler , proc .
27th intern .
physics on semiconductors icps-27 , flagstaff , arizona , usa , 26 - 30 july 2004 ; editors : jos mendez and chris g. van de walle , aip conference proceedings , vol . 772 ( published june 2005 ; isbn 0 - 7354 - 0257 - 4 ,
two volume , print ; 1697 pages ) , p.
1204 - 1205 .
m. s. shur , ieee electron device lett .
* 23 * , 511 ( 2002 ) . here
the concept of `` ballistic '' not a real mobility ( as it is known , it relates to the collision dominated regime ) was introduced in order to point out that the effective electron mobility in a short channel can be much smaller than in the long one .
concept of the 2deg bow - ties diodes and their demonstration for microwave detection : a. juozapaviius , l. ardaraviius , a. suiedlis , a. kozi , j. gradauskas , j. kundrotas , d. seliuta , e. irmulis , s. amontas , g. valuis , h. g. roskos , and k. khler , semicond .
* 19 * , s436 ( 2004 ) ; for detection of thz / subthz frequencies at room temperature : d. seliuta , e. irmulis , v. tamoinas , s. balakauskas , s. amontas , a. suiedlis , j. gradauskas , g. valuis , a. lisauskas , h. g. roskos , and k. khler , electron . lett . * 40 * , 631 ( 2004 ) .
the sensitivity was estimated scaling the spatial profile of the thz beam , calculating power density and multiplying it by area of the structure .
it was assumed that all the incident power is absorped by the sample . | electric field gradient effects induced by an asymmetrically in - plane shaped gaas / algaas modulation - doped structures of various design are investigated within 4300 k temperature range .
it is demonstrated that current voltage characteristics of such structures at low , 480 k , temperatures exhibit well - pronounced asymmetry arising due to a presence of two different gradients of the electric field in a two dimensional electron gas .
this phenomenon is caused by both , different accumulation of two - dimensional electrons due to asymmetrical shape of the structure and nonlocality in the electron drift velocity .
experiments are illustrated by a phenomenological model and monte carlo simulation .
possible applications of the effect to detect electromagnetic radiation of ghz and thz frequencies are discussed as well
. 1.0 truecm |
in the astrophysical literature @xcite one can find a number of phenomena which can be regarded as a possible manifestation of the existence of the so called _ strange quark matter _
( sqm ) ( in the form of lumps called strangelets ) , extremaly interesting possibility of a possible new stable form of matter .
they include , among others , anomalous cosmic ray burst from _
cygnus x-3 _ , extraordinary high luminosity gamma - ray bursts from the _ supernova remnant n49 _ in the large magellanic cloud or _
centauro _ type events .
there are also several reports suggesting direct candidates for the sqm .
in particular , anomalous massive particles , which can be interpreted as strangelets , have been apparently observed in cosmic ray experiments @xcite .
all this makes a search for other possible candidates or signals for sqm extremaly interesting topic . proceeding along this line
we would like to bring ones attention to the recent ( still unpublished , however ) data from the cosmic ray run of the aleph detector at cern - lep experiment .
the hypothesis which we shall discuss in what follows is that , if confirmed , the muon bundles of extremely high multiplicity observed recently by aleph in its dedicated cosmic - ray run @xcite can originate from the strangelets propagating through the atmosphere and interacting with the air nuclei .
why the cosmolep data are potentially so important ?
the reson is twofold .
first , the studies of high multiplicity cosmic muon events ( called muon bundles ) is potentially very important source of information about the composition of primary cosmic rays .
it is because muons transport ( in essentially undisturbed way ) significant information on the first interaction of the cosmic ray particle with atmosphere . in comparison
electromagnetic cascades are more calorimetric in nature and less sensitive to any model uncertainties , which could be important for establishing the primary spectrum .
the second point has to do with the fact that multi - muon bundles have never been studied with such precise detectors as provided by lep program at cern , nor have they been studies at such depth as at cern @xcite .
the underground location of the lep detectors ( between @xmath0 and @xmath1 meters ) is ideal for the muon based experiments because the corresponding muon momentum cut - off is then between @xmath2 and @xmath3 gev , i.e. , in the most sensitive range from the point of view of the primary interaction , where interaction and decay probabilities are equal at the starting point of the cascade .
the present situation is following .
data archives from the aleph runs have revealed a substantial collection of cosmic ray muon events .
more than @xmath4 muon events have been recorded in the effective run time @xmath5 seconds .
multi - muon events observed in the @xmath6 time - projection chamber with momentum cut - off @xmath3 gev have been analysed and good agreement with the monte carlo simulations ( performed using _ corsika _ code @xcite ) obtained for multiplicities @xmath7 between @xmath8 and @xmath9
. however , there are @xmath10 events with unexpectedly large multiplicities @xmath7 ( up to @xmath11 ) which rate can not be explained , even assuming pure iron primaries . they will be our central point of interest here .
for completeness we shall summarize now features of strangelets and their propagation through the atmosphere , which will be relevant to our further discussion .
the more detailed information can be found in @xcite .
typical sqm consists of roughly equal number of up ( @xmath12 ) , down ( @xmath13 ) and strange ( @xmath14 ) quarks and it has been argued to be the true ground state of qcd @xcite .
for example , it is absolutely stable at high mass number @xmath15 ( excluding weak interaction decays of strange quarks , of course ) and it would be more stable than the most tightly bound nucleus as iron ( because the energy per barion in sqm could be smaller than that in ordinary nuclear matter ) .
on the other hand it becomes unstable below some critical mass number @xmath16 , which is of the order of @xmath17 , depending on the various choices of relevant parameters @xcite . at this value of @xmath15 the separation energy ,
i.e. , the energy which is required to remove a single barion from a strangelets starts to be negative and strangelet decays rapidly by evaporating neutrons .
in @xcite we have demonstrated that the geometrical radii of strangelets @xmath18 are comparable to those of ordinary nuclei of the corresponding mass number @xmath15 ( i.e. , in both cases @xmath19 are essentially the same ) .
we have shown at the same place how it is possible that such big objects can apparently propagate very deep into atmosphere .
the scenario proposed and tested in @xcite was that after each collision with the atmosphere nucleus strangelet of mass number @xmath20 becomes a new one with mass number approximately equal @xmath21 and this procedure continues unless either strangelet reaches earth or ( most probably ) disintegrates at some depth @xmath22 of atmosphere reaching @xmath23 .
this results , in a first approximation ( in which @xmath24 ) , in the total penetration depth of the order of @xmath25 where @xmath26 is the usual mean free path of the nucleon in the atmosphere .
( 7.28,7.11 ) fig.1 integral multiplicity distribution of muons for the cosmolep data ( stars ) , monte carlo simulations for primary nuclei with `` normal '' composition ( dotted line ) and for primary strangelets with @xmath27 ( broken line ) .
full line shows the summary ( calculated ) distribution .
this is the picture we shall use to estimate the production of muon bundles produced as result of interaction of strangelets with atmospheric nuclei .
we use for this purpose the showersim @xcite modular software system specifically modified for our present purpose .
monte carlo program describes the interaction of the primary particles at the top of atmosphere and follows the resulting electromagnetic and hadronic cascades through the atmosphere down to the observation level .
muons with momenta exceeding @xmath3 gev are then registered in the sensitive area of @xmath6 ( randomly scattered in respect to the shower axes ) .
primaries initiated showers were sampled from the usual power spectrum @xmath28 with the slope index equal to @xmath29 and with energies above @xmath30 tev .
( 7.28,7.11 ) fig.2 lateral distribution of muons in the bundles with multiplicities @xmath31 , which originated from primary proton ( stars ) and primary strangelet with mass number @xmath27 ( squares ) ( both with energy @xmath32 tev per particle ) . the integral multiplicity distribution of muons from aleph data are compared with our simulations in fig .
we have used here the so called `` normal '' chemical composition of primaries @xcite with @xmath9 % of protons , @xmath33 % of helium , @xmath33 % of c - n - o mixture , @xmath34 % of ne - s mixture and @xmath34 % of fe .
it can describe low multiplicity ( @xmath35 ) region only . on the other hand , muon multiplicity from strangelet
induced showers are very broad .
as can be seen , the small amount of strangelets ( with the smallest possible mass number @xmath27 , i.e. , the one being just above the estimated critical one estimated to be @xmath36 here ) in the primary flux can accomodate experimental data .
taking into account the registration efficiency for different types of primaries one can estimate the amount of strangelets in the primary cosmic flux . to describe the observed rate of high multiplicity events
one needs the relative flux of strangelets @xmath37 ( at the same energy per particle ) .
( 7.28,7.11 ) fig.3 energy distribution of muons in the bundles with multiplicities @xmath31 , which originated from primary proton ( stars ) and primary strangelet with mass number @xmath27 ( squares ) ( both with energy @xmath32 tev per particle ) .
2 shows the lateral distribution and fig .
3 the energy distribution of muons in the bundles .
they allow us to test the origin of high multiplicity events .
note that , in order to obtain high @xmath7 tail from normal nuclei only , one needs much higher primary energies per nucleon than in the case where strangelets were also added .
distribution of muons from strangelets is broader and their energy spectrum softer in comparison to events with the same @xmath7 induced by protons .
it is interesting to observe that the high multiplicity events discussed here ( with @xmath38 recorded on @xmath6 ) correspond to @xmath39 muons with @xmath40 gev ( or @xmath41 muons with energies above @xmath42 gev ) .
these numbers are in suprisingly good agreement with results from other experiments like baksan valley , where 7 events with more than @xmath43 muons of energies exceeding @xmath42 gev were observed @xcite .
to recollect : we have demonstrated that the recently observed extremaly high multiplicity of muons can be most adequately described by relatively minuite ( of the order of @xmath44 of total primary flux ) admixture of strangelets of the same total energy .
this is precisely the flux we have estimated some time ago @xcite when interpreting direct candidates for strangelets and is fully consistent with existing experimental estimations provided by @xcite .
it accomodates also roughly the observed flux of centauro events as was shown in @xcite .
the cosmolep studies of muli - muon bundles will therefore significantly improve our understanding of the nature and importance of the sqm candidates .
+ acknowledgements : the partial support of polish committee for scientific research ( grants 2p03b 011 18 and 621/e-78/spub / cern / p-03/dz4/99 ) is acknowledged .
+ c.taylor et al . , ( cosmolep coll . ) , _ cosmolep and underground cosmic ray muon experiment in the lep ring _ cern / lep 99 - 5 ( 1999 ) lepc / p9 and _ cosmic multi - muon events in aleph as part of the cosmolep project _ , cosmolep report 1 ( 1999 ) ; cf . also : cern courier , vol .
39 - 8 , october ( 1999 ) 29 . | the hypothesis is discussed that muon bundles of extremaly high multiplicity observed recently by aleph detector ( in the dedicated cosmic - ray run ) can originate from the strangelets colliding with the atmosphere . |
photonic crystal fibers ( pcf ) are a new class of optical fibers which has revealed many surprising phenomena and also holds a big promise for future applications ( see _ e.g. _
these pcfs are made from pure silica with a cladding consisting of a regular lattice of air - holes running along the fiber axis .
depending on the arrangement of the air - holes the guiding of light can be provided by either modified total internal reflection @xcite or by the photonic band - gap effect @xcite and pcfs can even be endlessly single - mode @xcite because of the wavelength dependence of the cladding index . for the basic operation we refer to the review of broeng _ et al . _
@xcite .
understanding the shape and radiation pattern , as illustrated in fig .
[ fig1 ] , of the mode in the endlessly single - mode pcf is very important .
e.g. in tests and applications this is essential for estimations of coupling efficiencies and for determining the mode field diameter from the far - field distribution .
furthermore , it is fundamentally the simplest structure with a hexagonal cladding , and hence the understanding of this structure will be a natural basis for understanding the modes of more sophisticated pcf structures . in this paper
we present a semi - empirical model which is capable of explaining both the near and far - field distribution of the mode , but most importantly also accounts for the fine structure in the transition from the near to the far field .
the simplicity of the model allows for a phenomenological interpretation of the shapes of the near and far - field patterns .
the measurements reported are for a pcf with a triangular air - hole lattice with pitch of @xmath1 and air holes of diameter @xmath2 .
the measurements reported here were performed at a free - space wavelength of @xmath3 , where the light is guided in a single mode in the silica core of the fiber formed by a `` missing '' air hole . in panel
a of fig .
[ fig2 ] a micro - graph of the fiber structure can be seen .
the near - field distribution was measured using a microscope objective to magnify the mode onto a si - based ccd camera . in fig .
[ fig2]b the intensity distribution is shown at focus . by translating the fiber away from the focal plane
, the intensity distribution may be imaged at different distances between the near and the far field .
this is shown in panels b to h in fig .
as expected the mode at focus has a hexagonal shape , that extends in the six regions between the inner holes and is sharply confined at the six silica - hole interfaces .
however , when the image is defocused , the shape at first transforms into a nearly circular shape ( panel c ) followed by a hexagonal shape rotated by an angle of @xmath0 with respect to the focus ( panel d ) .
after this the shape again becomes close to circular ( panel e ) , and finally transforms into the original hexagonal orientation ( panel f ) with six satellites emerging from the distribution ( panels g and h ) .
it is noted that the orientation of the satellites is rotated by @xmath0 with respect to the six inner holes surrounding the core . in fig .
[ fig3 ] ( right ) the intensity distribution in the far - field limit is shown ( several centimeters from the fiber end - facet ) , obtained using a commercial far - field profiler . here , the satellites have fully developed and as shown in the cross sectional plot in fig .
[ fig3 ] ( left ) the peak intensities of the satellites are more than two orders of magnitude lower than the main peak .
hence , a reasonably accurate analysis of the far field may be performed considering only the main peak . apart from being a fascinating and intriguing evolution of the mode shape from the near to the far field , it is important to be aware of these transitions in any application that involves imaging of the modes .
_ e.g. _ for estimations of the mode field diameter and effective area based on near - field analysis , it is important to focus the mode correctly , and the positions corresponding to panel b and panel d in fig .
[ fig2 ] may easily be confused .
they both show the hexagonal shape and have only slightly different mode sizes . hence
, as a measurement procedure for determining the mode field diameter , a direct measurement of the near field may be even more tricky than it is for `` standard technology fibers '' with circular symmetry . in panel
a of fig .
[ fig4 ] two cross - sections of the measured near - field distribution are shown , one taken along a line passing through opposite hole centers ( 1 ) and the second taken along a line passing between the holes ( 2 ) ( rotated by an angle @xmath0 with respect to the first ) .
it is noted that a gaussian distribution is a significantly better fit to the intensity along line ( 2 ) , motivating a simple interpretation of the mode shape : the mode is a circular gaussian distribution from which a narrow distribution localized at each of the centers of the six inner holes is subtracted .
this simple interpretation is theoretically modeled in the following .
in order to simulate the radiated field we start from the fully - vectorial fields in the single - mode all - dielectric pcf @xmath4 where @xmath5 and @xmath6 are the transverse fields and the propagation constant , respectively .
these we calculate numerically by a fully - vectorial plane - wave method @xcite .
substantial insight in the physics of the radiation problem can be gained by expanding @xmath5 in gaussians . introducing the notation @xmath7 and using that the components of @xmath5 can be chosen either real or imaginary we consider @xmath8 for the radiation into free space this gives a linear combination of expanding gaussian beams and this is a well - studied problem , see _
e.g. _ @xcite . neglecting the small back - scattering from the end - facet , the gaussian @xmath9 at finite @xmath10 transforms as @xmath11,\ ] ] where @xmath12 and @xmath13 . in the following we consider a particular simple linear combination in the pcf ; @xmath14 where @xmath15 with @xmath16 being the center position of the six air holes nearest to the core . here ,
@xmath17 ( the radius of the silica core ) , @xmath18 ( the mode - field radius ) , and @xmath19 ( the radius of the air holes ) .
the first term gives the over - all gaussian intensity profile of the mode and with @xmath20 the additional six terms of opposite sign suppress the intensity at the six air - holes nearest to the core . for finite @xmath10
the intensity transforms as @xmath21 in panel b of fig .
[ fig4 ] we show an example of the intensity distribution in an ideal pcf with @xmath22 and @xmath23 at @xmath3 corresponding to experimental situation . for the dielectric function we have used @xmath24 for the air holes and for the silica we have used @xmath25 based on the sellmeier formula .
while eq .
( [ h_gaussian ] ) may seem too simplistic the good fits to gaussians strongly justify it and as we shall see it reproduces the physics observed experimentally . in fig .
[ fig5 ] we show the corresponding near field based on eq .
( [ h_gaussian ] ) .
the profile at the end - facet ( panel a ) first transforms into a close - to - circular profile ( panel c ) followed by a hexagonal shape rotated by @xmath0 ( panels d to f ) , a close - to - circular profile ( panel g ) , and finally a hexagonal shape ( panels h and i ) with the same orientation as at the end - facet ( panel a ) . comparing with fig .
[ fig2 ] this is qualitatively in excellent agreement with the experimental observations .
the fact that the fully coherent scattering description qualitatively reproduces the experimentally observed @xmath0 rotation gives strong indications of its nature ; it is a phenomena caused by an interference between the different gaussian elements used in the decomposition of the fundamental mode in the pcf . in fig .
[ fig6 ] we show the corresponding intensity distribution in the far - field limit which is in a very good agreement with the experiments , see fig .
it is seen that the satellites are reproduced and are in fact oriented in the same way as in the experiment .
moreover the relative intensities between the satellites and the main peak in fig .
[ fig6 ] ( left ) are very similar to the ones in fig .
[ fig3 ] ( left ) .
finally , for fibers where the air holes modify the overall gaussian profile sufficiently ( not shown ) we find indication of additional higher - order spots further away from the center of the intensity distribution which can also be seen experimentally .
the evolution of the mode shape of a pcf with a triangular cladding has been investigated in the transition from the near to the far field . when moving away from the near field at the focus of the fiber end - facet
, it has been observed that the hexagonal orientation is rotated two times by @xmath0 after which six satellites emerge in the radiation pattern . in the far - field limit
the satellites remain in the pattern , having a relative peak intensity more than two orders of magnitude less than the main peak and with an orientation rotated by @xmath0 relative to the six inner holes around the fiber core .
all these observations have been reproduced theoretically , by approximating the near - field distribution by a main gaussian peak from which six narrow gaussians located near the center of the six inner holes have been subtracted . from the simulations
it is concluded that the changes of shape in the radiation pattern are caused by an interference between the different gaussian elements used in the decomposition of the fundamental mode in the pcf .
the results presented here are very important for understanding and analysing the behavior of the mode in many optical systems based on photonic crystal fibers especially those involving imaging and focusing the mode .
furthermore , the successful idea of decomposing the near field of the mode into seven localized distributions can be adapted in future work aimed at a simple quantitative description of the near and far - field distributions , particularly for relating the measured far field to the physical structure around the fiber core that influences the near field .
the latter is very interesting in the field of fiber measurement procedures , since the far - field analysis of conventional optical fibers can not directly be adapted to pcfs because of the lack of cylindrical symmetry .
we thank j. riishede and t. p. hansen ( com , technical university of denmark ) , p. m. w. skovgaard and j. broeng ( crystal fibre a / s ) for technical assistance and useful discussions . | the transition from the near to the far field of the fundamental mode radiating out of a photonic crystal fiber is investigated experimentally and theoretically .
it is observed that the hexagonal shape of the near field rotates two times by @xmath0 when moving into the far field , and eventually six satellites form around a nearly gaussian far - field pattern .
a semi - empirical model is proposed , based on describing the near field as a sum of seven gaussian distributions , which qualitatively explains all the observed phenomena and quantitatively predicts the relative intensity of the six satellites in the far field . 10 opt .
express * 9 * , 674779 ( 2001 ) , http://www.opticsexpress.org/issue.cfm?issue id=124 .
j. opt .
a : pure appl
. opt . * 3 * , s103s207 ( 2001 ) .
j. c. knight and p. s. j. russell , `` applied optics : new ways to guide light , '' _ science _ * 296 * , 276277 ( 2002 ) .
j. c. knight , t. a. birks , p. s. j. russell , and d. m. atkin , `` all - silica single - mode optical fiber with photonic crystal cladding , '' _ opt .
lett . _
* 21 * , 15471549 ( 1996 ) .
j. c. knight , t. a. birks , p. s. j. russell , and d. m. atkin , `` all - silica single - mode optical fiber with photonic crystal cladding : errata , '' _ opt .
lett . _
* 22 * , 484485 ( 1997 ) .
j. c. knight , j. broeng , t. a. birks , and p. s. j. russell , `` photonic band gap guidance in optical fibers , '' _ science _ * 282 * , 14761478 ( 1998 ) .
r. f. cregan , b. j. mangan , j. c. knight , t. a. birks , p. s. j. russell , p.
j. roberts , and d. c. allan , `` single - mode photonic band gap guidance of light in air , '' _ science _ * 285 * , 15371539 ( 1999 ) . t. a. birks , j. c. knight , and p. s. j. russell , `` endlessly single mode photonic crystal fibre , '' _ opt .
lett . _
* 22 * , 961963 ( 1997 ) .
j. broeng , d. mogilevstev , s. e. barkou , and a. bjarklev , `` photonic crystal fibers : a new class of optical waveguides , '' _ opt .
fiber technol . _
* 5 * , 305330 ( 1999 ) .
s. g. johnson and j. d. joannopoulos , `` block - iterative frequency - domain methods for axwell s equations in a planewave basis , '' _ opt .
express _ * 8 * , 173190 ( 2000 ) , http://www.opticsexpress.org/abstract.cfm?uri=opex-8-3-173 .
a. k. ghatak and k. thyagarajan , _ introduction to fiber optics _
( cambridge , cambridge university press , 1998 ) . a. k. ghatak and k. thyagarajan , _ optical electronics _ ( cambridge , cambridge university press , 1989 ) . |
the currently favored model that describes the formation of structure in the universe is the @xmath4 cold dark matter ( lcdm ) paradigm . in this model ,
the initial density distribution of the universe was nearly homogenous , with small gaussian density perturbations imprinted during an inflationary epoch .
these fluctuations expand linearly , until the over - dense regions undergo non - linear gravitational collapse to form bound dark matter haloes .
these haloes form in a hierarchical fashion : small haloes form first , and then larger ones assemble later via merging . in the lcdm paradigm ,
baryons follow the dark matter .
since they can dissipate and cool , baryons condense , and eventually form observable galaxies in the centres of dark matter haloes .
the properties of dark matter haloes in the context of the lcdm paradigm have been studied in detail using numerical simulations over the past couple of decades with increasing resolution ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
this approach has been very fruitful in providing us with a detailed picture of the assembly and growth of structure in the universe .
these theoretical studies provide the framework within which the role of baryons and details of galaxy formation can be probed .
while collisionless dark matter in the lcdm paradigm interacts only gravitationally , baryons dissipate , have pressure , cool , form stars , and interact with radiation .
these , and other effects , introduce complications when trying to understand the properties of dark matter haloes such as their mass , angular momentum , shape , and density profiles from observations of the baryonic component .
there are , however two techniques that have allowed a more direct probe of the dark matter : gravitational lensing observations ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , and measurements of galaxy rotation curves ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
due to the difficulties and assumptions required to translate the observed baryonic properties to dark matter halo properties , cosmological n - body simulations offer a powerful tool to understand the properties and statistics of the dark matter haloes .
even with dark matter only numerical simulations , much has been learned about the assembly of dark matter haloes , including the halo mass function , halo clustering , halo shape and spin at low redshift ( see , e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
however , there have been few detailed studies of dark matter halo properties at high redshifts .
this is partly due to the number of particles required to resolve high redshift , low mass haloes , and still match observations of larger haloes at lower redshifts .
these restrictions until recently prevented the detailed study of a statistically significant sample of collapsed haloes at high redshifts . as the observational frontier is pushed to higher and higher redshifts with reports of the detection of galaxies out to @xmath5 @xcite , a deeper understanding of the properties of the dark matter haloes that host these most distant galaxies is critical as well as extremely timely .
a few recent studies have examined specific dark matter halo properties at higher redshifts .
@xcite , @xcite , and @xcite focus on the mass function of high redshift haloes .
@xcite trace the spatial distribution of dark matter halos from @xmath6 to the present day to understand their effect on galaxy mass haloes today .
@xcite use low resolution simulations to determine the spin and shape parameters of dark matter haloes at @xmath7 .
in a recent study @xcite we reported the results of the first high redshift and high resolution study to follow the growth of angular momentum in dark matter haloes in the mass range @xmath8 to @xmath9 from @xmath10 to @xmath11 , a period spanning 700 myrs of cosmic time .
we found that the spin distribution at these early epochs can be fit by a log - normal distribution as at lower redshifts .
in addition , we examined the two - point correlation function of haloes and found that higher spin haloes are more clustered by factors up to @xmath1 compared to their low spin counterparts at a given mass .
this finding extended across all mass bins and redshifts in our previous study , i.e. from @xmath12 and from @xmath13 .
this paper builds on our earlier work by investigating the role angular momentum and the environment play in the determination of structural properties of dark matter haloes at these epochs . in the lcdm paradigm ,
haloes acquire angular momentum by tidal torques from their neighbors @xcite .
this picture for the acquisition and growth of angular momentum has been shown to be generally accurate in n - body simulations wherein angular momentum initially grows linearly with time @xcite and then slows down at later times @xcite .
linear theory , however , overpredicts the angular momentum when compared to the fully non - linear n - body simulations @xcite .
in addition , as @xcite point out , linear theory predicts the angular momentum of a halo at a given redshift , but not the angular momentum of any particular progenitor at an earlier redshift .
thus , it becomes impossible with linear theory to trace the evolution of a halo s angular momentum in a hierarchical universe evolving via mergers .
@xcite all note that mergers do affect the spin of the halo in addition to the tidal torque model .
@xcite study mergers and spin evolution explicitly and argue that mergers only affect the spin of unrelaxed haloes , and find that relaxed , isolated haloes show no correlation between spin and merger history .
one way to study the acquisition of angular momentum is to correlate information about the environment with halo properties .
previous studies have shown that halo clustering strength depends on the angular momentum of the halo at low redshifts @xcite .
@xcite find that galaxy mass haloes inside clusters have smaller spins than haloes in the field or voids , and @xcite find low specific angular momentum in subhaloes near the central host halo .
observations using the sloan digital sky survey ( sdss ) show that the spin parameter , @xmath14 where @xmath15 is the total angular momentum , @xmath16 is the total energy , and @xmath17 the halo mass , has little dependence on the local environmental density @xcite .
it remains unclear , however , whether the spin parameter derived from the baryonic disk model used in interpreting sdss data correlates well with the host dark matter halo s spin which is what is assumed .
these results are found using galaxy neighbors to trace the large scale tidal field @xcite .
as an example of the difficulties in relating baryonic properties to the host dark matter halo , @xcite report how slight spatial offsets between the dark matter and the baryonic disk create disturbed lensing configurations which can be easily misinterpreted if the misalignment is not included in the mass reconstruction .
also , @xcite report the misalignment of light ellipticity with halo ellipticity in the sdss catalog .
these findings illustrate the complexities when inferring dark matter halo properties from baryonic observations .
assembly bias refers to the observation that the clustering strength of dark matter haloes depends on an additional parameter beyond just halo mass .
assembly bias has been studied in simulations by examining halo formation time , concentration , shape , triaxiality , velocity structure , and substructure content @xcite .
these previous works show convincingly that halo clustering depends on more than just halo mass .
however , all of these studies have been at low redshifts ( @xmath18 ) and for massive haloes ( @xmath19 ) . in this paper
, we extend previous work by examining the formation and growth of dark matter haloes at high redshift , with an emphasis on studying the role angular momentum and environment play in regulating the structural properties of dark matter haloes .
we limit ourselves to haloes in the mass range of @xmath20 to a few times @xmath21 , and the redshift range @xmath10 to @xmath11 .
this allows us to focus on the dark matter haloes that will likely host the first generation of stars and galaxies .
we outline our paper as follows . in section [ sec : lowz ] we summarize studies of angular momentum and assembly bias at low redshift to provide the framework for our findings at high redshift .
we describe our simulations in section [ sec : sims ] , and present the results of convergence tests in section [ sec : ct ] .
our results from the correlation of the spin parameter to the halo environment are presented in section [ sec : environ ] , and that of the effect of angular momentum on halo structure in section [ sec : structure ] .
we conclude with a discussion of the implications of our results for high redshift galaxy formation .
in this section , we summarize earlier findings pertaining to measurements of halo spin and clustering to provide the context for our findings .
many numerical simulations have shown that for massive haloes at low redshift , the distribution of the dimensionless spin parameter follows a log - normal distribution , @xmath22}\ ] ] with typical values of @xmath23 and @xmath24 ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
previous studies have also shown that dark matter haloes are generally triaxial with a preference for prolateness ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
@xcite find that nearly spherical haloes have smaller spins , while there is only a weak trend of halo triaxiality with spin . in studying halo concentration versus spin ,
it is seen that when unrelaxed haloes are eliminated from the sample there is only a very weak surviving ( if any ) correlation between these two parameters @xcite .
@xcite and @xcite all find that haloes with larger spins are more clustered than low spin haloes at a given mass .
however , @xcite and @xcite find in their simulations that haloes in cluster environments have smaller spins and are more spherical than their counterparts in the field , and @xcite find that haloes in filaments have larger spins than haloes of the same mass in clusters or in voids .
@xcite find that there is no environmental dependence on the spin parameter for haloes at a given mass .
thus , there appears to be some question as to the extent to which the environment affects the angular momentum properties of dark matter haloes . in this work
, we explore this relationship at high redshift using both the clustering strength and the local density to characterize the environment .
we run a series of n - body simulations to follow the growth of dark matter haloes from @xmath25 down to @xmath11 .
we choose the particle mass such that a @xmath26 dark matter halo has 100 particles . for @xmath27 particles
, this requirement sets the comoving box size at @xmath28 and the particle mass at @xmath29 .
the initial conditions are generated using a parallelized version of grafic @xcite , which calculates the gaussian random field for the dark matter particles .
we use gadget-2 @xcite to follow dark matter particles down to a redshift of @xmath11 , with output snapshots at @xmath30 .
we use the wmap3 ( \{@xmath31 } = \{0.238 , 0.762 , 0.0416 , 0.732 , 0.958 , 0.761 } , * ? ? ?
* ) and the wmap5 ( \{@xmath31 } = \{0.258 , 0.742 , 0.044 , 0.719 , 0.963 , 0.796 } , * ? ? ?
* ) cosmological parameters for our simulations .
the wmap3 cosmology was used for our first runs studying numerical convergence of our measurements of the angular momentum , and the wmap5 cosmology was used for the results presented in sections [ sec : environ ] and [ sec : structure ] .
table 1 shows the runs and the relevant parameters used in this paper .
gadget-2 uses a softening length , @xmath32 , to soften the gravitational force to prevent spurious 2-body interactions . [
cols="^,^,^,^,^,^ " , ] [ comp_stats ]
in our previous work @xcite , we found that at high redshift , higher spin haloes are more clustered than lower spin haloes at a given mass , using the halo - halo correlation function as a proxy for the environment .
as in our previous work , we again define distinct spin bins from the lognormal distribution , each with roughly one third of the haloes , and denote them as high , median , and low spin bins . using several different ways to characterize the environment , we report our findings of correlations between measures of the environment with the spin parameter for haloes in the mass range @xmath8 to @xmath33 .
we first use the distance to close neighbors as a measure of environment .
this distance is calculated as the separation between each halo s centre ( i.e. the densest particle ) .
we measure both the distance to the @xmath34 ( @xmath35 ) and @xmath36 ( @xmath37 ) nearest neighbors .
both distances show the same result : haloes with closer neighbors have a small tendency to have larger spins .
the second method that we use to measure environment is the over - density within ten times the virial radius .
we use as a proxy for the virial radius @xmath38 : the radius of a spherical region which has an average density of @xmath39 .
the factor @xmath40 is the overdensity criterion @xmath41 evaluated at @xmath11 . @xcite
give a fitting formula in a lcdm universe for @xmath41 , but at high redshifts , @xmath41 is approximately the same as it would be in an einstein - de sitter universe , such that @xmath42 . to calculate this radius
, we use all particles surrounding the centre of the potential for the given halo , regardless of whether or not the particles are part of the halo .
this is to account for the fact that not all haloes extended out far enough to reach the required density threshold . in other words ,
some haloes have @xmath38 outside the most distant particle assigned to the group .
we then calculate the over - density within @xmath43 , and find that haloes in over - dense regions have a slight tendency to have higher spins .
figure [ environ ] shows the trends with spin for both measures of environment in two mass bins : @xmath44 and @xmath45 , and histograms for the lowest mass bin after separating our sample by spin .
the histograms for the @xmath46 and @xmath33 bins show qualitatively the same trends .
we find no strong correlation between environment and spin it is possible to find low spin haloes in very dense environments , and high spin haloes in sparse environments .
however , we do find a small excess probability that haloes with lower spins are in less dense environments , as evidenced by the histograms in figure [ environ ] .
this is in contrast to @xcite , who report that galaxy scale haloes ( masses @xmath47 ) in clusters have lower spin than isolated haloes in the field .
interestingly , they find that the trend they report also disappears as they go back to @xmath48
. this may imply a trend with redshift where at high redshift , high spin haloes are in denser environments , and at very low redshift , high spin haloes are in less dense environments .
however , a strict comparison of these two numerical studies can not be made as we have significantly more haloes , probe significantly higher redshift slices and significantly lower halo masses .
@xcite only use the over - density criterion to quantify the environment , whereas our findings hold across three different measures of environment .
we note that the trends reported here have shallow slopes , similar to the @xcite work .
we conclude that the dependence of spin on environment is weak for the halo masses and epochs studied here .
in determining the effect of halo spin on halo structural properties , we first turn to the mass distribution within the halo .
we fit the nfw profile , @xmath49 @xcite , to each halo through a least squares fit to the radially averaged density profile calculated using all particles assigned to the halo by our group finding algorithm . in
doing the least squares fit , we seek to minimize @xmath50 , which is defined as @xmath51\ ] ] and where @xmath52 is the number of degrees of freedom .
we then use the concentration parameter , defined as @xmath53 , to characterize the density profile . in figure [ struct ] ( left column )
we show the measured concentrations versus spin parameters binned by mass .
there is only a small trend with spin , implying that the density structure has only a weak dependence on the angular momentum properties of the halo . in the bottom left panel of figure [ struct ] ,
we show the histogram of concentration values for haloes in the @xmath8 mass bin .
we report little to no dependence of @xmath54 on spin in these mass ranges .
we also look at the relationship between halo mass and @xmath54 .
we find that the fitting function in @xcite extrapolated down to these masses and up to these high redshifts , given by @xmath55 , where @xmath56 and @xmath57 is the typical collapsing mass at redshift @xmath58 , is not a good fit , and predicts values for the concentration that are too small compared to our estimates from the simulation .
this disagreement is not surprising as bullock et al .
studied higher mass haloes at lower redshifts ( @xmath59 that have had a significant amount of time to merge and virialize . observed rotation curves
are used to obtain mass estimates of galaxies and clusters .
the measured velocity of stars and gas reflects the combined gravitational potential of both the baryons and dark matter in the galaxy . to infer the total mass from an observationally determined rotation curve requires understanding how both types of matter are distributed spatially in galaxies .
@xcite report that for sb - im spirals , rotation curves can be represented by a universal function which is the sum in quadrature of two velocity curves : one from the disc and one from the dark matter halo .
this model is developed further in @xcite , where the dark matter velocity component is strictly a function of the virial mass of the halo .
if , however , the velocity curve of the dark matter halo is dependent on a second parameter , such as @xmath60 , we would expect to find a systematic error in a mass estimate for the galaxy derived from the observed rotation curve .
a relationship between @xmath60 and a characteristic circular velocity can be expected if we consider an alternative spin parameter given in @xcite : @xmath61 , where @xmath15 is the angular momentum inside some radius , @xmath62 , @xmath17 is the mass inside @xmath62 and @xmath63 is the circular velocity at @xmath62 . for the case of a truncated , singular isothermal sphere ,
@xmath64 at the virial radius .
thus it is expected that a relationship between the @xmath60 used in our work and a characteristic circular velocity should exist .
we find that the peak velocity , @xmath65 , in the measured circular velocity curve , defined as @xmath66 , systematically depends on spin . to measure this curve , we sort all particles in the halo by their distance to the halo centre . from this
, we can read off the circular velocity curve at the radius of each particle , and we define @xmath65 as the maximum value of this curve . figure [ struct ] ( right column ) shows that for all mass bins , there is a systematic offset in the peak velocity as a function of halo spin : high spin haloes have higher @xmath65 than their low spin counterparts .
the bottom panel shows two histograms of @xmath65 , binned by spin .
the peak of the high and low spin curves is offset by 20% in the @xmath8 mass bin , and 10% in the @xmath46 haloes ( not plotted ) . from a given @xmath65
one can infer a halo mass using the definition of circular velocity such that @xmath67 .
alternatively , one can use published relationships between @xmath65 and @xmath68 such as the power law @xmath69 where @xmath70 from @xcite ( see also * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
either way , when two halos of the same ( unknown ) mass have different spins , there is a bias in the resulting @xmath65 , which will lead to a bias in the mass estimated from the circular velocity .
* therefore two observed galaxies with differing values of @xmath71 could in fact inhabit dark matter haloes of the same mass but with different spin parameters .
alternately , two galaxies with the same value of @xmath65 could have different masses due to their different spin values .
* we can estimate the extra uncertainty in the halo mass due to the bias induced by @xmath60 . using the circular velocity definition
, we can translate a difference in velocity to a difference in mass : @xmath72 .
therefore the estimated total mass from an observationally measured value of @xmath65 can be off by 20 - 40% arising due to the unknown value of the spin parameter of the dark matter halo that hosts the galaxy .
another well explored correlation at low redshift is between the shapes of dark matter haloes and the spin parameter . here , we investigate the corresponding correlations for these high redshift haloes .
the existence of a correlation between halo shape and spin is likely to have important consequences for the formation of pre - galactic disks in these haloes .
we calculate the sphericity , @xmath73 , and triaxiality , @xmath74 from the eigenvalues ( @xmath75 ) of the normalized moment of inertia tensor , @xmath76 where @xmath77 is the distance to the halo center .
the tensor is calculated using only the particles assigned to the halo , not from all particles within the @xmath38 as is often done .
this allows for consistency with our measurements of @xmath60 .
we used the normalized tensor so that we do not weight particles on the outskirts of a halo stronger .
this helps to prevent the halo shape being dominated by residual tidal features from a recent merger or other irregularities found only in the outer regions which are not present in the halo interior .
we caution that both the normalized ( e.g. * ? ? ?
* ; * ? ? ?
* ) and unnormalized ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) tensor are used in the literature , and so caution must be used when comparing shape distributions .
we show in figure [ normshape ] histograms of @xmath78 and @xmath79 when calculated using both a normalized and unnormalized moment of inertia tensor .
we find that many more halos are significantly aspherical , and halos tend to be less prolate when the shape is found using the unnormalized tensor .
this is contrary to the study in @xcite , which reported no systematic offset in @xmath78 between the two methods .
we note , however , that they only used particles within @xmath80 , rather than all particles assigned to the halo as we do .
therefore , our unnormalized method will have a stronger bias because we include distant particles in the calculation of the halo shape . in
what follows , we use the normalized tensor to calculate the shape of the haloes .
we show in figure [ shape ] the relations between @xmath78 ( left column ) , @xmath79 ( right column ) , and @xmath60 , as well as the histograms of @xmath78 and @xmath79 binned by spin and mass . for higher spin haloes ,
the trend is for haloes to become less spherical , and more prolate .
this is in agreement with studies at low redshift and higher masses @xcite .
we also plot the two point correlation function , @xmath81 of the haloes in figure [ corr_shape ] , binning the haloes by their values of @xmath78 and @xmath79 .
the correlation function represents the excess probability of finding a halo at a distance @xmath82 when compared to a random distribution of haloes .
we use the same method as our previous work @xcite to calculate @xmath81 .
we calculate @xmath81 for haloes in four different bins , corresponding to large and small values of @xmath78 ( left column of figure [ corr_shape ] ) and @xmath79 ( right column ) .
the cuts were chosen so that one third of the haloes lie in each bin .
we also separate out haloes by mass so that we eliminate any mass effect on the value of @xmath81 . in the top row of figure [ corr_shape ]
we show @xmath81 for haloes with @xmath83 and in the bottom row we show @xmath8 haloes .
we find that nearly spherical haloes ( values of @xmath78 close to 1 ) are more clustered than the more aspherical haloes .
this trend is stronger for haloes in the @xmath46 mass range than in the @xmath8 mass range .
in the higher mass bin , we find an increase of @xmath2 in the correlation of nearly spherical haloes when compared to the extremely aspherical halo sample . in the lower mass range ,
we report a smaller increase of only @xmath84 .
thus we conclude that our results at high redshift follow trends found at lower redshift @xcite . unlike the sphericity parameter ,
when we separate haloes into high and low values of @xmath79 , we see little change in @xmath81 .
this is somewhat surprising , as @xcite find at low redshift an offset in @xmath81 when binning haloes by @xmath79 .
finally , we note that the clustering strength is weaker if an unnormalized moment of inertia tensor is used to calculate @xmath78 and @xmath79 of the haloes . in figure [ bias ] , we show the bias parameter , @xmath85 as a function of the halo peak height @xmath86 , where @xmath87 is the growth function @xcite .
this allows us to compare results from two redshifts ( @xmath7 and @xmath11 ) and from 5 mass bins at each redshift .
we split our haloes according to four properties : @xmath60 , @xmath78 , @xmath79 , and @xmath54 .
we report that the spin parameter has the largest effect on the bias , while the triaxiality has the least .
this implies that angular momentum has the strongest dependence on environment of all these considered variables and will be the source of the largest systematics due to differing environments between haloes .
therefore , baryonic properties that depend on the angular momentum of the host halo should have systematic offsets due to their local environment .
this may play a large role in the formation of the earliest galactic disks , as well as affect semi - analytic models which relate the baryonic spin to the dark matter halo spin ( * ? ? ?
* ; * ? ? ?
* e.g. ) , and estimates of the dark matter halo spin from the observed baryonic disk @xcite .
our key findings can be summarized as follows : * we have measured the spin , concentration , circular velocity , sphericity , and triaxiality parameter for a statistically large sample of dark matter haloes at high redshift ( @xmath88 ) . *
high spin haloes at high redshift are @xmath1 more clustered than their low spin counterparts at a given mass , and are more likely to be found in high density environments . * high spin haloes ( with masses @xmath89 ) have smaller maximum circular velocities than low spin haloes , leading to errors up to @xmath3 in the derived enclosed mass . *
high spin haloes at high redshift are more likely to be aspherical and prolate , similar to findings at low redshift . *
nearly spherical haloes are up to @xmath2 more clustered than extremely aspherical haloes , while there appears to be no difference in the clustering strength based on the triaxiality of the haloes .
our findings have an impact in two general areas : the role of angular momentum in halo structure and formation , and the role of assembly bias at high redshift .
our findings show that angular momentum has a measurable correlation with structural properties , including the concentration , sphericity , and triaxiality .
also , haloes with higher spin are preferentially found in higher density environments .
the finding that halo spin correlates with local environment at high redshift is important to the understanding of the evolving properties of the baryonic component of dark matter haloes .
a correlation between spin and baryonic properties , such as formation time , disk rotational speed , or disk size , would be the specific consequences of the correlation with the environment .
these correlations are likely to be significantly stronger at high redshift , before too many mergers have happened which could destroy any correspondence between the dark matter spin and the baryonic structure .
the role of assembly bias in halo evolution has been discussed at low redshifts .
our work extends the study of assembly bias to high redshifts , when the first galaxies form .
we see similar results to those at low redshift when studying the dependence of clustering on spin and sphericity of dark matter haloes .
however , our results differ when looking at the triaxiality of haloes .
we find that there is no difference in clustering strength between prolate and oblate haloes .
our findings are of particular importance now that galaxies are being found at these high redshifts @xcite .
the clustering of galaxies has been used to infer the masses of their host dark matter haloes ( e.g. , * ? ? ?
* ) . however , we find that properties other than mass - in particular halo spin - affect the measured correlation function
. this additional parameter will induce errors in mass estimates for dark matter haloes inferred purely from clustering measurements .
this fits in with the results of @xcite , who suggest that mass is likely not the only parameter that drives the interaction of haloes with their large - scale environment .
in addition to mass measurements , assembly bias will play an important role in feedback at these high redshifts . at the highest redshifts ,
simulations show that population iii stars have a large impact on their environment due to radiative and supernova feedback @xcite .
one consequence of our findings is that if angular momentum affects the formation and evolution of these pop iii stars , their feedback effects will show an environmental bias .
thus , the distribution of metals and reionization will be more clustered than otherwise expected .
we also expect that because these stars are the first baryonic objects to collapse , their properties can be expected to have a stronger relationship to their host dark matter halo than galaxies today , which have undergone multiple mergers .
we intend to pursue the consequences of our findings on baryonic results in future work .
our results suggest that the angular momentum properties of dark matter haloes likely have consequences for the properties of the first stars and galaxies hosted by them .
this work was supported in part by the facilities and staff of the yale university faculty of arts and sciences high performance computing center .
we also wish to thank zheng zheng , laurie shaw , jason tumlinson , elena donghia , and lars hernquist for helpful discussions and comments . | in this paper , we examine in detail the key structural properties of high redshift dark matter haloes as a function of their spin parameter . we perform and analyze high resolution cosmological simulations of the formation of structure in a lcdm universe .
we study the mass function , shapes , density profiles , and rotation curves for a large sample of dark matter haloes from @xmath0 .
we also present detailed convergence tests for individual haloes .
we find that high spin haloes have stronger clustering strengths ( up to @xmath1 ) at all mass and redshift ranges at these early epochs .
high redshift spherical haloes are also up to @xmath2 more clustered than extremely aspherical haloes .
high spin haloes at these redshifts are also preferentially found in high density environments , and have more neighbors than their low spin counterparts .
we report a systematic offset in the peak of the circular velocity curves for high and low spin haloes of the same mass . therefore , estimating halo masses without knowledge of the spin , using only the circular velocity can yield errors of up to @xmath3 .
the significant dependence of key structural properties on spin that we report here likely has important implications for studies of star formation and feedback from these galaxies .
cosmology : dark matter cosmology : early universe galaxies : high - redshift galaxies : formation |
tiny superconductors , accommodating only a few quantized magnetic flux lines , display a great variety of vortex structure different from the abrikozov vortex lattice in bulk superconductors @xcite .
examples include vortex polygons and concentric vortex rings called `` vortex shells '' , which are stable configurations of the repulsive flux lines under the geometrical confinement via the influence of the screening current flowing along the sample edge . the issue of vortex matter confined into small superconductors has been studied theoretically and numerically for many years , focusing mainly on how the vortices are distributed in disk @xcite@xcite@xcite@xcite@xcite@xcite , square @xcite , and triangle shaped small superconductors @xcite .
several studies reveal the field evolutions of the vortex structure in small superconductors , and argue which vortex configurations are energetically favorable and how the transition between different vortex states occurs . the obtained rule of shell filling and magic number configurations for consecutive new shells are relevant to phenomena observed in other systems , including the puzzling nucleation of vortices observed in rotating condensates of superfluid @xmath1he @xcite and cold dilute alkali - metal gases @xcite . the experimental investigations for vortex states in small superconductors were initiated by hall magnetometer measurements @xcite@xcite , followed by a multiple - small tunneling junction measurement @xcite .
kinks in the magnetization or jumps in the tunneling spectra mark transitions of vortex states in small superconducting dots .
they evidence changes of vorticity @xmath2 ( the number of vortices ) , not actual distributions of vortices in the dots .
the visualization of vortices in small superconductors was reported earlier in studies of the scanning superconducting quantum interference device ( squid ) microscopy @xcite@xcite and later in studies of the scanning hall probe microscopy @xcite@xcite , showing reasonable images of vortex configurations for @xmath2 up to 6 .
more vortices have been imaged in recent bitter decoration studies as patterns of magnetic particles deposited over micrometer - sized dots of nb films @xcite@xcite and nb mesas@xcite .
the experiments reveal the rule of the shell filling with magic numbers for @xmath2 up to 40 by combining vortex configurations visualized over many disks at different magnetic fields , although some features influenced by the bulk pinning and/or the roughness of the disks are involved .
the scanning squid microscope technique , which we employed in this study , is an alternative and complement method to visualize vortices in the superconducting dots . by scanning a small pick - up loop over the dots
, one can magnetically image the vortices without damaging the samples .
this allows us to study systematically how the vortex configuration evolves with magnetic field in a specific sample . while the scanning squid microscope has attained the best magnetic sensitivity ( better than 5 @xmath3 ) , the spatial resolution is limited typically to a few micrometer due to the size of the pick - up loop ( @xmath4 10 @xmath5 m ) @xcite .
this implies that the observable vortex density is low ( @xmath6 100 @xmath5 t ) and the sample becomes much larger than the loop size .
the earlier studies were made on ybco @xcite and nb dots @xcite with @xmath4 50 @xmath5 m in size , which is decades larger than the penetration depth @xmath7 or the coherence length @xmath8 of the superconducting materials . in such large dots , the effect of the geometrical confinement is weak and ( sparse ) vortices form likely disordered configurations due to the dominant influence of the bulk pinning in the superconducting materials @xcite@xcite .
if the influence of the bulk pinning is reduced by using weak pinning , superconducting materials , the situation can be different : the geometrical confinement dominates over the bulk pinning .
moreover , when the dots become thin in thickness @xmath9 @xmath10 , the interaction between vortices becomes long ranged and it decays slowly with the vortex - vortex spacing @xmath11 as @xmath12 for @xmath13 @xcite , in contrast to the short ranged interaction characterized by the exponential decay @xmath14 for @xmath15 in the bulk superconductors .
thus , even in the large dots the interplay between the vortex - vortex interaction and the geometrical confinement may lead to geometry - induced , symmetric configurations of the sparse vortices in the presence of the weak bulk pinning . here
, we report on the direct observation of vortices in large thin disks of an amorphous ( @xmath16 ) moge film by the scanning squid microscopy . due to the weak bulk
pinning of the amorphous film , we are able to observe ( quasi- ) symmetric configurations of polygons and concentric shells of vortices for @xmath2 up to 19 .
the results illustrate not only how the vortex configuration evolves with the magnetic field , but also how the vortex polygons change in size and rotate with respect to the disk center .
we present vortex images obtained in a disk with a pinning site and discuss how vortex configurations are altered by a pinned vortex .
this paper is outlined as follows : after describing experimental details in sec .
ii , we present vortex polygon states observed for small vorticities up to @xmath2= 5 in sec .
iii . in sec .
iv we present field evolutions of vortex shells observed for large vorticities @xmath17 6 .
the influence of a pinning site on vortex configurations is discussed in sec .
v. in sec .
vi comparison with other experiment is made .
our summary is given in sec .
we used a commercial scanning squid microscope ( sqm-2000 , sii nanotechnology ) with a dc squid magnetometer made of nb / al - alo@xmath18/nb josephson junctions and an inductively coupled , pick - up loop of a nb film @xcite .
the pick - up loop had 10 @xmath5 m in diameter with 2 @xmath5 m line width .
the dc squid and the pick - up loop were integrated on a square si chip of 3 @xmath19 3 mm@xmath20 in size , which was mounted on a phosphor bronze cantilever .
the sensor chip was tilted with respect to the sample stage by a shallow angle of @xmath4 10 degrees and a corner of the chip was softly in contact with the sample surface , in order to keep a short distance ( @xmath4 5 @xmath5 m at the minimum ) between the loop and the sample surface .
this ensures safe operation of the scanning squid microscope with reasonably high spatial resolutions ( @xmath4 4 @xmath5 m ) .
the whole assembly including the sensor chip and the sample stage was surround by a @xmath5-metal shield , resulting in a residual magnetic field ( ambient field ) of @xmath4 1 @xmath5 t .
a coil wound in our sample stage allowed us to apply small magnetic field @xmath21 perpendicular to the sample .
we scanned the sample stage in xy directions by dc stepper motors . in order to minimize the possible mechanical hysteresis effect
, we took one scan direction from the left side to the right side of the image .
the scanning speed was about 5 pixels / s and this resulted in an image of 64@xmath1964 pixels by typically 15 minutes .
the temperature for the squid sensor was 2 - 3 k , while that for the sample stage was varied .
we prepared weak pinning disks of an @xmath16-mo@xmath22ge@xmath23 ( @xmath24 78 @xmath25 ) thin film with a lithographic technique . after preparing a silicon substrate covered with a patterned - resist film
, we sputtered the @xmath16-moge film with 0.22 @xmath5 m thickness on top . employing the lift - off technique , we patterned the film into arrays of disks with four different diameters @xmath26 of 20 @xmath5 m , 34 @xmath5 m , 56 @xmath5 m and 106 @xmath5 m .
the spacing between the disks was chosen to be large enough ( @xmath27 200 @xmath5 m ) , so that the interdisk coupling can be excluded . in order to reduce possible damage during the scanning
, all the disks were covered with 0.1 @xmath5 m thick sio@xmath28 film as a protective layer .
the material parameters relevant to this study are as follows : the superconducting transition temperature @xmath29 is 6.0 k , the coherence length @xmath30 at @xmath31 is 4.5 nm , and the penetration depth @xmath32 at @xmath31 is 0.58 @xmath5 m ( @xmath33 ) @xcite . since the thickness is smaller than @xmath34 , the penetration depth is expanded to the effective one given by @xmath35 @xmath36 . for the present film @xmath37 3.1 @xmath5 m .
normalizing the disk size by the effective penetration depth , we find @xmath38 6 - 35 @xmath39 , indicating large mesoscopic disks . in weak pinning materials
, the distance @xmath40 between the pick - up loop and the sample surface is an important parameter to be controlled in order to obtain clear images of vortices . as demonstrated by plourde _
@xcite , the vortices may be dragged and/or shifted in positions by the pick - up loop during the scanning .
this coupled motion of vortices can be reduced when @xmath40 is increased by lowering the sample stage from the contact point with the sensor chip .
this procedure , however , results in a smeared image of vortices due to the spread of the magnetic flux above the sample surface @xcite .
we could not resolve individual vortices when @xmath40 became more than the loop diameter ( @xmath41 @xmath5 m ) .
hence , we set @xmath42 7 @xmath5 m as a typical distance .
all the images presented in this study were taken in the field - cool measurement , where the magnetic field was applied in some temperatures ( @xmath410 k ) above @xmath29 , followed by cooling the samples in the magnetic field to temperatures ( @xmath4 3 - 5 k ) below @xmath29 where scanning squid measurements were done .
this procedure results in the equilibrium distribution of vortices .
because of the possible motion of vortices during the scanning , each image was taken in a single scan made on a fresh vortex state prepared by the field - cool procedure . no image taken by repeated scans
is presented .
first we focus on vortex configurations observed for small vorticities up to 5 . figs .
1(a)-1(f ) show selected vortex images observed in a 34 @xmath5 m disk .
a color bar indicates the magnitude of the magnetic flux in the pick - up loop .
these images display clearly individual vortices with reasonable spatial and magnetic resolutions @xcite . the imaged vortex size is larger than 2@xmath43 7 @xmath5 m ) due to defocusing effect originating from the large distance @xmath44 7 @xmath5 m ) between the loop and the sample surface . details of the field profile of vortices
have been reported in refs .
we find that the vortices form symmetric configurations with respect to the disk center .
the vortex pattern evolves with field as follows ; after the meissner state where the magnetic field is expelled from the disk , a single vortex appears in the disk and it sits at the disk center ( fig .
1(a ) ) . then
, two vortices are located symmetrically with respect to the disk center ( fig .
1(b ) ) , followed by the formation of a triangle ( fig .
1(c ) ) , a square ( fig .
1(d ) ) and a pentagon ( fig .
1(e ) ) of vortices in the disk . although some distortion is visible for the square , these symmetric configurations are nearly identical to those reported in the bitter decoration @xcite and the theoretical studies @xcite@xcite@xcite .
an important remark for the observation is that some of the vortex patterns rotate with respect to the disk center .
this feature is clearly visible in the images of the vortex pair shown in figs . 1 ( b ) and 1(f ) . in fig .
1(b ) , the vortex pair is aligned nearly with the horizontal direction of the image , while in fig . 1
( f ) it rotates by @xmath4 45@xmath45 with respect to the disk center . other orientation of the pair is observed in other scan , indicating the rotational degree of freedom in the vortex configuration @xcite .
similar rotation is observed for the triangle pattern , but not clearly for the square and the pentagon patterns .
next we show how the number of vortices evolves with applied magnetic field .
counting the number of vortices in the images taken at various magnetic fields , we plot the results in fig .
2(a ) . here , instead of the field strength @xmath21
, we use the magnetic flux @xmath46 in the disk normalized by the flux quantum @xmath47 . as observed
, the vorticity @xmath2 shows a stair case behavior with field : after the meissner state ( @xmath48 0 ) , the single vortex state ( @xmath48 1 ) appears at @xmath49 2 , and it continues until @xmath49 4.5 , above which the vortex pair state ( @xmath48 2 ) appears . the vortex triangle ( @xmath48 3 ) , square ( @xmath48 4 ) and pentagon ( @xmath48 5 ) states appear at @xmath49 6.4 , 8.2 and 9.5 , respectively .
we note that the stair case behavior is well below the condition of @xmath50 where the magnetic field penetrates fully the disk and this is represented by a solid line .
the difference from the line becomes large on increasing field . at @xmath51 8 , for instance ,
@xmath52= 4 , which is half of @xmath53 .
this indicates a large diamagnetic response of the disk and it implies the presence of the large screening current near the disk edge .
the influence of the screening current upon vortex configurations is found by considering how the size of vortex polygons changes with magnetic field . as exemplified in the inset to fig .
2(b ) , we draw a circle circumscribing the positions of vortices and estimate the diameter @xmath54 of the polygon ring .
the results for this analysis made at different fields and different vorticities ( down to @xmath2= 2 ) are summarized in the main panel of fig .
as observed , @xmath54 varies largely with @xmath2 . for example
, @xmath54 at @xmath2= 2 is roughly 40 @xmath25 of the disk diameter , while that at @xmath2= 5 becomes expanded to roughly 65 @xmath25 of the disk diameter .
focusing on the results at each vorticity , we find @xmath54 decreases with increasing field .
the shrinkage of @xmath54 is observed to be 5 - 10@xmath25 of the disk diameter , depending on @xmath2 .
the obtained results are in excellent agreement with the theoretical considerations for vortex polygons confined by the screening current near the disk edge @xcite@xcite . in the london limit
, the diameter of a vortex polygon at a given magnetic field is written as @xmath55 where @xmath56 is the number of vortices on the polygon ring .
the behaviors for @xmath56= 2 - 5 are represented by dotted curves in fig .
our results for @xmath57 2 - 5 follow nicely those curves without any adjustable parameters . moreover , they are satisfactory for the condition that the polygon size should be below the maximum diameter above which the formation of the polygon ring is not stable .
this is represented by a solid curve and it follows the condition @xcite given by @xmath58 thus , we are convinced that the vortex configurations observed in the 34 @xmath5 m disk are dominantly influenced by the geometrical confinement via the screening current near the disk edge .
we note , to the best of our knowledge , that these quantitative agreements , together with the rotation of the vortex patterns in the same disk , are not reported in other experimental studies so far .
for vorticities more than 6 , we observe the formation of concentric shell rings of vortices .
since the vortices become dense and their magnetic images are overlapped each other , for clarity , we present here the results obtained in larger disks .
3(a)-3(c ) show vortex images for @xmath2= 6 - 8 observed in a disk with 56 @xmath5 m diameter .
one can clearly see the shell formation of vortices . in the image of fig .
3(a ) , for example , a vortex sits at the center of the disk , while the others form a ring around . using the standard notation ,
this configuration is referred to as ( 1,5 ) , meaning that the inner shell has one vortex and the outer shell has five vortices .
after the ( 1,5 ) configuration , we observe ( 1,6 ) and ( 1,7 ) configurations , which are shown in figs .
3(b ) and 3(c ) , respectively .
namely , the increase of vorticity in the disk results in the growth of the outer shell ring .
the growth of the inner shell occurs at even larger @xmath2 .
4 shows the field evolution of vorticity , together with vortex configuration , observed in the largest disk with 106 @xmath5 m diameter .
the vortex shells with a single vortex at the center are observed for @xmath2= 6 , 7 , and 8 .
this is consistent with the observation made on the 56 @xmath5 m disk discussed above . at @xmath2= 9
a vortex pair appears as the inner shell , while the others form the outer shell ring around , resulting in a ( 2,7 ) configuration .
the pair configuration of the inner shell is also observed at @xmath2= 10 . at @xmath2= 11
three vortices form a triangle as the inner shell , resulting in a ( 3,8 ) configuration .
the triangular inner shell is observed over some field range and persists up to @xmath2= 13 .
in contrast , the appearance of 4 vortices as the inner shell occurs only at @xmath2= 14 . at @xmath2= 15
five vortices form a pentagon as the inner shell .
the vortex shells with the pentagon inner ring appear in vorticities up to @xmath2= 17 above which three concentric shells are formed .
these results follow mostly the rule of the shell filling obtained in theoretical and numerical studies @xcite@xcite .
this includes the magic number configuration at @xmath2= 6 where the first formation of two concentric shells occurs , and also the formation of ( 2,7 ) at @xmath2= 9 where two configurations of ( 1,8 ) and ( 2,7 ) are energetically comparable but the latter one becomes the ground state for the large disk @xcite .
however , some differences are found for large vorticities ( @xmath59 ) .
the two shell configurations with 4 vortices in the inner ring are shown to appear for 13 @xmath60 15 , while we observe only the ( 4,10 ) configuration at @xmath61 . moreover , after the pentagon inner shell at @xmath48 16 the formation of three vortex shells is shown to occur at @xmath48 17 , which is not observed above .
a physical origin for the differences would be found in actual vortex configurations observed in the largest disk .
selected images are shown in figs .
3(d)-3(f ) . for clarity
, the corresponding vortex configurations are sketched in figs .
3(g)-3(i ) .
although some distortions are present , one can reasonably find quasi - symmetric shell formation of vortices even in the largest disk .
this is exemplified by figs .
3(d ) and 3(g ) , where three vortices form a triangle in the inner shell surrounded by 8 vortices as the outer shell ring around .
focusing on vortex configurations in the inner shell , we find the formation of a pair , the triangle and a pentagon , of which patterns are similar to those observed for small vorticities ( @xmath62 5 ) given in fig .
however , this is not the case for a square . as shown in figs .
3(e ) and 3(h ) , the vortices in the inner shell form a strongly deformed square like a rhombus . as discussed below , this originates from the effect of the bulk pinning quenched on the @xmath16-moge film .
we often observe a vortex ( or vortices ) close to the rim of the disk for large vorticities , which could be attributed to possible roughness of the disk boundary .
this is visible in the configurations of ( 4,10 ) ( figs .
3(e ) and 3(h ) ) and ( 1,5,13 ) ( figs .
3(f ) and 3(i ) ) .
as denoted by an open circle(s ) in the corresponding sketch , a vortex ( vortices ) appears close to the lower - right rim of the disk .
more rim vortices are occasionally observed when the field - cool procedure is repeated in the same magnetic field , indicating extra vortices trapped in the rim during the field - cool process .
if we neglect the rim vortex ( vortices ) , the differences between the experimental observation and the theoretical prediction in vortex configurations found for @xmath63 would be partly explained . in fig .
3(h ) , the rim vortex becomes a part of the outer shell ring .
if we neglect the vortex , the configuration becomes ( 4,9 ) .
this is consistent with the first configuration for 4 vortices in the inner shell given theoretically @xcite@xcite .
similarity is applicable to the ( 1,5,13 ) configuration shown in fig .
3(i ) , where two rim vortices are recognized .
thus , the configuration could be reduced to ( 1,5,11 ) , which is consistent with the magic number configuration for three vortex shells @xcite@xcite .
further studies would be needed to clarify the origin and the role of the rim vortices on the vortex configurations in details .
repeating field - cool measurements over some field range , we observe configurations with different vorticities at the same magnetic fields @xcite .
an example obtained on the 34 @xmath5 m disk is given to the inset to fig .
although we take the same procedure of the field cooling , both the @xmath2= 5 polygon and @xmath2= ( 1,5 ) shell are observed at the fields of @xmath64 10.4 and 10.9 , resulting in some field crossover in the transition between the @xmath2= 5 and @xmath2= ( 1,5 ) states .
similar crossover may be present in every step in the stair - case behavior of @xmath2 given to fig .
2(a ) and fig .
out of over 20 disks measured in this study , we found that some disks have a relatively strong pinning site(s ) @xcite , although no defect is visible under an optical microscope .
the pinning center(s ) is probably introduced during the sample preparation and it originates from the slight compositional inhomogeneity or subtle roughness in the disk .
the influence of the pinning site(s ) on vortex configurations depends highly on where it is present in the disk . as an example
, we present the results ( for @xmath2 up to 13 ) obtained on a 56 @xmath5 m disk with a pinning site near the disk center .
5(a)-5(i ) show selected vortex images observed in the disk .
we find the pinning site from the image at @xmath2= 1 shown in fig .
5(a ) , where a vortex is unnaturally off - centered in the disk @xcite .
this is against the geometrical confinement and it should be attributed to the influence of the pinning center .
thus , we identify the vortex position as the pinning site and mark a red dot on the image . no more pinning center is found in the disk .
the rest of the images indicate clearly how other vortices are distributed in the presence of the vortex trapped at the pinning site as marked by the red dot .
5(b ) shows the image of two vortices in the disk : one of them sits at the pinning site , while the other appears in the opposite side with respect to the disk center .
we do not observe any orientational changes in the pair even when the magnetic field is varied .
similarity is applicable to the image of a triangle configuration given to fig . 5 ( c ) .
thus , the influence of the pinning site fixes the vortex patterns and it breaks the rotational degree of freedom in the vortex configurations . the deformation of vortex polygons occurs for vorticities more than 4 . fig .
5(d ) shows the image of 4 vortices in the disk .
the vortices form a rhombus - shaped configuration , where the diagonal distance between the pinned vortex and the diagonal vortex is shorter than that for the other two vortices .
this is similar in shape to the disordered square in the inner shell observed in the largest disk ( see figs .
3(e ) and 3(h ) ) .
a unique configuration of vortices is observed in the image at @xmath2= 6 . as shown in fig .
5(f ) , one can see two rows of three vortices like a dice pattern of six with the pinned vortex in the middle of the upper row . this pattern is stable against some field change and no shell pattern
is observed at @xmath2= 6 .
a precursor of the formation of the rows can be seen in the image at @xmath2= 5 ( fig .
5(e ) ) , where the vortices form a pentagon with deformation inward by the pinned vortex ( the top of the pentagon ) and higher right three vortices try to be aligned .
the formation of the ( upper ) row remains in the image at @xmath2= 7 ( fig .
thus , the pinning site near the disk center triggers off the formation of the vortex row and stabilizes the dice like vortex pattern at @xmath2= 6 . because of the dice pattern of vortices at @xmath2= 6 , the first formation of two concentric shells occurs at @xmath2= 7 .
this is larger than @xmath2= 6 where the magic number configuration for two concentric shells is observed in the `` pin - free '' disks .
thus , the shell filling rule is altered by the pinning site near the disk center . for @xmath65 7 ,
the vortex shells are observed with the pinned vortex being a part of the inner shell ring .
as in the image of fig .
5(h ) , the pinned vortex is involved in the formation of a pair as the inner shell , while the other vortices form the outer vortex ring around . the corresponding sketch given in fig .
5(k ) allows us to identify the configuration as ( 2,6 ) . to our knowledge , this configuration has not been reported in other studies so far . for larger vorticities ,
we observe two shell configurations of ( 2,7 ) , ( 2,8 ) and ( 3,8 ) which are similar to those observed in the largest disk .
it is interesting to show the image at @xmath2= 13 given in fig .
5(i ) , where 4 vortices form the inner shell surrounded by 9 vortices as the outer shell .
as sketched in fig .
5(l ) , the vortices in the inner shell form a rhombus - shaped , deformed square , while the outer shell vortices form a ring around . again , focusing on the inner shell , we find that the deformation from the regular square is weaker than that observed at @xmath2= 4 ( fig .
5(d ) ) and the effect of the pinning site is opposite ; namely , the pinned vortex and the diagonal vortex gives the longer diagonal of the rhombus .
thus , the vortex square is stretched by the influence of the pinning site .
we note that this effect depends highly on where the pinning site is present and how large the square polygon is .
the regular square may appear if the pinning site is located on the circumference of the ring defined by the regular square vortices .
all the vortex configurations observed in the disk can be characterized by a line symmetry , defined by the pinned vortex and the disk center * o * , as sketched in fig .
drawing the line in the sketch of fig .
5(l ) , for instance , one can find that the vortex configuration is divided into halves that are nearly the mirror configurations of each other .
similarity is applicable to the other configurations including unique ones at @xmath2= 6 and 8 , although slight deviation is present at @xmath2= 5 and 7 .
note that the odd ( even ) number of vortices lies on the line when @xmath2 is odd ( even ) . in
the `` pin free '' disks no unique line symmetry characterizing all the vortex patterns with different vorticities appears .
thus , we are convinced that the line symmetry in the vortex configurations is induced by the off - centered pinning site in the disk .
a physical reason for the symmetry is not simple since the vortex patterns are determined by subtle balance of many competing interactions , including the interaction with the bulk pinning , the repulsive interaction between vortices , and the interaction with the disk boundary . to clarify this issue ,
a numerical simulation study would be needed .
let us estimate roughly the pinning force @xmath66 for the vortex at the pinning site .
we find that at 4.5 @xmath5 t , just below the step field between @xmath48 1 and @xmath48 2 states , the disk has only a single vortex , but it appears at the disk center .
this is distinct from the observation made at lower fields of 3.5 @xmath5 t and 4.0 @xmath5 t , where the vortex appears at the off - centered pinning site ( fig .
these facts indicate the important influence of the force on the vortex due to the geometrical confinement , and suggest that this force should be comparable to the pinning force in between 4.0 @xmath5 t and 4.5 @xmath5 t. a model described in refs .
@xcite , @xcite and @xcite shows that the force @xmath67 of vortex interactions with the shielding current and edge for a given magnetic field is written as @xmath68\cdot\frac{r}{d_d/2}\ ] ] where @xmath11 is the distance of the vortex position from the disk center and @xmath69
n / m ) is the unit of the force for a vortex per unit length . using the distance ( @xmath70 12 @xmath5 m ) for the off - centered vortex , we find the pinning force as @xmath71 1.8 @xmath72 .
this result , together with our observation where the pinning site accommodates only one vortex , are consistent with the numerical simulation made at the weak pinning force of @xmath73 2 @xcite .
further measurements are made on other disks with a pinning site near the disk edge .
it turns out that the pinned vortex near the disk edge gives only some distortion on the vortex ring(s ) .
no unique configuration is observed except for a highly off - centered vortex configuration at @xmath2= 1 .
these facts indicate that the pinning site near the disk center provides more strong influence upon the vortex configurations than that near the disk edge .
comparing the results in the present experiment to those found in the bitter decoration experiment @xcite , we find that both experiments share the general features of vortex configurations including the shell filling rule and the magic number , although magnetic fields and disk sizes are different in decades .
let us shortly discuss the experimental conditions made on these studies . in the decoration experiment the symmetric vortex shells were observed in smaller disks with 1 - 5 @xmath5 m diameter at higher magnetic fields ( @xmath4 mt ) . due to the small disk size , the geometrical confinement is strong and it dominates over the influence of the strong bulk pinning in the nb film .
meanwhile , in the present experiment , the disk size is @xmath4 10 times larger and the field range is two decades smaller . the corresponding average vortex spacing is @xmath4 10 @xmath5 m , which is larger than the effective penetration depth .
thus , the interaction between vortices as well as the influence of the edge current are weaker , but they remain effective in the vortex configurations because the bulk pinning in the amorphous film is weak .
this situation is validated in the observation of a reasonably ordered configuration of vortices in @xmath16-moge thin films at a low field of 10 @xmath5t@xcite . moreover , normalizing the disk size by the effective penetration depth , we find that our disk size ( 6 - 35 @xmath74 ) is not much different from that of 9 - 46 @xmath74 for the decoration experiment .
these facts not only link two experiments made in different sizes and different field ranges , but also confirm that our observation is made in the large _ mesoscopic _ regime .
we presented the scanning squid microscope images of vortices observed in the weak pinning , @xmath16-moge thin disks with different diameters .
the observed images illustrate clearly geometry - induced , ( quasi-)symmetric configurations of vortices for @xmath2 up to 19 . for small vorticities up to 5
, the vortices form the symmetric polygons with respect to the disk center .
the polygon size grows with @xmath2 , while it shrinks by the enhancement of the screening current .
some of the polygons rotate clearly with respect to the disk center , indicating the rotational degree of freedom in the vortex configurations .
the influence of the geometrical confinement is evident in the quantitative agreements between the field evolution of the polygon size and the theoretical considerations in the london limit . for large vorticities more than 6
, the vortices form the concentric shell rings .
the field evolutions of the inner and outer shell rings follow the rule of the shell filling obtained in the theoretical and numerical studies .
some differences are found in vortex configurations for large vorticities ( @xmath75 13 ) observed in the largest disk , and they could be explained partly by the influence of the weak bulk pinning and/or the edge roughness in the disk .
we also presented the vortex images obtained in the disk with the off - centered pinning site .
the results reveal the pinning induced , novel configurations of vortices : in addition to the deformed polygons , we observed unique vortex configurations of the dice pattern at @xmath2= 6 and ( 2,6 ) at @xmath2= 8 .
these lead to the pinning induced changes in the shell filling rule and the magic number .
all the configurations observed in the disk are nearly symmetric with respect to the line defined by the pinned vortex and the disk center .
n. k. thanks o. narikiyo , h. kawai , k. kadowaki , b. j. baelus , and v. r. misko for useful discussion . this work was supported partly by the grant in aid for scientific research from mext ( the ministry of education , culture , sports , science and technology ) , japan . corresponding author : n. kokubo , center for research and advancement in higher education , kyushu university , 4 - 2 - 1 , moto - oka , nishi - ku , fukuoka , fukuoka 819 - 0395 , japan ; e - mail : kokubo(at)rche.kyushu - u.ac.jp .
a. a. abrikosov , sov .
jetp * 5 * , 1174 ( 1957 ) .
a.i . buzdin and j.p .
brison , phys .
a * 196 * , 267 ( 1994 ) .
p. a. venegas and e. sardella , phys .
b * 58 * , 5789 ( 1998 ) .
j. j. palacios , phys .
lett . * 84 * , 1796 ( 2000 ) .
b. j. baelus , l. r. e. cabral , f. m. peeters , phys . rev .
b * 69 * , 064506 ( 2004 ) .
l. r. e. cabral , b. j. baelus , and f. m. peeters , phys .
b * 70 * , 144523 ( 2004 ) . v.
r. misko , b. xu , and f. m. peeters , phys .
b * 76 * , 024516 ( 2007 ) .
l. f. chibotaru , a. ceulemans , v. bruyndoncx , v. v. moshchalkov , nature * 408 * , 833 ( 2000 ) .
l. f. chibotaru , a. ceulemans , v. bruyndoncx , and v. v. moshchalkov , phys .
lett . * 86 * , 1323 ( 2001 ) .
a. k. geim , i. v. grigorieva , s. v. dubonos , j. g. s. lok , j. c. maan , a. e. filippo , f. m. peeters , nature * 390 * , 259 ( 1997 ) .
a. k. geim , s. v. dubonos , j. j. palacios , i. v. grigorieva , m. henini , j. j. schermer , phys .
lett . * 85 * , 1528 ( 2000 ) .
a. kanda , b. j. baelus , f. m. peeters , k. kadowaki , y. ootuka , phys .
* 93 * , 257002 ( 2004 ) .
y. hata , j. suzuki , i. kakeya , k. kadowaki , a. odawara , a. nagata , s. nakayama and k. chinone , physica c * 388 - 389 * , 719 ( 2003 ) .
t. nishio , s. okayasu , j. suzuki , k. kadowaki , physica c * 412 - 414 * , 379 ( 2004 ) .
t. nishio , q. chen , w. gillijns , k. de keyser , k. vervaeke , and v. v. moshchalkov , phys .
b * 77 * , 012502 ( 2008 ) .
m. r. connolly , m. v. milosevic , s. j. bending , j. r. clem and t. tamegai , euro phys . lett . * 85 * , 17008 ( 2009 ) . h. j. zhao , v. r. misko , f. m. peeters , v. oboznov , s. v. dubonos and i.v .
grigorieva , phys .
b * 78 * , 104517 ( 2008 ) .
grigorieva , w. escoffier , v. r. misko , b. j. baelus , f. m. peeters , l. y. vinnikov , and s. v. dubonos , phys . rev
. lett . * 99 * , 147003 ( 2007 ) .
j. r. kirtley , m. b. ketchen , k. g. stawiasz , j. z. sun , w. j. gallagher , s. h. blanton , and s. j. wind appl .
* 66 * , 1138 ( 1995 ) .
we estimate @xmath7 with the dirty - limit expression described in p. h. kes and c. c. tsuei , phys .
b * 28 * , 5126 ( 1983 ) .
the coherence length @xmath76 is estimated by @xmath77 with the second critical field @xmath78 at @xmath79 0 .
b. l. t. plourde , d. j. van harlingen , n. saha , r. besseling , m. b. s. hesselberth , and p. h. kes , phys .
b * 66 * , 054529 ( 2002 ) . v. g. kogan , v. v. dobrovitski , j. r. clem , y. mawatari and r. g. mints , phys . rev .
b * 63 * , 144501 ( 2001 ) .
the rms flux noise for our system referred to the pick - up loop is @xmath4 5 @xmath80 per pixel in the images .
f. tafuri , j. r. kirtley , p.g .
medaglia , p. orgiani , and g. balestrino , phys .
lett . * 92 * , 157006 ( 2004 ) .
t. nishio , s. okayasu , j. i. suzuki , n. kokubo , and k. kadowaki .
b * 77 * , 052503 ( 2008 ) .
we note that the observed orientations are some of the metastable states stabilized by subtle uneven boundary confinement due to the edge roughness and/or possible weak bulk pinning in the disk .
otherwise , during the scanning , the orientation can be aligned along the scan direction of the pick - up loop due to the weak coupling between the loop and the vortices .
we emphasize that the pinning site accommodates only one vortex and others are distributed around in the disk .
this is distinct from phenomena observed in the micrometer - sized nb mesas , where multi - fluxons are observed near the strong pinning site @xcite . | direct observation of vortices by the scanning squid microscopy was made on large mesoscopic disks of an amorphous moge thin film .
owing to the weak pinning nature of the amorphous film , vortices are able to form geometry induced , ( quasi-)symmetric configurations of polygons and concentric shells in the large disks .
systematic measurements made on selected disks allow us to trace not only how the vortex pattern evolves with magnetic field , but also how the vortex polygons change in size and rotate with respect to the disk center .
the results are in good agreement with theoretical considerations for mesoscopic disks with sufficiently large diameter .
a series of vortex images obtained in a disk with a pinning site reveals a unique line symmetry in vortex configurations , resulting in modifications of the shell filling rule and the magic number . #
1@xmath0#1 2 |
one of the long - sought goals in low - temperature physics is the creation of two interpenetrating superfluids .
early efforts were directed at mixtures of helium isotopes .
more recently , following the experiments with bose - einstein condensates ( bec ) of atomic gases @xcite , considerable efforts have been made to create systems where two species of atoms condense simultaneously .
this goal was partially achieved for two different hyperfine spin states of @xmath1rb , which were condensed in the same trap by the technique of sympathetic cooling @xcite .
later the dynamics of the complex relative motion of the condensates has been studied @xcite .
the possibility of the measurement of the relative phase between the two condensates has also been demonstrated @xcite . in these experiments
the two condensates have a substantial overlap , although they do not completely interpenetrate each other in the stationary state .
a similar state called `` spinor condensate '' has been observed for sodium gas @xcite .
theoretical investigation of two - component bose systems has started many decades ago with the construction of the phenomenological hydrodynamic equations in the spirit of the landau - khalatnikov two - fluid model for the one - component bec @xcite .
later , this construction has been put onto a microscopic basis @xcite .
recent experiments with alkali atoms have revived the interest in the theory of such systems .
hartree - fock theory has been successfully tested on the two - component @xmath1rb system @xcite .
the stability @xcite , ground - state properties @xcite and collective excitations @xcite have been studied theoretically by using the gross - pitaevskii equations .
many properties of two - component , or binary , bec can be understood from symmetry arguments . compared to one - component bose superfluids ,
two - component systems have more interesting pattern of symmetry and symmetry breaking .
condensation in binary bose systems corresponds to the spontaneous breaking of _ two _ ( instead of one ) global u(1 ) symmetries .
these symmetries are related , by noether s theorem , to the separate conservation of the number of atoms of each of the two species .
the quantum state of the binary bose system , therefore , is characterized by two phases of the two condensates . correspondingly , the physics of binary bec is also richer than of usual one - component systems .
the effects of a symmetry are often best exposed by violating the symmetry explicitly in a controlled fashion . a very interesting feature , specific to systems consisting of atoms of the same isotope in different spin states , is that it is possible to couple two condensates by a driving electromagnetic field tuned to the transition frequency . in this case atoms
can be interconverted between the two spin states and the numbers of atoms of each species are not conserved separately anymore ; only the total number of atoms is constant .
this implies that , in the presence of the coupling drive , only one u(1 ) symmetry remains exact , the other one is explicitly violated .
the preserved u(1 ) symmetry obviously comes from the conservation of the total number of atoms , and corresponds to changing the phases of the two condensates by the same amount ( i.e. , leaving the relative phase unchanged ) .
the violated u(1 ) corresponds to changing the relative phase between the two condensate .
the presence of the coupling drive lifts the degeneracy of the ground state with respect to the relative phase . in this work
, we show that a sufficiently small violation of the u(1 ) symmetry corresponding to the relative phase leads to the existence of a nontrivial static configuration a domain wall inside which the relative phase changes by @xmath0 . this configuration is a local minimum of the energy
. however , the domain wall is _ not _ topologically stable and can `` unwind '' itself . to unwind , however , the system must overcome an energy barrier .
thanks to this fact , the rate of the spontaneous decay of the domain wall is exponentially suppressed .
our paper is organized as follows .
section [ sec : l ] introduces the field - theoretical description of binary bec . in sec.[sec : solution ] we describe the domain wall configuration , whose physical interpretation is given in sec .
[ sec : interpretation ] .
section [ sec : boundary ] deals with the boundary of finite domain walls and the related phenomenon of `` vortex confinement '' .
section [ sec : conclusion ] contains concluding remarks . in appendix
[ sec : stability ] we find the metastability condition for the domain wall in the particular case when the densities of the two components are equal , and in appendix [ sec : decay ] two different mechanisms for the decay of the domain wall , operating at different temperature regimes , are considered .
in this section , we use field theory to describe general properties of binary bec .
our goal is to introduce notations and the formalism to lay the ground for the discussion of the domain walls in the next section .
a binary dilute bose system is described by a quantum field theory of two complex scalar fields @xmath2 and @xmath3 .
these fields have the meaning of the wave functions of the two condensates .
the dynamics of these fields is governed by the following lagrangian , @xmath4 where the hamiltonian @xmath5 has the form @xmath6 in eq .
( [ l ] ) @xmath7 are the chemical potentials of the two species , and @xmath8 are functions of coordinates .
we assume here that the trapping potentials are sufficiently wide so that these chemical potentials can be put to constants . ]
@xmath9 is the scattering amplitude , in the zero momentum limit , between an atom of the @xmath10-th species and that of the @xmath11-th species , and are proportional to the scattering lengths @xmath12 , @xmath13 and @xmath14 is the rabi frequency arising from the coupling drive . by varying the action @xmath15 with respect to @xmath16 ,
the familiar gross - pitaevskii equations are directly obtained : @xmath17 let us start by finding the ground state when the coupling drive is off , @xmath18 . in the superfluid ground state ,
both @xmath2 and @xmath3 have nonzero expectation values
. these can be found by minimizing the potential energy part in eq .
( [ l ] ) with respect to @xmath2 and @xmath3 .
this minimization procedure gives the equations determining the densities @xmath19 , @xmath20 in terms of the chemical potentials @xmath21 and @xmath8 , @xmath22 more conveniently , one could view eq .
( [ nmu ] ) as the equations fixing the chemical potentials for given values of the densities . strictly speaking , eqs .
( [ nmu ] ) only correspond to an _ extremum _ of the potential energy . for it to be a local minimum , the quadratic form @xmath23 needs to be positive definite : @xmath24 in fact , eq .
( [ stability ] ) is the condition for the mixture of the two bose superfluids to be thermodynamically stable against segregation @xcite . in this paper
we shall assume that ( [ stability ] ) is satisfied . in principle
, the constants @xmath25 , @xmath26 and @xmath27 can be arbitrary . in this paper
we limit ourselves to the regime when all three scattering lengths are close to each other , @xmath28 . in the case
when the two species are rubidium atoms in different hyperfine states , these lengths were found experimentally to differ by no more than a few percent . the assumption that @xmath29 also introduces considerable technical simplifications in our treatment .
we introduce the `` average '' scattering amplitude @xmath30 and the deviations from the average @xmath31 so that @xmath32 .
the stability condition ( [ stability ] ) implies that @xmath33 .
analogously , we introduce the average scattering lengths @xmath34 and the deviations @xmath35 and @xmath36 .
note that in the limit @xmath37 the hamiltonian has an su(2 ) symmetry which leads to interesting implications @xcite . with the lagrangian at hand
, the discussion of symmetry in the introduction can be made concrete . in the absence of the coupling drive , @xmath18 , the lagrangian ( [ l ] ) possesses a u(1 ) @xmath38 u(1 ) symmetry with respect to independent phase rotations of the fields , @xmath39 the corresponding conservation laws are those of the numbers of particles of each species , @xmath40 and @xmath41 . that @xmath42 and @xmath43 at @xmath18 are conserved separately is actually a basic assumption made when we wrote down the lagrangian ( [ l ] ) .
this assumption is not automatically satisfied : it requires that only elastic scattering between atoms is allowed ; inelastic scattering is forbidden . for binary bose systems made of rubidium atoms
, this appears to be a good approximation @xcite . once the coupling drive is turned on ( @xmath44 ) , the lagrangian is invariant only under a subset of the original u(1 )
@xmath38 u(1 ) rotations ; namely , those which rotate both @xmath2 and @xmath3 by the same angle , @xmath45 therefore one of the u(1 ) symmetries the system enjoyed at @xmath18 is explicitly violated . applying the goldstone theorem ,
we conclude that , at @xmath18 , there are two gapless excitations and only one of these modes remain gapless at @xmath44 .
the gapless modes at @xmath18 are the phonons of the two types of sounds .
one corresponds to the ordinary density wave ( b mode in our paper , see below ) , and another to the concentration wave ( a mode ) in which the densities of the two species oscillate relative to each other in such a way that the total density remains constant .
if we view the two components as being made of the same atoms , but in different hyperfine levels , then the mode a can be alternatively interpreted as a spin polarization wave , in which the density of nuclear spin is a function of space and time .
when the coupling drive is on , only the density wave remains gapless ; the phonon of the concentration ( spin density ) wave is gapped .
let us compute the sound speeds at @xmath18 . to this end
we write @xmath2 and @xmath3 as @xmath46 and expand eq .
( [ l ] ) to second order of @xmath47 and @xmath48 ( we will see that @xmath49 ) .
we find @xmath50 ( we have thrown away total derivatives ) .
the density fluctuations @xmath51 can be `` integrated out '' and replaced by the saddle point values @xmath52 . as a result ,
( [ lnphi ] ) becomes @xmath53 thus , the dispersion relations for the phonons are linear , @xmath54 , and the sound speed @xmath55 satisfies the characteristic equation @xmath56 when @xmath29 , the solutions are @xmath57 we see that when @xmath58 the speed of the concentration wave ( a ) is much smaller than that of the density wave ( b ) , @xmath59 .
the system is `` stiffer '' towards density fluctuations than towards fluctuations of concentration . that modes a and b are indeed concentration and density fluctuations respectively is seen from the corresponding eigenvectors .
the a mode corresponds to such fluctuations in which @xmath60 while the b mode corresponds to @xmath61 therefore , in the a sound @xmath62 and @xmath63 fluctuate in such a way that the overall density remains constant ( @xmath64 ) , while the b sound corresponds to density waves in which @xmath65 , or concentration , is unchanged ( @xmath66 ) .
the lagrangian ( [ lphi ] ) , in terms of the normal modes @xmath67 has the form @xmath68 + \left [ u_\b^{-2}(\d_t\phib)^2-(\nabla\phib)^2\right ] \biggr\}\,.\ ] ] when the coupling drive is on , @xmath44 , one should add to eq .
( [ lphi ] ) the potential energy term @xmath69 .
the b sound is not affected by this term since it corresponds to such fluctuations in which @xmath70 .
the phonons of the a sound acquire a gap @xmath71 for small values of @xmath14 . in our further discussion we shall need the formulas for the healing , or correlation , lengths , which are defined via the response of the system to a static source coupled locally to the particle density , @xmath72 . as in the case of the sounds , there are also two healing lengths , @xmath73 as seen from eq .
( [ coh_lengths ] ) , @xmath74 .
this is because @xmath75 is the correlation length of fluctuations of the overall density , while @xmath76 is the correlation length of the relative density . if we take the values typical for present day experiments : @xmath77 @xmath78 , @xmath79 , and @xmath80 , then @xmath81 @xmath82 m , and @xmath83 @xmath82 m . these lengths are smaller than the typical system size in experiments .
the existence of the domain wall of relative phase in binary bec with a coupling drive can be shown in a rather simple way .
let us first focus only on fluctuations of the fields on length scales much larger than the largest healing length @xmath76 . in this case
, the amplitudes @xmath62 and @xmath63 of @xmath16 can be regarded as `` frozen '' and the only important degrees of freedom are the phases , i.e. , @xmath84 and @xmath85 .
the energy of the system is a functional of @xmath84 and @xmath85 , @xmath86 = \int\ ! d^3\x\,\biggl [ { \hbar^2\over2m}(n_1(\nabla\varphi_1)^2 + n_2(\nabla\varphi_2)^2 ) - \hbar\omega\sqrt{n_1n_2}\cos(\varphi_1-\varphi_2 ) \biggr]\ , .
\label{h}\ ] ] the potential energy term @xmath87 has its minimum at @xmath88 .
the configuration in which @xmath89 over the whole space is obviously the global minimum of the total energy , and hence is the ground state .
the domain wall solution that we shall describe is , in contrast , a _ local _ minimum of the energy . to find the profile of the domain wall , we vary eq .
( [ h ] ) with respect to @xmath84 and @xmath85 and obtain the following equations @xmath90 for domain walls , @xmath84 and @xmath85 are functions of only one coordinate , say , @xmath91 , and at @xmath92 both approach constant values . a nontrivial solution to eq .
( [ walleq ] ) satisfying these conditions is @xmath93 the characteristic width of the domain wall is @xmath94 .
the tension of the domain wall is @xmath95 the relative phase @xmath96 changes from 0 to @xmath0 as @xmath91 runs from @xmath97 to @xmath98 .
note that @xmath96 is defined modulo @xmath0 so it goes a full circle as one passes through the wall .
therefore from the point of view of the energy functional ( [ h ] ) the domain wall ( [ profile ] ) is a topologically nontrivial configuration , which can not be continuously deformed into the @xmath89 configuration .
in fact , one can prove that the domain wall is a configuration with minimal energy defined in eq .
( [ h ] ) among those where @xmath96 changes by @xmath0 from @xmath99 to @xmath100 , and hence , can not decay away , _ as long as eq .
( [ h ] ) applies_. the domain wall is similar to the soliton of the sine - gordon model .
there is a small difference : the ground states on two sides of the domain wall are _ different _ : at @xmath99 @xmath101 , while the ground state at @xmath100 is @xmath102 , @xmath103 .
since @xmath104 are defined mod @xmath0 the latter is equivalent to @xmath105 . in reality , eq .
( [ h ] ) is not the full hamiltonian of the system : it is only an effective description valid at length scales larger than both healing lengths .
the full theory ( [ h - full ] ) contains , besides @xmath84 and @xmath85 , also the density fluctuations @xmath106 and @xmath107 . as a consequence ,
in the full theory , the domain wall ( [ profile ] ) _ can _ be continuously deformed into the trivial configuration @xmath89 .
such deformations necessarily pass through field configurations where either @xmath62 or @xmath63 vanishes at some points : at these points @xmath96 is ill - defined .
thus , the domain wall is not truly topological and can `` unwind '' , i.e. , decay away . the fact that the ground states on the two sides of the wall are different does not prevent the decay : it is possible to construct a field configuration interpolating between @xmath101 and @xmath105 with arbitrarily small energy per unit area .
although the domain wall is not a global minimum of the energy functional , it can still be a _
local _ minimum . in this case , to deform the domain wall into a `` topologically '' trivial configuration with @xmath89 one has to overcome an energy barrier .
the wall is in this case metastable . from our previous discussion
one can conclude that , roughly speaking , the wall is metastable when eq .
( [ h ] ) applies and is unstable when eq .
( [ h ] ) is not applicable . for eq.([h ] ) to be valid the wall has to be wider than the largest healing length , @xmath76 .
since the width of the wall decreases as one increases the rabi frequency @xmath14 , the wall is metastable only when @xmath14 is smaller than some critical value @xmath108 .
let us define @xmath109 as the value of the rabi frequency at which the width of the wall @xmath94 , as defined in eq .
( [ profile ] ) , is equal to to the longer healing length @xmath76 in eq.([coh_lengths ] ) : @xmath110 the wall is metastable when @xmath14 is less than some critical value @xmath108 of order @xmath109 , @xmath111 to find the exact value of @xmath108 one needs to perform a more refined stability analysis .
we present such an analysis for the special case @xmath112 ( i.e. , when the densities of the two species are equal ) in appendix [ sec : stability ] .
parametrically , the result is consistent with eq .
( [ omegac ] ) , and the ratio @xmath113 is found to be @xmath114 .
using the numerical values typical for experiments with rb : @xmath115 a few @xmath116 @xmath78 , @xmath117 , one finds @xmath118 hz . therefore to have a stable wall the rabi frequency needs to be smaller than about 100 hz .
the domain wall can not be thinner than the longer correlation length @xmath76 , which was estimated above to be a few @xmath82 m .
even when @xmath119 , a metastable wall can still spontaneously decay ( burst ) .
such a decay , as we have said , requires overcoming a potential barrier . at sufficiently large temperature , the mechanism for the decay is thermal activation ( over the barrier ) . at zero or small temperatures ,
the mechanism of the decay is quantum tunneling under the energy barrier .
the decay through thermal activation , which is relevant for the temperatures achieved in current experiments , is considered in appendix [ sec : decay ] , where it is shown that the decay rate is exponentially suppressed , since a global ( macroscopic ) fluctuation is required , unless @xmath14 is very close to @xmath108 .
the domain wall solution found in sec . [
sec : solution ] allows an interesting physical interpretation which suggests a possible way for their creation in experiments @xcite .
we first note that in a bec the superfluid velocity is proportional to the gradient of the phase . in a two - component bec , there are two such velocities .
@xmath120 equation ( [ walleq ] ) , in particular , implies that the total particle number current vanishes , @xmath121 .
individually , however , the particle number current of each species @xmath122 and @xmath123 are nonzero . from eq.([profile ] ) we find @xmath124 thus the domain wall is a configuration where the two components flow in opposite directions .
the flow is illustrated in fig.[fig : flow ] .
the velocities of the components are largest at the center of the wall ( @xmath125 ) and decrease as one moves toward the edge of the wall .
the flow is concentrated on the wall ; outside the wall ( @xmath126 ) there is essentially no flow . equation ( [ walleq ] ) can be rewritten in terms of the currents as @xmath127 for a stationary configuration as the one we are considering , eq.([conversion ] ) means that the number of particles in each species is not conserved .
it also implies that there is a _ conversion _ of atoms between the two energy levels due to the coupling drive . in the left half of the wall @xmath128 and there is a conversion of atoms of the second type to atoms of the first type . in the right half @xmath129 and
the conversion goes the opposite way ( fig.[fig : flow ] ) .
the rate of conversion is @xmath130 as is the flow , the conversion rate is also maximal near the wall . far from the wall ( @xmath126 ) there is essentially no conversion . the conversion rate ( [ conversionrate ] ) changes sign at @xmath131 .
since different species correspond to different energy levels of the atom , energy is absorbed in one half of the wall and released in the other .
plane accompanied by spin flips on both sides of it . ]
the interpretation of the domain walls given above suggests a possible method for their creation in experiments @xcite .
one starts with the coupling drive off ( @xmath18 ) and prepares a state where the two condensates flow in opposite directions ( for example , by manipulating the traps ) .
in such a state the relative phase @xmath132 is a linear function of the coordinate along the direction of motion ( say , @xmath91 ) .
one then slowly increases @xmath14 .
the domain walls will be created and the centers of the walls are located at the points where @xmath96 was an odd multiple of @xmath133 ( @xmath134 , @xmath135 , @xmath136 , etc . ) before @xmath14 was turned on . by changing the velocity of the initial relative motion of the condensates and the final values of
@xmath14 one can change the separation between the domain walls and their width .
such controlled creation of the domain walls , hopefully , can be achieved in future experiments .
so far we have always considered infinite domain walls which have no boundary .
it is also interesting , and perhaps more realistic , to consider domain walls with a boundary .
we shall show that the domain wall can be bounded by a vortex line .
suppose we have a finite domain wall whose boundary is a closed contour @xmath137 ( fig .
[ fig : boundary ] ) .
we shall assume that the length of @xmath137 is much larger than the width of the wall @xmath94 so one can view the domain wall as an infinitely thin membrane stretched on @xmath137 ( we shall call this picture the `` thin - wall approximation '' ) .
let us now take another , smaller contour which has a nontrivial linking with @xmath137 ( @xmath138 in fig .
[ fig : boundary ] ) .
as one goes along @xmath138 , one crosses the membrane once , so the relative phase @xmath96 changes by @xmath0 .
this is exactly what one expects from a vortex .
therefore , @xmath137 can be a vortex line .
we recall that the size of the core of the vortex is the healing length @xmath76 , which is smaller or of the same order as the width of the wall .
therefore , in the thin wall approximation , we have an infinitely thin membrane bounded by an infinitely thin vortex line .
one should note that such a bounded domain wall will tend to shrink to reduce its energy , which comes from the wall tension and the tension of the boundary vortex .
conversely , for small non - zero @xmath14 , a vortex must have a domain wall attached to it to minimize the energy due to the nontrivial phase @xmath96 winding around it .
this means that the energy of a single vortex per unit length is increasing linearly with the size of the system in the transverse direction .
this is in contrast to the situation at @xmath18 when the vortex tension has only a logarithmic dependence on the size of the system .
note that there are two types of vortices in binary bec .
those of the first type , which we shall call the @xmath84 vortices , have the condensate @xmath2 vanishing at the vortex center , while @xmath3 is nonzero . analogously
the @xmath85 vortices have @xmath139 and @xmath140 at their centers . as one goes around a @xmath84 vortex , @xmath84 changes by @xmath0 , while @xmath85 does not change , and vice versa for a @xmath85 vortex .
thus @xmath96 changes by either @xmath0 or @xmath141 for the two types of vortices , so the domain wall can be bounded by a vortex of either type .
in contrast to an individual vortex , a pair of @xmath84 and @xmath85 vortices , placed parallel to each other , will have energy per unit length which is only logarithmically divergent .
that is because the @xmath96 `` charges '' of the two vortices cancel each other , so @xmath96 is trivial at spatial infinity ( no winding ) .
the same situation occurs for a pair of parallel vortices of the same type ( @xmath84 or @xmath85 ) with opposite winding ( however , such vortex - antivortex pair can annihilate , while a @xmath142 pair can not ) . in a certain sense
, one can talk about the phenomenon of `` vortex confinement '' : vortices exist only in pairs .
this confinement should , in principle , be observable experimentally , in a rotating two - component bec , which can be already created in a laboratory @xcite .
the vortices are usually identified by the density depletion at their cores , but can be also seen as dislocations of interference fringes due to phase singularities @xcite . with the coupling drive off ( @xmath18 )
such a system contains an equal number of @xmath84 and @xmath85 vortices which are distributed in space with no particular correlation between @xmath84 and @xmath85 vortices .
as one turns on @xmath14 , the vortices will start to pair up and at some point the system will become a collection of composite objects , each being a bound state of a @xmath84 and a @xmath85 vortex ( fig .
[ fig : pairs ] ) .
the phenomenon of vortex confinement is very similar to that of quark confinement in the theory of strong interaction ( quantum chromodynamics ) .
similarly to our vortices , quarks and antiquarks do not exist as individual objects , but are confined into composite objects hadrons .
the analogy with quantum chromodynamics actually stretches further .
if one places a @xmath84 vortex and a @xmath85 vortex at a distance larger than @xmath94 , then a domain wall that connects these two vortices will be formed .
the tension of the domain wall is the force , per unit length , that attracts the two vortices .
the attractive force between the two vortices is thus independent of their separation , given that the latter is larger than the width of the domain wall .
this is analogous to the confining force between a quark and an antiquark , which is also constant at large distances . a confinement model which resembles most
the confinement of the vortices is the three - dimensional compact quantum electrodynamics considered in ref . @xcite . in this theory
the worldlines of electrically charged particles are analogous to the vortices in bec .
one can also imagine a system of two vortices which rotate around each other so that the confining force ( from the domain wall ) balances the centrifugal force .
such a system is analogous to the high - spin meson states in hadronic physics where a quark and an antiquark rotate around each other .
we have shown that in a system of two interpenetrating bec with a coupling drive , there exists a domain wall solution .
the relative phase between the two condensates changes by @xmath0 as one travels through the wall . the wall solution is formally similar to the kink in the sine - gordon field theory , yet it is not topologically stable and can decay . in this respect , the wall is more similar to a soap film , which can spontaneously burst . from the mathematical point of view , the domain walls discussed in this paper are similar to the ones which have been studied in particle physics .
such domain walls appear at least in two contexts : in the theory of the hypothetical axion @xcite and in high - density quark matter @xcite .
the similarity is that the domain wall solution arises from the spontaneous breaking of an approximate u(1 ) symmetry . in all cases
the domain wall exists only if the explicit violation of the u(1 ) symmetry ( determined by the value of the rabi frequency in our case ) is small enough .
the decay of the wall in all the examples occurs via hole nucleation . as to the experimental realization of the domain wall
, one could be optimistic since the estimated critical value of the rabi frequency , for densities and scattering lengths typical for the rubidium gas in recent experiments , is of the order 100 hz which is not too small .
the width of the wall , which might be as small as a few @xmath82 m if the rabi frequency is not much smaller than critical , can be also accommodated inside condensates of the size characteristic of present - day experiments .
the apparent immediate obstacle is still the creation of a system where two bec truly interpenetrate . in the recent experiments with trapped atomic gas
the region of overlap between the two bec is still small .
one can hope the technical problems of making a genuine two - component bec to be solved in the near future which would enable one to study the domain walls experimentally .
the authors thank richard friedberg for discussions .
we thank riken , brookhaven national laboratory , and u.s .
department of energy [ de - ac02 - 98ch10886 ] for providing the facilities essential for the completion of this work .
the work of dts is supported , in part , by a doe oji grant .
in the case when the densities of the two components are equal , @xmath112 , the maximal frequency @xmath108 , below which the wall configuration is still locally stable , can be found analytically .
for illustrative purposes we shall consider this particular case in details .
let us recall that when the scattering lengths @xmath144 , @xmath145 , and @xmath146 are approximately equal ( assuming @xmath147 ) there are two healing lengths @xmath76 and @xmath75 .
the healing length related to fluctuations of the overall density @xmath75 is much smaller than the one related to fluctuations of the relative density @xmath76 : @xmath148
. we shall be interested in the case when the width of the wall @xmath94 is much larger than @xmath75 ( but we shall not presume any relation between @xmath94 and @xmath76 ) . in this case
, the total density @xmath149 can be considered as frozen , and the system can be described in terms of three variables : @xmath150 , @xmath84 and @xmath85 , @xmath151 where @xmath150 runs from @xmath152 and @xmath133 . in terms of these three variables , the lagrangian has the form @xmath153-h\ , , \label{lthetaphi}\ ] ] where @xmath154 is the following hamiltonian @xmath155 + { 1\over2}(\delta\mu n - \delta g ' n^2)\cos\theta \nonumber\\ & & - { 1\over4}\delta gn^2\sin^2\theta - { 1\over2}\hbar\omega n\sin\theta\cos(\varphi_1-\varphi_2)\ , . \label{hthetaphi}\end{aligned}\ ] ] in eq .
( [ hthetaphi ] ) @xmath156 .
the ground state is found by minimizing the potential term in eq .
( [ hthetaphi ] ) . from here on we consider the special case when in the ground state the density of atoms of the two species are equal , @xmath157 ( or equivalently @xmath158 . )
this requires @xmath159 . in this case
@xmath154 possesses a discrete symmetry with respect to replacing @xmath160 , @xmath161 .
this symmetry is what makes it possible to find @xmath108 analytically .
it is more convenient to use , instead of @xmath104 , the normal modes @xmath162 defined in eqs .
( [ phiab ] ) , which in the case @xmath157 have the form @xmath163 in terms of these variables the hamiltonian becomes @xmath164 \nonumber \\ & & - { 1\over4}\delta gn^2\sin^2\theta - { 1\over2}\hbar\omega n\sin\theta\cos\phia\ , . \label{hphiab}\end{aligned}\ ] ] in order to find the domain wall configuration , we need to extremize the energy with respect to @xmath150 and @xmath165 .
varying with respect to @xmath166 , one finds @xmath167 this equation determines @xmath166 for given @xmath150 and @xmath96 .
the task of solving eq .
( [ phibthetaphia ] ) becomes much simpler if one assumes that all variables depend only one coordinate @xmath91 . in this case
@xmath168 which can be trivially solved : @xmath169 .
after eliminating @xmath166 , the energy functional one has to minimize is @xmath170 -{1\over4}\delta gn^2\sin^2\theta - { 1\over2}\hbar\omega n\sin\theta\cos\phia\ , . \label{hphia}\ ] ] it is easy to check that the following configuration is always a local extremum of eq .
( [ hphia ] ) : @xmath171 equation ( [ extremum1 ] ) can be guessed from the symmetry of eq.([hphia ] ) under @xmath160 . to see
if the domain wall solution is a local minimum , one needs to expand @xmath154 in the vicinity of ( [ extremum1],[extremum ] ) .
one writes @xmath172 to the second order in @xmath173 and @xmath174 the hamiltonian ( [ hphia ] ) is @xmath175 + { \hbar^2n\over8 m } \left[(\nabla\tilde\phia)^2 + k^2\cos\bar\phia\cdot\tilde\phia^2\right]\ , .
\label{h2ndorder}\ ] ] one has to find the eigenmodes of eq .
( [ h2ndorder ] ) : if there are no negative modes then the domain wall is a local minimum of the energy ; if there exist a negative mode then the domain wall is unstable . the second , @xmath150-independent , term in eq .
( [ h2ndorder ] ) does not have negative modes ( it has only one zero mode corresponding to the translation of the wall along the @xmath91 direction ) and does not lead to instability , and hence can be ignored . taking into account the explicit solution @xmath176 , the first term in eq.([h2ndorder ] ) is @xmath177\,.\ ] ] the well - known operator @xmath178 has the lowest eigenvalue equal to @xmath179 , corresponding to the eigenfunction @xmath180 , which implies that @xmath154 does not have a negative mode if @xmath119 , where @xmath181 when @xmath182 , the configuration ( [ extremum ] ) is not a local minimum of the energy functional : the domain wall does not exist .
the value ( [ omegac_exact ] ) is of the same order as @xmath109 in eq .
( [ omega0 ] ) : @xmath183 .
the numerical value @xmath184 is specific for the case @xmath112 ; if @xmath185 then @xmath113 is , in general , different .
as we have seen above , as long as @xmath119 , the wall minimizes the energy of the system with respect to small _ local _ variations of the condensates .
however , the global minimum of the energy is achieved when the phases of the condensates are constant in space . since ,
as discussed above , the wall configuration can be continuously deformed into the global minimum energy configuration , i.e. , the wall can decay . here
we shall estimate the lifetime of the wall due to such a decay .
since the wall minimizes the energy locally , such a deformation necessarily goes through a potential barrier .
thus the decay can only occur by a quantum tunneling through this barrier or , at finite temperature , by a thermal fluctuation over the barrier . in both cases ,
the decay rate is exponentially suppressed : in the first case by the wkb factor @xmath186 , and in the second case by the boltzmann factor @xmath187 , where @xmath188 is the height of the barrier .
the first formula applies at sufficiently low temperature , while the second one applies at higher temperatures .
a crude estimate ( see below ) suggests that the corresponding crossover temperature is quite small ( nanokelvins for parameters typical for present - day experiments ) , and is lower than temperatures achieved in present - day experiments .
thus we shall limit our discussion to the decay by thermal fluctuation .
this case is also simpler theoretically .
the quantum mechanical decay , relevant for very low temperatures , will be left beyond the scope of this paper .
assuming the temperature is much smaller than the critical temperature at which one of the condensates melts ( so that most atoms are still in the condensates ) , the rate of thermal activation across the barrier is @xmath189 , where @xmath188 is the height of the barrier at zero temperature .
since we are dealing with an infinite - dimensional configuration space , @xmath188 should be understood as the energy at the lowest point of the barrier .
this point is a saddle point ( the energy has a single negative curvature direction ) .
some information about the exponent @xmath190 can be obtained by a simple scaling argument without a detailed calculation . in terms of the dimensionless variables
@xmath191 defined as @xmath192 the energy becomes @xmath193 = { \hbar^3n^{1/2}\over m^{3/2}\delta g^{1/2}}\int\!d^3\widetilde \x\ , & & \biggl\ { { 1\over2 } \biggl [ { 1\over4}(\tilde\nabla\theta)^2 + \cos^2{\theta\over2}(\tilde\nabla\varphi_1)^2 + \sin^2{\theta\over2}(\tilde\nabla\varphi_2)^2\biggr ] \nonumber\\ & & -{1\over4 } ( \sin^2\theta+2\cos\theta_0\cos\theta ) -{1\over2 } { \hbar\omega\over\delta gn } \sin\theta\cos(\varphi_1-\varphi_2)\biggr\}\ , , \label{hrescale}\end{aligned}\ ] ] where @xmath194 determines the relative condensate densities in the ground state at @xmath18 : @xmath195 .
consider the dependence on @xmath14 at a given @xmath194 .
the energy functional eq.([hrescale ] ) is equal to a dimensionful constant times a dimensionless functional which depends on @xmath14 only via the ratio @xmath196 , where @xmath197 , same as in ( [ omega0 ] ) .
thus the saddle point of the energy is also a function of this dimensionless ratio : @xmath198 if , moreover , we define a temperature @xmath199 as @xmath200 which is of the same order as the critical temperature of the bose - einstein condensation , then the decay rate for a given @xmath65 ( or @xmath194 ) can be written as @xmath201 = \exp\biggl[-{1\over\sqrt{4\pi \delta an^{1/3 } } } { t_0\over t } f\biggl({\omega\over\omega_0}\biggr)\biggr]\ , .
\label{decayscaling}\ ] ] the form of the function @xmath202 can not be found from scaling arguments alone .
however , when @xmath203 one can expect @xmath204 , and then the exponent @xmath205 is large , since @xmath206 , and @xmath207 .
the function @xmath202 can be computed in the regime @xmath208 , where the saddle point configuration can be found in the `` thin - wall '' approximation . in this approximation field configurations
are described at length scales much larger than the width of the domain wall . from this point of view
the domain wall is an infinitely thin membrane .
the saddle point configuration is a membrane with a round hole in it ( fig .
[ fig : hole ] ) .
the radius of the hole @xmath209 must be much larger than the width of the wall for the thin wall approximation to be valid . as discussed in sec .
[ sec : boundary ] , the rim of the hole must be a vortex , since it is the boundary of the domain wall .
there are two contributions to the energy difference @xmath188 between the saddle point and the domain wall configurations .
one contribution is negative and comes from the hole ( since the hole is the absence of the wall ) .
another contribution is positive and comes from the rim .
therefore , @xmath210 where @xmath211 is the energy per unit length of the vortex ( the vortex tension ) and @xmath212 is the domain wall tension .
the energy ( [ enusigma ] ) has a maximum when the radius of the hole is @xmath213 this is the radius of the critical hole .
indeed , if a hole with a radius @xmath214 is nucleated , then it will expand and eventually eat up the whole wall . if , in contrast , the radius of the nucleated hole is less than critical , then the hole will shrink and disappear . substituting @xmath215 in eq .
( [ enusigma ] ) , we find the height of the energy barrier @xmath216 as there are two types of vortices , in eq .
( [ barrier ] ) @xmath211 refers to the vortex with the smaller tension .
the tension of a straight vortex is logarithmically divergent : @xmath217 where the index @xmath218 refers to the two types of vortices , and @xmath209 is the long - distance cutoff .
the role of @xmath209 is played by either the size of the critical hole @xmath219 or the width of the wall @xmath94 .
we shall see at the end of this section that the two lengths differ only by a logarithm , which does not affect eq .
( [ nu_i ] ) .
therefore , to logarithmic accuracy , the vortex tension is @xmath220 the argument of the logarithm is @xmath221 , and is large when @xmath208 , which justifies the use of the logarithmic approximation . without loss of generality
, we can assume that @xmath222 .
then the vortex of the first type has the smallest @xmath211 ( [ nu_i ] ) . substituting eqs.([nui ] ) and ( [ tension ] ) into eq .
( [ barrier ] ) , one finds that the barrier height has the form of eq .
( [ escaling ] ) , where @xmath223 the decay rate ( [ decayscaling ] ) can be rewritten in the following form @xmath224^{2/3 } \biggl({n\over n_2}\biggr)^{3/2 } { 1\over\sqrt{4\pi\delta an_1^{1/3}}}{t_{c1}\over t } \biggl({\omega_0\over\omega}\biggr)^{1/2 } \ln^2{\omega_0\over\omega}\biggr\}\ , , \label{gamma}\ ] ] where we introduced @xmath225 the critical temperature for the smallest of the two condensates ( @xmath62 by our choice ) .
the rate ( [ gamma ] ) is exponentially suppressed when @xmath208 and @xmath226 . to check the consistency of our assumptions , we note that the radius of the critical hole is @xmath227 comparing with the width of the wall in eq .
( [ profile ] ) , we find that @xmath219 is larger than @xmath94 by a factor @xmath228 , which is assumed to be parametrically large .
thus , the use of the thin wall approximation is justified . to make a good estimate of the crossover temperature , below which crossover proceeds via quantum tunneling , rather than via thermal activation
, one needs to estimate the action on the tunneling trajectory and compare it to the exponent in the thermal activation rate @xmath190 . for a very crude estimate we can take @xmath229
the exponents @xmath230 and @xmath190 become comparable at a temperature of order few nanokelvin , given typical parameters of present day experiments . | we consider a system of two interpenetrating bose - einstein condensates of atoms in two different hyperfine spin states .
we show that in the presence of a small coupling drive between the two spin levels , there exist domain walls across which the relative phase of the two condensates changes by @xmath0 .
we give the physical interpretation of such walls .
we show that the wall tension determines the force between certain pairs of vortices at large distances .
we also show that the probability of the spontaneous decay of the domain wall is exponentially suppressed , both at finite and at zero temperature , and determine the exponents in the regime of small rabi frequency .
we briefly discuss how such domain walls could be created in future experiments . |
with the growth of internet the relationship between online information and financial markets has become a subject of ever increasing interest .
online information offers with respect to its origin and purpose and reflects either interest of some profile of users in the form of query or knowledge about certain topic in the form of news blogs or reports .
financial markets are strongly information - driven and these effects can be seen by studying either search query volumes or social media sentiment .
many studies have analysed the effects of search query volumes of specific terms with movements in financial markets of related items @xcite .
bordino et al .
@xcite show that daily trading volumes of stocks traded in nasdaq 100 are correlated with daily volumes of yahoo queries related to the same stocks , and that query volumes can anticipate peaks of trading by one or more days .
dimpfl et al .
@xcite report that the internet search queries for term `` dow '' obtained from google trends can help predict dow jones realized volatility .
vlastakis et al . @xcite study information demand and supply using google trends at the company and market level for
30 of the largest stocks traded on nyse and nasdaq 100 .
chauvet et al .
@xcite devise an index of investor distress in the housing market , housing distress index ( hdi ) , also based on google search query data .
preis et al .
@xcite demonstrate how google trends data can be used for designing a market strategy or defining a future orientation index @xcite . in principle , different effects between information sources and financial markets are expected considering news , blogs or even wikipedia articles @xcite .
andersen et al .
@xcite characterize the response of us , german and british stock , bond and foreign exchange markets to real - time u.s . macroeconomic news .
zhang and sikena exploit @xcite blog and news and build a sentiment model using large - scale natural language processing to perform a study on how a company s media frequency , sentiment polarity and subjectivity anticipate or reflect stock trading volumes and financial returns .
chen et al .
@xcite investigate the role of social media in financial markets , focussing on single - ticker articles published on seeking alpha - a popular social - media platform among investors .
mao et al .
@xcite compare a range of different online sources of information ( twitter feeds , news headlines , and volumes of google search queries ) using sentiment tracking methods and compare their value for financial prediction of market indices such as the djia ( dow jones industrial average ) , trading volumes , and implied market volatility ( vix ) , as well as gold prices .
casarin and squazzoni @xcite compute bad news index as weighted average of negative sentiment words in headlines of three distinct news sources .
the idea of cohesiveness of news as a systemic financial risk indicator is related to recent works studying mimicry and co - movement in financial markets as phenomena reflecting systemic risk in financial systems @xcite .
harmon et al .
@xcite show that the last economic crisis and earlier large single - day panics were preceded by extended periods of high levels of market mimicry - direct evidence of uncertainty and nervousness , and of the comparatively weak influence of external news .
kennet et al .
@xcite define an index representing the balance between the stock correlations and the partial correlations after subtraction of the index contribution and study the dynamics of s&p 500 over the period of 10 years ( 1999 - 2010 ) .
the idea of cohesiveness as a measure of news importance is simple : if many sources report about same events then this should reflect their importance and correlate with the main trends in financial markets .
however , in order to capture the trends of systemic importance one must be able to track different topics over majority of relevant online news sources .
in other words , one needs : ( i ) an access to the relevant news sources and ( ii ) a comprehensive vocabulary of terms relevant for the domain of interest .
we satisfy the second prerequisite for systemic approach through the use of large vocabulary of financial terms corresponding to companies , financial institutions , financial instruments and financial glossary terms . to satisfy the first prerequisite , in our analysis we rely on financial news documents extracted by a novel text - stream processing pipeline newstream @xcite ( ` http://newstream.ijs.si/ ` ) , from a large number of web sources .
these texts are then filtered and transformed into a form convenient for computing nci for the particular period of time .
we show that importance of financial news can be measured in a more systemic way than via sentiment towards individual entities or number of occurrences of individual terms , and that strong cohesiveness in news reflects the trends in financial markets .
there is already a strong evidence linking co - movement of financial instruments to systemic risk in financial markets @xcite .
we hypothesize that cohesiveness of financial news reflects in some part this systemic risk .
our news cohesiveness index ( nci ) captures average mutual similarity between the documents and entities in the financial corpus .
if we represent documents as sets of entitites then there are two alternative views on similarity : ( i ) two documents are more similar than some other two documents if they share more entities and ( ii ) two entities are more similar than some other two entities if they co - occur in more documents .
we construct nci so that the overall similarity in a corpus of documents is equal regardless on the view we choose to adopt .
we analyse the nci in the context of different financial indices , their volatility , trading volumes , as well as google search query volumes .
we show that nci is highly correlated with volatility of main us and eu stock market indices , in particular their historical volatility and vix ( implied volatility of s&p500 ) .
furthermore , we demonstrate that there is a substantial difference between aggregate term occurrence and cohesiveness in their relations toward financial indices .
in order to measure the herding effects in financial news we introduce a news cohesiveness index ( nci ) - a systemic indicator that quantifies cohesion in a collection of financial documents . a starting point for the calculation of nci is a _ document - entity matrix _ that quantifies occurrences of entities in each individual document collected over certain period of time .
we use concept of entity ( instead of e.g. term ) to represent different lexical appearances of some concept in texts . in our case
we use a vocabulary of entities that includes financial glossary terms , financial institutions , companies and financial instruments .
the full taxonomy of entities is available in supplementary information section 3 .
we start with the definition of occurrence , which says whether some entity is present or not in some document , regardless of how many times it occurs in the document .
this makes document - entity matrix @xmath0 a binary matrix : @xmath1 @xmath0 is an @xmath2 matrix , where @xmath3 is the number of documents published in selected time period and @xmath4 is the total number of entities we monitor .
document - entity matrix @xmath0 also corresponds to a biadjacency matrix of a bipartite graph between documents and entities .
an edge between document @xmath5 and entity @xmath6 exists if the entity @xmath6 appears in the document @xmath5 .
the overall similarity in the collection of documents should be equal regardless whether we choose to view it as the similarity between the _ documents _ or between the _
entities_. to achieve this we define the similarity as the _ scalar product _ of either document pairs @xmath7 or entity pairs @xmath8 .
now we define nci as a frobenius norm of scalar similarity matrix between all pairs of documents @xmath9 or entities @xmath10 : @xmath11 where @xmath12 is either @xmath13 or @xmath14 .
frobenius norms of both document - document similarity matrix @xmath15 and entity - entity similarity matrix @xmath16 are equal , therefore the cohesion is conserved whether we measure it as the _ document _ or the _ entity _ similarity : @xmath17 in the network representation these two similarity matrices correspond to two projections of a bipartite graph of the original document - entity matrix , as illustrated in figure [ fig : projection - matrix - network - representation ] .
moreover , one can exploit properties of the frobenius norm of scalar similarity matrix and express cohesiveness as a function of singular values of document - entity matrix @xmath0 ( proof in the supplementary information section 2 ) : @xmath18 where @xmath19 are the @xmath20 largest singular values of matrix @xmath0 in a singular value decomposition : @xmath21 because singular values are calculated on the original document - entity matrix and not its document or entity projections , we claim that we capture an _ intrinsic _ property of the corresponding document - entity matrix that is invariant to projection .
this can also be inferred from the fact that the eigenvalues of similarity matrices @xmath22 and @xmath23 are equal and they correspond to the singular values of document - entity matrix @xmath0 .
this approach can be beneficial for large document - entity matrices as it is much more efficient in terms of time and memory compared to explicit calculation of similarity matrix .
we can calculate just first @xmath20 values incrementally , until we reach the desired accuracy of nci ( see supplementary information section 1 ) . in practice ,
only a small number of singular values is enough to calculate nci up to the desired precision .
as the number of documents is changing each day while the number of entities stays constant , all nci indices in our analyses are normalized with the number documents @xmath24 in the corpus .
we have statistically confirmed that the nci is largely above the level of fluctuations of cohesiveness random null model ( see supplementary information section 2 ) .
sometimes it is interesting to perform detailed analysis of which groups of entities or documents contribute the most to the overall cohesiveness . for this purpose we can divide entities or documents into groups using any appropriate semantic criteria and calculate cohesiveness for each group separately or between pairs of groups .
semantic partitions in the entity projection are created via grouping of entities in mutually disjoint groups , defined by their taxonomy labels ( hence semantic interpretation ) . on the other hand ,
semantic partitions in the document projection can be created via grouping of entities either by their temporal or source membership .
figure [ fig : semantic - partitioning ] illustrates the concept of partitioning in the context of different projections .
we can calculate cohesiveness separately for each semantic group or a combination of semantic groups .
note that even in this case we do not need to explicitly calculate similarity matrices ( see supplementary information section 1 ) .
following the taxonomy of entities described in supplementary information section 3 we defined four semantic groups : companies , regions , financial instrument and euro crisis terms .
figure [ fig : semantic_components_occurrences ] shows the most frequent entities in each of the semantic partitions , based on the news corpus collected over the period of analysis . in order to asses
nci s utility as a systemic risk indicator , we use correlations analysis and granger causality tests against the pool of financial market and information indicators .
the analysis should also provide deeper insight into the interplay between news , trends in financial markets and behaviour of investors .
we adopt terminology from @xcite , and treat our news based indicators ( nci variants and entity occurence ) as indicators of the information supply in online media , while volumes of google search queries will be treated as indicators of information demand or as a proxy of investor interest .
we group indicators as follows : * * inormation supply indicators : * - cohesiveness index based on all the news ( nci ) from newstream , cohesiveness index based only on filtered financial news ( nci - financial ) from newstream , total entity occurrences based on the aggregate from all news documents , and total entity occurrences based on strictly financial documents of newstream . * * information demand indicators : * - these are volumes of google search queries ( gsq ) for 4 finance / economy related categories from google finance ( from google domestic trends - finance&investment , bankruptcy , financial planning , business ) . * * financial market indicators : * - these include daily realized volatilities , historical volatilities and trading volumes of major stock market indices ( s&p 500 , dax , ftse , nikkei 225 , hang seng ) as well as implied volatilities of s&p500 ( vix ) .
details on preparation of individual indicators are given in methods section .
we start the analysis with a simple comparison of nci calculated on all news and nci calculated on filtered financial news .
figure [ fig : nci_vs_indices ] shows dynamics of nci , and nci - financial in comparison to vix ( implied volatility of s&p 500 , the so called `` fear factor '' @xcite ) .
scatter plots on the right show that correlation of vix and nci - financial is significantly higher than vix and nci .
this is a first illustration of the importance of the filtering the right content for the construction of indicators from texts . for more details on how filtering affects correlations with other indices
see supplementary information section 3 .
figure [ fig : corr_matrix ] shows pearson correlation coefficients between different information indicators and financial market indicators .
corresponding p - values are calculated using a permutation test and are available in supplementary information section 5 .
all correlations reported in this article have p - value @xmath25 unless explicitly stated .
interesingly , the correlations between total entity occurrences , nci and nci - financial are relatively low , confirming that cohesiveness captures very different signal from the entity occurrences .
furthermore , correlations between total entity occurrences , nci and financial indices are , on average , much lower than correlations between nci - financial and financial indices .
relatively low correlation between nci - financial and nci confirms importance of filtering out strictly financial market - related articles from the newstream , rather than having all the articles that contain some of the entities from the vocabulary .
we have performed a more detailed analysis of these effects by studying in parallel behavior of different variants of entity occurrences and nci - financial using different subsets of the vocabulary and the document space , independently .
the main insight gained was that entity occurences become more informative when a smaller vocabulary of the most frequent entities is used , but this requires use of the whole document space .
nci has proven to be much more robust to the choice of both vocabulary and document space ( details in supplementary information section 6 ) .
interestingly , the nci - financial index is highly correlated with implied volatility ( @xmath26 , figure [ fig : corr_matrix ] ) , as well as with historical and daily realized volatilities ( @xmath27 , figure [ fig : corr_matrix ] ) .
these correlations are much higher than the correlations of the gsq categories ( @xmath28 , figure [ fig : corr_matrix ] ) .
in contrast to nci - financial , gsq categories exhibit relatively stronger correlations with stock trading volumes ( @xmath29 , figure [ fig : corr_matrix ] ) .
google bankruptcy and google unemployment are significantly correlated with nci - financial ( correlation above 0.2 , figure [ fig : corr_matrix ] ) , which is most probably due to similarities in vocabulary used in constructing nci - financial and respective gsq indicators .
a more in depth picture of the news cohesiveness index is obtained when observing individual semantic components of nci - financial and their correlation patterns with financial and google search query indicators .
semantic components based on ` [ region ] ` and ` [ eurocrisis ] ` taxonomy categories all have similar correlation patterns to nci - financial ( with correlation above 0.7 for ` [ eurocrisis ] ` and above 0.5 for ` [ region ] ` , figure [ fig : corr_matrix ] ) ; this also shows that these categories are most important for the behavior of nci - financial . on the other hand , semantic components based on ` [ company ] ` and ` [ instrument ] ` exhibit quite different , in many parts , opposite correlation patterns ( with correlations close to 0 or even negative ) .
it is interesting to note that both the nci - financial and gsq indicators have strong negative correlation with nikkei 225 volatility and trading volume ( up to -0.4 for nci - financial and up to -0.5 for gsq - unemployment ) .
the granger - causality test ( gc test ) is frequently used to determine whether a time series @xmath30 is useful in forecasting another time series @xmath31 .
the idea of the gc test is to evaluate if @xmath31 can be better predicted using both the histories of @xmath31 and @xmath30 rather than using only the history of @xmath31 ( i.e. @xmath30 granger - causes @xmath31 ) .
the test is performed by regressing @xmath31 on its own time - lagged values and on those of @xmath30 included .
an f - test is used in examining if the null hypothesis that @xmath31 is not granger - caused by @xmath30 can be rejected . in table
[ fig : table_gc_bidirect ] we show results of pairwise g - causality tests between information supply and demand indicators and financial indicators .
cells of the table give both directionality ( @xmath32 , @xmath33 or bidirectional @xmath34 ) and significance at two levels of f - test ( p - values @xmath35 ;
@xmath36 ) . besides gc testing nci - financial and its semantic components at higher taxonomy levels ,
we show also results obtained for nci ( non - filtered news nci ) and total entity occurences as a baseline . the results in table [ fig : table_gc_bidirect ] paint a much different picture than the correlation study .
firstly , granger causality seems to be almost exclusively directed from financial to information world , with single bidirectional exception between ` [ region]x[eurocrisis ] ` semantic component of nci - financial and hang seng daily realized volatility .
our financial news indicator nci - financial seems to be g - caused solely by ftse daily volatility .
this finding is in contrast with the fact that the nci - financial is strongly correlated with several other indicators like implied volatility ( vix ) ( @xmath26 , table [ fig : corr_matrix ] ) .
however , two of the semantic components ` [ eurocrisis]x[eurocrisis ] ` and ` [ region]x[eurocrisis ] ` are strongly g - caused by implied volatility , historical and daily volatilities of most of the major stock market indices .
on the other hand , the gsq categories seem to be mostly gc driven by trading volumes , almost exclusively of us and uk financial market ( s&p 500 and ftse ) .
gsq indicators seem to be divided in two groups by their g - causality : ( i ) those that are g - caused mainly by trading volumes ( business and industrial , bankruptcy , financial planning and finance and investment ) and total entity occurrences in the news , and ( ii ) those that are strongly g - caused by all other gsq categories ( unemployment ) .
interestingly , total entity occurrence in the news , seem to be the strongest g - causality driver of the gsq volumes , while two of the semantic components of nci - financial are g - caused by gsq finance and investment and financial planning .
this work introduces a new indicator of financial news importance based on a concept of cohesiveness of texts , from large corpora of news and blogs sources .
in contrast to indicators introduced by other authors which are based on sentiment modelling , nci measures cohesiveness in the news by approximating the average similarity between texts .
our correlation results confirm the main hypothesis that cohesiveness of the financial news is a signal that is strongly correlated with systemic financial market indicators in particular volatilities of major stock exchanges .
the analysis of granger causality tests over a pool of financial and information related indicators suggests that nci - financial is mainly related with the volatility of the market . in our analysis
most important semantic components of nci - financial are mainly g - caused by implied ( vix ) , historical and daily volatilities .
this implies effects from both short term and long term risks in the financial market .
the only exception ( bidirectional causality between _ [ region]x[eurocrisis ] _ and hang seng daily volatility ) might be plausibly explained as a time zone effect .
this does not seem to be the case for gsq indicators which are mainly driven by trading volumes , with the exception of gsq unemployment , which seems to be driven mostly by search volumes of other gsq categories .
similar to findings of some previous studies @xcite , in which aggregate sentiment or financial headline occurrence were used as measures of state of the financial market , information supply co - movement as measured by nci - financial , seem to be primarily caused by trends on the financial market rather than the opposite .
we find that similar results holds also for the gsq categories which approximate information demand side in our case .
g - causality patterns show , similarly to correlation , that cohesiveness captures quite different signal with respect to total entity occurrence ; the results also suggest the presence of somewhat circular interplay between information supply and information demand indicators .
for example , total entity occurrence is g - causing three of the gsq categories ( business and industry , bankruptcy and financial planning ) , while financial planning and unemployment are g - causal for semantic components ` [ instrument]x[eurocrisis ] ` and ` [ eurocrisis]x[eurocrisis ] ` , which suggests feedback mechanisms between news and search behaviour . in comparison with the findings of studies which used simpler measures of news importance or sentiment
, we find that financial news cohesiveness reflects the level of the volatility in the market and is gc driven both by current level of volatility and implied volatility , while gsq volumes are driven mainly by trading volumes .
impact of news cohesiveness and gsq volumes in the reverse direction , as determined by gc tests , is only weakly implied in case of semantic components of nci - financial and hang seng index .
this is not in line with previous works that report predictive utility ( mostly for gsq volumes ) with respect to certain financial instruments .
however , one has to bear in mind that the results of gc tests reflect average of lagged correlations between indicators over the specific period in time ( in our case oct 2011 - jul 2013 ) .
it is possible that direction of causality between information and financial indicators changes in time , but this was hard to detect in our data due to the limited length of time series .
another possible reason for different results is that most of previous works were based on a limited number of google search query terms , typicaly more closely related to the particular stock market index of interest . in principle , this is different from volumes of gsq term categories in our case , which reflect aggregates over larger number of different query terms .
gsq categories closely resemble the concept of semantic components and it is possible that the application of the concept of cohesiveness , if adapted to gsq category volumes , may produce signals more predictive with respect to financial market trends .
access to structured information about the financial markets with its various instruments and indicators is available for several decades , but systematic quantification of unstructured information hidden in news from diverse web sources is of relatively recent origin .
we base our analyses on a newly created text processing pipeline - newstream @xcite , designed and implemented within the scope of eu fp7 projects first ( ` http://project-first.eu/ ` ) and foc ( ` http://www.focproject.eu/ ` ) .
newstream continuously downloads articles from more than 200 worldwide news sources , extracts the content and stores complete texts of articles .
it is a domain independent data acquisition pipeline , but biased towards finance by the selection of news sources and the taxonomy of entities relevant for finance . for the purpose of filtering , efficient
storing and analytics , expert based financial taxonomy and vocabulary of entities and terms have been created , containing names of relevant financial institutions , companies , finance and economics specific terms , etc .
the newstream pipeline has been collecting data since october 2011 . in our analyses
we use text corpora from october 2011 to june 2013 and we have filtered over 1,400,000 financially related texts stored in the form of document - entity matrices .
full structure of the taxonomy is in the supplementary information section 3 , and the list of the domains from which most documents were downloaded in the supplementary information section 4 . [
[ filtering - of - financial - documents ] ] filtering of financial documents + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + newstream pipeline downloads articles from more than 200 web sites of online news and blogs .
moreover , despite the selection of financial news sites , there are many articles which are only indirectly related to finance , such as politics or even sport . to obtain a clean collection of strictly financial texts , we have developed a rule - based model utilizing taxonomy categories as features to describe documents , and a gold standard of human labelled documents ( @xmath37 3500 documents ) . a machine learned rule - based model is used as a filter for extracting strictly finance related texts from a corpus .
this model has a recall of over 50% , with precision of well over 80% .
the rule - based model for filtering financial documents can be found in supplementary information section 3 .
[ [ financial - indicators ] ] financial indicators + + + + + + + + + + + + + + + + + + + + we analyse nci in comparison to the financial market indicators of worldwide markets and google search query volumes . for that purpose
we have downloaded stock market indices from yahoo finance web service ( ` http://finance.yahoo.com/ ` ) : high , low , open , close prices and volume of s&p 500 , dax , ftse , nikkei 225 and hang seng index .
we also use implied volatility of s&p500 ( vix ) .
implied volatily is calculated for the next 30 days by chicago board options exchange ( cboe , ` http://www.cboe.com/ ` ) using current prices of indices options .
historical ( realized ) volatilities are calculated from the past prices of the indices themselves .
we use daily prices of individual indices to calculate a proxy of daily realized volatility .
@xmath38 historical ( realized ) volatilities are calculated as standard deviations of daily log returns in the appropriate time window : @xmath39 where @xmath40 are daily prices , and @xmath4 is time window . in our analyses
we used a window of 21 working days . [
[ google - search - query - volumes ] ] google search query volumes + + + + + + + + + + + + + + + + + + + + + + + + + + + almost all previous studies used search query volumes of specific terms .
instead , we used google search query volumes of predefined term categories from google finance website .
we have chosen five categories from google domestic trends that are related to financial market : business and industrial , bankruptcy , financial planning , finance and investing , unemployment .
we downloaded yoy ( year over year ) change values for these categories from google finance web service ( ` https://www.google.com/finance ` ) .
we have used functions of the r packages _ tseries _ , _ lmtest _ , _ vars _ , _ urca _ to download and calculate indices , construct joint time series dataset , determine correlations and study granger causality relations .
we have followed the methodology of toda and yamamoto @xcite for granger causality testing of non - stationary series .
details of the procedure are given in supplementary information section 5 .
10 url # 1`#1`urlprefix[2]#2 [ 2][]#2 , , , & . in _ _ , wsdm 12 , ( , ) .
_ et al . _ . _ _ * * , ( ) . &
_ _ ( ) . & .
_ _ * * , ( ) . , & ( ) .
http://ssrn.com/abstract=2148769 . ,
_ * * ( ) . , , & .
_ _ * * ( ) .
_ et al . _ . _
_ * * ( ) . , , & .
_ _ * * , ( ) . & .
( ) . , , & .
_ _ ( ) . http://ssrn.com/abstract=1807265 . ,
_ _ ( ) .
_ _ * * , ( ) . , , , &
_ _ * * ( ) . , , & .
_ _ * * ( ) . , & .
_ _ * * ( ) .
_ et al . _
( ) . http://ssrn.com/abstract=1829224 .
et al . _ . _
_ * * , ( ) . , & .
_ _ ( ) .
( ) . http://ssrn.com/abstract=2311964 . , &
( ) . http://rady.ucsd.edu/faculty/directory/engelberg/pub/portfolios/fears.pdf .
* * , ( ) .
this work was supported in part by the european commission under the fp7 projects foc ( forecasting financial crises , measurements , models and predictions , grant no .
255987 and foc inco , grant no .
297149 ) and by the croatian ministry of science , education and sport project `` machine learning algorithms and applications '' .
we would like to thank the following people for helpful discussions : stefano battiston , vinko zlati , guido caldarelli , michelangelo puliga , tomislav lipi and matej miheli .
all authors contributed to the writing and editing of the manuscript .
mp , naf and ts carried out the modelling and the analyses .
pkn , i m and mg were involved in gathering and processing of the data .
iv and ts were involved in interpretation of the results .
* supplementary information * accompanies this paper . +
* competing financial interests : * the authors declare no competing financial interests .
+ * dataset availability : * all data and codes that we used in our analysis is freely available on ` http://lis.irb.hr/foc/data/data.html ` .
correspondence to tomislav muc , + email : ` [email protected] ` | motivated by recent financial crises significant research efforts have been put into studying contagion effects and herding behaviour in financial markets . much less has been said about influence of financial news on financial markets .
we propose a novel measure of collective behaviour in financial news on the web , news cohesiveness index ( nci ) , and show that it can be used as a systemic risk indicator .
we evaluate the nci on financial documents from large web news sources on a daily basis from october 2011 to july 2013 and analyse the interplay between financial markets and financially related news .
we hypothesized that strong cohesion in financial news reflects movements in the financial markets .
cohesiveness is more general and robust measure of systemic risk expressed in news , than measures based on simple occurrences of specific terms .
our results indicate that cohesiveness in the financial news is highly correlated with and driven by volatility on the financial markets .
= 1 |
one of the major goals of ultrarelativistic nuclear collision studies is to obtain information on the qcd phase diagram @xcite .
currently , one of the most direct approaches is the investigation of hadron production .
hadron yields measured in central heavy ion collisions from ags up to rhic energies can be described very well @xcite within a hadro - chemical equilibrium model . in our approach
@xcite the only parameters are the chemical freeze - out temperature @xmath0 and the baryo - chemical potential @xmath1 ( and the fireball volume @xmath2 , in case yields rather than ratios of yields are fitted ) .
other approaches @xcite employ ( several ) other , non - thermal , parameters . for a review see @xcite .
the main result of these investigations was that the extracted temperature values rise rather sharply from low energies on towards @xmath310 gev and reach afterwards constant values near @xmath0=160 mev , while the baryochemical potential decreases smoothly as a function of energy .
this limiting temperature @xcite behavior suggests a connection to the phase boundary and it was , indeed , argued @xcite that the quark - hadron phase transition drives the equilibration dynamically , at least for sps energies and above . for the lower energies , the quarkyonic state of matter @xcite could complement this picture by providing a new phase boundary at large @xmath1 values .
the conjecture of the tricritical point @xcite was put forward in this context .
the importance of measurements at very high energies to obtain information on the existence of a limiting temperature of excited hadronic matter produced in nuclear collisions was pointed out early @xcite based on analysis of particle spectra at the bevalac ( see also the review @xcite ) , from pions to heavier complex nuclei . at first glance , it may seem inappropriate to use the chemical freeze - out concept for light nuclei , as their binding energies are a few mev , much less than the chemical freeze - out temperatures of 100 - 170 mev .
we note , however , that the relative yield of particles composed of nucleons is determined by the entropy per baryon , which is fixed at chemical freeze - out .
this has been first recognized already 30 years back @xcite and was subsequently further substantiated in @xcite , constituting the basis of thermal analyses of yields of light nuclei @xcite .
it is entropy conservation , and not the difference between the binding energy and temperature of the system , which governs the production yields in this case after chemical freeze - out , entropy is conserved .
it was also noted then that the yields obtained within the thermal model are in close agreement to those from coalescence models @xcite .
the thermal model studies were already at that time extended to nuclei carrying strangeness ( hyperons in replacement of nucleons ) and even hypothetical objects with roughly equal number of up , down and strange quarks ( strangelets ) . at the same time , a vigorous line of theoretical investigations on the existence of multi - strange hypernuclei , or memos @xcite was established .
recently , the first measurement of the lightest ( anti)hypernucleus , ( anti)hyper - tritium , in high - energy nucleus - nucleus collisions was achieved by the star experiment at the rhic @xcite .
this measurement opens up a very interesting new regime for tests of particle production at chemical equilibrium . at relatively low beam energies , where the baryo - chemical potential and , hence , the baryon density is maximum ( fair energy regime ) objects with a large number of baryons and moderate strangeness
may be abundantly produced @xcite . at rhic and lhc energies production of objects with moderate
( anti)baryon number and large strangeness content may be expected . in this paper
we investigate and predict within the thermal model the production yields of heavy baryons and anti - baryons and in particular of hypernuclei and their antiparticles and confront these calculations with all presently available data ranging from ags to rhic energies .
, @xmath4 , @xmath5 ) , for the rhic energy of @xmath6=200 gev . the ratio @xmath7 , recently measured by the star experiment @xcite is included in the fit . ]
the measurement of the production yields of light nuclei ( and anti - nuclei ) without strangeness in central nuclear collisions provides significant constraints on thermal model parameters , in particular on the value of the baryo - chemical potential @xmath1 .
this is most easily seen when one recognizes that yield ratios such as @xmath8 scale like @xmath9}$ ] . in fig .
[ fig1 ] we show the updated thermal fit to the hadron yield data measured at rhic ( @xmath6=200 gev ) including the newly - measured @xcite yield ratio @xmath7 . including this ratio significantly narrows the range of possible @xmath1 values , while @xmath0 and @xmath2 of the new fit remain unchanged ( @xmath0=164 mev , @xmath2=1960 @xmath10 ) compared to our earlier fit @xcite .
quantitatively , the new fit leads to @xmath1=24@xmath112 mev , while without the ratio @xmath7 , @xmath1=30@xmath114 mev @xcite .
the quality of the present fit is similar to that of the earlier one ( which had @xmath12/dof=29.7/12 ) .
this result supports previous findings at lower energies @xcite .
we stress that the agreement between the experimental value and the calculated one for the ratio @xmath7 is a powerful argument that indeed entropy conservation governs the production of light nuclei .
if one were to use a temperature comparable to the binding energy per nucleon , that is @xmath0=5 mev , the calculated ratio would be 3.1@xmath1310@xmath14 , while it is 0.415 for @xmath0=164 mev , see fig .
[ fig1 ] . in fig .
[ fig2 ] we show that predictions using the thermal model can be used to describe quantitatively the measured energy dependence of @xmath15 , @xmath16 , and @xmath17 yield ratios over a very wide energy range the calculations , here and in the following , are performed using the parametrizations for @xmath0 and @xmath1 established in @xcite based on fits of midrapidity data in central collisions .
the penalty against anti - particles at lower energies , well described within the thermal approach , is also drastically exhibited in this figure .
note , in particular , the very good agreement between the model and the measurements at ags @xcite for the @xmath16 ratio , which extends the the range of the model agreement over 8 orders or magnitude ( for the other measurements see references in @xcite ) .
it is thus natural to extend such analyses to light nuclei containing strangeness .
note that , at the lowest energies , the canonical strangeness suppression is important and is incorporated in our model as described in @xcite . in the following
we will therefore use the thermal model with the parameters as discussed to analyze the production of hypernuclei and their anti - particles and confront model predictions with the now available data .
in table [ tab1 ] we show a comparison of the measured data @xcite and model calculations for the best fit thermal parameters discussed above .
the yield ratio @xmath18 is as well reproduced by the model as the ratio @xmath7 . on the other hand , the measured ratios of ( anti)hyper - tritium to ( anti)@xmath19 are larger than predicted in the model by about two standard deviations , using the statistical and systematic uncertainties quoted for the data
. .ratios at rhic energy , @xmath6=200 gev .
the experimental values are from the star experiment @xcite and contain statistical ans systematic errors .
the errors for the model calculations correspond to the errors of the fit for the baryochemical potential , @xmath20 mev . [
cols="<,^,^",options="header " , ] to shed more light on the situation we turn now to the energy dependence of ( strange ) baryon production . in fig .
[ fig3 ] we show the experimental energy dependence of the @xmath21 and @xmath22 ratios and confront these data with our thermal model predictions .
the degree of agreement between data and calculations is impressive .
we also include in this figure thermal model predictions for the energy dependence of the ratio @xmath23 and @xmath24 .
the broad maximum around @xmath35 gev for the ratio @xmath24 has the same origin as the maximum in the @xmath25 ratio , namely it arises as a consequence of strangeness neutrality condition , imposed in our model , and a competition between rising @xmath0 and decreasing @xmath1 @xcite .
we also note that a slightly less prominent maximum is likely to survive even if one relaxes the condition of strangeness neutrality , as demonstrated in @xcite . at high energies the value for the two ratios approach each other , as expected for decreasing values of @xmath1 at a nearly constant temperature .
ratio ( see text ) . ] in fig .
[ fig4 ] we show the measured energy dependence for the @xmath26 and the @xmath27 ratio .
this double ratio was suggested by the authors of @xcite in the expectation that dividing out the strange to non - strange baryon yield should result in a value near unity .
the data are compared to thermal model predictions .
note that there is negligible feed - down from heavier states into states with baryon number 3 .
as expected , the measured energy dependence of the @xmath26 is well reproduced by the model calculations . on the other hand ,
the discrepancy between thermal model predictions and data for the @xmath27 ratio is apparent ( red line ) .
it is important to realize that the ratio @xmath21 is significantly influenced by feed - down from strong decays of excited baryonic states , leading to a value for the double ratio significantly below unity . the red line in fig .
[ fig4 ] contains such feed - down , the blue dashed line represents a calculation where the feeding is artificially left out .
the feed - down from strong decays increases the @xmath21 ratio and , hence , reduces the overall ratio .
the star collaboration actually measures a ratio close to 1 , above the thermal model prediction by twice the error quoted by the experiment .
interestingly , results from the e864 collaboration @xcite ( as shown in ref .
@xcite ) at ags energy are , albeit with large uncertainties , consistent with the thermal model prediction .
the discrepancy at rhic energy , if experimentally established , would point to a new production mechanism not contained in the thermal approach and not present at lower beam energies .
the possible existence of an excited @xmath28 state of ( anti)hyper - tritium has been recently pointed out to us @xcite . this excited state could contribute via decay to the ground state , and would lead to close agreement between model and data .
further measurements at rhic and , very soon , lhc energy are eagerly awaited to shed light on the situation . yield ratios . ] to complete our studies we show , in fig .
[ fig5 ] , predictions within the thermal model for the energy dependence of the production yield of multistrange light hypernuclei @xcite relative to @xmath29 hyperons .
these ratios exhibit a pronounced maximum in the fair energy regime , which is the consequence of a competition between a strong increase ( followed by saturation ) of @xmath0 and a strongly decreasing @xmath1 ( see also the discussion above ) .
in addition , the canonical suppression , arising from the condition of local strangeness conservation , leads to reduced yields at low energies . in case of hyper - tritium production
, there is no maximum , since it is mainly determined by the strong energy dependence of @xmath1 at low energies .
it is larger at the ( low ) fair energies by two to three orders of magnitude compared to rhic and lhc energies .
even larger are the differences between low and high energies for the production of the exotic multi - hyperon states .
central collisions .
the predicted yields of @xmath30 and @xmath31 nuclei are included for comparison , along with the corresponding anti - nuclei ( dashed lines ) . ] in fig .
[ fig6 ] we show , as a function of energy , predictions of yields at midrapidity per one million central collisions .
the volume at chemical freeze - out is that from our fits of yields @xcite . at fair energy
the production yields of exotic nuclei is maximal , although the absolute yields are still rather small . as an example , for @xmath32 , the rate of production at 10@xmath33 central pb+pb collisions per second is about 60 per month , for a reasonable duty factor of the accelerator . assuming a reconstruction efficiency of the order of a percent
, this implies a few candidates per year of data taking , clearly at the edge of achievability . at the lhc , ( anti-)@xmath34he and their corresponding hypernuclei are experimentally accessible . for the lhc energy of 2.76 tev of the present data taking ,
we predict ratios @xmath35he/@xmath34he and @xmath36/@xmath37 of 2.76@xmath38 and 2.70@xmath38 , respectively ( to be compared to the corresponding values for the rhic energy of 200 gev of 3.13@xmath38 and 2.37@xmath38 ) .
these predictions can also be used as guideline for expectation in pp collisions at lhc energy , where one could estimate , in the grand - canonical limit , yields reduced by a factor of the order of 200 - 400 compared to pb+pb collisions .
this is compensated by the much larger number of pp collisions ( about 10@xmath39 events ) which can be inspected at lhc ( for a running time of 10@xmath40 s per year of operation ) .
this should allow , at the lhc , a measurement of the yields of produced ( anti-)hypernuclei up to mass number 4 in pp and pb - pb collisions and provide a detailed test of our predictions .
we have demonstrated that the yield of light nuclei and their anti - particles are well reproduced with thermal model calculations employing parameters established from the analysis of general hadron production in relativistic nuclear collisions . as shown above
, such ratios can be used to provide a precision constraint of the baryo - chemical potential @xmath1 .
we have furthermore shown that the newly measured yield ratio @xmath18 is also well described with the thermal approach , while the ratio @xmath41 which is reproduced at ags energy is significantly underpredicted at rhic energy .
the origin of this discrepancy is currently not clear and needs further study .
our studies have also indicated interesting energy dependence in such yields and ratios . in particular , particles with large baryon number and moderate strangeness are produced in significant numbers at fair energy .
the hyper - nuclei program , started by the star experiment at rhic , has made these studies very topical .
although significant questions remain , it is clear that the study of the production of complex nuclei with and without strangeness in relativistic nuclear collisions can open a new chapter in the quest to understand the relation of particle production to the qcd phase boundary .
the thermal model predictions can hopefully soon be tested also at the lhc energy with the data already collected in 2010 . | we present , using the statistical model , an analysis of the production of light nuclei , hypernuclei and their antiparticles in central collisions of heavy nuclei . based on these studies
we provide predictions for the production yields of multiply - strange light nuclei . , , , |
in past two decades , the study of heat conduction in low dimensional model systems has enriched our understanding on heat conduction from microscopic point of view@xcite . in turn
, the study has also lead to some interesting inventions for heat control and management devices .
for example , the heat conduction in nonlinear lattice models @xcite demonstrates rectification phenomenon , namely , heat flux can flow preferably in one direction .
furthermore , the _ negative differential thermal resistance _ is also found and based on which , a thermal transistor model has been constructed.@xcite most recently , the two segment model of thermal rectifier proposed in ref has been experimentally realized by using gradual mass - loaded carbon and boron nitride nanotubes.@xcite these works have opened a new era for heat management and heat control in microscopic level .
on the other hand , there have been increasing studies on heat conduction in real nano scale systems@xcite .
for example , the thermal conduction of single walled carbon nanotubes ( swcnts ) has attracted both theoretical @xcite and experimental @xcite attentions . almost all experiments and numerical simulations
have payed their attention to the extremely high thermal conductivity of swcnts .
the dependence of the thermal conductivity on the length or chirality of swcnts is also extensively studied theoretically .
therefore , one may asks , how can we make heat controlling devices , i.e. , the thermal diode or the thermal transistor , from swcnts and their derivatives ? in the practical applications , swcnts may have many kinds of impurities such as element impurities , isotopic impurities@xcite , topological impurities , etc . among these numerous derivatives of swcnts , the swcnt intramolecular junctions ( imjs ) , which are formed by introducing the pentagon - heptagon rings in swcnts ,
have been expected to be an important structure in the applications .
it is well - known that the electronic properties of the swcnt imjs have a close relationship with their geometrical or topological characteristics.@xcite it is reasonable that the thermal transport behavior of swcnt imjs can also be altered by their geometrical characteristics .
compared with the electronic properties , the thermal transport behavior may have not so close relationship with the detailed local geometrical arrangements because it is mainly influenced by the long wavelength phonons .
but the non - equilibrium thermal transport behavior of imjs can be affected by the high - frequency optical phonon modes,@xcite which can reflect the information of the local defects .
so it is interesting to investigate the non - equilibrium thermal transport of different swcnt imjs . in this work ,
the thermal rectification in ( @xmath0 , 0)/(2@xmath0 , 0 ) imjs and its dependencies on tube radius , imj length , and external stress , e.g. , tensile and torsional stress are studied . a complex structure of ` peapod ' structure and imj
is also studied to investigate the effect of periodical potential on the thermal rectification .
the paper is organized as the follows . in sec .
ii we introduce the basic structure of ( n , 0)/(2n , 0 ) imj and our numerical method . in sec .
iii , the main numerical results are discussed . in sec iv
, we show that the external stress can improve the rectification . in the same section
we also show the results of ( n , 0)/c@xmath1@(2n , 0 ) structures , which can be regarded as a combination of peapod structures and ( n , 0)/(2n , 0 ) imjs . in sec .
v , we give some concluding remarks .
a typical ( @xmath0 , 0)/(2@xmath0 , 0 ) imj structure is depicted in fig . 1
, in which the index @xmath0 equals to 8 .
the structure contains two parts , namely , a segment of ( @xmath0 , 0 ) swcnt and a segment of ( 2@xmath0 , 0 ) swcnt . for simplicity , the lengths of the two segments are almost equal in our calculations .
the two segments are connected by @xmath2 pairs of pentagon - heptagon defects . because ( @xmath0 , 0 ) and ( 2@xmath0 , 0 ) tubes have the common rotational symmetry @xmath3 , thus when @xmath0 can be divided exactly by @xmath2 , we can adjust the defects to make the two segments connect straightly . especially , if @xmath4 , the pentagon - heptagon defects can be arranged in the connection part so that the rotational symmetry of the imj is @xmath5 . on the other hand ,
it is necessary to define a basic system length because the thermal conductivity depends on the system length.@xcite after setting the lengths of ( @xmath0 , 0 ) and ( 2@xmath0 , 0 ) tubes to be the same , we can define a basic total length @xmath6 when both of the two segments contain 24 periods .
all the structures are fully optimized before further molecular dynamic ( md ) calculations , and then we can obtain @xmath7 nm .
[ fig1 ] in this kind of structure , the outmost one period of each heads ( colored by blue and marked as region ` i ' in fig . 1 ) are fixed in md process . then
two periods of each ends ( illustrated by orange color and marked as region ` ii ' in fig . 1 ) are put in the heat baths , which are realized by the nos - hoover thermostat.@xcite the temperatures of the thermostats at left and right heads are @xmath8 and @xmath9 , respectively . for convenient , here we introduce two quantities : @xmath10 , and @xmath11 . in this work
, @xmath12 is always kept at 290 k. the c - c bonding interactions are described by the second - generation reactive empirical bond order ( rebo ) potential,@xcite which is the most recent version of the tersoff - brenner type potential , combining advantages of the two sets of parameters in the earlier version.@xcite the velocity verlet method is employed to integrate the equations of motion with the time step of 0.51 fs .
the typical total md process is 5@xmath1310@xmath14 steps , which is about 2.55 ns , and the statistic averages of interesting quantities start from half of the md process , i.e. , 2.5@xmath1310@xmath14 steps are used to relax the system to a stationary state .
the instant temperature of atoms @xmath15 is defined as @xmath16 , where @xmath17 is the time - dependent velocity , @xmath18 the mass and @xmath19 the boltzmann constant .
this is a result of energy equipartition theorem .
the thermal flux is obtained from the thermostats , i.e. , the total work from the thermostats can be regarded as the heat flux runs from thermostats to the system .
if the work is positive , then the heat flux is also positive . in the scheme of the nos - hoover thermostat,@xcite
the equation of motion for the particle @xmath15 in heat bath is : @xmath20 where @xmath21 is the momentum and @xmath22 is the force applied on the atom .
@xmath23 is the number of degrees of freedom of the atoms in the thermostat .
@xmath24 , where @xmath25 is the relaxation time . in our simulation
, @xmath26 is kept as 4 ps . the heat bath acts on the particle with a force @xmath27 , thus the power of heat bath is @xmath28 , which can also be regarded as the heat flux from heat bath , i.e. , @xmath29 the total heat flux from the heat bath to the system can be obtained by @xmath30 , where the subscript @xmath15 runs over all the particles in the thermostat . the final thermal conduction is @xmath31 , where @xmath32 is the heat flux density , @xmath33 the area of cross - section , and @xmath34 the system length .
in fact this definition is equivalent to the usual definition : @xmath35 , @xmath36 is the instant total energy of particle @xmath15 , but eq .
( [ eq2 ] ) is much simpler for computational simulation .
first of all , we investigate the effect of defects in connection region .
the defects in connection region determine the property of the interface between two tubes , and as a result , they will affect the heat transport property of the system .
however , the question whether they can affect the rectification is still open .
we study the ( 8 , 0)/(16 , 0 ) structures as examples . generally speaking
, they are many methods to connect an imj , thus we only consider two conditions here .
the first one is the coaxal straight imjs .
the possible number @xmath2 of pentagon - heptagon defect pairs for a coaxal imj is 8 , 4 or 2 , but @xmath37 makes the cross - section of the tube part highly deformed from circle , so we only consider @xmath38 and 8 .
the second condition is the simplest imj , i.e. , two tubes are connected with each other by using only one pair of heptagon - pentagon defects ( @xmath39).@xcite the top views and side view of the three structures ( @xmath40 , 4 and 1 ) are plotted in fig .
it can be found even in the simplest zigzag / zigzag imj ( @xmath41 ) , there is still a small angle between the two segments .
[ fig2 ] the total heat fluxes @xmath42 versus the normalized temperature difference @xmath43 is plotted in fig .
positive @xmath42 means the heat flux flows from ( @xmath0 , 0 ) tube to ( 2@xmath0 , 0 ) tube .
we do not show the heat flux density here , because it is difficult to define the area of the cross - section of the whole imj .
is the number of pentagon - heptagon defects .
error bars are also plotted.,title="fig : " ] [ fig3 ] it can be found from fig .
3a that the total heat flux for @xmath44 is larger than that for @xmath45 .
it is straightforward because the pentagon - heptagon defects are topological defects , they scatter the phonons and cause larger interfacial thermal resistance . as a result , more topological defects will reduce the thermal conductance and the heat flux .
but the total heat flux for @xmath41 is close to that for @xmath44 .
this is due to the big difference of the connecting region of these imjs .
the length of the connecting region in imj with @xmath46 is much longer than those in the imjs with @xmath38 or 8 .
the long connecting region can cause big thermal resistance , as a result , the thermal resistance in @xmath41 imj is not much smaller than that in @xmath44 imj .
when the temperature difference is very small , the heat flux only shows weak asymmetry .
the asymmetry becomes larger when temperature difference is larger .
this characteristic is seen clearer in fig .
3b , in which the thermal rectification @xmath47 is drawn . in fig .
3b , @xmath48 means that the heat flux is of negative sign , _ vice versa _
, @xmath49 means that the heat flux is of positive sign .
it can be seen that , an increase in temperature gradient @xmath50 leads to an increase of thermal rectification .
this is because the rectification is a result of non - equilibrium transport , which is mainly determined by optical phonons . only when the temperature difference is large enough
, the optical phonons can be excited and contribute to the heat conduction .
more interestingly , when the temperature difference is large enough ( for @xmath51 ) , the rectification begins to decrease .
but , the temperature difference is so large that this result may not be hold in more restrict calculations due to lose of local thermal equilibrium and the quantum effect in the lower temperature head .
another meaningful feature is that , although the absolute values of the heat fluxes are completely different , rectifications are almost the same in imjs with @xmath44 and 8 , whose connecting region is short .
namely , the rectification is weakly dependent on the detailed structure of the interface when the connecting region is short enough .
and the rectification in @xmath41 imj is largest among the three kinds imjs . in order to understand above rectification phenomenon
, we consider the power spectra of the atoms around the connecting part .
the atoms to be studied are illustrated in the fig .
4a . they are selected to be next to the defects along the axes .
the corresponding phonon spectra are presented in the fig .
the overlapping of the phonon spectra of the two atoms around the connecting parts have been emphasized by shadows . ) , where the atoms whose power spectra are recorded are emphasized by balls .
( b ) the corresponding phonon power spectra . the overlapping of the phonon spectra of the two particles around the connecting parts have been emphasized by shadows and bordered by dark lines.,title="fig : " ] [ fig4 ] it is found that when the direction of the temperature gradient is exchanged , the area of the overlap region is also changed .
more importantly , the overlap area is larger when @xmath52 is negative .
this result corresponds with the trend of the heat flux .
so we can state that in this real system , the relationship between the overlap area and the absolute value of the heat flux is same as that in the one dimensional nonlinear lattice systems@xcite , in which it is found that matching / mismatching of the energy spectra around the interface is the underlying mechanism of the rectification . on the other hand , the most obvious change appears at 400@xmath531000 @xmath54 and 1600@xmath551800 @xmath54 .
the @xmath44 imj is also studied , and similar conclusion can be obtained .
so the optical phonon modes are important for the rectification .
in addition , in ref . , the authors show that in carbon nanotubes with finite length where the long - wavelength acoustic phonons behave ballistically , even optical phonons can play a major role in the non - fourier heat conduction .
the dispersion relations of the swnt show that , in the intermediate range of the normalized wave vector @xmath56 , some of the phonon branches , especially the ones with relatively low frequency , have group velocity comparable to the acoustic branches .
furthermore , the thermal conductivity can be expressed as @xmath57 , where @xmath58 is a set of quantum numbers specifying a phonon state , @xmath59 is the phonon mean free path , @xmath60 is the magnitude of the phonon group velocity along the direction of the heat flow , and @xmath61 ^ 2}$ ] is the thermal capacity of lattice wave with wave vector @xmath62 and angular frequency @xmath63 .
thus , the contribution from the high frequency modes is weak because their group velocities are much smaller than those of low frequency modes .
so the modes at 400@xmath551000 @xmath54 are more important than the modes at 1600@xmath551800 @xmath54 .
the frequencies of the low frequency modes have been found to be inversely proportional to the tube radius.@xcite thus , the low frequency modes , e.g. , the radial breathing mode , have different exciting temperatures in ( 16 , 0 ) and ( 8 , 0 ) tubes . on the other words ,
their participation in transport process occurs at different temperatures .
this fact further explains why these low frequency optical phonon modes are important for the rectification . based on above discussions
, now we can understand the following phenomena .
firstly , the rectification almost does not change in the @xmath44 and 8 imjs .
secondly , the rectification in @xmath41 imj is largest in the three imjs .
as has been mentioned , the modes at 400@xmath551000 @xmath64 are most important for the thermal rectification .
these modes have larger wave lengths than the high frequency modes .
the connecting parts of the two imjs with @xmath38 and 8 are short , as a result , the defects have weak effect on the effective optical modes . in ref .
, it is found that the high frequency vibrational modes caused by topological defects are localized states . and the higher frequency localized modes show smaller spatial dimension , which means they can hardly entangle with the transport phonon modes .
thus in the short connecting imjs ( @xmath38 and 8) , the arrangement of the detailed defects has weak influence on the rectification .
in contrast , the connecting region is much longer in the @xmath39 imj , and the distance between the pentagon and heptagon rings is relatively long , which means the connecting region can affect those middle frequency phonon modes , which have longer wave lengths .
therefore , the rectification is largest in the case of @xmath39 imj .
however , the rectification in current structures is still small , which might limit the application of the imj as a thermal rectifier .
so we will try to find the factors which can help improve the rectification .
above all , we investigate the thermal conductance of the imjs with the different index @xmath0 numerically . , 0)/(2@xmath0 , 0 ) swcnt imjs , @xmath0 = 7 , 8 and 9 .
( a ) the heat flux versus temperature difference .
( b ) the thermal rectification versus temperature difference . in this figure , @xmath4 .
error bars are also plotted.,title="fig : " ] [ fig5 ] in order to make the results comparable with each other , the number @xmath2 of pentagon - heptagon defect pairs are set to be @xmath0 . considering the computational consumption , only @xmath65 , 8 and 9 are considered in fig .
5 . according to fig .
5a , the total heat flux increases when @xmath0 increases , but this does not mean that the thermal conductivity for big @xmath0 is also large because it should be scaled by the area of cross - section .
in fact , it has been shown that the thinner swcnts have higher thermal conductivity in literatures .
another interesting feature is that , when index @xmath0 increases , the asymmetric heat flux ratio does not always increase . in fig .
5b , the largest rectification appears when @xmath66 .
so this reminds us that the heat flux rectification is not only induced by the radius difference of two segments .
when tube radius increases to a sufficient large value , i.e. , @xmath0 is large , the vibrational density of states ( vdos ) of carbon nanotube approaches to that of the graphite .
but when @xmath0 is small , the vdos of the swnt deviates from that of the graphite due to the periodical boundary along the circumferential direction and the curvature effect , and the amplitude of the deviation is almost inversely proportion to @xmath0 . as a result , when ( @xmath0 , 0 ) and ( 2@xmath0 , 0 ) tubes connect with each other , the difference between the vdoss of the two tubes is obvious if @xmath0 is small , and _
vice versa _ , the difference is not so obvious if @xmath0 is large .
recalling the fact that the rectification can be related with the mismatching between the vdoss around the connecting region , we can conclude that the rectification of the imj is not obvious when @xmath0 is large . .
@xmath6 is defined in text .
error bars are also plotted.,title="fig : " ] [ fig6 ] next , we investigate the length dependence of heat conduction . from fig .
6a , it can be seen that the total heat flux decreases as the system length increases from @xmath6 to @xmath67 .
however , the thermal conductivity , @xmath68 , increases with the increasing of the system length .
this result agrees with the previous studies .
moreover , there still exists asymmetry in the heat flux , which is clearly illustrated in fig .
6b . where the rectification at different @xmath69 is plotted for different system lengths .
obviously , the increase of system length weakens rectification .
it can be understood as follows . as the system length increases ,
long - wave phonons modes contribute more to the heat conduction .
this kind of heat conduction is symmetric because long - wave phonon modes are hardly scattered by the junction or any local defects .
moreover , the asymmetric thermal conductance is mainly controlled by the asymmetric interface , i.e. , the connection part , which does not change when system length increases . as a result
, although the absolute value does not change , the rectification appears to be smaller . based on above study
, we can conclude that changes in radial and axial direction have different influence on rectification .
the increase of structural asymmetry around the interface can increase the rectification . in the following section we will discuss other alternatives to improve the rectification .
first , it is possible to increase the difference in force constants around the interface by applying external stress . because the young s modulus has been proved to be almost the same in different tubes using the tersoff - brenner potential,@xcite it can be expected that the ( @xmath0 , 0 ) tube and ( 2@xmath0 , 0 ) tube will give different response to the same external stress .
thus it is possible that one can use this property to fabricate devices .
here we consider two kinds of stresses , tensile and torsional stresses .
it has been proven in the framework of tight - binding @xcite that these two stresses can change the electronic structure of swcnts greatly . here
the ( 8 , 0)/(16 , 0 ) swcnt imj ( @xmath70 is taken as an example . the original length is @xmath6 .
the stresses considered in this work are all small enough to ensure the system does not undergo any structural transformations @xcite or defects.@xcite for tensile stress , we elongate the imj along the axial direction by 3% . for torsional stress , one head of imj is rotated relative to the other by 30 around the axis .
the results are shown in fig .
error bars are also plotted.,title="fig : " ] [ fig7 ] it can be seen that the torsional deformation can change neither the absolute value of heat flux nor the asymmetric behavior largely .
however , the tensile stress can greatly change the heat transport . in the elongated structure ,
the heat flux becomes smaller than that of undeformed structure , but the rectification becomes larger .
above results are reasonable .
first of all , the deformation of the thicker tube is smaller than that of thinner ones under same external stress .
this means that when thick and thin tubes are connected with each other to form imj , they will appear different axial deformation under the same stress , which will further increase the structural difference between the two tubes .
secondly , the heat transport in carbon nanotubes are mainly contributed by the four acoustic phonon modes .
these phonon modes have weak dependence on the local structure , but they can be affected by the shape of the cylinder surface . when tensile stress is applied to the swcnts , the swcnts will shrink in the radial direction and elongate in the axial direction .
this means their cylinder surface will change , so that the acoustic phonon modes will also be changed .
but when swcnts are twisted , the cylinder surface almost does not change , so that the thermal conductance almost does not change . comparing fig .
7a and 3a , it can be found that the heat flux decreases in elongated structure .
the reason is straightforward . as mentioned above
, the thermal conductivity can be expressed as @xmath71 , where @xmath72 ^ 2}$ ] .
for the low frequency ( @xmath73 modes , which are most important for the thermal transport , @xmath74 .
furthermore , @xmath75 and @xmath76 are the increasing functions of the force constants @xmath77 , thus the thermal conductivity @xmath78 is also an increasing function of the force constants @xmath77 . in fact , some numerical simulations show that @xmath79 in the weak coupled lattice models.@xcite when the system is elongated , the bonds along the axis are also elongated . then according to the abell - tersoff formalism of the tersoff - brenner potential , the corresponding force constants will decrease . finally , the thermal conductivity decreases in the elongated structure .
to elongate an imj is not very difficult for the state - of - the - art experiments , so we believe that tensile stress can be a possible method to improve the rectification . next , we will try to modify the structure to change rectification .
it is suggested @xcite that a suitable on - site potential can induce large rectification in two - segment model .
so it is quite interesting that what will happen if some periodical potential is introduced into the swcnt imjs . on the other hand ,
a nanotube with sufficiently large diameter can be filled with spherical c@xmath1 fullerene molecules to build up a new hybrid structure referred to as a ` peapod',@xcite with spherical fullerenes representing peas and the carbon nanotube representing a pod .
the thermal conductivity of the peapod structure has been investigated by using md recently.@xcite so it is possible to combine these two structures , i.e. , imjs and peapod , into a new structure .
[ fig8 ] the structure ( see fig .
8) contains two parts , i.e. , a segment of ( @xmath0 , 0 ) swcnt and a segment of ( 2@xmath0 , 0 ) swcnt with some c@xmath1 balls inserted into its center , which is typically called as swcnt peapod structure .
the lengths of these two segments are almost the same .
they are connected by @xmath2 pairs of pentagon - heptagon defects .
because the closest distances between the adjacent c@xmath1 balls are almost same , 10.05 , we can consider that they exert an external periodic potential on the outside swcnt . here
, the interaction between two atoms on c@xmath1 balls and outside swcnt is modeled by lennard - jones potential : @xmath80,\ ] ] with parameters , @xmath81 mev and @xmath82 . by default , the total length of the structure is @xmath6 , which permits 10 c@xmath1 balls in the ( 2@xmath0 , 0 ) segment . in this complex structure ,
the two heads are closed to ensure the c@xmath1 balls not running out in the md calculations . and several periods of two heads ( presented by blue color and marked as region ` i ' in fig .
8) are fixed in the md simulation . then
two periods of each ends ( emphasized by orange color and marked as region ` ii ' in fig .
8) are put in the nos - hoover thermostat.@xcite in fig .
9 , the heat transports of ( 8 , 0)/c@xmath1@(16 , 0 ) structure with different lengths are shown .
@(16 , 0 ) structures with different length .
( a ) the heat flux versus temperature difference .
( b ) the thermal rectification versus temperature difference .
@xmath6 is defined in text .
error bars are also plotted.,title="fig : " ] [ fig9 ] comparing fig . 9a with fig .
6a , one can find that the heat flux of peapod structure is smaller than that of the empty imj , which seems to be in contradict with the result in ref . .
but it should be pointed out that in our structure the peapod region is closed at the ends .
in fact , this prevents the possible mass transport ( via fullerenes ) which in turn reduces the assistance transport effect of c@xmath1 balls .
furthermore , because of the coupling between the phonon modes ( mainly the radial breathing modes ) of carbon nanotube ( cnt ) and c@xmath1 balls , the conduction phonon will be scattered and the thermal conductivity of outside cnt will be reduced . on the other hand , we can notice that the length of error bars in fig .
9b is much larger than that in fig .
this means that the existence of c@xmath1 balls increases the fluctuations in the heat flux . in ref .
, it is found that the temperature fluctuations of peapod structure are larger than that of empty nanotube .
it is explained by the weak van der waals forces between fullerenes and the outside cnt .
another result is that the existence of c@xmath1 balls can hardly change the asymmetric behavior of heat flux .
this is somehow disappointing for us , because it is natural to expect that the periodical potential introduced by the c@xmath1 balls can enhance the thermal rectification according to the results based on lattice models .
however , our numerical results are also reasonable because of following facts .
the periodical potential introduced by the c@xmath1 balls is intrinsically van der waals potential , so its strength is much weaker than the binding potential among the carbon atoms in swcnt .
the period of the periodical potential introduced by c@xmath1 balls is about 10 , which is too large compared with the c - c bond length in swcnt , thus only low frequency vibrational phonons are affected .
as has been mentioned , the asymmetric heat flux is controlled by the optical phonon modes , so the change of low frequency vibrational modes can hardly change the asymmetric behavior . according to our result ,
if one wants to build a thermal rectifier by applying periodical external potential , van der waals force is not a good choice .
in this work , heat rectification in ( @xmath0 , 0)/(2@xmath0 , 0 ) imjs has been studied .
it is found that the rectification depends weakly on the detailed structure of connection part , but depends strongly on the temperature gradient .
the rectification increases as the temperature gradient increases .
moreover , we have found that the tensile stress can increase the rectification , while torsional stress can not .
it is due to the different effects of the stresses on imj structure .
it can be believed that the tensile stress will play an important role in building of devices with highly asymmetric transport behavior .
the ` peapod ' structure is combined with imj structure to investigate the effect of periodical potential on the rectification .
our numerical result shows that the new structure can hardly change the rectification .
f. bonetto , j. l. lebowitz , and l. rey - bellet , in _ mathematical physics 2000 _ , edited by a. fokas , a. grigoryan , t. kibble , and b. zegarlinsky ( imperial college press , london , 2000 ) pp .
128 - 150 ; s. lepri , r. livi , and a. politi , phys . rep . * 377 * , 1 ( 2003 ) ; b. li , j. wang , l. wang , and g. zhang , chaos * 15 * , 015121 ( 2005 ) .
g. zhang and b. li , j. chem .
phys,*123 * , 014705(2005 ) ; c. w. chang , a. m. fennimore , a. afanasiev , d. okawa , t. ikuno , h. garcia , deyu li , a. majumdar , and a. zettl , phys . rev . lett . * 97 * , 085901 ( 2006 ) . | we study heat conduction in ( @xmath0 , 0)/(2@xmath0 , 0 ) intramolecular junctions by using molecular dynamics method .
it is found that the heat conduction is asymmetric , namely , heat transports preferably in one direction .
this phenomenon is also called thermal rectification .
the rectification is weakly dependent on the detailed structure of connection part , but is strongly dependent on the temperature gradient .
we also study the effect of the tube radius and intramolecular junction length on the rectification .
our study shows that the tensile stress can increase rectification .
the physical mechanism of the rectification is explained . |
trapping and cooling bose - fermi mixtures of dilute quantum gases has opened a wide area of research in atomic physics .
the interactions between bosonic and fermionic species interconnect two systems of fundamentally different quantum statistics .
the diluteness of the gaseous mixtures allows one to treat the interactions between particles in terms of binary collisions . in consequence
we can replace the real inter - atomic potential by a pseudo - potential characterized by only one parameter , the @xmath0-scattering length .
the latter is experimentally@xcite tunable by exploiting optically or magnetically induced feshbach resonances@xcite . despite its simplicity the interaction potential ( mathematically ill defined@xcite ) of the ultra - cold multi - component gases confined in optical lattices
is responsible for a wealth of novel quantum phases@xcite including charge density waves ( cdw)@xcite , as well as supersolid behavior@xcite .
the nature of the phase transition and qualitative phase diagram for one - component bosonic system can be inferred based on very simple arguments@xcite . when tunneling between lattice sites of the bosons is suppressed , compared with point - like interaction between them , the system can undergo a quantum phase transition between a superfluid ( sf ) phase ( characterized by large number fluctuations at each lattice site ) , and a mott insulating ( mi ) phase where each lattice site is occupied by precisely an integer number of bosons without any number fluctuations . adding to a bosonic system a fermionic ingredient and allowing for the mutual repulsion or attraction between species of different statistics strongly affects the equilibrium properties . increasing the boson - fermion repulsion drives the system towards spatial separation whereas attraction gives rise to implosion@xcite .
the dynamics underlying the phase transitions in the bose - fermi mixtures is produced by the small changes of the bosonic density which induce a modulation of the fermionic density .
as a consequence of the feedback of the fermionic perturbation a shift of the bosonic energy occurs , thereby inducing an additional attraction or repulsion that changes the original interaction between bosons@xcite . in the case of deep optical lattices and small densities ,
the coherent description of the system provided by the gross - pitaevskii equation is not reliable due to rising effects of correlation .
the experimental data clearly demonstrated that adding a fermionic cloud to strongly interacting bosons always results in a decay of visibility of the interference pattern in time - of - flight images@xcite .
moreover , the scale of disappearance of the coherence in the mixtures is different for attractive and repulsive scattering lengths@xcite . to predict such behavior
theoretically one can include the more general , than one - component bose - hubbard ( bh ) , multi - band model . in mentioned approach
if the higher - band renormalization of the boson parameters is dominant over the fermion screening of the interaction , the mott - insulating lobes in the bose - hubbard phase diagram are enhanced for either sign of the bose - fermi interactions@xcite . on the other hand inclusions of the retardation effects@xcite ( which arises from the presence of very low energy excitations in a fermi sea )
give rise to so - called orthogonality catastrophe@xcite .
another approach to quantum mixtures of particles of unequal masses , when the difference in the tunneling amplitudes between heavy bosons and light fermions is large enough to neglect quantum nature of the bosons , provides to description of the system in fermi - bose version of the falicov - kimball model@xcite .
the mutual interactions of bosons and fermions can affect the spectrum of collective excitations in the collisionless regime as the mixture goes toward either demixing or collapse@xcite .
it has been shown that mode - mode coupling effects may arise when sound velocity of the bose gas is comparable to the fermi velocity of the fermions@xcite .
the energy spectra of the bosonic and fermionic mixtures and phase diagram were also obtained by the field theory methods@xcite , however calculated phase boundary does not change the structure and only shifts the chemical potential .
the properly constructed effective theory lead to an effective , fermion mediated , long - range interaction between bosons with alternating sign that is the origin of the cdw and can explain the mi - cdw phase separation@xcite .
there is also another possibility of the analysis of the mixtures of atoms with different statistics where the second species is strongly localized on random sites which can lead to random shifts of the on - site energies and , in consequence , the disorder with discrete probability distribution is created @xcite .
recently the experiments @xcite on a harmonically trapped mixtures of atomic bose - bose gases show that the presence of relevant fraction of the @xmath1k bosonic species modifies the quantum phase transition occurring in _ _ @xmath2__@xmath3rb inducing a significant loss of coherence similarly to bose - fermi systems that can be explained in the mean - field theory framework@xcite .
the aim of this work is to study the superfluid to mott - insulator zero - temperature phase transition by means of the bose - fermi - hubbard model in two- ( @xmath4 ) and three - dimensional ( @xmath5 ) optical condensates . in order to find a phase boundary for bf mixtures
very sophisticated methods and calculations are required . only several theoretical works concentrated on the inherent difficulty of dealing with bf hubbard hamiltonian originates from the non - perturbative nature of the model and retardation effects . to elucidate the quantum phase transition in optical lattices , where the kinetic energy scale is less than the dominating repulsive
energy and density - density coupling between species with different statistics comes into play , we have adopted a theoretical approach for strongly interacting fermions @xcite to the bf hubbard model in a way to include the effects of particle number fluctuations and make the qualitative phase diagrams more quantitative @xcite . to facilitate this task
, we employ a functional integral formulation of the theory that enables to perform functional integration over fields defined on different topologically equivalent classes of the @xmath6 group , i.e. , with different winding numbers .
an inclusion of the winding numbers ( comes from periodicity of the phase variables ) is unavoidable in order to properly construct the phase diagram and the poisson re - summation formula turns out to be very useful for derivation of the topological term of the partition function .
the quantum rotor representation method we use is deeply rooted in the gauge symmetries of the model .
we construct an invariant theory introducing an appropriate @xmath6 gauge transformation . in sec .
ii we review the hamiltonian for the system and show the connections of the parameters to the experimentally measured quantities in optical lattices . sec .
iii contains description of the method we use and can serve a guidance to obtain the critical line equation presented in sec .
before showing the phase diagrams for the bose - fermi hubbard model in the quantum rotor description we make some general remarks in sec . v concern the phase boundary equation and compare our results with the diagrammatic perturbation approach to the one - component bose - hubbard model for experimentally accessible densities of the particles .
vi presents the discussion of the ground state phase diagrams for the bose - fermi hubbard hamiltonian calculated within quantum rotor approach .
is devoted to some concluding remarks .
the appendixes contain the derivation of relevant formulas of the main text and are introduced to keep the text self - contained .
for bosons confined in optical lattices the two main energy scales are set by the hopping amplitude proportional to @xmath7 ( which sets the kinetic energy scale for bosons ) due to the particles tunneling , and the on - site interaction @xmath8 .
for @xmath9 the phases of the superfluid order parameter on individual lattice sites are well defined . on the other hand , for sufficiently large repulsive energy @xmath10 , the quantum phase fluctuations lead to complete suppression of the long - range phase coherence even at zero temperature .
the competition between the kinetic energy , which is gained by delocalizing bosons over lattice sites and the repulsive interaction energy , which disfavors having more than one particle at any given site , can be modeled by quantum bose - hubbard hamiltonian@xcite .
the physics of the bosonic and non - interacting spin - polarized ( collisions in the @xmath0-wave channel are forbidden by their statistics ) fermionic mixtures with density - density interaction between species @xmath11 leads to bose - fermi - hubbard hamiltonian @xcite : @xmath12 where @xmath13 and @xmath14 stand for the bosonic ( fermionic ) creation and annihilation operators @xmath15 , @xmath16 is the boson ( fermion ) number operator on the site @xmath17 , and the reduced chemical potential @xmath18 controls the number of bosons and @xmath19 fermions respectively .
here , @xmath20 identifies summation over the nearest - neighbor sites .
furthermore , @xmath21 is the hopping matrix element for bosons ( fermions ) . for simplicity , we neglect the inhomogeneous magnetic trap potential .
if the on - site boson - fermion coupling strength @xmath11 becomes very strong the dilute gaseous mixtures are unstable to phase separation @xmath22 or to collapse of the phase separated configuration @xmath23 @xcite .
the presence of the lattice will introduce kinetic energy scales @xmath24 competing with @xmath11 stabilizing the system .
we assume that an optical lattice created by the counter - propagating laser beams is deep enough and we can restrict ourselves to the lowest bloch bands .
the corresponding experimental parameters can be estimated by following relations@xcite@xmath25,\\ u_{x } & \simeq & \sqrt{\frac{8}{\pi}}ka_{x}e_{r}^{x}\left(\frac{v_{0}}{e_{r}^{x}}\right)^{3/4},\end{aligned}\ ] ] ( subscript @xmath26 means @xmath27 bosons and @xmath28 fermions respectively ) where boson - boson @xmath29 , fermion - fermion @xmath30 and boson - fermion @xmath31 @xmath32 scattering lengths can be continuously tune in the experiments@xcite inducing attractive or repulsive interaction between species .
the @xmath33 is the wavelength of the laser and @xmath34 is the recoil energy and @xmath35 is the atomic mass .
we write the partition function of the system switching from the particle - number representation to the conjugate phase representation of the bosonic degrees of freedom using the bosonic and fermionic path - integral over the complex fields @xmath36 and @xmath37 depending on the `` imaginary time '' @xmath38 with @xmath39 being the temperature:@xmath40e^{-\mathcal{s}\left[\bar{b},b,\bar{f},f\right]}.\ ] ] the action @xmath41 is given by@xmath42+\int_{0}^{\beta}d\tau\mathcal{h\left(\tau\right)},\ ] ] where@xmath43 & = & \sum_{i}\int_{0}^{\beta}d\tau\bar{b}_{i}\left(\tau\right)\frac{\partial}{\partial\tau}b_{i}\left(\tau\right)\nonumber \\ & + & \sum_{i}\int_{0}^{\beta}d\tau\bar{f}_{i}\left(\tau\right)\frac{\partial}{\partial\tau}f_{i}\left(\tau\right).\end{aligned}\ ] ] in the next section we will integrate over the fermionic fields since the action is quadratic in @xmath37 variables .
we attempt to reduce the large number of degrees of freedom in the partition function to the few which dominate the low energy physics . before integrating out of the fermionic degrees of freedom we write the action in the form : @xmath44 & = & \int_{0}^{\beta}d\tau\left\ { \sum_{i}\left[\bar{b}_{i}\left(\tau\right)\frac{\partial}{\partial\tau}b_{i}\left(\tau\right)+\frac{u_{b}}{2}n_{bi}^{2}\left(\tau\right)\right]\right.\nonumber \\ & - & \left.\sum_{\left\langle i , j\right\rangle } t_{bij}\bar{b}_{i}\left(\tau\right)b_{j}\left(\tau\right)-\bar{\mu}_{b}\sum_{i}n_{bi}\left(\tau\right)\right\ } , \nonumber \\ \mathcal{s}_{f}\left[\bar{f},f , n_{f}\right ] & = & \int_{0}^{\beta}d\tau\left[\sum_{i}\bar{f}_{i}\left(\tau\right)\frac{\partial}{\partial\tau}f_{i}\left(\tau\right)\right.\nonumber \\ & + & \left.\sum_{\left\langle i ,
j\right\rangle } t_{fij}\bar{f}_{i}\left(\tau\right)f_{j}\left(\tau\right)-\mu_{f}\sum_{i}n_{fi}\left(\tau\right)\right],\nonumber \\ \mathcal{s}_{\mathrm{int}}\left[n_{b},n_{f}\right ] & = & u_{bf}\sum_{i}\int_{0}^{\beta}d\tau n_{bi}\left(\tau\right)n_{fi}\left(\tau\right).\end{aligned}\ ] ] we notice that adding the inter - species interaction term to the fermionic part of the action@xmath45_{ij}f_{j}\left(\tau'\right).\ ] ] allows one to integrate over fermionic fields , because the action is gaussian in @xmath37 operators .
resulting partition function takes the form@xmath46e^{-\mathcal{s}_{b}\left[\bar{b},b , n_{b}\right]}e^{\mathrm{tr}\ln\hat{g}_{f+\mathrm{int}}^{-1}}.\label{partitionfunction1}\ ] ] we will be looking for solutions obeying translational invariance in the `` imaginary time '' direction , i.e. , such that the partition function depends only on the difference @xmath47 . expanding the trace of the logarithm in eq .
( [ partitionfunction1 ] ) we have @xmath48 with@xmath49_{ij } & = & \left[\left(\frac{\partial}{\partial\tau}-\mu_{f}\right)\delta_{ij}-t_{fij}i_{ij}\right]\delta\left(\tau-\tau'\right),\nonumber \\ \left[\hat{g}_{\mathrm{int}}\left(\tau,\tau'\right)\right]_{ij } & = & u_{bf}\bar{b}_{i}\left(\tau\right)b_{j}\left(\tau'\right)\delta_{ij}\delta\left(\tau-\tau'\right).\end{aligned}\ ] ] we defined @xmath50 if @xmath51 are the nearest neighbors and equals zero otherwise .
trace over first term of the expansion gives constant contribution of the fermions , in the non - interacting system , to the action .
second one induces a shift in the chemical potential of bosons . the third term after exploiting fourier - matsubara transform@xmath52 where @xmath53 ( @xmath54 ) with ( @xmath55 ) are bose(fermi)-matsubara frequencies respecting periodic ( anti - periodic ) boundary conditions of the bosonic ( fermionic ) field operator , reduces to@xmath56 where @xmath57 and @xmath58 is called lindhard function and imaginary @xmath59 part of the local dynamic lindhard function for square ( @xmath4 ) and cubic lattice ( @xmath5 ) , in the low - temperature limit , as a function of frequency , where the normalized fermionic potential @xmath60 is equal zero .
the normalized value of the critical frequency @xmath61 shows where induced , frequency - dependent , effective part of the interaction @xmath62 changes character from attractive to repulsive . ]
@xmath63-f\left[t_{f\boldsymbol{k}+\boldsymbol{q}},\mu_{f},\beta\right]}{t_{f\boldsymbol{k}}-t_{f\boldsymbol{k}+\boldsymbol{q}}-i\nu_{\ell}}.\label{lindhard}\ ] ] in the above @xmath64=1/\left\ { \exp\left[\beta\left(t_{f\boldsymbol{k}}-\mu_{f}\right)\right]+1\right\ } $ ] is the fermi distribution function and @xmath65 is the fermionic dispersion relation . to stay in the local ( momentum integrated ) regime
we perform @xmath66 and * k * integration over the first brillouin zone and in the @xmath67 limit using an analytic continuation @xmath68@xmath69 we obtain imaginary part @xmath70 of the local dynamic lindhard function ( see appendix for details)@xmath71\nonumber \\ & \times & \rho\left(x\right)\rho\left(x-\omega'\right)\end{aligned}\ ] ] that satisfies sum rule which is just the conservation of the number of particles . in the above @xmath72 $ ]
is the density of states and @xmath73 is the unit step function .
therefore , the corresponding real part @xmath74 can be easily deduced from kramers - krnig relation @xmath75 later , without any loss of generality , we drop the chemical potential dependence writing @xmath76 and @xmath77 .
finally the `` imaginary - time '' partition function with integrated out fermionic degrees of freedom , in the local approximation ( see appendix for details ) , can be written as:@xmath46e^{-\mathcal{s}_{\mathrm{eff}}\left[\bar{b},b\right]}\ ] ] with the effective action @xmath78 & = & \int_{0}^{\beta}d\tau\left\ { \sum_{i}\bar{b}_{i}\left(\tau\right)\frac{\partial}{\partial\tau}b_{i}\left(\tau\right)\right.\nonumber \\ & + & \frac{u_{\mathrm{eff}}}{2}\sum_{i}n_{bi}^{2}\left(\tau\right)-\sum_{\left\langle i , j\right\rangle } t_{bij}\bar{b}_{i}\left(\tau\right)b_{j}\left(\tau\right)\nonumber \\ & - & \left.\bar{\mu}_{b}\sum_{i}n_{bi}\left(\tau\right)\right\ } \label{effective action}\end{aligned}\ ] ] expressed in terms of bosonic degrees of freedom only .
we want to emphasize that applying a local , in the matsubara - imaginary time , approach we neglect any dissipation effects . whereas locality in real space rules out some parts of interesting physics such as the charge density wave , namely an insulating phase with modulated density or the supersolid phase , presenting the coexistence of superfluidity and a periodic spatial modulation of the density , different from that of the lattice . on the other hand , the long - range character of the fermion mediated interaction between bosons with the fermion - induced mean field potential can lead to spatially homogeneous regions of commensurate cdw @xcite .
the motivation of the local approximation was the idea that response of an interacting system can be pictured as the response of a non - interacting system to an effective self - consistent field , that depends on global properties such as the particle densities .
a question of both fundamental and practical interest is , to what extent can the physics of the exact non - local interaction be captured by an approximate local theory ?
it seems that local approximations often work surprisingly well , yielding energies very accurately , without suffering from some of the characteristic drawbacks of non - locality @xcite . from eq .
( [ effective action ] ) it is concluded that there is a striking resemblance to the one - component bose - hubbard action with the original repulsive interaction replaced now by @xmath79 which is the induced , frequency - dependent , effective interaction between bosons . from eq .
( [ u effective ] ) we see that integrating out fermionic field from bf hubbard hamiltonian provides an additional interaction among bosons , which is _ not affected _ by the attractive or repulsive nature of the inter - species interaction @xmath80 . before we proceed with further calculations
let us make some remarks .
the substitution we introduced in eq .
( [ u effective ] ) is deceptively simple and can lead to the assumption that the phase diagram of the bfh model can be easily derived from the critical line of the one - component bh hamiltonian , which already has been obtained in several approximations .
unfortunately , as we will see in the next section , @xmath81 is _ not _ the only one ingredient to the final equation for the critical line
. the additional part , which influences the phase boundary line condition , comes from the inter - species interaction and number of fermions added to the system has an impact on amplitude of the order parameter . moreover , the chemical potential for fermions @xmath82 is shifted ( we postpone calculations of it now and show proper formula later ) because of the induced effective interaction between them . part of the local dynamic lindhard function for cubic lattice in the space of the parameters : normalized frequency @xmath83 and fermionic chemical potential @xmath82 in the zero - temperature limit . ]
part of the local dynamic lindhard function for cubic lattice in the space of the parameters : normalized frequency @xmath83 and fermionic chemical potential @xmath82 in the zero - temperature limit . ]
next step in the calculations depends on the ratio @xmath84 of the masses of bosons and fermions which can be seen from eq .
( [ ubfexp ] ) .
consequence of the latter is the fact that the speed of the bogoliubov sound @xmath85 for bosons differs from the first sound @xmath86 of the ideal fermi gas . in typical experimental realizations @xmath87-@xmath88 systems
the acoustic long - wavelength ( boson ) @xmath89 and fermion @xmath86 velocities are comparable or @xmath90 ( boson mass is larger than fermion ) .
therefore we do not restrict our calculations to the static limit but consider also local dynamical response function ( thus including the retardation effects ) .
if @xmath89 is much larger than @xmath86 ( in the @xmath87-@xmath91 species we have @xmath92 ) the resulting interaction between bosons is _ instantaneous _ and _ always _ attractive ( with @xmath93 ) so using the static approximation is justified with an error which involves the small parameter @xmath94 .
the imaginary part of the lindhard response function rises to a broad peak before falling and the real part takes zero when changes in @xmath95 are the biggest ( see fig .
[ fig1 ] ) .
the real part @xmath96 is negative for the frequencies @xmath97 for square and @xmath98 for cubic lattice ( see fig .
[ fig1 ] ) .
we normalized the frequency by the width of the band for non - interacting fermions to show the scale of the energy .
nevertheless , for higher values of the normalized frequency @xmath99 the induced part of the interaction @xmath100 can be _ positive _ and _ increase _ the effective interaction between bosons , in consequence provide stronger localization these species on lattice sites . in the case of large fermion
hopping and commensurate filling with the lattice the effective long - range density - density interaction between bosons has alternating sign and is the origin of the charge density - wave phases @xcite .
the higher values of the normalized chemical potential for the fermions @xmath82 decreases the values of both ( fig .
[ fig2 ] and fig .
[ fig3 ] ) real and imaginary part of the local lindhard function .
that leads to the situation where the terms containing explicitly the average density of fermions @xmath101 will acquire more significance than terms with exclusively the inter - species interaction @xmath11 .
unfortunately the effective action is not quadratic in bosonic fields @xmath102 and we have to decouple the effective interaction term in eq .
( [ effective action ] ) by a gaussian integration over the auxiliary scalar potential fields @xmath103 with static @xmath104 and periodic part @xmath105 where @xmath106 is the bose - matsubara frequency .
we observe now that effective bf hubbard hamiltonian has a local @xmath6 gauge symmetry , when expressed in terms of the underlying boson variables .
this points out a possibility of an emergent dynamical @xmath6 gauge field as a fluctuating complex field attached to bosonic variables , which is dynamically generated , by interacting bosons .
thus , the periodic part @xmath107 couples to the local particle number through the josephson - like relation @xmath108 , where @xmath109 the quantity @xmath110 is the @xmath6 _ phase _ field and satisfies the periodicity condition @xmath111 as a consequence of the periodic properties of the @xmath112 field in eq .
( [ periodic part v ] ) .
next , we perform the local gauge transformation to the new bosonic variables@xmath113=\left[\begin{array}{cc } e^{i\phi_{i}\left(\tau\right ) } & 0\\ 0 & e^{-i\phi_{i}\left(\tau\right)}\end{array}\right]\left[\begin{array}{c } a_{i}\left(\tau\right)\\ \bar{a}_{i}\left(\tau\right)\end{array}\right]\ ] ] that removes the imaginary term @xmath114 from all the fourier modes except at zero frequency . from the above we deduce
bosons have a composite nature made of bosonic part @xmath36 and attached `` flux '' @xmath115 $ ] .
due to such @xmath6 gauge invariance , the fluctuations and the phase have the dynamics of @xmath6 gauge field . by integrating out the auxiliary static field @xmath117 we calculate the partition function with an effective action expressed in the form of the propagator @xmath118@xmath119e^{-\sum_{i}\int_{0}^{\beta}d\tau\left[\frac{1}{2u_{\mathrm{eff}}}\dot{\phi}_{i}^{2}\left(\tau\right)+\frac{1}{i}\frac{\bar{\mu}_{b}}{u_{\mathrm{eff}}}\dot{\phi}_{i}\left(\tau\right)\right]+\mathrm{tr}\ln\hat{g}^{-1}},\label{partition function propagator}\ ] ] where @xmath120 is the shifted reduced bosonic chemical potential . in the above @xmath121 and the determinant takes the form@xmath122\exp\left\ { -\sum_{\left\langle i , j\right\rangle } \int_{0}^{\beta}d\tau\right.\nonumber \\ & \times & \bar{a}_{i}\left(\tau\right)\left[\left(\frac{\partial}{\partial\tau}+\bar{\mu}_{b}\right)\delta_{ij}\right.\nonumber \\ & - & \left.\left.e^{i\phi_{i}\left(\tau\right)}t_{bij}e^{-i\phi_{j}\left(\tau\right)}\right]a_{i}\left(\tau\right)\right\ } .\end{aligned}\ ] ] we parametrize the boson fields @xmath123 and incorporate fully our calculations to the phase fluctuations governed by the gauge group @xmath6 . assuming nonfluctuating amplitude at low temperatures @xmath124
, we drop the corrections , which was proved to be justified in the large @xmath125 limit we are interested in @xcite .
the amplitude fluctuations are massive one and do not play important role in the low energy scales .
it is very convenient to define the order parameter @xmath126\right\rangle = a_{0}\psi_{\mathrm{b}},\label{order parameter definition}\ ] ] which signals the emergence of the superfluid phase and vanishes in the mott - insulator state .
the sf state is characterized by spontaneously breaking of the @xmath116 symmetry of bose - fermi - hubbard hamiltonian .
note , that a nonzero value of the amplitude @xmath127 in eq .
( [ order parameter definition ] ) is _ not sufficient _ for superfluidity . to achieve this ,
also the phase variables @xmath128 in eq .
( [ order parameter definition ] ) , must become stiff and coherent , which implies @xmath129 .
as we see in the next sections the presence of the fermions and density - density interactions @xmath11 between species of different statistics can also change the amplitude of the order parameter . after mentioned assumption the inverse of the propagator becomes@xmath130 the explicit value of the amplitude @xmath127 in eq .
( [ order parameter definition ] ) can be obtained from minimization of the hamiltonian @xmath131 .
therefore , we write@xmath132 expanding the trace of the logarithm in eq .
( [ partition function propagator ] ) and making use the above we obtain up to the second order in the amplitude of the order parameter eq .
( [ order parameter definition ] ) @xmath133 trace over first term of the expansion , as previously , not containing any fluctuating field variables , gives an inessential constant contribution to the action .
let us consider the second order term in more detail@xmath134}\delta\left(\tau-\tau'\right),\end{aligned}\ ] ] where@xmath135 the hopping matrix elements are re - normalized by the amplitude of the order parameter .
we see that in comparison to pure bosonic case there is an additional shift @xmath136 that depends on the average of the fermion concentration and normalized inter - species interaction .
the above was also obtained in the effective bosonic model and recognized as a mean - field contribution @xcite .
finally , the partition function eq .
( [ partition function propagator ] ) becomes@xmath137e^{-\mathcal{s}_{\mathrm{phase}}\left[\phi\right]}\end{aligned}\ ] ] with an effective action expressed _ only _ in the _ phase _ fields variable@xmath138 & = & \int_{0}^{\beta}d\tau\left\ { \sum_{i}\left[\frac{1}{2u_{\mathrm{eff}}}\dot{\phi_{i}^{2}}\left(\tau\right)+\frac{1}{i}\frac{\bar{\mu}_{b}}{u_{\mathrm{eff}}}\dot{\phi_{i}}\left(\tau\right)\right]\right.\nonumber \\ & - & \left.\tilde{t}_{b}\sum_{i , j}e^{\phi_{i}\left(\tau\right)}i_{ij}e^{-\phi_{j}\left(\tau\right)}\right\ } , \label{action only phase}\end{aligned}\ ] ] where @xmath139
. the total time derivative berry phase imaginary term in eq .
( [ action only phase ] ) is nonzero due to topological phase field configurations with @xmath140 ( @xmath141 ) that results in topological ingredients to the correlator we will see below .
therefore , we concentrate on closed paths in the `` imaginary time '' @xmath142 labeled by the integer winding numbers @xmath143 . the path - integral @xmath144 ...
\equiv\sum_{\left[m_{i}\right]}\int_{0}^{2\pi}\left[\mathcal{d}\phi\left(0\right)\right]\int_{_{\phi_{i}\left(0\right)}}^{\phi_{i}\left(\tau\right)+2\pi m_{i}}\left[\mathcal{d}\phi\left(\tau\right)\right] ... ,\ ] ] includes a summation over @xmath143 and in each topological sector the integration goes over the gauge potentials .
therefore , we do not ignore the compactness of the gauge fields .
to proceed , we replace the phase degrees of freedom by the uni - modular scalar complex field @xmath145 which satisfies the quantum periodic boundary condition @xmath146 .
this can be conveniently done using the fadeev - popov method with dirac delta functional resolution of unity , where we take @xmath145 as a continuous but constrained ( on the average ) variable to have the uni - modular value .
we introduce @xmath147\delta\left(\sum_{i}\left|\psi\left(\tau\right)\right|^{2}-n\right)\nonumber \\ & \times & \delta\left(\psi_{i}-e^{i\phi_{i}\left(\tau\right)}\right)\delta\left(\psi_{i}^{*}-e^{-i\phi_{i}\left(\tau\right)}\right)\end{aligned}\ ] ] and@xmath148 where @xmath149 is the number of lattice sites . introducing the lagrange multiplier @xmath150 , which adds the quadratic terms ( in the @xmath145 fields ) to the action we can solve for the constraint .
the partition function can be rewritten to the form@xmath151\nonumber \\ & \times & \exp\left\ { -\sum_{i , j}\int_{0}^{\beta}d\tau d\tau'\psi_{i}\left[\left(\tilde{t}_{b}i_{ij}+\lambda\delta_{ij}\right)\delta\left(\tau-\tau'\right)\right.\right.\nonumber \\ & + & \left.\left.\gamma_{ij}\left(\tau,\tau'\right)\right]\psi_{j}^{*}\right\ } , \end{aligned}\ ] ] where @xmath152\right\ } \right\rangle\ ] ] is the two - point phase correlator associated with the order parameter field , where @xmath153 denotes averaging with respect to the action in eq .
( [ action only phase ] ) . because the values of the phases @xmath128 which differ by @xmath154 are equivalent we decompose phase field in terms of a periodic field and term linear in @xmath155:@xmath156 with @xmath157 as a result the phase correlator factorizes as the product of a topological term depending on the integers @xmath143 and non - topological one:@xmath158 performing the poisson re - summation formula in @xmath159}e^{-i\frac{2\pi}{\beta}\left(\tau-\tau'\right)m_{i}}e^{-\frac{2\pi}{\beta}\sum_{i}\left[\frac{\pi}{u_{\mathrm{eff}}}m_{i}^{2}+\frac{\beta}{i}\frac{\bar{\mu}_{b}}{u_{\mathrm{eff}}}m_{i}\right]}}{\sum_{\left[m_{i}\right]}e^{-\frac{2\pi}{\beta}\sum_{i}\left[\frac{\pi}{u_{\mathrm{eff}}}m_{i}^{2}+\frac{\beta}{i}\frac{\bar{\mu}_{b}}{u_{\mathrm{eff}}}m_{i}\right]}}\ ] ] and the functional integration over the phase variables @xmath160e^{-i\left[\varphi_{i}\left(\tau\right)-\varphi_{j}\left(\tau'\right)\right]}e^{-\sum_{i}\frac{1}{2u_{\mathrm{eff}}}\int_{0}^{\beta}d\tau\dot{\varphi_{i}^{2}}\left(\tau\right)}}{\int\left[\mathcal{d}\varphi\right]e^{-\sum_{i}\frac{1}{2u_{\mathrm{eff}}}\int_{0}^{\beta}d\tau\dot{\varphi_{i}^{2}}\left(\tau\right)}}\ ] ] the final formula of the correlator takes the form @xmath161 where @xmath162 is the jacobi theta function , which comes from the topological contribution - summation over integer winding numbers . the function @xmath162 is defined by @xmath163 and is @xmath164-periodic in the `` imaginary time '' as well in the variable @xmath165 with the period of unity which emphasizes the special role of its integer values .
after fourier transforming one obtains @xmath166}\frac{e^{-\frac{u_{\mathrm{eff}}\beta}{2}\sum_{i}\left(m_{i}+\frac{\bar{\mu}_{b}}{u_{\mathrm{eff}}}\right)^{2}}}{1 - 4\left(\sum_{i}m_{i}+\frac{\bar{\mu}_{b}}{u_{\mathrm{eff}}}-i\frac{\omega_{\ell}}{u_{\mathrm{eff}}}\right)^{2}},\label{topological contribution}\ ] ] where@xmath167}e^{-\frac{u_{\mathrm{eff}}\beta}{2}\sum_{i}\left(m_{i}+\frac{\bar{\mu}_{b}}{u_{\mathrm{eff}}}\right)^{2}}\ ] ] is the partition function for the set of quantum rotors . the action eq .
( [ action only phase ] ) , with the topological contribution eq .
( [ topological contribution ] ) , after fourier transform , is written as @xmath168=\frac{1}{n\beta}\sum_{\mathbf{k},\ell}\bar{\psi}_{\mathbf{k}}\left(\omega_{\ell}\right)\mathrm{\gamma}_{\mathbf{k}}^{-1}\left(\omega_{\ell}\right)\psi_{\mathbf{k}}\left(\omega_{\ell}\right),\ ] ] where @xmath169 is the inverse of the propagator and @xmath170 is the fourier transform of the bosonic hopping matrix elements for two- @xmath171 and three - dimensional @xmath172 lattice .
within the phase coherent state the order parameter @xmath173 is evaluated in the thermodynamic limit @xmath174 by the saddle point method @xmath175 and the uni - modular condition of the @xmath6 phase variables translates into the equation@xmath176 with@xmath177 the phase boundary is determined by the divergence of the order parameter susceptibility @xmath178@xmath179 which determines the critical value of the lagrange parameter @xmath180 and stays constant in the whole global coherent phase . to proceed
, it is desirable to introduce the density of states @xmath181 because the analytical expressions we use can be advantageous in evaluating sums over momenta . the corresponding formulas for square lattice can be written as@xmath182 and for simple cubic geometry takes form @xmath183 with @xmath184 and @xmath185 ; @xmath186 is the elliptic function of the first kind.@xcite . after summation over bose - matsubara frequency and for zero temperature limit
@xmath187 we can rewrite the critical line equation to the form that represents solution of the bf hubbard model in terms of re - normalized pure bose - hubbard hamiltonian in the quantum rotor approach : @xmath188 in the above @xmath189 where @xmath190 $ ] is the fractional part of the number and @xmath191 $ ] is the floor function which gives the greatest integer less than or equal to @xmath192 ; @xmath193 stands for the maximum value of the bosonic dispersion spectrum @xmath170 and @xmath194 is the lattice coordination number .
the renormalization parameters are defined as : @xmath195 and allow us to see how adding free fermions to strongly interacting bosons confined in optical lattice influences the phase boundary .
the zero - temperature phase diagram of the bose - fermi - hubbard model eq .
( [ hamiltonian ] ) can be calculated from eq .
( [ critical line-1 ] ) and usually is plotted as a function of @xmath196 , with the density of the bosons controlled by a chemical potential @xmath197 .
the presence of the fermions implicates two additional different parameters that can by varied namely @xmath198 and @xmath199 in eq .
( [ parameters1 ] ) and eq .
( [ parameters2 ] ) .
the strength of the inter - species interaction influences both of them , however the sign of @xmath11 and the average density of fermions @xmath101 added to the system affects only @xmath199 .
besides , in the local dynamic approach the sign of the density - density interaction depends also on the normalized frequency @xmath83 . for a general choice of parameters , eq .
( [ critical line-1 ] ) is easy to solve , however considerations of special cases can provide more insights into the solution of the problem . in discussion we will follow the scheme@xmath200 firstly choosing the sign of the @xmath199 and later @xmath198 in the static ( @xmath201 ) or dynamic ( @xmath202 ) limit ( see fig .
[ fig1 ] ) .
before we proceed with analysis let us introduce the notation for the maximum of the critical value for parameter @xmath196 ( as a function of the normalized chemical potential @xmath197 ) at the tip of the @xmath203th ( @xmath204 ) mi lobe for different lattice geometries and model parameters @xmath198 and @xmath199 as follows @xmath205 -@xmath197 ) for the square ( @xmath4 ) lattice for different @xmath206 ( higher panel , @xmath207 ) , @xmath208 ( @xmath99 ) and @xmath209 ( negative scattering length @xmath210 ) .
dashed line stands for the phase boundary of one - component bose - hubbard model . within the lobes the mott insulator phase takes place with @xmath211.[fig4 ] ] -@xmath197 ) for the cubic ( @xmath5 ) lattice for different @xmath206 ( higher panel , @xmath207 ) , @xmath208 ( @xmath99 ) and @xmath209 ( negative scattering length @xmath210 ) . dashed line stands for the phase boundary of one - component bose - hubbard model . within the lobes the mott insulator phase takes place with @xmath211.[fig5 ] ] the above determines when the transition from mi to sf occurs .
values @xmath212 and @xmath213 stand for the one - component bosonic case . in table [ comparison ] we show comparison of @xmath214 for higher densities of the particles calculated in the quantum rotor approach ( qra ) to very accurate , recently developed , diagrammatic perturbation approach@xcite to bose - hubbard hamiltonian .
the results for @xmath5 bh model obtained in both theories are very close and also comparison to quantum monte - carlo ( qmc ) numerical calculations@xcite indicates that methods we use are able to properly catch the interesting physics of strongly interacting systems
. however , we want to analyze the phase boundary for number of particles per lattice sites higher than one @xmath215 that adds another dimension to the analysis and is difficult for the qmc to catch .
the phase boundary for square lattice shows that qra works well also in low - dimensional geometries , especially for higher densities . nevertheless , the structure of eq .
( [ critical line-1 ] ) can cause some problems when _ both _ @xmath198 and @xmath199 are nonzero and the number of bosons is equal one per lattice sites .
we expect that for @xmath216 some artificial effects may arise for values of the normalized chemical potential @xmath217 close to zero .
.comparison of the maximum of the critical value for @xmath218 ( as a function of the normalized bosonic chemical potential @xmath197 ) at the tip of the @xmath203th ( @xmath219 and @xmath220 ) mott insulator lobe for the square ( @xmath4 ) and cubic ( @xmath5 ) lattice in the one - component bose - hubbard model : dpt - diagrammatic perturbation theory@xcite ) , qra - our calculations using quantum rotor approach ) .
[ cols="^,^,^,^,^,^",options="header " , ]
for different @xmath198 and @xmath221 ( very small amount of fermions ) for square ( @xmath4 ) and cubic ( @xmath5 ) lattice with @xmath222 . vertical solid line stands for the @xmath223 obtained in one - component bose - hubbard model with one particle per lattice site . above the curves the superfluid phase takes place with @xmath224.[fig6 ] ] in the experiments@xcite for a degenerate mixtures of @xmath225 @xmath88 bosons and @xmath226 @xmath87 fermions the scattering length @xmath31 ( and in consequence interaction @xmath11 ,
( [ ubfexp ] ) ) can be continuously tune between @xmath227 below and between @xmath228 above feshbach resonance , where @xmath127 is the bohr radius .
the form of the parameters we choose eq .
( [ parameters1 ] ) and eq .
( [ parameters2 ] ) allows for its interpretation .
the periodicity of the phase diagram can be easily deduced from the periodic properties of the propagator eq .
( [ topological contribution ] ) and strongly depends on @xmath198 however , the interaction between species does not generate additional mott lobes in the phase diagram .
the above is in contrast to the strong coupling expansion and exact diagonalization method applied to the system of two alkali - metal atoms with different masses where ( for very small lattice sizes and quenched disorder ) the mi phases with integer filing factors disappear for boson - impurity interaction energy larger than on - site atom - atom interaction energy itself and also the mi phase exists for incommensurate bosonic filling @xcite .
if we fix the number of fermions @xmath101 and inter - species interaction @xmath11 in eq .
( [ parameters2 ] ) still there is a dynamic part of the local lindhard function @xmath229 we have to take into account . in the static limit @xmath230 ( where the lindhard response function is purely real )
there is nothing unexpected in the behavior of the critical line ( see discussion below ) .
however , we must stress that even we left the frequency dependence apart , there is still very interesting part of physics remained , because the lindhard response function for the system with regular density of states shows logarithmic divergence as temperature goes to zero .
these singularities give rise to instabilities in the system towards two new ground states a phase separated state or a supersolid phase @xcite . on the other hand the oscillation of the induced effective interaction between bosons is the origin of the formation of charge density waves @xcite . taking @xmath207 we recover the previous theoretical results where , after adding fermions to the system , the effective interaction @xmath231 becomes smaller than repulsive energy @xmath10 for bosons only ( see fig .
[ fig4 ] and fig .
[ fig5 ] ) and superfluid phase increases .
the above is best shown on fig .
[ fig6 ] where for very small amount of fermions @xmath101 the parameter @xmath232 causes decreasing the mott insulator region of the phase diagram in comparison to the pure bosonic case .
however , in the local dynamic limit , when @xmath99 the mott insulator phase is becoming stronger and bosons tend to localize on the lattice sites in both @xmath4 and @xmath5 cases ( see fig .
[ fig6 ] , fig .
[ fig7 ] and fig .
[ fig8 ] ) .
-@xmath197 ) for the square ( @xmath4 ) lattice for different @xmath206 ( higher panel , @xmath207 ) , @xmath208 ( @xmath99 ) and @xmath233 ( positive scattering length @xmath234 ) . dashed line stands for the phase boundary of one - component bose - hubbard model . within the lobes the mott insulator phase takes place with @xmath211.[fig7 ] ] -@xmath197 ) for the cubic ( @xmath5 ) lattice for different @xmath206 ( higher panel , @xmath207 ) , @xmath208 ( @xmath99 ) and @xmath233 ( positive scattering length @xmath234 ) .
dashed line stands for the phase boundary of one - component bose - hubbard model . within the lobes the mott insulator phase takes place with @xmath211.[fig8 ] ] with @xmath222 for different @xmath235 , @xmath236 , @xmath237 , @xmath238 , @xmath239 , @xmath240 , @xmath241 , @xmath242 for square ( @xmath4 ) and cubic ( @xmath5 ) lattice .
vertical solid line stands for the @xmath223 obtained in the one - component bose - hubbard model with one particle per lattice site . above the curves the superfluid phase takes place with @xmath224.[fig9 ] ] one may argue that the lindhard response function depends not only on the frequency @xmath83 but also on the chemical potential @xmath82 for fermions and so far we did not restrict ourselves to any particular value of it .
in many approaches it is a little tricky to handle because despite of the absence of any direct interaction between fermions @xmath243 the density - density fluctuations can indeed induce some effective interaction between fermionic species @xcite .
therefore problem becomes complex and many theories just take half - filled band with @xmath244 so that @xmath60 .
alternative approach comes from partial particle - hole symmetry hamiltonian eq .
( [ hamiltonian ] ) possesses . to make our approach self - consistent we can calculate how does a particular value of the fermionic chemical potential change in the effectively interacting system .
we remind that the amplitude of the order parameter was obtained from minimization condition , assuming nonfluctuating bosonic amplitude at low temperatures . by operating a similar procedure
we get a shift of the chemical potential for fermions @xmath245 that in the non - interacting case @xmath246 reduces to that of free particles obeying fermionic commutation relations ( see also appendix ) .
there is the limit where the system containing gaseous mixtures has the same value of @xmath247 as in the case of only bosons confined in optical lattice ( see fig .
[ fig9 ] ) .
again , we take advantage of the choice of the parameters , that suits well our goal , and make notation of the condition very simple . if @xmath248 we have@xmath249 the above , in terms of the original variables , leads to @xmath250 . for cubic lattice
the formula eq .
( [ condition ] ) seems to not hold ( fig .
[ fig9 ] ) however is accurate with a numerical error less than @xmath251 percent .
therefore , if the number of fermions added to the system is equal to the inter - species interaction then bosons behaves as if were unaffected by the presence of fermions . as a matter of fact we have to remember about sign of the scattering length @xmath31 and normalized frequency @xmath83 that also modifies the introduced condition
the recent experiment@xcite shows that there is an asymmetry in profiles of visibility of the interference pattern ( recorded by absorption imaging ) versus the inter - species scattering length that increases with lattice depth .
presented data indicate that visibility shows a maximum at the position consistent with @xmath252 .
besides , there is an asymmetry in a shift of the mi to sf transition boundary .
our calculations can reproduce latter however if the sign of the inter - species interaction is negative the mi phase diminishes and quite oppositely for positive scattering length and , as we expected , some anomaly appears at the point with @xmath253 ( see fig .
[ fig7 ] and fig .
[ fig8 ] ) .
there is no physical reason for the phase boundary to change a position where the chemical potential for bosons is zero and the repulsive interactions are very strong @xmath254 .
moreover , in that case the value obtained from eq .
( [ critical line-1 ] ) at mentioned point ( @xmath253 and fixed @xmath255 ) is constant in whole nonzero range of the parameter @xmath198 and depends only on the considered topology of the system .
the similar to fermion - boson loss of coherence for the boson - boson species was found using the gutzwiller mean - field approach @xcite .
the main effect of the addition different species of the same statistics is that the new structure of wedding cake appear but the oscillatory behavior of the relevant condensate fraction does not necessarily result in increase of the overall coherence of other species .
the later is limited exclusively to the shallow lattice depth and was never observed in the experiments . we want to stress that one have to be careful with the analysis of the phase diagrams .
the summary of our results for square lattice ( we omit qualitatively similar results for cubic geometry ) is presented on fig ( [ fig10 ] ) and fig ( [ fig11 ] ) . for the static and dynamic limit , but below the critical frequency @xmath207 the mott insulator region on the phase diagram broadens only when the scattering length is positive ( the part of the surface above the plane of the critical value of @xmath256 for the one - component bh model ) .
when we must take into account the difference in the inter - species masses @xmath257 the sign of the real part of the local lindhard response function may become positive and , in consequence , @xmath198 parameter takes values above one fig .
( [ fig11 ] ) leading to higher repulsive energy between bosons
even if measured scattering length is negative . in that case for
@xmath258 there is always a shift for higher values for the parameter @xmath259 results in stronger localization of the bosons after adding fermions to the system .
with the negative real part of the local dynamic lindhard function @xmath2@xmath260 ( @xmath207 ) , for cubic ( @xmath5 ) lattice in the space of the dimensionless parameters @xmath261 with @xmath222 .
the flat surface stands for @xmath262 ( see also eq . [ condition ] ) in case of the one - component bose - hubbard model with one particle per lattice site .
the dashed line stands for the condition where the system of gaseous mixtures has the same value @xmath223 as only bosons confined in optical lattice . above the surfaces the superfluid phase takes place with @xmath224 . ] with the positive value of the real part of the local dynamic lindhard function @xmath2@xmath263 ( @xmath99 ) , for cubic ( @xmath5 ) lattice in the space of the dimensionless parameters @xmath261 with @xmath222 .
the flat surface stands for @xmath262 ( see also eq . [ condition ] ) in case of the one - component bose - hubbard model with one particle per lattice site .
the dashed line stands for the condition where the system of gaseous mixtures has the same value @xmath223 as only bosons confined in optical lattice . above the surfaces the superfluid phase takes place with @xmath224 . ]
it is well known that the ground state of a system of repulsively interacting bosons confined in a periodic potential can be either in a superfluid or in a mott - insulting state , characterized by integer boson densities . because the phase of the order parameter and the particle number , as conjugate variables , are subject to the uncertainty principle @xmath264 ,
so the bosons can either be in the eigenstate of particle number or phase .
the eigenstate of phase is a superfluid and that of particle number is a localized mott insulator .
therefore , the quantum mi - sf phase transition takes place as the particle density is shifted thus facilitating emergence of the superfluid from the mott insulating state . adding to bosons particles of different statistics and allowing for the mutual repulsion or attraction between species
strongly affects the equilibrium properties .
we presented a field - theoretic study of the ground - phase diagram in quantum two- and three - dimensional gaseous bose - fermi condensates where mentioned emulation takes place .
we calculated the phase diagram using the quantum rotor approach that can reproduce the asymmetry in a shift of the mi to sf transition boundary for positive and negative inter - species scattering length .
analysis of the local dynamic lindhard function revealed the critical value of the frequency for the collective excitations , where the real part of the response function ( and in consequence the interaction between bosons and fermions ) alters sign .
the choice of the parameters of the model led to simple condition for the experimentally accessible parameters within the phase diagram for bose - fermi mixtures is qualitatively the same as for one - component repulsively interacting bose system .
we also compared the maximum of the critical value for @xmath196 parameter ( as a function of the normalized chemical potential @xmath197 ) at the tip of the @xmath203th mi lobe for square and cubic lattice with numerical diagrammatic method and found them in a good agreement especially for higher , experimentally realizable , filling factors .
the nice feature of presented approach , described in details above , is that all the expressions and handling are analytic .
it is also worth to notice that provided local approximation can be very useful in various situations whenever the retardation effects has to be taken into account and we are not interested in effects caused by non - locality .
the third term of the trace ( eq .
[ trace ] ) can be written after fourier transform in form @xmath265 in the above we picked up some special value of the frequency @xmath266 .
now , doing the inverse fourier transform and using gradient expansion@xmath267\ ] ] we obtain local , in the matsubara - imaginary time , quadratic form of the trace in the bosonic variables @xmath268^{2}.\ ] ] we performed an expansion were not the @xmath269 degree of freedom itself but rather its gradients of @xmath270 are assumed to be small .
the explicit formula of the imaginary part for the dynamical lindhard function is calculated in the next section .
to stay in the local regime and using an analytic continuation @xmath68@xmath69 , where @xmath271 comes from the causality relation of the response function @xmath272 where symbol @xmath273 denotes the cauchy principal value which prevents divergence when @xmath274 , we calculate the explicit value of the imaginary part that is somewhat easier to obtain than the real part @xmath275-f\left[t_{f\boldsymbol{k}+\boldsymbol{q}},\mu_{f},\beta\right]}{t_{f\boldsymbol{k}}-t_{f\boldsymbol{k}+\boldsymbol{q}}-\omega+i\epsilon}\nonumber \\ & = & \lim_{\epsilon\rightarrow0^{+}}\frac{1}{\left(2\pi\right)^{4}}\int_{-\pi}^{+\pi}d\boldsymbol{q}d\boldsymbol{k}\frac{f\left[t_{f\boldsymbol{k}},\mu_{f},\beta\right]-f\left[t_{f\boldsymbol{k}+\boldsymbol{q}},\mu_{f},\beta\right]}{t_{f\boldsymbol{k}}-t_{f\boldsymbol{k}+\boldsymbol{q}}-\omega+i\epsilon}\\ & = & \lim_{\epsilon\rightarrow0^{+}}\int_{-\infty}^{+\infty}dx\int_{-\infty}^{+\infty}dy\frac{f\left(x,\mu_{f},\beta\right)-f\left(y,\mu_{f},\beta\right)}{x - y-\omega+i\epsilon}\rho\left(x\right)\rho\left(y\right)\nonumber \\ & = & \int_{-\infty}^{+\infty}d\omega'\frac{a\left(\omega',\mu_{f}\right)}{\omega'-\omega}\nonumber \end{aligned}\ ] ] in the low temperature limit @xmath67 the fermi distribution becomes @xmath276 and we write @xmath277\nonumber \\ & \times & \rho\left(x\right)\rho\left(x-\omega'\right)\end{aligned}\ ] ] that satisfies sum rule @xmath278 .
therefore , we can also calculate the real part @xmath279 from these results one finds that @xmath229 is an even function of frequency while @xmath280 is odd .
in the non - interacting case the number of fermions on the lattice can be calculated as follows @xmath281 + 1}\nonumber \\ & \overset{t\rightarrow0}{= } & \frac{1}{n}\sum_{\boldsymbol{k}}\int_{-\infty}^{+\infty}d\xi\delta\left(\xi - t_{f\boldsymbol{k}}\right)\left[1-\theta\left(\xi-\mu_{f}\right)\right]\nonumber \\ & = & 1-\int_{-\infty}^{+\infty}d\xi\rho\left(\xi\right)\theta\left(\xi-\mu_{f}\right).\end{aligned}\ ] ] with @xmath282 being the density of states for chosen lattice geometry . introducing a shift eq .
( [ shift ] ) in the above we are able to obtain the particular value of the chemical potential for fermions with effective interaction induced by the coupling with bosonic species .
we are grateful to n. teichmann for providing the diagrammatic perturbation theory data used in table .
we thank r. micnas for fruitful , stimulating discussions that allowed to improve some parts of the paper . | we study the ground state phase diagram of a mixture of bosonic and fermionic cold atoms confined on two- and three - dimensional optical lattices
. the coupling between bosonic fluctuations and fermionic atoms can be attractive or repulsive and has similarities with electron - phonon coupling in crystals .
we investigate behavior of the mixtures in the limit , where the bogoliubov sound velocity that dictates bosonic dynamics is comparable to the fermi velocity , hence the retardation effects are important part of the physics .
the dynamic lindhard response function of the fermionic density to changes in the bosonic number of particles above some critical frequency can alter the sign and in consequence the inter - species interaction between particles becomes repulsive in contrast to the static limit ( instantaneous and always attractive ) . considering the above we show that the structure of the phase diagrams crucially depends on the difference in masses of the bosons and fermions .
we discuss the situations where integrating out fermionic field provides an additional interaction that can decrease or increase bosonic coherence . |
an anomaly in field theory occurs if a symmetry of the action or the corresponding conservation law , valid in the classical theory , is violated in the quantized version .
this surprising feature of quantum theory discovered by adler @xcite , bell and jackiw @xcite , and by bardeen @xcite in 1969 plays a fundamental role in physics ( for details see refs.@xcite
@xcite ) . physically there is a difference between external and internal symmetries .
the breakdown of an external symmetry is not dangerous for the consistency of the theory , on the contrary , it provides for instance the physical explanation for the @xmath1-decay @xcite , @xcite or the solution to the @xmath2 problem in qcd @xcite . on the other hand
, the breakdown of an internal symmetry ( i.e. gauge symmetry ) leads to an inconsistency of the quantum theory , the anomalous ward identities destroy the renormalizability of the theory @xcite , and also the unitarity of the @xmath3-matrix may be lost @xcite . to avoid such anomalies imposes severe restrictions to the physical content of a theory .
for instance , in the famous @xmath4 standard theory for electroweak interactions one had to demand the existence of the top quark long before it was discovered .
gravitation regarded as a gauge theory also suffers from anomalies .
the gauges are the general coordinate transformations ( diffeomorphisms ) or the rotations in the tangent frame ( lorentz transformations ) or the conformal transformations ( weyl transformations ) .
then in the quantum case the classical conservation law of the energy - momentum tensor can be broken an einstein anomaly occurs or an antisymmetric part of the energy - momentum tensor can exist a lorentz anomaly occurs or the trace of the tensor is nonvanishing a weyl anomaly arises ( for details see e.g. refs.@xcite,@xcite , @xcite ) .
whereas the anomaly in the tensor trace has been found already in the seventies @xcite @xcite the study of the gravitational anomalies started with the pioneering work of alvarez - gaum and witten @xcite in the eighties .
the anomalies have been first found within perturbation theory , they are local polynomials in the connection and curvature . the authors @xcite @xcite have calculated ( ultraviolet divergent ) feynman diagrams where the external gravitational field couples to a fermion loop via the energy - momentum tensor .
of course , the anomaly reflecting the deep laws of quantum physics must show up within other approaches too . so they have been calculated by the heat kernel method @xcite , @xcite , by fujikawa s path integral approach @xcite , @xcite , and by modern mathematical techniques such as differential geometry and cohomology @xcite @xcite and topology ( index theorems ) @xcite , @xcite , @xcite , ( for an overview see ref.@xcite ) . deeply related to anomalies
are the socalled schwinger terms @xcite @xcite ( for an introduction see e.g. refs .
@xcite , @xcite ) . in a yang - mills gauge theory schwinger terms
( st ) show up as additional terms ( extensions ) in the canonical algebra of the equal time commutators ( etc ) of the gauss law operators ( see e.g. @xcite
cohomologically they are described by the faddeev - mickelsson cocycle @xcite , @xcite , and geometrically they can be related to a berry phase in the vacuum functional @xcite @xcite .
st are frequently calculated within perturbation theory where the bjorken - johnson - low limit @xcite , @xcite works very well .
however , the definition of a point - splitting method turns out to be more subtle and might not lead to the correct result @xcite , @xcite . in gravitation schwinger terms
occur in the etc of the energy - momentum tensors , c - number terms that are proportional to derivatives of the @xmath0-function .
they can be related to the gravitational anomalies @xcite , and they have been calculated explictly via the invariant spectral function and via cohomological techniques @xcite . furthermore there exists an interesting relation of the st to the curvature of the determinant line bundle @xcite , @xcite .
our work deals with the calculation of the gravitational anomalies , specifically the einstein anomaly and the weyl anomaly .
the lorentz anomaly is not independent of the einstein anomaly , both types of anomalies can be shifted into each other by a suitable counterterm @xcite . for convenience
we choose the case where the lorentz anomaly is vanishing .
we also calculate the gravitational schwinger terms .
the purpose of our work is to show that all these anomalous features are easily obtained by the method of dispersion relations , a less familiar but very useful approach .
some of our results we have briefly presented in refs .
@xcite , @xcite .
already since their first introduction into quantum field theory @xcite dispersion relations ( dr ) proved to be a very valuable tool . in connection with anomalies
dr have been fomulated by dolgov and zakharov @xcite and also by kummer @xcite . in the following several authors
@xcite @xcite used successfully dr to determine the anomalies in the chiral current .
recently hoej ' i and schnabl @xcite have applied the method to the well - known trace anomaly @xcite , @xcite which is related to the broken dilatation ( or scale ) invariance .
we extend in our work the method of dr to the case of pure gravitation .
so we consider chiral fermion loops coupled to gravitation for their evaluation it is enough to use gravitation as an external or nonquantized field and we show that the gravitational anomalies and the schwinger terms are , in fact , completely determined by dispersion relations .
all calculations are performed in two dimensions .
conceptuelly the dr approach is an independent and complementary view of the anomaly phenomenon as compared to the ultraviolet regularization procedures . within
dr the anomaly manifests itself as a very peculiar infrared feature of the imaginary part of the amplitude .
but as we shall show there is a link between the two approaches , the dr method and the n - dimensional regularization precedure . our paper is organized as follows . in section 2
we present the general structure of the considered ( pseudo- ) tensor amplitude and we discuss the ward identities which we have to study . in section 3 we introduce the dispersion relations for the relevant formfactors of the amplitude and calculate their imaginary parts via the cutkosky rule . in order to reproduce the dr results in a definite ultraviolet regularization scheme
we have worked out in detail the t hooft - veltman regularization procedure in section 4 and we have compared the several amplitudes with the results of tomiya @xcite and alvarez - gaum and witten @xcite .
the equivalence between the dispersive approach and the dimensional regularization procedure is given in section 5 . in section 6
we derive the anomalous ward identities and explain the source of the anomaly in the dr approach . from the ward identities we deduce the linearized gravitational anomalies
the einstein- and the weyl anomaly and we also determine their covariant versions , a comparison with the exact results is given .
the gravitational schwinger terms occuring in the etc of the energy - momentum tensors we calculate in section 7 , where we adapt the dispersive approach to a method proposed by klln @xcite .
finally we summarize our main results in section 8 .
in two dimensions the lagrangian describing a weyl fermion in a gravitational background field can be written as @xmath5 where @xmath6 is the zweibein and @xmath7 its inverse @xmath8 .
the determinant of the zweibein is @xmath9 and @xmath10 is the covariant derivative with the spin connection @xmath11 .
we use the following conventions in 2 dimensions : * for the flat metric @xmath12 * for the epsilon tensor @xmath13 * for the dirac matrices @xmath14 the einstein and the weyl anomaly are determined by the one - loop diagram in fig.[loop ] and it is sufficient to use the linearized gravitational field @xmath15 since in two dimensions the spin connection @xmath11 does not contribute ( see e.g. ref.@xcite ) we find the following linearized interaction lagrangian ( for convenience @xmath16 is absorbed into @xmath17 , @xmath18 acts only on @xmath19 ) @xmath20 and @xmath21 from this expression follow the feynman rules for the vertices in the loop diagram @xmath22 and the explicit form of the ( symmetric ) energy - momentum tensor @xmath23 where we have dropped the terms proportional to @xmath24 as they do not contribute to the amplitude .
+ then the whole amplitude is given by the two - point function @xmath25 due to lorentz covariance and symmetry the general structure of the amplitude can be written in the following way @xmath26 @xmath27 @xmath28 t_{8}(p^{2 } ) \ , , \label{formfactors3}\end{aligned}\ ] ] where we have separated the amplitude into its pure tensor part @xmath29 ( coming from the vector piece of the chirality projection in eq.([energy - momentum - tensor ] ) ) and into its pseudo - tensor part @xmath30 ( coming from the axial piece in eq.([energy - momentum - tensor ] ) ) . the functions @xmath31 are the formfactors that are to be evaluated . +
classically the energy - momentum tensor has the following properties : 1 .
@xmath32 , symmetric 2 .
@xmath33 , conserved 3 .
@xmath34 , traceless which lead to the canonical ( naive ) ward identities : 1 .
@xmath35 2 .
@xmath36 3 .
@xmath37 .
we are interested in the pure einstein anomaly therefore we demand the quantized energy - momentum tensor to be symmetric .
this is always possible due to the bardeen zumino theorem @xcite which states that the gravitational anomaly can be shifted from pure lorentz type to pure einstein type , and vice versa .
( we disregard here leutwyler s point of view @xcite who emphasizes his preference for the lorentz anomaly ) .
thus the symmetry property 1 . ) of the amplitude is fulfilled and an other symmetry @xmath38 is trivially satisfied .
however , the naive ward identities 2 . ) and 3 . )
need not be satisfied , they can be broken by the einstein and weyl anomalies respectively .
the canonical ward identities we re - express by the formfactors @xmath39 { } \nonumber\\{}&&+ \varepsilon_{\rho \tau } p^{\tau}\left[p_{\nu}p_{\sigma } ( p^{2 } t_{6 } + t_{7 } + t_{8 } ) + g_{\nu \sigma } p^{2 } t_{8}\right ] { } \nonumber\\{}&&+ \varepsilon_{\sigma \tau } p^{\tau}\left[p_{\nu}p_{\rho } ( p^{2 } t_{6 } + t_{7 } + t_{8 } ) + g_{\nu \rho } p^{2 } t_{8}\right ] \label{pta}\end{aligned}\ ] ] and @xmath40 where @xmath41 is the dimension .
we call from now on ward identity ( wi ) the property 2 . ) and trace identity ( ti ) the property 3 . ) . for the pure tensor part of the amplitude
the wi may be written as @xmath42 and the ti as @xmath43 of course , relations ( [ ti2 ] ) and ( [ ti1 ] ) are not independent of each other , ( [ ti2 ] ) follows from ( [ ti1 ] ) and ( [ vwi1 ] ) ( [ vwi3 ] ) . in the following we shall use a renormalization procedure which keeps the wi in the pure tensor part ( [ vwi1 ] ) ( [ vwi3 ] ) so that the anomaly occurs only in the axial pieces representing the pseudotensor part of the amplitude . + calculating the amplitude with massive fermions ( the loop in fig .
[ loop ] ) with help of the feynman rules gives @xmath44 \frac{1\pm\gamma_{5}}{2}{}\nonumber\\&&{}\times \frac{/\hspace{-6.5pt}p + /\hspace{-6.5pt}k+m}{(p+k)^{2}-m^{2}+i \varepsilon } \left [ \gamma_{\rho } ( p+2k)_{\sigma } + \gamma_{\sigma } ( p+2k)_{\rho}\right ] \frac{1\pm\gamma_{5}}{2 } \frac{/\hspace{-6pt}k+m}{k^{2}-m^{2}+i \varepsilon } \biggr\}.\end{aligned}\ ] ] for convenience we split the pure tensor piece of the loop in the following way @xmath45 where @xmath46 represents the loop with the identity instead of the chirality projectors and @xmath47 denotes the part proportional to @xmath48 .
we separate the no interchange amplitudes as @xmath49 finally , the axial part of the amplitude is connected to the vector part due to relation @xmath50 ( valid only in 2 dimensions for our conventions ( [ definition of the flat metric ] ) - ( [ definition of the gamma matrices ] ) ) .
a symmetric decomposition turns out to be useful @xmath51
the formfactors of the amplitude ( [ formfactors1 ] ) ( [ formfactors3 ] ) are analytic functions in the complex @xmath52 plane except a cut on the real axis starting at @xmath53 . due to cauchy s theorem
they can be expressed by dispersion relations which relate the real part of the amplitude to its imaginary part .
the imaginary parts of the amplitude can be easily calculated via cutkosky s rule @xcite .
it states to replace in the amplitude each propagator by its discontinuity on mass shell @xmath54 { } \nonumber\\ { } & & \!\!\!\times \delta(k^{2}-m^{2 } ) \delta\bigl((p+k)^{2}-m^{2}\bigr ) \theta(-k^{0 } ) \theta(k^{0}+p^{0 } ) \label{abstract i m tni}\\ i m t^{dv ,
ni}_{\mu \nu \rho \sigma}(p ) & = & \frac{1}{32 } m^{2 } g_{\mu\rho } \int \!d^{2}k ( p+2k)_{\nu } ( p+2k)_{\sigma}{}\nonumber\\ { } & & \!\!\!\times \delta(k^{2}-m^{2 } ) \delta\bigl((p+k)^{2}-m^{2}\bigr ) \theta(-k^{0 } ) \theta(k^{0}+p^{0 } ) \ , .
\label{abstract i m tdvni}\end{aligned}\ ] ] the explicit integration gives @xmath55 p_{\mu}p_{\nu}p_{\rho}p_{\sigma}{}\nonumber\\ { } & & \qquad + \left [ -\frac{1}{6}p^{2}+\frac{4}{3}m^{2}-\frac{8}{3}\frac{m^{4}}{p^{2}}\right ] \left(p_{\mu}p_{\nu } g_{\rho \sigma } + p_{\mu}p_{\sigma } g_{\nu \rho } + p_{\nu}p_{\rho } g_{\mu \sigma } + p_{\rho}p_{\sigma } g_{\mu \nu}\right ) { } \nonumber\\ { } & & \qquad + \left [ \frac{1}{3}p^{2}-\frac{2}{3}m^{2}-\frac{8}{3}\frac{m^{4}}{p^{2}}\right ] p_{\mu}p_{\rho } g_{\nu \sigma } + \left[\frac{1}{3}p^{2}-\frac{5}{3}m^{2}+\frac{4}{3}\frac{m^{4}}{p^{2}}\right ] p_{\nu}p_{\sigma } g_{\mu \rho } { } \nonumber\\ { } & & \qquad + \left[\frac{1}{6}(p^{2})^{2}-\frac{4}{3}p^{2}m^{2}+\frac{8}{3}m^{4}\right ] \left(g_{\mu \nu } g_{\rho \sigma } + g_{\mu \sigma } g_{\nu \rho}\right ) { } \nonumber\\{}&&\qquad + \left[-\frac{1}{3}(p^{2})^{2}+\frac{5}{3}p^{2}m^{2}-\frac{4}{3}m^{4}\right ] g_{\mu \rho } g_{\nu \sigma } \biggr\ } \\ & & i m t^{dv , ni}_{\mu \nu \rho \sigma}(p ) = \frac{1}{32 } j_{0 } m^{2 } g_{\mu\rho } \biggl\ { \left ( -p^{2}+4m^{2 } \right ) g_{\nu\sigma } + \left(1 - 4\frac{m^{2}}{p^{2 } } \right ) p_{\nu}p_{\sigma } \biggr\},\end{aligned}\ ] ] with the threshold function @xmath56 from which we quickly find all imaginary parts of the formfactors in the total amplitude @xmath57 : @xmath58 considering in addition the amplitude @xmath46 we have the following imaginary parts of the formfactors which we denote by : @xmath59 clearly , the imaginary parts ( [ i m a1 ] ) ( [ i m a5 ] ) of the amplitude @xmath46 satiesfy the wi ( [ vwi1])([vwi3 ] ) with @xmath60 , and the subtraction procedure we choose in the following keeps this property for the entire formfactors @xmath61 .
+ now we start with an unsubtracted dispersion relation ( dr ) for the formfactors @xmath62 and we observe that , for instance , the integral for @xmath63 is convergent whereas for @xmath64 it is logarithmically divergent and needs to be subtracted once , and for @xmath65 it is linearly divergent and needs to be subtracted twice
. we can infer already from the @xmath52 behaviour of the imaginary parts which kind of dispersion relation we have to use , see table [ subtraktionen ] .
.the formfactors and the used type of dispersion relations [ cols= " < , > " , ] so for the formfactors @xmath66 an unsubtracted dr is sufficient and we get @xmath67 \label{t1 in dr}\end{aligned}\ ] ] with @xmath68 a once subtracted dr defined by @xmath69 we use for the following formfactors ( see table 1 ) @xmath70 for the remaining formfactors a twice subtracted dr defined by @xmath71 is necessary and we find @xmath72\\ t^{r}_{5}(p^{2 } ) & = & \frac{p^{4}}{\pi } \int\limits^{\infty}_{4m^{2}}\!\!\frac{dt}{t - p^{2 } } \frac{1}{t^{3}}\left(1-\frac{4m^{2}}{t}\right)^{\hspace{-3pt}-\frac{1}{2 } } \hspace{-3pt } \left ( -\frac{1}{96}t^{2 } + \frac{1}{48}t m^{2}+\frac{1}{12 } m^{4 } \right ) { } \nonumber\\{}&= & p^{2}\left[-\frac{5}{288\pi } - \frac{1}{24\pi } \frac{m^{2}}{p^{2 } } + \frac{1}{48\pi } \bigl(1 + 2 \frac{m^{2}}{p^{2 } } \bigr ) a(p^{2})\right]\\ a^{r}_{5}(p^{2 } ) & = & \frac{p^{4}}{\pi } \int\limits^{\infty}_{4m^{2}}\!\!\frac{dt}{t - p^{2 } } \frac{1}{t^{3}}\left(1-\frac{4m^{2}}{t}\right)^{\hspace{-3pt}-\frac{1}{2 } } \hspace{-3pt } \left ( -\frac{1}{48}t^{2 } - \frac{1}{12}t m^{2}+\frac{2}{3}m^{4}\right ) { } \nonumber\\{}&= & p^{2}\left[-\frac{1}{72\pi } - \frac{1}{3\pi } \frac{m^{2}}{p^{2 } } + \frac{1}{24\pi } \bigl(1 + 8 \frac{m^{2}}{p^{2 } } \bigr ) a(p^{2})\right ] .
\label{a5r in dr}\end{aligned}\ ] ] with these explicit expressions for the formfactors we have determined the whole amplitude @xmath57 , eqs.([amplitude expressed by t - product])([formfactors3 ] ) , from which the correct ward identities will follow .
although the method of dispersion relations appears quite different to the methods of regularizations we can reproduce its results in a definite regularization scheme , namely in the n - dimensional regularization procedure of t hooft veltman @xcite .
it is instructive to work it out in more detail .
we start with amplitude ( [ amplitude by dimensional regularization with massive fermions ] ) , calculate the @xmath73-matrices and follow the standard procedure by inserting the feynman parameter integral @xmath74^{2 } } \ , \ , .\ ] ] then we obtain for the ` no interchange ' amplitudes of the pure tensor pieces @xmath75^{2 } } { } \nonumber\\{}&&\quad \times \bigl [ ( l+px)_{\mu}(l - p(1-x))_{\rho } + ( l+px)_{\rho}(l - p(1-x))_{\mu } - g_{\mu\rho } ( l+px)^{\lambda}(l - p(1-x))_{\lambda } \bigr]\end{aligned}\ ] ] and @xmath76^{2 } } \ , \ , , \ ] ] with @xmath77 calculating the t hooft veltman integrals @xmath78^{\alpha } } = \frac{(-1)^{\alpha}i } { ( 4\pi)^{\omega } } \frac{\gamma(\alpha - \omega)}{\gamma(\alpha ) } \delta^{\omega-\alpha } \\
p_{1}^{\mu\nu } & = & \int\!\ ! \frac{d^{2\omega}l}{(2\pi)^{2\omega } } \frac{l^{\mu}l^{\nu}}{[l^{2}-\delta]^{\alpha } } = \frac{\delta}{2(\omega - \alpha + 1 ) } g^{\mu\nu } p_{0 } \label{integral p1}\\ p_{2}^{\mu\nu\rho\sigma } & = & \int\!\ ! \frac{d^{2\omega}l}{(2\pi)^{2\omega } } \frac{l^{\mu } l^{\nu } l^{\rho } l^{\sigma } } { [ l^{2}-\delta]^{\alpha } } { } \nonumber\\{}&= & \frac{\delta^{2}}{4(\omega - \alpha + 1)(\omega - \alpha + 2 ) } ( g^{\mu\nu } g^{\rho\sigma } + g^{\mu\rho } g^{\nu\sigma } + g^{\mu\sigma } g^{\nu\rho } ) p_{0 } \label{integral p2}\end{aligned}\ ] ] with @xmath79 provides the following expressions @xmath80 p_{\nu}p_{\sigma } g_{\mu \rho } { } \nonumber\\ { } & & \qquad + 2 \frac{\delta^{2 } } { \omega(\omega-1 ) } ( g_{\mu \nu } g_{\rho \sigma } + g_{\mu \sigma } g_{\nu \rho } ) + \left [ -2\omega \frac{\delta^{2 } } { \omega(\omega-1 ) } + 2x(1-x)p^{2 } \frac{\delta}{\omega-1 } \right ] g_{\mu \rho } g_{\nu \sigma } \biggr\}\label{explicit tni in massive dim reg}\\ & & t^{dv , ni}_{\mu \nu \rho \sigma}(p ) = - i \frac{2^{\omega}}{32 } m^{2 } g_{\mu\rho } \int\limits_{0}^{1 } \!dx p_{0 } \biggl\ { 2 \frac{\delta}{\omega-1 } g_{\nu\sigma } + ( 1 - 2x)^{2 } p_{\nu}p_{\sigma } \biggr\}\label{explicit tdvni in massive dim reg}.\end{aligned}\ ] ] these expressions we compare now with the formfactor decompositions ( [ formfactors2 ] ) and ( [ formfactors3 ] ) ( recall eqs.([without interchange ] ) and ( [ axial part with gamma5 symmetrically distributed ] ) ) then we obtain all formfactors of the amplitude @xmath57 explicitly @xmath81 the formfactors of the amplitude @xmath46 follow via eq.([pure vector amplitude ] ) @xmath82 as we can see the formfactors @xmath83 and @xmath84 ( eqs.([t1 in massive dim reg])and ( [ a1 in massive dim reg ] ) ) are finite in the limit @xmath85 and explicitly we find @xmath86 \label{t1r in massive dim reg}\end{aligned}\ ] ] with @xmath87 given by eq.([definition of a ] ) .
however , the remaining formfactors are divergent in the limit @xmath85 and we have to renormalize them in an appropriate way .
we separate the formfactors into a divergent part and a finite part @xmath88 so that we can extract the finite result by a suitable prescription .
the formula we need for this procedure is @xmath89 - \frac{1}{2\pi}\left.\frac{df}{d\omega}\right|_{\omega=1}+ o(1-\omega ) \ , .\end{aligned}\ ] ] it is the pole part @xmath90 of the formfactor which tells us which kind of renormalization we have to choose , for instance , for a constant a simple subtraction is sufficient , for a term proportional to @xmath91 a double subtraction . in this way
the following formfactors are determined by a simple subtraction @xmath92 whereas for the remaining formfactors a double subtraction is needed @xmath93 { } \nonumber\\ { } & = & p^{2}\left [ \frac{1}{18\pi } - \frac{1}{6\pi } \frac{m^{2}}{p^{2 } } - \frac{1}{24\pi } \bigl(1 - 4 \frac{m^{2}}{p^{2 } } \bigr ) a(p^{2})\right]\label{t4r in massive dim reg}\\ t^{r}_{5}(p^{2 } ) & = & t_{5}(p^{2 } ) - t_{5}(0 ) - p^{2 } \left.\frac{d}{dp^{2}}t_{5}(p^{2})\right|_{p^{2}=0 } { } \nonumber\\ { } & = & \frac{1}{16\pi } \int\limits^{1}_{0}\!dx \ p^{2}x(1-x ) \ln \frac{m^{2}-p^{2}x(1-x)}{m^{2 } } { } \nonumber\\{}&= & p^{2}\left[-\frac{5}{288\pi } - \frac{1}{24\pi } \frac{m^{2}}{p^{2 } } + \frac{1}{48\pi } \bigl(1 + 2 \frac{m^{2}}{p^{2 } } \bigr ) a(p^{2})\right]\\ a^{r}_{5}(p^{2 } ) & = & a_{5}(p^{2 } ) - a_{5}(0 ) - p^{2 } \left.\frac{d}{dp^{2}}a_{5}(p^{2})\right|_{p^{2}=0 } { } \nonumber\\ { } & = & \frac{1}{8\pi } \int\limits^{1}_{0}\!dx \biggl[\bigl ( m^{2 } + p^{2}x(1-x ) \bigr ) \ln \frac{m^{2}-p^{2}x(1-x)}{m^{2 } } + p^{2}x(1-x)\biggr ] { } \nonumber\\{}&= & p^{2}\left[-\frac{1}{72\pi } - \frac{1}{3\pi } \frac{m^{2}}{p^{2 } } + \frac{1}{24\pi } \bigl(1 + 8 \frac{m^{2}}{p^{2 } } \bigr ) a(p^{2})\right ] .
\label{a5r in massive dim reg}\end{aligned}\ ] ] of course , the formfactors @xmath94 of the tensor amplitude @xmath46 satisfy the wi ( [ vwi1 ] ) ( [ vwi3 ] ) and as the difference @xmath47 to the pure tensor amplitude is proportional to @xmath48 ( recall eqs.([pure vector amplitude]),([explicit tdvni in massive dim reg ] ) ) the amplitude @xmath29 will also satisfy the wi ( [ vwi1 ] ) ( [ vwi3 ] ) in the limit @xmath95 .
this procedure works in complete analogy to the dispersion relation approach ( see table 1 ) and we clearly obtain the same results for the formfactors as one can see by comparing eqs.([t1r in massive dim reg])([a5r in massive dim reg ] ) with ( [ t1 in dr])([a5r in dr ] ) . even more , as we shall show in the next chapter , the two approaches are equivalent .
but before we want to consider the case of massless fermions , the limit @xmath95 , in order to compare our results with the ones of other authors .
taking the limit @xmath95 in the @xmath96 expansion ( eq.([pole terms ] ) ) of the formfactors we find the following results @xmath97 \label{t2 for wi } \\ t_{3}(p^{2 } ) & = & -\frac{1}{p^{2}}t_{5}(p^{2 } ) = \mp 4\ t_{8}(p^{2 } ) { } \nonumber\\{}&= & -\frac{1}{96\pi } \biggl [ \frac{1}{\omega-1 } + \gamma - \frac{2}{3 } + \ln\left|\frac{p^{2}}{2\pi\mu^{2}}\right| - i\pi \biggr ] \label{t3 for wi}.\end{aligned}\ ] ] which agree with the expressions we obtain when starting with @xmath98 from the very beginning ( we also introduced here the mass @xmath99 to keep the correct dimensionality , see ref.@xcite ) .
now , for comparison tomiya @xcite in his work on the gravitational anomaly defines the amplitude with the covariant @xmath100-product and calculates the following formfactors @xmath101 they also satiesfy the wi ( [ vwi1 ] ) ( [ vwi3 ] ) , however , they ( some of them ) differ from ours ( [ t1 for wi ] ) ( [ t3 for wi ] ) , which is not surprising as the covariant @xmath100-product differs from the ordinary @xmath102-product by suitable seagull terms . but as we shall show below his and our results lead to the same amplitudes and consequently to the same anomaly expression . using on one hand the wi ( [ vwi1 ] ) ( [ vwi3 ] ) for the formfactors @xmath103 in the massless limit , @xmath95 , and on the other the fact that @xmath104 we find for the several amplitudes ( [ formfactors1 ] ) ( [ formfactors3 ] ) the following expressions @xmath105 as we can see all amplitudes depend only on the ( convergent ) formfactor @xmath63 , explicitly given by eq.([t1 for wi ] ) , and are independent of a second ( divergent ) formfactor , say @xmath64 , which one might have expected to contribute too ( since we have six restrictive relations between the eight formfactors @xmath103 ) .
so the amplitudes do not determine the formfactors uniquely ! therefore @xmath64 can be chosen at will and tomiya @xcite in his work renormalized the formfactors conveniently such that @xmath64 came out to be zero .
this is in accordance with the amplitude result of deser and schwimmer @xcite , @xcite who consider fields without chirality . on the other hand ,
alvarez - gaum and witten @xcite consider the gravitational anomaly within light - cone coordinates and use a specific regularization prescription to calculate the following amplitude @xmath106 their result is consistent with our calculations as we have @xmath107 { } \nonumber\\{}&= & \frac{1\mp1}{24\pi p^{2}}p_{+}^{4 } \ , \ , .\end{aligned}\ ] ]
the dimensional regularization procedure is a method which satisfies the wi ( [ vwi1 ] ) ( [ vwi3 ] ) . on the other hand , within the dispersion relation approach we also have effectively renormalized in a way which keeps this wi property ( [ vwi1 ] ) ( [ vwi3 ] ) .
if we had subtracted more often or at some other point than @xmath108 we would have lost this property .
the dispersive approach and the dimensional regularization procedure are equivalent in the following sense . for
the formfactors that were convergent or logarithmically divergent the corresponding integrals in both approaches do not have just identical values but we even can transform them into each other by a suitable substitution . more precise , we transform the feynman parameter integrals over x in the dimensional regularization procedure into the corresponding dispersion integrals over t. for the formfactors that were linearly divergent the situation is as follows .
accidentially for @xmath109 there exists such a substitution , however , not for @xmath110 and @xmath111 .
but anyway , for these formfactors the integrals are identical , what is actually sufficient .
the substitutions which link the two approaches are @xmath112 to show the equivalence we need the following relations which are valid for any function @xmath113 with suitable differentiation properties @xmath114
now we turn to the calculation of the ward identities and gravitational anomalies .
we consider the massless limit , @xmath95 , where the formfactors @xmath115 ( i = 1, ... ,5 ) fulfill the wi ( [ vwi1 ] ) ( [ vwi3 ] ) .
this means that the wi for the pure tensor part ( [ ptv ] ) is satiesfied @xmath116 calculating next the wi for the pseudo tensor part ( [ pta ] ) we use the formfactor identities ( [ axial vector ] ) and we find @xmath117 in the flat space - time limit @xmath118 we have the relation @xmath119 and we finally obtain the anomalous result @xmath120 as we discovered already before when calculating the amplitudes , the anomalous wi depends only on the finite formfactor @xmath121 with its explicit result ( [ t1 for wi ] ) .
so the anomaly is independent of a specific renormalization procedure ( as long as it preserves the wi ( [ vwi1 ] ) ( [ vwi3 ] ) ) and we clearly agree with the anomaly results of tomiya @xcite and alvarez - gaum and witten @xcite .
we want to emphasize that our subtraction procedure is _ the _ ` natural ' choice dictated by the @xmath122behaviour of the imaginary parts @xmath123 of the formfactors . since on general grounds the imaginary parts of the amplitudes fulfill the wi ( [ vwi1 ] ) ( [ vwi3 ] ) ( in the limit @xmath95 ) the naturally chosen dispersion relations @xmath124=0 \\ & & \frac{1}{\pi}\int\limits^{\infty}_{4m^{2}}\!\!\frac{dt}{t - p^{2}}\frac{p^{4}}{t^2}\left[t\ , im\,t_{2}(t)+im\,t_{4}(t)\right]=0 \\ & & \frac{1}{\pi}\int\limits^{\infty}_{4m^{2}}\!\!\frac{dt}{t - p^{2}}\frac{p^{4}}{t^2}\left[t\ , im\,t_{3}(t)+im\,t_{5}(t)\right]=0\end{aligned}\ ] ] imply the pure tensor wi ( [ vwi1 ] ) ( [ vwi3 ] ) for the renormalized formfactors ( in the limit @xmath95 ) @xmath125 therefore our subtraction procedure automatically shifts the total anomaly into the pseudotensor part of the wi ( [ pta ] ) .
+ what is the origin of the anomaly in this dispersive approach ?
the source of the anomaly is the existence of a superconvergence sum rule for the imaginary part of the formfactor @xmath63 @xmath126 the anomaly originates from a @xmath0-function singularity of @xmath127 when the threshold @xmath128 approaches zero ( the infrared region ) @xmath129 the limit must be performed in a distributional sense .
+ then the unsubtracted dispersion relation for @xmath63 , eq.([unsubtracted dr ] ) , provides the result ( [ t1 for wi ] ) .
this threshold singularity of the imaginary part of the relevant formfactor is a typical feature of the dispersion relation approach for calculating the anomaly ( see e.g. refs.@xcite , @xcite
@xcite , and the appendix b in ref .
+ next we turn to the energy - momentum tensor . from the anomalous
wi @xmath130 we can deduce the linearized consistent einstein ( or diffeomorphism ) anomaly @xmath131 for comparison we demonstrate that result ( [ linearized consistent einstein anomaly ] ) is indeed the linearization of the exact result that follows from differential geometry and topology ( see for instance ref.@xcite ) .
the exact einstein anomaly in two dimensions is given by @xmath132 where @xmath133 therefore we get @xmath134 considering the linearized gravitational field , eq.([linearization of the vielbein with upper indices ] ) , the christoffel symbols become ( with explicit @xmath16-dependence ) @xmath135 so that we find as linearization of the exact result ( [ exact consistent einstein anomaly ] ) @xmath136 this agrees with our result ( [ linearized consistent einstein anomaly ] ) since in two dimensions we have the identity @xmath137 now what about the covariant einstein anomaly ?
it arises when considering the covariantly transforming energy - momentum tensor @xmath138 which is related to our tensor definition ( [ energy - momentum - tensor ] ) by the bardeen - zumino polynomial @xmath139 @xcite @xmath140 this polynomial is calculable and explicitly we find ( in two dimensions ) @xcite @xmath141 ( @xmath142 denotes the ricci scalar ) which leads to the covariant einstein anomaly @xmath143 linearizing the ricci scalar ( with explicit @xmath16-dependence ) @xmath144 and using ( [ linearization of the christoffel symbols ] ) , ( [ epsilon identity ] ) we get @xmath145 so that we find for the linearized covariant einstein anomaly @xmath146 it is twice the linearized consistent result ( [ linearized consistent einstein anomaly ] ) as it should be .
+ finally we also calculate the trace identity , eqs.([gtv ] ) and ( [ gta ] ) . using again relations ( [ axial vector ] )
we find @xmath147,\ ] ] and taking into account the wi ( [ vwi1 ] ) provides us the anomalous result @xmath148.\ ] ] the anomalous ti depends only on the finite formfactor @xmath121
what we know already from the previous discussion
so that it is independent of a specific renormalization procedure ( as long as it preserves the wi ( [ vwi1 ] ) ( [ vwi3 ] ) ) . inserting the formfactor , eq.([t1 for wi ] ) , gives @xmath149,\ ] ] which implies the following linearization of the weyl ( or trace ) anomaly @xmath150.\ ] ] again , we compare our calculation with the exact result for the weyl anomaly which is given by ( see for instance ref.@xcite ) @xmath151 and consequently we have @xmath152 considering the linearizations of the spin connection ( with explicit @xmath16-dependence ) @xmath153 and of the ricci scalar eq.([linearization of r ] ) , we see that our result ( [ linearized weyl anomaly ] ) is indeed the linearization of the exact expression ( [ exact consistent weyl anomaly ] ) .
adding last but not least the bardeen - zumino polynomial @xmath139 @xmath154 with its linearization @xmath155 we find for the covariant trace anomaly @xmath156 and @xmath157 for its linearized version .
clearly these results are in agreement with ref .
in quantum field theory the energy - momentum tensors form an algebra which is generally not closed but has central extensions , socalled schwinger terms , for example @xmath158 the schwinger terms , the c - number terms @xmath159 , can be determined by considering the vacuum expectation value of the etc . to evaluate the st we work with a technique that has been introduced by klln @xcite and is closely related to the dispersive approach used before .
this technique has been applied already by skora @xcite to compute the st for currents in yang - mills theories .
our aim is to generalize this procedure to the case of gravitation , where the current is replaced by the energy - momentum tensor . from the lagrangian ( [ lagrangian ] ) describing a weyl fermion in a gravitational background field in two dimensions we get the following classical energy - momentum tensor @xmath160 where @xmath161 means normal ordering . using relation ( [ elimination of gamma5 ] ) and the equations of motions we can express the pseudo tensor part of the enery - momentum tensor by the pure tensor part ( recall that the tensor is symmetric ) @xmath162 in two dimensions we have the identity @xmath163 so that we find @xmath164\vert 0\rangle = \frac{1}{4 } \biggl\ { \langle 0\vert [ t^{v}_{\mu\nu}(x),t^{v}_{\rho\sigma}(0)]\vert 0\rangle + \langle 0\vert [ t^{v}_{\rho\nu}(x),t^{v}_{\mu\sigma}(0)]\vert 0\rangle { } \nonumber\\ { } & & \qquad- g_{\mu\rho } \langle 0\vert [ t^{v\lambda}_{\nu}(x),t^{v}_{\lambda\sigma}(0)]\vert 0\rangle \mp \varepsilon_{\mu}^{\ \lambda}\langle 0\vert [ t^{v}_{\lambda\nu}(x),t^{v}_{\rho\sigma}(0)]\vert 0\rangle \mp \varepsilon_{\rho}^{\ \lambda } \langle 0\vert [ t^{v}_{\mu\nu}(x),t^{v}_{\lambda\sigma}(0)]\vert 0\rangle \biggr\}.\end{aligned}\ ] ] let us define @xmath165 by inserting the completeness relations @xmath166 and using the translation invariance @xmath167 we obtain @xmath168 where the sum runs over many - particle states @xmath169 with positive energy and momentum @xmath170 . + we may write @xmath171 where @xmath172 from lorentz covariance and symmetry we get the same decompositon into formfactors as for @xmath29 ( see eq .
( [ formfactors2 ] ) @xmath173 making use of @xmath174 provides the ward identity @xmath175 that can be expressed by the formfactors @xmath176 now let us explicitly evaluate @xmath177 . as we are considering the energy - momentum tensor as a free ( non interacting ) tensor analogously to the case of free currents we only need to sum up states that consist of one fermion - antifermion pair in eq .
( [ symbolic gmunurhosigma ] ) .
we get @xmath178 where @xmath179 let us assume that the fermion fields are canonically quantized @xmath180\\ \bar\psi(x)&= & \frac{1}{(2\pi)^{1/2}}\int\!dp\sum_{s=1}^{2}\sqrt{\frac{m}{e_{p}}}\left [ b^{\dagger(s)}(p)\bar u^{(s)}(p)e^{ipx}+d^{(s)}(p)\bar v^{(s)}(p)e^{-ipx}\right],\end{aligned}\ ] ] then we find @xmath181 this provides @xmath182 without the interchanges we call this @xmath183 .
+ if we use the completeness relations @xmath184 and the integral equation @xmath185 we obtain @xmath186.\end{aligned}\ ] ] integrating next over the first @xmath0-function and evaluating the trace gives @xmath187.\end{aligned}\ ] ] if we compare this with ( [ abstract i m tni ] ) and ( [ abstract i m tdvni ] ) ( @xmath188 ) , we see that @xmath177 is given by the imaginary part of the amplitude of the pure vector loop @xmath189 this is the important relation which links the schwinger terms to the gravitational anomalies . from eq .
( [ i m a1 ] ) ( [ i m a5 ] ) we read off the formfactors @xmath190 now we consider the commutator @xmath191 in terms of g } \langle 0\vert [ t^{v}_{\mu\nu}(x ) , t^{v}_{\rho\sigma}(0 ) ] \vert 0\rangle = f_{\mu\nu\rho\sigma}(x)-f_{\rho\sigma\mu\nu}(-x ) = \int\!d^{2}p\ e^{-ipx } \varepsilon(p^{0 } ) g_{\mu\nu\rho\sigma}(p).\ ] ] if we remove the mass , @xmath192 , that acted as an infrared cutoff , we get @xmath193 from eq .
( [ [ t , t ] in terms of g ] ) we explicitly find @xmath194 \vert 0\rangle_{et } & = & \lim_{x_{0}\rightarrow 0 } -\frac{1}{24\pi^{2}}\int\!d^{2}p\ e^{-ipx}p_{1}^{4}\ \varepsilon(p^{0 } ) \delta(p^{2 } ) = 0 \\ \langle 0\vert [ t^{v}_{11}(x ) , t^{v}_{11}(0 ) ] \vert 0\rangle_{et } & = & \lim_{x_{0}\rightarrow 0 } -\frac{1}{24\pi^{2}}\int\!d^{2}p\ e^{-ipx}p_{0}^{4}\ \varepsilon(p^{0 } ) \delta(p^{2 } ) = 0 \\ \langle 0\vert [ t^{v}_{00}(x ) , t^{v}_{11}(0 ) ] \vert 0\rangle_{et } & = & \langle 0\vert [ t^{v}_{01}(x ) , t^{v}_{01}(0 ) ] \vert 0\rangle_{et } { } \nonumber\\{}&= & \lim_{x_{0}\rightarrow 0 } -\frac{1}{24\pi^{2}}\int\!d^{2}p\ e^{-ipx}p_{0}^{2}p_{1}^{2}\ \varepsilon(p^{0 } ) \delta(p^{2 } ) = 0 \\ \langle 0\vert [ t^{v}_{00}(x ) , t^{v}_{01}(0 ) ] \vert 0\rangle_{et } & = & \lim_{x_{0}\rightarrow 0 } -\frac{1}{24\pi^{2}}\int\!d^{2}p\ e^{-ipx}p_{0}p_{1}^{3}\ \varepsilon(p^{0 } ) \delta(p^{2 } ) \\
\langle 0\vert [ t^{v}_{01}(x ) , t^{v}_{11}(0 ) ] \vert 0\rangle_{et } & = & \lim_{x_{0}\rightarrow 0 } -\frac{1}{24\pi^{2}}\int\!d^{2}p\ e^{-ipx}p_{0}^{3}p_{1}\ \varepsilon(p^{0 } ) \delta(p^{2 } ) \ , .\label{[t01,t11 ] no interchange } \end{aligned}\ ] ] the first three expressions vanish because @xmath195 is antisymmetric . to evaluate the next two we use the pauli - jordan function @xmath196 with the properties @xmath197 and we find @xmath198 so we conclude @xmath199 \vert 0\rangle_{et } & = & \langle 0\vert [ t^{v}_{01}(x ) , t^{v}_{11}(0 ) ] \vert 0\rangle_{et } { } \nonumber\\{}&= & \lim_{x_{0}\rightarrow 0}-\frac{1}{12\pi}\partial_{1}^{3}\partial_{0}\triangle(x ) = \frac{i}{12\pi}\partial_{1}^{3}\delta(x^{1 } ) \ , .\end{aligned}\
] ] with relation ( [ symbolic vev of commutator ] ) we finally obtain the schwinger terms in the etc of the energy - momentum tensors @xmath200 \vert 0\rangle_{et } & = & \langle 0\vert [ t_{11}(x ) , t_{11}(0 ) ] \vert 0\rangle_{et } = \mp\frac{i}{24\pi}(\partial_{1})^{3}\delta(x^{1})\\ \langle 0\vert [ t_{00}(x ) , t_{11}(0 ) ] \vert 0\rangle_{et } & = & \langle 0\vert [ t_{01}(x ) , t_{01}(0 ) ] \vert 0\rangle_{et } = \mp\frac{i}{24\pi}(\partial_{1})^{3}\delta(x^{1})\\ \langle 0\vert [ t_{00}(x ) , t_{01}(0 ) ] \vert 0\rangle_{et } & = & \langle 0\vert [ t_{01}(x ) , t_{11}(0 ) ] \vert 0\rangle_{et } = \frac{i}{24\pi}(\partial_{1})^{3}\delta(x^{1 } ) \ , .\end{aligned}\ ] ] our result agrees with the one of tomiya @xcite who works with an invariant spectral function method and in addition uses a cohomological approach .
it also coincides with the result of ebner , heid and lopes - cardoso @xcite who derive the schwinger terms directly from the gravitational anomaly . in our approach
( [ g and tpv ] ) is the basic relation .
it connects the schwinger terms , determined by @xmath201 , with the ( linearized ) gravitational anomalies given by @xmath202 in our dispersion relations procedure .
we have investigated the gravitational anomalies , specifically the pure einstein anomaly and the weyl anomaly .
so we demanded the quantized energy - momentum tensor to be symmetric no lorentz anomaly occurs which is a possible choice .
the relevant amplitude , the two - point function of the energy - momentum tensors @xmath57 , we have separated into its pure tensor part @xmath29 and into its pseudo - tensor part @xmath30 , eqs.([amplitude expressed by t - product ] ) ( [ formfactors3 ] ) , and we have decomposed the amplitudes into a general structure of tensors containing 8 formfactors @xmath31 .
these formfactors we have expressed by dispersion relations where we had to calculate only the imaginary parts via the cutkosky rules .
our subtraction procedure for the formfactors no subtraction for @xmath83 , one subtraction for @xmath203 and two subtractions for @xmath204 is _ the _ natural choice dictated by the @xmath122behaviour of the imaginary parts @xmath123 .
it implies that the pure tensor wi ( [ vwi1 ] ) ( [ vwi3 ] ) for the renormalized formfactors is satisfied ( in the limit @xmath95 ) , so that the total anomaly is automatically shifted into the pseudotensor part of the wi ( [ pta ] ) .
it turns out that the anomalous ward identity and the anomalous trace identity depend only on the finite formfactor @xmath121 , with its explicit result ( [ t1 for wi ] ) , demonstrating such the independence of a special renormalization procedure . from the anomalous ward identity and the anomalous trace identity
we could deduce the linearized gravitational anomalies
the linearized einstein- and weyl anomaly and determine their covariant versions .
the origin of the anomaly is the existence of a superconvergence sum rule for the imaginary part of the formfactor @xmath63 . in the zero mass limit the imaginary part of the formfactor approaches a @xmath0-function singularity at zero momentum squared , exhibiting in this way the infrared feature of the gravitational anomalies
this is an independent and complementary view of the anomalies as compared to the ultraviolet regularization procedures . if we compare , however , the dr approach with the n - dimensional regularization procedure of t hooft veltman we find an equivalence .
the two approaches are linked by the substitutions ( [ substitution dis rel - dim reg ] ) .
we have also calculated the gravitational schwinger terms which occur in the etc of the energy - momentum tensors .
we have adapted our dispersive approach to the method of klln . as a result
all gravitational schwinger terms are determined by the formfactor @xmath205 which in the zero mass limit approaches a @xmath0-function singularity at zero momentum squared as in the case of anomalies .
so also the schwinger terms show this peculiar infrared feature of the anomalies .
we have performed all calculations in two dimensions , where already all essential features of the dr approach show up , analogously to the chiral current case . from a practical point of view
the method appears quite appealing .
all one has to calculate is the imaginary part of an amplitude ( formfactor ) , which is an easy task .
however , this computational simplicity is a special ( and convenient ) feature of the two space - time dimensions . in higher dimensions
the amplitude will contain more formfactors and we have to calculate dispersion relations for higher loop diagrams , which is a much more delicate task , but nevertheless we expect the method to work here too .
r. jackiw , _ field theoretic investigations in current algebra , topological investigations of quantized gauge theories _ , in : _ current algebra and anomalies _ ,
treiman , r. jackiw , b. zumino and e. witten ( eds . ) , p.81 , and p.211 , world scientific , singapore ( 1985 ) .
p. van nieuwenhuizen , _ anomalies in quantum field theory : cancellation of anomalies in @xmath206 supergravity .
leuven notes in mathematical and theoretical physics _ , vol . 3 , leuven university press ( 1988 ) . | we are dealing with two - dimensional gravitational anomalies , specifically with the einstein anomaly and the weyl anomaly , and we show that they are fully determined by dispersion relations independent of any renormalization procedure ( or ultraviolet regularization ) .
the origin of the anomalies is the existence of a superconvergence sum rule for the imaginary part of the relevant formfactor . in the zero mass limit the imaginary part of the formfactor approaches a @xmath0-function singularity at zero momentum squared , exhibiting in this way the infrared feature of the gravitational anomalies .
we find an equivalence between the dispersive approach and the dimensional regularization procedure .
the schwinger terms appearing in the equal time commutators of the energy momentum tensors can be calculated by the same dispersive method .
although all computations are performed in two dimensions the method is expected to work in higher dimensions too . |
the migration of low mass protoplanets ( @xmath7 ) under the action of disk tides is long known to be a fast process in disks with power law surface density profiles @xcite .
the fast migration timescale estimates of these objects even constitutes a bottleneck for the core accretion scenario , which implies a slow build up of a solid core until it reaches the mass threshold ( @xmath8 ) above which rapid gas accretion begins . indeed ,
the solid core build up time is @xmath9 yrs @xcite , while the migration timescale of a @xmath10 planet is @xmath11 yrs @xcite and scales inversely proportionally to the planet mass .
the existence of gaseous giant planets at large distances ( @xmath12 au ) from their central star therefore constitutes a puzzle .
recent work by @xcite has shown that the core build up time scale can be lowered by taking migration into account , which prevents the depletion of the core feeding zone . however , these authors find that the most up to date type i migration timescale estimate , which includes three dimensional effects and the co - rotation torque @xcite , still needs to be lowered by a factor @xmath13 in order to allow for the solid core survival
. the total torque exerted by the disk onto the planet can be split into two parts : the differential lindblad torque , that corresponds to the torque of the spiral wake that the planet excites in the disk , and the corotation torque , exerted by the material located in the planet coorbital region .
the role of the corotation torque has often been overlooked in migration rate estimates .
the two main reasons for that is that it is harder to evaluate than the differential lindblad torque , and that it saturates ( i.e. tends to zero ) in the absence of dissipation .
the corotation torque scales with the radial gradient of @xmath14 , where @xmath15 is the disk surface density and @xmath16 is the second oort s constant , or half the disk flow vorticity vertical component in a non - rotating frame .
this scaling makes the corotation torque a quantity very sensitive to local variations of the disk surface density or rotation profile .
here we investigate the behavior of the total ( lindblad + corotation ) tidal torque exerted on a planet in the vicinity of a surface density radial jump , in order to investigate a suggestion by @xcite that localized , positive surface density jumps may be able to halt migration .
we assume that the surface density transition occurs on a length scale @xmath17 of a few pressure scale heights @xmath18 .
we consider the case in which the surface density is larger on the outside of the transition , but we do not limit ourselves to the case where the surface density on the inside is negligible compared to its value on the outer side .
the case of a virtually empty central cavity has already been contemplated by @xcite in the context , different of ours , of giant planet migration .
they conclude that giant planet migration is halted or considerably slowed down once the planet is inside the cavity and in @xmath19 resonance with the cavity edge , as beyond this resonance the disk torque onto a planet on a circular orbit becomes negligible . in section [ sec : analytic ]
we provide simple analytical estimates of the lindblad and corotation torques at a surface density transition .
we show that the corotation torque , which is a positive quantity there , is likely to overcome the lindblad torque if the transition is localized enough , i.e. @xmath20 a few @xmath18 . in section [ sec : setup ]
we describe the numerical setup that we used to check this prediction with a numerical hydro - code . in section [ sec : num ] , we present the results of our numerical simulations which indeed exhibit for a wide range of parameters a fixed point at the transition , i.e. a point where the corotation and lindblad torques cancel each other and where planetary migration stops
. we also discuss in this section the issue of the saturation of the corotation torque and the need of turbulence to prevent it , and the conditions under which turbulence is able or not to unlock a planet from the transition .
we then discuss in section [ sec : discuss ] where in protoplanetary disks such surface density transitions can be found , and what are the consequences of these planet traps on giant planet formation .
a protoplanet embedded in a gaseous protoplanetary disk excites in the latter a one - armed spiral wake @xcite , as a result of the constructive interference of propagative density waves excited at lindblad resonances with the planet .
this wake exerts a torque on the planet , which can be decomposed into the outer lindblad torque ( @xmath21 ) , which is negative and that is exerted by the outer arm , and the inner lindblad torque ( @xmath22 ) , which is positive and that is exerted by the inner arm .
these two torques do not cancel out .
the residue @xmath23 , called the differential lindblad torque , is negative @xcite , thereby leading to inward migration .
if one calls one - sided lindblad torque the arithmetic mean of the absolute values of the outer and inner lindblad torques : @xmath24 then the differential lindblad torque is a fraction of this torque which scales with the disk aspect ratio @xmath25 , where @xmath18 is the pressure scale height or disk thickness and @xmath26 the radius . in particular , for a disk with uniform surface density and aspect ratio @xcite : @xmath27 as noted by @xcite , in a nebula with @xmath28 , the differential lindblad torque is a sizable fraction of the one - sided lindblad torque .
this is of some importance for our concern : denoting @xmath29 the lindblad torque on an empty cavity edge ( it then amounts to the outer lindblad torque , i.e. @xmath30 ) and @xmath31 the lindblad torque further out in the disk where we assume the surface density and aspect ratio profiles to be flat , we have : @xmath32 which means that the lindblad torque on a cavity edge is at most @xmath33 times larger than further out in the disk , for @xmath34 .
in addition to the wake torque , the material in the orbit vicinity exerts on the planet the so - called corotation torque
. this torque scales with the gradient of the specific vorticity ( or vortensity ) @xcite : @xmath35 this torque comes from the exchange of angular momentum between the planet and fluid elements that perform u - turns at the end of their horseshoe streamlines @xcite .
it has been estimated that the co - rotation torque amounts to a few tens of percent @xcite of the differential lindblad torque for a disk with a power - law surface density profile .
@xcite give the following estimate for a three dimensional disk with flat surface density and temperature profiles : @xmath36 note that the corotation torque cancels out for a surface density profile @xmath37 , while it is a positive quantity for a shallower surface density profile .
we now estimate the ratio of the corotation torque at a cavity edge and further out in the disk ( assuming a flat surface density profile ) .
we assume that the surface density inside of the transition is @xmath38 , while it is @xmath39 on the outside .
the transition from @xmath38 to @xmath40 occurs on a length scale @xmath41 . in the outer disk
where @xmath42 , we have : @xmath43 while at the transition , @xmath44 is a sharply peaked function with a maximum that scales as : @xmath45 similarly , retaining only the terms that vary most rapidly with @xmath26 at the transition : @xmath46 where @xmath47 is the angular velocity .
the last factor of the above equation can also be expressed , using the disk rotational equilibrium and again keeping only the terms that vary most rapidly with @xmath26 , as : @xmath48 the last factor of the r.h.s .
is a function of @xmath26 that scales as @xmath49 , and that has , for a sufficiently smooth , monotonic transition from @xmath38 to @xmath40 , two minima and one maximum , the exact locations and amplitudes of which depend on the surface density profile .
the ( inverse of ) the specific vorticity logarithmic gradient is therefore : @xmath50 where @xmath51 and @xmath52 are numerical functions of @xmath26 of order unity that depend on the exact shape of the surface density profile . in particular , one sees immediately that if @xmath53 , the contributions of both terms ( the surface density gradient and the vorticity gradient ) to the corotation torque are of similar magnitude . the corotation torque at the transition @xmath54 can therefore be larger than the corotation torque @xmath55 further out in the disk by a factor : @xmath56 where @xmath57 is a numerical factor of order unity that depends on the surface density profile . using eqs .
( [ eqn : lindbladratio ] ) , ( [ eqn : crlindbladratio ] ) and ( [ eqn : crratio ] ) , we find that : @xmath58 which means that the ( positive ) corotation torque can counteract the ( negative ) lindblad torque at a cavity , provided that the edge of the latter be narrow enough ( @xmath59 ) .
we check by means of numerical simulations that it is indeed possible to halt type i migration on the edge of a surface density transition .
we performed a large number of two - dimensional hydrodynamical simulations that we detail below .
the code used was fargomasset / fargo ] .
it is a staggered mesh code on a polar grid , with upwind transport and a harmonic , second order slope limiter @xcite .
it also uses a change of rotating frame on each ring that enables one to increase significantly the time step @xcite .
the surface density transition was set in the following manner .
we call @xmath60 the target surface density ratio between the inside and the outside of the cavity : @xmath61 . in order to realize this ratio , we adopt a fixed profile of kinematic viscosity with an inverse ratio : @xmath62 .
if we call @xmath63 the mean radius of the transition , then for @xmath64 , @xmath65 , for @xmath66 , @xmath67 , and for @xmath68 , @xmath69 is an affine function of @xmath26 .
we then relax the surface density profile during a preliminary 1d calculation over @xmath70 orbits , so that the profile can safely be considered as steady state in the 2d calculations . in our calculations we use
non - reflecting inner and outer boundary conditions , while we impose the disk surface density at the inner and outer boundary . in all our calculations
the disk aspect ratio is @xmath71 , the disk is not self - gravitating and it has a locally isothermal equation of state .
the length unit is @xmath63 .
the mesh inner and outer boundaries are respectively at @xmath72 and @xmath73 .
the mass unit is the mass @xmath74 of the central object , and the gravitational constant is @xmath75 .
the time unit is @xmath76 .
whenever we quote a planet mass in earth masses , we assume the central object to have a solar mass .
the mesh size in all our runs is @xmath77 and @xmath78 , hence the radial resolution is @xmath79 .
the horseshoe zone width for a @xmath80 planet is @xmath81 , so it is @xmath82 zones wide .
the error on the corotation torque due to this finite width therefore amounts to at most @xmath83 % [ see @xcite , fig .
we also made calculations with a @xmath84 mass planet for which the horseshoe zone is only @xmath85 zones wide .
this resolution may seem low but the corotation torque is very well captured at low resolution for small mass planets which hardly perturb the surface density profile [ see discussion in @xcite , appendix a ] .
we contemplated the fashionable issue of whether the roche lobe content must be included or not when integrating the torque over the disk material @xcite .
we found that taking that material into account or not makes essentially no difference in our case .
we make the following technical comments related to this issue : * @xcite have envisaged the case of jupiter or saturn mass planets , for which there is a significant roche lobe with radius @xmath86 ( @xmath51 being the planet semi - major axis ) . for the mass range studied here ( @xmath7 )
, the roche lobe size is much smaller than this simple estimate . in particular , in the linear limit , the flow does not exhibit any circumplanetary libration , while the two ends of the horseshoe region reconnect through a single stagnation point located on the orbit ( masset et al .
, in prep . ; tanaka , private communication ) , that is offset from the planet as a result of the gaseous disk slight sub - keplerianity .
the issue of roche lobe sampling in that case is therefore irrelevant .
* @xcite have found that resolution may alter the torque exerted on the planet when the roche lobe material is taken into account .
this behavior , which appears at very high resolution , is related to the build up of a gaseous envelope around the point - like planet , while the mass of this envelope heavily depends upon the equation of state , which is poorly known .
the envelope mass may be considerable for the giant protoplanets considered by @xcite , as it may be several times the planet mass , but it is presumably a much smaller fraction of the planet mass for the planet masses that we consider [ @xmath7 ; see also @xcite ] . in 2d simulations one has to introduce a potential smoothing length @xmath87 meant to mimic the disk vertical extent .
the planet potential expression is then @xmath88 , where @xmath89 is the gravitational constant and @xmath26 the distance to the planet . in our calculations we use a potential smoothing length @xmath90 .
this is quite a large value , which leads to underestimate the corotation torque @xcite , as it lowers the horseshoe zone width , while the corotation torque is proportional to the fourth power of this width .
our calculations are therefore conservative , and it is likely that 3d calculations would yield a torque cancellation for an even wider set of disk parameters than what we found .
we consider in this section a disk with a target surface density ratio between the inside and the outside of the transition set to @xmath91 . as
this value is large compared to one , we call the inside part of the transition a cavity .
we performed two sequences of 75 calculations in which a planet was set on a fixed circular orbit with radius @xmath92 ( @xmath93 ) .
for these calculations we report as a function of @xmath94 the total torque exerted on the planet [ averaged over @xmath95 orbits , in order for the libration flow inside the co - orbital region to reach a steady state @xcite ] . the left plot of fig .
[ fig : cavitprop ] shows the relaxed surface density profile after @xmath96 orbits of the preliminary 1d calculation , and the right plot shows the quantities involved in the corotation torque scaling .
the right panel of fig .
[ fig : cavitprop ] shows that the vortensity gradient peak is only partially due to the surface density gradient , since the gradient of the flow vorticity plays a role of comparable importance , as suggested in section [ sec : analytic ] . the total torque measured from numerical simulations is presented on fig . [ fig : tqnumcavit ]
. both plots , which correspond to two different planet masses , exhibit similar features : * on the left side ( for @xmath97 ) and on the right side ( for @xmath98 ) , the torque profile is flat and negative .
this corresponds to the total negative torque acting on a planet embedded in a flat surface density disk .
since the surface density is @xmath91 times lower on the inside , the torque value is much smaller on that side . * in the range @xmath99 , the torque as a function of radius exhibits a shape very similar to the vortensity gradient shown on fig .
[ fig : cavitprop ]
. in particular the torque has a maximum value near @xmath100 and a minimum value near @xmath101 .
this last observation unambiguously indicates that the corotation torque largely dominates the total torque at the surface density transition , as expected from the analytic estimate . in particular
, there are two positions in the disk where the total torque cancels out ( and therefore where the migration is halted ) : one at @xmath102 and the other one at @xmath103 .
the first of these fixed points is unstable : moving the planet away from it yields a torque that moves it further away , while the second one ( i.e. the outermost one ) is stable : moving the planet to a higher radius yields a negative torque that tends to bring the planet back to its former position , and vice - versa for a motion towards smaller radii .
the fixed point at @xmath104 should therefore trap all type i migrating embryos .
we note in passing that the torque as a function of radius does not have exactly the same shape between a @xmath80 and a @xmath84 planet .
there are two reasons for this : the onset of non - linear effects between @xmath84 and @xmath80 , and the inaccuracy of the corotation torque estimate for the @xmath84 case , for the resolution that we have adopted .
however , we see that even in the @xmath84 case , the value of the torque maximum ( @xmath105 ) is much larger than the torque absolute value away from the cavity ( for @xmath98 , this value is @xmath106 ) . therefore , even if the error on the corotation torque amounts to @xmath0 % [ see @xcite ] , one still gets a torque cancellation and a stable fixed point .
we have performed a series of ten calculations with planets of masses @xmath107 to @xmath80 in geometric sequence , with the same disk cavity as in the previous section ( @xmath91 ) . in these calculations
the planets were released at @xmath108 , i.e. beyond the cavity edge , and then allowed to freely migrate under the action of the disk torque .
the results are presented at fig .
[ fig : freemig ] .
we see how the transition efficiently traps all these objects at the fixed point at @xmath104 identified at the previous section .
we note in passing that on the lower side of the mass range considered , the small width of the horseshoe region may lead to large uncertainties on the corotation torque .
nevertheless , the migration of all objects stops at locations which are all very close to @xmath109 .
this indicates that the corotation torque at the edge , even misrepresented by a too low resolution , still strongly dominates the differential lindblad torque . for masses lower than a few earth masses ,
the hill radius is much smaller than the disk thickness and the disk response is linear . in the linear regime ,
the corotation to lindblad torque ratio is independent of the planet mass .
as we have stressed already , our potential smoothing length is large and our corotation torque estimate is conservative .
this indicates that any small mass in - falling embryo will be trapped at the cavity edge that we modeled .
disk profiles with an extremum of the specific vorticity are prone to a linear rossby wave instability that eventually leads to the formation of vortices @xcite .
this is the case of the transitions that we have simulated .
the arc - shaped over - dense region located near @xmath110 in the right plot of fig .
[ fig : freemig ] corresponds to such a vortex .
it leads to a torque modulation over the synodic period between the planet and the vortex , that is seen on the migration curves of the left plot of fig .
[ fig : freemig ] . as a planet approaches the transition , the synodic period increases , and so does the period of the semi - major axis modulations .
we have performed a series of calculations similar to those of section [ sec : numcavit ] , except that we adopted a much smaller target outer to inner surface density ratio : @xmath111 instead of @xmath112 .
the surface density jump is therefore a very shallow one .
the results are presented on fig .
[ fig : cavitshallow ] .
they show that again there is a fixed point at which migration is halted , at @xmath113 .
we have not searched further the limit inner to outer surface density ratio that leads to migration reversal , as this quantity depends on the jump radial size , on the profile adopted for the kinematic viscosity and on the potential smoothing length .
we just stress that a factor @xmath114 and a jump size of @xmath115 suffices to halt migrating bodies . as can be seen on fig .
[ fig : tqnumcavit ] or [ fig : cavitshallow ] , the torque maximum value is positive and large .
this implies that , if the cavity radius evolves with time , either inwards or outwards , it drives any planet trapped on its edge so that its migration follows the edge evolution .
more precisely , a planet will be held at a location where the specific torque acting upon it endows it with a drift rate equal to that of the cavity radius . as the torque maximum in absolute value is at least of the order of the total torque in the outer disk , this means that the edge expansion or contraction can be as fast as the type i drift rate of its trapped objects and still keep these objects trapped .
we illustrate this shepherding on fig .
[ fig : trap ] .
we performed a calculation in which an initially fixed cavity such as the one considered at the previous sections ( here with @xmath116 and a transition over @xmath117 ) traps a @xmath80 in - falling planet and then is endowed with an expansion rate @xmath118 .
this is achieved by imposing a kinematic viscosity profile as described in section [ sec : setup ] with a radius @xmath63 that varies linearly with time .
one can see that the planet semi - major axis follows the cavity radius evolution , so that the planet remains trapped on the cavity edge .
the corotation torque , in the absence of any angular momentum exchange between the horseshoe region and the rest of the disk , is prone to saturation : its value tends to zero after a few libration periods @xcite .
the disk turbulence , which endows the disk with an effective viscosity , can help to prevent the corotation torque saturation .
here we estimate the amount of viscosity required to prevent saturation , as a function of the planet mass .
the maximal amount of saturation for which one still has a torque cancellation depends on the planet mass and on the detailed disk profiles .
here we just give an order of magnitude of the viscosity needed to prevent saturation by assuming a @xmath0 % saturation .
this amount of saturation is reached when the viscous timescale across the horseshoe region is equal to the horseshoe libration timescale @xcite : @xmath119 where @xmath120 is the horseshoe zone half width , @xmath69 is the disk effective kinematic viscosity , @xmath121 is the planet orbital frequency and @xmath51 its semi - major axis . in order to express @xmath120 as a function of the planet mass
, we note that in the linear ( hence fully unsaturated ) limit the corotation torque can be expressed as @xcite : @xmath122 while from @xcite we have : @xmath123 where @xmath124 .
( [ eqn : tq1 ] ) and ( [ eqn : tq2 ] ) yield : @xmath125 eqs .
( [ eqn : sat ] ) and ( [ eqn : xs ] ) give the minimum effective viscosity to prevent the corotation torque saturation : @xmath126 or , in terms of the @xmath127 coefficient ( @xmath128 ) : @xmath129 the value of @xmath127 required to prevent the corotation torque saturation of an earth mass planet is therefore much smaller than the one inferred from disk lifetimes of t tauri stars , namely @xmath130 @xcite .
note that this also holds for large mass solid cores ( @xmath131 ) .
the corotation torque acting on these objects should therefore be a large fraction of the unsaturated estimate , so the analysis presented above for the torque cancellation is essentially valid .
we also note that the planets of figs .
[ fig : freemig ] and [ fig : trap ] remain trapped over a timescale much longer than the horseshoe libration timescale , so that their corotation torque partial saturation is weak enough .
the disk effective viscosity at the trap location is in these calculations @xmath132 ( i.e. @xmath133 ) .
we comment that the corotation torque saturation could be different in a laminar disk with kinematic viscosity @xmath69 and in a turbulent disk ( we discuss more in detail this topic in the next section ) with the same effective viscosity , when the horseshoe region width is smaller than the turbulence length scale .
the saturation should be easier to remove in this last case , which suggests that our saturation estimates is a conservative one .
since the likely physical origin of the disk effective viscosity is turbulence , in particular the turbulence generated by the magneto - rotational instability ( mri ) @xcite , the torque acting upon the planet does not have a value constant in time as in a laminar disk , but rather displays large temporal fluctuations arising from the density perturbations in the planet vicinity .
it is important to assess the impact of these fluctuations on the mechanism outlined here . in the bulk of a disk with a smooth surface density profile
, these torque fluctuations can yield an erratic behavior of the planet semi - major axis referred to as `` stochastic migration '' @xcite . in
what follows we use the torque fluctuation estimates of @xcite and @xcite in order to determine whether despite of its semi - major axis jitter the planet is confined to the trap vicinity , or whether it is unlocked the from the trap position .
we use a very simple prescription , that we detail in appendix [ turbmc ] , to perform monte - carlo calculations of the planet semi - major axis over a period of time @xmath134 .
we place initially the planet at the outer stable fixed point , and we declare it trapped if over that amount of time it never goes through the inner , unstable fixed point , while we declare it not trapped if it happens to go through the inner fixed point . while appendix [ turbmc ] presents the technical details of our calculations , we discuss below a number of issues relevant to these calculations .
* can the torque acting upon the planet be considered as the sum of the torque in a laminar disk and the fluctuations arising from turbulence ?
@xcite finds that this might not be the case , at least as long as the differential lindblad torque is concerned , since the corotation region is not sufficiently resolved in his calculations ; as an alternative explanation to the lack of the torque convergence to the laminar value , @xcite suggests the existence of very low frequency components of the torque fluctuations . as this problem is essentially open ( and totally unaddressed as far as the corotation torque is concerned ) , we chose to describe the torque acting upon the planet by the sum of the torque in the laminar case and of the torque fluctuations due to the turbulence . * the amplitude @xmath135 of the torque fluctuations can be estimated by considering a density fluctuation of order unity with a length scale of order @xmath18 , at a distance @xmath18 from the planet @xcite and found to amount to @xmath136 , where @xmath89 is the gravity constant and @xmath26 the distance to the central object .
this is the value that we adopt for our calculations . *
the time scale of the fluctuations is of particular importance to assess the torque convergence towards its laminar value .
@xcite finds in his calculations that the torque fluctuations due to turbulence contain significant power at very low frequency , and suggests that this could be linked to global communication across the disk on the viscous timescale .
here our situation is rather different , as the planet lies on a radially localized structure , with a radial extent @xmath17 .
fluctuations with frequency lower than @xmath137 would imply turbulent structures larger than the trap width , and therefore would imply a radial displacement of the trap itself along the lines of section [ sec : trap ] . here
we choose a correlation timescale for the turbulence of @xmath138 for @xmath139 .
* both the tidal torque and the torque fluctuations scale with the disk surface density , so that the time needed for the time averaged torque to reach convergence does not depend on the latter .
however , the spread of the probability density of the planet semi - major axis after that timescale does depend on the disk surface density .
if the expectancy of the distance of the planet to its initial location is larger than the trap width , then the planet can certainly be lost , while if it is much smaller than the trap width , the planet is certainly trapped .
our study therefore amounts to a disk critical mass search . for a given trap profile , a given planet mass and given fluctuation characteristics @xmath140 and @xmath141 ,
there exists a critical disk mass so that a lighter disk keeps the planet trapped , while a heavier disk is unable to lock it and the planet is likely to go through the inner unstable fixed point .
we perform this search in a dichotomic manner until we reach a precision of @xmath5 % on the disk mass , and we repeat the calculation with different random seeds , so as to get an idea of the precision on the disk critical mass estimate .
our results are displayed on fig .
[ fig : trapturb ] .
the `` disk outer mass '' is defined as @xmath142 , where @xmath143 is the disk surface density on the outside of the surface density transition .
we see on this figure that the deep cavity with @xmath91 unconditionally traps @xmath144 to @xmath80 protoplanets in a disk with a density profile of the minimum mass solar nebula , at any distance up to @xmath145 au , while the shallow surface density transition would be able to retain in falling bodies only in the very central regions ( @xmath146 au , if the nebula extends that far in ) of such a nebula . further out in the disk , at @xmath147
au , a shallow transition in the mmsn would only capture embryos with @xmath148 , etc .
it is not unreasonable to assume that these simple calculations are conservative , for the following reasons : * we have assumed the torque fluctuations to scale only with the disk surface density , which implies that all the disk material in the trap vicinity is turbulent .
there is at least one realization of the trap , which is the inner edge of a dead zone ( detailed at the next section ) , for which this statement is wrong . in our example
, the trap stable fixed point corresponds to a location with a large surface density , therefore located rather on the `` dead '' side of the transition , i.e. where turbulent transport of angular momentum is inefficient and hence where the turbulence should be much weaker than contemplated in our simple monte carlo calculations .
* we recall that our large potential smoothing length tends to significantly underestimate the corotation torque , while the trap ability to lock infilling bodies , for a given turbulence characteristics , strongly depends on the corotation torque peak value at the surface density transition .
we conclude this section with two remarks .
firstly , the numerical simulations undertaken by @xcite and @xcite neglect , for reasons of computational speed , the gravitational potential @xmath149 dependency , which means that their disks have no vertical stratification .
future numerical experiments with vertical stratification may lead to different characteristics of the torque fluctuations . secondly
, other sources of turbulence have been suggested in protoplanetary disks , such as the global baroclinic instability @xcite , which may have different torque fluctuations statistics , and for which the trap efficiency may therefore be different . to date
this is essentially an unaddressed issue .
the mechanism exhibited here is generic , since it plays a role at any positive disk surface density transition , even shallow ones , with the cautionary remark that some turbulence is required to prevent the corotation torque saturation , and that a condition is required on the disk mass to prevent the stochastic migration induced by the turbulence to overcome the trap effect .
although our knowledge of the detailed properties of protoplanetary disks is still incipient , we already know a number of locations for transitions exhibiting the above properties , which constitute potential traps for in - falling embryos : * at the inner edge of a dead zone @xcite , the surface density is expected to be significantly larger in the dead zone than in the ionized inner disk , while the inner disk turbulence provides the dissipation required to maintain the corotation torque . if the dead zone inner radius changes with time ( since layered accretion involving a dead zone is thought not to be a steady process ) the objects trapped at the fixed point follow the transition radius , as shown in section [ sec : trap ] .
the edges of dead zones are thought , in the totally different context of runaway or type iii migration @xcite , to be able to halt or reverse a runaway episode @xcite . here ,
in our context of small mass , type i migrating objects , we argue that the inner edge of a dead zone halts type i migrating bodies and efficiently traps them . *
sizable surface density transitions can also be found at the transition between an inner disk region where jets are launched and the outer viscous disk @xcite .
indeed , the accretion velocity in the jet emitting disk ( hereafter jed ) region is entirely due to the jet torque and becomes much larger than in a standard accretion disk ( hereafter sad ) .
mass flux conservation then leads to a surface density jump at the transition radius between the sad and the jed .
this radius has been inferred to vary from 0.3 to a few astronomical units in some t tauri stars @xcite .
as all known embedded objects have jets , the existence of such a transition radius provides a natural trap for in - falling embryos over the jet lifetime .
* we finally mention the outer edge of a gap opened by a preexisting giant protoplanet .
the latter can trap planetesimals at mean motion resonances @xcite , but also at the large positive surface density gradient found on its gap outer edge .
we note that the first two possibilities might not coexist in an accretion disk .
not all t tauri stars show indeed evidence of high velocity jets @xcite . for these jet - less stars ,
the whole disk is likely to resemble a sad with a dead zone in its innermost region .
the protoplanet trap would then be realized at the inner radius of the dead zone , namely only if there is a sad settled at smaller radii .
objects with a dead zone extending down to the star would have no trap . on the contrary
, young stars producing jets require the presence of a jed in the inner disk region . in that case
, the trap is located at the outer jed radius ( the jed itself may extend down to the star with no impact on the trap ) . note also that since a jed is less dense than a sad , it is unlikely that a dead zone would establish ( ferreira et al . in prep . ) .
the likely existence of such planetary traps in protoplanetary disks has important implications for the formation of giant planets .
first , it constitutes a workaround to the timescale conflict between a solid core build up ( which is slow ) and its migration to the central object ( which would be fast , if the object was not stopped at a trap ) .
second , a large amount of solid embryos should accumulate at the trap location .
this should speed up the build up of a critical mass core ( @xmath150 ) which could accrete the nebula gas and convert itself into a giant planet .
the dynamics of the set of embryos trapped at the jump is however beyond the scope of this communication , and will be presented in a forthcoming work .
we have seen in section [ sec : analytic ] that the corotation torque value is boosted not only by the large surface density gradient but also by the alteration of the rotation profile , which yields a large gradient of @xmath16 , with both large minimal and maximal values since the gradient of @xmath16 involves the second derivative of the angular velocity .
one may wonder whether at a disk opacity transition , in which there is a disk temperature jump and hence a pressure gradient radial jump , the alteration of the rotation profile is sufficient to yield a peak value of the corotation torque that counteracts the lindblad torque .
we have undertaken subsidiary calculations in which we have a power law surface density profile and a sound speed profile with a localized radial jump .
we found that we need a jump of the aspect ratio of @xmath151 over @xmath152 in order to halt migration .
this ratio is too large and is unrealistic in a sad . indeed , in a standard accretion disk in which energy is transported vertically by radiative diffusion , the central temperature scales as @xmath153 , where @xmath154 is the optical depth .
as a consequence , even in the vicinity of an opacity transition , the temperature profile exhibits a smooth behavior and the corotation torque is not sufficiently strong to counterbalance the lindblad torque .
nevertheless , @xcite have found that the opacity transitions may induce a significant reduction of the differential lindblad torque , but they explicitly neglected the corotation torque in their study .
it would be of interest to investigate whether the corotation torque boost at these locations can be sufficient to halt migration .
remarkably , the scale height of a jed is always smaller than that of the sad , so that the transition between these two flows naturally yields a jump in the disk aspect ratio .
this is due to two effects .
first , a jed feeds its jets with most of the released accretion power so that only a small fraction of it is radiated away at the disk surface : the disk is cooler .
second , in addition to gravity , the large scale magnetic field vertically pinches the disk @xcite .
we also stress that at the inner dust puffed - up rims of protoplanetary disks @xcite , there can be a large radial gradient of the gas temperature , which may be sufficient to trap migrating planets .
we finally emphasize that the mechanism presented here does not allow low mass objects to proceed inside of a cavity , but rather traps them on the cavity edge .
this is quite different from what has been studied in the context of gap opening planets , which migrate further inside until they reach the @xmath19 resonance with the cavity edge @xcite .
we have shown that the type i migration of @xmath155 planets can be halted at radially localized disk surface density jumps ( the surface density being larger on the outside of the jump ) .
some turbulence is then required to prevent the corotation torque saturation .
the latter tends to induce stochastic migration of the embryos , and therefore tends to act against the trap .
the lighter the disk , the easier it is to keep an embryo trapped , since its radial jitter over the timescale needed to reach torque convergence is small and can be confined to the vicinity of the trap stable fixed point .
inversely , the lighter the embryo , the more difficult it is to trap it : in the zero mass limit , the embryo has no tidal torque and only feels the torque fluctuations arising from turbulence . we have shown that a jump with a large surface density ratio efficiently traps embryos above one earth mass up even in relatively massive disks , such as the mmsn at @xmath156 au , while shallow jumps ( a @xmath157 % surface density increase over @xmath158 disk scale heights ) will retain embryos of a few earth masses and more only in light disks , such as the mmsn at @xmath147 au .
we have illustrated by means of numerical simulations that a surface density jump traps in - falling embryos , and that it drives them along with it if it moves radially as the disk evolves .
such a jump can occur at different locations in a protoplanetary disk : it can be found at the transition between the standard accretion disk solution ( sad ) and the jet emitting disk ( jed ) , thought to be at a distance @xmath159 to @xmath160 au from the central object .
it can also be found at the inner edge of a dead zone , where one expects a very large ratio of outer to inner surface densities .
it may also be found at the outer edge of a protoplanetary gap if one giant planet already exists in the disk .
we have stressed that these planetary traps have two important consequences : ( i ) they can hold the type i migration of protogiant cores over the timescale needed to overcome the critical mass ( @xmath161 ) over which rapid gas accretion begins onto these cores , which are then quickly converted into giant planets that should enter the slow , type ii migration regime ; ( ii ) as solid embryos accumulate at the trap , the build up of critical cores should be faster than elsewhere in the disk .
finally , we mentioned that although temperature radial jumps in principle also yield large peaks of the corotation torque , they are unlikely to be able to counteract the lindblad torque by themselves , as the temperature gradients required for that are unrealistically large in a standard accretion disk .
the planet semi - major axis is evolved in time using an euler method : @xmath162a(t)},\ ] ] where @xmath163 is the planet semi - major axis at date @xmath164 , @xmath165 is the torque as measured in a laminar disk for a planet in circular orbit of radius @xmath51 , linearly interpolated from the results displayed either in fig .
[ fig : tqnumcavit]a or [ fig : cavitshallow]a , @xmath166 is the torque fluctuation due to turbulence , and where a keplerian profile is assumed for @xmath16 ( this is grossly wrong as far as its radial derivative is concerned , as we saw in the main text , but this is largely sufficient for our purpose here ) . the torque fluctuation is evaluated as follows : @xmath167a(t)[2v-1],\ ] ] where @xmath168=\max[\sigma(r)]_{r\in[a - h , a+h]}$ ] , and where @xmath169 is a random variable uniformly distributed over @xmath170 $ ] .
the value of @xmath166 is kept constant over @xmath141 .
we use the above prescription for @xmath168 $ ] since the planet is located in a region where the surface density gradient is large .
the largest torque fluctuations may arise from regions located within @xmath171 to @xmath172 from the central object , and where the surface density may be significantly different from the one at the planet location .
each realization of the random variable is independent of the previous ones .
we anticipate that the results should not depend on the torque fluctuation law , since the central limit theorem states that the cumulative torque law tends towards a normal law @xcite .
alibert , y. , mordasini , c. , benz , w. , & winisdoerffer , c. 2005 , , 434 , 343 dangelo , g. , bate , m. r. , & lubow , s. h. 2005 , , 358 , 316 artymowicz , p. 2004 , asp conf .
ser . 324 : debris disks and the formation of planets , 324 , 39 balbus , s. a. , & hawley , j. f. 1991 , , 376 , 214 balmforth , n. j. , korycansky , d. g. 2001 , , 326 , 833 coffey , d. , bacciotti , f. , woitas , j. , ray , t. p. , & eislffel , j. 2004 , , 604 , 758 ferreira , j. 1997 , , 319 , 340 ferreira , j. , & pelletier , g. 1995 , , 295 , 807 fromang , s. , terquem , c. , & balbus , s. a. 2002 , , 329 , 18 gammie , c. f. 1996 , , 457 , 355 goldreich , p. , & tremaine , s. 1979 , , 233 , 857 hartigan , p. , edwards , s. , & ghandour , l. 1995 , , 452 , 736 isella , a. , & natta , a. 2005 , , 438 , 899 hahn , j. m. , & ward , w. r. 1996 , lunar and planetary institute conference abstracts , 27 , 479 hartmann , l. 2001 , in accretion processes in star formation , cambridge university press klahr , h. 2004 , , 606 , 1070 korycansky , d. g. , & pollack , j. b. 1993 , icarus , 102 , 150 kuchner , m. j. , & lecar , m. 2002 , , 574 , l87 li , h. , finn , j. m. , lovelace , r. v. e. , & colgate , s. a. 2000 , , 533 , 1023 li , h. , colgate , s. a. , wendroff , b. , & liska , r. 2001 , , 551 , 874 lin , d. n. c. , & papaloizou , j. 1986 , , 307 , 395 lin , d. n. c. & papaloizou , j. c. b. 1986 , , 309 , 846 masset , f. s. 2002 , , 387 , 605 . 2001 , , 558 , 453 . 2000a , , 141 , 165
2000b , in disks , planetesimals and planets , asp conference series , vol .
219 , 7580 masset , f. s. , & papaloizou , j. c. b. 2003 , , 588 , 494 menou , k. , & goodman , j. 2004 , , 606 , 520 nelson , r. p. 2005
, , 443 , 1067 nelson , r. p. , & papaloizou , j. c. b. 2004 , , 350 , 849 ogilvie , g. i. , & lubow , s. h. 2002 , , 330 , 950 pesenti , n. , dougados , c. , cabrit , s. , ferreira , j. , casse , f. , garcia , p. , & obrien , d. 2004 , , 416 , l9 pollack , j. b. , hubickyj , o. , bodenheimer , p. , lissauer , j. j. , podolak , m. , & greenzweig , y. 1996 , icarus , 124 , 62 tanaka , h. , takeuchi , t. & ward , w. r. 2002 , , 565 , 1257 van leer , b. 1977 , journal of computational physics , 23 , 276 ward , w. r. 1986 , icarus , 67 , 164 .
1997 , icarus , 126 , 261 . 1992 , abstracts of the lunar and planetary science conference , 23 , 1491 . 1991 , abstracts of the lunar and planetary science conference , 22 , 1463 | the tidal torque exerted by a protoplanetary disk with power law surface density and temperature profiles onto an embedded protoplanetary embryo is generally a negative quantity that leads to the embryo inwards migration .
here we investigate how the tidal torque balance is affected at a disk surface density radial jump .
the jump has two consequences : * it affects the differential lindblad torque . in particular
if the disk is merely empty on the inner side , the differential lindblad torque almost amounts to the large negative outer lindblad torque . *
it affects the corotation torque , which is a quantity very sensitive to the local gradient of the disk surface density . in particular
if the disk is depleted on the inside and if the jump occurs radially over a few pressure scale - heights , the corotation torque is a positive quantity that is much larger than in a power - law disk .
we show by means of customized numerical simulations of low mass planets embedded in protoplanetary nebulae with a surface density jump that the second effect is dominant , that is that the corotation torque largely dominates the differential lindblad torque on the edge of a central depletion , even a shallow one .
namely , a disk surface density jump of about @xmath0 % over @xmath1 disk thicknesses suffices to cancel out the total torque . as a consequence the type i migration of low mass objects reaching the jump should be halted , and all these objects should be trapped there provided some amount of dissipation is present in the disk to prevent the corotation torque saturation .
as dissipation is provided by turbulence , which induces a jitter of the planet semi - major axis , we investigate under which conditions the trapping process overcomes the trend of turbulence to induce stochastic migration across the disk .
we show that a cavity with a large outer to inner surface density ratio efficiently traps embryos from @xmath2 to @xmath3 @xmath4 , at any radius up to @xmath5 au from the central object , in a disk which has same surface density profile as the minimum mass solar nebula ( mmsn ) .
shallow surface density transitions require light disks to efficiently trap embryos . in the case of the mmsn
, this could happen in the very central parts ( @xmath6 au ) .
we discuss where in a protoplanetary disk one can expect a surface density jump .
this effect could constitute a solution to the well known problem that the build up of the first protogiant solid core in a disk takes much longer than its type i migration towards the central object . |
the first - kind associated legendre functions @xmath3 , or the ferrers versions @xmath4 , are classical .
( @xmath3 and @xmath4 are continuations of each other , with respective real domains @xmath5 and @xmath6 . ) the roles they play when the degree @xmath7 and order @xmath8 equal integers @xmath9 are familiar .
the legendre , or ferrers polynomials @xmath10 , @xmath11 , are orthogonal on @xmath12 $ ] and are used in series expansions .
the spherical harmonics @xmath13 are orthogonal on the symmetric space @xmath14 , and appear in harmonic analysis based on the lie group @xmath15 .
it is less well known that ferrers functions @xmath4 of a fixed order @xmath8 , and degrees that may be non - integral but are spaced by integers , can also be used in series expansions .
the fundamental relation , due to love and hunter @xcite , is one of biorthogonality : @xmath16 which holds if ( i ) @xmath17 , and ( ii ) the degrees @xmath18 differ by a nonzero even integer and are not half - odd - integers . for suitable @xmath19 , this makes possible bilateral expansions of the form @xmath20 and in particular , the calculation of the coefficients @xmath21 as inner products in @xmath22 $ ] .
( this is the usual legendre expansion if @xmath23 , as @xmath24 for all @xmath7 . ) for conditions on @xmath25 sufficient for ( interior ) pointwise convergence , see @xcite .
the restriction to @xmath17 comes from the requirement that the expansion functions lie in @xmath22 $ ] .
if the order @xmath8 is not a positive integer , @xmath4 will have leading behavior as @xmath26 proportional to @xmath27 , but in general its leading behavior as @xmath28 comprises two terms : one proportional to @xmath29 , and one to @xmath30 .
the consequences for convergence of the integral in ( [ eq : orthgonalitylove ] ) are obvious .
these asymptotics have motivated the suggestion by pinsky @xcite that when @xmath31 , the series ( [ eq : serieslove ] ) should really be viewed as an expansion of @xmath32^{\mu/2}f(z)$ ] in the functions @xmath33^{\mu/2}\,\allowbreak{\rm
p}_{\nu_0 + 2n}^\mu(z)$ ] .
this enables a discussion of endpoint convergence , because the latter functions do not diverge as @xmath34 .
it is not usually the case that @xmath4 and @xmath35 are elementary functions , unless of course @xmath7 and @xmath8 are integers .
this may be why such expansions as ( [ eq : serieslove ] ) have been used infrequently . in this paper
, we derive explicit , trigonometrically parametrized formulas for several families of legendre functions , expressing @xmath4 , @xmath35 , and their second - kind counterparts @xmath36 , @xmath37 , as elementary functions . in each family ,
@xmath38 are non - integral but are spaced by integers : @xmath39 for some fractional @xmath40 .
the simplest example is @xmath41 where @xmath42 . here
, @xmath43 is an ` octahedral ' rational function that if @xmath44 is a polynomial of degree @xmath45 in @xmath46 ; in the base case @xmath47 , it equals unity .
it satisfies differential recurrences on @xmath48 and @xmath49 , and three - term non - differential recurrences , as well .
the function @xmath50 has a hypergeometric representation in the gauss function @xmath51 : it equals @xmath52 . but
@xmath53 , which according to ( [ eq : formulaone ] ) , appears in series of the form ( [ eq : serieslove ] ) when @xmath54 , is less classical .
it satisfies a second - order differential equation on the riemann @xmath46-sphere with four singular points , not three ; so ( if @xmath55 ) it is a _
heun polynomial _
, not a hypergeometric one . the functions @xmath56 are mutually orthogonal on the @xmath46-interval @xmath57 $ ] , in a sense that follows from ( [ eq : orthgonalitylove ] ) , but the orthogonality is of an unusual sturm liouville kind .
it is clear from ( [ eq : formulaone ] ) that for any @xmath58 , the function @xmath59 depends _ algebraically _ on @xmath60 , and can be evaluated using radicals .
each of the function families considered in this paper is similarly algebraic , and because any legendre function can be written in terms of @xmath51 , the results below are really trigonometric parametrizations of families of algebraic @xmath51 s . to see a link to prior work , recall from frobenius theory that each legendre function of degree @xmath7 and order @xmath8 satisfies a differential equation on the riemann sphere with three singular points , the characteristic exponent differences at which are @xmath61
it is a classical result of schwarz ( see @xcite , and for more recent expositions , @xcite , @xcite and @xcite ) that any such equation will have _ only algebraic solutions _ only if the ( unordered , unsigned ) triple of exponent differences falls into one of several classes .
the triples from @xmath62 , as in ( [ eq : formulaone ] ) , are @xmath63 , and they lie in schwarz s octahedral class v. the families treated below include octahedral ones , with @xmath64 , and tetrahedral ones , with @xmath65 or @xmath66 ; the schwarz classes for the latter being ii and iii .
the resulting legendre functions are octahedral or tetrahedral in the sense that their defining differential equation , on the riemann @xmath60-sphere , has as its projective monodromy group a finite subgroup of the mbius group @xmath67 , which is octahedral or tetrahedral
. this will not be developed at length , but there is a strong geometric reason why @xmath68 deserve to be called octahedral functions , or ( when @xmath69 ) polynomials . for general @xmath70 , the lifted function @xmath71 turns out to satisfy an equation on the riemann @xmath72-sphere with @xmath73 singular points .
these include @xmath74 , which are the six vertices of an octahedron inscribed in the sphere ; and also , the centers of its eight faces . up to normalization , the doubly indexed functions @xmath75 are identical to specializations of triply - indexed ones introduced by ochiai and yoshida in their groundbreaking work on algebraic @xmath51 s @xcite . for schwarz classes such as the octahedral and tetrahedral , they considered the effects of displacing the triple of exponent differences , not by @xmath76 as in the legendre case , but by general elements of @xmath77 .
it is a key result of the present paper that in the legendre case , when the triple has only two degrees of freedom , it is far easier to derive and solve recurrences on exponent displacements .
schwarz s classification of algebraic @xmath51 s also includes a dihedral class ( class i ) and a related ` cyclic ' class ( unnumbered but called class o here ) .
legendre functions lie in class i when the order @xmath8 is a half - odd - integer , and in class o when the degree @xmath7 is an integer .
we obtain explicit formulas for the legendre ( and ferrers ) functions in the respective families , of the first and second kinds .
the simplest dihedral example is @xmath78 where @xmath79 and @xmath80 is arbitrary . here
, @xmath81 is the jacobi polynomial of degree @xmath82 , and @xmath83 signifies the even part under @xmath84 . when @xmath85 , this becomes a trigonometric version of a well - known algebraic formula @xcite ) ; and when @xmath86 , it expresses @xmath87 in terms of the @xmath49th chebyshev polynomial of the third kind .
but the general jacobi representation ( [ eq : jacobirep ] ) is new .
there is a significant literature on ` dihedral polynomials ' appearing in dihedrally symmetric @xmath51
s @xcite , and vidnas has shown they can be expressed as terminating appell series @xcite . focusing on the legendre case , when two of the three exponent differences are equal , leads to such simpler formulas as ( [ eq : jacobirep ] ) , for both the dihedral and cyclic families . constructing bilateral ferrers series of the form ( [ eq : serieslove ] ) is facilitated by the explicit formulas derived below for the legendre and ferrers functions in the several families .
but the functions @xmath88 , @xmath89 , and corresponding spherical harmonics @xmath90 , do not fit into conventional @xmath15-based harmonic analysis unless @xmath91 , when the latter are the usual surface harmonics on @xmath14 . in the octahedral and tetrahedral families ( and also the dihedral and cyclic , if @xmath92 resp .
@xmath93 is rational ) , it is nonetheless the case that each spherical harmonic can be viewed as a _
finite - valued _
function on @xmath94 .
this is due both to @xmath40 being rational , and to the algebraicity of @xmath95 in its argument @xmath60 , as seen in ( [ eq : formulaone ] ) . to begin to relate the present results to harmonic analysis
, we interpret in lie - theoretic terms the recurrences satisfied by any family @xmath96 or @xmath97 , @xmath98 , which are based on first - order differential operators that perform ladder operations .
it is well known that for any @xmath99 , there is an infinite - dimensional representation of the lie algebra @xmath100 on @xmath101 .
( in the case when @xmath102 , this includes the familiar @xmath103-dimensional representation on @xmath104 .
) there is also a representation of @xmath105 on @xmath106 .
the real algebras @xmath107 are real forms of the complex lie algebra @xmath108 .
as we explain , the ` order ' and ` degree ' algebras generate a 10-dimensional , rank-2 complex lie algebra isomorphic to @xmath1 , which for any @xmath99 , acts differentially on the family @xmath109 , @xmath98 , of _ generalized solid harmonics _ on @xmath110 .
the root system of @xmath1 , of type @xmath111 , comprises the eight displacement vectors @xmath112 , which yield four ladders on @xmath113 ; and for each ladder , there are differential operators for raising and lowering , a differential recurrence satisfied by @xmath114 , and a three - term non - differential recurrence .
the ones coming from the roots @xmath115 , such as the ` diagonal ' recurrences @xmath116\,{\rm p}_\nu^\mu(z ) \\
+ \bigl[(\nu+\tfrac12)\pm(\mu-\tfrac12)\bigr]\ , \bigl[(\nu+\tfrac12)\pm(\mu-\tfrac32)\bigr]\ , \sqrt{1-z^2}\:\,{\rm p}_{\nu\mp1}^{\mu-1}(z ) = 0,\end{gathered}\ ] ] may be given here for the first time .
( here , @xmath117 may be replaced by @xmath118 . ) connections between associated legendre / ferrers functions , or spherical harmonics , and the complex lie algebra @xmath1 [ or its real forms @xmath119 and
@xmath120 are known to exist .
( see @xcite and ( * ? ? ?
* chaps .
3,4 ) , and @xcite in the physics literature ; also @xcite for hyperspherical extensions . )
but , most work has focused on functions of integral degree and order . the octahedral , tetrahedral , dihedral , and cyclic
families yield explicit infinite - dimensional representations of @xmath1 and its real forms , which are carried by finite - valued solid harmonics on @xmath110 . when @xmath121 or @xmath122 , the representation of @xmath2 turns out to be equivalent to a known skew - hermitian one , of the dirac singleton type ( the ` di ' or the ` rac ' one , respectively ) .
but in general , these representations are new , non - skew - hermitian ones .
it is shown below that irrespective of @xmath99 , any of these representation of @xmath1 [ or any of its real forms ] , carried by a harmonic family @xmath109 , @xmath98 , is of a distinguished kind , in the sense that it has the same two values for the two casimir invariants of the algebra ; cf .
@xcite .
this paper is structured as follows . in [ sec : prelims ] , facts on legendre / ferrers functions that will be needed are reviewed . in [ sec : octa ] , the key results on the octahedral functions @xmath123 are stated , and explicit formulas for octahedral legendre / ferrers functions are derived .
these are extended to the tetrahedral families in [ sec : tetra1 ] and [ sec : tetra2 ] . in [
sec : proofs ] , the results in [ sec : octa ] are proved . in [ sec : biorthog ] , love
hunter biorthogonality is related to sturm
liouville biorthogonality . in
[ sec : cyclicdihedral ] , formulas for legendre / ferrers functions in the cyclic and dihedral families are derived , and love
hunter expansions in dihedral ferrers functions are briefly explored . in [ sec : last ] , recurrences on the degree and order , valid for any @xmath99 , are derived , and are given a lie - theoretic interpretation : @xmath1 and its real forms are introduced , and their representations carried by solid harmonics are examined .
the ( associated ) legendre equation is the second - order differential equation @xmath124 + \left[\nu({\nu}+1 ) -
\frac{{\mu^2}}{1-z^2}\right]{u}=0 \label{eq : legendre}\ ] ] on the complex @xmath60-plane .
for there to be single - valued solutions , the plane is cut along the real axis either from @xmath125 to @xmath126 ( the legendre choice ) , or from @xmath125 to @xmath127 and from @xmath126 to @xmath128 ( the ferrers choice ) .
the respective solution spaces have @xmath129 and @xmath130 as bases , except in degenerate cases described below . at fixed real @xmath8 , eq .
( [ eq : legendre ] ) can be viewed as a singular sturm
liouville equation on the real ferrers domain @xmath12 $ ] , the endpoints of which are of weyl s ` limit circle ' type if @xmath131 .
( see @xcite . ) in this case , all solutions @xmath132 lie in @xmath22 $ ] , irrespective of @xmath7 ; but the same is not true when @xmath133 , which is why such orthogonality relations as ( [ eq : orthgonalitylove ] ) can only be obtained if @xmath131 , or more generally if @xmath17 .
further light on endpoint behavior is shed by frobenius theory .
equation ( [ eq : legendre ] ) has regular singular points at @xmath134 and @xmath135 , with respective characteristic exponents expressed in terms of the degree @xmath7 and order @xmath8 as @xmath136 ; @xmath136 ; and @xmath137 .
the exponent differences are @xmath61 .
the functions @xmath138 are frobenius solutions associated to the exponent @xmath139 at @xmath140 , and the second legendre function @xmath141 is associated to the exponent @xmath142 at @xmath143 .
( the second ferrers function @xmath144 is a combination of two frobenius solutions . )
these functions are defined to be analytic ( or rather meromorphic ) in @xmath38 @xcite , the legendre functions having the normalizations @xmath145 by convention @xcite .
the notation @xmath146 will be used henceforth ; it removes an awkward @xmath147 factor . the formulas ( [ eq : pasympt]),([eq : qasympt ] ) apply if the gammas are finite ; the asymptotics when they are not are given in @xcite .
one such degenerate case is when @xmath148 .
then , @xmath149 .
a familiar example is when the degree is a non - negative integer @xmath150 .
then , @xmath151 if the order is an integer @xmath152 ; though not if @xmath153 .
another degenerate case is when @xmath154 is a negative integer .
if so , @xmath155 is undefined , as ( [ eq : qasympt ] ) suggests ; except when @xmath156 .
then , @xmath157 are defined .
the ferrers functions are related to the legendre ones on their common domains , which are the upper and lower half - planes @xmath158 , by @xmath159 thus @xmath138 are related by analytic continuation , up to phase .
going from @xmath160 to @xmath161 typically involves replacing a factor @xmath162 by @xmath163 ; for instance , @xmath164 are @xmath165 , @xmath166 .
also , owing to ( [ eq : lincomb2 ] ) , @xmath144 is undefined iff @xmath167 is
. equation ( [ eq : legendre ] ) is invariant under @xmath168 , @xmath169 , and @xmath170 , so that in nondegenerate cases , the legendre and ferrers functions with @xmath7 replaced by @xmath171 , @xmath8 by @xmath172 , and/or @xmath60 by @xmath173 , can be expressed as combinations of any two of @xmath174 , at most
. some connection formulas of this type , which are needed below , are @xmath175 , @xmath176 , @xmath177 the @xmath178 reduction @xmath179 the @xmath180 reduction @xmath181 and the @xmath182 reduction @xmath183 ( see @xcite . )
it follows from ( [ eq : rmptormq ] ) that if @xmath184 , then @xmath185 .
the functions @xmath186 are known to have the hypergeometric representations [ eq : reps ] @xmath187 ( for @xmath161 , replace @xmath188 in the prefactor on the right of ( [ eq : prep ] ) by @xmath189 ; the alternative expression ( [ eq : altprep ] ) for @xmath190 comes from ( [ eq : prep ] ) by a quadratic hypergeometric transformation . ) here , @xmath191 is the gauss function with parameters @xmath192 , defined ( on the disk @xmath193 , at least ) by the maclaurin series @xmath194\,x^k$ ] . in ( [ eq : reps ] ) and below
, the notation @xmath195 is used for the rising factorial , i.e. , @xmath196^{-1},&k =- k'\le0 .
\end{cases}\end{aligned}\ ] ] ( the second half of this definition , which extends the usual @xmath195 to negative @xmath197 so that @xmath198^{-1}$ ] for all @xmath199 , will be needed . ) in each @xmath191 in ( [ eq : reps ] ) , it is assumed that the lower parameter @xmath200 is not a non - positive integer .
the gauss equation satisfied by @xmath191 has the three singular points @xmath201 , with respective exponent differences @xmath202 .
taking into account either of ( [ eq : prep]),([eq : qrep ] ) , one sees that this triple is consistent with the exponent differences @xmath61 at the singular points @xmath203 of the legendre equation ( [ eq : legendre ] ) .
schwarz s results on algebraicity were originally phrased in terms of the gauss equation , its solutions such as @xmath51 , and the ( unordered , unsigned ) triple @xmath204 ; but they extend to the legendre equation , its solutions , and the triple @xmath61 .
this section and sections 4 and 5 derive parametric formulas for legendre and ferrers functions that are either octahedral or tetrahedral ( two types ) .
the formulas involve the octahedral polynomials , or functions , @xmath68 .
section [ subsec : polys ] defines these rational functions and states several results , the proofs of which are deferred to [ sec : proofs ] .
[ def : rdef ] for @xmath58 , the rational functions @xmath43 and their ` conjugates ' @xmath205 are defined implicitly by @xmath206^{-1/12-{m}-{n}}\,r^{m}_{n}(s),\nonumber\\ & { } _ 2f_1\left ( { { \tfrac5{24}+\frac{m}2-\frac{n}2,\,\tfrac{17}{24}+\frac{m}2-\frac{n}2}\atop{\tfrac54+{m } } } \biggm| r(s)\right ) = \nonumber\\ & \qquad\qquad\qquad\qquad\qquad\qquad\qquad \bigl[(1+s)(1 - 34s+s^2)\bigr]^{5/12+{m}-{n}}(1-s)^{-1 - 4{m}}\,\overline r^{m}_{n}(s),\nonumber \end{aligned}\ ] ] which hold on a neighborhood of @xmath207 . here , @xmath208 ^ 2 } = 1 - \frac{(1 + 14s+s^2)^3}{\bigl[(1+s)(1 - 34s+s^2)\bigr]^2},\\ & { : = }
-108\,\frac{p_{\rm v}(s)}{p_{\rm e}(s)^2 } = 1 - \,\frac{p_{\rm f}(s)^3}{p_{\rm e}(s)^2 } , \end{aligned}\ ] ] where @xmath209 are the polynomials @xmath210 , @xmath211 , @xmath212 .
equivalently , @xmath213 where @xmath214 .
it is clear from the definition that @xmath215 are analytic at @xmath207 , at which they equal unity ; though it is not obvious that they are rational in @xmath46 .
but it is easily checked that the gauss equations satisfied by the two @xmath216 functions have respective exponent differences ( at the singular points @xmath201 ) equal to @xmath217 and @xmath218 .
these triples lie in schwarz s octahedral class iv , so each @xmath216 must be an algebraic function of @xmath219 .
the definition implicitly asserts that if these algebraic @xmath51 s are _ parametrized _ by the degree-@xmath220 rational function @xmath221 , the resulting dependences on @xmath46 will be captured by certain _ rational _
@xmath43 , @xmath222 . in the terminology of @xcite , these are octahedral functions of @xmath46 .
[ thm:1 ] ( i ) for @xmath44 , @xmath75 is a polynomial of degree @xmath223 in @xmath46 , to be be called the octahedral polynomial indexed by @xmath224 .
its coefficient of @xmath225 is unity , and its coefficient of @xmath226 is @xmath227 ( ii ) for unrestricted @xmath89 , @xmath75 is a rational function that equals unity at @xmath207 and is asymptotic to @xmath228 as @xmath229 .
the conjugate function @xmath230 is related to @xmath231 by @xmath232 so that if @xmath44 , @xmath230 is a reversed version of the polynomial @xmath231 , scaled to equal unity at @xmath207 .
the functional form of the octahedral functions that are not polynomials , which are indexed by @xmath233 , is not complicated .
[ thm : nonpoly ] for any @xmath44 , @xmath234 where on each line , @xmath235 signifies a polynomial of degree @xmath197 in @xmath46 , with the coefficient of @xmath225 equal to unity , and a coefficient of @xmath236 that comes from the preceding theorem . on their indices
@xmath9 , the @xmath231 satisfy both differential recurrences and three - term non - differential recurrences .
the former are given in [ sec : proofs ] ( see theorem [ thm : diffrec ] ) , and the latter are as follows .
[ thm : recs ] the octahedral functions @xmath43 , indexed by @xmath89 , satisfy second - order ( i.e. , three - term ) recurrences on @xmath49 and @xmath48 , namely @xmath237 where @xmath209 are the polynomials in @xmath46 , satisfying @xmath238 , that appeared in definition [ def : rdef ] . moreover , they satisfy @xmath239r^{m}_{n } \\ & \qquad\qquad\qquad\qquad{}+ 9(12{m}+12{n}+1)(12{m}+12{n}-11)\,p_{\rm v}(s)\,p_{\rm f}^3(s)\ , r^{{m}-1}_{{n}-1 } = 0,\\ & 3(4{m}-3)(4{m}+1)\,p_{\rm f}^3(s)\,r^{{m}+1}_{{n}-1 } - ( 4{m}-3)\left[(12{m}-12{n}-1)p_{\rm f}^3(s ) + 4(3{n}+1)p_{\rm e}^2(s)\right]r^{m}_{n } \\ & \qquad\qquad\qquad\qquad{}+ 9(12{m}-12{n}-7)(12{m}-12{n}-19)\,p_{\rm v}(s)\ , r^{{m}-1}_{{n}+1 } = 0 , \end{aligned}\ ] ] which are second - order ` diagonal ' recurrences . from the first two recurrences in this theorem
, one can compute @xmath231 for any @xmath58 , if one begins with @xmath240 , which are low - degree polynomials in @xmath46 computable ` by hand . ' in fact , @xmath241 by specializing to @xmath242 ( at which @xmath209 equal @xmath243 ) , one can prove by induction that @xmath244 if @xmath245 .
examples of octahedral functions that are not polynomials because they have at least one negative index , illustrating theorem [ thm : nonpoly ] , are @xmath246 these also follow from the recurrences of theorem [ thm : recs ] .
the recurrences are non - classical , not least because they are bilateral : they extend to @xmath247 .
it is shown in [ sec : proofs ] that for @xmath44 , the degree-@xmath248 polynomials @xmath231 in @xmath46 are ( generalized ) heun polynomials , rather than hypergeometric ones ; they are not orthogonal polynomials in the conventional sense . a useful third - order ( i.e. , four - term ) recurrence on @xmath197 for the coefficients @xmath249 of @xmath231
is given in theorem [ thm : genheunrep ] .
an important degenerate case is worth noting : the case @xmath250 . for any @xmath251 , there are hypergeometric representations in @xmath51 for the degree-@xmath252 octahedral polynomials @xmath253 and @xmath254 , namely @xmath255 these follow by a sextic hypergeometric transformation of the @xmath51 s in definition 1 , as well as by the methods of [ sec : proofs ] .
the first can also be deduced from the @xmath250 case of the recurrence on @xmath49 in theorem [ thm : recs ] .
these representations extend to @xmath256 .
the following two theorems ( theorems [ thm : oct1 ] and [ thm : oct2 ] ) give trigonometrically parametrized formulas for the legendre / ferrers functions @xmath257 when @xmath258 and @xmath259 , with @xmath42 .
the respective triples of exponent differences , @xmath260 , are @xmath63 and @xmath261 , which lie in schwarz s octahedral class v. an interesting application of these formulas to the evaluation of certain mehler
dirichlet integrals appears in theorem [ thm : mehler ] .
let hyperbolic - trigonometric functions @xmath262 , positive on @xmath263 , be defined by @xmath264 so that @xmath265 .
this choice is motivated by definition [ def : rdef ] : if @xmath266 and @xmath267 are alternatively parametrized as @xmath268 and @xmath269 , respectively , it is not difficult to verify that the three polynomials in @xmath46 that appear in definition [ def : rdef ] will have the parametrizations @xmath270 moreover , and more fundamentally , @xmath271 .
also in this section , let @xmath272 signify @xmath273 ; so that when @xmath44 , @xmath272 is a scaled version of the octahedral polynomial @xmath231 , with its leading rather than its trailing coefficient equal to unity . equivalently , @xmath274 .
[ thm : oct1 ] the formulas @xmath275\\ & \qquad\qquad\qquad\qquad\qquad{}\times(\sinh\xi)^{-1/4-{m } } a_+^{1/4 + 3{m}+3{n}}\ , r^{m}_{n}(-a_-/a_+),\\ & { p}_{-\frac16+{n}}^{-\frac14-{m}}(\cosh\xi)=\left[(-)^{n}2^{-2{m}-3{n}}\,3^{3/4 + 3{m}}\,\gamma(\tfrac54+{m})^{-1}\right]\\ & \qquad\qquad\qquad\qquad\qquad{}\times(\sinh\xi)^{-1/4-{m } } a_-^{1/4 + 3{m}+3{n}}\ , \hat r^{m}_{n}(-a_+/a_-)\end{aligned}\ ] ] hold for @xmath89 and @xmath276 .
these formulas follow from the hypergeometric representation ( [ eq : altprep ] ) of @xmath277 , together with the implicit definitions of @xmath215 ( see definition [ def : rdef ] ) . if @xmath278 , the argument @xmath279 of the right - hand @xmath51 in ( [ eq : altprep ] ) will equal @xmath268 .
this is why it is natural to parametrize definition [ def : rdef ] by letting @xmath266 equal @xmath268 , with the just - described consequences . in deriving the formulas
, one needs the representation ( [ eq : parampolyedgemidpt ] ) ; and for the second formula , the definition ( [ eq : ddef ] ) of @xmath280 . in the following ,
the circular - trigonometric functions @xmath281 , positive on @xmath282 , are defined by @xmath283 so that @xmath284 .
[ thm : oct2 ] the formulas @xmath285\\ & \qquad\qquad\qquad\qquad\qquad{}\times(\sin\theta)^{-1/4-{m } } b_+^{1/4 + 3{m}+3{n}}\ , r^{m}_{n}(b_-/b_+),\\ & { \rm p}_{-\frac16+{n}}^{-\frac14-{m}}(\cos\theta)=\left[2^{-2{m}-3{n}}\,3^{3/4 + 3{m}}\,\gamma(\tfrac54+{m})^{-1}\right]\\ & \qquad\qquad\qquad\qquad\qquad{}\times(\sin\theta)^{-1/4-{m } } b_-^{1/4 + 3{m}+3{n}}\ , \hat r^{m}_{n}(b_+/b_-)\end{aligned}\ ] ] hold for @xmath89 and @xmath286 . by analytic continuation of theorem [ thm : oct1 ] ; or in effect , by letting @xmath287 .
because @xmath288 and @xmath289 , theorems [ thm : oct1 ] and [ thm : oct2 ] also supply formulas for @xmath290 and @xmath291 . by exploiting the @xmath292 and @xmath293 reductions ( [ eq : ptoq ] ) and ( [ eq : rmptormq ] ) , one easily obtains additional formulas , for @xmath294 and @xmath295 .
theorems [ thm : oct1 ] and [ thm : oct2 ] permit certain mehler
dirichlet integrals to be evaluated in closed form .
for example , consider @xcite and 3.7(27 ) ) @xmath296}{(\cosh\xi - \cosh t)^{\mu+\frac12}}\,{\rm d}t , \label{eq : md1 } \\ { \rm p}_\nu^\mu(\cos\theta ) & = \sqrt{\frac2{\pi}}\ : \frac{(\sin\theta)^\mu}{\gamma(\frac12-\mu)}\ , \int_0^\theta \frac{\cos[(\nu+\frac12)\phi]}{(\cos \phi - \cos \theta)^{\mu+\frac12}}\,{\rm d}\phi , \label{eq : md2}\end{aligned}\ ] ] which hold when @xmath297 for @xmath276 and @xmath286 .
these integral representations of the legendre and ferrers functions of the first kind are classical @xcite .
[ thm : mehler ] the formulas @xmath298}{{(\cosh\xi-\cosh t)}^{\frac14-m}}\,{\rm d}t & = ( -1)^n k_n^m \ , a_-^{1/4 + 3 m + 3n } \,\hat r_n^m(-a_+/a_-),\\ \int_0^\theta \frac{\cos[(\frac13+n)\phi]}{{(\cos \phi-\cos \theta)}^{\frac14-m}}\,{\rm d}\phi & = k_n^m \ , b_-^{1/4 + 3 m + 3n } \,\hat r_n^m(b_+/b_- ) , \end{aligned}\ ] ] with @xmath299 hold when @xmath82 is a non - negative integer and @xmath150 an integer , for @xmath276 and @xmath286 .
substitute the second formulas of theorems [ thm : oct1 ] and [ thm : oct2 ] into the @xmath300 specializations of ( [ eq : md1 ] ) and ( [ eq : md2 ] ) .
the following theorem gives trigonometrically parametrized formulas for the second legendre function @xmath155 when @xmath301 or @xmath302 , with @xmath89 .
the respective triples of exponent differences , @xmath260 , are @xmath303 and @xmath304 , which lie in schwarz s tetrahedral class ii .
the hyperbolic - trigonometric functions @xmath305 on @xmath263 are defined as in ( [ eq : adef ] ) .
[ thm : tetr1 ] the formulas @xmath306\\ & \qquad\qquad\qquad{}\times ( \sinh\xi)^{1/4-{m } } \left [ -(-)^{n } \sqrt{\sqrt3 + 1}\:a_+^{1/4 + 3{m}+3{n}}\,r^{m}_{n}(-a_-/a_+ ) \right],\\ & ( 2/\pi)\hat q_{-\frac14+{m}}^{-\frac13-{n}}(\coth\xi ) = \left [ 2^{11/4 - 2{m}-3{n}}\,3^{-3/8}\left(\tfrac14\right)_{m}\left(\tfrac{13}{12}\right)_{{m}+{n}}^{-1}\gamma(\tfrac43)^{-1 } \right]\\ & \qquad\qquad\qquad{}\times ( \sinh\xi)^{1/4-{m } } \left [ + ( -)^{m } \sqrt{\sqrt3 - 1}\:a_-^{1/4 + 3{m}+3{n}}\,r^{m}_{n}(-a_+/a_- ) \right]\end{aligned}\ ] ] hold for @xmath89 and @xmath276 . combine whipple s @xmath292 transformation @xcite ) , @xmath307 with the results in theorem [ thm : oct1 ] . the symmetrical forms of the right - hand prefactors are obtained with the aid of the gamma - function identities @xmath308 of vidnas @xcite . because @xmath309 are proportional to each other ( see ( [ eq : qtoq ] ) ) , theorem [ thm : tetr1 ] also supplies formulas for @xmath310 .
moreover , it leads to the following two theorems .
[ thm : tetr2 ] the formulas @xmath311 \\ & \qquad\qquad\qquad{}\times ( \sinh\xi)^{1/4-{m}}\,\bigl [ ( -)^{n } \sqrt{\sqrt3 - 1}\:a_+^{1/4 + 3{m}+3{n}}\,r^{m}_{n}(-a_-/a_+ ) \\ & \qquad\qquad\qquad\qquad\qquad\qquad\qquad{}- ( -)^{m } \sqrt{\sqrt3 + 1}\:a_-^{1/4 + 3{m}+3{n}}\,r^{m}_{n}(-a_+/a_- ) \bigr],\\ & p_{-\frac34-{m}}^{\frac13+{n}}(\coth\xi ) = \left [ ( -)^{n } 2^{-1/4 - 2{m}-3{n}}\,3^{-3/8}\left(\tfrac14\right)_{m}\left(\tfrac{5}{12}\right)_{{m}-{n}}^{-1}\gamma(\tfrac23)^{-1 } \right ] \\ & \qquad\qquad\qquad{}\times ( \sinh\xi)^{1/4-{m}}\,\bigl [ ( -)^{n } \sqrt{\sqrt3 + 1}\:a_+^{1/4 + 3{m}+3{n}}\,r^{m}_{n}(-a_-/a_+ ) \\ & \qquad\qquad\qquad\qquad\qquad\qquad\qquad{}+ ( -)^{m } \sqrt{\sqrt3 - 1}\:a_-^{1/4 + 3{m}+3{n}}\,r^{m}_{n}(-a_+/a_- ) \bigr]\end{aligned}\ ] ] hold for @xmath89 and @xmath276 .
combine the @xmath178 reduction ( [ eq : qtop ] ) with the results in theorem [ thm : tetr1 ] . because @xmath288 , theorem [ thm : tetr2 ] also supplies formulas for @xmath312 .
in theorem [ thm : tetr3 ] , the hyperbolic - trigonometric functions @xmath313 , positive on @xmath314 , are defined by @xmath315 so that @xmath316 .
[ thm : tetr3 ] the formulas @xmath317 \\ & \qquad\qquad\qquad{}\times ( \cosh\xi)^{1/4-{m}}\,\bigl [ -(-)^{n } \sqrt{\sqrt3 - 1}\:c_+^{1/4 + 3{m}+3{n}}\,r^{m}_{n}(-c_-/c_+ ) \\ & \qquad\qquad\qquad\qquad\qquad\qquad\qquad{}+ ( -)^{m } \sqrt{\sqrt3 + 1}\:c_-^{1/4 + 3{m}+3{n}}\,r^{m}_{n}(-c_+/c_- ) \bigr],\\ & { \rm p}_{-\frac34-{m}}^{\frac13+{n}}(\tanh\xi ) = \left [ ( -)^{n } 2^{-1/4 - 2{m}-3{n}}\,3^{-3/8}\left(\tfrac14\right)_{m}\left(\tfrac{5}{12}\right)_{{m}-{n}}^{-1}\gamma(\tfrac23)^{-1}\right]\\ & \qquad\qquad\qquad{}\times ( \cosh\xi)^{1/4-{m}}\,\bigl [ ( -)^{n } \sqrt{\sqrt3 + 1}\:c_+^{1/4 + 3{m}+3{n}}\,r^{m}_{n}(-c_-/c_+ ) \\ & \qquad\qquad\qquad\qquad\qquad\qquad\qquad{}+ ( -)^{m } \sqrt{\sqrt3 - 1}\:c_-^{1/4 + 3{m}+3{n}}\,r^{m}_{n}(-c_+/c_- ) \bigr]\end{aligned}\ ] ] hold for @xmath89 and @xmath318 . by analytic continuation of the results in theorem [ thm : tetr2 ] ; or in effect , by replacing @xmath319 by @xmath320 . because @xmath321 , theorem [ thm : tetr3 ] also supplies formulas for @xmath322 . by exploiting the @xmath293 reduction ( [ eq : rmptormq ] ) ,
one easily obtains additional formulas , for @xmath323 .
the following theorems give parametrized formulas for the legendre / ferrers functions @xmath324 when @xmath325 and @xmath326 , with @xmath327 .
the respective triples of exponent differences , @xmath260 , are @xmath328 and @xmath329 , which lie in schwarz s tetrahedral class iii .
[ thm : tetr2_1 ] the formulas @xmath330 where expressions for the right - hand legendre functions are provided by theorem [ thm : tetr2 ] , hold for @xmath331 and @xmath276 .
[ thm : tetr2_2 ] the formulas @xmath332 where expressions for the right - hand ferrers functions are provided by theorem [ thm : tetr3 ] , hold for @xmath331 and @xmath318 .
[ thm : tetr2_3 ] the formulas @xmath333 where expressions for the right - hand ferrers functions are provided by theorem [ thm : tetr3 ] , hold for @xmath331 and @xmath318 . because @xmath334 and @xmath289 , theorems [ thm : tetr2_1 ] and [ thm : tetr2_2 ] also supply formulas for @xmath335 . and
because @xmath309 are proportional to each other ( see ( [ eq : qtoq ] ) ) , theorem [ thm : tetr2_3 ] also supplies formulas for @xmath336 . by exploiting the @xmath293 reduction ( [ eq : rmptormq ] ) ,
one easily obtains additional formulas , for @xmath337 .
the formulas in theorems [ thm : tetr2_1 ] , [ thm : tetr2_2 ] , and [ thm : tetr2_3 ] are straightforward reparametrizations of the @xmath338 and @xmath339 cases of the identities @xmath340 which hold when @xmath276 and @xmath286 .
these appear as identities @xmath341 , @xmath342 , and @xmath343 in @xcite , and are really quadratic hypergeometric transformations in disguise .
the octahedral and tetrahedral formulas in [ subsec : formulas ] , [ sec : tetra1 ] , and [ sec : tetra2 ] followed from the theorems in [ subsec : polys ] on the octahedral functions @xmath344 , which were stated without proof .
the present section provides proofs , in some cases sketched , and obtains a few additional results .
these are theorem [ thm : diffrec ] ( on the differential equation and differential recurrences satisfied by @xmath231 ) , and theorems [ thm : hypergrep ] , [ thm : heunrep ] , and [ thm : genheunrep ] ( on the interpretation of @xmath231 when @xmath44 as a hypergeometric , heun , or generalized heun polynomial ) .
this section also reveals the origin of the degree-@xmath220 rational function @xmath221 in definition [ def : rdef ] .
consider a riemann sphere @xmath345 , parametrized by @xmath72 and identified by stereographic projection with the complex @xmath72-plane .
( as usual , @xmath346 is at the bottom and @xmath347 is at the top ; points with @xmath348 are taken to lie on the equator . )
let a regular octahedron ( a platonic solid ) be inscribed in the sphere , with its six vertices @xmath349 at @xmath74 , i.e. , at the five roots of @xmath350 and at @xmath347 . by some trigonometry @xcite , one can show that the twelve edge - midpoints @xmath351 of the octahedron , radially projected onto the sphere , are located at @xmath352 and @xmath353 , which are the roots of @xmath354 .
similar , its eight face - centers @xmath355 , when radially projected , are located at @xmath356 , which are the roots of @xmath357 . the polynomials @xmath358 are ( relative ) invariants of the symmetry group of the octahedron , which is an order-@xmath359 subgroup of the group of rotations of the riemann @xmath72-sphere . the well - known _ octahedral equation _ states that @xmath360 .
the validity of this identity ( a syzygy , in the language of invariant theory ) suggests considering the degree-24 rational function @xmath361 equal to @xmath362 , i.e. , @xmath363 ^ 4}{\bigl[(1+u^4)(1 - 34u^4+u^8)\bigr]^2 } = 1 - \frac{(1 + 14u^4+u^8)^3}{\bigl[(1+u^4)(1 - 34u^4+u^8)\bigr]^2}.\ ] ] on the @xmath72-sphere , @xmath364 equals @xmath365 at ( respectively ) the vertices , the face - centers , and the edge - midpoints .
it is an _ absolute invariant _ of the symmetry group of the octahedron .
( its derivative @xmath366 can be written as @xmath367 but is only a relative invariant . ) the covering @xmath368 given by @xmath369 is ramified above @xmath201 , and its ramification structure can be written as @xmath370 : each of the six points above @xmath346 ( i.e. , the vertices ) appears with multiplicity @xmath371 , etc .
following and extending schwarz @xcite , consider the effect of lifting the gauss hypergeometric equation satisfied by @xmath372 , the @xmath216 appearing in definition [ def : rdef ] , from the @xmath219-sphere to the @xmath72-sphere , along @xmath373 .
it should be recalled that the gauss equation satisfied by @xmath374 is the fuchsian differential equation ( @xmath375 ) @xmath376d_xf + \left [ \frac{ab}{x(x-1 ) } \right]f=0,\ ] ] which has characteristic exponents @xmath377 , @xmath378 , @xmath379 at its singular points @xmath201 , with ( unsigned ) exponent differences @xmath380 ; @xmath381 ; @xmath382 . ( the function @xmath191 is the frobenius solution associated to the zero exponent at @xmath383 . )
the effects of the ramified lifting by @xmath369 are conveniently expressed in the classical notation of riemann p - symbols , which display the exponents at each singular point @xcite . for the @xmath51 of definition [ def : rdef ] , one can write @xmath384 because any pair of characteristic exponents at a point @xmath385 beneath a ramification point @xmath386 of order @xmath197 is multiplied by @xmath197 when lifted . this function of @xmath72 satisfies a differential equation on the @xmath72-sphere with the indicated singular points and exponents .
if @xmath387 is any fuchsian differential equation on the @xmath72-sphere , the modified equation @xmath388 obtained by the change of dependent variable @xmath389 has its exponents at @xmath386 shifted upward by @xmath80 , and those at @xmath347 shifted downward by the same . as an application of this , one deduces from ( [ eq : preprep ] ) that [ eq : upsym ] @xmath390^{\frac1{12}+{m}+{n}}\,{}_2f_1\left(x=\tilde r(u)\right)\\ & = p \left\ { \begin{array}{cccccc|c } v_1,\dots , v_5 & & f_1,\dots , f_8 & & e_1,\dots , e_{12 } & \infty & \,u \\
\hline 0 & & 0 & & 0 & -1 - 12{m}-12{n } & \\ 1 + 4{m } & & 1 + 3{n } & & 1 & -8{m}-12{n } & \end{array } \right\}\\ & = p \left\ { \begin{array}{cccc|c } v_1,\dots , v_5 & & f_1,\dots , f_8 & \infty & \,u \\
\hline 0 & & 0 & -1 - 12{m}-12{n } & \\ 1 + 4{m } & & 1 + 3{n } & -8{m}-12{n } & \end{array } \right\ } , \label{eq : upsym3}\end{aligned}\ ] ] because @xmath391 are the roots of @xmath392 , and @xmath393 .
the left - hand function @xmath394 , which by examination is @xmath395 , will be the solution of a ` lifted and shifted ' differential equation on the @xmath72-sphere , with the indicated exponents .
after the shifting , the edge - midpoints @xmath351 cease being singular points , because the new exponents at each are @xmath396 , which are those of an ordinary point .
it is straightforward if tedious to compute the differential equation satisfied by @xmath397 explicitly , by applying to the appropriate gauss equation of the form ( [ eq : gausshe ] ) the changes of variable that perform ( i ) the lifting along @xmath398 , and ( ii ) the multiplication by @xmath399^{\frac1{12}+{m}+{n}}$ ] .
one finds that @xmath400 satisfies @xmath401 , where @xmath402\,d_u\\ & \qquad\qquad{}+ 4(12{m}+12{n}+1)\ , \frac{u^3\left[(2n+3m)(u^8 + 14u^4 + 1 ) - 12m(3u^4 + 1)\right ] } { u(u^4 - 1)(u^8 + 14u^4 + 1)}.\nonumber\end{aligned}\ ] ] that the singular points of this operator are the roots @xmath403 of @xmath404 ( plus @xmath393 ) , and the roots @xmath405 of @xmath406 , is clear ; as is the fact that their exponents are as shown in the p - symbol ( [ eq : upsym3 ] ) .
the degenerate case @xmath407 is especially interesting .
as one expects from the p - symbol , the operator @xmath408 is simply @xmath409 , the kernel of which is spanned by @xmath410 . for @xmath411 ,
it is easy to rule out any admixture of the latter solution by examining definition [ def : rdef ] ; and because @xmath412 equals unity at @xmath207 , the base octahedral function @xmath413 must be identically equal to unity . because @xmath414 , it follows from definition [ def : rdef ] that the hypergeometric function appearing in the definition of @xmath231 when @xmath407 , which is @xmath415 , must be algebraic in its argument @xmath219
this is essentially the 1877 result of schwarz @xcite , the proof of which was later restated in a p - symbol form by poole @xcite .
however , it is not trivial to extend this result on @xmath413 to a constructive proof that @xmath43 is a rational function of @xmath46 for any @xmath89 .
this is because the differential equation @xmath416 , as one sees from ( [ eq : ude ] ) , is far more complicated than @xmath417 ( laplace s equation ) when @xmath418 . a constructive proof is best based on _ contiguity relations _ between adjacent @xmath419 , i.e. , recurrences in the spirit of gauss , derived as follows .
first , simplify the lifting along the covering @xmath420 , i.e. , along the degree-24 map @xmath369 .
each octahedral function @xmath421 turns out to ` factor through ' @xmath422 , so it suffices to lift the gauss hypergeometric equation from the @xmath219-sphere @xmath423 to the @xmath46-sphere @xmath424 , along the degree-@xmath220 map @xmath221 of definition [ def : rdef ] , i.e. , @xmath425 where @xmath426 replacing the lifted variable @xmath72 by @xmath422 quotients out an order-4 cyclic group of rotations of the @xmath72-sphere ( and hence of the octahedron ) , about the axis passing through its north and south poles .
the syzygy becomes @xmath427 .
the roots @xmath428 of @xmath429 , and @xmath430 , correspond to the south - pole vertex , the four equatorial ones , and the north - pole one .
the three roots @xmath431 of @xmath432 correspond to the four edge - midpoints in the northern hemisphere , the four on the equator , and the four to the south .
the two roots @xmath433 of @xmath434 correspond to the four face - centers in the north , and the four in the south .
the covering @xmath435 is still ramified above @xmath201 , but its ramification structure is @xmath436 .
taking the multiplicities in this ramification structure into account , one finds that if the @xmath51 of definition [ def : rdef ] is lifted along @xmath221 rather than @xmath369 , the p - symbol identity ( [ eq : upsym ] ) is replaced by @xmath437^{\frac1{12}+{m}+{n}}\,{}_2f_1\left(x = r(s)\right)\\ & = p \left\ { \begin{array}{cccccc|c } 0 & & & 1 & -(2\pm\sqrt3)^2 & \infty & s \\ \hline 0 & & & 0 & 0 & -\tfrac14 - 3{m}-3{n } & \\
\frac14 + { m } & & & 1 + 4{m } & 1 + 3{n } & -2{m}-3{n } & \end{array } \right\}. \label{eq : displayedp}\end{aligned}\ ] ] this p - symbol has five singular points ( at most ; fewer if @xmath438 or @xmath439 ) . by the preceding explanation ,
the points @xmath440 represent @xmath441 vertices of the octahedron , and each of @xmath442 represents a cycle of four face - centers .
[ thm : diffrec ] the octahedral function @xmath43 satisfies the fuchsian differential equation @xmath443 , where @xmath444\,d_s\\ & \qquad\qquad\quad{}+(1 + 12{m}+12{n})\ : \frac{(2{m}+3{n})(s^2 + 14s+1 ) - 12{n}(3s+1)}{4\,s(s-1)(s^2 + 14s+1 ) } , \end{aligned}\ ] ] the p - symbol of which appears in eq .
( [ eq : displayedp ] ) .
the function @xmath231 is the frobenius solution associated to the zero characteristic exponent of the singular point @xmath207 .
it satisfies eight differential recurrences of the form @xmath445,\ ] ] in which @xmath446 , @xmath447 and @xmath115 . for each recurrence , the exponents @xmath448 , the exponents @xmath449 , and the prefactor @xmath450 are listed in table [ tab:1 ] .
the differential equation comes by applying to the appropriate gauss equation of the form ( [ eq : gausshe ] ) the changes of variable that perform ( i ) the lifting along @xmath451 , and ( ii ) the multiplication by @xmath452^{\frac1{12 } + { m } + { n}}$ ] . or
, one can merely substitute @xmath422 into the equation @xmath453 , with @xmath454 as in ( [ eq : ude ] ) .
the differential recurrences of jacobi , which shift the parameters @xmath455 of the function @xmath191 by integers , are well known .
( see @xcite2.8(27 ) ) . ) and if @xmath456 , the @xmath457 in the definition of @xmath75 has its parameters shifted by integers .
( see definition [ def : rdef ] . ) by some calculus , one can change the independent variable in the relevant differential recurrences of jacobi from @xmath221 to @xmath46 , thereby obtaining the final four recurrences in table [ tab:1 ] ( the diagonal ones ) .
the change uses the fact that @xmath458 equals @xmath459 , and the details are straightforward .
the first four recurrences in the table , with @xmath460 , come a bit less easily , because they shift the @xmath51 parameters in definition [ def : rdef ] by half - integers rather than integers .
but by examination , definition [ def : rdef ] is equivalent to @xmath461 where @xmath462 this follows by a quadratic hypergeometric transformation , @xmath46 being related to @xmath463 by @xmath464 , and @xmath465 to @xmath466 by @xmath467 . when @xmath468 or @xmath447 , the parameters of the @xmath51 in ( [ eq : new2f1 ] ) are shifted by integers , and
the same technique can be applied .
llll @xmath469 & @xmath470 & @xmath471 & @xmath450 + @xmath472 & @xmath473 & @xmath474 & @xmath475 + @xmath476 & @xmath477 & @xmath478 & @xmath479 + @xmath480 & @xmath481 & @xmath474 & @xmath482 + @xmath483 & @xmath484 & @xmath485 & @xmath486 + @xmath487 & @xmath488 & @xmath489 & @xmath475 + @xmath490 & @xmath491 & @xmath492 & @xmath493 + @xmath494 & @xmath495 & @xmath496 & @xmath475 + @xmath497 & @xmath498 & @xmath499 & @xmath500 + the four three - term non - differential recurrences in theorem [ thm : recs ] follow by a familiar elimination procedure from the differential recurrences of theorem [ thm : diffrec ] , taken in pairs .
they are analogous to the contiguity relations ( or ` contiguous function relations ' ) of gauss , for @xmath51 , which follow by elimination from the differential recurrences of jacobi , though gauss did not derive them in this way . the explicit formulas for the functions @xmath231 with small @xmath9 given in [ subsec : polys ] ( see eqs .
( [ eq : byhand1]),([eq : byhand2 ] ) ) also follow from the differential recurrences of theorem [ thm : diffrec ] .
[ thm : hypergrep ] for any @xmath256 , the octahedral function @xmath253 has the hypergeometric representation @xmath501 thus when @xmath251 , @xmath253 is a degree-@xmath252 hypergeometric polynomial.moreover , for any @xmath44 , @xmath231 is a polynomial of degree @xmath223 . when @xmath250 , @xmath502 loses two singular points and degenerates to a gauss hypergeometric equation of the form ( [ eq : gausshe ] ) , with independent variable @xmath46 and parameters @xmath503 , @xmath504 , @xmath505 .
hence @xmath50 has the claimed representation , and if @xmath251 , is a degree-@xmath252 polynomial in @xmath46 .
it follows from the differential recurrence with @xmath506 that @xmath75 is a polynomial in @xmath46 for all @xmath55 .
it must be of degree @xmath223 , because in the p - symbol of @xmath507 [ see ( [ eq : displayedp ] ) ] , the only characteristic exponent at @xmath430 that is a ( nonpositive ) integer is @xmath508 .
the statement of this theorem includes additional claims that were made in [ subsec : polys ] .
the following related theorem mentions the heun function @xmath509 , for which see @xcite .
this is a frobenius solution ( at @xmath510 ) of a fuchsian differential equation that has four singular points , namely @xmath511 , and an ` accessory ' parameter @xmath512 that unlike @xmath513 , does not affect their characteristic exponents .
it has a convergent expansion @xmath514 , where the @xmath515 satisfy a second - order recurrence with coefficients quadratic in @xmath197 .
[ thm : heunrep ] for any @xmath327 , the octahedral function @xmath516 has the heun representation @xmath517 and the equivalent expansion @xmath518 , where @xmath519 satisfy the second - order recurrence @xmath520\,a_k \label{eq : heunlike}\\ & \qquad\qquad\qquad\qquad\qquad { } + ( k-3{n}-1)(4k-12{n}-5)\,a_{k-1 } = 0 , \nonumber\end{aligned}\ ] ] with @xmath521 , @xmath522 .
thus when @xmath55 , @xmath516 is a degree-@xmath523 heun polynomial .
if @xmath85 , the @xmath242 singular point of @xmath507 drops out , i.e. , becomes ordinary , and @xmath524 \,d_s + \frac{3{n}(1 + 12{n})}4 \ , \frac{s+3}{s(s^2 + 14s+1)}.\ ] ] the substitution @xmath525 reduces @xmath526 to the standard heun equation @xcite , with the stated values of @xmath527 .
the recurrence ( [ eq : heunlike ] ) , based on a second - order difference operator , comes by substituting @xmath528 into @xmath526 . for general @xmath89 , @xmath502 has five singular points .
the theory of such generalized heun equations is underdeveloped at present , but the coefficients of their series solutions are known to satisfy third - order ( i.e. , four - term ) recurrences .
[ thm : genheunrep ] for any @xmath89 , the octahedral function @xmath231 has the expansion @xmath518 , where @xmath519 satisfy the third - order recurrence @xmath529\,a_k\nonumber\\ & \qquad { } -\left[(k-1)(52k-276{m}-144{n}-65)+2(14{m}+3{n})(12{m}+12{n}+1)\right]\,a_{k-1 } \nonumber\\ & \qquad { } -(k-2{m}-3{n}-2)(4k-12{m}-12{n}-9)\,a_{k-2}=0,\nonumber \end{aligned}\ ] ] with @xmath521 , @xmath522 , @xmath530 .
thus when @xmath44 , @xmath75 is a degree-@xmath531 generalized heun polynomial .
the recurrence comes by substituting @xmath532 into @xmath533 .
it can be shown that if @xmath85 , the third - order ( i.e. generalized heun ) difference operator in ( [ eq : gheunrec ] ) has the second - order ( i.e. heun ) difference operator in ( [ eq : heunlike ] ) as a right factor ; and if @xmath250 , it has a first - order ( i.e. hypergeometric ) difference operator as a right factor , which is why the representation in theorem [ thm : hypergrep ] exists .
the coefficients of all these difference operators are quadratic in @xmath197 .
as stated in theorem [ thm : nonpoly ] , it is not merely the case that when @xmath44 , the rational function @xmath43 is a polynomial of degree @xmath534 . in each quadrant of the @xmath535-plane ,
it is the quotient of a polynomial of known degree ( the numerator ) by a known polynomial ( the denominator ) . to obtain the formulas in theorem [ thm : nonpoly ] that refer to quadrants other than the first , reason as follows .
consider the second formula : it says that if @xmath69 , @xmath536 equals a polynomial of degree @xmath537 , divided by @xmath538 .
this is proved by induction on @xmath48 , the base case ( @xmath250 ) being @xmath539 which comes from ( [ eq : hyprep ] ) by euler s transformation of @xmath51 .
the inductive step uses the differential recurrence with @xmath540 , as in the proof of theorem [ thm : hypergrep ] . in the same way
, the third and fourth formulas follow from the @xmath541 recurrence .
one sees from the four formulas in theorem [ thm : nonpoly ] that irrespective of quadrant , @xmath542 , which partially confirms the claims of theorem [ thm:1](i , ii ) .
a consequence of this asymptotic behavior is that besides being the frobenius solution associated to the exponent @xmath543 at @xmath207 , @xmath231 is the frobenius solution associated to the exponent @xmath544 at @xmath430 , which appeared in the p - symbol ( [ eq : displayedp ] ) .
theorem [ thm:1 ] states specifically that @xmath545 , with @xmath280 defined in ( [ eq : ddef ] ) .
this too is proved by induction .
the base case ( @xmath250 ) has sub - cases @xmath251 and @xmath546 , which follow by elementary manipulations from ( [ eq : hyprep ] ) and ( [ eq : nneg ] ) , respectively .
the inductions toward @xmath55 and @xmath547 come from the differential recurrences with @xmath548 , the @xmath229 asymptotics of which yield expressions for @xmath549 .
as one can check , these two expression agree with what ( [ eq : ddef ] ) predicts .
the only claim in [ subsec : polys ] remaining to be proved is theorem [ thm:1](iii ) : the statement that the conjugate function @xmath205 is related to @xmath43 by @xmath550 , or equivalently @xmath551 .
( the constant of proportionality comes from @xmath552 equalling unity . ) just as one derives the differential equation @xmath553 of theorem [ thm : diffrec ] by ` lifting and shifting , ' one can derive an equation @xmath554 on the @xmath46-sphere from the definition of @xmath230 , given in definition [ def : rdef ] ; and a further differential equation satisfied by @xmath555 .
the latter turns out to be @xmath556 , i.e. , to be identical to the equation of theorem [ thm : diffrec ] . but
@xmath557 , analytic at @xmath207 , is the frobenius solution associated to the exponent @xmath543 at @xmath207 of @xmath558 .
hence , @xmath559 is the frobenius solution associated to the exponent @xmath544 at @xmath430 of @xmath556 .
but as was noted three paragraphs ago , this is @xmath231 ; so @xmath560 , and theorem [ thm:1](iii ) follows .
the octahedral functions @xmath43 , which are polynomials if @xmath44 , satisfy recurrences , such as the three - term ones of theorem [ thm : recs ] , that are quite unlike the ones satisfied by the classical orthogonal polynomials .
but at least if @xmath561 , it can be shown that the family @xmath562 displays orthogonality on the @xmath46-interval @xmath57 $ ] , or rather a form of biorthogonality .
the biorthogonality is best expressed in terms of the lifted functions @xmath563 of the last section , the full domain of which is the riemann @xmath72-sphere in which the defining octahedron is inscribed .
these functions are solutions of @xmath401 , where the operator @xmath454 was defined in ( [ eq : ude ] ) . by inspection , it has the simpler representation @xmath564\frac{q_{\rm v}^2}{q_{\rm f}^2 } \right\ } , \label{eq : slop}\ ] ] where @xmath565 and @xmath566 are the usual polynomials that equal zero at the five finite vertices and the eight face - centers of the octahedron . by ( [ eq : slop ] ) , @xmath454 is conjugated by a similarity transformation to a formally self - adjoint operator of the schrdinger type . for any fixed @xmath49 , the calculation of the eigenfunctions @xmath567 of the latter , on the @xmath72-interval @xmath57
$ ] , can be viewed as solving a sturm liouville problem .
the coefficient @xmath568 in ( [ eq : slop ] ) plays the role of the sturm
liouville eigenvalue , and @xmath569 that of the sturm liouville weight function . because the coefficient function @xmath570 diverges at the endpoints @xmath571 , this sturm
liouville problem is typically a singular one . to avoid a discussion of endpoint classifications and boundary conditions , it is best to derive orthogonality results not from @xmath454 , but rather from the love hunter biorthogonality relation ( [ eq : orthgonalitylove ] ) , i.e. , @xmath572 which holds if @xmath131 and @xmath18 differ by a nonzero even integer .
( see ( * ? ? ? * appendix ) for a proof . )
equation ( [ eq : orthgonalitylove2 ] ) is a relation of orthogonality between the eigenfunctions of a singular boundary value problem based on ( [ eq : legendre ] ) , the associated legendre equation ( i.e. , @xmath573 , @xmath331 ) , and the eigenfunctions of the _ adjoint _ boundary value problem ( i.e. , @xmath574 , @xmath331 ) .
the first problem is non - self - adjoint because the boundary conditions that single out @xmath573 , @xmath331 , as eigenfunctions are not self - adjoint .
however , one feature of the operator @xmath454 must be mentioned .
if @xmath394 solves @xmath401 , then so does @xmath575 .
this claim can be verified by a lengthy computation , but its correctness is indicated by the p - symbol of @xmath454 , which appeared in ( [ eq : upsym3 ] ) .
the map @xmath576 is a @xmath577 rotation of the @xmath72-sphere , and hence of the inscribed octahedron , around the axis through its equatorial vertices @xmath578 .
this rotation takes vertices to vertices , edges to edges , and faces to faces .
the multiplication by @xmath579 shifts the characteristic exponents at the affected vertices to the values they had before the rotation .
[ thm : biorthog ] for @xmath85 and @xmath580 , the lifted family @xmath581 is biorthogonal on the @xmath72-interval @xmath57 $ ] in the following sense : the inner product integral @xmath582 \\ & \qquad{}\times\left [ q_{\rm v}^{-2{m } } q_{\rm f}^{-3{n}'/2}(u)\cdot ( 1-u)^{1 + 12{m}+12{n}}\ : \tilde r^{m}_{{n}'}\!\left(\frac{1+u}{1-u}\right ) \right ] \,\frac{q_{\rm v}^2}{q_{\rm f}^2}(u ) \,{\rm d}u\end{aligned}\ ] ] equals zero if @xmath583 differ by a nonzero even integer .
substitute the @xmath561 cases of the formulas for @xmath584 in theorem [ thm : oct2 ] into ( [ eq : orthgonalitylove2 ] ) , and change the variable of integration from @xmath585 to @xmath586 , and then to @xmath587 .
the involution @xmath170 corresponds to @xmath588 .
this biorthogonality theorem is formulated so as to indicate its close connection to the above - mentioned sturm liouville problem : evaluating the integral over @xmath589 computes the inner product of the two square - bracketed factors in the integrand , which are an eigenfunction and an adjoint eigenfunction that come from @xmath590 and @xmath591 , with respect to the weight function @xmath592 .
the two factors are solutions of the same differential equation on @xmath593 , but satisfy different boundary conditions . the theorem can not be extended to general @xmath594 , because the integral diverges unless @xmath85 or @xmath580 , owing to rapid growth of one or the other of the bracketed factors at each of the endpoints @xmath571 .
this divergence follows readily from the results on @xmath231 given in theorems [ thm:1 ] and [ thm : nonpoly ] .
alternatively , the divergence arises from @xmath161 not lying in @xmath22 $ ] when @xmath8 is non - integral , unless @xmath595 . the formulas for the _ tetrahedral _ ferrers functions @xmath596 given in theorem [ thm : tetr3 ] ( cases @xmath561 )
can also be substituted usefully into the love
hunter relation ( [ eq : orthgonalitylove2 ] ) . but the resulting statement of biorthogonality is more complicated than theorem [ thm : biorthog ] and is not given here .
this section derives parametric formulas for legendre and ferrers functions that are cyclic or dihedral .
the formulas involve the jacobi polynomials @xmath597 , and are unrelated to the octahedral and tetrahedral ones in [ sec : octa ] , [ sec : tetra1 ] , and [ sec : tetra2 ] .
they are of independent interest , and subsume formulas that have previously appeared in the literature .
as used here , ` cyclic ' and ` dihedral ' have extended meanings .
the terms arise as follows .
the associated legendre equation ( [ eq : legendre ] ) has @xmath598 as its ( unordered , unsigned ) triple of characteristic exponent differences . by the results of schwarz on the algebraicity of hypergeometric functions ,
this differential equation will have _ only algebraic solutions _ if @xmath599 lies in @xmath600 or @xmath601 , for some positive integer @xmath197 .
these restrictions cause the equation to lie in schwarz s cyclic class ( labelled o here ) , resp .
his dihedral class i. the terms refer to the projective monodromy group of the equation , which is a ( finite ) subgroup of @xmath67 .
however , the formulas derived below are more general , in that they allow @xmath197 to be arbitrary : they are formulas for _ continuously _ parametrized families of legendre and ferrers functions , which are generically transcendental rather than algebraic . because of this , we call a legendre or ferrers function cyclic , resp . dihedral , if @xmath599 lies in @xmath602 , resp . @xmath603 ; the asterisk denoting an unspecified value
. that is , the degree @xmath7 should be an integer or the order @xmath8 a half - odd - integer , respectively .
explicit formulas in terms of jacobi polynomials are derived in [ subsec : explicitcyclicdihedral ] , and how dihedral ferrers functions can be used for expansion purposes is explained in [ sec : biorthog2 ] .
the jacobi polynomials @xmath604 are well known @xcite .
they have the hypergeometric and rodrigues representations @xmath605\end{aligned}\ ] ] and are orthogonal on @xmath12 $ ] with respect to the weight function @xmath606 , if @xmath607 and the weight function is integrable .
legendre and ferrers functions that are cyclic ( i.e. , of integer degree ) are easily expressed in terms of jacobi polynomials .
[ thm : cyclic1 ] the formulas @xmath608^{\mu}\ , p_n^{(-\mu,\mu)}(\cosh\xi ) \end{aligned}\ ] ] hold when @xmath150 is a non - negative integer , for @xmath5 and @xmath276 .
( in the degenerate case when @xmath609 is a positive integer , @xmath610 . ) compare the representations ( [ eq : prep ] ) and ( [ eq : jacobirep2 ] ) .
[ thm : cyclic2 ] the formulas @xmath611^{\mu}\ , p_n^{(-\mu,\mu)}(\cos\theta ) \end{aligned}\ ] ] hold when @xmath150 is a non - negative integer , for @xmath6 , @xmath318 , and @xmath286 .
( in the degenerate case when @xmath609 is a positive integer , @xmath612 . ) by analytic continuation of theorem [ thm : cyclic1 ] ; or in effect , by letting @xmath287 . by exploiting the @xmath292 and @xmath293 reductions (
[ eq : ptoq ] ) and ( [ eq : rmptormq ] ) , one easily derives additional formulas from theorems [ thm : cyclic1 ] and [ thm : cyclic2 ] , for @xmath613 and @xmath614 .
legendre and ferrers functions that are dihedral ( i.e. , are of half - odd - integer order ) are the subject of the following theorems . for conciseness , a special notation is used : @xmath615_\pm$ ] signifies @xmath616 , resp .
@xmath617 , in the @xmath618 , resp .
@xmath619 case ; and @xmath620 , where @xmath621 depends on @xmath80 , signifies the even or odd part of @xmath621 under @xmath84 , i.e. , @xmath622 $ ] .
[ thm : dihedral1 ] the formulas @xmath623_\pm \\ & \qquad\qquad\qquad\qquad{}\times ( z^2 - 1)^{-1/4}\,\bigl(z+\sqrt{z^2 - 1}\bigr)^{-\alpha}\,p_m^{(\alpha,-\alpha)}\!\left(\frac{z}{\sqrt{z^2 - 1}}\right),\\ & \hat q_{-\frac12+\alpha}^{\pm(\frac12+m)}(\cosh\xi ) =
\sqrt{\frac{\pi}2}\ : m!\ , \left[1 \bigm | \frac1{(\alpha - m)_{2m+1}}\right]_\pm \\ & \qquad\qquad\qquad\qquad{}\times ( \sinh\xi)^{-1/2}\ , { \rm e}^{-\alpha\xi}\ : p_m^{(\alpha,-\alpha)}(\coth\xi ) \end{aligned}\ ] ] hold when @xmath82 is a non - negative integer , for @xmath5 and @xmath276 . combine whipple s @xmath292 transformation @xcite ) , which appeared as eq .
( [ eq : whipple ] ) , with the results in theorem [ thm : cyclic1 ] ; and write @xmath82 for @xmath150 , and @xmath624 for @xmath8 . in these formulas ,
the proportionality of @xmath625 to each other is expected ; cf .
( [ eq : qtoq ] ) .
also , the division in the ` minus ' case by @xmath626 which equals zero if @xmath627 , is not unexpected . as was noted in [ sec : prelims ] , @xmath155 is undefined if @xmath154 is a negative integer , except when @xmath156 , in which case @xmath628 are defined .
this implies that for @xmath79 , @xmath629 is defined for all @xmath80 , and that @xmath630 is undefined if and only if @xmath631 . [ thm : dihedral2 ] the formulas @xmath632_\pm \\ & \qquad\qquad\qquad\qquad{}\times ( z^2 - 1)^{-1/4}\,\left\{\bigl(z+\sqrt{z^2 - 1}\bigr)^{-\alpha}\,p_m^{(\alpha,-\alpha)}\!\left(\frac{z}{\sqrt{z^2 - 1}}\right)\right\}_{\alpha,\pm},\\ & p_{-\frac12+\alpha}^{\pm(\frac12+m)}(\cosh\xi ) = \sqrt{\frac2{\pi}}\ : m!\ , \left[(-1)^m \bigm | \frac{(-1)^{m+1}}{(\alpha - m)_{2m+1}}\right]_\pm \\ & \qquad\qquad\qquad\qquad{}\times ( \sinh\xi)^{-1/2}\,\left\{{\rm e}^{-\alpha\xi}\:p_m^{(\alpha,-\alpha)}(\coth\xi)\right\}_{\alpha,\pm } \end{aligned}\ ] ] hold when @xmath82 is a non - negative integer , for @xmath5 and @xmath276 ; it being understood in the ` minus ' case that when @xmath631 and there is an apparent division by zero , each right - hand side requires the taking of a limit .
combine the @xmath178 reduction ( [ eq : qtop ] ) with the results in theorem [ thm : dihedral1 ] .
[ thm : dihedralpq ] the formulas @xmath633_\pm \\ & \qquad\qquad{}\times ( \sin\theta)^{-1/2}\,\left\{{\rm e}^{{\rm i}\alpha\theta}\:p_m^{(\alpha,-\alpha)}({\rm i}\,\cot\theta)\right\}_{\alpha,\pm},\\ { \rm q}_{-\frac12+\alpha}^{\pm(\frac12+m)}(\cos\theta ) & = \sqrt{\frac{\pi}2}\ : m!\ , \left[{\rm i}^{m+1 } \bigm | \frac{{\rm i}^{-m}}{(\alpha - m)_{2m+1}}\right]_\pm \\ & \qquad\qquad{}\times ( \sin\theta)^{-1/2}\,\left\{{\rm e}^{{\rm i}\alpha\theta}\:p_m^{(\alpha,-\alpha)}({\rm i}\,\cot\theta)\right\}_{\alpha,\mp}\end{aligned}\ ] ] hold when @xmath82 is a non - negative integer , for @xmath286 . in the sub - cases @xmath631 of the ` minus ' case , the apparent division by zero in the first formula is handled by interpreting its right - hand side in a limiting sense ; but the division by zero in the second formula causes both its sides to be undefined .
the first formula follows by analytic continuation of the last formula in theorem [ thm : dihedral2 ] ; in effect , by letting @xmath634 .
the second formula then follows from the @xmath635 reduction ( [ eq : rmptormq ] ) , after some algebraic manipulations . as was noted in
[ sec : prelims ] , @xmath144 is undefined if and only if @xmath155 is .
it was also noted that if @xmath184 , then @xmath636 .
it follows that in the ` plus ' case of the second formula of the theorem , the right - hand side must equal zero if @xmath631 .
this yields the interesting jacobi - polynomial identity @xmath637 which holds for @xmath79 , when @xmath638 . in this subsection , we show that an expansion in dihedral ferrers functions can be , in effect , an expansion in chebyshev polynomials ( of the fourth kind ) ; and as an application , show that the result of @xcite on the convergence of love
hunter expansions can be slightly extended .
the first formula on dihedral ferrers functions in theorem [ thm : dihedralpq ] specializes when @xmath439 to the known pair of formulas @xcite @xmath639 these hold for @xmath286 , the @xmath640 case of the former requiring the taking of a limit .
hunter biorthogonality , i.e. , the orthogonality of the functions @xmath4 and @xmath591 in @xmath22 $ ] when ( i ) @xmath641 and ( ii ) @xmath18 differ by an even integer and are not half - odd - integers , specializes when @xmath642 and @xmath585 to @xmath643 and thus to @xmath644\,{\rm d}\theta = 0,\ ] ] which holds if @xmath645 differ by an even integer .
( by continuity , the restriction to @xmath645 that are not integers can be dropped . )
the orthogonality in ( [ eq : notwellknown ] ) is not well known .
a love hunter expansion of an ` arbitrary ' function @xmath646 on @xmath647 is a bilateral expansion of @xmath25 in the ferrers functions @xmath648 , of the form ( [ eq : serieslove ] ) , in which the coefficients @xmath649 are computed as inner products , i.e. , @xmath650 existing results on the convergence of such expansions @xcite require that @xmath651 , or in the real case , @xmath652 .
it is of interest to examine whether convergence results can also be obtained in the boundary cases @xmath653 .
to treat the case when @xmath654 , define the indexed ( @xmath331 ) and continuously parametrized ( @xmath655 ) functions @xmath656}{\sqrt{\sin\theta}},\qquad \chi_{n}^{(\alpha)}(\theta ) = \frac{\cos[(2n+\alpha)(\pi-\theta)]}{\sqrt{\sin\theta}}\qquad\ ] ] on @xmath657 , which are biorthogonal with respect to the weight function @xmath658 .
( they differ only in normalization from @xmath659 and @xmath660 . ) in terms of the first , one has a formal @xmath642 love
hunter expansion @xmath661 where @xmath662 the denominator inner product equals @xmath663 for all @xmath150 , by examination ; hence the restriction @xmath664 must obviously be imposed .
this expansion is not fully satisfactory , because each @xmath665 diverges as @xmath666 ; though it converges to zero , asymmetrically , as @xmath667 .
the underlying problem is that if @xmath668 , the function @xmath4 has leading behavior as @xmath26 proportional to @xmath669 , but its leading behavior as @xmath28 comprises two terms : one proportional to @xmath29 , and one to @xmath30 . in order ( i ) to make endpoint behavior more symmetrical and less divergent , and ( ii ) to study endpoint convergence , pinsky @xcite has proposed modifying love
hunter expansions by treating @xmath32^{\mu/2}\,\allowbreak{\rm p}_{\nu}^\mu(z)$ ] rather than @xmath590 as the expansion function . by ( [ eq : prep ] ) , this amounts to replacing each @xmath4 by the @xmath51 function in terms of which it is defined ; i.e. , performing a _ hypergeometric _ expansion . adopting the suggestion of @xcite when @xmath642 amounts to replacing @xmath670 by versions that are multiplied by @xmath32^{-1/4}$ ] , which equals @xmath671 . with a trivial change in normalization , these are the functions @xmath672}{\sin(\theta/2)},\qquad
\hat\chi_{n}^{(\alpha)}(\theta ) = \frac{\cos[(2n+\alpha)(\pi-\theta)]}{\sin(\theta/2)}\qquad\ ] ] ( @xmath331 ) on @xmath657 , which are biorthogonal with respect to the weight function @xmath673 .
each @xmath674 has a finite , nonzero limit as @xmath667 and @xmath666 , and as a function of @xmath585 is proportional to @xmath675 in terms of these trigonometric functions @xmath674 , one has ( formally ) a bilateral expansion of an arbitrary function @xmath646 defined on @xmath647 , namely @xmath676 for all @xmath286 , where @xmath677 the denominator in ( [ eq : hasden ] ) equals @xmath663 for all @xmath150 , as before .
one can clearly restrict @xmath80 from @xmath678 to the interval @xmath679 , without losing generality .
if @xmath646 is piecewise continuous on @xmath680 , then in the symmetric case @xmath86 , the bilateral series in ( [ eq : pinskyexp ] ) will converge as @xmath681 to @xmath682 at all points of continuity , including the endpoints , and in general to @xmath683/2 $ ] .
the chebyshev polynomials @xmath684 of the fourth kind , for @xmath685 , are given by @xcite @xmath686}{\sin(\frac12\theta ) } = \sum_{m =- j}^j { \rm e}^{{\rm i}m\theta}.\ ] ] it follows from ( [ eq : psichihatdefs ] ) that when @xmath11 , both of @xmath687 equal @xmath688 ; and when @xmath689 , they equal @xmath690 .
the bilateral expansion in ( [ eq : pinskyexp ] ) thus reduces if @xmath86 to a _
unilateral _ expansion in the polynomials @xmath684 , @xmath685 .
the chebyshev polynomials of the first kind , @xmath691 , @xmath692 , are defined by @xmath693/(\cos\theta)$ ] . by standard fourier series theory ,
the expansion of @xmath694 in the @xmath695 , when @xmath696 is piecewise continuous on @xmath697 , will converge to @xmath696 at all points of continuity , and in general to @xmath698/2 $ ] .
but ( see @xcite ) , if one writes @xmath699 ( so that @xmath700 if @xmath585 ) , then @xmath701 equals @xmath702 .
therefore an expansion of @xmath646 in the fourth - kind @xmath701 on @xmath680 is effectively an expansion of @xmath703 on @xmath704 in the first - kind @xmath695 , all of the even-@xmath197 terms of which must vanish .
the theorem follows .
it is useful to compare this convergence result , which refers to an expansion of @xmath25 in the ferrers functions @xmath705 , with the pointwise convergence result of @xcite . the latter deals with an expansion in the functions @xmath706 , where @xmath92 is arbitrary and @xmath652 .
however , it requires that @xmath25 be piecewise smooth , not merely piecewise continuous .
as the above theorem reveals , this assumption can be relaxed ; at least , in the seemingly difficult ` corner ' case when @xmath707 .
whether smoothness can also be dropped as a hypothesis for the pointwise convergence of love
hunter expansions with @xmath652 , or with @xmath708 when @xmath709 , remains to be explored .
it must be mentioned that the octahedral and tetrahedral formulas of theorems [ thm : oct2 ] and [ thm : tetr3 ] facilitate the calculation of the coefficients in love
hunter expansions of the form ( [ eq : serieslove ] ) , with @xmath710 equal to @xmath711 and @xmath712 , respectively . because these values satisfy @xmath652 , the convergence result of @xcite applies .
in the preceding sections , explicit formulas for the legendre and ferrers functions in the octahedral , tetrahedral , dihedral , and cyclic families were derived .
each such family ( in the first - kind ferrers case ) is of the form @xmath713 , where @xmath40 are or may be fractional , and @xmath419 ranges over @xmath714 . in this section , the connection between such a family and conventional @xmath15-based harmonic analysis on the sphere @xmath14 , coordinatized by the angles @xmath715 , is briefly explored .
the connection goes through the corresponding family of generalized spherical harmonics , @xmath716 , with @xmath39 .
but the connection is not as strong as one would like .
if @xmath40 are rational but not integral , these harmonic functions will not be single - valued on the symmetric space @xmath94 .
( in the cases of interest here , each @xmath4 in the family is algebraic in @xmath60 , and they can be viewed as finite - valued . ) they may not be square - integrable , because the leading behavior of @xmath590 as @xmath26 is proportional to @xmath669 unless @xmath8 is a positive integer . for these reasons , the focus is on the action of _ lie algebras _ ( of ` infinitesimal transformations ' ) on a function family of this type , specified by @xmath99 , rather than the action of a lie group such as @xmath15 .
the space spanned by the classical spherical harmonics @xmath717 , with @xmath718 and @xmath594 , admits an action of the rotation group @xmath15 .
the lie algebra @xmath100 of @xmath719 real skew - symmetric matrices can be represented by differential operators on @xmath94 , with real coefficients , and acts on the space of spherical harmonics .
the resulting infinite - dimensional representation is reducible : for @xmath11 , it includes the usual @xmath720-dimensional representation on the span of @xmath721 .
but , @xmath100 is not the only lie algebra to be considered . a larger lie algebra than @xmath100 acts naturally on the spherical harmonics , or rather , on the ( regular ) _ solid harmonics _ @xmath722 , which satisfy laplace s equation on @xmath110 .
( see @xcite . )
this is the @xmath723-dimensional real lie algebra @xmath724 that is generated by ` ladder ' operators that increment and decrement the degree @xmath150 , as well as the order @xmath82 .
they are represented by differential operators on @xmath110 , with real coefficients .
the real span of these operators exponentiates to the lie group @xmath725 , which contains as subgroups ( i ) the @xmath726-parameter group @xmath15 of rotations about the origin , ( ii ) a @xmath726-parameter abelian group of translations of @xmath110 , ( iii ) a @xmath126-parameter group of dilatations ( linear scalings of @xmath110 ) , and ( iv ) a @xmath726-parameter abelian group of ` special conformal transformations . '
the last are quadratic rational self - maps of @xmath110 ( or rather the real projective space @xmath727 , because they can interchange finite and infinite points ) .
the preceding results , now standard , are extended below to any family of _ generalized _ solid harmonics @xmath728 , with @xmath729 for specified @xmath40 . in [ subsec : last1 ] , the differential and non - differential recurrences on @xmath7 and @xmath8 are derived .
( see theorems [ thm : diffrecs ] and [ thm : nondiffrecs ] . ) in [ subsec : last2 ] , it is shown that the operators in the differential recurrences generate a @xmath723-dimensional real lie algebra , and an isomorphism from this algebra not to @xmath724 but to @xmath2 is exhibited
. the treatment closely follows celeghini and del olmo @xcite , but the explicit isomorphism in theorem [ thm : newso32 ] is new . in the setting of special function identities , which typically involve _ real _ linear combinations of differential operators ,
@xmath2 arises more naturally than does @xmath724 .
but by a limited form of complexification , @xmath2 can be converted to @xmath724 , and indeed to @xmath730 .
these are alternative real forms of the rank-2 complex lie algebra @xmath1 , to which they complexify , and the eight displacement vectors @xmath731 , @xmath447 , @xmath115 can be identified with the roots of @xmath1 .
in [ subsec : last3 ] , it is shown that irrespective of @xmath99 , the representation of @xmath2 [ or of @xmath724 or @xmath730 ] carried by the solid harmonics @xmath733 with @xmath39 is of a special type : its quadratic casimir operator takes a fixed value , and its quartic one vanishes .
( see theorem [ thm : mostdegenerate ] . )
the former fact was found in @xcite , but the latter is new .
the representation of @xmath2 on the solid harmonics of integer degree and order , and its representation on the ones of half - odd - integer degree and order , are identified as the known dirac singleton representations of @xmath2 . in any family @xmath734 , where @xmath117 can be taken as any of @xmath735 , any three distinct members are linearly dependent , over the field of functions that are rational in @xmath60 and @xmath736 ( ferrers case ) or @xmath737 ( legendre case )
. in particular , any three contiguous members are so related , by a three - term ladder recurrence .
the underlying recurrences are differential ones , which permit any single @xmath738 and its derivative to generate any member contiguous to it , as a linear combination ; and by iteration , to generate any @xmath739 in which @xmath740 .
[ thm : diffrecs ] the ferrers functions @xmath741 satisfy eight differential recurrences , divided into four ladders , i.e. , @xmath742-pairs , with @xmath743 , @xmath744 , @xmath745 , and @xmath746 .
each pair is of the form @xmath747,\ ] ] and for each pair , the exponents @xmath748 , the exponents @xmath749 , and the prefactor @xmath750 are given in table [ tab:2 ] .
the second - kind functions @xmath144 satisfy identical recurrences .
the legendre functions @xmath751 [ the latter _
unnormalized _ ,
i.e. , the functions @xmath752 satisfy recurrences obtained from the preceding by ( i ) multiplying the right - hand side by a sign factor , equal to @xmath753 ; and ( ii ) replacing @xmath754 by @xmath755 .
the four non - diagonal recurrences on the order and degree , with @xmath743 and @xmath744 , are classical and can be found in many reference works @xcite .
they can be deduced from the differential recurrences of jacobi , which increment or decrement the parameters of the function @xmath191 .
( see @xcite2.8(27 ) ) . ) the final four diagonal ones , at least for @xmath738 when @xmath38 are integers , are due to celeghini and del olmo @xcite .
each can be derived from the non - diagonal ones by a tedious process of elimination , but the process can be systematized as the calculation of the commutator of two differential operators .
( see [ subsec : last2 ] , below . )
llll @xmath756 & @xmath748 & @xmath749 & @xmath750 + @xmath757 & @xmath758 & @xmath396 & @xmath759_\pm$ ] + @xmath744 & @xmath760 & @xmath761 & @xmath762_\pm$ ] + @xmath745 & @xmath763 & @xmath764 & @xmath765_\pm$ ] + @xmath746 & @xmath766 & @xmath764 & @xmath767_\pm$ ] + [ eq : rewrite ] @xmath768 { \rm p}_\nu^{\mu},\\ \alpha_\pm { \rm p}_{\nu\pm1}^{\mu } & = \left\{\pm(\sin\theta)d_\theta + \left[(\nu+\tfrac12)\pm\tfrac12\right]\cos\theta \right\ } { \rm p}_\nu^{\mu},\\ \alpha_\pm { \rm p}_{\nu\pm1}^{\mu\pm1 } & = \left\{\pm ( \cos\theta)d_\theta - \mu\csc\theta + \left[-(\nu+\tfrac12)\mp\tfrac12\right]\sin\theta\right\ } { \rm p}_\nu^{\mu},\label{eq : rplusminus}\\ \alpha_\pm { \rm p}_{\nu\mp1}^{\mu\pm1 } & = \left\{\pm ( \cos\theta)d_\theta - \mu\csc\theta + \left[+(\nu+\tfrac12)\mp\tfrac12\right]\sin\theta\right\ } { \rm p}_\nu^{\mu},\label{eq : splusminus}\end{aligned}\ ] ] which are satisfied by @xmath769 . here , @xmath770 , and the four prefactors @xmath750 are given in the last column of table [ tab:2 ] , in order . the recurrences ( [ eq : rplusminus ] ) and ( [ eq : splusminus ] ) have appeared in the literature but are not well known ; the only appearances that we have found , with @xmath38 restricted to integer values , are in @xcite and ( * ? ? ? * a.2 ) .
the similarity between ( [ eq : rplusminus ] ) and ( [ eq : splusminus ] ) is attributable to the fact that @xmath771 for all @xmath38 .
that is , @xmath161 is unaffected by the negating of the shifted degree parameter @xmath772 .
the three - term ladder recurrences derived from the four pairs of differential recurrences are given in the following theorem .
the diagonal ones , coming from the ladders with @xmath773 and @xmath746 , appear to be new .
[ thm : nondiffrecs ] the ferrers functions @xmath774 satisfy second - order ( i.e. , three - term ) recurrences on the order @xmath8 and degree @xmath7 , namely @xmath775 and the two diagonal recurrences @xmath776\,{\rm p}_\nu^\mu(z ) \\ + \bigl[(\nu+\tfrac12)\pm(\mu-\tfrac12)\bigr]\ , \bigl[(\nu+\tfrac12)\pm(\mu-\tfrac32)\bigr]\ , \sqrt{1-z^2}\:\,{\rm p}_{\nu\mp1}^{\mu-1}(z ) = 0
. \end{gathered}\ ] ] the second - kind functions @xmath144 satisfy identical second - order recurrences . the legendre functions @xmath751 ( the latter _
unnormalized _ , as above ) , satisfy recurrences obtained from the preceding by ( i ) multiplying each term containing a function of order @xmath777 and a coefficient proportional to @xmath778^\alpha$ ] by a sign factor , equal to @xmath779 ; and ( ii ) replacing @xmath736 by @xmath737 .
eliminate the derivative terms from the recurrences of theorem [ thm : diffrecs ] .
this is the procedure used to derive gauss s three - term , nearest - neighbor ` contiguous function relations ' for @xmath191 from jacobi s differential recurrences on @xmath192 .
it was noted in [ sec : prelims ] that if @xmath780 is a negative integer , @xmath144 and @xmath781 are generally undefined ( though there are exceptions ) .
the recurrences for @xmath144 and @xmath781 in theorems [ thm : diffrecs ] and [ thm : nondiffrecs ] remain valid in a limiting sense even when @xmath113 is such that one or more of the functions involved is undefined .
the raising and lowering of the degree and order , in any doubly indexed family of ( generalized ) solid harmonics @xmath782 where @xmath39 , can be performed by differential operators that do not need to depend explicitly on @xmath99 if they are allowed to involve , instead , the derivative operators @xmath783 in addition to @xmath784
. the basic idea is due to miller @xcite , and there is freedom in its implementation : either or both of the factors @xmath785 and @xmath786 could include an ` @xmath787 ' in its exponent , and the ferrers functions @xmath788 could be replaced by the legendre ones @xmath789 . with the choices made in ( [ eq : gensolid ] ) , @xmath790 can be viewed as a ( typically multi - valued ) function of the spherical coordinates @xmath791 , which satisfies laplace s equation on @xmath110 .
define ladder operators by [ eq : jkrs ] @xmath792,\\ k_\pm & = r^{\pm1 } \left [ \pm(\sin\theta)d_\theta + ( \cos\theta)(rd_r+\tfrac12\pm\tfrac12)\right],\\ r_\pm & = r^{\pm1}{\rm e}^{\pm{\rm i}\varphi}\left[\pm(\cos\theta)d_\theta + { \rm i}(\csc\theta)d_\varphi - ( \sin\theta)(rd_r+\tfrac12\pm\tfrac12)\right],\\ s_\pm & = r^{\mp1}{\rm e}^{\pm{\rm i}\varphi}\left[\pm(\cos\theta)d_\theta + { \rm i}(\csc\theta)d_\varphi + ( \sin\theta)(rd_r+\tfrac12\mp\tfrac12)\right ] . \end{aligned}\ ] ] [ eq : jkrsplus ] @xmath793_\pm \,{\mathcal{s}}_\nu^{\mu\pm1 } , \\ k_\pm\ , { \mathcal{s}}_\nu^\mu & = [ \nu-\mu+1,\nu+\mu]_\pm \,{\mathcal{s}}_{\nu\pm1}^\mu , \\ r_\pm\ , { \mathcal{s}}_\nu^\mu & = [ 1,(\nu+\mu)(\nu+\mu-1)]_\pm \,{\mathcal{s}}_{\nu\pm1}^{\mu\pm1 } , \\ s_\pm\ , { \mathcal{s}}_\nu^\mu & = [ 1,(\nu-\mu+1)(\nu-\mu+2)]_\pm \,{\mathcal{s}}_{\nu\mp1}^{\mu\pm1}. \end{aligned}\ ] ] for each @xmath99 , the solid harmonics @xmath794 , or more accurately their real linear span , carry a representation of @xmath795 . generating a lie algebra by working commutators out ,
one finds ( with @xmath796 $ ] signifying @xmath797 ) @xmath798,\qquad\quad s_\pm = \mp[j_\pm , k_\mp],\ ] ] which explains why the diagonal recurrences in theorems [ thm : diffrecs ] and [ thm : nondiffrecs ] can be most efficiently obtained by commutator calculations , as claimed .
it is useful additionally to define ` labeling ' operators @xmath799 so that @xmath800 by further calculations , one finds that the real lie algebra generated by @xmath801 _ closes _ , in the sense that it is finite - dimensional .
in particular , @xmath802&=\pm j_\pm,&&\qquad\qquad & [ j_+,j_-]&=2j_3 , \\ [ k_3,k_\pm]&=\pm k_\pm,&&\qquad\qquad & [ k_+,k_-]&=-2k_3 , \\ [ r_3,r_\pm]&=\pm 2r_\pm,&&\qquad\qquad & [ r_+,r_-]&=-4r_3 , \\ [ s_3,s_\pm]&=\pm 2s_\pm,&&\qquad\qquad & [ s_+,s_-]&=-4s_3,\end{aligned}\ ] ] where @xmath803 and @xmath804 . to interpret these , recall that any real linear space with basis @xmath805 , given a lie algebra structure by @xmath806=\pm x_\pm,\qquad\qquad [ x_+,x_-]=2\sigma x_3,\ ] ] is isomorphic to @xmath100 if @xmath807 , and to @xmath105 ( or equivalently @xmath808 ) if @xmath809 . hence , the real lie algebras spanned by @xmath810 , @xmath811 , @xmath812 , and @xmath813 , coming from the ladders with @xmath743 , @xmath744 , @xmath745 , and @xmath746 , are isomorphic to @xmath100 ( the first ) and @xmath105 ( the remaining ones ) .
the last two turn out to commute . the real lie algebra generated by @xmath801 , of which these copies of @xmath100 and @xmath105 are subalgebras , is @xmath723-dimensional and is spanned over @xmath814 by @xmath815 .
it of course has real structure constants . for any @xmath99
, its representation by differential operators on @xmath110 , as above , is carried by the real span of the solid harmonics @xmath816 , @xmath89 .
this result was obtained by celeghini and del olmo @xcite , though they confined themselves to integer @xmath38 , i.e. , in effect to @xmath91 .
are our @xmath817 .
also , in 5 , @xmath818 $ ] and @xmath819 $ ] should be emended to read @xmath820 $ ] and @xmath821 $ ] . ] to identify this @xmath723-dimensional real algebra , it is useful to relabel its basis elements .
first , let @xmath822 in each of which the three elements commute .
the algebra can then be viewed as the span over @xmath814 of @xmath823 ; @xmath824 ; @xmath825 and @xmath826 , which will be written as @xmath827 henceforth .
define also , for @xmath829 , define the ` skew - cartesian ' elements @xmath830 so that @xmath831 . the algebra will then be the real span of @xmath832 ; @xmath833 ; @xmath834 , or equivalently of @xmath832 ; @xmath835 ; @xmath827 .
[ eq : commreps1 ] @xmath840 & = \{-1,+1,-1\}_k\:\mathcal{j}_k,\\ [ \mathcal{j}_i , \mathcal{pc}^\pm_j ] & = \{-1,+1,-1\}_k\:\mathcal{pc}_k^\pm,\\ [ \mathcal{pc}^\pm_i , \mathcal{pc}^\pm_j ] & = \mp\{-1,+1,-1\}_k\:\mathcal{j}_k,\end{aligned}\ ] ] where @xmath841 is any cyclic permutation of @xmath842 , with @xmath843 meaning @xmath455 when @xmath844 .
also , the @xmath719 matrix of commutators @xmath845 $ ] indexed by @xmath846 equals @xmath847 .
additionally , @xmath848 = 0 , \qquad\qquad [ d,\mathcal{pc}_i^\pm ] = - \mathcal{pc}_i^\mp,\ ] ] for @xmath839 .
these identities specify the structure of the algebra . now , recall that the real lie algebra @xmath849 with @xmath850 has the following defining representation .
if @xmath851 , with @xmath512 @xmath852 s and @xmath853 @xmath127 s , then @xmath849 comprises all real @xmath854 matrices @xmath616 for which @xmath855 is skew - symmetric .
there is a sign convention here , and a @xmath856 symmetry ; without loss of generality , @xmath857 will be assumed .
it is sometimes useful to permute the @xmath852 s and @xmath127 s .
more concretely , @xmath849 can be realized as the real span of the @xmath854 matrices @xmath858 , @xmath859 , where @xmath860 . in this , @xmath861 is the @xmath862 matrix with a @xmath126 in row @xmath863 , column @xmath864 , and zeroes elsewhere .
one often extends the size-@xmath865 basis @xmath866 to a ` tensor operator , ' i.e. , a skew - symmetric @xmath862 matrix of elements @xmath867 , by requiring that @xmath868 for @xmath869 . the commutation relations @xmath870
= g_{ad}{\bf m}_{bc } + g_{bc}{\bf m}_{ad } - g_{ac}{\bf m}_{bd } - g_{bd}{\bf m}_{ac}\ ] ] are easily checked .
are the ` compact ' and ` non - compact ' subspaces .
( the terms refer to the subgroups of @xmath875 to which they exponentiate . )
real algebras isomorphic to @xmath724 and @xmath730 can be obtained by weyl s trick of redefining some or all of the basis elements of @xmath876 to include ` i ' factors . in doing this
, a slightly changed notation will be useful . for @xmath829
, define the ` cartesian ' elements @xmath877 so that @xmath878 and @xmath879 .
[ thm : newso41 ] ( i ) the real span of @xmath880 , or equivalently of @xmath881 , is a real lie algebra isomorphic to @xmath724 , an isomorphism being specified by the tensor operator @xmath882 with @xmath883 .
the real span of @xmath884 is a real lie algebra isomorphic to @xmath2 , an isomorphism being specified by the tensor operator @xmath885 with @xmath886 .
( i ) multiply the second row and the second column of the @xmath887 in theorem [ thm : newso32 ] by ` @xmath787 ' , and ( innocuously ) interchange the second and third rows , and the second and third columns .
( ii ) continuing ( or in a sense reversing ) , multiply the last row and the last column by ` @xmath787 ' . with ` @xmath787 ' factors in basis elements , the @xmath2 , @xmath724 and @xmath730 of theorems [ thm : newso41 ] and [ thm : newso5r ]
look awkward .
but it follows from ( [ eq : jkrs ] ) , ( [ eq : j3k3 ] ) , and ( [ eq : cartesian ] ) that each basis element in the @xmath724 of theorem [ thm : newso41](i ) is realized by a differential operator in @xmath791 with _ real _ coefficients .
this is not the case for the basis elements of @xmath2 and @xmath730 . where @xmath895 and @xmath841 is a cyclic permutation of @xmath842 .
that is , the @xmath896 generate rotations about the origin , the @xmath897 generate translations , and @xmath827 generates dilatations ( linear scalings of @xmath110 ) .
the @xmath898 generate special conformal transformations , which are degree-2 rational maps of @xmath110 ( or rather @xmath727 ) to itself .
the commutation relations [ eq : lastcr ] @xmath899 & = { \rm i}\epsilon_{ijk } j_k,\\ [ j_i , pc_j^\pm ] & = { \rm i}\epsilon_{ijk}pc_k^\pm,\\ [ pc_i^\pm , pc^\pm_j ] & = \mp{\rm i}\epsilon_{ijk}j_k,\\ [ pc_i^+,pc_j^- ] & = -\delta_{ij}d,\\ [ d , j_i ] & = 0,\\ [ d , pc_i^\pm ] & = - pc_i^\mp,\end{aligned}\ ] ] written in terms of @xmath900 , follow either from ( [ eq : commreps1]),([eq : commreps2 ] ) , from ( [ eq : commreps3 ] ) , or from ( [ eq:41ops ] ) . here , the summation convention of tensor analysis is employed .
the levi - civit tensor @xmath901 is skew - symmetric in all indices , with @xmath902 , and @xmath903 is the kronecker delta . together with @xmath904 the formulas the operators of ( [ eq:41ops ] ) span ( over @xmath814 ) the lie algebra of _ conformal _ differential operators on @xmath110 , which is known to have an @xmath724 structure .
( see miller @xcite . )
this @xmath724 can be viewed as acting on any suitable space of functions on @xmath110 , and exponentiates to the group @xmath725 of conformal transformations , realized as flows on @xmath110 ( or @xmath727 ) .
but the starting point used here was their action on the span of the generalized solid harmonics @xmath816 with @xmath89 , which are ( multi - valued ) solutions of laplace s equation . in the physics literature on conformal lie algebras and groups , the terms ` @xmath907 ' in ( [ eq : ci ] ) and ` @xmath908 ' in ( [ eq : d ] ) often appear as @xmath909 and @xmath910 respectively , where @xmath910 is the so - called scaling dimension ; though the resulting commutation relations do not involve @xmath910 .
the value @xmath911 is specific to the symmetry algebra of laplace s equation .
there are many variations on the technique of using differential recurrences to construct real lie algebras , realized by differential operators , which are isomorphic to the real forms of @xmath1 .
the solid harmonics @xmath790 that were employed here are extensions to @xmath110 of the ( surface ) spherical harmonics @xmath912 on the symmetric space @xmath14 . if not ferrers but legendre functions were used , the starting point would be the hyperboloidal ones @xmath913 , defined using coordinates @xmath914 on the hyperboloid @xmath915 , i.e. , the surface @xmath916 .
their extensions to @xmath110 satisfy the @xmath917-dimensional wave equation , rather than laplace s equation .
( see ( * ? ? ? * chap . 4 ) and @xcite . ) but isomorphic algebras could be constructed . in [ subsec : last2 ] , it was shown that for any @xmath99 , there are representations of the real lie algebras @xmath2 , @xmath724 , @xmath730 that are carried by the span of the family of ( generically multi - valued ) solid harmonics @xmath918 , @xmath39 .
these arise from the action of the ladder operators on the ferrers functions @xmath788 .
solid harmonics are harmonic functions on @xmath110 , satisfying laplace s equation ; and the ones in the octahedral , tetrahedral , dihedral , and cyclic families are or can be finite - valued .
these infinite - dimensional representations are restrictions of the representation of the common complexification @xmath1 , which is also carried by the ( complex ) span of the family .
they are generically irreducible , and are also generically non - skew - hermitian , so that except in special cases , they do not exponentiate to unitary representations of the corresponding lie groups , even formally .
this will now be investigated .
each of the three real lie algebras is of rank @xmath0 , so the center of its universal enveloping algebra is generated by two elements , called casimir invariants ; and any irreducible representation must represent each casimir by a constant .
the analysis of such representations resembles the unified classification of the irreducible representations of @xmath105 and @xmath100 , the real forms of @xmath108 , which is well known .
( see , e.g. , ( * ? ? ? * chap .
3 ) . ) in this , representations are classified by the value taken by their ( single ) casimir , and by their reductions with respect to a ( @xmath126-dimensional ) cartan subalgebra .
this leads to an understanding of which representations are skew - hermitian and which are finite - dimensional .
however , no comparable unified approach to all representations of @xmath2 , @xmath724 , and @xmath730 seems to have been published .
the literature has dealt almost exclusively with the skew - hermitian ones .
( @xmath2 and @xmath724 are treated separately in @xcite and @xcite , and @xmath724 and @xmath730 are treated together in @xcite . )
the starting point is the complexification @xmath1 , which is generated over @xmath919 by @xmath920 and @xmath921 , the non - diagonal ladder operators on the order and degree .
it is the complex span of @xmath922 , each of which is represented by an infinite matrix ; see ( [ eq : jkrsplus ] ) and ( [ eq : j3k3plus ] ) .
the elements @xmath923 span a cartan subalgebra ( an abelian subalgebra of maximal [ complex ] dimension , here @xmath0 ) , which is represented diagonally : @xmath924 when the representation of @xmath1 is reduced with respect to this subalgebra , it splits into an infinite direct sum of @xmath126-dimensional representations , indexed by @xmath925 .
the corresponding cartan subalgebras of the @xmath2 and @xmath724 of theorems [ thm : newso32 ] and [ thm : newso41](i ) are the real spans of @xmath926 and @xmath927 . for the @xmath2 and @xmath730 of theorems [ thm : newso41](ii ) and [ thm : newso5r ] , they are the real span of @xmath928 .
( recall that @xmath929 . ) only for the last two will the cartan subalgebra be represented by skew - hermitian matrices ; in fact , by imaginary diagonal ones .
it is readily verified that @xmath922 can serve as a cartan
weyl basis of @xmath1 , their complex span .
that is , when the adjoint actions of @xmath930 and @xmath931 on this lie algebra are simultaneously diagonalized , the common eigenvectors ( ` root vectors ' ) include @xmath795 .
the associated roots @xmath932 are @xmath0-tuples of eigenvalues , which can be identified with the displacements @xmath933 , i.e. , @xmath757 , @xmath744 , @xmath745 , @xmath746 .
these form the @xmath111 root system .
one can write @xmath934 = 0 , \qquad [ h_i , e_\alpha]=\alpha_ie_\alpha,\ ] ] where @xmath935 is the root vector associated to root @xmath80 .
the commutators @xmath936 $ ] also prove to be consistent with the @xmath111 root system .
the casimir invariants of @xmath1 and its three real forms can be computed from the commutation relations of the cartan weyl basis elements .
( for instance , the killing form for the algebra yields a quadratic casimir . ) but it is easier to express them using the tensor operator @xmath858 of any of theorems [ thm : newso32 ] , [ thm : newso41 ] , and [ thm : newso5r ] . as elements of the universal enveloping algebra , the two casimirs , quadratic and quartic ,
are defined thus @xcite : @xmath937 where @xmath938 and the summation convention is employed , indices being raised and lowered by the tensors @xmath939 and @xmath940 .
the levi - civit tensor @xmath941 is skew - symmetric in all indices , with @xmath942 .
the normalization and sign conventions are somewhat arbitrary .
[ thm : mostdegenerate ] in the representation of the universal enveloping algebra of any of the real lie algebras @xmath2 , @xmath724 , and @xmath730 on the span of the generalized solid harmonics @xmath943 , @xmath89 , the casimirs @xmath944 and @xmath945 are represented by the constants @xmath946 and @xmath543 , irrespective of @xmath40 . where @xmath949 is the anti - commutator .
this expresses @xmath944 in terms of @xmath923 and the root vectors . from their representations ( [ eq : jkrsplus]),([eq : j3k3plus ] ) as infinite matrices , one calculates that @xmath944 ( like @xmath923 ) is diagonal in @xmath419 , with each diagonal element equaling @xmath950 . for @xmath730 , which is representative ,
the five components of @xmath951 include ( i ) the scalar @xmath952 , ( ii ) the three components of the vector @xmath953 , and ( iii ) the scalar @xmath954 .
these expressions , involving the scalar and vector product of three - vectors , must be interpreted with care : any product @xmath955 of two lie algebra elements signifies the symmetrized product @xmath956 .
but by direct computation , one finds from ( [ eq : lastneeded1]),([eq : lastneeded2 ] ) and the infinite matrix representations ( [ eq : jkrsplus]),([eq : j3k3plus ] ) that each component of @xmath951 is represented by the zero matrix , even ( surprisingly ) without symmetrization .
this result is plausible , if not expected . in any unitary representation of a semi - simple lie group @xmath957 on @xmath958 ,
@xmath466 being a homogeneous space @xmath959 of rank @xmath126 , all casimir operators except the quadratic one must vanish .
( see @xcite and ( * ? ? ? *
x ) . ) however , the present representations of @xmath724 , by real differential operators acting on multi - valued , non - square - integrable functions , are non - skew - hermitian , and can not be exponentiated to _
unitary _ representations of @xmath725 of this ` most degenerate ' type .
each of these representations of a real lie algebra @xmath960 , @xmath724 , or @xmath730 , or its universal enveloping algebra @xmath961 , on the solid harmonics @xmath816 , @xmath89 , can be viewed linear - algebraically : as a homomorphism @xmath962 of real vector spaces , taking @xmath963 ( or @xmath961 ) into the space of infinite matrices indexed by @xmath419 . for each basis element @xmath964 , @xmath965
is determined by ( [ eq : jkrsplus]),([eq : j3k3plus ] ) ; and because the basis elements given in theorems [ thm : newso41 ] and [ thm : newso5r ] include ` @xmath787 ' factors , its elements may be complex . to show that certain of these representations are equivalent to known ones by infinite matrices that are skew - hermitian , consider the effect of replacing the family @xmath966 by @xmath967 , where @xmath968 that is , @xmath969 , where @xmath970 when @xmath971 , with @xmath972 , this @xmath973 is the classical ( complex ) spherical harmonic on @xmath94 , of degree @xmath7 and order @xmath8 . , appropriate for an inner product on @xmath94 ,
is omitted .
but this @xmath974 automatically includes the so - called condon
shortley factor , owing to the definition of @xmath975 used here ( see [ sec : prelims ] ) . ] in this case , the square root factor equals @xmath976^{-1/2}$ ] and is positive by convention ; a discussion of how to interpret it in other cases is deferred .
at least formally , the representation of @xmath963 or @xmath961 on @xmath977 comes from that on @xmath978 by a diagonal similarity transformation .
the formulas and @xmath980 replace ( [ eq : jkrsplus ] ) and ( [ eq : j3k3plus ] ) ; the actions of the elements @xmath981 , @xmath982 , @xmath983 , @xmath827 are still defined in terms of these by ( [ eq : jiddef ] ) and ( [ eq : pcidef ] ) . [ thm : last ] if @xmath91 or @xmath984 , the representation @xmath962 of @xmath960 on @xmath985 , obtained by identifying @xmath2 with the real span of @xmath986 as in theorem [ thm : newso41](ii ) , is defined on the subspace with @xmath11 and @xmath987 ; and every element is represented by an infinite matrix , indexed by @xmath419 , that is skew - hermitian . by ( [ eq : jkrsnew ] ) , if @xmath988 then @xmath989 give zero when acting on @xmath990 , and if @xmath991 then @xmath992 give zero .
thus @xmath962 restricts to the subspace on which @xmath993 , @xmath11 , and @xmath987 . on this subspace ,
the similarity transformation performed by the square root factor in ( [ eq : replacement ] ) is not singular : one of the gammas is infinite only if @xmath994 is a negative integer . for @xmath99
equal to either of @xmath122 and @xmath984 , this square root factor simply equals @xmath976^{-1/2}$ ] . by ( [ eq : jkrsnew ] ) and ( [ eq : j3k3plusnew ] ) , each of @xmath995 is a real matrix , the plus and minus versions being transposes of each other , and @xmath996 being symmetric ( and diagonal ) .
it follows from ( [ eq : jiddef ] ) and ( [ eq : pcidef ] ) that the @xmath997 , the @xmath998 , and @xmath999 are hermitian , and the @xmath1000 are skew - hermitian .
the claim follows .
being skew - hermitian , the two infinite - dimensional representations of @xmath2 in theorem [ thm : last ] exponentiate to ( irreducible ) unitary representations of the so - called anti- sitter group @xmath875 , or its universal cover . the latter have been classified @xcite , and the ones arising from the theorem can be identified .
they are the remarkable dirac singleton representations , with whimsical names @xcite : the @xmath121 one is ` di ' and the @xmath91 one is ` rac . ' for the dirac singletons , the casimirs @xmath1001 are known to equal @xmath1002 .
( see @xcite . )
they are singletons in the sense that if they are reduced with respect to the subalgebra @xmath1003 and split into representations of @xmath1004 , each of the latter that appears , does so with unit multiplicity .
the ones that appear are labeled uniquely by @xmath993 , @xmath11 .
the rac representation of @xmath2 is realized by differential operators on @xmath110 ( with complex coefficients ) , expressions for which follow immediately from ( [ eq:41ops ] ) .
they act on the span of the _ classical _ solid harmonics , @xmath1005 with @xmath11 and @xmath1006 .
kyriakopoulos @xcite in effect discovered that the rac has such a realization , before the name was coined , and extended this result to higher dimensions .
the di is realized by the same operators , acting on the span of @xmath1007 with @xmath11 and @xmath1008 .
the existence of this realization seems not to be known .
this is perhaps because the ` spinorial ' solid harmonics of half - odd - integer degree and order are double - valued on @xmath110 , are typically non - square - integrable , and are based on the little - known dihedral ferrers functions .
the expressions for the dihedral ferrers functions in terms of jacobi polynomials , given in theorem [ thm : dihedralpq ] above , are new . the representations of @xmath2 carried by the octahedral and tetrahedral families of solid harmonics , @xmath1009 with @xmath1010 equal to @xmath711 , @xmath712 , and @xmath1011 , are not skew - hermitian , even up to diagonal equivalence .
changing the basis to @xmath967 does not help matters , because only if @xmath40 are both integers or both half - odd - integers does it permit the representation to be restricted to a subspace on which @xmath993 , @xmath11 , and @xmath987 . for general @xmath99
, there is accordingly no restriction on the index @xmath89 of the basis functions of the representation , and the square roots in ( [ eq : replacement]),([eq : replacement2]),([eq : jkrsnew ] ) may be square roots of negative quantities .
irrespective of what sign convention for the square root is adopted , the resulting imaginary factors will interfere with skew - hermiticity ; and upon integration of the representation , with unitarity . in fact
, the familiar definition of the ( surface ) spherical harmonic @xmath1012 given in ( [ eq : replacement2 ] ) , incorporating the square root factor , seems to be useful only when the degree @xmath7 and the order @xmath8 are both integers or both half - odd - integers .
bohm , a. , loewe , m. , magnollay , p. , tarlini , m. , aldinger , r.r .
, biedenharn , l.c . , van dam , h. : quantum relativistic oscillator .
contraction between the algebras of @xmath1014 and the three - dimensional harmonic oscillator .
d * 32*(10 ) , 28282834 ( 1985 ) dobrev , v.k . ,
sezgin , e. : spectrum and character formul of so(3,2 ) unitary representations . in : j.d .
hennig , w. lcke , j. tolar ( eds . ) differential geometry , group representations , and quantization , no . 379 in lecture notes in physics , pp .
springer - verlag , new york / berlin ( 1991 ) kuriyan , j.g . ,
mukunda , n. , sudarshan , e.c.g . : master analytic representations and unified representation theory of certain orthogonal and pseudo - orthogonal groups . commun .
* 8 * , 204227 ( 1968 ) olver , f.w.j . ,
lozier , d.w . ,
boisvert , r.f . ,
clark , c.w .
( eds . ) : nist handbook of mathematical functions .
department of commerce , national institute of standards and technology , washington , dc ( 2010 ) | trigonometric formulas are derived for certain families of associated legendre functions of fractional degree and order , for use in approximation theory .
these functions are algebraic , and when viewed as gauss hypergeometric functions , belong to types classified by schwarz , with dihedral , tetrahedral , or octahedral monodromy .
the dihedral legendre functions are expressed in terms of jacobi polynomials .
for the last two monodromy types , an underlying ` octahedral ' polynomial , indexed by the degree and order and having a non - classical kind of orthogonality , is identified , and recurrences for it are worked out .
it is a ( generalized ) heun polynomial , not a hypergeometric one . for each of these families of algebraic associated legendre functions , a representation of the rank-@xmath0 lie algebra @xmath1
is generated by the ladder operators that shift the degree and order of the corresponding solid harmonics .
all such representations of @xmath1 are shown to have a common value for each of its two casimir invariants .
the dirac singleton representations of @xmath2 are included . |
there are two approaches to characterizing spacetime singularities in a cosmological context .
the first approach may be called _ geometric _ and consists of finding sufficient and/or necessary conditions for singularity formation , or absence , _ independently _ of any specific solution of the field equations under general conditions on the matter fields .
methods of this sort include those based on an analysis of geodesic congruences in spacetime and lead to the well known singularity theorems , cf .
@xcite , as well as those which are depend on an analysis of the geodesic equations themselves and lead to completeness theorems such as those expounded in cbc02 , and the classification of singularities in @xcite . the second approach to the singularity problem can be termed _ dynamical _ and refers to characterizing cosmological singularities in a geometric theory of gravity by analysing the dynamical field equations of the theory _ _ _ _ it uses methods from the theory of dynamical systems and can be _ global _ , referring to the asymptotic behaviour of the system of field equations for large times , or _
local _ , giving the behaviour of the field components in a small neighborhood of the finite - time singularity . in this latter spirit
, we present here a local method for the characterization of the asymptotic properties of solutions to the field equations of a given theory of gravity in the neighborhood of the spacetime singularity .
we are interested in providing an asymptotic form for the solution near singularities of the gravitational field and understanding all possible dominant features of the field as we approach the singularity .
we call this approach the _ method of asymptotic splittings . _ in the following sections , we give an outline of the method of asymptotic splittings with a view to its eventual application to cosmological spacetimes in different theories of gravity . for the sake of illustration , in the last section
we analyze the asymptotic behaviour of a friedmann - robertson - walker ( frw ) universe filled with perfect fluid in einstein s general relativity , which provides the simplest , nontrivial cosmological system .
it is advantageous to work on any differentiable manifold @xmath0 , although for specific applications we restrict attention to open subsets or @xmath1 we shall use interchangeably the terms vector field @xmath2 and dynamical system defined by @xmath3 on @xmath0 , @xmath4 , with @xmath5 .
also , we will use the terms integral curve @xmath6 of the vector field @xmath3 with initial condition @xmath7 , and solution of the associated dynamical system @xmath8 passing through the point @xmath9 , with identical meanings . given a vector field @xmath3 on the @xmath10-dimensional manifold @xmath11 , we define the notion of a _ general _ solution of the associated dynamical system as a solution that depends on @xmath10 arbitrary constants of integration , @xmath12 .
these constants are uniquely determined by the initial conditions in the sense that to each @xmath13 we can always find a @xmath14 such that the solution @xmath15 is the unique solution passing through the point @xmath13 .
therefore , a property holds _ independently _ of the initial conditions if and only if it is a property of a general solution of the system .
a _ particular _ solution of the dynamical system is any solution obtained from the general solution by assigning specific values to at least one of the arbitrary constants .
the particular solutions containing @xmath16 arbitrary constants can be viewed as describing the evolution in time of sets of initial conditions of dimension @xmath16 strictly smaller than @xmath10 .
a particular solution is called an _
exact _ solution of the dynamical system when @xmath17 .
thus , in our terminology , a particular solution is a more general object than any exact solution , the latter having the property that all arbitrary constants have been given specific values .
the hierarchy : exact ( no arbitrary constants ) to particular ( strictly less than maximum number of arbitrary constants ) to general solutions , will play an important role in what follows .
general , particular , or exact solutions of dynamical systems can develop _ finite - time singularities _ ; that is , instances where a solution @xmath18 , misbehaves at a finite value @xmath19 of the time @xmath20 .
this is made precise as follows .
we say that the system @xmath21 ( equivalently , the vector field @xmath3 ) has a _ finite - time singularity _ if there exists a @xmath22 and a @xmath23 such that for all @xmath24 there exists an @xmath25 such that @xmath26for @xmath27 . here
@xmath28 , @xmath29 for some @xmath30 , and @xmath31 ( resp .
@xmath32 ) . note also , that @xmath33 is an arbitrary point in the domain @xmath34 and may be taken to mean ` now ' . alternatively , we may set @xmath35 , @xmath36 , and consider the solution in terms of the new time variable @xmath37 , @xmath38 , with a finite - time singularity at @xmath39 .
we see that for a vector field to have a finite - time singularity there must be at least one integral curve passing through the point @xmath13 of @xmath40 such that at least one of its @xmath41 norms diverges at @xmath42 .
we write @xmath43to denote a finite - time singularity at @xmath44 .
one of the most interesting problems in the theory of singularities of vector fields is to find the structure of the set of points @xmath13 in @xmath45 such that , when evolved through the dynamical system defined by the vector field , the integral curve of @xmath3 passing through a point in that set satisfies property ( [ sing2 ] ) .
another important question , of special interest in relativistic cosmology , is to discover the precise relation between the finite - time singularities of vector fields that arise as reductions of the field equations and those that emerge in the form of geodesic incompleteness .
the difficulty here is that the finite - time singularities of vector fields appear to be unconnected to geodesic incompleteness and conversely , singularities which arise through the formation of conjugate points do not seem to demand , or require , any dynamical description .
finite - time singularities of ( general or particular ) solutions of linear dynamical systems are located at the singularities of their coefficients and are _ fixed _ because they are known from the singularities of the coefficients of the system .
the fixed singularities in a solution are therefore independent of the choice of initial conditions . in contrast
, solutions of nonlinear systems can develop finite - time singularities that are either fixed or movable .
a singularity is _ fixed _ if it is a singularity of @xmath46 for all @xmath47 ; otherwise , we say it is a _
movable singularity_. note that any fixed finite - time singularity of a particular solution ( @xmath48 ) can not be a fixed singularity of a general solution since at least one of the constants appearing in the general solution has been set to zero , and so this singularity is not one of a @xmath46 for _ all _ @xmath49 , that is independent of the initial conditions .
hence , fixed finite - time singularities _ in a general solution _ can not be understood by studying fixed singularities in particular solutions . however , movable singularities of a particular solution , if they _
are _ singularities ( in the sense of the definition above ) of the general solution , will always be movable ones .
therefore , movable finite - time singularities in particular solutions make a nonzero contribution to the singularity pattern of the vector field and must be taken into consideration in the general study of its singularities .
it may also happen that a dynamical system has no movable singularities in the general solution but still has a singular or particular solution with a movable singularity .
hence , a fixed ( or movable ) singularity in a general solution can be a fixed ( or movable ) singularity of some particular solution but not vice versa .
in general , movable singularities are more interesting , since the issue of choosing initial conditions plays an important role for them ; consequently , we shall restrict our attention to them almost exclusively in what follows .
how should we tackle the geometric problem of describing the behaviour of vector fields and their integral curves solutions of the associated dynamical system in the neighborhood of a finite - time movable singularity ?
assume that we are given a vector field , and we know that at some point , @xmath44 , a system of integral curves , corresponding to a particular or a general solution , has a ( future or past ) finite - time singularity in the sense of definition ( [ sing1 ] ) .
the approach we take in this paper is an asymptotic one .
the vector field ( or its integral curves ) can basically do two things sufficiently close to the finite - time singularity , namely , it can either show some dominant feature or not . in the latter case ,
the integral curves can ` spiral ' in some way around the singularity _ ad infinitum _ so that ( [ sing1 ] ) is satisfied and the dynamics are totally controlled by the subdominant ( lower - order in terms of weight - see below ) terms , whereas in the former case solutions share a distinctly dominant behaviour on approach to the singularity at @xmath44 determined by the most nonlinear terms . to describe both cases
invariably , we decompose the vector field into simpler , component vector fields and examine whether the most nonlinear one of these shows a dominant behaviour while the rest become subdominant in some exact sense . using this picture
, we then built a system of integral curves corresponding , where feasible , to the general solution , and sharing exactly its characteristics in a sufficiently small neighborhood of the finite - time singularity .
the construction of these solutions is given as a formal asymptotic series expansion around the singularity and is done term - by - term .
we begin by introducing some useful notation and terminology .
we write @xmath50 to denote the function @xmath51 where @xmath52 and @xmath53 a function of the form ( [ scale function ] ) is _ scale invariant _ in the sense that a change in the time scale , @xmath54 , @xmath55 , reveals that we must also have @xmath56 , and conversely .
now we demand that a scale invariant function @xmath57 of the form ( [ scale function ] ) is an integral curve of the vector field @xmath3 passing through @xmath13 , or equivalently , the associated dynamical system @xmath21 has a particular solution that is scale invariant , valid in a neighborhood of the assumed finite - time singularity .
this means that @xmath58the notation @xmath59 stands for the monomial @xmath60 and is valid for each @xmath61 ( range convention , no summation ) .
we say that a vector field @xmath3 ( or the associated dynamical system ) is _ scale invariant _ if it satisfies @xmath62more generally , a vector field is called _ weight - homogeneous _ with _ weighted degree _
@xmath63 if there is a vector @xmath64 , called _ the weight _ , and a vector @xmath65 such that @xmath66we denote the degree by @xmath67 .
when @xmath68 the vector field is called _ homogeneous _ of degree @xmath63 . note that any scale invariant vector field is a weight - homogeneous field @xmath69 such that each component @xmath70 has degree @xmath71 , from which it follows that @xmath72 has weight @xmath73 .
if @xmath3 has degree @xmath63 , then @xmath74 has degree @xmath75 , with @xmath76 the @xmath77-th unit vector .
there are weight - homogeneous vector fields that are not scale invariant ; for instance , all linear dynamical systems with constant coefficients which are homogeneous vector fields . using eqs .
( [ basic1 ] ) , ( [ basic2 ] ) we see that given any nonzero vector @xmath78 , a scale - invariant vector field @xmath3 admits a scale - invariant integral curve provided that the inhomogeneous linear system @xmath79has nontrivial solutions for @xmath80 .
note that there may be nontrivial solutions for @xmath80 with some components zero .
when at least one of the components of @xmath80 is nonzero , the corresponding solution of the form ( [ scale function ] ) is a particular solution of the dynamical system .
therefore , in this case , we know the exact asymptotic behaviour of the vector field @xmath3 in the neighborhood of a finite - time singularity , given by the solution ( [ scale function ] ) with suitable @xmath81 s and @xmath78 s .
unfortunately , most vector fields are neither scale invariant nor weight - homogeneous . however , since any _ analytic _ vector field @xmath82 can be expanded in a power series in some domain @xmath83 , by taking any @xmath84 for any @xmath85 distinct from @xmath86 any such vector field can be decomposed into weight - homogeneous _ components _ by taking for instance the first @xmath87 terms in its taylor expansion around @xmath9 .
a simpler example of a vector field admitting a weight - homogeneous splitting is to take @xmath3 to be any polynomial vector field ; then , a possible decomposition is to split it such that each component is a suitable combination of monomials .
we say that the nonlinear vector field @xmath3 on @xmath0 admits a _ weight - homogeneous decomposition _ with respect to a given vector @xmath78 if it splits as a combination of the form @xmath88where the _ components _
@xmath89 are weight - homogeneous vector fields , namely , @xmath90for some non - negative numbers @xmath91 and all @xmath80 in some domain @xmath92 of @xmath93 . in terms of individual components ,
condition ( [ dec2 ] ) reads @xmath94 in a slightly vague but suggestive manner we can say that a weight - homogeneous decomposition splits the original vector field in parts starting by collecting together the most nonlinear part and then proceeding down to the ` weakest ' component such that each term in the splitting is ` less nonlinear ' than the previous one in a precise sense .
there are two important features of such a vector - field decomposition ( dec1 ) , ( [ dec2 ] ) .
firstly , it is not unique . in general , for the given vector field @xmath3 , many different vectors @xmath78 can be found which each lead to different weight - homogeneous decompositions of the vector field in the same domain @xmath95 .
secondly , since the _ subdominant exponents _ @xmath96 can be ordered , @xmath97the degrees of the component vector fields in the decomposition ( [ dec2 ] ) are also ordered and ( only ) the first vector field @xmath98 appearing in the decomposition is scale invariant .
therefore , a weight - homogeneous decomposition of a vector field with respect to a vector @xmath78 is a splitting into @xmath87 weight - homogeneous components each with degree @xmath99 such that the lowest - order vector field in the decomposition is scale invariant .
the lowest - order vector field , @xmath98 , in a splitting is sometimes called _ the dominant part _ of @xmath3 and includes the most nonlinear terms in @xmath3 , whereas the remaining sum of parts , @xmath100 , is called _ the subdominant part_. given a vector field @xmath101 it is very important to have the complete list of all possible weight - homogeneous decompositions it admits ; in other words , to know all the possible dominant and subdominant ways it can be split . the asymptotic method we employ to trace the behaviour of vector fields and their integral curves in a neighborhood of a movable finite - time singularity ( of a particular or general solution ) , begins by finding all weight - homogeneous decompositions of the vector field valid in that neighborhood .
suppose that there exists a decomposition @xmath102into ( @xmath87 ) weight - homogeneous components , such that the dominant part @xmath103 is scale invariant , and each subdominant component @xmath104 is weight - homogeneous . the scale invariant solution @xmath105is a solution only , as complex ones do not describe the behaviour near finite time singularities in the sense of eq .
( [ sing2 ] ) . ] of the dominant part @xmath106 of the vector field provided that @xmath107 .
thus , some of the components of the vector @xmath80 may vanish .
we call the components of the vector @xmath108 the _ dominant exponents . _ in this case , we sometimes say that ( [ dec2 ] ) is an _ asymptotic solution _ of the original system @xmath109
. on the other hand , the @xmath16 subdominant components @xmath110 satisfy @xmath111with the subdominant exponents @xmath112 which are ordered and strictly positive .
dividing both sides by @xmath113 and taking the limit as @xmath114 , the true meaning of the subdominant exponents is revealed , and we have _ 0=0 , which proves that the subdominant part of the vector field , @xmath115 , is less dominant than the dominant part , @xmath116 , which of course asymptotes as @xmath117 .
we say that the pair @xmath118 is a _
( dominant ) balance _ for the vector field @xmath3 if the latter admits a decomposition satisfying eqs .
( [ dec1])-([dec3 ] )
. there may be several different balances for any particular decomposition of @xmath3 and a way to classify them is by means of their order .
the _ order _ of a balance @xmath119 is the number of the nonzero components of the vector @xmath80 . for a vector field on @xmath0 ,
the highest order of any possible balance is @xmath10 and in this case the scale - invariant solution ( [ dec2 ] ) corresponds to a possible dominant behaviour of a _ general _ solution of the original system @xmath120 near the singularity . on the other hand , balances of a lower order than @xmath10
describe possible asymptotics of _ particular _ solutions .
there is an elegant convex - geometric explanation of the dominant balances of a vector field @xmath3 .
this requires us first to express the vector field in the so - called quasi - monomial form .
then , a dominant balance of order @xmath121 corresponds precisely to a @xmath121-dimensional face of the newton - puiseux - bruno polyhedron associated to @xmath3 ( cf .
@xcite , @xcite ) .
we can now study an important square matrix , the so - called _ ( k-)ovalevskaya matrix _
, associated with a given vector field @xmath3 . consider the dominant part , @xmath116 , of @xmath3 which admits an exact solution of the form ( [ dec2 ] ) , described by the dominant balance @xmath118 ( of any nonzero order ) . the _ k - matrix of the vector field @xmath3 at the balance _
@xmath118 is the square matrix @xmath122the _
( k-)ovalevskaya exponents _ associated with the balance @xmath123 are the @xmath10 eigenvalues @xmath124 of @xmath125 . when the order of the balance is @xmath10 ( @xmath126 @xmath61 ) , the _ k_-exponents are called the _ resonances _ of @xmath127 . setting @xmath128 and differentiating with respect to @xmath129 we have @xmath130while by the chain rule @xmath131where the last equality is most easily understood if we expand the derivative to obtain @xmath132 , for the @xmath133-th component .
thus , from the last two equations , we find @xmath134that is , the k - matrix always has the vector @xmath135 as an eigenvector with eigenvalue equal to @xmath136 .
we say that a balance is _ hyperbolic _ if the remaining @xmath137 k - exponents have positive real parts .
suppose now that we know all the eigenvectors @xmath138 and eigenvalues @xmath139 of the k - matrix . by simply inspecting the form of the two sides in the variational equation for the dominant part of the vector field , namely , the equation @xmath140we can write the set of fundamental solutions @xmath141 of this linear equation for @xmath142 in the form @xmath143where , depending on whether or not @xmath127 is semi - simple , the @xmath144 s are the eigenvectors @xmath141 or , in general , polynomials in @xmath145 .
hence , the solutions of the variational equation will be appropriate sums of terms by the form ( [ var ] ) .
therefore , any solution of the original system will be well approximated ( cf .
@xcite , p.299 , for the precise conditions ) by a solution of the form ( this is a taylor estimate ) : @xmath146where @xmath147where the @xmath148 s are , in general , polynomials in @xmath145 .
furthermore , we arrive at the interesting conclusion that in any particular or general solution , the arbitrary constants characterizing it will first appear in those terms whose coefficients have indices equal to a k - exponent . in a general solution , the arbitrary constants normally appear at different places in an expansion , and consequently , a solution in which a k - exponent ( different from the @xmath149 value which as we have shown always exists ) is either negative , or has nontrivial multiplicity , may or may not be a general solution .
we shall see in the next section how we can use eq .
( [ main1 ] ) to obtain an important series representation of the solutions to the original dynamical system near a finite - time singularity .
apart from the principal use made here in unravelling the nature of finite - time singularities , there are various important connections between the k - exponents , first integrals , integrability properties of hamiltonian systems and complex algebraic geometry , cf .
@xcite , @xcite .
in the previous subsection , we derived the formal expansion ( [ main1 ] ) for the solution of the dynamical system @xmath150 near a finite - time singularity by including only the first terms in a power - series expansion of @xmath151 .
that solution corresponds to a given dominant balance @xmath152 and depends on the k - exponents @xmath139 associated with the dominant part , @xmath116 , of the original vector field @xmath153 .
writing eq .
( [ main1 ] ) in full , that is including terms of all orders , can lead to very complicated expansions in general , and the essential feature is the appearance of logarithmic terms .
we note that there are a number of general theorems guaranteeing the existence and convergence of such expansions see , for instance , @xcite for a review .
let @xmath154 be the set of all k - exponents with positive real parts . for simplicity
we assume that all k - exponents in @xmath154 are rational , and define the number @xmath155 to be the least common multiple of the denominators of the numbers in the set @xmath156 . in the case where any log terms are absent ( for instance , when k is semi - simple ) , we can write , by ( [ main1 ] ) , the full expansion of the general solution around the finite - time singularity in the form of a _
puiseux series _ ,
@xmath157where , as we know already from the previous section , each of the @xmath10 arbitrary constants in ( [ main2 ] ) will first appear in the term with coefficient @xmath158 and @xmath159 .
hence , finding the final form of the solution ( general or particular , depending on the number of arbitrary constants appearing in the series expansion ) can be now reduced to knowing the coefficients @xmath160 in the expansion .
these coefficients are computed by inserting the puiseux series ( [ main2 ] ) into the original system .
this leads to a set of _ recursion relations _ , a linear system for the coefficients @xmath160 . for the @xmath77th - order coefficient
we find @xmath161 where the forms @xmath162 are polynomial in its variable , read off from the original equation .
there is an important consistency condition to be satisfied for the above analysis to be valid . multiplying both sides of ( [ main3 ] ) by @xmath163 , an eigenvector of the k - matrix
, we see that when @xmath164 , an eigenvalue of the k - matrix , we must have the following _ compatibility condition _
( @xmath165 denotes the adjoint eigenvector of @xmath166 ) : @xmath167therefore , if the above compatibility condition is _ violated _ at some eigenvalue , then we conclude that no solution in the form of a puiseux series can exist and we have to search for more general solutions which may contain logarithmic terms .
such a more general series will be of the form of a _
@xmath168-series _ ( cf .
@xcite for this terminology ) : a direct generalization of the form ( [ main2 ] ) [ main4 ] = ^(+_i=1^_j=1^ _ ij^i / s(^)^j / s ) , where @xmath169 is the first k - exponent for which the compatibility condition is not satisfied and @xmath170 as defined above .
the procedure for the calculation of the coefficients in this more general case is the same as before , leading again to the form of the general solution in a suitable neighborhood of the finite - time singularity as a @xmath168-series .
suppose that ( [ dec2 ] ) is an asymptotic solution of the vector field @xmath171 .
we set @xmath172and imagine that an equation of the form @xmath173regards the coefficient @xmath80 of a given balance @xmath174 of the vector field @xmath3 as an equilibrium point of the new system given , in terms of the new variables @xmath175 , by @xmath176where @xmath177 and @xmath178 is the least common multiple of the denominators of the subdominant exponents in the original system .
we call the dynamical system ( [ comp ] ) the _ companion system _ of the original vector field @xmath21 .
thus , the transformation ( [ comp ] ) to the companion system associates a different system of this sort to each one of the balances @xmath118 of the original system provided that any given pair of balances has different dominant exponents @xmath179 .
consider now the _
linearized system _
@xmath180which is the variational equation of the companion system around the equilibrium point @xmath181 .
this is a constant - coefficient , linear system . from the fundamental theorem of such systems
, it follows that the general solution passing through the initial condition @xmath182 is given by @xmath183now we know ( cf .
@xcite ) that for any @xmath184 matrix @xmath185 : @xmath186where @xmath187 are the eigenvalues of @xmath185 and @xmath188 are @xmath189-valued polynomials , the latter being constant if and only if @xmath190 is semi - simple ( diagonalizable ) .
so @xmath191 in terms of the @xmath37-time .
hence , using this form of the local solution to the companion system , we can express the local solution to the original system around its finite - time singularity in the form @xmath192which is precisely the taylor estimate leading eventually to the @xmath193-series representation we arrived at in previous sections .
we therefore reach the interesting conclusion that around the equilibrium point @xmath194 , origin of the companion system ( [ comp ] ) , the eigenvalues of that system are simply @xmath195 ; that is , the dominant exponents of the asymptotic solution of the original system together with the number @xmath178 characterizing the subdominant part of the vector field . moreover , around the equilibrium point @xmath196 of the companion system ( [ comp ] ) , the eigenvalues @xmath197 of the companion system are @xmath198 ; that is , they are precisely the k - exponents of the original system together with the subdominant number @xmath199 .
we have an interpretation of the results of the previous sections concerning the local behaviour of the original system around its finite - time singularities from a dynamical systems perspective .
using the companion transformation , a local analysis of the companion system @xmath200 _ around its equilibrium points _ ( with @xmath201 necessarily ) will provide the local analysis of the solutions of all possible balances of the original system _ around its singularities _
( @xmath202 since @xmath203 always , for an acceptable decomposition ) .
note that , since we need @xmath204 , we are only interested in the unstable manifold ( eigenvalues with positive real parts ) of the equilibrium points of the companion system . therefore , the negative k - exponents ( corresponding to the stable manifold of the companion system ) are not connected to the behaviour of solutions of the original system at the finite - time singularity , but are associated with its behaviour as @xmath205 . for a discussion of the companion transformation in connection with integrability and complex dynamics , see @xcite and refs . therein .
the results in the previous subsections suggest a general procedure to uncover the nature of singularities by constructing series expansion representations of particular or general solutions of dynamical systems in suitable neighborhoods of their finite - time singularities .
this method consists of building splittings of vector fields that are valid asymptotically and trace the dominant behaviour of the vector field near the singularity .
a resulting series expansion connected to a particular dominant balance helps to decide whether or not the arrived solution is a general one and to spot the exact positions of the arbitrary constants as well as their role in deciding about the nature of the time singularity .
the method we suggest below is analogous to the so - called ars procedure connected with the painlev and integrability properties of dynamical systems @xcite , but here the whole approach and viewpoint are completely different .
we are not concerned with notions of integrability but solely with the problem of the nature of finite - time singularities .
also , our systems are real - valued with a real time variable . to apply this
_ method of asymptotic splittings _ to a particular dynamical system in an effort to discover the nature of its time singularities , we must follow this recipe : 1 .
write the system of equations in the form of a dynamical system @xmath206 with @xmath207 , and identify the vector field @xmath208 .
2 . find all the different weight - homogeneous decompositions of the system ; that is , the splittings of the form @xmath209and choose one of these splittings to start the procedure .
3 . substitute the scale - invariant solution @xmath210into the equation @xmath211 . study the resulting algebraic systems , and find all dominant balances @xmath118 together with their orders .
4 . identify the non - dominant exponents , that is the positive numbers @xmath212 , such that @xmath213 5 .
construct the k - matrix @xmath127 : @xmath214 6 .
compute the spectrum of @xmath127 , @xmath215is @xmath127 semi - simple ?
are the balances hyperbolic ? 7 .
find the eigenvectors @xmath138 of @xmath127 .
identify @xmath170 as the multiplicative inverse of the least common multiple of all the subdominant exponents and positive k - exponents .
[ puiseux ] substitute the puiseux series @xmath216into the original system .
identify the polynomials @xmath217 and solve for the final recursion relations which give the unknown coefficients @xmath218 .
check the compatibility conditions at the k - exponents , @xmath219 12 .
if the puiseux series is valid , then the method is concluded for this particular splitting .
otherwise , if compatibility conditions are violated at the eigenvalue @xmath220 , restart from step [ puiseux ] by substituting the logarithmic series ( [ main4 ] ) . 13 .
get coefficient at order @xmath220 .
write down the final expansion with terms up to order @xmath220
verify that compatibility at @xmath221 is now satisfied .
repeat whole procedure for each of the other possible decompositions .
note that although the whole spectrum of possible behaviours of the system near a time singularity is concluded once we find valid series expansions corresponding to each balance in each particular decomposition , an additional analysis of the phase space of the companion systems corresponding to each one of the balances of the original system may lead to valuable insights as to the geometric structure of the phase space _ how the orbits behave _ in the neighborhood of the finite - time singularity . following the above steps even up to that of calculating a dominant balance in one particular decomposition , can be very useful since it gives you one particular possible asymptotic behaviour of the system near the time singularity . in this respect , the whole method expounded here is truly generic since it helps to decide the generality of any behaviour found in an exact solution that is , how many arbitrary constants there are in the final solution that shares that behaviour ( particular or general solution ) .
it is rare that a puiseux series is inadequate to describe the dynamics ( semi - simplicity of k ) , but in such uncommon cases one must resort to the more complex logarithmic solutions .
as a concrete example of the above analysis , consider the homogeneous and isotropic frw cosmological equations in general relativity for a perfect fluid of pressure @xmath222 , density @xmath223 and equation of state @xmath224 , where @xmath225 is a constant .
they read @xmath226where @xmath227 is the scale factor , @xmath228 the hubble expansion rate and @xmath16 the constant spatial curvature . setting @xmath229 , this system reads @xmath230with @xmath231and is subject to the integral constraint ( the friedmann equation ) @xmath232this system is weight - homogeneous .
the unique balance is determined by @xmath233and the parameters @xmath234 depend on @xmath225 via eq .
( [ 7 ] ) .
since the field is weight - homogeneous ( i.e. , @xmath235 ) , this balance corresponds to an exact , scale - invariant solution of the original system .
note that there is one arbitrary coefficient , and so we expect that one of the k - exponents will be zero . for the vector field
@xmath236 the associated k - matrix , @xmath237 has characteristic equation @xmath238hence the k - exponents are @xmath239we can use the integral constraint ( [ 8 ] ) to get a value for the coefficient @xmath240 .
balancing the terms in ( [ 8 ] ) leads to @xmath241 in the case when all components of the vector @xmath80 are real , the solution @xmath242 of the dynamical system experiences a finite - time singularity .
when @xmath243 , that is when either @xmath244 or @xmath245 , solutions are general ; while , in the range @xmath246 we have only behaviours corresponding to particular solutions of the system .
note that when @xmath247 , we have @xmath248 and we find a behaviour similar to that of dust , but when @xmath249 , @xmath250 the behaviour is that of the standard radiation models .
further , calculating the recursion relations to compute the coefficients of the series expansion term by term , we find the following asymptotic solution for radiation , @xmath251while the dust - dominated expansion is found to be @xmath252 the method that we have described in detail here provides a toolkit for the investigation of the general form of a range of finite - time singularities in general relativistic cosmologies . in particular , the sudden singularities introduced by one of us @xcite , in which @xmath253 and @xmath169 remain finite but @xmath254 and @xmath255 at finite time in situations where no functional relationship is assumed between @xmath256 and @xmath257 , have been widely studied @xcite .
such situations , allow sudden singularities to develop at finite time without violating the strong - energy condition of general relativity and are require less severe conditions than future big rip singularities @xcite with @xmath258 .
elsewhere , we will report on the results of applying these methods to determine the general behaviour on approach to these singularities in general relativity and in higher - order gravity theories .
we thank yvonne choquet - bruhat , alain goriely and antonis tsokaros for useful discussions .
the work of s.c . was partly supported by the ministry of education and religious affairs ( 25% ) and by e.u.(75% ) under the grant `` pythagoras '' .
99 s. w. hawking and g. f. r. ellis , _ the large - scale structure of space - time _ , ( cup , 1973 ) .
y. choquet - bruhat and s. cotsakis , j. geom .
43 ( 2002 ) 345 - 350 ( arxiv : gr - qc/0201057 ) ; see also , y .
choquet - bruhat , s. cotsakis , in _ recent developments in gravity _ ,
proceedings of the 10th hellenic relativity conference , k. d. kokkotas and n. stergioulas eds . , ( world scientific 2003 ) , pp .
145 - 149 .
taylor , _ partial differential equations , basic theory _ , ( springer , 1996 ) , p. 19 .
barrow , class .
quantum gravity * 21 * ( 2004 ) l79 ; j.d .
barrow , class .
quantum grav .
* 21 * ( 2004 ) 5619 ; j.d .
barrow and c.g .
tsagas class .
quantum grav .
* 22 * ( 2005 ) 1563 .
s. cotsakis and i. klaoudatou , j. geom .
@xmath259 ( 2005 ) 306 _ _ ; _ _ p.f . _
_ _ _ gonzlez - daz , phys . rev__. _ _ @xmath260 ( 2003 ) 021303(r ) ; g. calcagni , phys . rev .
d @xmath261 ( 2004 ) 103508 ; v. gorini , a. kamenshchik , u. moschella and v. pasquier , phys .
_ _ _ _ d @xmath261 ( 2004 ) 123512 ; g. kofinas , r. maartens and e. papantonopoulos , jhep @xmath262 ( 2003 ) 066 ; v. sahni and y. shtanov , class .
quantum grav__. _ _ @xmath263 ( 2002 ) l101 ; s. nojiri and s. d. odintsov , phys .
lett__. _ _ b @xmath264 ( 2004)1 ; m. dabrowski , t. stachowiak and m. sydlowski , phys .
d @xmath260 ( 2003 ) 067301 ; p. elizalde and j. quiroga , mod .
lett . a @xmath263 ( 2004 ) 29 ; p.f .
gonzlez - daz , phys .
b @xmath265 ( 2004 ) 1 ; a. feinstein and s. jhingan , _ _ _ _ mod . phys .
_ _ _ _ a _
_ @xmath263 _ _ ( _ _ 2004 ) 457 ; l.p . chimento and r. lazkoz , _ _ _ _ phys .
lett . _ _ @xmath266 ( 2003 ) 211301 ; e. elizalde , s. nojiri and s.d .
odintsov , phys .
d @xmath267 ( 2004 ) 043539 ; l.p .
chimento and r. lazkoz , mod .
_ _ _ _ a @xmath261 ( 2004 ) 123512 . | we define the notion of a finite - time singularity of a vector field and then discuss a technique suitable for the asymptotic analysis of vector fields and their integral curves in the neighborhood of such a singularity . having in mind the application of this method to cosmology
, we also provide an analysis of the time singularities of an isotropic universe filled with a perfect fluid in general relativity . |
recent developments in the design of spintronic devices favor perpendicular magnetization , increasing the interest in materials with perpendicular magnetic anisotropy @xcite .
one advantage is that devices with the same thermal stability can be switched more easily if the magnetization is perpendicular than if it is in plane @xcite .
since magnetostatic interactions favor in - plane magnetization for a thin film geometry , perpendicular magnetic anisotropy requires materials and interfaces that have strong magnetocrystalline anisotropy .
numerous computational studies @xcite show the importance of interfaces on magnetocrystalline anisotropy . the theory developed by bruno @xcite provides an insightful explanation of the surface magnetocrystalline anisotropy originating from spin - orbit coupling and
the theory has been confirmed by experiments @xcite .
the cases for which the bruno s theory does not apply @xcite require a case by case study through first - principles calculations , making it difficult to get much insight .
some insight into perpendicular magnetic anisotropy can be gained by studying it within a simple model .
one such model is the two - dimensional rashba model @xcite .
a two - dimensional rashba model includes only minimal terms imposed by symmetry breaking .
as extensive theoretical studies have shown , a two - dimensional rashba model can capture most of the qualitative physics of spin - orbit coupling with broken inversion symmetry , such as the intrinsic spin hall effect @xcite , the intrinsic anomalous hall effect @xcite , the fieldlike spin - orbit torque @xcite , the dampinglike spin - orbit torque @xcite , the dzyaloshinskii - moriya interaction @xcite , chiral spin motive forces @xcite , and corrections to the magnetic damping @xcite , each of which has received attention because of its relevance for efficient device applications . despite the extensive studies , exploring magnetocrystalline anisotropy within the simple model
is still limited .
magnetocrystalline anisotropy derived from a two - dimensional rashba model may clarify the correlations between it and various physical quantities listed above .
there are recent theoretical and experimental studies on the possible correlation between the magnetic anisotropy and the rashba spin - orbit coupling strength .
the theories @xcite report a simple proportionality relation between perpendicular magnetic anisotropy and square of the rashba spin - orbit coupling strength and argued its connection to the voltage - controlled magnetic anisotropy @xcite .
however , these experiments require further interpretation .
et al . _ @xcite report the positive correlation between the rashba spin - orbit coupling strength and the perpendicular magnetic anisotropy while kim _
et al . _
@xcite report an enhanced perpendicular magnetic anisotropy accompanied by a reduced dzyaloshinskii - moriya interaction in case of ir / co . considering that the dzyaloshinskii - moriya interaction and the rashba spin - orbit coupling are correlated according to ref .
[ ] , the perpendicular magnetic anisotropy and the rashba spin - orbit coupling vary opposite ways in the latter experiment .
these inconsistent observations imply that the correlation is , even if it exists , not a simple proportionality . in such conceptually confusing situations , simple models , like that in this work ,
may provide insight into such complicated behavior . in this paper , we compute the magnetocrystalline anisotropy within a two - dimensional rashba model in order to explore the correlation between the magnetocryatalline anisotropy and the rashba spin - orbit coupling .
we start from rashba models added to different kinetic dispersions ( sec .
[ sec : model ] ) and demonstrate the following core results .
first , a two - dimensional ferromagnetic rashba model with a free electron dispersion results in _ exactly _ zero anisotropy once the fermi level is above a certain threshold value ( sec .
[ sec : result - a ] ) .
this behavior suggests that the simple model is not suitable for studying the magnetic anisotropic energy in that regime .
second , simple modifications of the model do give a finite magnetocrystalline anisotropy proportional to the square of the rashba parameter ( sec .
[ sec : result - b ] ) .
we illustrate with tight - binding hamiltonians that a rashba system acquires perpendicular magnetic anisotropy for most parameter ranges .
this demonstrates that the absence of magnetic anisotropy is a peculiar feature of the former free - electron rashba model and we discuss the similarity of this behavior to the intrinsic spin hall conductivity @xcite .
third , we show that the magnetocrystalline anisotropy of the modified rashba models strongly depends on the band filling ( sec . [
sec : result - b ] ) .
the system has in - plane magnetic anisotropy for low band filling .
as the electronic states are occupied , the anisotropy evolves from in - plane to perpendicular and back to in - plane for high electron density .
this suggests that it may be possible to see such behavior in systems in which the interfacial charge density can be modified , for example through a gate voltage .
we make further remarks in sec .
[ sec : result - c ] and summarize the paper in sec .
[ sec : conclusion ] .
we present the analytic details in appendix .
we first present the model and formalism for a quadratic dispersion and then generalize the model to a tight - binding dispersion . in this paper , we call a rashba model with a quadratic dispersion a free - electron rashba model " and call a rashba model with a tight - binding dispersion a tight - binding rashba model " .
all the models include ferromagnetism in the same manner . , the number of minority electrons per unit area that satisfies @xmath0 .
@xmath1 is given by the area enclosed by the constant energy contour of @xmath2 .
@xmath3 , the number of majority electrons per unit area that satisfies @xmath4 , has the similar meaning ( not shown in the figure).,width=241 ] a ferromagnetic free - electron rashba model is described by the following hamiltonian .
@xmath5 where @xmath6 is the momentum operator of itinerant electrons , @xmath7 is the effective electron mass , @xmath8 is the exchange energy between conduction electrons and the magnetization , @xmath9 is the vector of the pauli spin matrices , @xmath10 is the rashba parameter , @xmath11 is the interface normal direction perpendicular to the two - dimensional space , and @xmath12 is a unit vector along the direction of magnetization .
the terms in eq .
( [ eq : rashba_quadratic ] ) reflect the quadratic kinetic energy , the exchange interaction , and the rashba spin - orbit coupling , respectively .
the second and third term originate respectively from the time - reversal symmetry breaking ( magnetism ) and the space - inversion symmetry breaking ( interface ) .
thus , the rashba model is a minimal model taking account of the symmetry breaking features of the system .
there are various types of rashba models depending on the momentum dependence of spin - orbit coupling hamiltonian @xcite .
we confine the scope of the paper to the linear rashba model that is linear in @xmath6 [ the last term in eq .
( [ eq : rashba_quadratic ] ) ] and is the most widely used form .
diagonalization of eq .
( [ eq : rashba_quadratic ] ) gives the single particle energy spectrum of the free - electron rashba model .
for a homogeneous magnetic texture , @xmath13 commutes with @xmath6 , thus @xmath14 is a good quantum number . in terms of @xmath15 , diagonalization of the @xmath16 hamiltonian gives the energy eigenvalues @xmath17 of @xmath13 for spin majority and minority bands , where @xmath18 and @xmath19 refer to minority and majority bands respectively .
@xmath20 where @xmath21 .
since the system has rotational symmetry around @xmath11 axis @xcite , we assume @xmath22 from now on .
the total electron energy is given by summing up single particle energies at all electronic states below the fermi level .
to do this , we define @xmath23 , the number of minority / majority electrons per unit area that satisfies @xmath24 .
the geometrical meaning of @xmath23 is the area enclosed by the constant energy contour @xmath25 ( fig .
[ fig : electron_density ] ) . with this definition ,
the density of states for each band is given by @xmath26 .
therefore , the expression of the total energy per unit area is given by @xmath27 where @xmath28 is the band bottom energy of each band , below which @xmath29 . @xmath30 if @xmath31 so that there is no occupied minority state , and @xmath32 otherwise .
such a factor is absent for the first term because we only consider the fermi level @xmath33 above @xmath34 . otherwise , the magnetocrystalline anisotropy is trivially zero since there is no occupied state .
the total energy density depends on the direction of magnetization in general .
we then compute the magnetocrystalline anisotropy by the difference of the total energy density for perpendicular and in - plane magnetizations ; @xmath35 . to compute @xmath36 from eq .
( [ eq : total energy ] ) , the fermi levels for @xmath37 and @xmath38 need to be specified . since the energy dispersion [ eq .
( [ eq : quadratic e_pm ] ) ] is in general dependent on @xmath12 , the fermi level also changes as a function of @xmath12 , because the total electron density does not change for an isolated magnetic system .
thus , we fix the total electron density as a constraint . to fix the total electron density as a constraint
, we define the total electron density below energy @xmath39 . @xmath40
the domain of @xmath39 is @xmath41 so that @xmath42 .
since @xmath43 is a strictly increasing function of @xmath39 in the domain , it has an inverse function in @xmath44 .
we denote the inverse function by @xmath45 .
@xmath46 has the same physical meaning as the fermi level @xmath33 for a given electron density @xmath47 .
however , we use the different symbols to emphasize that @xmath46 is given by a _ function _ of the electron density while @xmath33 is just a _ given constant_. with this definitions , the magnetocrystalline anisotropy is given by @xmath48 this is the central equation of the formalism to compute the magnetocrystalline anisotropy .
we now compute the magnetic anisotropy for a tight - binding rashba model . to construct a tight - binding hamiltonian
, we discretize eq .
( [ eq : rashba_quadratic ] ) @xcite . in the main text
, we use a tight - binding hamiltonian for a two - dimensional square lattice as an example .
the construction and the results of a tight - binding hamiltonian for a two - dimensional hexagonal lattice ( equivalently a triangular lattice ) are presented in appendix [ sec(a):hexagonal ] . for simplicity , we use a two - band tight - binding hamiltonian including spin degrees of freedom only , but ignoring all orbital degrees of freedom .
the tight - binding hamiltonian we construct here is given by @xmath49 where @xmath50 , @xmath51 , and @xmath52 are the discretized versions of the kinetic energy , the exchange energy , and the rashba hamiltonian , respectively .
@xmath50 is constructed by the hopping terms to the nearest neightbor sites .
@xmath53 where @xmath54 is the lattice constant , @xmath55 and @xmath56 are the site indicies , and @xmath57 is the electron annihilation operator at site @xmath58 with spin @xmath59 .
. refers to hermitian conjugate of all the terms in front of it .
each term in the summand corresponds to hopping to @xmath60 and @xmath61 directions respectively .
the hopping parameter @xmath62 is determined by matching the energy dispersion with the free electron dispersion @xmath63 in the continuum limit @xmath64 .
@xmath51 is constructed by on - site energy that mixes the spin degree of freedom . @xmath65\cdot{\bm{\mathrm{m}}},\ ] ] where @xmath66 is the matrix element of the pauli matrices .
@xmath52 is constructed as following .
we impose a hopping term from a site to a neighboring site , along a direction @xmath67 .
since @xmath67 corresponds to the electron momentum direction , the term acquires a spin pauli matrix @xmath68 .
then , a hopping term along the @xmath61 direction acquiring @xmath69 is given by @xmath70 , where @xmath71 is a real hopping parameter .
after considering all the neighboring hopping terms satisfying the hermiticity condition , we determine the hopping parameter by taking continuum limit up to @xmath72 and matching the energy dispersion with eq .
( [ eq : quadratic e_pm ] ) . in this way
, we end up with @xmath73+\mathrm{h.c.}.\end{aligned}\ ] ] [ eq : tb hamiltonian ] for more details of determining the hopping parameters , see the example in appendix [ sec(a):hexagonal ] for a two - dimensional hexagonal lattice .
now we use the same formalism [ eq .
( [ eq : fixed electron density ] ) ] .
we use the discrete translational symmetry of the lattice to use the bloch theorem and compute the energy dispersion relation as a function of the crystal momentum .
one difference is that the brillouin zone and the band width for a tight - binding hamiltonian are finite ( fig .
[ fig : bz ] ) , while these are infinite for the free electron model eq .
( [ eq : rashba_quadratic ] ) .
therefore , the domain of the integration in eq .
( [ eq : total energy ] ) is not only limited by the fermi contour , but also limited by the brillouin zone boundary .
we show in sec .
[ sec : result - b ] that the finite band width is a crucial feature for emergence of perpendicular magnetic anisotropy for wide ranges of parameters . ) .
we denote the @xmath74 and @xmath75 points for later purpose . , width=287 ]
although the free electron model we present above [ eq . ( [ eq : rashba_quadratic ] ) ] has a simple form , it still requires complicated mathematics to assess the magnetocrystalline anisotropy predicted by the model since a constant energy contour given by eq .
( [ eq : quadratic e_pm ] ) is a quartic curve . in this section ,
we first discuss results of a perturbative analysis , which assumes @xmath10 to be small and keeps terms only up to @xmath76 . in this regime ,
a constant energy contour is a quadratic curve which allows the magnetocrystalline anisotropy to be calculated analytically .
the analytic results shall give useful insight into the model .
we then go beyond the perturbative regime and discuss exact results in the nonperturbative regime with arbitrary @xmath10 .
in particular , we check if the conclusions from the perturbative analysis remain valid in the nonperturbative regime . as we show in appendix .
[ sec : detail - a ] , expanding eq .
( [ eq : quadratic e_pm ] ) up to @xmath76 , we obtain a quadratic equation with respect to @xmath77 .
the contour @xmath25 forms an ellipse , by which the area enclosed is exactly computable . since
@xmath23 is exactly given in a simple way , calculating eq .
( [ eq : total energy ] ) is straightforward . in this perturbative regime , the relation between electron number density and the fermi level eq .
( [ eq : total number density ] ) is linearly given so inverting @xmath43 is also straightforward .
then , the magnetic anisotropy @xmath78 [ eq . ( [ eq : fixed electron density ] ) ]
is evaluated after simple algebra .
there are two different regimes ; @xmath31 and @xmath79 .
for the first case , there are no minority electrons . for this case , @xmath30 in eq .
( [ eq : total energy ] ) . in the second case
, the minority band is also occupied . for this case , @xmath32 in eq .
( [ eq : total energy ] ) .
we examine the cases one by one . when only majority band is occupied ( @xmath30 ) , the magnetocrystailline anisotropy [ eq . ( [ eq : fixed electron density ] ) ] is @xmath80 here @xmath81 is the electron density when the fermi level touches the bottom of the minority band .
the result shows that the magnetocrystalline anisotropy is at least quadratic in @xmath10 .
below we show this is a result of symmetry that the magnetocrystalline anisotropy should be an even function of @xmath10 .
equation ( [ eq : mca majority only ] ) is valid only when there is no minority electrons @xmath82 .
we show in appendix [ sec : detail - a ] that @xmath83 , which is independent of @xmath12 term here since eq .
( [ eq : mca majority only ] ) is already proportional to @xmath84 . ] . since @xmath85 , eq .
( [ eq : mca majority only ] ) predicts the magnetocryatalline anisotropy to be negative .
the sign corresponds to in - plane magnetic anisotropy , which is counter to the nave expectation that the rashba spin - orbit coupling generates the perpendicular magnetic anisotropy .
however , this observation does not contradict experimental observations showing perpendicular magnetic anisotropy since experimental results are usually obtained when both bands are occupied .
next we examine the second regime where both bands are occupied ( @xmath32 ) .
strikingly , the same formalism leads us @xmath86 regardless of @xmath47 .
there is no magnetocrystalline anisotropy for this case .
an intuitive way to understand this striking behavior is observing the absence of angular dependence of @xmath47 as a function of the fermi level . in appendix [ sec : detail - a ] , we show that , once both bands are occupied , @xmath87 which has no @xmath12 dependence . therefore , when we increase the number of electrons slightly by @xmath88 , the contribution to the additional magnetocrystalline anisotropy is @xmath89dn_e$ ] .
since this is independent of the direction of magnetization , adding electrons does not change the magnetocrystalline anisotropy at all . by noting that eq .
( [ eq : mca majority only ] ) vanishes @xmath90 , we end up with eq .
( [ eq : absence mca ] ) .
there is a recent theory @xcite which predicts perpendicular magnetic anisotropy with the free - electron rashba model . in that work
, the magnetocrystalline anisotropy is expressed by a characteristic energy denoted by @xmath91 . here
we show that @xmath91 takes a value within that model such that the anisotropy is strictly zero . to summarize this section , by using a perturbative approach , we make the following observations .
first , the free - electron rashba model model gives the magnetocrystalline anisotropy that is at least quadratic in @xmath10 .
second , the model does not give perpendicular magnetic anisotropy .
third , the magnetocrystalline anisotropy vanishes unless only a single band is occupied .
we summarize the result in fig .
[ fig : summary_quadratic ] . .
on the other hand , the magnetocrystalline anisotropy vanishes once both bands are partially occupied in the ground state.,width=317 ] so far , we examined the properties of the free - electron rashba model in the perturbative regime .
the perturbative approach allows gaining insight into the model easily but it works only for small @xmath10 . in this section , we go beyond the perturbative regime to see if the conclusions we made in the previous section change when @xmath10 is not small .
we prove that the qualitative results from the perturbative analysis remain valid for large @xmath10 as well .
first we prove that the magnetocrystalline anisotropy is at least quadratic in @xmath10 .
for this , we consider the sign reversal of @xmath10
. this does not affect the energy eigenvalue spectrum of the hamiltonian at all since the energy eigenvalue satisfies the property , @xmath92 [ see eq .
( [ eq : quadratic e_pm ] ) ] .
since the total energy density can not change by a rotational transformation , it should be invariant under @xmath93 .
therefore , the magnetocrystalline anisotropy may be expanded as a power series of @xmath84 with the leading order term proportional to @xmath84 . ] . when @xmath10 becomes larger , higher order terms in @xmath84
can contribute . in fig .
[ fig : quadratic_numerical ] , we numerically compute the magnetocrystalline anisotropy divided by @xmath84 .
we see that the first three curves almost overlap with each other .
however , when @xmath10 becomes larger so @xmath94 is comparable to @xmath95 , the magnetocryatalline anisotropy divided by @xmath84 varies as @xmath10 changes , implying the breakdown of the perturbative result [ eq .
( [ eq : mca majority only ] ) ] .
although the perturbation theory breaks down quantitatively , qualitative features remain the same for a wide range of @xmath10 . in particular , fig .
[ fig : quadratic_numerical ] shows that the magnetocrystalline anisotropy predicted by the free - electron rashba model is negative ( in - plane magnetic anisotropy ) for low electron density and vanishes ( within the numerical error of our calculation ) once the total electron density goes above threshold value .
perpendicular magnetic anisotropy is never generated . within the free - electron rashba model .
the results show in - plane magnetic anisotropy for wide range of the rashba parameter and zero anistropy after a certain threshold within the numerical error .
note that the magnetocrystalline anisotropy is proportional to @xmath84 for small rashba parameters , confirming the result of our perturbation approach .
we use @xmath96 for the simulation.,width=325 ] it turns out that eq .
( [ eq : absence mca ] ) can be rigourously proven for arbitrary @xmath10 .
due to its complexity , here we sketch the proof only briefly .
the detailed proof is presented in appendix [ sec : detail - b ] . the proof proceeds as follows .
first , we consider a situation where both bands are occupied for both @xmath37 and @xmath38 , which occurs if and only if @xmath97 but not for @xmath38 .
this case is not eligible for the proof .
the assumption requires that the minority band should be occupied regardless of the direction of magnetization . for more information ,
see appendix [ sec : detail - b ] . ] .
we then use the cauchy integral formalism for complex contour integrals to show that eq .
( [ eq : ne vs ef ] ) holds beyond the perturbative regime . as discussed in the previous section , eq .
( [ eq : ne vs ef ] ) implies that the magnetocrystilline anisotropy is independent of the fermi level when @xmath98 .
next , we show that @xmath36 vanishes in the large @xmath33 limit .
when combined together these features prove that @xmath36 should be exactly zero for @xmath98 , which is nothing but eq .
( [ eq : absence mca ] ) . here
we emphasize that although eq .
( [ eq : absence mca ] ) holds for arbitrary @xmath10 , it is very unstable with respect to the model variation since it is crucially dependent on @xmath47 being independent of the magnetization @xmath12 [ eq . ( [ eq : ne vs ef ] ) ] , which holds only for the idealized free - electron rashba model [ eq . ( [ eq : rashba_quadratic ] ) ] .
various types of modification of the rashba model which make it more realistic can break this independence and result in the violation of eq .
( [ eq : absence mca ] ) .
possible deformations include the change of dispersion from strictly quadratic and truncation of the infinite band with to finite width . in the next section ,
we consider a tight - binding rashba model , which is more realistic than the idealized free - electron rashba model in the sense that the former has finite band width whereas the latter has infinite band width .
this model shows that eq .
( [ eq : absence mca ] ) is indeed violated and perpendicular magnetic anisotropy emerges . in passing ,
we note that not only the magnetocrystalline anisotropy but also other properties of the idealized free - electron rashba model are peculiar .
a well known example is the intrinsic spin hall conductivity @xcite . for the idealized free - electron rashba model
, it vanishes identically when both bands are partially filled but for slightly modified rashba models @xcite , it is finite .
we consider the tight - binding rashba model for a square lattice . from eq .
( [ eq : tb hamiltonian ] ) , we use the discrete crystal symmetry and the bloch theorem .
we define @xmath99 , where @xmath100 is the total number of sites and @xmath15=@xmath77 is the crystal momentum within the brillouin zone in fig .
[ fig : bz ] , which diagonalizes the hamiltonian .
we define the reduced @xmath16 hamiltonian @xmath101 by @xmath102_{\sigma',\sigma}c_{{\bm{\mathrm{k}}},\sigma}$ ] , where @xmath103_{\sigma',\sigma}$ ] is the matrix element of @xmath101 in the @xmath16 spin space .
since @xmath101 is a @xmath16 matrix , we compute the eigenvalues exactly .
@xmath104^{1/2}.\label{eq : energy dispersion - tight binding}\end{aligned}\ ] ] then , the same formalism eq .
( [ eq : fixed electron density ] ) allows to compute the magnetocrystalline anisotropy . in this section
, we present the results for a two - dimensional square lattice only .
the result for a two - dimensional hexagonal lattice is presented in appendix [ sec(a):hexagonal ] . .
we use the rashba parameters @xmath105nm,@xmath106 nm , and @xmath107 nm .
the results show in - plane magnetic anisotropy ( i m a ) for very low and very high electron occupation . for a wide range of the intermediate electron density ,
it shows perpendicular magneitc anisotropy ( pma ) .
( b ) the peak values of the magnetocrystalline anisotropy as a function of the rashba parameter .
the blue circles are the simulation results and the solid line is a quadratic fitting result .
@xmath108 , @xmath109 , and @xmath110 .
the area of the two - dimensional system in this simulation is @xmath111 where @xmath112 is the length of each direction.,width=325 ] figure [ fig : tb - numerical](a ) shows the relation between the magnetocrystalline anisotropy and the electron density ( normalized to one when both majority and minority bands are completely filled ) . for low electron density ( @xmath113 ) ,
the system acquires in - plane magnetic anisotropy .
this is understandable in that a parabolic approximation of the dispersion relation [ eq .
( [ eq : energy dispersion - tight binding ] ) ] is equivalent to that of the free - electron rashba model [ eq .
( [ eq : quadratic e_pm ] ) ] .
however , as the electron density increases , the parabolic approximation breaks down , thus the system can acquire perpendicular magnetic anisotropy from the point where the effective mass becomes negative ( @xmath114 ) .
after this point , the perpendicular magnetic anisotropy persist widely , until @xmath115 , covering the whole regime where the two spin bands overlap , which is in distinct contrast to the prediction [ eq . ( [ eq : absence mca ] ) ] of the idealized free - electron rashba model . our computation shows a similar behavior to a first - principles calculation @xcite on the band filling dependence of the magnetocrystalline anisotropy .
although a simple rashba model can not be exact , it provides much insight into the system .
changing the electron density by means of substituting atoms or an external voltage can change not only the magnitude of the magnetocrystailline anisotropy but also its sign .
there are two key differences between the tight - binding rashba model and the free - electron rashba model that give rise to finite perpendicular magnetic anisotropy .
the first difference is the deviation of the dispersion from a quadratic .
it allows a nonzero magnetocrystalline anisotropy for a wide range of band filling , due to breakdown of eq .
( [ eq : ne vs ef ] ) .
once the relation between @xmath47 and @xmath33 has a magnetization dependence , a finite magnetocrystalline anisotropy can arise even if both bands are occupied .
the second difference is finiteness of band width ( or brillouin zone ) .
it plays a crucial role for the sign of the magnetocrystalline anisotropy .
since the band width is finite , there must be both maximum ( band top ) and minimum ( band bottom ) energies .
near the band bottom ( the @xmath116 point in fig .
[ fig : bz ] ) , the dispersion is electron - like with a positive effective mass .
thus , the theory in sec .
[ sec : result - a ] is relevant , and the sign of the magnetocrystalline anisotropy corresponds to in - plane magnetic anisotropy for low electron density . on the other hand , near the band top ( the @xmath117 point in fig .
[ fig : bz ] ) , the dispersion is holelike with a negative effective mass .
since the behavior is opposite to the electron - like part , the sign of the magnetocrystalline anisotropy can correspond to perpendicular magnetic anisotropy . as a result
, the magnetocrystalline anisotropy near the band top of the majority band corresponds to perpendicular magnetic anisotropy [ fig .
[ fig : tb - numerical](a ) ] .
figure [ fig : tb - numerical](b ) indicates that the magnetocrystalline anisotropy is proportional to @xmath84 in a reasonable range of @xmath10 .
we argue analytically in sec .
[ sec : result - a ] that the magnetocryatalline anisotropy can be expanded in terms of @xmath84 .
the same argument applies to this tight - binding rashba model .
we discuss below in sec .
[ sec : result - c ] the implication of the sign independence on experimental observation of the correlation between the magnetocrystalline anisotropy and other spin - orbit coupling phenomena . , and 0.9 ev with a fixed rashba parameter @xmath118 nm .
@xmath119 kg , and @xmath120 nm are used .
the stronger the exchange energy is , the higher the magnetocrystalline anisotropy is . on the other hand ,
the weaker the exchange energy is , the wider the range of the electron density that acquires perpendicular magnetic anisotropy . , width=275 ] we now fix the rashba parameter and compute the magnetocrystalline anisotropy for various exchange strengths . figure [ fig : mca_jsd ] shows the result .
the general behaviors discussed above remain the same .
the weaker @xmath95 is , the wider the range of the emergence of perpendicular magnetic anisotropy is .
this is because the band overlap between the majority and minority bands increases as @xmath95 decreases . on the other hand ,
the stronger @xmath95 is , the stronger the magnetocrystalline anisotropy is .
therefore , we conclude that materials with strong @xmath95 are advantageous to achieve a strong magnetocrystalline anisotropy with high controllability under an external voltage .
on the other hand , materials with weak @xmath95 are advantageous for perpendicular magnetic anisotropies that stably exists over a wide range of the electron density .
the mirror symmetry of the magnetocrystalline anisotropy in fig . [ fig : tb - numerical](a ) originates from the symmetry between electrons at the @xmath74 point and holes at the @xmath75 point . from eq .
( [ eq : energy dispersion - tight binding ] ) , the total energy density for completely filled bands is @xmath121d^2k=0 $ ] , thus the magnetocrystalline anisotropy at high electron density can be computed by hole contributions near the @xmath75 point . in other words
, @xmath122 is the same as the contribution from @xmath47 number of holes .
equation ( [ eq : energy dispersion - tight binding ] ) shows the symmetry between the electron - like @xmath74 point and holelike @xmath75 point , @xmath123 , which implies @xmath124 . this is a model - specific property .
for instance ,
in appendix [ sec(a):hexagonal ] , we start from a two - dimensional hexagonal lattice for which the dispersion does not have such symmetry [ eq .
( [ eq(a):hexagonal dispersion ] ) ] and shows that this mirror symmetry around @xmath125 is not general .
there are four kinks in the magnetocrystalline anisotropy in fig .
[ fig : tb - numerical](a ) .
we observe that the two kinks around @xmath126 and @xmath127 correspond to the bottom of the minority band and the top of the majority band , respectively .
since the minority band starts to be occupied from @xmath128 , the behaviors of the magnetocrystalline anisotropy below and above this value are different .
similarly , the majority band is no longer occupied above @xmath127 .
there are two more kinks near @xmath129 and @xmath130 .
we see that these occur near the point where each band are half filled .
near these points , electrons at the fermi level is near inflection points of the energy dispersion so the effective mass changes its sign .
the existence of kinks is quite general as presented in fig .
[ fig : mca_jsd ] and appendix [ sec(a):hexagonal ] . to summarize this section
, we perform tight - binding calculations for the magnetocrystalline anisotropy within a discretized rashba model .
the deviation from a quadratic dispersion allows a nonzero magnetocrystalline anisotropy even when both bands are occupied .
the finite band width allows emergence of perpendicular magnetic anisotropy over a wide range of the total electron density .
the resulting magnetocrystalline anisotropy is proportional to @xmath84 for a reasonable range of @xmath10 .
even though @xmath10 becomes larger than that , the magnetocrystalline anisotropy is independent of the sign of @xmath10 due to symmetry , and is constrained by symmetry to be even powers of @xmath10 .
the implications of the sign independence and comparison with experiments are discussed in the next section .
we perform similar calculations for a two - dimensional hexagonal lattice as well as a square lattice discussed here .
the results are present in appendix [ sec(a):hexagonal ] .
the dependence of the magnetocrystalline anisotropy on @xmath10 differs from the corresponding dependence of many other phenomena of spin - orbit coupling origin . in the previous sections ,
we show by symmetry that the magneocrystalline anisotropy is independent of the sign of @xmath10 . as a result , it is quadratic in @xmath10 for a reasonable range of @xmath10 .
on the other hand , other phenomena of spin - orbit origin such as spin - orbit torque and dzyaloshinskii - moriya interaction have a linear contribution in @xmath10 .
this feature has clear experimental implications . when a magnetic layer has two interfaces with opposite rashba parameters , the total spin - orbit torques and the total dzyaloshinskii - moriya interaction arising from the both interfaces are zero since they are odd in @xmath10 and the contributions from the two interfaces mutually cancel each other .
however , such cancellation does not occur for the magnetocrystalline anisotropy and the contributions from the two interfaces add up since the anisotropy is even in @xmath10 .
a similar phenomenon persists even when only one interface of a magnetic layer is subject to strong inversion asymmetry , if there are multiple energy bands .
it is demonstrated @xcite that multiple bands for a given interface may experience different signs of the rashba spin - orbit coupling . in such a situation , it is possible that contributions of those bands to the magnetocrystalline anisotropy can add up whereas their contributions to the linear spin - orbit phenomena such as spin - orbit torque and dzyaloshinskii - moriya interaction tend to cancel out .
this observation indicates that simple proportionality analysis in experiments may fail to capture the correlation between the magnetocrystalline anisotropy and other phenomena of spin - orbit coupling origin . in this sense , our observation can be consistent with a recent experiment @xcite reporting the opposite behaviors of the dzyaloshinskii - moriya interaction and the perpendicular magnetic anisotropy in ir / co / alo@xmath131 multilayers for various thickness of co. according to the work , the dzyaloshinskii - moriya interaction reduces as the thickness of co increases , while the perpendicular magnetic anisotropy increases .
this difference may originate from multiple origins of the spin - orbit coupling phenomena , such as multiple interfaces and multiple orbital bands . as the thickness of co increases , the contributions to the dzyaloshinskii - moriya interaction may cancel out while those to the magnetocrystalline anisotropy should add up .
one remark is in order .
although our theory demonstrate that the positive correlation between the magnetocrystalline anisotropy and other spin - orbit coupling phenomena may breakdown , it is not necessarily the explanation of the breakdown observed in ref .
[ ] because there are other sources of magnetocrystalline anisotropy .
we observe that the magnetocrystalline anisotropy depends on the total electron density [ fig .
[ fig : tb - numerical](a ) ] and it can even change its sign . the strong dependence of magnetocrystaline anisotropy on the total electron density is another feature that requires a well - controlled experiment to observe the correlation .
when one varies the experimental conditions to obtain systems with various spin - orbit coupling parameters , the total electron density at the interface may change , which disturbs clear interpretation of the dependence of the magnetocrystalline anisotropy on the spin - orbit coupling parameter .
the density - dependent magnetocrystalline anisotropy opens another route of the voltage - controlled magnetic anisotropy @xcite .
the voltage - controlled magnetic anisotropy received considerable attention due to its significant potential to enhance the efficiency of spintronic devices .
there are previous theories @xcite suggesting that modulating the rashba parameter by applying an external voltage is a possible route of the voltage - controlled magnetic anisotropy . however , it is unlikely to be a main mechanism in metallic ferromagnetic films in which a nominal potential gradient is not a main mechanism generating rashba parameters @xcite .
an external electric field is shielded by electron screening in the metal , thus it is difficult to change the rashba parameter significantly . on the other hand , density variations by doping or
an external voltage can change the electron density at the interface , changing the interfacial contributions to the magnetocrystalline anisotropy significantly .
the conclusion from the simple model is consistent with first - principle studies @xcite .
in conclusion , we compute the magnetoctrystalline anisotropy for simple ferromagnetic rashba models .
the properties dramatically change depending on the dispersion relations . for a free electron ( quadratic ) dispersion
, the system does not acquire perpendicular magnetic anisotropy at all .
more interestingly , we analytically show that the magnetocrystalline anisotropy is _ exactly _ zero regardless of the rashba coupling strength if both majority and minority bands are partially occupied in the ground state .
this result is not consistent with experimental observations , implying that a free electron dispersion is not suitable for studying perpendicular magnetic anisotropy arising from the rashba interaction .
we thus generalize the model to have a finite band width , which necessarily generates deviation from the free electron dispersion .
we start from tight - binding hamiltonians and conclude that the system acquires perpendicular magnetic anisotropy over wide range of parameters , consistent with experimental observations .
a finite band width is a crucial feature of the tight - binding hamiltonians that gives rise to perpendicular magnetic anisotropy .
we also observe that the magnetocrystalline anisotropy depends on the band filling and it can even change its sign .
we argue that the interface electron density modulation by voltage is a more important cause of voltage - controlled magnetic anisotropy than the voltage - controlled modulation of the rashba parameter is .
our results show the possibility of breakdown of positive correlation between perpendicular magnetic anisotropy and other spin - orbit coupling phenomena .
in particular , if there are multiple sources of spin - orbit coupling phenomena , such as multiple interfaces and multiple orbital bands , experimental observation of the correlation requires careful analysis .
the authors acknowledge r. d. mcmichael , c .- y .
you , p. m. haney , and j. mcclelland for critical reading of the manuscript .
k.w.k . acknowledges support under the cooperative research agreement between the university of maryland and the national institute of standards and technology , center for nanoscale science and technology , grant no .
70nanb10h193 , through the university of maryland .
was also supported by center for nanoscale science and technology , national institute of standards and technology , based on a collaborative research agreement with basic science research institute , pohang university of science and technology .
k.j.l was supported by the national research foundation of korea ( grant no . 2011 - 028163 , 2015m3d1a1070465 ) .
was supported by the national research foundation of korea ( grant no .
2013r1a2a2a05006237 ) and the ministry of trade , industry and energy of korea ( grant no .
10044723 ) .
the two - dimensional hexagonal lattice we use here is presented in fig .
[ fig(a):hexagonal ] .
we construct a tight - binding hamiltonian by the same way illustrated in sec .
[ sec : model ] .
first we define the electron annihilation operator @xmath57 at the site @xmath132 and spin @xmath59 .
the indices of a site are defined by assigning its position to be @xmath133 where @xmath134 and @xmath135 .
then , the hamiltonian is @xmath49 where @xmath136\cdot{\bm{\mathrm{m}}},\\ h_r&=it_r\sum_{pq\sigma\sigma'}\left[c_{p+1,q,\sigma}^\dagger[({\hat{\bm{\mathrm{u}}}}_1\times{\hat{\bm{\mathrm{z}}}})\cdot{\bm{\mathrm{\sigma}}}]c_{p , q,\sigma}\right.\nonumber\\ & \phantom{it_r\sum_{pq\sigma\sigma'}(}\quad+c_{p , q+1,\sigma}^\dagger[({\hat{\bm{\mathrm{u}}}}_2\times{\hat{\bm{\mathrm{z}}}})\cdot{\bm{\mathrm{\sigma}}}]c_{p , q,\sigma}\nonumber\\ & \phantom{it_r\sum_{pq\sigma\sigma'}(}\quad\left.+c_{p+1,q-1,\sigma}^\dagger[(({\hat{\bm{\mathrm{u}}}}_1-{\hat{\bm{\mathrm{u}}}}_2)\times{\hat{\bm{\mathrm{z}}}})\cdot{\bm{\mathrm{\sigma}}}]c_{p , q,\sigma}\right]+\rm h.c.,\end{aligned}\ ] ] where @xmath137 and @xmath138 are hopping parameters to be determined . by using bloch theorem ,
the hamiltonian can be written by 2@xmath1392 matrix , of which the eigenvalues are exactly given .
@xmath140 is the lattice constant , @xmath55 and @xmath56 are the position indices .
there are two principal directions @xmath141 and @xmath142 .
then , the position vector of each site is @xmath133 .
the labeled sites by @xmath143 , @xmath144 , and @xmath145 are the neighboring hopping sites .
the other three sites are captured by adding hermitian conjugates of these . , width=275 ] the next step is determining @xmath137 and @xmath138 . for
a continuum limit up to @xmath72 , @xmath146 this should coincide with the continuum dispersion eq .
( [ eq : quadratic e_pm ] ) ( up to a constant energy shift ) .
therefore , we obtain @xmath147 and @xmath148 . ) .
the overall features are the same as a two - dimensional square lattice , except the absence of the mirror symmetry around the normalized electron density @xmath125 .
, width=287 ] we now compute the magnetocrystalline anisotropy by the same way .
the result is shown in fig .
[ fig(a):hexagonal_mca ] .
the features discussed in the main text are valid , except the model - specific property of a square lattice that the magnetocrystalline anisotropy is mirror symmetric around @xmath125 .
the aim of this section is to present the mathematical derivations of eqs .
( [ eq : mca majority only])([eq : ne vs ef ] ) from eq .
( [ eq : quadratic e_pm ] ) up to @xmath76 . throughout this section
, we discard all terms beyond @xmath76 .
first we compute @xmath23 from the energy dispersion . expanding eq .
( [ eq : quadratic e_pm ] ) up to @xmath76 , the dispersion relation is approximated by the following quadratic function .
@xmath149 where @xmath150 is defined by @xmath151 , the spin - dependent band shift is given by @xmath152 , and the renormalized masses are @xmath153 then , @xmath154 is given by the area of the contour of @xmath155 in @xmath15 space ( see fig . [
fig : electron_density ] ) .
since @xmath155 forms an ellipse , the area is analytically computable .
@xmath156 from eq .
( [ eq : quadratic e_pm approx ] ) , we obtain the band bottom energies @xmath157 by substituting @xmath158 .
@xmath159 we are now ready to compute eq .
( [ eq : total energy ] ) .
equation ( [ eq : mca majority only ] ) is derived by putting @xmath160 and @xmath30 . then @xmath161 . inverting the function
, we obtain the fermi level as a function of the total electron density @xmath162 . equation ( [ eq : total energy ] ) ( as a function of @xmath47 ) is @xmath163 here , at the first line , we change the variable from @xmath39 to @xmath47 by eq .
( [ eq : npm vs e ] ) . keeping in mind
that the renormalized masses and @xmath34 have angular dependence , we end up with the magnetocrystalline anisotropy @xmath164 where @xmath165 from eqs .
( [ eq : npm vs e ] ) and ( [ eq : eminpm ] ) .
this is eq .
( [ eq : mca majority only ] ) . to derive eq .
( [ eq : absence mca ] ) , we start from taking derivative of eq .
( [ eq : npm vs e ] ) with respect to @xmath39 , @xmath166 therefore , eq . ( [ eq : fixed electron density ] ) for @xmath32 is given by , after some algebra , @xmath167 which is independent of @xmath12 .
we then combine eq .
( [ eq : npm vs e ] ) with eq .
( [ eq : eminpm ] ) to end up with @xmath168=\frac{4\pi m_e(m_e\alpha_r^2+e_f\hbar^2)}{\hbar^4},\ ] ] which is nothing but eq .
( [ eq : ne vs ef ] ) . inverting the function , @xmath169 combining eqs .
( [ eq : e_tot(e_f ) both bands ] ) and ( [ eq : e_f both bands ] ) , @xmath170 has no angular dependence .
thus eq .
( [ eq : fixed electron density ] ) is @xmath171 when both bands are occupied .
this is eq .
( [ eq : absence mca ] ) .
the purpose of this section is to show that eq .
( [ eq : absence mca ] ) holds regardless of how large @xmath10 is .
the flow of the proof is sketched in sec .
[ sec : result - a ] . we first show that i ) eq .
( [ eq : ne vs ef ] ) is exact above the total electron density at which both bands are occupied .
this implies that the magnetocrystalline anisotropy is independent of @xmath33 in this density range , which amounts to @xmath98 .
then we show that ii ) @xmath172 .
we prove this by showing that @xmath36 goes @xmath173 at most for large @xmath33 limit . combining i ) and
ii ) , we end up with the result that the magnetocrystalline anisotropy is exactly zero [ eq .
( [ eq : absence mca ] ) ] .
we prove eq .
( [ eq : ne vs ef ] ) by using the contour integral technique , mainly , the cauchy integral theorem .
we do not assume that @xmath10 is small .
we assume that both bands are occupied for all @xmath12 .
we first prove that this is equivalent to @xmath98 .
to show the forward part of this equivalence , we take @xmath38 . then , @xmath174 , thus @xmath33 should be greater than or equal to @xmath95 for the minority band to be occupied . to prove the backward part
, we assume @xmath98 . for @xmath175 , @xmath176 .
therefore , @xmath175 state is occupied for both bands .
one corollary from the proof is that @xmath175 is always occupied when @xmath98 .
we start from eq .
( [ eq : quadratic e_pm ] ) with @xmath177 for @xmath178 .
we change the variables @xmath77 to a single complex variable @xmath179 . in terms of @xmath180 , @xmath181 where @xmath182 . for a given @xmath33 , @xmath183
is given by the area enclosed by @xmath184 ( fig .
[ fig : electron_density ] ) . by green s theorem ,
the area is given by @xmath185 where @xmath186 is the set of occupied states and @xmath187 is the boundary of @xmath188 , that is , the contour of the fermi level . to perform the integration
, we express @xmath189 as a function of @xmath180 . by equating @xmath184 and solving @xmath189 , @xmath190,\label{eq : zstar}\\ r(z)&=[m_e\alpha_r^2(z - w)+e_fz\hbar^2]^2-z^2\hbar^4(e_f^2-j^2+\alpha_r^2wz).\label{eq : rz}\end{aligned}\ ] ] here
@xmath191 are functions of @xmath180 which satisfy @xmath192 on @xmath193 .
we denote the subscript by @xmath194 or @xmath195 since it is ambiguous which one corresponds to the majority band and the minority band
. however , it does not affect the final result .
the total electron density is then given by @xmath196 the cauchy integral theorem implies that the complex contour integrals in eq .
( [ eq : ne complex expression ] ) is equivalent to those around nonanalytical points only . from eq .
( [ eq : zstar ] ) , there are two types of nonanalytic points of @xmath191 .
the first one is the pole at @xmath197 .
we call this the trivial pole .
we show at the beginning of this section that @xmath198 is occupied for both bands .
that is , the trivial pole @xmath197 is always in @xmath188 ( see fig .
[ fig : contour ] ) .
the second type comes from the square root function .
since the square root function is multivalued in the complex plane , there are branch cuts which connect the branch points that are defined by the zeros of @xmath199 .
the whole branch cuts are nonanalytic points .
thus , it is important to see the behavior of the zeros of @xmath199 .
since @xmath199 is a cubic polynomial , there are three zeros of @xmath199 .
below we present three properties of the three zeros without proofs .
the proofs are presented in appendix [ sec(a):lemma ] .
the first property is that i ) _ all three zeros of @xmath199 are real and nonnegative if @xmath98_. we call the zeros @xmath200 , @xmath201 , and @xmath202 , satisfying @xmath203 .
another important result is that ii ) _
@xmath204 is equivalent to @xmath205_. intuitively , we may say that , if @xmath206 is inside the contour @xmath207 , it is also inside the contour @xmath208 is not simple . ] . since @xmath209 , one direction of
the proof is obvious , but the other direction is not .
the last property is that iii ) _ no or two zeros of @xmath199 are in @xmath188 ( or inside @xmath193)_. as a result , the situation is summarized in fig .
[ fig : contour ] .
we observe that @xmath210 is analytic .
therefore , when we shrink the integral contour by using the cauchy integral theorem , we can end up with the same contour @xmath211 for both terms in eq .
( [ eq : ne complex expression ] ) . by using the cauchy integral theorem , both terms in eq .
( [ eq : ne complex expression ] ) share the same integral contour .
@xmath212 if no zeros of @xmath199 is in @xmath188 , @xmath213 is the only relevant contour .
however , we below show that contributions from @xmath214 are cancelled out when we add up @xmath215 and @xmath216 .
one remark is in order .
the situation becomes complicated if any of @xmath206 is exactly on @xmath193 .
for this case , defining @xmath193 bypassing @xmath206 with an infinitesimally small radius does not change the result .
another resolution is using continuity of @xmath217 . since one of @xmath206 can be exactly on @xmath193 only at particular values of @xmath33 , we may exclude the particular points in the proof and use the continuity to get @xmath217 for the whole domain .
the result greatly simplifies the situation .
the complicated @xmath218 terms in @xmath215 and @xmath216 are cancelled out when they are added up .
@xmath219 which is exactly eq .
( [ eq : ne vs ef ] ) . at the second line
, we use the cauchy s residue theorem .
the importance of the assumption that both bands are occupied in this proof is twofold .
first , the condition is equivalent to @xmath98 so that the zeros of @xmath199 satisfy the properties proven in appendix [ sec(a):lemma ] .
the properties guarantee that the integrands in eq .
( [ eq : ne complex expression ] ) are analytic in @xmath210 so that we can shrink the integral contours for both bands to the same contour .
second and more importantly , the complicated contributions from @xmath220 are cancelled out when we add up the contributions from both bands . therefore , we can use the cauchy s residue theorem for the trivial pole @xmath197 only .
for extremely large @xmath33 , the contour of the fermi level is simple .
therefore , we can define fermi momenta for each band as a function of the azimuthal angle of the momentum .
we write @xmath221 .
then , the fermi momentum @xmath222 is defined by @xmath223 . for simplicity of equations ,
we assume @xmath224 , but the flow of the proof is the same for general @xmath10 . from eq .
( [ eq : quadratic e_pm ] ) and by putting @xmath177 , @xmath225 by using the polar coordinate , the total energy density below the fermi sea is @xmath226 we can expand the integrand with respect to @xmath227 and integrate term by term since @xmath222 is @xmath228 .
after tedious algebra , we end up with @xmath229 therefore , @xmath230 at most , which proves that @xmath172 .
in this section , we prove some important properties of zeros of @xmath199 defined by eq .
( [ eq : rz ] ) .
since @xmath199 is a cubic polynomial , it has three zeros .
we call these @xmath206 for @xmath231 .
we below show that all of @xmath206 are real .
therefore , we can denote @xmath206 by the order of its magnitude @xmath203 .
this section consists of three subsections each of which corresponds to each property that we mention in the main text .
zeros of a cubic polynomial @xmath234 are all real if and only if @xmath235 is nonnegative . after some algebra , @xmath236t,\end{aligned}\ ] ] where @xmath237 , @xmath238 and @xmath239 .
we treat @xmath240 as a function of @xmath71 .
@xmath241 is quadratic and the domain of @xmath71 is @xmath242 .
after some algebra , @xmath243\ge0,\\ \tilde{\delta}_{\rm ext}&=\frac{4}{27}[3e_f^2 - 3j^2+(\alpha+e_f)^2]^3>0,\end{aligned}\ ] ] if @xmath98 .
here @xmath244 is the extremum value of @xmath241 evaluated at the value @xmath71 satisfying @xmath245 . since the boundary values and the extremum value are all nonnegative , @xmath240 ( thus @xmath246 ) is nonnegative on @xmath242 , proving all of @xmath206 are real . to show @xmath247 for all @xmath248 , we see the signs of the coefficients in eq .
( [ eq(a):r coefficients ] ) .
it is easy to see that @xmath249 for any real and positive @xmath180 .
therefore , @xmath199 has no negative real zero .
this statement is equivalent to that any branch point of @xmath191 can not be in @xmath210 .
it is one of the most important properties that allows us to draw fig .
[ fig : contour ] . since @xmath209 , @xmath250 is straightforward , but the other direction is not . to prove this , we use the definition of @xmath188 that @xmath251 is equivalent to @xmath252 .
we start from the following identity .
@xmath253[e_-(z)-e_f]&=\left(\frac{m_e\alpha_r^2}{\hbar^2}\frac{z - w}{z}+e_f-\frac{\hbar^2z^*z}{2m_e}\right)^2\nonumber\\ & \quad-\frac{r(z)}{z^2\hbar^4}.\end{aligned}\ ] ] since @xmath254 , the second term in the right - hand side is zero when @xmath255 .
in addition , we show that @xmath206 should be real in the previous section .
therefore , the first term in the right - hand side is nonnegative when @xmath255 .
@xmath256[e_-(r_i)-e_f]\ge 0.\label{eq(a):e+-ef e -- ef}\ ] ] in the main text , we exclude the case where any @xmath206 is exactly on @xmath193 .
thus , we may assume @xmath257 .
under this assumption , eq .
( [ eq(a):e+-ef e ef ] ) implies that @xmath258 is equivalent to @xmath259 . in other words
, @xmath205 is equivalent to @xmath204 for any @xmath206 satisfying @xmath254 . in @xmath188 satisfying eq .
( [ eq : rz condition ] ) . since @xmath260 , the only possible domains in which @xmath261 can be present are ( a ) @xmath262 and ( b ) @xmath263 .
the number of @xmath206 less than @xmath261 is the number of @xmath206 in @xmath188 .
therefore , either no or two of @xmath206 are in @xmath188 , leading us to fig .
[ fig : contour ] .
, width=287 ] in the previous section , we show that the branch points of the integral eq .
( [ eq : ne complex expression ] ) ( @xmath206 ) are not in @xmath210 .
what is important is not only the branch points but also the branch cuts .
the branch cuts are defined by connecting a pair of branch points ( including the complex infinity if the number of branch points are odd ) . to show that any branch cut does not have
an intersection with @xmath210 , only even number of @xmath206 should be in @xmath188 ( see fig .
[ fig : contour ] ) .
the following lemma is useful for the proof : _
@xmath251 is equivalent to @xmath264_. this lemma is a corollary of the previous section . with this lemma
, we do not need to compute @xmath265 and compare to @xmath33 in order to check @xmath251 .
instead , we only compare @xmath206 to @xmath266 plays a crucial role for deducing this . ] .
therefore , it provides a useful criterion to check @xmath251 .
we first prove @xmath267 . since @xmath251 , @xmath268 , which is the desired result .
we next prove @xmath269 .
note that @xmath270 , thus @xmath204 . in the previous section ,
we show that @xmath204 is equivalent to @xmath205 .
therefore , @xmath251 , which completes the proof . as a result , the statement that we want to show is equivalent to the statement that only even number of @xmath206 satisfy @xmath271 where @xmath272 . " after some algebra , @xmath273 therefore , @xmath274 is positive unless @xmath275 and @xmath276 .
the latter case is not our interest because of the following argument .
note that @xmath277 , thus , @xmath278 is a zero of @xmath199 when @xmath275 .
since @xmath276 , @xmath278 is exactly at the fermi level ( on @xmath193 ) . in the main text
, we exclude this situation . as a result ,
we now have @xmath279 since @xmath199 has three real and nonnegative zeros , there are only two possibilities as presented in figs .
[ fig(a):rz_roots](a ) and [ fig(a):rz_roots](b ) respectively .
figure [ fig(a):rz_roots ] shows that either no or two of @xmath206 satisfy @xmath271 , which is the desired result .
99 p. f. carcia , a. d. meinhaldt , and a. suna , perpendicular magnetic anisotropy in pd / co thin film layered structures , appl .
* 47 * , 178 ( 1985 ) .
j. g. draaisma , w. j. m. de jonge , and f. j. a. den broeder , magnetic interface anisotropy in pd / co and pd / fe multilayers , j. magn .
* 66 * , 351 ( 1987 ) .
s. monso , b. rodmacq , s. auffret , g. casali , f. fettar , b. gilles , b. dieny , and p. boyer , crossover from in - plane to perpendicular anisotropy in pt / cofe / alox sandwiches as a function of al oxidation : a very accurate control of the oxidation of tunnel barriers , appl .
. lett . * 80 * , 4157 ( 2002 ) .
s. ikeda , k. miura , h. yamamoto , k. mizunuma , h. d. gan , m. endo , s. kanai , j. hayakawa , f. matsukura , and h. ohno , a perpendicular - anisotropy cofeb
mgo magnetic tunnel junction , nat .
* 9 * , 721 ( 2010 ) .
w jung , w. kim , t .- d .
lee , k .- j .
lee , and h .- w .
lee , current - induced domain wall motion in a nanowire with perpendicular magnetic anisotropy , appl .
. lett . * 92 * , 202508 ( 2008 ) .
m. nakayama , t. kai , n. shimomura , m. amano , e. kitagawa , t. nagase , m. yoshikawa , t. kishi , s. ikegawa , and h. yoda , spin transfer switching in tbcofe / cofeb / mgo / cofeb / tbcofe magnetic tunnel junctions with perpendicular magnetic anisotropy , j. appl .
* 103 * , 07a710 ( 2008 ) .
o. g. heinonen and d. v. dimitrov , switching - current reduction in perpendicular - anisotropy spin torque magnetic tunnel junctions , j. appl . phys .
* 108 * , 014305 ( 2010 ) .
r. sbiaa , s. y. h. lua , r. law , h. meng , r. lye , and h. k. tan , reduction of switching current by spin transfer torque effect in perpendicular anisotropy magnetoresistive devices ( invited ) , j. appl . phys . *
109 * , 07c707 ( 2011 ) .
d. c. worledge , g. hu , d. w. abraham , j. z. sun , p. l. trouilloud , j. nowak , s. brown , m. c. gaidis , e. j. osullivan , and r. p. robertazzi , spin torque switching of perpendicular tacofebmgo - based magnetic tunnel junctions , appl .
. lett . * 98 * , 022501 ( 2011 ) .
g. h. o. daalderop , p. j. kelly , and f. j. a. den broeder , prediction and confrmation of perpendicular magnetic anisotropy in co / ni multilayers , phys .
lett . * 68 * , 682 ( 1992 ) .
a. sakuma , first principle calculation of the magnetocrystalline anisotropy energy of fept and copt ordered alloys , j. phys .
jpn * 63 * , 3053 ( 1994 ) .
h. x. yang , m.chshiev , b. dieny , j. h. lee , a. manchon and k. h. shin , first - principles investigation of the very large perpendicular magnetic anisotropy at femgo and comgo interfaces , phys .
b * 84 * , 054401 ( 2011 )
. k. h. khoo , g. wu , m. h. jhon , m. tran , f. ernult , k. eason , h. j. choi , and c. k. gan , first - principles study of perpendicular magnetic anisotropy in cofe / mgo and cofe / mg@xmath280b@xmath281o@xmath282 interfaces , phys . rev .
b * 87 * , 174403 ( 2013 ) .
d. odkhuu , s. h. rhim , n. park , and s. c. hong , extremely large perpendicular magnetic anisotropy of an fe(001 ) surface capped by 5d transition metal monolayers : a density functional study , phys .
b * 88 * , 184405 ( 2013 ) .
k. nakamura , t. akiyama , t. ito , m. weinert and a. j. freeman , role of an interfacial feo layer in the electric - field - driven switching of magnetocrystalline anisotropy at the fe / mgo interface , phys .
b * 81 * , 220409 ( 2010 ) .
a. hallal , h. x. yang , b. dieny , and m. chshiev , anatomy of perpendicular magnetic anisotropy in fe / mgo magnetic tunnel junctions : first - principles insight , phys .
b * 88 * , 184423 ( 2013 ) .
p. bruno , tight - binding approach to the orbital magnetic moment and magnetocrystalline anisotropy of transition - metal monolayers , phys .
b * 39 * , 865 ( 1989 ) . p. bruno and j .- p .
renard , magnetic surface anisotropy of transition metal ultrathin films , appl .
a * 49 * , 499 ( 1989 ) .
d. weller , j. sthr , r. nakajima , a. carl , m. g. samant , c. chappert , r. mgy , p. beauvillain , p. veillet , and g. a. held , microscopic origin of magnetic anisotropy in au / co / au probed with x - ray magnetic circular dichroism , phys .
lett . * 75 * , 3752 ( 1995 ) . j. m. shaw , h. t. nembach , and t. j. silva , measurement of orbital asymmetry and strain in co@xmath283fe@xmath284/ni multilayers and alloys : origins of perpendicular anisotropy , phys . rev .
b * 87 * , 054416 ( 2013 ) . c. andersson , b. sanyal , o. eriksson , l. nordstrm , o. karis , and d. arvanitis , influence of ligand states on the relationship between orbital moment and magnetocrystalline anisotropy phys . rev .
99 * , 177207 ( 2007 ) .
y. a. bychkov and e. i. rashba , properties of a 2d electron gas with lifted spectral degeneracy , jetp lett . *
39 * , 78 ( 1984 ) .
j. sinova , d. culcer , q. niu , n. a. sinitsyn , t. jungwirth , and a. h. macdonald , universal intrinsic spin hall effect , phys .
* 92 * , 126603 ( 2004 ) .
j. inoue , g. e. w. bauer , and l. w. molenkamp , suppression of the persistent spin hall current by defect scattering , phys .
b * 70 * , 041303(r ) ( 2004 ) . j. inoue , t. kato , y. ishikawa , h. itoh , g. e. w. bauer , and l. w. molenkamp , vertex corrections to the anomalous hall effect in spin - polarized two - dimensional electron gases with a rashba spin - orbit interaction , phys . rev . lett . * 97 * , 046604 ( 2006 ) . a. manchon and s. zhang , theory of nonequilibrium intrinsic spin torque in a single nanomagnet , phys .
b * 78 * , 212405 ( 2008 ) .
a. matos - abiague and r. l. rodrguez - surez , spin - orbit coupling mediated spin torque in a single ferromagnetic layer , phys .
b * 80 * , 094424 ( 2009 ) .
x. wang and a. manchon , diffusive spin dynamics in ferromagnetic thin films with a rashba interaction , phys . rev .
* 108 * , 117201 ( 2012 ) .
seo , j. ryu , k .- j .
lee , and h .- w .
lee , magnetization dynamics induced by in - plane currents in ultrathin magnetic nanostructures with rashba spin - orbit coupling , phys .
b * 85 * , 180404(r ) ( 2012 ) .
d. a. pesin and a. h. macdonald , quantum kinetic theory of current - induced torques in rashba ferromagnets , phys .
b * 86 * , 014416 ( 2012 ) .
h. kurebayashi , j. sinova , d. fang , a. c. irvine , t. d. skinner , j. wunderlich , v. novk , r. p. campion , b. l. gallagher , e. k. vehstedt , l. p. zrbo , k. vborn , a. j. ferguson , and t. jungwirth , an antidamping spin - orbit torque originating from the berry curvature , nature nanotechnol .
* 9 * , 211 ( 2014 ) .
i. j. dzyaloshinsky , a thermodynamic theory of weak ferromagnetism of antiferromagnetics , phys .
solids * 4 * , 241 ( 1958 ) .
t. moriya , anisotropic superexchange interaction and weak ferromagnetism , phys .
rev . * 120 * , 91 ( 1960 ) .
a. fert and p. m. levy , role of anisotropic exchange interactions in determining the properties of spin - glasses , phys .
rev . lett . * 44 * , 1538 ( 1980 ) .
kim , h .- w .
lee , k .- j .
lee , and m. d. stiles , chirality from interfacial spin - orbit coupling effects in magnetic bilayers , phys .
lett . * 111 * , 216601 ( 2013 ) .
kim , j .- h .
moon , k .- j .
lee , and h .- w .
lee , prediction of giant spin motive force due to rashba spin - orbit coupling , phys .
* 108 * , 217202 ( 2012 ) .
g. tatara , n. nakabayashi , and k .- j .
lee , spin motive force induced by rashba interaction in the strong coupling regime , phys .
b * 87 * , 054403 ( 2013 ) .
s. e. barnes , j. ieda , and s. maekawa , rashba spin - orbit anisotropy and the electric field control of magnetism , sci . rep . * 4 * , 4105 ( 2014 ) . l. xu and s. zhang , electric field control of interface magnetic anisotropy , j. appl .
phys . * 111 * , 07c501 ( 2012 ) .
m. weisheit , s. fhler , a. marty , y. souche , c. poinsignon , and d. givord , electric field - induced modification of magnetism in thin - film ferromagnets , science * 315 * , 349 ( 2007 ) .
g duan , j. p. velev , r. f. sabirianov , z. zhu , j. chu , s. s. jaswal , and e. y. tsymbal , surface magnetoelectric effect in ferromagnetic metal films , phys .
* 101 * , 137201 ( 2008 ) .
t. maruyama , y. shiota , t. nozaki , k. ohta , n. toda , m. mizuguchi , a. a. tulapurkar , t. shinjo , m. shiraishi , s. mizukami , y. ando , and y. suzuki , large voltage - induced magnetic anisotropy change in a few atomic layers of iron , nat .
* 4 * , 158 ( 2009 ) . w .-
wang , m. li , s. hageman , and c. l. chien , electric - field - assisted switching in magnetic tunnel junctions , nat .
mater . * 11 * , 64 ( 2012 ) .
y. shiota , f. bonell , s. miwa , n. mizuochi , t. shinjo , and y. suzuki , opposite signs of voltage - induced perpendicular magnetic anisotropy change in cofebmgo junctions with different underlayers , appl .
. lett . * 103 * , 082410 ( 2013 ) .
l. e. nistor , b. rodmacq , c. ducruet , c. portemont , i. l. prejbeanu , and b. dieny , correlation between perpendicular anisotropy and magnetoresistance in magnetic tunnel junctions , ieee trans . mag , * 46 * , 1412 ( 2010 ) .
kim , j. jung , j. cho , d .- s .
han , y. yin , j .- s .
kim , h. j. m. swagten , and c .- y .
you , interfacial dzyaloshinskii - moriya interaction , surface anisotropy energy , and spin pumping at spin orbit coupled ir / co interface , appl .
. lett . * 108 * , 142406 ( 2016 ) . l. g. gerchikov and a. v. subashiev , spin splitting of size - quantization subbands in asymmetric heterostructures , sov .
. semicond .
* 26 * , 73 ( 1992 ) .
k. garello , i. m. miron , c. o. avci , f. freimuth , y. mokrousov , s. blgel , s. auffret , o. boulle , g. gaudin , and p. gambardella , symmetry and magnitude of spin?orbit torques in ferromagnetic heterostructures , nat . nanotechnol .
* 8 * , 587 ( 2013 ) .
f. mireles and g. kirczenow , ballistic spin - polarized transport and rashba spin precession in semiconductor nanowires , phys .
b * 64 * , 024426 ( 2001 ) .
t. p. pareek and p. bruno , spin coherence in a two - dimensional electron gas with rashba spin - orbit interaction , phys .
b * 65 * , 241305(r ) ( 2002 ) .
s. morakami , absence of vertex correction for the spin hall effect in p - type semiconductors , phys .
b * 69 * , 241202(r ) ( 2004 ) .
k. nomura , j. sinova , n. a. sinitsyn , and a. h. macdonald , dependence of the intrinsic spin - hall effect on spin - orbit interaction character , phys .
b * 72 * , 165316 ( 2005 ) . j
park , c. h. kim , h .- w .
lee , and j. h. han , orbital chirality and rashba interaction in magnetic bands , phys . rev .
b * 87 * , 041301(r ) ( 2013 ) . k. kyuno , j .-
g . ha , r. yamamoto and s. asano , first - principles calculation of the magnetic anisotropy energies of ag / fe(001 ) and au / fe(001 ) multilayers , j. phys .
. jpn . * 65 * , 1334 ( 1996 ) . | we compute the magnetocrystalline anisotropy energy within two - dimensional rashba models . for a ferromagnetic free - electron rashba model ,
the magnetic anisotropy is exactly zero regardless of the strength of the rashba coupling , unless only the lowest band is occupied . for this latter case
, the model predicts in - plane anisotropy . for a more realistic rashba model with finite band width
, the magnetic anisotropy evolves from in - plane to perpendicular and back to in - plane as bands are progressively filled .
this evolution agrees with first - principles calculations on the interfacial anisotropy , suggesting that the rashba model captures energetics leading to anisotropy originating from the interface provided that the model takes account of the finite brillouin zone .
the results show that the electron density modulation by doping or an external voltage is more important for voltage - controlled magnetic anisotropy than the modulation of the rashba parameter . |
a precise measurement of the stellar initial mass function ( imf ) and its functional dependence on environmental conditions would impact astronomy over a wide range of physical scales .
it would be of great help to theorists in untangling the mysteries of star formation and it is a key input in spectral synthesis models used to interpret the observed properties of galaxies both nearby and in the early universe . the current question is whether the imf is _ universal_
the same regardless of time and environmental conditions .
@xcite concisely states the current understanding of imf universality .
it is difficult to believe that the imf is universal given the diversity of galaxy types , environments , star formation rates , and populations within galaxies over the range of observable lookback times . on the other hand , while imf measurements do vary they are all consistent with a universal imf within measurement errors and sampling statistics . the only way to proceed then is to strive for smaller measurement errors and improving sample sizes . a definitive theoretical derivation of the imf does not yet exist .
theoretical approaches to the imf usually center around the jeans mass , @xmath9 , the mass at which a homogeneous gas cloud becomes unstable . at first the collapse of a cloud
is isothermal and the jeans mass decreases which leads to fragmentation of the cloud @xcite .
both @xcite and @xcite suggested that at some point during the cloud collapse the line cooling opacity becomes high enough that the collapse is no longer isothermal . at this point the jeans mass increases and fragmentation stops .
the minimum jeans mass is the smallest fragment size at this point and it provides a lower limit to size of the stars formed .
authors have calculated jeans masses and minimum jeans masses using a variety of methods . in the classical derivation of the jeans
mass @xmath10 .
@xcite finds that @xmath11 where @xmath12 is the mass of gas atoms or molecules and @xmath13 is the opacity at the final fragmentation .
more recently turbulence in clouds has been studied .
@xcite found @xmath14 where @xmath15 is the number density and @xmath16 is the velocity dispersion of the gas .
other investigators have looked at the hierarchical fractal geometry of molecular clouds , thought to arise from turbulence , as a generator of the imf ( e.g. @xcite ) . on a related note @xcite point out that molecular clouds exhibit structure on all resolvable spacial scales suggesting that as no characteristic density exists for the clouds neither does a single jeans mass .
they develop a semi - empirical model for determining the final masses of stars from the initial conditions of molecular clouds without invoking jeans mass arguments and use it to construct imf models .
the key components of their model are sound speed and rotation rate of cloud cores and the idea that stars help determine their final masses through winds and outflows .
but from the beginning the study of the imf has been driven by measurements . in 1955
salpeter was the first to make a measurement of the imf inferring it from his observed stellar luminosity function @xcite .
we parameterize the imf by : @xmath17 following @xcite .
salpeter found that @xmath18 .
it is often overlooked that his measurement only covered masses for which @xmath19 .
nonetheless his original measurement is surprisingly consistent with modern values over a wide range of masses .
the imf in equation [ eq : imfbaldry ] is similar to the salpeter imf for @xmath20 .
the difference is that there are fewer stars with masses less than 0.5 @xmath1 .
we adopt a two part power law as there is agreement amongst several authors that there is a change in the imf slope near 0.5 @xmath21 @xcite .
the technique we will use is not sensitive to the imf at low masses so we assume a constant value in that regime .
salpeter s idea continues to be used today in imf measurements of resolved stellar populations .
the technique can be applied to field stars as well as clusters .
however salpeter s method has several inherent limitations .
the nature of stars presents a challenge . on the high mass end stars are very luminous , but live only a few million years , while on the low end stars are faint but have lifetimes many times longer than the current age of the universe .
there are very few star clusters which are both young and close enough to allow us access to the imf over the full mass range .
in addition , the main sequence mass - luminosity relationship is a function of age , metallicity , and speed of rotation in addition to mass .
it is not yet well - known at the low and high mass extremes .
unresolved binaries can also affect the measured imf @xcite .
the light from unresolved binaries is dominated by the more massive of the pair . as a result
the less massive star is typically not detected which leads to a systematic under - counting of low mass stars .
@xmath22 of main sequence m stars @xcite and 43% of main sequence g stars @xcite are primary stars in multiple star systems .
these are both lower limits as some companions may have eluded detection .
as roughly half of stars are in multiple systems it has a potentially large effect on the observed luminosity function . except for at the high mass end field stars in the solar neighborhood
offer the best statistics for imf measurements .
however the solar neighborhood imf is found to be deficient in massive stars when compared to other galaxies .
for example the @xcite and @xcite solar neighborhood imfs are rejected by integrated light approaches , i.e. @xcite , hereafter k83 , @xcite ( ktc94 ) , and @xcite .
analysis of individual star clusters can be used to detect imf variations . as methods and
data quality can vary between authors comparisons between individual clusters are difficult .
however @xcite studied eight young open clusters with the same technique . on the extremes they measured @xmath23 for ngc 663 and @xmath24 for ngc 581 over masses from around 1 to 12 m@xmath25 .
@xcite considered the case in which one could have perfect knowledge of the masses of all stars in a cluster .
clusters have a finite size so even with no measurement errors uncertainty arises from sampling the underlying imf .
he shows that the observed scatter in imf power law values above 1 m@xmath25 can be accounted for by sampling bias for clusters with between @xmath26 and @xmath27 members ( the @xcite clusters have memberships in this range ) .
furthermore , dynamical evolution of clusters can affect measurements of the imf .
this can happen by preferentially expelling lower mass stars from the cluster and by breaking up binary systems within the cluster .
in total @xcite finds that for stars with @xmath28 the spread in the observed imf power - law slopes in clusters is around 1 when both binary stars and sampling bias are considered even when the underlying imfs are identical .
stochastic processes can also influence the ability to detect systematic imf variations .
o and b stars produce ionizing photons and cosmic rays which affect the surrounding nebula .
however the probability of creating one of these massive stars is comparatively small .
if one of these massive stars happens by chance to form first it may drive up the nebular temperature and depress the formation of less massive stars compared to regions without massive stars @xcite .
an alternative to imf measurements based on the stellar luminosity functions of resolved stellar populations is to infer the characteristics of stellar populations from the integrated light of galaxies using spectral synthesis models .
the advantage of using integrated light techniques is that many of the problems plaguing imf investigations of resolved stellar populations are avoided .
stochastic effects are washed out over a whole galaxy .
unresolved multiple star systems are irrelevant .
the number of observable galaxies is large and their environments span a much larger range than those of milky way clusters .
integrated light techniques can be applied to the high redshift universe .
this creates a strong motivation to develop imf techniques and test them for galaxies in the local universe which can later be used to probe earlier stages of galaxy formation .
the assumption of a universal imf has a huge influence on the interpretation of the high redshift universe .
the reionization of the universe and the madau plot the global star formation rate of the universe over cosmic history are two areas where imf variations could impact the current picture of galaxy formation and evolution . on the downside conclusions are dependent on the stellar evolution models used , which are not well constrained at high masses or with horizontal branch stars at low metallicities .
the biggest problem is that changes in the imf , star formation history ( sfh ) or age of a galaxy model can have similar effects in the resulting spectral energy distribution ( sed ) .
any integrated light technique needs to address these degeneracies .
it is also difficult to probe the imf at sub - solar masses using integrated light . on the level of galaxies the concept of a universal
imf has recently become more complicated .
a number of recent studies have investigated the effect of summing the imf in individual clusters over a galaxy in the presence of power law star cluster mass functions .
@xcite argue that the integrated galaxial imf will appear to vary as function of galactic stellar mass even if the stellar imf is universal .
however @xcite argues that the galaxy wide imf should not differ from the imf in individual clusters based on analytical arguments and monte carlo simulations .
our approach can only measure the imf averaged over whole galaxies and can not address this distinction .
even so systematic variations of any kind from the salpeter slope have not been measured outside of some evidence for non - standard imfs in low surface brightness galaxies ( lsb ) and galaxies experiencing powerful bursts of star formation @xcite .
either way , observational evidence for systematic variations of the imf in galaxies would be highly valuable .
the plan of this paper is as follows : 2 explains our method for constraining the imf . in
3 we describe the sdss data and our sample selection .
4 details our modeling scheme . in 5
we discuss our statistical techniques .
6 reports our results . in 7 we check our results against the h@xmath29 distribution of the data and 8 presents our conclusions .
this paper revisits the `` classic '' method of k83 ( and the subsequent extension ktc94 ) to constrain the imf of integrated stellar populations .
the method takes advantage of the sensitivity of the h@xmath0 equivalent width ( ew ) to the imf .
k83 showed that model imf tracks can be differentiated in the @xmath30 plane .
the total flux of a galaxy at 6565 is the combination of the underlying continuum flux plus the flux contained in the h@xmath0 emission line .
the h@xmath0 flux and the continuum flux have different physical origins , both of which can be used to gain physical insights into galaxies . in the absence of agn activity
the h@xmath0 flux is predominantly caused by massive o and b stars which emit ionizing photons in the ultraviolet .
o and b stars are young and found in the regions of neutral hydrogen in which they formed . in case b recombination , where it is assumed that these clouds are optically thick ,
any emitted lyman photons are immediately reabsorbed .
after several scattering events the lyman photons are converted into lower series photons ( including h@xmath0 ) and two photon emission in the @xmath31 continuum .
these photons experience smaller optical depths and can escape the cloud .
the transition probabilities are weakly dependent on electron density and temperature and can be calculated . through this process
the measured h@xmath0 flux can be converted into the number of o and b stars currently burning in an integrated stellar population .
however case b recombination is an idealized condition and it is possible that ionizing photons can escape the cloud without this processing , a situation known as lyman leakage .
as such the h@xmath0 flux is a lower limit on the number of o and b stars present .
the continuum flux of a galaxy is due to the underlying stellar population . at 6565
the continuum is dominated by red giant stars in the 0.7 - 3 m@xmath25 range while the h@xmath0 flux comes from stars more massive than 10 m@xmath25 .
the ew is defined as the width in angstroms of an imaginary box with a height equal to the continuum flux level surrounding an emission or absorption line which contains an area equal to the area contained in the line .
this is effectively the ratio of the strength of a emission or absorption line to the strength of the continuum at the same wavelength .
given the physical origins of the h@xmath0 flux and the continuum at 6565 the h@xmath0 ew is the ratio of massive o and b stars to stars around a solar mass . therefore the h@xmath0 ew is sensitive to the imf slope above around 1 m@xmath32 and can be used to probe the imf in galaxies . as mentioned in the introduction several degeneracies plague the study of the imf from the integrated light properties of galaxies .
variations in the imf , age , metallicity , and sfh of galaxy models can all yield similar effects in the resulting spectra .
for example , increasing the fraction of massive stars , reducing the age of a galaxy , lowering the metallicity , and a recent increase in the star formation rate will all make a galaxy bluer .
metallicity effects were not discussed in either k83 or ktc94 and galaxy ages were assumed , 15 gyr for k83 and 10 gyr for ktc94 .
the sfh in k83 is addressed by calculating models with exponentially decreasing sfhs for a range of e - folding times , as well as a constant and a linearly increasing sfh . in the @xmath30 plane
the effect of varying the sfh e - folding time is orthogonal to imf variations .
however this is only true for smoothly varying exponential and linear sfhs .
discontinuities , either increases ( bursts ) or decreases ( gasps ) , in the star formation rate can affect the h@xmath0 ew relative to the color in ways similar to a change in the imf
. along with the exponential sfhs ktc94 also uses models with instantaneous bursts on top of constant sfhs .
however this was done to access high ews at an age of 10 gyr rather than to fully flesh out the effects of sfh discontinuities .
the assumption of smoothly varying sfhs is a key assumption in our analysis .
most late - type galaxies are thought to form stars at a fairly steady rate over much of recent time although bursts of star formation may play a significant role in low mass galaxies @xcite . for these galaxies
smoothly varying sfhs are justified .
however there are other galaxies clearly in the midst of a strong burst of star formation ( e.g. m82 ) and dwarf galaxies with complex sfhs , e.g. ngc 1569 @xcite , for which this assumption is a poor one .
the effects of violations of our sfh assumptions are described in detail in the results section .
k83 cites four major sources of error all of which are improved upon or eliminated by the high , uniform quality of sdss spectroscopic and photometric data .
the h@xmath0 fluxes in his sample can be contaminated by nonthermal nuclear emission . in the updated investigation , ktc94
, this problem is addressed by removing objects with known seyfert or liner activity and luminous agn . the sdss spectra allow measurements of emission line ratios which can be used to separate star forming galaxies from agn @xcite .
the second problem is that the h@xmath0 emission flux will be underestimated if the underlying stellar absorption of h@xmath0 is not taken into consideration .
the narrow band filter photometry of @xcite could not measure this effect for individual galaxies so a fixed ratio was assumed for all galaxies .
the sdss spectroscopic pipeline does not take this into account either .
we use the h@xmath0 fluxes measured from the sdss spectra by @xcite which fit the continua with stellar population models to more accurately measure the h@xmath0 emission . while the sdss pipeline method is sufficient for strong emission lines the h@xmath0 absorption ew can be as large as 5 which is significant for weaker emission lines .
thirdly , their narrow band h@xmath0 imaging includes [ ] emission which is corrected for by assuming a constant [ ] /h@xmath0 ratio .
the h@xmath0 and [ ] emission lines are resolved in the sdss spectra so there is no need for a correction .
this is a significant improvement .
@xcite found from a literature survey that the mean value of the h@xmath0/(h@xmath0 + [ ] ) ratio is @xmath33 for spiral galaxies and @xmath34 in irregular galaxies .
these mean corrections were applied uniformly to the k83 data . in ktc94
a uniform correction was applied using [ ] /h@xmath0 = 0.5 . for comparison in our sample the mean value of
the h@xmath0/(h@xmath0 + [ ] ) ratio is a strikingly similar 0.752 and the mean [ ] /h@xmath0 ratio is 0.340 .
however [ ] /h@xmath0 ranges from 0.0 to 0.6 .
if our h@xmath0 and [ ] lines were blended , applying a fixed correction would introduce errors of as much as 25% in the h@xmath0 ews of individual galaxies .
lastly , in both k83 and ktc94 extinction corrections were addressed by plotting data alongside models which were either assumed an average value for the extinction or were unextincted .
the balmer decrement ( h@xmath0/h@xmath35 ) can be measured from sdss spectra which allows for extinction corrections for individual galaxies .
another major advantage of this study is that the sample size is much larger than that of k83 and ktc94 .
the ktc94 sample contains 210 galaxies , whereas ours has @xmath36 .
the large sample size allows us to investigate imf trends as functions of galaxy luminosity , redshift and aperture fraction with subsamples larger than the entire ktc94 sample .
there is a key disadvantage to this method as well .
k83 and ktc94 were able to adjust the sizes of their photometric apertures to contain the entire disk of individual galaxies to a limiting isophote of 25@xmath37 given by the rc2 catalog @xcite .
the advantage of narrow band measurements of h@xmath0 ew is that they can cover a much larger aperture and match the broadband measurements set to match the physical size of individual galaxies .
the sdss has fixed 3 `` spectroscopic apertures .
this problem is partly offset by using matching 3 '' photometry apertures from the sdss .
however this introduces aperture effects as observed galaxies have a wide range of angular sizes due to the range of physical sizes and distances present in the local universe .
this is significant as radial metallicity gradients ( e.g. @xcite ) are observed in spiral galaxies which could incorrectly be interpreted as radial imf gradients .
even so , this method can constrain the imf within the sdss apertures . for our program galaxies 23% of the total
light falls in the sdss apertures .
the fact that the more distant galaxies are more luminous and tend to be larger helps to balance out the larger physical scales of the fixed aperture size at greater distances . on average
17% of the light falls in the aperture for the faintest galaxies while it is 25% for the brightest bin . in spite of their limited size
the sdss apertures still contain a great diversity of stellar populations which make this data set an excellent test bed for imf universality .
we will present extensive tests of aperture effects below .
the sample is selected from sloan digital sky survey data .
the project goal of the sdss is to image a quarter of the sky in five optical bands with a dedicated 2.5 m telescope @xcite . from the imaging
@xmath38 galaxies and @xmath39 quasars will be selected for spectroscopic followup .
photometry is done in the @xmath40 filter system described by @xcite .
magnitudes are on the ` arcsinh ` system @xcite which approaches the ab system with increasing brightness .
spectra are taken with a multi - object fiber spectrograph with wavelength coverage from 3800 to 9200 and @xmath41 @xcite .
our sample is a sub - sample of the main galaxy sample from sdss dr4 @xcite .
the main galaxy sample ( mgs ) targets galaxies with @xmath42 in petrosian magnitudes @xcite .
all galaxies in the mgs are strong detections so the differences between luptitudes and the ab system can be ignored . in order to avoid fiber crosstalk in the camera an upper brightness limit of @xmath43 , @xmath44 and @xmath45 is imposed .
targets are selected as galaxies from the imaging by comparing their psf magnitudes to their de vaccouleur s and exponential profile magnitudes .
exposure times for spectroscopy are set so that the cumulative median signal - to - noise satisfies @xmath46 at @xmath47 and @xmath48 fiber magnitudes .
the time to achieve this depends on observing conditions but always involves at minimum three 15 minute exposures . due to the construction of the spectrograph fibers can not be placed closer than 55 " to each other .
this may be a source of bias in the sample .
cluster galaxies may be preferentially excluded from the sample .
the sdss collaboration has plans to quantify this effect in the near future @xcite .
lsbs are excluded from the mgs with a surface brightness cut which may also bias the sample @xcite .
there is some evidence that lsbs may have imfs which differ from a universal imf .
@xcite find that the comparatively high mass - to - light ratios of lsbs can be explained with an imf deficient in massive stars relative to normal galaxies .
our sample begins with the fourth data release ( dr4 ) of the sdss @xcite .
dr4 covers 4783 deg@xmath49 in spectroscopy for a total of 673,280 spectra , 567,486 of which are galaxy spectra .
429,748 of these have flags set indicating they are part of the mgs .
dr4 also includes special spectroscopic observations of the southern stripe which were not selected by the standard algorithm but which nonetheless have the ` target_galaxy ` flag set in ` primtarget ` which usually indicates membership in the mgs .
these objects are identified by comparing their spectroscopic plate number to the list of special plates in @xcite and rejected .
this leaves 423,285 spectra . the first round of cuts to our sample address general data quality .
while the overall quality of sdss data is high there are a handful of objects with pathological values for one or more parameters .
first we require that all parameters of interest have reasonable , real values .
this means that the petrosian and fiber magnitudes must be between 0 and 25 and have errors smaller than 0.5 in all five @xmath40 bands .
line flux errors are capped at @xmath50 @xmath51 and equivalent widths at @xmath39 . for most of the parameters
less than 1% of objects fail this loose requirement .
however 6.6% of the objects fail the petrosian magnitude error requirement .
this is most likely due to the photometric pipeline having a difficult time defining the petrosian radius .
as such this constraint is potentially biased against lsbs or galaxies with unusual morphologies . in defense of this cut
we later bin our data by luminosity and aperture fraction both of which are determined in part by the petrosian magnitudes and also by @xmath52-corrections determined from them .
in addition , limiting flux errors to 50% is hardly unreasonable .
altogether 391,160 galaxies pass these combined requirements , which is 92.4% of the mgs .
galaxies from photometry run 1659 are removed because of a known problem with the photometry .
this excludes 4,485 galaxies ( 1.1% ) from a continuous strip on the sky and should not be a source of bias .
we place a further constraint on the @xmath53 band fiber magnitude requiring that the error be less than 0.15 .
the @xmath53 band generally suffers from the most noise so this requirement ensures that the fiber magnitude quality is good enough to minimize the chance of erroneous @xmath52-corrections which could affect our colors .
only 2,866 ( 0.7% ) mgs galaxies fail this test . combining the general data quality requirements , the run 1659 rejection and the fiber @xmath53 band error limit leaves 386,647 galaxies ( 91.3% of the mgs ) . the next round of cuts to our sample , while necessary , have clear astrophysical implications .
many of the objects in the mgs have agn components .
as we are interested in studying only the underlying stellar populations of these objects agn must be removed .
this is done using the classical @xcite diagram comparing the logarithms of the [ @xmath54/h@xmath35 and [ @xmath55/h@xmath0 emission line ratios .
we used the criterion of @xcite where objects for which @xmath56/\rm h\beta ) > \frac{0.61}{\log([\rm{n\ ii}]/h\alpha)-0.05}+1.3 \label{eq : kbpt}\ ] ] are classified as agn and rejected .
following @xcite we require the s / n of the h@xmath0 , h@xmath35 , [ ] and [ ] lines to be at least 3 to properly classify a galaxy as a star forming one .
131,807 galaxies ( 34.1% of mgs objects surviving our first round of cuts ) survive this cut .
the above cut automatically rejects any galaxies with weak [ ] and [ ] lines .
this excludes galaxies with weak star formation .
to a lesser extent metal poor galaxies which also have weak [ ] emission are rejected as well .
@xcite also define a low s / n star forming class of galaxies which we identify and keep in our sample .
these are the galaxies which have not already been classified as star forming or agn by strong lines and equation [ eq : kbpt ] , nor have been identified as low s / n agn by [ @xmath55/h@xmath0 @xmath57 0.6 with s /
n @xmath57 3 in both lines , yet still have h@xmath0 with s /
n at least 2 . 79,548 ( 20.6% ) of the sample
falls into the low s / n star forming galaxy category . combining the two classes 211,355 ( 54.7% ) of the galaxies survive the agn cut .
the agn cut also has a strong luminosity bias for two reasons .
galaxies with agn components tend to be brighter .
the bimodal distribution of galaxies in color - magnitude space @xcite also plays a role .
the most luminous galaxies are predominantly red with minimal star formation and thus weak emission lines .
luminous galaxies are rejected both for having agn components and for having low s / n emission lines . over 95% of galaxies fainter than @xmath58
meet this criteria , but by @xmath59 the fraction is only 38.9% . a color bias is also introduced by the agn cut . over 95% of galaxies
bluer than @xmath60 pass , which drops to 24% by @xmath61 .
this is mainly due to the s / n requirement for the emission lines .
redder galaxies tend to have weak emission lines and are rejected .
the balmer decrement is used for the extinction correction so h@xmath0 and h@xmath35 s / n are required to be at least 5 to reduce errors .
214,912 galaxies ( 55.6% ) have h@xmath35 s / n @xmath62 which is the more restrictive of the two criteria .
this cut has a clear luminosity bias .
roughly 85% of galaxies with @xmath63 satisfy this requirement , but this fraction decreases with increasing luminosity until only 12.9% survive at @xmath64 .
this is again due to the bimodal distribution of galaxies .
the luminous red galaxies with weak emission lines are rejected .
there is also a color bias .
over 93% of the bluest galaxies blueward of @xmath65 survive the cut while only 23% of the reddest pass this requirement .
as previously mentioned , by nature the reddest galaxies have weak balmer lines as a result of their low sfrs and are preferentially rejected .
a redshift cut of @xmath66 is applied to ensure that peculiar velocities do not dominate at low redshift and to limit the range of galaxy ages .
384,349 galaxies ( 99.4% ) meet this criteria .
nearly all galaxies from @xmath67 to @xmath68 survive this cut . on the low luminosity end only 27.4% of @xmath69 galaxies are distant enough to pass and on the high end 93.1% of @xmath70 galaxies are close enough to survive .
only 63% of the bluest galaxies pass .
many of the blue galaxies that fail are actually regions of local group galaxies which are treated as their own objects by the sdss pipeline so removing them actually improves the integrity of our sample .
the stellar populations of galactic bulges can be significantly different from those in the spiral arms .
the sdss fibers have a fixed aperture of 3 so over the large range of luminosities and distances in the mgs aperture affects can become very important . to remove outliers
we require that at least 10% of the light from a galaxy falls within the spectroscopic aperture .
this is done by comparing the petrosian magnitude to a fixed 3 aperture fiber magnitude .
both of these quantities are calculated for all objects in the sdss by the photometric pipeline .
371,777 ( 96.2% ) galaxies survive this cut .
this cut rejects proportionally more faint galaxies ; 99.2% pass at @xmath70 compared to 50.9% at @xmath69 .
the aperture cut has low sensitivity to color .
the intersection of the agn , balmer line s / n , redshift and aperture fraction cuts leaves 140,598 galaxies 36.4% of the high quality mgs data defined by our first round of cuts and 33.2% of the mgs as a whole . at this point
three final cuts are applied to the sample .
one galaxy is removed for surviving all criteria , but having a negative h@xmath0 ew .
the extinction of individual galaxies is estimated using the balmer decrement ( h@xmath0/h@xmath35 emission flux ratio ) for each galaxy . given case b recombination , a gas temperature of 10,000k and density of 100 @xmath71 the balmer decrement
is predicted to be 2.86 @xcite .
this ratio is weakly dependent on nebular temperature and density .
@xcite lists values down to 2.74 for case b recombination in environments where both the temperature and electron density are high .
3.2% of the galaxies suffer from the problem that the balmer decrement is less than 2.86 and 2.1% have a balmer decrement below 2.74 .
538 galaxies ( 0.4% ) have balmer decrements more than 3@xmath72 below 2.74 , which is around 6 times more than expected .
this is does not suggest a problem with the case b assumption as the predicted balmer decrements for case a recombination are nearly identical in each temperature regime . to understand the reason behind this problem around 100 of the offending spectra
were inspected revealing a few different causes for the problem . around 80 galaxies are at redshifts where the telluric @xmath73 line affects the measurement of h@xmath35 .
many of these galaxies have very strong emission lines with extremely weak stellar absorption .
using the sdss pipeline values instead of the @xcite values yields acceptable balmer decrements .
the rest are galaxies with low flux where the balmer lines are in absorption .
this shows there are a few cases where attempting to fit the underlying stellar absorption lines fails and this failure is not reflected in the error values .
these galaxies are rejected without any apparent introduction of bias .
all of these cuts combined leave 140,060 galaxies , which is 33% of the mgs and 36% of the high quality mgs data . after removing duplicate observations
there are 130,602 galaxies in our sample . of these objects
1.7% overlap with the luminous red galaxy sample .
overall the bulk of the galaxies are removed by agn rejection and the h@xmath35 s / n requirement , with the rest of the cuts having little effect .
both of these cuts are necessary .
agn must be removed to ensure that the h@xmath0 emission represents the underlying stellar population and not an accretion disk .
our method requires that the galaxies have measurable balmer emission lines .
this coupled with the need for accurate extinction corrections justifies the balmer line s / n requirement .
our cuts bias our sample by preferentially excluding galaxies at both luminosity and both color extremes .
the faintest galaxies are most affected by the redshift and aperture cuts while the luminous galaxies succumb to the h@xmath35 s / n requirement . at the red extremes it is the h@xmath35 s / n and agn requirements that play equally large roles , while the bluest objects are primarily rejected by the hubble flow redshift requirement .
although our sample is biased by our cuts each one is a necessary evil .
we do not attempt to correct this bias , but we remind the reader that the following results are only representative of actively star forming galaxies without any agn activity . the aim of this paper , however , is to test imf _
universality_. if the imf is truly universal it should be universal in any subsample of galaxies .
the fact that our sample is slightly biased with respect to luminosity and color is not a significant barrier to achieving our goal .
the sdss includes a number of different calculated magnitudes .
we use the _ fiber _ magnitudes which are 3 fixed aperture magnitudes .
although it was not the case in earlier versions of the photometric pipeline , fiber magnitudes are now seeing corrected @xcite .
the fiber magnitudes were not originally intended for science purposes but rather to get an idea of how bright an object will appear in the spectrograph .
we use the fiber magnitudes to reduce the aperture effects arising from comparing a 3 spectroscopic aperture to petrosian magnitudes .
originally the sdss used `` smear '' exposures to correct spectra for light falling outside the 3 aperture due seeing , guiding errors and atmospheric refraction @xcite .
the smear technique was later found to be an improvement only for high s / n point sources and its use was discontinued @xcite .
the spectra here are not seeing corrected . after paring the sample to its final size a number of corrections
must be made to both the photometric and spectroscopic data .
galactic reddening from the milky way must be corrected for .
sdss database includes the @xcite dust map values for each photometry object .
the extinction of individual galaxies is estimated using the balmer decrement for each galaxy .
the data is corrected assuming that 2.86 is the true value of the balmer decrement using the milky way dust models of @xcite .
the assumption of milky way dust is not significant as models of the dust attenuation in the milky way , smc and lmc are nearly identical in the @xmath74 band and redward .
as aforementioned a few percent of our galaxies have balmer decrements below 2.86 .
our solution is to set the emission line extinction to @xmath75 magnitudes for these galaxies . massive young stars and their surrounding ionized nebulae tend to be embedded in their star forming regions more so than older , lower mass stars which have had time to migrate from their birth regions .
as such nebular emission lines will experience more extinction than the continuum .
@xcite find the ratio of emission to continuum line extinction is @xmath76 .
we assume this value to be 2.0 and correct the continuum and emission lines separately .
this is the same value used by k83 and ktc94 in their extinction corrected models .
we note that the spatial geometry of the dust can influence the extinction law , but this complication is beyond the scope of this paper .
galaxy photometry is @xmath52-corrected to @xmath77 using version 4.1.4 of the code of @xcite .
this redshift is roughly the median of the sample and is selected to minimize errors introduced by the @xmath52-corrections .
the @xmath78 colors we use are the @xmath79 colors we would observe if the galaxies were all located at @xmath80 .
stated explicitly the @xmath78 color is @xmath81 where @xmath82 and @xmath83 are @xmath52-corrections , @xmath84 and @xmath85 are milky way reddening values , and 1.153 and 0.834 relate the v band extinction to the @xmath74 and @xmath86 bands assuming a milky way dust model . the corrected equivalent width is obtained as follows @xmath87^{-1 } \label{eq : width}\ ] ] where @xmath88 is the measured , uncorrected equivalent width .
the @xmath89 arises from the fact that the total flux in the h@xmath0 line is not affected by cosmological expansion of the universe but the flux per unit wavelength of the continuum is depressed by a factor of @xmath89 .
the following term is the extinction correction .
the 0.775 relates the v band extinction to the h@xmath0 line assuming a milky way dust model and the @xmath90 is due to the fact that the emission line and continuum experience different amounts of extinction as previously explained .
figure [ fig : grha_data ] shows the distribution of the galaxies in the color vs. h@xmath0 ew plane . in order to conduct a likelihood analysis
we need error estimates which take into account both the errors induced by the photometry and spectroscopy and those by the aforementioned corrections .
the error in the corrected color , @xmath91 , is given by @xmath92 where @xmath93 and @xmath94 are the poisson errors in the observed @xmath74 and @xmath86 band photometry , @xmath95 is the ratio of the emission line to continuum extinction , and @xmath96 is the error introduced by the @xmath52-corrections .
the terms inside the square root in equation [ eq : sigmac ] are obtained through a straight error propagation of equation [ eq : color ] .
the 0.03 outside the square root is due to the systematic zero point errors of the sdss filter system .
following @xcite @xmath95 is fixed at 2.0 and @xmath97 is set to 0.4 . the error in emission line @xmath98
is given by @xmath99 which is dependent on the fractional uncertainty of the h@xmath0 and h@xmath35 emission line fluxes .
@xcite reports a 16% error in their milky way reddening values . because a dust model is assumed reddening in the @xmath74 and @xmath86 bands are linearly related , so the error in @xmath100 is a function of @xmath84 .
this relationship combined with the 16% error yields the 0.0440 in equation [ eq : sigmac ]
. the median value of @xmath74 band reddening is 0.10 so the errors introduced by the mw reddening correction are insignificant for the majority of objects .
errors in the redshift determination are negligible , typically 0.01% .
the value of @xmath96 is estimated to be 0.02 by visual inspection of a plot of @xmath101 as a function of redshift . for a typical galaxy
the term involving @xmath97 is the largest contributor to the extinction corrected color error .
the median poisson error from the photometry is 0.01 in both bands .
the median value of @xmath91 is 0.085 .
the error in the corrected h@xmath0 equivalent width , @xmath102 , is given by @xmath103 \label{eq : sigmaw}\ ] ] equation [ eq : sigmaw ] is the result of propagating the errors in equation [ eq : width ] , neglecting the insignificant redshift errors .
again , the term involving @xmath97 is the largest contributor to the error for typical galaxies .
the median error in the equivalent width is 17% .
the median error bars for the sample are shown in figure [ fig : grha_data ] . for comparison ,
k83 reports equivalent width errors of around 10% , but the extinction is uncertain at the 20 - 30% level .
model galaxy spectra were calculated using the publicly available pegase.2 spectral synthesis code @xcite .
models are calculated for ages from 1 myr to 13 gyr .
25 smoothly varying sfhs generated by analytic formulae are considered . the sfhs range from 19 exponentially decaying sfhs with time constants from 1.1 to 35 gyr , a constant sfr , and four increasing sfhs which are proportional to @xmath104 where @xmath105 is the time constant .
the precise values of the time constants were selected to smoothly sample the h@xmath0 ew vs. @xmath78 plane .
the metallicity of the stars is assumed to be constant with respect to time and calculated for @xmath106 = 0.005 , 0.010 , 0.020 and 0.025 .
galactic winds , galactic infall and dust extinction are turned off .
the dust extinction is not modeled because we have applied an extinction correction to the data .
nebular emission is calculated from the strength of the lyman continuum .
emission line ratios are fixed .
the model spectra are redshifted to @xmath80 to match the redshift range of the data .
the model parameter of interest is the imf .
implicit in equation [ eq : imfbaldry ] is the assumption that the imf does not vary as a function of time .
the continua of galaxies are weakly influenced by low mass stars in the optical .
this method is sensitive to the imf for masses above around 1 m@xmath25 so the slope is fixed below 0.5 m@xmath25 . above 0.5 m@xmath25 models
are calculated for @xmath107 , where @xmath108 is incremented by 0.05 between models .
we treat our imf model as though it has only one degree of freedom @xmath108 above 0.5 m@xmath25 . in truth
it has three more as written : the lower and upper mass cutoffs and the point at which the slope changes .
so how well are our assumptions justified ?
we have parameterized the imf as a piecewise power law with two components .
piecewise power laws are motivated by empirical fits to data starting with @xcite , which had only one component . by contrast
the power law formulation of the @xcite imf has 24 components .
the log - normal distribution is a more physical choice as it can arise from stochastic processes .
@xcite were the first to fit an observational measurement of the imf with a log - normal distribution .
the log - normal distribution is normalizable as it goes to zero smoothly at both extremes without any awkward truncation .
its main drawback is that it can not fit any structure in the imf over small mass ranges .
log - normal distributions have three degrees of freedom .
this is less than the four our model has .
however , we are not sensitive to imf over the full range of masses which makes it much more difficult to fit the parameters of the log - normal distribution .
instead we use this piecewise model and lower the degrees of freedom through physical arguments .
many investigators find a change in slope in the imf around 0.5 m@xmath25 .
our fixed lower end of the imf is designed to be consistent to this .
as our technique is not sensitive to this regime this assumption does not impact the results .
the imf must be normalizable because the total mass of a stellar population is finite . in our parameterization
this is achieved by truncation at 0.1 and 120 m@xmath25 .
this seems unphysical as the existence of brown dwarfs suggests that the imf should continue below the hydrogen burning limit .
however , stars at 0.1 m@xmath25 do not contribute much to the integrated light of galaxies .
as we are not sensitive to stars in this mass range this choice is not unreasonable .
in fact , truncating the imf at 0.5 m@xmath25 yields models which are at worse differ by 0.002 in @xmath78 and 3% in h@xmath0 ew from those with low mass stars .
the slope below 0.5 m@xmath25 has essentially no effect on our results .
only when the imf is truncated at 0.9 m@xmath25 do the models differ at the level of the errors in the data .
on the high mass end the choice of limit does matter .
there is a physical upper limit to the size of stars associated with the eddington limit .
the value of this theoretical limit is not widely agreed upon .
the largest stellar mass measured reliably , via analysis of a binary system , is @xmath109 m@xmath25 @xcite .
@xcite argue that given the large mass and youth of the star forming cluster r136 in the large magellanic cloud stars in excess of 750 m@xmath25 should be present given a salpeter imf with no upper mass limit to stars , whereas no stars above 150 m@xmath25 are observed .
an analysis of the arches cluster , the youngest observable cluster , gives an upper limit of 150 m@xmath25 based on monte carlo simulations although stars above 130 m@xmath25 are not detected @xcite .
the pegase model tracks only extend up to 120 m@xmath25 so this is the cutoff used .
another issue is that the physics and evolution of such high mass stars is not well known so the models themselves may be a significant source of error in this regime . the left half of figure [ fig : oddimf ] shows the effects of varying the high mass cutoff in the imf .
the effect in the @xmath78-h@xmath0 ew plane is seen to be very similar to increasing the value of @xmath108 .
lowering the upper mass cutoff from 120 to 90 m@xmath25 has roughly the same effect as increasing @xmath108 from 1.35 to 1.45 .
the relationship between the change in the upper mass cutoff ( from 120 m@xmath25 ) and the apparent change in @xmath108 is @xmath110 and is roughly linear for upper mass cutoffs down to 50 m@xmath25 .
the coefficient in the relationship is a week function of the age of the population ranging from 0.004 for 13 gyr old populations to 0.006 for 100 myr old populations .
the right half of figure [ fig : oddimf ] shows the affect of adding a second break in the imf at 10 m@xmath25 . reducing the value of @xmath108 over the 0.5 - 10 m@xmath25 range
while keeping it fixed at @xmath111 above 10 m@xmath25 has a similar effect to decreasing @xmath108 in a two component model .
this illustrates one of the limitations of this model . at this point
it is not possible to detect fine structure in the imf slope or to state precise values for the imf slope . in this limited space of observables
the imf models themselves are degenerate .
what it does provide is a framework with which to detect variations in the imf .
although we can construct similar tracks from different imf models , we can still detect the differences between two groups of galaxies .
while we will report our results as a function of @xmath108 it must always be kept in mind that it is degenerate with the upper mass cutoff and other fine structure in the imf at high stellar masses .
the assumption of smoothly varying sfhs is of great consequence . in the event that a galaxy is experiencing or
has recently experienced a burst our sfh assumption can lead to measured @xmath108 values that are off by as much as 0.5 .
the effects of bursts are more closely examined in a later section . within the assumption of smoothly varying sfhs
much can be said about the effects of the imf , metallicity , age , and sfh in the color - h@xmath0 ew plane .
figure [ fig : models ] demonstrates these relationships . in both panels
the ages of the models decrease from the upper left to lower right .
the effects of the age of a stellar population are largely orthogonal to those of imf variations . in figure
[ fig : models]a the effects of changing the functional form of the smoothly varying sfh with fixed metallicity and @xmath108 are shown .
sfh variation is degenerate with the imf .
however the effect is relatively small over a wide range of sfhs .
the solid lines are exponentially decreasing sfhs with @xmath112 gyr where the bulk of the star formation occurs early in the galaxy s life .
the dashed lines have sfhs that are increasing with time where most star formation occurs at late ages .
the effect of variations in the form of smooth sfhs is larger at later ages and higher values of @xmath108 but does not dominate the effects of imf variations . with all other parameters fixed the range of smooth sfhs cause systematic uncertainties at the level of @xmath4 in @xmath108 . in figure
[ fig : models]b the effects of metallicity variations with fixed sfh and @xmath108 are shown .
the metallicity variations are also degenerate with imf variations . with all other parameters fixed between colors of @xmath113 the systematic uncertainty due
to metallicity is less than 0.05 in @xmath108 .
this uncertainty increases to 0.35 at @xmath114 and 0.7 . as aforementioned
, the extinction correction is another potential problem .
the arrows in figure [ fig : models ] show the length and direction of the extinction correction for typical galaxies in our sample . it is assumed that @xmath115 , but @xmath116 and @xmath117
are also plotted to show the potential effect of variations in @xmath95 .
the reddening vectors for @xmath115 and 4 are fortuitously orthogonal to the imf variations . only when the continuum and emission extinctions are equal , when @xmath116 , do the extinction correction and variations in @xmath95 become a larger concern than metallicity and sfh
. however such low @xmath95 ratios are not observed in galaxies ( @xcite and section [ sec : fratio ] ) .
figure [ fig : all135 ] shows all 18,480 model points with @xmath111 .
models are interpolated in sfh history for fuller coverage of the color - h@xmath0 ew plane .
for each value of @xmath108 the models cover a stripe rather than a single line .
it can be seen in the lower right of figure [ fig : all135 ] that the model become degenerate in @xmath108 for old , red galaxies with weak current star formation .
the data and models are compared using a pseudo-@xmath118 " minimization . for various reasons ( detailed below )
the classical @xmath118 estimator assumptions are violated so we can not use traditional tables for error estimates but we can still use the @xmath118 as a statistical estimator as long as the confidence regions are calibrated by monte carlo ( mc ) techniques as we will do .
we proceed as follows : for each galaxy @xmath119 we have a measured @xmath78 color , @xmath120 , and an h@xmath0 ew , @xmath121 , and measurement errors @xmath122 and @xmath123 , given by equations [ eq : sigmac ] and [ eq : sigmaw ] .
we also have model values @xmath124 and @xmath125 for a range of imf slopes @xmath108 , metallicities @xmath106 , ages @xmath126 and sfhs @xmath127 .
we can then construct a @xmath118 value as @xmath128 which is calculated by brute force .
the goal of this paper however is to investigate the imf with relatively simple measurements of the h@xmath0 ew and a broadband color . while making crude measurements of the mean stellar metallicities of individual galaxies is possible , disentangling age and sfh effects on an individual basis is a daunting task . assuming that it is possible to do , it does not scale up well to the high redshift universe where observations will be of lower quality .
it does not make sense to minimize @xmath118 over all galaxies for a particular set of @xmath129 because we have no a priori reason to think that all of the galaxies should have the same metallicity or sfh .
in fact we expect that they would not .
the solution is to marginalize @xmath118 over metallicity , age and sfh for each galaxy such that @xmath130 _
{ z , t,\psi } \label{eq : chisquare2}\ ] ] this is somewhat unorthodox , because for some galaxies the data points are over - fitted , i.e. there will be a stripe in ( @xmath131 ) space corresponding to a given @xmath108 and we can get @xmath132 values very close to zero ( but not exactly because of the discrete nature of the model grid ) for galaxies within the stripe .
we note we also have partial degeneracy between parameters such as age and metallicity they both shift the tracks in similar directions largely orthogonal to @xmath108 ( though not completely which is why we have a stripe in parameter space not a line ) .
this makes it difficult to calculate the traditional number of degrees of freedom . "
despite these limitations it is clear that galaxies inconsistent with a particular @xmath108 will _ still _ have large values of @xmath132 for example a very blue galaxy with a low equivalent width in figure [ fig : all135 ] .
the complication is that the stripes for similar @xmath108 values overlap , and for red galaxies with low equivalent widths the stripes for vastly different @xmath108 values overlap .
as such , the imf for an individual galaxy is only broadly constrained .
measuring a precise best imf for an individual galaxy boils down to random chance and the discrete nature of the models .
however , by summing @xmath133 over many galaxies the imf is narrowly constrained for the sample being summed over as long as we are careful in our confidence region analysis .
because of this over - fitting and partial degeneracy we can not apply the textbook notions of the @xmath118 distribution , calculate degrees of freedom and choose @xmath134 contours for different confidence regions .
further to this @xmath120 and @xmath121 are not truly independent variables .
both the colors and ews are subject to the same extinction and reddening corrections which tie the errors together . for galaxies with @xmath135
the h@xmath0 line is in the observed @xmath86 band , although this only affects a relatively small number of galaxies in the sample almost all of which have @xmath136 .
also the direct statistical interpretation of @xmath118 is predicated on the assumption of normal errors .
equations [ eq : sigmac ] , [ eq : sigmaav ] and [ eq : sigmaw ] reveal that our errors are complicated mixtures of individual measurements which are most likely poisson distributed .
thus @xmath91 and @xmath102 are unlikely to be normally distributed .
bursty sfhs can potentially create outliers which are statistically significant due to the fact that neither @xmath91 or @xmath102 include a term for this difficult to quantify effect .
the problem is even worse if the errors are non - symmetric which could potentially arise from the aforementioned bursty sfhs . in the case of non - symmetric errors the best value of @xmath108
could erroneously be pulled away from the true value .
given all this we abandon the direct statistical interpretation of @xmath118 and regard it as an estimator of the goodness of fit whose confidence regions have to be calibrated empirically .
we do this via mc simulations ( as recommended by @xcite ) where we simulate data points for a given @xmath108 with the correct error distributions and propagate everything through the analysis in the same way as for the actual data . for each of our mc simulations we add poisson errors to the @xmath40 fiber magnitudes , h@xmath0 and h@xmath35 fluxes and the observed h@xmath0 ew .
we assume that these observed quantities have poisson dominated errors as members of the mgs they are high s / n measurements .
the entire analysis described above is repeated , including a new extinction measurement and a recalculation of the @xmath52-corrections . for each value of @xmath108
we run 100 mc simulations to estimate the 95% confidence interval .
setting up the mc architecture in this way has the further advantage that we can use the same machinery to test the effect of systematic errors such as the violation of our smooth sfh assumptions , as we will do later .
the main downside of course is that this approach is computationally intense .
run times for the 100 mc simulations are typically 18 days on a 2 ghz desktop pc for the samples considered here . in practice
it turns out that @xmath137 is still a smooth well - behaved function with , not surprisingly , a quadratic minimum which has the advantage that we can then interpolate it to increase the resolution in the best - fitting @xmath108 without incurring the additional computational expense .
this arises of course from the fact that our estimator is similar to a traditional @xmath118 and is a good reason to stick with this similarity over some more exotic goodness of fit measure .
an estimate of the systematic errors is discussed later .
regardless of how poorly a sample is modeled by a universal imf the above method will still find a best fitting @xmath108 and corresponding confidence region .
we still expect @xmath118 to be small for a model that is a good fit and large for one that is not .
one nuance in comparing @xmath118 between different sub - samples , as we will do , is that the samples are often of considerably different sizes . because of this we choose instead to use the _ mean _ @xmath118 , @xmath138 instead , as a sample metric
this has the advantage that absolute @xmath138 values and confidence regions are more similar between the sub - samples , though we note that the confidence regions on @xmath138 are still determined directly from our mc simulations .
figure [ fig : all - like ] shows the results of our analysis for the full sample of galaxies using just the observed data set .
the `` x '' marks the best fitting imf , where @xmath139 and @xmath140 with @xmath141 . at @xmath142 @xmath143 while steeper imfs are more heavily rejected with @xmath144 at @xmath145 . for comparison several `` classic '' imfs
are also plotted in figure [ fig : all - like ] at their approximate equivalent values of @xmath108 . with @xmath146
the @xcite solar neighborhood imf is a particularly bad fit .
two more recent solar neighborhood imfs , @xcite and @xcite , yield @xmath147 and 1.98 respectively .
these results reinforce the conclusions of k83 , ktc94 and @xcite that the solar neighborhood is not representative of galaxies on the whole as far as the imf is concerned .
on the other hand the @xcite imf , established from a review of star cluster imf studies in the literature , is a better fit than the best @xmath108 value in our parameterization with @xmath148 .
this result highlights the degeneracy of the imf models themselves in the color - h@xmath0 ew plane two considerably different imfs ( one with one break and the other with two ) fit nearly equally as well .
the results of the mc simulation show that the 95% confidence region is @xmath149 for the data set as a whole .
the mc simulation shows that additional data will not improve the overall results as the random errors are already small . clearly and not surprisingly systematic errors , which are discussed later , dominate .
the overall result of @xmath150 is steeper than the original salpeter value of @xmath111 .
it is also steeper than the @xcite value of @xmath151 derived from galaxy luminosity densities in the uv to nir .
it is however well within their 95% confidence limit of @xmath152 as well as their measurement of @xmath153 based on the h@xmath0 luminosity density .
the difference between their two results suggests that the h@xmath0 and mid - uv to optical fluxes may have different sensitivities to massive stars .
@xcite estimated the uncertainty in @xmath108 , either due to measurement uncertainties , real imf variations or both , in his star cluster imf based on the spread of results in the literature .
our result is well within his range of uncertainty in both mass regimes @xmath154 for @xmath155 and @xmath156 for @xmath157 .
the luminosity of a galaxy could potentially have an effect on the imf within it .
for one the ambient radiation field is likely higher in more luminous galaxies .
figure [ fig : allmagr ] shows the best fitting imf and @xmath138 values as a function of @xmath158 for all 130,602 galaxies .
the galaxies have been binned in @xmath158 such that there are 500 objects in each bin .
the bin size was chosen to maximize coverage in @xmath158 yet still keep the random errors in each bin small .
the solid lines represent the lower and upper 95% confidence region determined from the mc simulation .
figure [ fig : allmagr ] reveals a constant value of @xmath7 for galaxies with @xmath158 between @xmath159 and @xmath160 with linear increases in @xmath108 for both brighter and fainter galaxies .
there is also a sudden downturn in @xmath108 values for galaxies fainter than @xmath161 .
given the sizes of the random errors the differences in @xmath108 between @xmath162 and @xmath160 are substantial , from 1.59 to 1.41 , and statistically significant .
the agreement with the salpeter slope is the best for galaxies between @xmath159 and @xmath160 in @xmath158 . in many ways
it is not surprising that previous investigators have not found this trend .
the milky way is thought to have a luminosity of @xmath163 @xcite ; the @xmath164 and @xmath165 filter curves cover roughly the same wavelengths . at comparable luminosities
our results are similar to salpeter .
the galaxies in the k83 sample have a median @xmath166 with only 16% ( 18 objects ) fainter than @xmath167 @xcite .
this is a significant bias toward more luminous galaxies where our results are in agreement with a universal imf .
by contrast 30% of our sample is fainter than @xmath168 @xcite .
we have a sample of 39,350 galaxies fainter than @xmath169 .
the lower panel of figure [ fig : allmagr ] shows that the relative quality of the fits rapidly deteriorates as the luminosity of the galaxies decrease . for the brightest galaxies @xmath138 floats around 0.15 , while in the faintest bin it is over 6 . for comparison @xmath170 for the large magellanic cloud and @xmath171 for the small magellanic cloud @xcite
this trend could indicate that a universal imf is a good fit to the most luminous galaxies , but dwarf galaxies can not be described by a universal imf , even if a different universal slope is allowed .
however it could have a more mundane explanation .
it could be that errors are over or underestimated as a function of luminosity .
it also could arise from deviations from our assumption of smoothly varying sfhs .
we can not bin our data by stellar mass without assuming an imf which is contrary to the goals of the project .
we can repeat the analysis of figure [ fig : allmagr ] using @xmath172 in the place of @xmath158 .
@xmath172 , being redder , is a better proxy to stellar mass .
the resulting plot is nearly identical to figure [ fig : allmagr ] which shows that the relationship persists across several wavebands .
figure [ fig : allmagr ] reveals a clear , statistically significant trend in @xmath108 and @xmath138 with respect to luminosity .
the rest of this section focuses whether this trend is a manifestation of true imf variations or if it is the result of sample biases or poor assumptions .
if the imf is truly universal and our method successfully probes the imf any subsample of galaxies that we could choose should yield the same @xmath173 value as any other in spite of any selection biases or aperture effects .
figure [ fig : allmagr ] clearly shows that the preceding statement is false . in this section
we set aside the possibility of imf variations and search for biases in our sample .
figure [ fig : allmagr ] shows that the overall result of @xmath174 is really a weighted average .
the sdss mgs is a magnitude limited sample , one defined by flux limits , with both upper and lower limits .
table [ tab : lumbins ] gives the number of objects in each luminosity bin .
there are 41,411 galaxies with @xmath175 but only 28 for which @xmath176 .
as such the overall result is heavily biased by more luminous galaxies .
malmquist bias will affect any magnitude limited sample . because brighter objects can be seen at greater distances a magnitude limited sample contains bright objects from a greater volume of space than fainter objects .
the result is that the ratio objects by luminosity in a magnitude limited sample differs from the true ratio in nature ; brighter objects are over - represented . to eliminate malmquist bias volume limited bins where subsamples are complete for a range of luminosities are constructed .
figure [ fig : vl - def ] details the construction of these bins .
given both the upper and lower flux limits of the mgs ( @xmath177 ) only a factor of 13 in luminosity falls in the sample at any given redshift .
the redshift limits of each volume limited bin are defined such that no galaxies within the magnitude limits of the bin are affected by the flux limits of the mgs . within each box in figure
[ fig : vl - def ] the true ratio of galaxy luminosities is preserved and is thus free of malmquist bias .
figure [ fig : vl - plot ] shows the results for volume limited magnitude bins .
most error bars are smaller than the plotting symbols due to the larger number of galaxies , 329 to 29,701 as given in table [ tab : lumbins ] , in each bin . figures [ fig : allmagr ] and [ fig : vl - plot ] show the exact same trends . the largest difference in @xmath108 between the whole and volume limited samples is 0.0116 in the @xmath67 bin .
the other notable difference is that the fainter galaxies have larger @xmath138 values in the volume limited case .
however malmquist bias across bins is not responsible for the luminosity trends in figure [ fig : allmagr ] .
another effect of magnitude limited samples is that the faintest galaxies are much closer than the most luminous ones .
the mean redshifts of the volume limited magnitude bins range from @xmath178 for @xmath67 to @xmath179 for @xmath180 .
this corresponds to a difference in age of around 1.8 gyr . as aforementioned
, model tracks reveal that age is largely orthogonal to the imf in our parameter space , but there could be other effects tied to age and distance .
in addition the imf could evolve with time .
the large number of galaxies in our sample affords us the luxury of investigating the effects of luminosity and redshift simultaneously to obtain a better understanding of what role , if any , the redshift plays in our analysis .
figure [ fig : redshift ] shows our fitted parameters for all 130,602 galaxies in bins that are 0.25 magnitudes wide in luminosity and 0.005 wide in redshift .
the upper left panel shows the best fitting @xmath108 for each two dimensional bin . on the upper right the width of the 95% confidence region in @xmath108 for each bin
is shown . at bottom left
is the @xmath181 and at bottom right is the log of the number of galaxies in each bin .
the white contour demarcates the region in which each bin contains at least 50 galaxies .
the number 50 is arbitrary but it shows the region where poisson errors are expected to be small .
the black areas are regions where there are no galaxies with the given parameters . using the plot of @xmath108 at the upper left we can look for potential redshift biases .
this is complicated by the fact that at a fixed luminosity there is a limit to the range of redshifts in the sample due to the flux limits of the sample described earlier .
looking at vertical slices through the plot at any fixed luminosity there is a trend towards larger values of @xmath108 with increasing redshift .
however , for horizontal slices of fixed redshift the same relationship between @xmath108 and luminosity that is present for the whole sample is seen modulo a normalization factor .
the right half of figure [ fig : redshift ] shows a strong relationship between the number of galaxies per bin and the width of the 95% confidence region in @xmath108 .
this simply reflects the fact that larger samples are less affected by poisson errors .
the lower left panel of figure [ fig : redshift ] provides an excellent example of why our metric of fit quality , @xmath138 , is so important .
bins with similar numbers of galaxies and @xmath182 values can have vastly different values of @xmath138 .
it is worth reminding that the contours in @xmath138 are logarithmic . at fixed luminosity
the galaxies are better fit by a universal imf at higher redshift .
similar to the sample as a whole the quality of fit improves with luminosity .
while there does appear to be some weak trending of @xmath108 and @xmath138 with redshift , redshift effects are not driving the relationship seen between imf and luminosity as it persists at fixed redshifts .
one explanation for the trend in @xmath108 with redshift is aperture effects .
again , if the imf is truly universal aperture effects should not exist .
the sdss spectra have a fixed aperture of 3 for all galaxies . depending on the angular extent and distance to a galaxy
a different fraction of the total light of the galaxy will fall into the aperture .
the problem is mitigated by the fact that the most distant galaxies are the most luminous and more likely to have a larger physical size . as the physical area contained in the aperture increases with distance ,
so too does the size of the galaxies being observed .
however , the two effects do not exactly balance out .
table [ tab : lumbins ] shows that the mean aperture fraction for the @xmath183 bin is 0.20 and increases to 0.27 at @xmath184 . on average
35% more of the most luminous galaxies fall within the aperture compared to the faintest .
figure [ fig : aperture ] shows the behavior of our fitted parameters for two dimensional bins of luminosity and aperture fraction in the same manner as figure [ fig : redshift ] did for luminosity and redshift . for fixed luminosities increasing aperture fraction leads to decreasing values of @xmath108 .
however at fixed aperture fraction the qualitative imf - luminosity trend remains .
the @xmath138 values are a strong function of luminosity , but @xmath138 does increase with aperture fraction at each fixed luminosity .
the trend with aperture fraction is the exact opposite of what would be expected in the presence of a systematic effect operating given the redshift result in figure [ fig : redshift ] .
the nearest galaxies should have the smallest aperture fractions in a particular luminosity bin .
the nearest galaxies in figure [ fig : redshift ] have the smallest values of @xmath108 while the smallest aperture fractions in figure [ fig : aperture ] have the largest values of @xmath108 .
figure [ fig : aperture ] suggests that the measured imf is more dependent on the aperture fraction than the redshift .
there are several possible physical explanations for imf trends with the aperture fraction , all of which are related to radial gradients in disk galaxies .
@xcite make the theory based claim that the imf should be a function of the original local temperature of the star - forming molecular clouds .
metallicity gradients are also known to exist in disk galaxies , including the milky way @xcite .
@xcite measure a linear , radial light metal ( c , o , mg & si ) abundance gradient of @xmath185 in the disk of the milky way .
given the increased efficiency of cooling with metal lines we would expect the most low mass stars where metallicity is the highest on average towards the center of galaxies .
the trend in @xmath108 in figure [ fig : aperture ] is qualitatively consistent with this idea . if there are radial imf gradients in galaxies one would expect the fits to decrease in quality with increasing aperture fraction .
a blend of imfs will not be fit as well as a universal one given our technique .
this idea is consistent with the results in figure [ fig : aperture ] .
however , in well - resolved stellar populations there is no evidence for a relationship between the imf and metallicity , except perhaps at masses lower than those probed by our method @xcite .
if metallicity plays a role in determining the imf the effects are only being revealed as a global trend in our large sample of integrated stellar populations . for individual clusters metallicity must play a secondary role to stochastic effects .
the extinction correction is another potential source of bias .
it is possible that there is a second order correction that our fairly simple extinction correction fails to take into account .
this could potentially lead to an erroneous imf trend with extinction correction .
this affects the luminosity results because more luminous galaxies tend to be dustier , as evidenced in table [ tab : lumbins ] .
the problem is further complicated by the fact that dust is thought to play an integral part in star formation so it is not unreasonable that an observed imf trend with extinction may be real .
figure [ fig : extinction ] , similar to figures [ fig : redshift ] and [ fig : aperture ] , shows the results of our analysis for two dimensional bins of luminosity and the extinction correction that was measured and applied .
vertical slices through the upper left panel of figure [ fig : extinction ] show that @xmath108 does depend on @xmath186 , trending towards lower @xmath108 values with increased extinction over the region where the poisson error in @xmath108 is reasonable . yet
again , horizontal cuts of fixed extinction show the imf - luminosity relationship .
the decreasing @xmath108 values with increasing extinction is counter - intuitive .
dustier regions tend to be more metal rich .
if metal cooling plays a significant role in the imf the dustiest regions should have the steepest imfs . at fixed luminosity
@xmath138 increases with extinction .
as aforementioned our calculated errors in color and ew ( equations [ eq : sigmac ] and [ eq : sigmaw ] ) have a functional dependence on the observed emission line extinction . in both cases
it is the term proportional to @xmath187 which is on average the major contributor to the calculated error .
because luminous galaxies tend to be more heavily extincted they will also be more likely to have larger errors .
this is potentially problematic for our observed imf trend with luminosity . if we are unknowingly underestimating the errors for faint galaxies with low extinction the source of the poor fit qualities of these galaxies
could be systematic instead of astrophysical .
however the lower left panel of figure [ fig : extinction ] shows that the most extincted galaxies have the poorest fits where such a bias would suggest that they should fit the best due to the large accommodating errors .
it is also possible that biases in our @xmath108 measurements could depend on two parameters simultaneously .
figure [ fig : avap ] shows the measured galaxy of @xmath108 as a function of both aperture fraction and measured emission line extinction for six volume limited luminosity bins . when holding all other parameters fixed , increasing the aperture fraction leads to lower values of @xmath108 in all statistically significant areas of figure [ fig : avap ] .
this is the same relationship found in the earlier section on aperture fraction . decreasing values of @xmath108
are also seen for increasing extinction when all other parameters are constant . a notable exception to this
is that galaxies with large extinction and small aperture fractions favor higher @xmath108 values .
most importantly when looking at a particular combination of aperture fraction and extinction the imf becomes shallower with increasing luminosity until the highest luminosities where it becomes steeper again . even in the narrowest slices of the data
set the same imf - luminosity trend is seen , albeit with slightly different absolute values of @xmath108 . as discussed
previously we have allowed two classes of star forming galaxies into our sample .
111,806 galaxies ( 86% ) fall in the star forming class and the other 18,796 ( 14% ) belong to the low s / n star forming class where the or lines are weak , but the h@xmath0 and h@xmath35 lines still have s / n @xmath188 . by comparing the results from these two subsamples we can investigate a possible bias of the results with respect to the level of star formation .
figure [ fig : sfclass ] shows the results for both classes as a function of luminosity . as it comprises 86% of the total sample it is not surprising that the results for the star forming class are similar to those of the sample as a whole in figure [ fig : allmagr ] .
the low s / n class exhibits a similar qualitative behavior to the set as a whole with a few notable differences .
the @xmath108 values are offset by at least 0.08 towards larger @xmath108 .
the measured imf turns toward steeper values at lower luminosities than for the sample as a whole .
the @xmath138 are several times lower as well . while the galaxies in low s / n class meet the same requirement of s / n @xmath57 5 in the balmer lines as the star forming class they are biased towards noisier h@xmath0 line measurements .
this corresponds to lines that are either weak ( low sfr ) or weak compared to the continuum ( low present sfr compared to the past ) both of which lead to low ews .
another issue at play is that the relationship between the h@xmath0 line flux and the sfr is dependent on the imf . at a fixed metallicity and sfr increasing @xmath108 by 0.05 reduces the h@xmath0 flux by 20% .
in fact the h@xmath0 flux of a galaxy with @xmath142 will be 33 times larger than a galaxy with the same sfr and @xmath145 . in the presence of real imf variations at any fixed luminosity
the low s / n class will be biased towards galaxies with steeper imfs . both low sfrs and steep imfs potentially lead to low s / n h@xmath0 flux .
however it is difficult to determine the level of influence of each effect .
as shown in figure [ fig : grha_data ] low h@xmath0 ews lead to larger values of @xmath108 for any fixed color .
as the low s / n class tends toward noisier h@xmath0 fluxes and therefore ews it is easy to see from equations [ eq : sigmac ] , [ eq : sigmaav ] and [ eq : sigmaw ] that the errors for this class will tend to be larger .
this in turn leads to lower @xmath118 values .
as the qualitative imf - luminosity trend occurs in both star forming classes the strength of star formation is unlikely to be a significant bias on our results . as mentioned in the data corrections section , the @xmath95 ratio is ratio of the extinction experienced by the nebular emission lines to that experienced by the stellar continuum .
the assumption of a value for @xmath95 could potentially bias our results
. an alternative way of looking at the same problem is that our data in the color - h@xmath0 ew plane can be used to constrain the @xmath95 ratio by assuming a universal salpeter imf .
figure [ fig : f - all ] gives the results of this analysis for the data set as a whole .
a value of @xmath189 is found with @xmath190 .
this is in good agreement with to the @xcite value of @xmath191 .
the quality of the fit in the best case is worse than in figure [ fig : all - like ] .
part of the reason for this is that the errors used were slightly smaller as the @xmath97 terms in equations [ eq : sigmac ] and [ eq : sigmaw ] are set equal to 0 .
the quality of the fit drops sharply below @xmath115 and more gradually for @xmath192 .
values of @xmath95 near 1 are heavily rejected .
however in this particular plot the results are dominated by luminous galaxies .
the values of @xmath95 as a function of luminosity are shown in figure [ fig : f - vl ] .
for galaxies @xmath193 and brighter the best value of @xmath95 is consistent with @xmath115 .
the faintest two bins the prefer an @xmath95 ratio closer to 2.5 .
however the lower panel shows that this new @xmath95 value does not translate to improved fit quality .
in fact the faintest galaxies have in general smaller measured extinctions and are therefore less susceptible to changes in @xmath95 .
the same qualitative trend of worsening fits with decreasing luminosity seen when allowing @xmath108 to float is seen with a varying @xmath95 value .
together these two @xmath95 ratio plots provide a number of insights . for one
it shows that our choice of the @xmath95 ratio is very sensible and provides an independent confirmation of other @xmath95 ratio measurements .
the fact that our best fitting @xmath108 values are at least 0.05 above the salpeter value can not be reconciled by changing the geometry of the dust screen .
it provides further evidence that the relationship between @xmath108 and luminosity is not a function of extinction or a byproduct of our extinction correction . in this section
we have investigated several possible sources of bias to account for our observed trend between the imf and luminosity .
relationships between the imf and redshift , aperture fraction , extinction and star formation strength have been uncovered .
two parameter biases were also found . in all cases in narrow slices through the data where potential biases are held fixed the qualitative imf - luminosity relationship appears .
the parameters primarily act to offset the value of @xmath108 at a particular luminosity .
the ratio of continuum to emission line extinction , @xmath95 , was found to be a sensible choice and the results are not sensitive to small changes in this value .
there are two possible interpretations to the relationships between the imf and potential biases .
one is that they are systematic effects due to some problem with our measurement of @xmath108 .
the second is that they are real physical effects .
it is not clear from the data which of these statements is more correct . in the previous section
several possible sources of bias were investigated , but none were able to account for our observed trend in @xmath108 with luminosity or the inability of a universal imf to fit low luminosity galaxies .
figure [ fig : hagr1723 ] shows the distribution in color - h@xmath0 ew space for the least and most luminous bins , @xmath67 and @xmath68 . from this figure it is apparent that the most luminous galaxies lie roughly parallel to the imf tracks while the faintest galaxies are more perpendicular to the tracks . in the low luminosity bin
there are galaxies which are simultaneously blue and have low ews .
these galaxies are not consistent with a universal imf with @xmath111 and as mentioned before are not consistent with a universal imf with a different slope .
in addition the faintest galaxies have the lowest extinctions they are the least sensitive to dust and @xmath95 ratio issues . before concluding that this is evidence for imf variations we must first consider whether our model assumption of smooth sfhs is justified . the sfh of individual galaxies is the most problematic aspect of the k83 analysis .
a sudden burst on top of a smoothly varying background will immediately increase the h@xmath0 ew .
this is due to the formation of o and b stars which indirectly increase the h@xmath0 flux through processing of their ionizing photons .
the new presence of o and b stars also makes the color of the galaxy bluer .
both of these effects are proportional to the size of the burst .
after the burst is over the h@xmath0 ew is smaller and the colors are redder than they would be if the burst had not occurred .
the h@xmath0 ew drops because there is no longer an excess of o and b stars and their ionizing photons which reduces the h@xmath0 flux to pre - burst levels .
however there is now an excess of red giants due to the less massive stars from the burst leaving the main sequence .
this increases the continuum around the h@xmath0 line which further drops the ew in addition to making the galaxy colors redder . after enough time has elapsed after the burst the galaxy returns to the same position in the color - h@xmath0 ew plane it would have occupied had no burst occurred , although it will have taken longer to get there .
figure [ fig : burst ] gives one example of this cycle . a solar metallicity galaxy with @xmath111 and an exponentially decreasing sfh with @xmath194 gyr experiences a burst of star formation at an age of 4.113 gyr which lasts 250 myr and forms 10% of the stellar mass .
the black dots , spaced at 100 myr intervals , show that comparatively more time is spent below the nominal track than above it .
the peak h@xmath0 ew is reached just 5 myr after the start of the burst . if you happen to be observing the galaxy during the burst a shallower imf
will be measured ( assuming a burst - free sfh ) , after the burst for 1 gyr a steeper imf will be measured and after that the effects of the burst largely disappear , although the galaxy will appear younger than it actually is .
figure [ fig : burstfit ] gives the best fitting imf slope from our analysis , which assumes no bursts , as a function of age for the galaxy in figure [ fig : burst ] .
the jitter in the best fit @xmath108 values is due to the discrete nature of our model grid , and the fact that the model tracks for different imfs run together at large ages . within a 300 myr period during and just after the burst the best fitting imf slope is anywhere from @xmath195 to 1.95 .
this shows that even if a universal imf exists the sfh can mimic a huge range of imf models .
roughly 1 gyr after the burst the measured imf is back to its true value .
a galaxy with a bursty sfh viewed at a random time will be biased toward a steeper imf than what is the true imf .
while one model is not an exhaustive study of the effects of the sfh on imf measurements it does give a good sense of what issues arise . to eliminate this uncertainty we investigated cutting the sample on sfh . in order to detect a relative lack or excess of present star formation
it is necessary to measure both the present and past star formation rates or at least be able to compare the two in some way .
the problem lies in the fact that conversions of observables into star formation rates assume an imf to do so .
our aim is to measure the imf so we can not make strong a priori assumptions about it .
instead of biasing or results from the start we fit all galaxies and then try to determine the affect of sfhs on our conclusions .
the simplest burst model is that of a single burst at a random time on top of our smooth exponential sfhs .
a grid of 1000 sfhs was constructed by first selecting one of the 24 smoothly varying sfhs at random .
a burst lasting 200 myr was superposed on the sfh at a time selected uniformly at random over a range of 12.5 gyr .
the strength of the burst was randomly selected up to 40% of the total stellar mass , with preference given to smaller bursts .
the colors and ews of these sfhs were calculated at 1 myr intervals over 12.5 gyr for the imfs @xmath18 and 1.80 and a fixed metallicity of @xmath196 .
this metallicity choice is based on the luminosity - metallicity relationship in @xcite for @xmath197 galaxies .
plotting these models in the color - h@xmath0 ew plane shows that all observed data points are covered by either imf . to test whether the observed distribution of points in the @xmath162 bin can be explained by bursting sfhs the mc techniques of 5 were used .
the models of 4 were replaced with the grid of single burst models .
the 100 mc simulations were constructed as described earlier , but using only the 329 @xmath162 galaxies as a basis .
the analysis yielded @xmath198 for @xmath111 and @xmath199 for @xmath200 , both of which are over fits .
this shows than an individual galaxy can be fit with an arbitrary imf given the freedom to choose a sfh .
however , our advantage is that we have _ many _ galaxies and the distribution of the properties of the best fit models can be shown to be implausible .
figure [ fig : chibursts ] demonstrates the problem with the single burst model . on the left of the figure the distribution of the best fitting burst strengths are plotted as a fraction of the total stellar mass formed .
at right is the distribution of the best fitting times from the burst onset .
for example , a galaxy which best fits a model with a burst at 1.000 gyr at an age of 1.211 gyr has a time from burst onset of 211 myr .
this measure is used because in an investigation of the effects of bursts the age relative to the burst is more important than the age given that the bursts occur at different , random times across the models . for both @xmath111 and @xmath200 the number of objects best fit by a model prior to the burst
is @xmath201 , or 3% , and are not plotted . in both cases the distributions of the best fitting ages and ages at which the burst begins are roughly uniform .
given this fact it is expected that half of the galaxies should be best fit by a pre - burst model .
furthermore the right hand panel shows sharp , significant discontinuities in the distribution of best fitting time from burst onset .
again , viewed at random times this distribution should be uniform but is highly peaked in the 25 myr at the start of the burst and the 25 myr just after the burst ends . in both cases
the errors bars show that the discontinuities are significant . in the case of @xmath111 , 3.5 times as many galaxies are in the 200 myr after the burst ends than the 200 myr during it and this 400 myr accounts for 57% of all galaxies .
although our sample is @xmath86-band selected the stars in the 0.7 to 3 m@xmath25 range which dominate the red continuum in the red giant phase do not start to leave the main sequence for 300 myr .
the sharp increase in galaxies fit at 200 myr after the burst can not be due to a selection effect . assuming a universal imf this points to a strong coordination of sfhs across a population of galaxies unrelated in space .
these arguments show that while a single burst model can fit the data extremely well , it does not do so in a physically self - consistent fashion . to find a physical motivation for sfh models for low luminosity galaxies we look towards the local group .
there have been a number of recent studies of the sfhs of local dwarf galaxies which use hst to get color - magnitude diagrams ( cmd ) of resolved stellar populations .
the sfh is determined by fitting isochrones to the cmd .
the sample here is biased by local group membership and by what galaxies have been observed to date .
the galaxies mentioned here give a point of reference rather than a well - defined distribution of sfhs . the blue compact dwarf ( bcd )
ugca 290 was found to quiescently form stars over the past gyr up until a ten - fold increase in sfr from 15 to 10 myr ago which more recently has decreased to a quarter of its peak value @xcite .
the dwarf irregular ic 1613 , which is relatively isolated and non - interacting , was found to have sfr enhanced by a factor of 3 from 3 to 6 gyr ago without evidence of strong bursts @xcite .
the dwarf irregular ngc 6822 , also relatively isolated , is found to have a roughly constant sfh @xcite .
the bcd ngc 1705 is found to be _
gasping_- a sfh marked by moderate activity punctuated by short periods of decreased star formation @xcite .
the authors also note that ngc 1705 is best fit by an imf with @xmath202 .
ngc 1569 likely experienced three strong bursts in the last gyr as well as a quiescent phase from 150 to 300 myr ago @xcite .
inspired by the preceding local group sfhs we constructed six sfh classes with multiple bursts .
these sfh classes are described in table [ tab : sfhs ] .
each class starts with an underlying smooth sfh .
sfhs # 1 & 2 have no star formation , # 3 & 4 have a constant sfh and # 5 & 6 have exponentially decreasing sfhs like those previously described .
star formation discontinuities are then superimposed on top of the smooth sfhs .
these discontinuities are in the form of increased ( bursts ) or decreased ( gasps ) star formation for periods of 200 myr .
the time and spacing of the discontinuities is random with the mean interval between bursts listed in table [ tab : sfhs ] . for each sfh class described above we randomly generated 1,000 sfhs .
colors and ew widths were calculated for each sfh using @xmath111 and @xmath196 . according to the sdss mass - metallicity relationship @xcite galaxies at @xmath67 will on average have @xmath196 .
we then repeated our @xmath118 analysis with the 100 mc simulations of the @xmath162 galaxies in the same manner as for the single burst models .
the results of our analysis for each of the six sfh classes are shown in figure [ fig:6sfh ] .
sfh 5 , the gasps on top of exponential sfhs , is the best fit with @xmath203 .
extended periods of no star formation punctuated by bursts ( sfhs # 1 & 2 ) do not fit the data . as was the case for the single burst models an unreasonable fraction of galaxies
are best fit by sfhs in the 20 myr immediately following a burst or the first 20 myr of a gasp .
if the sfh models are reasonable we should see roughly equal numbers of galaxies in each time bin .
there is no reason why all of the low luminosity galaxies across the large volume of space in the sdss footprint should have experienced coordinated bursts .
however each panel of figure [ fig:6sfh ] has at least 40% of the galaxies in one 20 myr bin .
one explanation for this is that it is an artifact of our sample being selected in the @xmath86 band .
however spectral synthesis models show that for instantaneous bursts of star formation the @xmath165 magnitude is brightest at the burst time and decays smoothly for a range of @xmath108 . if anything it is more likely to catch galaxies during a burst rather than after or after a gasp instead of during one .
regardless of the sfh model the presence of blue galaxies with low h@xmath0 ew requires a recent discontinuity in the sfr for @xmath111 .
based on the evolution of the @xmath165 band luminosity we expect to see a similar number of galaxies with excess h@xmath0 ews .
the fact that these galaxies are missing shows that the discontinuous sfh models do not match our observations .
therefore imf variations are a more likely explanation for the observed distribution of @xmath67 galaxies . as a last exercise
the best fitting sfh models from the previous section can be run forward to see if the correct imf can be recovered . for each sfh
model grids 10,000 data points were chosen by selecting a random sfh and a _ uniformly distributed _ age .
normal errors were added using the error characteristics of the @xmath67 bin .
this synthetic data was analyzed in the same fashion as the real data in the earlier sections .
for the single burst models the recovered imf models for @xmath204 and 1.80 were 1.34 and 1.79 respectively with best fitting @xmath205 0.80 and 0.65 .
for sfh 5 the recovered imf was also @xmath206 with @xmath207 .
in all three cases the correct imfs were recovered although the fit was worsened by by the burst activity .
this reinforces the difficulty in producing enough blue , low h@xmath0 ew galaxies to match the observed data with simple sfh models .
in the previous sections we have expanded on the k83 method and exploited the h@xmath0 and color information as much as possible . in the bias section we found that various possible biases do not fully explain either the increased values of @xmath108 or the poor fit to a single imf in the lowest luminosity bins . in the sfh section we found that an arbitrary @xmath108 value coupled with a plausible sfh with bursts or gasps can account for the position of any individual galaxy in the color - h@xmath0 ew plane . however taking the population of @xmath67 galaxies together necessitates an incredibly unlikely coordination of sfhs across the disparate group of objects .
this points to the extraordinary conclusion that while the imf may be universal across luminous galaxies , it is not in fact universal in low luminosity galaxies .
such an extraordinary claim would ideally be backed by extraordinary evidence . in this section
we take a look beyond the k83 method for some reinforcement of our result .
the h@xmath208 absorption feature can be used to gain additional insight into the nature of stellar populations .
absorption is due to absorption lines form stellar photospheres .
the balmer absorption lines are most prominent in a stars and weaken due to the saha equation for both hotter and cooler stars .
as such the h@xmath208 absorption is a proxy for the fraction of light of a stellar population being supplied by a stars , and to a lesser extent b and f type stars . in a stellar population of a uniform age the h@xmath208 absorption will peak after the o and b stars burn out , but before the a stars leave the main sequence .
for this reason the strength of h@xmath208 can be used help determine the age of a population or to detect bursts of star formation which occurred around 1 gyr in the past .
@xcite describes two different methods for measuring h@xmath208 absorption .
the h@xmath209 definition is tuned to most accurately measure the h@xmath208 absorption from f stars .
the h@xmath29 definition has a wider central bandpass to match the line profiles of a stars .
they state that the h@xmath29 definition is better to use for galaxies because it is less noisy in low s / n galaxies and velocity dispersion acts to widen absorption features . on the downside the narrower h@xmath209 definition is much more sensitive to population age where it can be used .
@xcite observationally determines the range of h@xmath29 values to be from 13 for a4 dwarf stars to -9 for m - type giant stars .
h@xmath29 values can be measured from the sdss spectra . like
the h@xmath0 ew values our h@xmath29 values come from @xcite instead of the sdss pipeline .
figure [ fig : hdfig ] compares the distribution of the h@xmath29 values for the @xmath67 and @xmath180 luminosity bins . for reference recall that figure [ fig : hagr1723 ] plots the color and h@xmath0 ews for these two bins .
the difference between the high and low luminosity bins is clear .
gaussian profiles can be fit to both distributions . for the @xmath180 bin
the standard deviation of the profile is twice as large as the measurement error in h@xmath29 suggesting that the true distribution has a range of values .
assuming both the errors and underlying distribution are gaussian the distribution of h@xmath29 for the luminous galaxies is centered at h@xmath210 with a standard deviation of 1.5 . by contrast , the gaussian fit to the @xmath67 bin has the same standard deviation as the median error in the galaxies .
this is consistent with nearly all of the galaxies having h@xmath211 .
this shows that the fainter galaxies on average have significantly larger fractions of a - type stars amongst their stellar populations .
it also shows a seemingly unlikely coordination of h@xmath29 in the low luminosity galaxies , reminiscent of the earlier suggestion of coordinated sfhs .
the question then becomes why ? for a possible explanation we look again to the models .
modeling the behavior of h@xmath29 requires an extra step .
the standard pegase.2 models do not have the required resolution to accurately measure the h@xmath29 index .
this is remedied with the use of the pegase - hr code @xcite which uses a library of echelle spectra of 1503 stars to calculate spectral synthesis models with r=10,000 over the range of 4000 to 6800 .
using pegase - hr in the low resolution mode yields the same results as pegase.2 , and the same input parameters are used for both codes .
the models here are the same as those described earlier in the models section , but have been recalculated using pegase - hr to allow h@xmath29 measurements .
figure [ fig : hdmodplot ] shows the behavior of the h@xmath29 index as a function of age for four different imf models , @xmath142 , 1.40 , 1.70 and 2.00 . for each
imf models of all metallicities and smooth sfhs are plotted for each age .
the qualitative behavior of all models is the same .
the h@xmath29 holds steady for the first 20 to 40 myr before increasing to a peak value at 700 myr to 1 gyr and then falls off .
prior to reaching the peak value for each imf the metallicity has the strongest effect on the h@xmath29 index .
at this point the lowest metallicity galaxies have the highest h@xmath29 .
after the peak the sfh has the strongest effect with the constant and increasing sfhs maintaining higher h@xmath212 values .
after a few gyr the differences in h@xmath29 between models with different imfs disappear . prior to that
there are three main differences .
first , the peak h@xmath29 values are higher for larger values of @xmath108 . for @xmath145 h@xmath29
reaches a maximum value of nearly 8 and the maximum value is similar for all metallicities . for @xmath142
the maximum value ranges from 4.5 to 6 depending on the metallicity .
this is due to the fact that the steeper imfs have fewer luminous massive stars to dilute the h@xmath208 absorption features from the a star population .
secondly , the low @xmath108 models have h@xmath29 values that start their initial increases at a later time .
lastly the low @xmath108 models reach their peak values later in time than those with fewer massive stars . to compare the h@xmath29 values for the @xmath180 and @xmath213 bins to the models the range of the middle 90% values from figure [ fig : hdfig ] for each bin
are overlaid on figure [ fig : hdmodplot ] .
once again the @xmath180 bin is in good agreement with our assumption of a universal imf and smooth sfh .
the range of h@xmath29 values can be accomplished with a single salpeter - like @xmath214 imf with only the proviso that most galaxies be older than a few gyr or younger than 300 myr .
however the exact same statement can be made for @xmath215 so h@xmath29 provides only a constraint on the age of the most luminous galaxies , but not the imf . for the @xmath67 galaxies the distribution of h@xmath29 can not be achieved with the shallowest imfs investigated under the assumption of smooth sfhs .
however salpeter and steeper imfs can be accomplished .
what changes is the range of ages over which the models have the correct h@xmath29 .
steeper imfs require that the galaxies be either older than a few gyr or 200 myr old to accommodate the observed h@xmath29 values .
most troubling is that there does not appear to be any reason why h@xmath29 should stack up at 6.1 for the @xmath67 galaxies .
h@xmath29 values can also be calculated for the same sfh class models used earlier .
unfortunately this does not provide any added constraints .
values remain elevated for several 100 myr after the start of a burst or the end of a gasp of sfh .
the behavior of h@xmath29 in the presence of sfh discontinuities provides no need for galaxies to stack up in the narrow 20 myr intervals seen in the earlier multiple burst section , nor do these models suggest why the low luminosity galaxies are consistent with a single value of h@xmath29 .
there is no satisfactory model to account for the h@xmath29 distribution of the low luminosity galaxies .
however the sfh results from the previous section strongly suggest that the incredible coordination of discontinuities is highly unlikely .
steeper imfs do allow for higher h@xmath29 values for longer periods of time , thus relaxing the sfh coordination requirement .
this agrees with our earlier results for faint galaxies .
the low luminosity galaxies are most likely the result of a mix of imfs which are on average steeper than salpeter .
the goal of this paper was to revisit the k83 method for inferring the imf from integrated stellar populations and to harness the richness of the sdss data , improved spectral synthesis models and greater computational power available today to make a state - of - the - art measurement of the imf .
the quality of the sdss spectroscopy allowed us to address several of the limitations of k83 and ktc94 we resolve the [ ] lines ( a significant improvement in the accuracy of the h@xmath0 ews of individual galaxies ) , eliminate contamination from agn , make extinction corrections for individual galaxies and fit underlying stellar absorption of h@xmath0 .
we succeeded in achieving more accurate ews for individual galaxies .
the median total ew error for our sample is 17% compared to a 10% uncertainty in ews combined with a 20 - 30% uncertainty in the extinction correction for k83 .
we expanded the grid of models to allow for a range of ages and metallicities .
we used @xmath118 minimization to go beyond differentiating between two or three imf models to actual fitting for the best imf slope .
the vast size of the sdss sample allowed us to both drive down random errors and to cut the data into narrow parameter ranges which were still statistically viable .
the size of the dr4 sample yielded @xmath216 95% confidence region due to random error for the sample as a whole .
even the volume limited @xmath67 luminosity bin with only 329 objects has a random error of @xmath217 . only in bins with fewer than 10 objects
do the random errors become significant .
our imf fitting is therefore dominated by systematics .
originally we believed that our systematics would be dominated by the effects of sfh discontinuities .
however we conducted several experiments where we selected populations of galaxies from models with bursting or gasping sfhs and gave them measurement errors consistent with those in the @xmath162 luminosity bin . to our surprise
our @xmath118 minimization revealed the true imfs with @xmath218 .
the main effect was to reduce the quality of the fits .
this is due to the fact that h@xmath0 ews return to nominal levels in a relatively short time after sfh discontinuities . another way to estimate
the size of the systematic errors is to look at the trends of the @xmath219 and @xmath160 luminosity bins , because they have the largest membership , in figures [ fig : redshift ] , [ fig : aperture ] and [ fig : extinction ] . assuming that the imf is universal and that our method is perfect we should get the same answer for any subset of the data we might choose .
the largest ranges are @xmath220 for redshift binning , @xmath221 by aperture and @xmath221 by extinction .
conservatively then the systematic error is @xmath4 .
there are two points to be kept in mind about this estimate of the systematic error . for one
it is the systematic error in the exact value of @xmath108 . even in figure
[ fig : avap ] where more narrow bands of measured extinction and aperture fraction are considered the same trends with luminosity are seen as with the sample as a whole .
the relative systematics between luminosity bins in these narrow slices is much smaller .
the second thing to remember is the way in which we empirically defined our systematic error discounts the possibility of imf variations .
what we have called systematics could actually be science .
if galaxies have radial imf gradients or if dust content plays a strong role in star formation the systematic error could be much smaller .
the main area in which we were unable to improve upon the k83 and ktc94 studies is that they were able to match the aperture size to the galaxies which avoids the issue of aperture effects .
first , for galaxies brighter than @xmath222 the best fitting imfs are salpeter - like ( @xmath7 ) .
in addition the assumption of a universal imf and smoothly varying sfhs is a good fit .
this is reassuring as it follows the conventional wisdom and provides confidence that the method works .
secondly , galaxies fainter than @xmath222 are best fit by steeper imfs with larger fractions of low mass stars . for these galaxies a universal imf and
smooth sfh is a poor assumption .
this result is in qualitative agreement with evidence that lsbs have bottom - heavy imfs @xcite .
it is worth mentioning the main caveat of our @xmath108 values again .
as illustrated in figure [ fig : oddimf ] imf parameterizations are themselves degenerate in our parameter space . increasing the imf slope has a similar effect to lowering the highest mass stars that are formed or increasing the fraction of intermediate mass stars .
this method can not explicitly determine if two populations have the same underlying imf .
figure [ fig : all - like ] shows that for the sample as a whole the @xcite three part power law yields nearly the same result as our two part power law .
however our method is sensitive in many cases if the imfs are different . in terms of star
forming cloud temperatures the harsher ambient radiation and larger number of sources of cosmic rays present in more luminous galaxies agree qualitatively with our results . with the extra energy hitting the star forming clouds
larger masses may be needed for contraction and fractionization may end sooner , suppressing the formation of less massive stars @xcite .
@xcite find that while the regions of the luminous grand design spiral ngc 5457 ( m 31 ) can be reproduced by a single salpeter imf , for the low luminosity flocculent galaxy ngc 4395 a blend of two imfs is required .
however , such trends are not seen in studies of well - resolved stellar populations @xcite .
another explanation for the absence of massive stars is that the massive stars are there , but are not visible .
extinction to the center of star forming regions , where massive stars preferentially exist , can reach @xmath223 @xcite .
however the low luminosity galaxies have the lowest observed extinctions ( see table [ tab : lumbins ] ) which is the opposite of what would be expected given our imf results .
it is also possible that the imf is in fact universal , but the way in which it is sampled in embedded star clusters leads to an integrated galaxial imf which varies from the true imf .
@xcite use a universal imf with the assumption that stars are born in clusters where the maximum cluster mass is related to the star formation rate . for a range of models
this leads to a narrow range for the apparent imf in high mass galaxies . for low mass galaxies
the imf is steeper with a wider range of slopes .
the results here are in qualitative agreement for some of the integrated galactic imf scenarios in @xcite given that there should be a rough correlation between galaxy luminosity and mass .
once again , @xcite argues that the galaxy wide imf should be the same as the imf in clusters regardless . in light of the theory of @xcite , whether the results of this paper speak to a relationship between environment and the formation of individual stars is open to interpretation .
however the impact on the modeling and interpretation of the properties of galaxies is clear .
@xcite note that the integrated galaxial imf is the correct imf to use when studying global properties of galaxies .
even if the imf of stars is in truth universal it may currently be misused in the modeling of galaxies .
furthermore a varying integrated galaxial imf could open the door to new insights in galaxy evolution .
for instance , @xcite suggest that the observed mass - metallicity relationship in galaxies naturally arises from a variable integrated galaxial imf similar to the results of this paper
. future work will expand in several directions .
the success constraining the imf with only two observed parameters ( albeit carefully chosen to be orthogonal to systematic errors ) motivates a more expansive analysis with more parameters .
information from wavelength regimes beyond the sdss can be used .
for instance the absorption strength and p - cygni profile shape of the @xmath224 line due to massive stars is sensitive to the imf slope and upper mass cutoff @xcite .
however it could be contaminated by absorption from the interstellar medium and would need to be disentangled @xcite .
a full markov chain monte carlo analysis could be implemented fitting to multiple spectral features and marginalizing over a range of sfhs .
in addition to luminosity , surface brightness and gas phase metallicity can be tested for systematic imf variations .
while the results for luminous galaxies are already dominated by systematics , the continued progress of the sdss can provide better statistics for a more detailed analysis of what physical processes are behind the imf variations in faint galaxies .
e.a.h . and k.g .
acknowledge generous funding from the david and lucile packard foundation .
we would also like to thank c. tremonti , g. kauffmann and t. heckman for an early look at their sdss spectral line catalogs . e.a.h .
would like to thank johns hopkins for funding from various sources as well as c. tremonti , e. peng and a. pope for invaluable assistance with the nuances of the sdss .
several plots were created with the use of publicly available idl routines written by d. schlegel .
funding for the creation and distribution of the sdss archive has been provided by the alfred p. sloan foundation , the participating institutions , the national aeronautics and space administration , the national science foundation , the u.s .
department of energy , the japanese monbukagakusho , and the max planck society .
the sdss web site is http://www.sdss.org/.
the participating institutions are the university of chicago , fermilab , the institute for advanced study , the japan participation group , the johns hopkins university , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , princeton university , the united states naval observatory , and the university of washington .
abazajian , k. , et al .
2004 , , 128 , 502 adams , f. c. , & fatuzzo , m. 1996 , 464 , 256 adelman - mccarthy , j. k. , et al .
2006 , , 162 , 38 angeretti , l. , et al .
2005 , , 129 , 2203 annibali , f. , et al .
2003 , , 126 , 2752 baldry , i. k. & glazebrook , k. 2003 , , 593 , 258 baldry , i. k. , et al .
2004 , , 600 , 681 baldwin , j. a. , phillips , m. m. , & terlevich , r. 1981 , , 93 , 5 blanton , m. r. , et al . 2003a , , 125 , 2348 blanton , m. r. , et al .
2003b , , 592 , 819 bonanos , a. z. et al .
2004 , , 611 , l33 brinchmann , j. , et al .
2004 , , 351 , 1151 calzetti , d. , kinney , a. l. , & storchi - bergmann , t. 1994 , , 429 , 582 cedrs , b. , cepa , j. & tomita , a. 2005 , , 634 , 1043 courteau , s. , & van den bergh , s. 1999 , , 118 , 337 crone , m. m. , et al .
2002 , , 567 , 258 delhaye , j. 1965 , in stars and stellar systems , vol . 5 , galactic structure , ed .
a. blaaun & m. schmidt ( chicago : university of chicago press ) , 61 duquennoy , a. , & mayor , m. 1991 , , 248 , 485 efstathiou , g. , ellis , r. s. & peterson , b. a. 1988 , , 232 , 431 elmegreen , b. g. 1997 , , 486 , 944 elmegreen , b. g. 2006 , , 648 , 572 engelbracht , c. w. , rieke , m. j. , rieke , g. h. , kelly , d. m. , & achtermann , j. m. 1998 , , 505 , 639 figer , donald f. 2005 , , 434 , 192 fioc , m. , & rocca - volmerange , b. 1997 , , 326 , 950 fischer , d. a. , & marcy , g. w. 1992 , , 396 , 178 fukugita , m. , et al .
1996 , , 111 , 1748 hoyle , f. 1953 , , 118 , 513 kauffmann , g. , et al .
2003 , , 346 , 1055 kennicutt , r. c. 1983 , , 272 , 54 ( k83 ) kennicutt , r. c. 1998a , in asp conf .
142 , the stellar initial mass function , ed . , g. gilmore & d. howell ( san francisco : asp ) , 1 kennicutt , r. c. 1998b , , 36 , 189 kennicutt , r. c. , & kent , s. m. 1983 , , 88 , 1094 kennicutt , r. c. , tamblyn , p. , & congdon , c. w. 1994 , , 435 , 22 ( ktc94 ) kppen , j. , weidner , c. , & kroupa , p. 2007
, , 375 , 673 kroupa , p. 2001
, , 322 , 231 kroupa , p. 2002 , science , 295 , 82 kroupa , p. , tout , c. a. , & gilmore , g. 1993 , , 262 , 545 larson , r. b. 1998 , , 301 , 569 le borgne , d. , et al .
2004 , , 425 , 881 lee , h. , et al .
2004 , , 353 , 113 leitherer , c. , robert , c. , & heckman , t. m. 1995 , , 99 , 173l low , c. , & lynden - bell , d. 1976 , , 176 , 367 lupton , r. h. , gunn , j. e. & szalay , a. s. 1999 , , 118 , 1406 mayor , m. 1976 , , 48 , 301 miller , g. e. & scalo , j. m. 1979 , , 41 , 513 osterbrock , d. e. 1989 , astrophysics of gaseous nebulae and active galactic nuclei ( sausalito , ca : university science books ) padoan , p. , nordlund , . , & jones , b. j. t. 1997 , , 288 , 145 pei , y. c. 1992 , , 395 , 130 phelps , r. l. , & janes , k. a. 1993 , , 106 , 1870 press , w. h. , teukolsky , s. a. , vetterling , w. t. , & flannerty , b. p. 1992
, numerical recipes in c : the art of scientific computing , second edition ( cambridge : cambridge university press ) rees , m. j. 1976 , , 176 , 483 robberto , m. et al .
2004 , in imf@50 : the initial mass function 50 years later , ed .
, e. corbelli , f. palla & h. zinnecker ( kluwer academic publishers ) rolleston , w. r. j. , smartt , s. j. , dufton , p. l. & ryans , r. s. i. 2000 , , 363 , 537 salpeter , e. e. 1955 , , 121 , 161 scalo , j. m. 1986 , fund . cos . phys . , 11 , 1 scalo , j. m. 1998 , in asp conf .
142 , the stellar initial mass function , ed . ,
g. gilmore & d. howell ( san francisco : asp ) , 201 schlegel , d. j. , finkbeiner , d. p. , & davis , m. 1998 , , 500 , 525 shapley , a. e. , steidel , c. c. , pettini , m. , & adelberger , k. l. 2003 , , 588 , 65 skillman , e. d. , et al . 2003 , , 596 , 253 stoughton , c. , et al . 2002 , , 123 , 485 tremonti , c. a. , et al . 2004 , , 613 , 898 uomoto , a. , et al . 1999 , , 31 , 1501 vila - costas , m. b. & edmunds , m. g. 1992 , 259 , 121 weidner , c. & kroupa , k. 2004 , , 348 , 187 weidner , c. & kroupa , k. 2005 , , 625 , 754 worthey , g. & ottaviani , d. l. 1997 , 111 , 377 wyder , t. k. 2003 , , 125 , 3097 york , d. g. , et al . 200 , , 120 , 1579 crcccccccrcc -14 & - & - & - & - & - & - & - & - & 28 & 1.3835 & 7.43 + -15 & - & - & - & - & - & - & - & - & 188 & 1.3892 & 9.86 + -16 & - & - & - & - & - & - & - & - & 406 & 1.5461 & 4.64 + -17 & 329 & 1.6000 & 1.6045 & 1.6086 & 3.41 & 0.013 & 0.38 & 0.20 & 1,304 & 1.5879 & 2.57 + -18 & 1,555 & 1.5338 & 1.5370 & 1.5424 & 2.05 & 0.022 & 0.50 & 0.21 & 4,327 & 1.5326 & 1.57 + -19 & 4,935 & 1.4772 & 1.4788 & 1.4813 & 1.17 & 0.032 & 0.70 & 0.21 & 10,375 & 1.4813 & 0.96 + -20 & 12,951 & 1.4306 & 1.4320 & 1.4330 & 0.63 & 0.050 & 0.91 & 0.22 & 24,851 & 1.4436 & 0.56 + -21 & 28,633 & 1.4051 & 1.4057 & 1.4063 & 0.32 & 0.077 & 1.15 & 0.24 & 41,411 & 1.4064 & 0.31 + -22 & 29,701 & 1.4036 & 1.4042 & 1.4050 & 0.19 & 0.116 & 1.32 & 0.25 & 38,406 & 1.4084 & 0.20 + -23 & 8,049 & 1.4545 & 1.4556 & 1.4568 & 0.15 & 0.168 & 1.55 & 0.27 & 9,106 & 1.4505 & 0.16 + -24 & - & - & - & - & - & - & - & - & 192 & 1.5329 & 0.12 + all * & 130,602 & 1.4432 & 1.4437 & 1.4443 & 0.50 & 0.090 & 1.05 & 0.25 & - & - & - + [ tab : lumbins ]
sfh 1 & burst & 200 & - & 3 & none + sfh 2 & burst & 200 & - & 1 & none + sfh 3 & burst & 200 & 4.0 & 3 & constant + sfh 4 & gasp & 200 & 0.0 & 1.5 & constant + sfh 5 & gasp & 200 & 0.1 & 1.5 & exponential + sfh 6 & burst & 200 & 5.0 & 1.5 & exponential + [ tab : sfhs ] | this paper revisits the classical kennicutt method for inferring the stellar initial mass function ( imf ) from the integrated light properties of galaxies . the large size
, uniform high quality data set from the sloan digital sky survey dr4 is combined with more in depth modeling and quantitative statistical analysis to search for systematic imf variations as a function of galaxy luminosity .
galaxy h@xmath0 equivalent widths are compared to a broadband color index to constrain the imf .
this parameter space is useful for breaking degeneracies which are traditionally problematic .
age and dust corrections are largely orthogonal to imf variations .
in addition the effects of metallicity and smooth star formation history e - folding times are small compared to imf variations . we find that for the sample as a whole the best fitting imf slope above 0.5 @xmath1 is @xmath2 with a negligible random error of @xmath3 and a systematic error of @xmath4 .
galaxies brighter than around @xmath5 ( including galaxies like the milky way which has @xmath6 ) are well fit by a universal @xmath7 imf , similar to salpeter , and smooth , exponential star formation histories ( sfh ) .
fainter galaxies prefer steeper imfs and the quality of the fits reveal that for these galaxies a universal imf with smooth sfhs is actually a poor assumption .
several sources of sample bias are ruled out as the cause of these luminosity dependent imf variations .
analysis of bursting sfh models shows that an implausible coordination of burst times is required to fit a universal imf to the @xmath8 galaxies .
this leads to the conclusions that the imf in low luminosity galaxies has fewer massive stars , either by steeper slope or lower upper mass cutoff , and is not universal . |
space time explorer & quantum equivalence space test ( ste - quest ) is a medium - sized mission candidate for launch in 2022/2024 in the cosmic vision programme of the european space agency .
after recommendation by the space science advisory committee , it was selected to be first studied by esa , followed by two parallel industrial assessment studies .
this paper gives a brief summary of the assessment activities by astrium which build on and extend the preceding esa study as described in@xcite .
ste - quest aims to study the cornerstones of einstein s equivalence principle ( eep ) , pushing the limits of measurement accuracy by several orders of magnitude compared to what is currently achievable in ground based experiments@xcite . on the one hand
, experiments are performed to measure the gravitational red - shift experienced by highly accurate clocks in the gravitational fields of earth or sun ( space time explorer ) . on the other hand , differential accelerations of microscopic quantum particles
are measured to test the universality of free fall , also referred to as weak equivalence principle ( quantum equivalence space test ) .
these measurements aim at finding possible deviations from predictions of general relativity ( gr ) , as postulated by many theories trying to combine gr with quantum theory .
examples include deviations predicted by string theory @xcite , loop quantum gravity@xcite , standard model extension@xcite , anomalous spin - coupling@xcite , and space - time - fluctuations@xcite , among others .
the ste - quest mission goal is summarized by the four primary science objectives@xcite which are listed in tab.[tab : mission_objectives ] together with the 4 measurement types geared at achieving them .
lll primary mission objective & measurement accuracy & measurement strategy + measurement of & to a fractional frequency & ( a ) space - clock comparison + earth gravitational red - shift & uncertainty better than & to ground clock at apogee + & @xmath0 & ( b ) space - clock comparison + & & between apogee and perigee + measurement of & to a fractional frequency & ( c ) comparison between + sun gravitational red - shift & uncertainty better than & two ground clocks via + & @xmath1 , goal : @xmath2 & spacecraft + measurement of & to a fractional frequency & ( c ) comparison between + moon gravitational red - shift & uncertainty better than & two ground clocks via + & @xmath3 , goal : @xmath4 & spacecraft + measurement of & to an uncertainty in the & ( d ) atom interferometer + weak equivalence principle & etvs param .
smaller & measurements at perigee + & @xmath5 & +
the ste - quest mission is complex and comprises a space - segment as well as a ground segment , with both contributing to the science performance .
highly stable bi - directional microwave ( x / ka - band ) and optical links ( based on laser communication terminals ) connect the two segments and allow precise time - and - frequency transfer .
the space - segment encompasses the satellite , the two instruments , the science link equipment , and the precise orbit determination equipment .
the ground - segment is composed of 3 ground terminals that are connected to highly accurate ground - clocks . in order to fulfil the mission objectives ,
various measurement types are used that are shown in fig.[fig : measurement_principle ] .
we shall briefly discuss them .
_ earth gravitational red - shift measurements : _ the frequency output of the on - board clock is compared to that of the ground clocks . in order to maximize the signal , i.e. the relativistic frequency offset between the two clocks , a highly elliptical orbit ( heo ) is chosen .
when the spacecraft is close to earth during perigee passage , there are large frequency shifts of the space - clock due to the strong gravitational field . when it is far from earth during apogee passage , there are only small gravitational fields and therefore small frequency shifts . whilst measurement type ( a ) compares the space - clock at apogee to the ground - clock , relying on space - clock accuracy , measurement type ( b ) compares the frequency variation of the space - clock over the elliptical orbit , which requires clock stability in addition to visibility of a ground terminal at perigee ( see also section [ sec : orbit_ground ] ) . , scaledwidth=100.0% ]
+ _ sun gravitational red - shift measurements : _ the relativistic frequency offset between a pair of ground clocks is compared while the spacecraft merely serves as a link node between the ground clocks for a period of typically 5 to 7 hours , as described by measurement type ( c ) . in that case , the accuracy of the space - clock is of minor importance whereas link stability and performance over a long period are essential . as
this requirement is also hard to fulfil by the microwave links alone , the optical links play an important role .
however , availability and performance of optical links are strongly affected by weather conditions which in turn depend on the location and altitude of the ground terminals .
+ _ moon gravitational red - shift measurements : _ as in the preceding case , the relativistic frequency offset between a pair of ground clocks is compared while the spacecraft merely serves as a link node between the ground clocks .
the potential signals for a violation of the eep with the moon as the source mass can be easily distinguished from those with the sun as the source mass due to the difference in frequency and phase .
it is important to point out that unless the eep is violated the measured frequency difference between two distant ground clocks ( due to the sun or the moon ) is expected to be zero up to small tidal corrections and a constant offset term from the earth field@xcite .
+ _ atom interferometer measurements : _ these measurements do not require contact to ground terminals but must be performed in close proximity to earth where the gravity gradients are large and the etvs parameter ( see eq.[equ : otvos ] below ) becomes small .
the latter defines the magnitude of a possible violation of the weak equivalence principle ( wep ) , as required for mission objective 4 . for this reason , the atom interferometer operates during perigee passage at altitudes below 3000 km , where gravity accelerations exceed @xmath6 .
the ste - quest payload primarily consists of two instruments , the atomic clock and the atom interferometer , as well as equipment for the science links ( microwave and optical ) and for precise orbit determination ( pod ) . whilst a detailed description of the proposed payload elements
is found in @xcite , here we shall only give a brief overview over the payload as summarized in table [ tab : payload ] .
note that some payload elements , such as the optical links and the onboard clock , may become optional in a future implementation of ste - quest , which is discussed in the official esa assessment study report ( yellow book)@xcite .
lll payload element , & subsystem , & components + instrument & unit & + & pharao ng & caesium tube , laser source + * atomic clock * & & + & microwave - optical local & laser head , frequency comb , + & oscillator ( molo ) & highly - stable cavity + & & + & mw synthesis & frequency & ultra - stable oscillator , frequency gen . ,
+ & distribution ( msd ) & comparison and distribution ( fgcd ) + & & + & instr .
control unit ( icu ) & control electronics , harness + & physics package & vacuum chamber , atom source , + * atom interf .
* & & magnetic coils and chip , laser + & & interfaces , mu - metal shields + & & + & laser subsystem & cooling , repumping , raman , + & & detection , optical trapping lasers + & & + & electronics & ion pump control , dmu , magnetic coil + & & drive , laser control electronics + & microwave links & mw electronics , x-/ka - band antennas + * equipment * & & + & optical links & 2 laser communication terminals + & & ( lcts ) , synchronization units + & gnss equipment & gnss receiver and antennas + * equipment * & & + & laser ranging equipment & corner cube reflectors ( optional ) + + _ the atomic clock
_ is essentially a rebuilt of the high - precision pharao clock used in the aces mission@xcite , with the additional provision of an optically derived microwave signal ( mw ) for superb short term stability which is generated by the molo . in pharao
, a cloud of cold caesium atoms is prepared in a specific hyperfine state and launched across a vacuum tube .
there , the cloud is interrogated by microwave radiation in two spatially separated ramsey zones before state transition probabilities are obtained from fluorescence detection . repeating the cycle
while scanning the mw fields yields a ramsey fringe pattern whose period scales inversely proportional to the flight time t between ramsey interactions . in the molo , a reference laser
is locked to an ultra - stable cavity , which provides superb short - term stability in addition to the long - term stability and accuracy obtained from pharao by locking the reference signal to the hyperfine transition of the caesium atoms .
the reference laser output is frequency - shifted to correct for long term drifts and then fed to the femtosecond frequency comb@xcite to convert the signal from the optical to the microwave - domain .
core components of the molo , such as the highly stable cavity@xcite , the frequency comb@xcite , and optical fibres @xcite are currently being developed or tested in dedicated national programs .
additionally , other components of optical clocks have been investigated in the ongoing space - optical - clock program ( soc)@xcite of the european space agency .
+ _ the atom interferometer_@xcite uses interference between matter waves to generate an interference pattern that depends on the external forces acting on the atoms under test . in order to reject unwanted external noise sources as much as possible ,
differential measurements are performed simultaneously with two isotopes of rubidium atoms , namely @xmath7 and @xmath8 .
the measured phases for both isotopes are subtracted to reveal any possible differences in acceleration under free fall conditions of the particles .
this way , the common mode noise which is present on both signals is rejected to a high degree@xcite . in the proposed instrument
@xcite a number of the order of @xmath9 atoms of both isotopes are trapped and cooled by means of sophisticated magnetic traps in combination with laser cooling .
thus , the residual temperature and consequently the residual motion and expansion of the atomic clouds is reduced to the nk regime , where the cloud of individual atoms makes the transition to the bose - einstein condensed state .
the atoms are then released from the trap into free fall before being prepared in superpositions of internal states and split into partial waves by pulses of lasers in a well defined time sequence . by using a sequence of 3 laser pulses separated by a period of @xmath10 , a mach - zehnder interferometer in space - time
is formed .
the distribution of the internal atomic states is then detected either by fluorescence spectroscopy or by absorption imaging .
the distribution ratio of the internal atomic states provides the information about the phase shift which is proportional to the acceleration acting on the atoms .
the etvs parameter @xmath11 ( see mission objective 4 ) depends on the ratio of inertial mass @xmath12 to gravitational mass @xmath13 for the two isotopes and can be found from the measured accelerations @xmath14 as follows : @xmath15 where subscript 1 refers to physical parameters of the first isotope ( @xmath7 ) and subscript 2 to those of the second isotope ( @xmath8 ) .
+ the programmatics of european space atom interferometers , as proposed for ste - quest strongly builds on the different national and european - wide activities in this direction , in particular the space atom interferometer ( sai)@xcite and the nationally funded projects quantus@xcite and ice@xcite . in these activities innovative technologies are developed and tested in parabolic flights , drop tower campaigns and sounding rocket missions .
+ _ the science links _ comprise optical and microwave time & frequency links to compare the on - board clock with distributed ground clocks .
similarly , the links allow to compare distant ground clocks to one another using the on - board segment as a relay .
the microwave link ( mwl ) is designed based on re - use and further development of existing mwl technology for the aces mission@xcite , while implementing lessons learned from aces .
this applies to the flight segment and the ground terminals , as well as for science data post - processing issues down to the principles and concept for deploying and operating a network of ground terminals and associated ground clocks .
digital tracking loops with digital phase - time readout are used to avoid beat - note concepts .
different to aces and given the highly eccentric orbit , ka - band is used instead of ku - band , x - band instead of s - band for ionosphere mitigation , and the modulation rate is changed from 100 mchip / s to 250 mchip / s .
the mwl flight segment comprises four receive channels to simultaneously support comparisons with 3 ground terminals and on - board calibration .
the optical link baseline design relies on two tesat laser communication terminals ( lcts)@xcite on - board the satellite , both referenced to the space - clock molo signal , and the ground lcts to perform two time bi - directional comparisons .
the lcts operate at 1064 nm with a 5 w fiber laser , similar to the edrs lct@xcite , and a 250 mhz rf modulation .
a telescope aperture of 150 mm in space and 300 mm on ground was chosen .
a preliminary performance estimate has been made by extrapolation from existing measurements with more detailed investigations following now in dedicated link studies .
the optical links are found to be particularly sensitive to atmospheric disturbances and local weather conditions .
these particularly affect common view comparisons between two ground terminals which are typically located in less favorable positions concerning local cloud coverage . + _ the precise - orbit - determination ( pod ) equipment _ is required to determine the spacecraft position and velocity , which allows calculating theoretical predictions of the expected gravitational red - shifts and the relativistic doppler , and compare them to the measured ones .
our studies found that a gnss - receiver with a nadir - pointing antenna suffices to achieve the required accuracy@xcite .
additional corner - cube reflectors to support laser ranging measurements , such as also used for the giove - a satellite@xcite , may optionally be included as a backup and for validation of gnss - derived results .
_ the baseline orbit _ is highly elliptical with a typical perigee altitude of 700 km and an apogee altitude of 50000 km , which maximizes gravity gradients between apogee and perigee in order to achieve mission objective 1 ( earth gravitational red - shift ) .
the corresponding orbit period is 16h , leading to a 3:2 resonance of the ground track as depicted in fig.[fig : orbit ] left .
some useful orbit parameters are summarized in tab .
[ tab : baseline_orbit ] .
llllll perigee alt . & apogee alt . & orbit period & inclination & arg .
perigee & raan drift + 700 - 2200 km & @xmath16 50 000 km&16 h & @xmath17 & @xmath18 & @xmath19 + contrary to the old baseline orbit@xcite , where the argument of perigee was drifting from the north to the south during the mission , the argument of perigee for the new baseline orbit is frozen in the south . therefore , contact to the ground terminals located in the northern hemisphere can only be established for altitudes higher than 6000 km , which considerably impairs the modulation measurements of the earth gravitational red - shift ( measurement type b in table[tab : mission_objectives ] ) . on the pro side , for higher altitudes and in particular around apogee , common view links to several terminals and over
many hours can be achieved which greatly benefits the sun gravitational red - shift measurements ( type c ) .
the baseline orbit also features a slight drift of the right - ascension of the ascending node ( raan ) due to the @xmath20 term of the non - spherical earth gravitational potential@xcite , which leads to a total drift of approximately 30@xmath21 per year .
this drift being quite small , the orbit can be considered as nearly - inertial during one year so that seasonal variations of the incident solar flux are expected ( see also fig.[fig : solar_flux ] ) .
the major advantage of the new baseline orbit over the old one is a greatly reduced @xmath22v for de - orbiting , which allows reducing the corresponding propellant mass from 190 kg to 60 kg .
the reason for that is the evolution of the perigee altitude due to the 3-body interaction with the sun and the moon@xcite .
the perigee altitude starts low at around 700 km , goes through a maximum of around 2200 km at half time , and then drops back down below 700 km in the final phase , as depicted in fig.[fig : orbit ] right .
the low altitude at the end of mission facilitates de - orbiting using the 6 x 22-n oct thrusters during an apogee maneuver that is fully compliant with safety regulations . + _ the ground segment _ comprises 3 ground terminals , each equipped with microwave antennas for x- and ka - band communication as well as ( optional ) optical terminals based on the laser communication terminals ( lct ) of tesat@xcite .
the terminals must be distributed over the earth surface with a required minimum separation of five thousand kilometers between them .
this maximizes their relative potential difference in the gravitational field of the sun in pursuit of mission objective 2 .
additionally , the terminals must be in the vicinity of high - performance ground clocks ( distance @xmath23 km ) to which they may be linked via a calibrated glass - fibre network .
this narrows the choice of terminal sites to the locations of leading institutes in the field of time - and frequency metrology .
factoring in future improvements in clock technology and design , those institutes are likely able to provide clocks with the required performance of a fractional frequency uncertainty on the order of @xmath24 ) at the start of the mission , considering that current clock performance comes already quite close to the requirement@xcite .
accounting for all the aspects mentioned above , boulder ( usa ) , torino ( italy ) , and tokyo ( japan ) were chosen as the baseline terminal locations .
the spacecraft has an octagonal shape with a diameter of 2.5 m and a height of 2.7 .
it consists of a payload module ( plm ) , accommodating the instruments as well as science link equipment , and a service module ( svm ) for the spacecraft equipment and propellant tanks ( see fig .
[ fig : spacecraft_overview ] ) . the laser communication terminals ( lcts )
are accommodated towards the nadir end of opposite skew panels .
the nadir panel ( + z - face ) accommodates a gnss antenna and a pair of x - band / ka - band antennas for the science links . the telemetry & tele - commanding ( tt&c ) x - band low gain and medium gain antennas
are accommodated on y - panels of the service module .
the two solar array wings are canted at @xmath25 with respect to the spacecraft y - faces and allow rotations around the y - axis .
radiators are primarily placed on the + x/-x faces of the spacecraft and good thermal conductivity between dissipating units and radiators is provided through the use of heat - pipes which are embedded in the instrument base - plates .
a set of 2 x 4 reaction control thrusters ( rcts ) for orbit maintenance and slew maneuvers after the perigee measurement phase is used as part of the chemical propulsion system ( cps ) together with 1 x 6 orbit control thrusters ( octs ) for de - orbiting .
the latter is becoming a stringent requirement for future mission designs to avoid uncontrolled growth of space debris and reduce the risk of potential collisions between debris and intact spacecraft@xcite .
the micro - propulsion system ( mps ) is based on gaia heritage@xcite and primarily used for attitude control , employing a set of 2x8 micro - proportional thrusters to this purpose . as primary sensors of the attitude and orbit control subsystem ( aocs ) ,
the spacecraft accommodates 3 star trackers , 1 inertial measurement unit ( imu ) , and 6 coarse sun sensors . the payload module of fig .
[ fig : payload_module ] shows how the two instruments are accommodated in the well protected central region of the spacecraft , which places them close to the center - of - mass ( com ) and therefore minimizes rotational accelerations whilst also providing optimal shielding against the massive doses of radiation accumulated over the five years of mission duration .
the pharao tube of the atomic clock is aligned with the y - axis .
this arrangement minimizes the coriolis force which acts on the cold atomic cloud propagating along the vacuum tube axis when the spacecraft performs rotations around the y - axis .
these rotations allow preserving a nadir - pointing attitude of the spacecraft as required for most of the orbit ( see also section [ sec : aocs ] ) .
the sensitive axis of the atom interferometer must point in the direction of the external gravity gradient .
it is therefore aligned with the z - axis ( nadir ) and suspended from the spacecraft middle plate into the structural cylinder of the service module which provides additional shielding against radiation and temperature variations .
the total wet mass of the spacecraft with the launch adapter is approximately 2300 kg , including unit maturity margins and a 20% system margin .
the total propellant mass ( cps+mps ) is 260 kg , out of which 60 kg are reserved for de - orbiting .
the total power dissipation for the spacecraft is 2230 w , including all required margins , if all instruments and payload units operate at full power .
the spacecraft is designed to be compatible with all requirements ( including launch mass , envelope , loads , and structural frequencies ) for launch with a soyuz - fregat from the main esa spaceport in kourou .
the two instruments pose stringent requirements on attitude stability and non - gravitational accelerations which drive the design of the attitude and orbit control system ( aocs ) and the related spacecraft pointing strategy . to ensure full measurement performance
, the atom interferometer requires non- gravitational center - of - mass accelerations to be below a level of @xmath26 . among the external perturbations acting on the spacecraft and causing such accelerations , air drag is by far the dominant one .
it scales proportional to the atmospheric density and therefore decreases with increasing altitudes .
figure [ fig : drag_and_rotation ] ( left ) displays the level of drag accelerations for various altitudes and under worst case assumptions , i.e. maximal level of solar activity and maximal cross section offered by the spacecraft along its trajectory .
the results indicate that below altitudes of approximately 700 km drag accelerations become intolerably large .
we conclude that , for the baseline orbit , drag forces are generally below the required maximum level and therefore need not be compensated with the on - board micro - propulsion system , although such an option is feasible in principle . another requirement is given by the maximal rotation rate of @xmath27 during perigee passage when sensitive atom interferometry experiments are performed .
figure [ fig : drag_and_rotation ] ( right ) plots the spacecraft rotation rate ( black solid line ) against the time for one entire orbit , assuming a nadir - pointing strategy .
the plot demonstrates that the requirement ( red broken line ) would not only be violated by the fast rotation rates encountered at perigee , even for the comparatively slow dynamics at apogee the rotation rates would exceed the requirement by more than one order of magnitude .
we therefore conclude that atom interferometer measurements are incompatible with a nadir - pointing strategy throughout the orbit , requiring the spacecraft to remain fully inertial during such experiments . considering that atom interferometry is generally performed at altitudes below 3000 km
, the following spacecraft pointing strategy was introduced for ste - quest ( see figure [ fig : pointing_strategy ] ) : * * perigee passage : * for altitudes below 3000 km , lasting about @xmath28 each orbit , the atom interferometer is operating and the spacecraft remains inertial .
the atom interferometer axis is aligned with nadir at perigee and rotated away from nadir by 64 degrees at 3000 km altitude ( see figure [ fig : pointing_strategy ] ) .
the solar array driving mechanism is frozen to avoid additional micro - vibrations . * * transition into apogee : * for altitudes between 3000 km and 7000 km , right after the perigee passage , the spacecraft rotates such that it is nadir pointed again ( slew maneuver ) .
the solar arrays are unfrozen and rotated into optimal view of the sun . *
* apogee passage : * for altitudes above 7000 km the spacecraft is nadir pointed and rotational accelerations are generally much slower than during perigee passage . the solar array is allowed to rotate continuously for optimal alignment with the sun .
atom interferometer measurements are generally not performed . * * transition into perigee : * for altitudes between 7000 km and 3000 km , right after the apogee passage , the spacecraft rotates such that it is aligned with nadir at perigee .
the solar arrays are frozen . during the science operations at perigee and apogee
, the attitude control relies solely on the mps .
the slew maneuvers , i.e. @xmath29 pitch rotations , are performed by means of the cps in @xmath30 at a rate of @xmath31 .
the cps is also used to damp out residual rotational rates which are caused by solar array bending modes and fuel sloshing due to the fast maneuvers . when the perigee passage is entered , the mps is activated to further attenuate the residual rate errors coming from the minimum impulse bit and sensor bias ( ca .
@xmath32 ) , which is shown in fig.[fig : rate_errors]b .
closed - loop simulations , based on a dynamical computer model of the spacecraft in - orbit and simulated aocs equipment , demonstrated that the damping requires about 500 s. the complete transition , including maneuvers , lasts about @xmath33 . in order to mitigate the effects of external disturbances during the perigee passage
, an aocs closed - loop control has been designed that is based on cold - gas thrusters and includes roll - off filtering to suppress sensor noise while maintaining sufficient stability margins .
this allows compensating external torques from drag forces , solar radiation pressure and the earth magnetic field ( among others ) , yielding rotation rates well below the required @xmath27 after the initial damping time of approximately @xmath34 s , as can be seen in in fig.[fig : rate_errors]b .
the associated mean relative pointing error ( rpe ) is found to be smaller than @xmath35 in a 15 s time window corresponding to one experimental cycle .
note that the spacecraft rotation rates grow to intolerably large levels of several hundred @xmath36 , if the spacecraft is freely floating without control loops to compensate the environmental torques ( see fig.[fig : rate_errors]a ) .
for orbit maintenance and de - orbiting , use of the cps is foreseen .
is outside the plot scale.,scaledwidth=100.0% ] the problems associated with micro - vibrations lie at the heart of two key trades for our current spacecraft configuration .
one major cause of micro - vibrations are reaction wheels . to investigate the option of using them instead of the mps for attitude control , extensive simulations were performed with a finite element model of the spacecraft which provided the transfer functions from the reaction wheels to the atom interferometer interface .
based on this model and available measurement data of low - noise reaction wheels ( bradford w18e ) , the level of micro - vibrations seen at instrument interface could be compared against the requirements .
the results of the analyses are summarized in fig .
[ fig : micro_vibrations ] left which clearly shows a violation of the requirement for frequencies above 20 hz within the measurement band [ 1 mhz , 100 hz ] .
the major contribution to the micro - vibrations comes from the h1 harmonic when the disturbance frequency corresponds to the wheel rotation rate up to the maximum wheel rate of 67 hz .
further micro - vibrations are induced by the amplification of solar array flexible modes through the wheel rotation .
the violation of the requirement for the maximum level of tolerable micro - vibrations when using reaction - wheels , as found from our simulations , lead us to use a micro - propulsion system for attitude control in our baseline configuration . } $ ] .
note that for ease of comparison , requirements for the spectral density have been transformed into rms requirements by integrating over the measurement bandwidth.,scaledwidth=100.0% ] similar analyses were performed to determine the micro - vibrations induced by the solar array driving mechanism ( sadm ) on the molo interface , where a complex model of the sadm stepper motor together with the spacecraft finite - element - model ( fem ) with deployed solar arrays was combined in a matlab simulink environment .
the stepper motor generates two kinds of torque perturbations , one relating to the magnetic detent effect and the other relating to the commanded micro - stepping of the motor .
one particular resonance of the spacecraft structural model , corresponding to a torsional mode of the solar arrays around their rotation axis ( y - axis ) , was found to be the most critical one .
when this resonance was hit by a harmonic disturbance originating from the detent torque of the solar arrays driving mechanism , the resulting response displayed a low - frequency isolated resonance peak at the excitation frequency of 1.3 hz , with a few more peaks at higher frequencies due to harmonic excitation from the micro - stepping ( see fig .
[ fig : micro_vibrations ] right ) . restricting the solar array rotation rates to below 0.13
deg / s , the level of micro - vibrations is found to be fully compliant throughout the specified frequency range , as can be seen in figure [ fig : micro_vibrations ] . allowing for higher rates up to 0.08 deg / s , which are required to follow the fast spacecraft rotations with the solar array at low altitudes below 10000 km , we find that only one specific rotation rate at 0.044 deg / s leads to a violation of the requirement .
however , this is easily mitigated by quickly moving over the resonance , therefore avoiding to rotate the solar arrays at this particular rate . for earth gravitational red - shift measurements of type ( a ) ,
the measured frequency shift of the space - clock with respect to a ground clock is compared to the predicted frequency - shift . in order to compute the theoretical prediction ,
it is necessary to know the precise position and velocity of the space - clock ( and therefore the precise orbit ) , as both quantities enter the equation describing the de - phasing of the space - clock proper time @xmath37 with respect to the coordinate time @xmath38 , where the latter is referenced to an inertial earth - centered coordinate system .
this dependence is expressed in eq.[equ : proper_time ] , which is derived from a post - newtonian expansion to second order in @xmath39 @xcite .
@xmath40 where @xmath41 denotes the gravitational potential , @xmath42 the spacecraft position , and @xmath43 its velocity .
note that the first term in brackets corresponds to the gravitational time dilation whereas the second term corresponds to the special relativistic time dilation .
+ _ clock and link uncertainty : _ considering only the gravitational time dilation , the frequency shift @xmath44 between a space clock at apogee and a ground clock may then be written as @xmath45/c^2=6.2\times10^{-10}$ ] , where @xmath46 and @xmath47 are the earth radius and the radial distance to apogee , respectively .
it follows that the space - clock and the science links must have a fractional frequency uncertainty of better than @xmath48 to allow measuring the gravitational red - shift of the earth to the specified accuracy . from similar deliberations
one finds that the ground clock and science link performance must be even better , on the order of @xmath24 , to measure the gravitational red - shift of the sun to the specified accuracy .
this requires very long contact times between the spacecraft and ground terminals , on the order of @xmath49 , which is only achieved after integration over many orbits .
+ the above requirements can be directly translated into the required stability for the link , expressed in modified allan deviation@xcite , and into the frequency uncertainty for a space - to - ground or ground - to - ground comparison .
a detailed performance error budget was established to assess all contributions to potential instabilities .
for the microwave links we found that , provided the challenging requirements for high thermal stability are met , the aces design with the suggested improvements can meet the specifications .
the assessment of the optical links showed that a significantly better performance than the state - of - the - art is required .
a major issue for further investigation is the validity of extrapolating noise power spectral densities over long integration times in presence of atmospheric turbulence , which is required for comparing the estimated link performance to the specified link performance expressed in modified allan deviation . + _ orbit uncertainty : _ as the fractional frequency uncertainty is specified to be around @xmath48 for the space - clock , a target value of @xmath50 for the orbit uncertainty does not compromise the overall measurement accuracy .
although preliminarily a constant target value of 2 m in position and 0.2 mm / s in velocity accuracy has been derived in @xcite , these values can be significantly relaxed at apogee@xcite .
a final analysis will have to build on a relativistic framework for clock comparison as outlined in @xcite .
+ _ gnss tracking : _ the proposed method of precise orbit determination ( pod ) relies on tracking gps and galileo satellites with a gnss receiver supporting all available modulation schemes , e.g. bpsk5 , bpsk10 and mboc@xcite , and an antenna accommodated on the nadir panel . considering that gnss satellites broadcast their signal towards the earth and not into space , they become increasingly difficult to track at altitudes above their constellation ( 20200 km altitude ) , when the spacecraft can only receive signals from gnss satellites behind the earth . in order to increase the angular range of signal reception ,
the receiver is required to support signal side - lobe tracking .
this way , a good signal visibility for six hours during each orbit and for altitudes up to 34400 km can be achieved , with a corresponding pvt ( position - velocity - time ) error of 1.7 m at 1 hz output rate@xcite .
the transition maneuvers split the visibility region ( green arch in fig .
[ fig : gnss_visibility ] ) into separate segments , which must be considered separately in the orbit determination . for both regions
we find orbit errors less than 10 cm in position and less than 0.2 mm / s in velocity .
extrapolation of the spacecraft trajectory from the end of the visibility arch as far as apogee introduces large orbit errors ( @xmath51 ) due to large uncertainties in parameters describing the solar radiation pressure .
nonetheless , these errors are sufficiently small to achieve the specified frequency uncertainty for the space - clock measurement if one considers the arguments for relativistic clock comparison derived in @xcite . a brief discussion of the possibility to relax the ste - quest orbit accuracy close to apogee is also given in @xcite . finally , it shall be pointed out , that the accuracies obtained from gnss - based pod are sufficient to probe the flyby anomaly , an inexplicable momentum transfer which has been observed for several spacecraft careening around the earth .
these investigations can be performed without further modifications to the baseline hardware or mission concept , as described in ref .
the major focus of the assessment study was to define preliminary designs of the spacecraft subsystems in compliance with all mission requirements .
to this end , key aspects such as pointing strategy of spacecraft and solar arrays , variability of solar flux and temperature stability , measurement performance and pointing stability , instrument accommodation and radiative shielding , and launcher specifications , among others , must be reflected in the custom - tailored design features .
additional provisions for ease of access during integration and support of ground - based instrument testing on spacecraft level are also important considerations .
as we intend to give a concise overview in this paper , we will limit our discussion on a few distinctive design features .
+ _ thermal control subsystem ( tcs ) : _ the ste - quest payload dissipates a large amount of power which may total up to more than 1.7 kw ( including maturity and system margins ) if all payload equipment is active .
this represents a major challenge for the thermal system which can only be met through dedicated heat - pipes transporting the heat from the protected accommodation region in the spacecraft center to the radiator panels .
also problematic is the fact that the baseline orbit is not sun - synchronous and features a drift of the right - ascension of the ascending node ( raan ) , which leads to seasonally strongly variable thermal fluxes incident on the spacecraft from all sides .
this is mitigated by optimal placement of the radiators in combination with a seasonal yaw - rotation by @xmath52 to minimize the variability of the external flux on the main radiators .
the variation of the thermal flux during the 6 years of mission is plotted in fig.[fig : solar_flux ] ( left ) .
the time when the largest temperature fluctuations occur during one orbit , defined as the maximum temperature swing , is close to the start of the mission and coincides with the time of the longest eclipse period .
the additional use of heaters in proximity to the various payload units allows us to achieve temperature fluctuations of the payload units of less than @xmath53c , whilst average operating points consistently lie within the required range from @xmath54c to @xmath55c . during periods when only a minimum of solar flux is absorbed by the spacecraft , e.g. cold case of fig.[fig : solar_flux ]
, the heaters of the thermal control system may additionally contribute up to 150 w to the power dissipation .
these results were found on the basis of a detailed thermal model of the spacecraft implemented with esatan - tms , a standard european thermal analysis tool for space systems@xcite .
+ _ electrical subsystem & communications : _ similar to the tcs , the design of the electrical subsystem is driven by the highly variable orbit featuring a large number of eclipses ( more than 1800 during the mission with a maximum period of 66 minutes ) , the high power demand of the instruments , and the satellite pointing strategy in addition to the required stability of the spacecraft power bus . in order to meet the power requirements , two deployable solar arrays at a ca
nt angle of 45 degrees are rotated around the spacecraft y - axis .
this configuration , in combination with two yaw - flips per year , ensures a minimum solar flux of approximately @xmath56 on the solar panels ( see fig.[fig : solar_flux ] right ) , which generates a total power of 2.4 kw , including margins .
the option of using a 2-axis sadm has also been investigated in detail and could be easily implemented .
however , it is currently considered too risky as such mechanisms have not yet been developed and qualified in europe and they might cause problems in relation to micro - vibrations . as far as telemetry and tele - commanding ( tt&c ) is concerned , contact times to a northern ground station ( cebreros ) are close to 15 h per day on average for the baseline orbit , which is easily compliant with the required minimum of 2 h per day .
this provides plenty of margin to download an estimated 4.4 gbit of data using a switched medium gain antenna ( mga ) and low gain antenna ( lga ) x - band architecture which is based on heritage from the lisa pathfinder mission@xcite .
the switched tt&c architecture allows achieving the required data volume at apogee on the one hand , whilst avoiding excessive power flux at perigee to remain compliant with itu regulations on the other hand .
+ _ mechanical subsystem , assembly integration and test ( ait ) and radiation aspects : _ the spacecraft structure is made almost entirely from panels with an aluminum honeycomb structure ( thickness 40 mm ) and carbon - fibre reinforced polymer ( cfrp ) face sheets , which is favorable from a mechanical and mass - savings perspective .
a detailed analysis based on a finite - element - model ( fem ) of the spacecraft structure revealed a minimum eigenfrequency of 24 hz in transverse and 56 hz in axial direction , well above the launcher requirements of 15 hz and 35 hz , respectively .
the first eigenmode of the atom interferometer physics package , corresponding to a pendulum mode in radial direction , was found at 67 hz .
the spacecraft structure is designed to optimize ait and aspects of parallel integration . to this end
, the two instruments can be integrated and tested separately before being joined on payload module level without the need to subsequently remove any unit or harness connection .
the heat - pipe routing through the spacecraft structure allows functional testing of both instruments on ground by operating the heat - pipes either in nominal mode or re - flux mode ( opposite to the direction of gravity ) . as an important feature in the integration process
, payload and service module components are completely separated in their respective modules up to the last integration step , when they are finally joined and their respective harness connected on easily accessible interface brackets .
the second major aspect in structural design and instrument accommodation has been to minimize radiation doses of sensitive instrument components .
as the spacecraft crosses the van allen belt twice per orbit and more than 5000 times during the mission , it sustains large radiation doses from trapped electrons and solar protons .
dedicated design provisions and protected harness routing ensure that the total ionizing doses ( tids ) are below 30 krad for all instrument units and below 5 krad for the sensitive optical fibres . the non - ionizing energy loss , which is primarily caused by low - energy protons , is found to be well below @xmath57 , expressed in terms of 10 mev equivalent proton fluence .
these results were obtained from the total dose curves accumulated over the orbit during the mission in combination with a dedicated sectoring analysis performed with a finite - element - model of the spacecraft .
ste - quest aims to probe the foundations of einstein s equivalence principle by performing measurements to test its three cornerstones , i.e. the local position - invariance , the local lorentz - invariance and the weak equivalence principle , in a combined mission .
it complements other missions which were devised to explore the realm of gravity in different ways , including the soon - to - be - launched aces@xcite and microscope@xcite missions and the proposed step@xcite mission .
the ste - quest measurements are supported by two instruments .
the first instrument , the atomic clock , benefits from extensive heritage from the aces mission@xcite , which therefore reduces associated implementation risks .
the other instrument , the atom interferometer , would be the first instrument of this type in space and poses a major challenge which is currently met by dedicated development and qualification programs .
the mission assessment activities summarized in this paper yielded a spacecraft and mission design that is compliant with the challenging demands made by the payload on performance and resources .
several critical issues have been identified , including low maturity of certain payload components , potential unavailability of sufficiently accurate ground clocks , high sensitivity of the optical links to atmospheric distortions and associated performance degradation , and high energy dissipation of the instruments in addition to challenging temperature stability requirements .
however , none of these problems seems unsurmountable , and appropriate mitigation actions are already in place .
the work underlying this paper was performed during the mission assessment and definition activities ( phase 0/a ) for the european space agency ( esa ) under contract number 4000105368/12/nl / hb .
the authors gratefully acknowledge fruitful discussions and inputs from the esa study team members , in particular martin gehler ( study manager ) , luigi cacciapuoti ( project scientist ) , robin biesbroek ( system engineer ) , astrid heske , ( payload manager ) , florian renk ( mission analyst ) , pierre waller , ( atomic clock support ) , and eric wille ( atom interferometer support ) .
we also thank the astrium team members for their contributions to the study , in particular felix beck , marcel berger , christopher chetwood , albert falke , jens - oliver fischer , jean - jacques floch , sren hennecke , fabian hufgard , gnter hummel , christian jentsch , andreas karl , johannes kehrer , arnd kolkmeier , michael g. lang , johannes loehr , marc maschmann , mark millinger , dirk papendorf , raphael naire , tanja nemetzade , bernhard specht , francis soualle and michael williams .
the authors are grateful for important contributions from our project partners mathias lezius ( menlo systems ) , wolfgang schfer , thorsten feldmann ( timetech ) , and sven schff ( astos solutions ) . finally , we are indebted to rdiger gerndt and ulrich johann ( astrium ) for regular support and many useful discussions .
martin gehler , luigi cacciapuoti , astrid heske , robin biesbroek , pierre waller , and eric wille .
the esa ste - quest mission study - space mission design to test einstein s equivalence principle . in _ proceedings of the aiaa space
2013 conference , paper number : 2013.5464 , doi : 10.2514/6.2013 - 5464_. american institute of aeronautics and astronautics , 2013 .
s. schiller and et al .
pace time explorer and quantum equivalence principle space test ( ste - quest ) : proposal in response to esa m3 call 2010 for the cosmic vision programme , 2010 .
available at _ http://www.exphy.uni - duesseldorf.de / publikationen/2010/ste - quest_final.pdf_. mp plattner , u hugentobler , d voithenleitner , m hseinze , v klein , k kemmerle , and s bedrich . optical clock technology for optimized satellite navigation . in _
eftf 2010 , the 24th european frequency and time forum , 13 - 16 april 2010 _ , volume 1 , page 2 , 2010 .
m lezius , k predehl , w stower , a turler , m greiter , ch hoeschen , p thirolf , w assmann , d habs , alexander prokofiev , et al .
radiation induced absorption in rare earth doped optical fibers .
, 59(2):425433 , 2012 .
c schubert , j hartwig , h ahlers , k posso - trujillo , n gaaloul , u velte , a landragin , a bertoldi , b battelier , p bouyer , et al .
differential atom interferometry with rb87 rb85 for testing the uff in ste - quest . , 2013 .
d aguilera , h ahlers , b battelier , a bawamia , a bertoldi , r bondarescu , k bongs , p bouyer , c braxmaier , l cacciapuoti , et al .
-test of the universality of free fall using cold atom interferometry . , 2013 .
fiodor sorrentino , kai bongs , philippe bouyer , luigi cacciapuoti , marella de angelis , hansjoerg dittus , wolfgang ertmer , antonio giorgini , jonas hartwig , matthias hauth , et al . a compact atom interferometer for future space missions .
, 22(4):551561 , 2010 .
guillaume stern , baptiste battelier , rmi geiger , gal varoquaux , andr villing , frdric moron , olivier carraz , nassim zahzam , yannick bidel , w chaibi , et al .
light - pulse atom interferometry in microgravity .
, 53(3):353357 , 2009 .
mp hess , j kehrer , m kufner , s durand , g hejc , h fruhauf , l cacciapuoti , r much , and r nasca .
status and test results . in _
frequency control and the european frequency and time forum ( fcs ) , 2011 joint conference of the ieee international _ , pages 18 .
ieee , 2011 .
a seidel , mp hess , j kehrer , w schafer , m kufner , m siccardi , l cacciapuoti , i aguilar sanches , and s feltham .
the aces microwave link : instrument design and test results . in _
frequency control symposium , 2007 joint with the 21st european frequency and time forum .
ieee international _
, pages 12951298 .
ieee , 2007 .
m gregory , f heine , h kmpfner , r meyer , r fields , and c lunde .
laser communication terminal performance results on 5.6 gbit coherent inter satellite and satellite to ground links . in _ international conference on space optics _ , volume 4 , page 8 , 2010 .
gerald hechenblaikner , jean - jacques floch , francis soualle , and marc - peter hess . -based precise orbit determination for a highly eccentric orbit in the ste - quest mission . in _ proceedings of the 26th international technical meeting of the satellite division of the institute of navigation ( ion gnss 2013 ) , nashville , september 2013_. institute of navigation , 2013 .
c urschl , g beutler , w gurtner , u hugentober , and m ploner .
orbit determination for giove - a using slr tracking data . in
_ extending the range .
proceedings of the 15th international workshop on laser ranging _ , pages 4046 , 2008 . till rosenband , db hume , po schmidt , cw chou , a brusch , l lorini , wh oskay , re drullinger , tm fortier , je stalnaker , et al . frequency ratio of al+ and hg+ single - ion optical clocks ; metrology at the 17th decimal place . , 319(5871):18081812 , 2008 .
g matticari , g noci , l ceruti , l fallerini , a atzei , c edwards , g morris , and k pfaab .
old gas micro propulsion for european science missions : status on development and realization activities for lisa pathfinder , microscope and forthcoming euclid , following the successful delivery of the first fm for gaia program .
49th aiaa / asme / sae / asee joint propulsion conference & exhibit .
t. j. sumner , j. anderson , j .-
blaser , a. m. cruise , t. damour , h. dittus , c. w. f. everitt , b. foulon , y. jafry , b. j. kent , n. lockerbie , f. loeffler , g. mann , j. mester , c. pegrum , r. reinhardt , m. sandford , a. scheicher , c. c. speake , r. torii , s. theil , p. touboul , s. vitale , w. vodel , and p. w. worden . . ,
39:254258 , 2007 . | ste - quest is a fundamental science mission which is considered for launch within the cosmic vision programme of the european space agency ( esa ) .
its main scientific objectives relate to probing various aspects of einstein s theory of general relativity by measuring the gravitational red - shift of the earth , the moon and the sun as well as testing the weak equivalence principle to unprecedented accuracy . in order to perform the measurements ,
the system features a spacecraft equipped with two complex instruments , an atomic clock and an atom interferometer , a ground - segment encompassing several ground - terminals collocated with the best available ground atomic clocks , and clock comparison between space and ground via microwave and optical links .
the baseline orbit is highly eccentric and exhibits strong variations of incident solar flux , which poses challenges for thermal and power subsystems in addition to the difficulties encountered by precise - orbit - determination at high altitudes .
the mission assessment and definition phase ( phase - a ) has recently been completed and this paper gives a concise overview over some system level results . |
aberration and retardation effects ( a / r hereafter ) can be observed in a pulsar profile as a shift of the position of conal components with respect to the core component towards earlier longitudinal phases ( see for e.g. malov & suleimanova ( 1998 ) , gangadhara & gupta 2001 , g&ga hereafter ) .
such effects should occur if different components are emitted at different distances from the pulsar surface ( emission radii ) , as well as from the pulsar spin axis .
aberration is caused by bending of radiation beam due to the polar cap rotation , while retardation is based on a path difference for radiation from different conal emission regions to reach an observer .
if we assume that the emission from the core component arises relatively close to the star surface , then it should not be strongly affected by either of the two above mentioned effects .
this will be our initial assumption . to determine a / r shifts the pulsar profile has to meet certain requirements .
it has to be a high resolution profile with high signal to noise ( s / n ) ratio .
the core and the conal components have to be clearly identified within the profile .
multifrequency data is recommended , so one can follow the profile evolution throughout all frequencies , which can help to identify different emission components .
when values of a / r shifts are determined , then the heights of the radiation emission region ( emission altitudes hereafter ) can be calculated ( see g&ga and dyks et .
al 2004 ) .
it is believed that at different frequencies the emission arises at different heights above the pulsar surface ( kijak & gil 1998 , 2003 and mitra & rankin 2002 ) .
the results of this analysis can be used to verify the existence of a radius to frequency mapping .
all observational limits for emission altitude hence can be crucial for understanding the physical mechanism of generation of pulsar coherent radio emission .
the relativistic beaming model initially proposed by blaskiewicz , cordes & wasserman ( 1991 , bcw hereafter ) clearly demonstrated that aberration and retardation effects play an important role in pulsars .
this study was primarily based on evidence which followed from the effects of a / r as seen in the inflexion point of the pulsar s polarisation position angle ( hereafter pa ) traverse , which lags the midpoint of the total intensity profile centre .
a similar effect of a / r was reported by g&ga and gupta & gangadhara ( 2003 , g&gb hereafter ) in a handful of pulsars where the core emission was seen to lag behind the profile centre . in this paper
we have undertaken a careful study to establish the a / r effect observed by g&ga and g&gb for a large sample of pulsars observed at multiple frequencies .
most of the data are new observations from the giant meterwave radio telescope ( gmrt hereafter ) and the arecibo observatory ( ao hereafter ) .
we have also used some archival data from the european pulsar network ( epn hereafter ) archive . in section ( [ sec2 ] )
we discuss various methods used to find emission heights in pulsars , in section ( [ sec3 ] ) we discuss various factors affecting a / r measurements in pulsars and section ( [ sec4 ] ) deals with the observation and data analysis methods used in this paper . as a result of our analysis presented in section ( [ sec5 ] )
we found that out of 23 pulsars in our sample 7 clearly show the a / r effect , 12 show a clear tendency towards this effect , while the remaining 4 are counter examples .
however , as argued in section ( [ sec3 ] ) , all problematic cases ( pulsar profiles at all or some frequencies not showing the a / r effect ) can be attributed to a number of effects like incorrect identification of the core component or missing conal emission .
we can conclude that a / r effects are seen to operate in pulsars , which we discuss in section ( [ sec6 ] ) .
radio emission heights in pulsars are primarily obtained by two methods ; the geometrical method and heights estimation based on a / r effects . here
we briefly mention the essential ingredients of the methods used , and a detailed discussion of the various methods used can be found in mitra & li ( 2004 ) .
radio emission geometry is determined by several parameters : @xmath0 an inclination angle of the magnetic dipole with respect to the rotation axis , @xmath1 the minimum angle between an observer s line of sight and magnetic axis ( impact angle ) , @xmath2 an opening angle of the radio emission beam , @xmath3 a radial distance of the radio emission region measured from the centre of the neutron star ( emission altitude ) .
the opening angle @xmath2 of the pulsar beam corresponding to the pulse width @xmath4 is given by : @xmath5 where @xmath6 , @xmath1 , @xmath4 and @xmath2 are measured in degrees ( gil et al .
the opening angle @xmath2 is the angle between the pulsar magnetic axis and the tangent to magnetic field lines at points where the emission corresponding to the apparent pulse width @xmath4 originates . for dipolar
field lines : @xmath7 ( gil & kijak 1993 ) , where @xmath8 is a mapping parameter which describes the locus of corresponding field lines on the polar cap ( @xmath9 at the pole and @xmath10 at the edge of the polar cap ) , @xmath11 is the distance of the given magnetic open field line from the dipolar magnetic axis ( in cm ) , @xmath12 is the polar cap radius ( in cm ) and @xmath13 is the pulsar period in seconds .
the radio emission altitude can be obtained using eqn .
( [ rho ] ) : @xmath14 in this equation parameter @xmath10 is used which corresponds to the bundle of last open magnetic field lines .
kijak and gil ( 1997 ) also derived a semi empirical formula for emission height which was slightly modified by kijak & gil ( 2003 ) by using larger number of pulsars and broadening the frequency coverage in their analysis .
they estimated the emission heights for a number of pulsars at several frequencies using the geometrical method ( eqs . [ rho_1 ] and [ r_geo ] ) based on the above assumption that the low intensity pulsar radiation ( corresponding to the profile wings ) is emitted tangentially to the bundle of the last open dipolar field lines . in the method
, they used low
intensity pulse width data at about 0.01 % of the maximum intensity ( kijak & gil 1997 and kijak & gil 2003 ) .
the most recent formula has a form @xmath15 where @xmath16 is observing frequency in ghz and @xmath17 is the period derivative .
the formula @xmath18 sets lower and upper boundaries for emission height values in the statistical sense . for calculating the emission heights one can take a different approach .
the a / r effects have been observed in the total intensity profiles as discussed by malov & suleimanova ( 1998 ) , g&ga , g&gb , dyks et al ( 2004 ) . in turn , they have been used to calculate radio emission heights .
observationally the a / r effects are obtained by measuring the positions of the conal emission components and identifying the central core emission component .
the shape of the observed profile depends on the structure of radio beam , which is thought to be composed of a central core emission surrounded by nested cones ( rankin 1983 , mitra & deshpande 1999 ) .
the line of sight passing through the central region of this beam will hence cut the cone emission at the edges including the central core region . following this ,
pulsar profiles can be classified due to the number and type of visible components ( rankin 1983 ) .
if we assume that in a certain profile the core component and both the conal components are present , we can define the phase difference : @xmath19 where @xmath20 and @xmath21 are phase positions ( in degrees ) of the peaks of the leading and trailing components with respect to the core phase ( @xmath22 ) .
following the refinement of g&ga by dyks et al .
( 2004 ) one can estimate the radio emission altitude and @xmath1 , and the corrections in obtaining emission heights turns out to be of the order of 2% to 10% for normal pulsars .
we have however not used this formula since all the pulsars in our sample are normal pulsars and , further , the error in estimating emission heights are often more than 10% . ] : @xmath23 where @xmath24 ( where @xmath25 is speed of light ) is the light cylinder radius .
the a / r method has an edge over the geometrical method since it only depends weakly on the geometrical angles of the star which are difficult to find .
generally the emission altitudes found using both these methods are of the order of few hundred kilometers above the surface of the neutron star . in this paper
we use equation ( [ r_del ] ) for estimating emission altitudes following the method outlined by g&ga .
the various factors affecting usage of this method are discussed in section ( [ sec3 ] ) .
the crucial aspect for estimating radio emission heights using the a / r technique is to identify the location of the meridional plane i.e. the plane containing the magnetic axis and the pulsar rotation axis .
generally polarisation observations have been used to identify the meridional plane as per the rotating vector model ( rvm ) proposed by radhakrishnan & cooke ( 1969 ) . according to the rvm the variation of the pa of the linear polarisation across the pulse reflects the underlying structure of the dipolar magnetic field if one is located on the pulsar frame . in this case the steepest gradient or the inflexion point of the pa traverse is the point contained in the meridional plane , and should be coincident with the midpoint of the pulse profile .
however in the observers frame , as was pointed out by bcw , dyks et al ( 2004 ) , the a / r effects causes the pa inflexion point to appear delayed w.r.t .
the midpoint of the pulse profile .
the magnitude of the delay is proportional to the height of emission from the neutron star surface and inversely proportional to the pulsar period .
hence in the observer s plane the inflexion point of the pa traverse is no longer the point containing the meridional plane , rather the meridional plane lies somewhere in between the profile midpoint and the inflexion point of the pa traverse .
we use the method employed by g&ga , b to find emission heights due to a / r effects using eq .
if the mid point of the outer conal peaks leads the core peak , it would imply that the conal emission is higher than the core emission and the shift can be used to find emission heights w.r.t .
this argument however assumes that the meridional plane passes through the peak of the core emission .
while this assumption is not rigorously verified , there are few pulsars which have been used to justify this argument .
one such pulsar is psr b1857@xmath2626 ( see fig .
[ b1857 ] ) where the peak of the core , the inflexion point and the sign changing circular , are all close to each other ( see mitra & rankin 2008 for a detailed discussion ) .
if one assumes that the core emission is curvature radiation which fills the full polar cap as was suggested by rankin ( 1990 ) , then the meridional plane would lie in the plane containing the inflexion point , peak of the core and sign changing circular .
although a critical look at psr b1857@xmath2626 reveals that the peak of the core is delayed by a degree or so w.r.t .
the inflexion point ( see mitra & rankin 2008 ) .
mitra & li ( 2004 ) have shown that for a few other pulsars the peak of the core lags the inflexion point by about a degree .
the shifts expected due to a / r effects for slow pulsars are also often small ( e.g. for emission altitude of 300 km and a 0.7 sec pulsar , the expected shift is of the order of a degree ) and is comparable to core shifts seen w.r.t . the inflexion point .
hence , it is important to note that to demonstrate the a / r effects as suggested by g&ga and g&gb , the meridional plane should pass through the peak of the core emission ( which is currently not verified ) and the conal emission should be rising higher than the core emission .
identification of the core emission is not trivial in pulsars . under the core
cone model of pulsar beaming the core is centrally located and surrounded by nested cones .
this implies that the core is located in the centre of a pulse profile , which is very close to the inflexion point of the pa .
since the impact parameter needs to be small to cut through the core emission region in a pulsar , the corresponding pa traverse as per the rvm should show a significant rotation across the pulse .
core emission is often associated with a sign changing circular polarisation and has a steeper spectral index than the conal emission .
there are situations where several core like emission characteristics are seen in a pulse component .
for example in fig ( [ b1831 ] ) psr b1831@xmath2604 has a bright central component which is identified as core emission by rankin ( 1993a , 1993b ) .
this component has a steep spectral index and a sign changing circular as seen in gould & lyne s ( 1998 , gl98 hereafter ) multi frequency polarisation profiles .
on the other hand the pa traverse is flat , showing no significant rotation ( except 90@xmath27 jumps ) and the profile is extremely wide ( @xmath28 150@xmath27 ) .
such emission properties are consistent with the observer crossing the pulsar emission beam along the conal emission zone and the observer is likely to miss the core emission .
hence for psr b1831@xmath2604 , the bright centrally located component can be interpreted as conal emission rather than core emission .
further there are situations where the pa below the core emission can deviate from the rvm model ( see mitra , rankin & gupta 2007 ) , and hence the location of the inflexion point is difficult to determine .
a careful examination using multi frequency single pulse polarisation data is hence essential to find the location of the core emission . for our data set
we have used the gl98 polarisation data sets wherever possible to find the location of the core emission which we discuss in the appendix . however incorrect identification of the core
would clearly affect demonstration of the a / r effect . in several cases
the central emission close to the core is weaker compared to the outer cones
. methods employed to find the peak of the core ( like the gaussian fitting scheme discussed in section [ sec4 ] ) have problems in establishing the location of the core component ( like psr b1738@xmath2608 ) , and hence cause complication in discerning the a / r effect .
subpulse modulation ( or drifting ) provides another means to identify core and conal components in pulsars .
the conal components tends to show subpulse drifting seen as a spectral feature in phase resolved fluctuation spectra , while in the core emission this phenomenon is generally absent ( rankin 1986 ) .
it is sometimes possible that the central component of the pulsar shows drifting properties , in which case the central component can not be core emission .
we have verified our sample pulsars for such drifting of the central component based on the dataset of pulsar subpulse modulation published by weltevrede et al .
( 2006 , 2007 ) .
three pulsars from our sample show central component modulation feature out of which two pulsars psr b1508 + 55 and b1917 + 00 has a broad modulation feature which is consistent with core emission , while psr b1944 + 17 shows clear drifting properties making this pulsar unsuitable for a / r analysis . in several pulsars
only a part of the emission cone is visible and these pulsars were identified as partial cones by lyne & manchester ( 1988 ) .
the pa inflexion point in these pulsars lies towards one edge of the profile .
this kind of pulsar however poses a problem for finding the a / r effect , since estimation of the midpoint of the total intensity profile is incorrect due to missing emission .
one such example in our sample is perhaps psr b1859 + 03 as shown in fig ( [ b1859 ] ) .
the pa inflexion point coincides with the peak of the trailing component and has a sign changing circular , which is suggestive of core emission , however there is no emission component seen in the trailing part of the profile .
single pulse study by mitra , rankin & sarala ( 2008 ) recently has shown that the missing partial cone regions occasionally flare up , and these flared profiles can be used to demonstrate the a / r effect . 26 polarisation profile .
the upper panel of the plot shows the intensity ( solid line ) , linear polarisation ( dotted line ) and the circular polarisation ( dashed line ) .
the bottom panel shows the pa with error bars .
this profile is an example where the core component , the inflection point and the sign changing circular are close to each other ( see section 3 for details).,height=245 ] 04 polarisation profile ( same as in fig [ b1857 ] ) .
although this pulsar shows the sign change in circular it has a flat pa and a wide profile . hence it is likely that the central bright emission component is not a core ( see section 3 for details),height=245 ] ) .
this pulsar has a clear sign changing circular indicating the location of the core emission .
however the emission towards the trailing parts of the profile seems to be missing ( see section 3 for details).,height=245 ]
for our analysis we selected profiles of 28 pulsars from three sources ( see table [ tabelapulsarow ] ) : gmrt , ao and epn .
the gmrt ( swarup et al . 1991 ) operating at meter wavelengths is an array of 30 antennas with alt
az mounts , each of 45 meter diameter , spread out over a 25 km region located 80 km north of pune , india .
it is designed to operate at multiple frequencies ( 150 , 235 , 325 , 610 and 1000 1450 mhz respectively ) having a maximum bandwidth of 32 mhz split into upper and lower sidebands of 16 mhz each .
it is primarily an aperture synthesis interferometer as well as a phased array instrument .
the gmrt observations reported in this paper are done in the phased array mode of operation at rf 225 and 325 mhz , respectively , using the upper side band and a bandwidth of 16 mhz . in the phased array mode of the gmrt ( sirothia , s. 2000 ,
gupta et al . 2000 ) , signals from any selected set of the antennas are added together after an appropriate delay and phase compensation to synthesise a single , larger antenna with a narrower beam .
signals of 16 mhz from each side band of each antenna are available across 256 channels which can be summed in the gmrt array combiner .
the summed voltage signal output from the phased array is then processed like any other single dish telescope signal , by the pulsar backend . at the frequencies of interest here , gmrt has two linearly polarised outputs for each antenna which are further mixed in a quadrature hybrid to produce two circular polarised signals which go further into the front end electronics and rest of the signal chain .
the raw signals are finally post integrated by the pulsar backend and the data was recorded with the final sampling time of 0.512 msec . the time resolution for all the analysis done in this paper is the pulsar period divided by the sampling time .
all the pulsars were observed for about 2000 pulse periods to obtain stable pulse profiles . the off line analysis involved interference rejection , dedispersion and folding to obtain the average pulse profiles .
the ao data set is from observations at the arecibo observatory , carried out between december 2001 and june 2002 .
the pulsars were observed at different frequencies using the following receivers : 430 mhz , l band ( 1175 mhz ) , s low ( 2250 mhz ) , s high ( 3500 mhz ) , c band ( 48505000 mhz ) .
the pulsar back
end used most frequently was the wapp ( wideband arecibo pulsar processor ) , though the pspm ( penn state pulsar machine ) was also used for some of the observations .
the bandwidths of the observations were typically 25/50 mhz at 430 mhz and 100 mhz for all the higher frequency observations .
the time resolution of the wapp raw data was typically 128 , 256 or 512 microsec .
the data were analysed using the sigproc analysis package to get the average folded profiles .
the data from the epn archive were mostly obtained by gl98 using 76m lovell telescope at jodrell bank observatory , except b1039@xmath2619 at 4850 mhz , which was obtained by kijak et al .
( 1997 ) using effelsberg observatory . generally ,
resolution of epn data is around 1@xmath29 , which is much lower than that of gmrt or ao data . [
cols="^,^,^,^,^,^,^,^,^",options="header " , ] our analysis followed these steps : ( 1 ) identification of core components in profiles .
we have used several established methods for discriminating the core components : ( i ) quasi centrality of the component flanked by one or two pairs of conal components , ( ii ) associating the core component with the inflexion point of the polarisation position angle traverse and sense reversing signature of circular polarisation under the core component ( although the polarimetry was not always available in our data ) , ( 2 ) identification of conal pairs as leading and trailing components , primarily based on symmetric location of components on either side of the core one , ( 3 ) estimation of phase shifts according to eqn .
( [ delta_phi ] ) , and ( 4 ) estimation of the emission heights according to eqn .
( [ r_del ] ) .
this simple procedure is not always straightforward in practice .
there are cases when components merge ( especially as a function of the observing frequency ) and their positions are not clearly determined . in these cases we decided to use a gaussian fitting method introduced by kramer et .
al ( 1994 ) . for consistency
, we applied this method to all our profiles , even those in which all components were clearly distinguishable .
the gaussian fitting procedure was as follows : fix gaussian parameters of identifiable components ( position , amplitude , width ) , put additional gaussian with some initial parameters and allow the fitting method to find the best position , amplitudes and widths . for detailed information about gaussian fitting method
see kramer et .
al ( 1994 ) .
after applying this method to unresolved components we obtained a full and consistent set of component positions to be used in further analysis .
our preliminary data set was reduced to 23 pulsars , because it was hard to perform a reasonable fitting on 5 pulsars ( b0950 + 08 , b1859 + 03 , b1946 + 35 , b2020 + 28 , b2315 + 21 ) due to either low s / n or problems with determining positions of core and/or conal components , even using a gaussian fitting method .
the determined positions of all components in a pulse profile , using the already mentioned gaussian fitting procedure , were used for a further analysis .
firstly , we estimate the phase shifts due to the a / r effect , and find the emission height @xmath30 using eqn .
( [ r_del ] ) . for calculating the phase
shifts using eqn .
( [ delta_phi ] ) , we first identify the leading and trailing component of a cone of emission using the location of the peak of the gaussian fitted component as @xmath20 and @xmath21 ( this profile measure is often called as peak to peak component separation ( ppcs ) , see for e.g. g&ga and g&gb , dyks et al .
2004 , mitra & li 2004 ) .
correspondingly we denote the phase shift as @xmath31 . wherever possible ,
@xmath31 was estimated for the inner and outer cones , separately . to estimate errors of single phase position measurements we used the formula introduced by kijak & gil ( 1997 ) : @xmath32 where @xmath33 is the observing resolution , @xmath34 is the root mean square of total intensity , and i is the measured signal level .
since @xmath4 estimations need two measurements ( left and right side of profile at particular intensity level ) and @xmath35 estimations need three measurements ( @xmath36 , @xmath37 and @xmath38 ) we took @xmath39 @xmath40 as an error for @xmath4 and @xmath35 respectively .
the best example of a pulsar whose apparent components are very well reproduced by the gaussian fitting method is psr b1237 + 25 presented in fig .
( [ b1237 ] ) at a number of frequencies .
all b1237 + 25 profiles are fitted perfectly with 5 gaussian components , being an ideal example of applicability of the gaussian fitting technique used in this pulsar .
this pulsar clearly shows the a / r effects at all frequencies , consistently over the spectrum from about 5 ghz to about 0.16 ghz ( see also table a.5.2 in section a.5 in the on line appendixchriss / krzeszowski_2009_appendix.ps ] ) .
the profiles are arranged in panels with ascending frequency .
the original profiles are plotted with solid lines , while the fitted gaussian components are drawn with dashed lines .
the profiles are aligned with respect to the peak of the central gaussian component identified as the core component .
all the profiles of pulsar b1237 + 25 show a clear a / r shift .
srostlik & rankin ( 2005 ) have found that pa steepest gradient traverse lags the sign changing circular polarisation zero point by about 0.4@xmath29 .
however if one takes this value into account , b1237 + 25 still shows the a / r shift . our main results , which will be discussed later on , show the commonness of a / r effects in pulsar profiles . within our sample of 23 pulsars
7 show clear a / r effects , 12 of them ( doubtful cases ) show a clear tendency towards this effect , while the remaining 4 are so called counter examples .
when it comes to single @xmath41 measurements we found that 45 measurements show negative shifts , which is expected for a / r effects , 23 measurements are doubtful within uncertainties , while remaining 19 show opposite shifts .
however taking into account only the outer cones ( where more than one cone is seen ) reveals that 30 shift measurements show a / r effects , 18 are doubtful cases whereas 14 show no a / r effects at all . for the majority of non
a / r cases , the problem lies mostly in misidentifying core components .
some of the cases are partial cones which also makes it difficult to measure a / r shift properly . for a detailed discussion of various issues regarding
a / r calculations see section ( [ sec3 ] ) . for a discussion of the results
see section ( [ sec6 ] ) .
note that all individual results are available in the form of figures and detailed tables in the electronic form as an appendix to this paper .
observationally , the peak of the core component lagging the centre of the outer conal component has always been considered as an evidence for a / r effects operating in pulsars .
those effects have been demonstrated in a few pulsars so far : psr b0329 + 54 by malov & suleimanova ( 1998 ) and g&ga between 103 mhz to 10.5 ghz and psr s b0450@xmath2618 , b1237 + 25 , b1821 + 05 , b1857@xmath2626 , b2045@xmath2616 and psr b2111 + 46 at 325 mhz by g&gb . in this work ,
our aim was to verify the a / r effects for a large sample of pulsars .
while our sample consists of 23 pulsars , we note that our analysis is aided by using multi frequency measurements , where each frequency serves as an independent data set . hence , although our sample consists of three pulsars common with g&gb , the measurements are done at different frequencies and can be treated as independent check for their results . in our data set of 62 pulse profiles measured , the a / r effects
are clearly seen in 30 profiles .
( [ hist ] ) shows the distribution of the measured phase shifts @xmath31 for our sample of pulsars ( for the data see the appendix ) .
the tendency for the negative shift , which is the signature for a / r effect , is clearly visible . out of the four counter
examples noted in our sample , the most outstanding one is psr b1831@xmath2604 ( fig . [ b1831 ] ) .
we think that the central component in this pulsar might be incorrectly identified as the core emission ( as discussed in section 3 ) .
the flat pa traverse seems to suggest that the line of sight is not cutting the overall beam centrally ( see another example of broad profile pulsar b0826@xmath2634 discussed by gupta , gil , kijak et al .
2004 ) , so the core emission might be missing .
if this is the case then the central component could arise from grazing the third , innermost cone of emission , although its non - central location is a concern in this model .
however , if our hypothetical 3rd cone interpretation is incorrect , then b1831@xmath2604 remains the strongest counter example of a / r effects .
the three other counter
examples b1738@xmath2608 , b1848 + 04 and b1859 + 03 show a marginal deviation from the a / r effects and are probably also related to our inability of identifying the location of the core component . in the first case ( b1738@xmath2608 ) the core component is not apparent ,
although a blind gaussian fitting method indicates a central core " component ( see fig .
a.9 in the on line appendix ) .
the second pulsar ( b1848 + 04 ) is again a typical example of a broad profile pulsar ( see fig .
a.13 in the on line appendix ) , where the core component is likely to be missed . the third pulsar ( b1859 + 03 ) ( see fig .
a.4 in the on line appendix ) is a clear example of a pulsars with missing component ( in this case the trailing one ) .
therefore , we conclude , that all 4 cases of strong counter examples reported are the result of mis identification of the core component .
sometimes weak conal emission or merged conal components leads to wrong identification of the emission components .
for example in psr b1541 + 09 the leading conal component at 1225 and 2330 mhz is difficult to model . in cases like psr b1944 + 17 and b1737 + 13
the weak conal emission poses difficulties in determining the location of the components .
it is likely that the same set of problems occur in the group of doubtful cases ( i.e. pulsars which show a / r effects at some frequencies but not in others ) .
this strengthens our general conclusions that a / r effects are apparent in pulsar profiles . in cases where a / r effects are seen between the core and conal components in pulsars , implies that the core emission necessarily has to originate closer to the stellar surface than the conal emission ( see for e.g. g&ga , dyks et al .
the delay emission heights obtained by eqn .
( [ r_del ] ) are essentially the conal emission heights with respect to the core emission .
alternatively geometrical methods can be used to find the conal emission height @xmath18 as specified by eqn .
( [ rkg ] ) . a comparison between the conal emission heights measured by these two methods is shown in fig .
( [ r_em_plot ] ) .
this plot can be compared to fig .
( 29 ) of bcw where a similar effect is noted , although the delay heights there were measured from polarisation observations and they used log log axes . in this figure
the clear a / r cases are marked with filled circles , doubtful cases are marked with open circles and the so called counter examples are marked with stars .
note that many of the negative @xmath30 values , in fig .
( [ r_em_plot ] ) correspond to the counter
examples in our sample .
it is noteworthy that statistically the relation @xmath42 seems to hold well .
there are primarily two reasons which can explain this effect .
firstly the @xmath30 values can be underestimated if the putative core emission arises from a definite height above the polar cap . in order to square the two height estimates by making the cluster of points in fig .
( [ r_em_plot ] ) to be located on the 45@xmath43 line , a nominal average core emission height of about 100 km is needed .
alternatively , the geometrical emission height @xmath18 might be overestimated , since the assumption prevailing in estimating these heights is that the measured widths entail the last open field lines , i.e. @xmath44 ( see eqn .
[ rho ] ) .
if the parameter @xmath45 was around 0.7 , the two height estimates would square with each other . as discussed by mitra & li ( 2004 ) , a careful and systematic study of the relation of the core emission w.r.t .
the polarisation position angle traverse might help to resolve this issue . .
note that a clear tendency for negative shifts is seen in the sample .
, width=377 ] given by eqn .
( [ r_del ] ) and the semi empirical geometrical height estimate @xmath18 given by eqn .
( [ rkg ] ) .
note that statistically @xmath42 as seen by the excess of points lying above the @xmath46 line in the figure ( see legend to distinguish different groups of pulsars ) .
see text for details .
, width=377 ]
we thank the gmrt operational staff for observational support .
the gmrt is run by the national centre for radio astrophysics of the tata institute of fundamental research .
we acknowledge the usage of the gaussian fitting program given to us by michael kramer which has been extensively used in this paper .
jg and jk acknowledge partial support of the polish grant n n203 2738 33 .
jk and kk acknowledge the polish grant n 203 021 32/2993 .
this paper was also partially supported by the polish grant n n203 3919 34 .
we would like to thank an anonymous referee whose comments and suggestions helped us to improve our paper .
we thank urszula maciejewska for critical reading of the manuscript .
1991 blaskiewicz , m. , cordes , j. m. & wasserman , i. 1991 , apj 370 , 643 ( bcw ) dyks , j. , rudak , b. & harding , a. k. 2004 , apj 607 , 939 gangadhara , r. t. & gupta , y. 2001 , apj 555 , 31 ( g&ga ) gangadhara , r. t. , 2005 , apj , 628 , 923 gil , j. , gronkowski , p. & rudnicki , w. 1984 , a&a 132 , 312 gil , j. a. & kijak , j. 1993 , a&a 273 , 563 gould , d. m. & lyne , a. g. 1998 , mnras 301 , 235 ( gl98 ) gupta , y. , gothoskar , p. , joshi , b. c. , vivekanand , m. , swain , r. , sirothia , s. , bhat , n. d. r. , 2000 pulsar astronomy 2000 and beyond , asp conference series , vol .
202 ; proceedings of the 177th colloquium of the iau held in bonn , germany , 30 august 3 september 1999 .
( san francisco : asp ) .
edited by m. kramer , n. wex , and n. wielebinski , p. 277
gupta , y. & gangadhara , r. t. 2003 , apj 584 , 418 ( g&gb ) gupta , y. , gil , j. , kijak , j. , sendyk , m. , a&a , 426 , 229 kijak , j. & gil , j. 1997 , mnras 288 , 631 kijak , j. , kramer , m. , wielebinski , r. & jessner , a. 1997 , a&a 318 , l63 kijak , j. & gil , j. 1998 , mnras 299 , 855 kijak , j. & gil , j. 2003,a&a 397 , 969 kramer , m. , wielebinski , r. , jessner , a. , gil , j. a. & seiradakis , j. h. 1994 , a&a sup 107 , 515 lyne , a. g. & manchester , r. n. 1988 , mnras 234 , 477 malov , i. f. , suleimanova , s. a. , 1998 , astronomy reports , 42 , 388 mitra , d. & deshpande , a. a. 1999 , a&a 346 , 906 mitra , d. & rankin , j. m. 2002 , apj 577 , 322 mitra , d. , rankin , j. m. , sarala , s. , 40 years of pulsars : millisecond pulsars , magnetars and more .
aip conference proceedings , volume 983 , pp .
106 mitra , d. & li , x. h. 2004 , a&a 421 , 215 mitra , d. , rankin , j. m. , gupta , y. , 2007 , mnras 379 , 932 mitra , d. , rankin , j. m. , 2008 , mnras , 385 , 606 rankin , j. m. , 1983 , apj 274 , 333 rankin , j. m. , 1986 , apj 301 , 901 rankin , j. m. , 1990 , apj 352 , 247 rankin , j. m. , 1993a , apj 405 , 285 rankin , j. m. , 1993b , apj sup 85 , 145 radhakrishnan , v. , cooke , d. j. , 1969 , astrophysical letters , vol .
3 , p.225 sirothia , s. 2000 , m.sc . thesis , university of pune srostlik , z. , rankin , j. m. , 2005 , mnras 362 , 1121 swarup , g. , ananthakrishnan , s. , kapahi , v. k. , rao , a. p. , subrahmanya , c. r. , kulkarni , & v. k. 1991 , current science 60 , 95 .
weltevrede , p. , stappers , b. w. , & edwards , r. t. 2006 , a&a , 445 , 243 weltevrede , p. , edwards , r. t. , & stappers , b. w. 2007 , a&a , 469 , 607 | the magnetospheric locations of pulsar radio emission region are not well known .
the actual form of the so called radius to frequency mapping should be reflected in the aberration
retardation ( a / r ) effects that shift and/or delay the photons depending on the emission height in the magnetosphere .
recent studies suggest that in a handful of pulsars the a / r effect can be discerned w.r.t the peak of the central core emission region . to verify these effects in an ensemble of pulsars we launched a project analysing multi
frequency total intensity pulsar profiles obtained from the new observations from the giant meterwave radio telescope ( gmrt ) , arecibo observatory ( ao ) and archival european pulsar network ( epn ) data .
for all these profiles we measure the shift of the outer cone components with respect to the core component which is necessary for establishing the a / r effect . within our sample of 23 pulsars
7 show the a / r effects , 12 of them ( doubtful cases ) show a tendency towards this effect , while the remaining 4 are obvious counter examples . the counter examples and doubtful cases may arise from uncertainties in determination of the location of the meridional plane and/or the core emission component .
it hence appears that the a / r effects are likely to operate in most pulsars from our sample .
we conclude that in cases where those effects are present the core emission has to originate below the conal emission region .
[ firstpage ] pulsars : general stars : neutron radiation mechanisms : non - thermal methods : data analysis |
binary star systems are of astrophysical importance for various reasons .
first , they compose an important portion of stars in the galaxy @xcite and thus theories about stellar formation and evolution should account for the binary nature of stars .
second , binary stars allow us to directly measure the masses of their component stars . the determined masses in turn allow other stellar parameters , such as radius and density , to be indirectly estimated .
these physical parameters help us to understand the processes by which binary stars form @xcite .
in particular , the separation and mass of a binary system tell us about the amount of angular momentum in the system . because it is a conserved quantity , binaries with measured angular momentum
give us important clues about the conditions under which the stars were formed . despite the importance ,
broad ranges of separations , distances , and component masses make it hard to detect and measure all binaries .
nearby systems with wide separations may be directly resolved using high - resolution imaging , while systems with small separations can be detected as eclipsing or spectroscopic binaries .
however , binaries with intermediate separations are difficult to be detected by the conventional methods .
in addition , it is difficult to detect binaries if they are located at large distances or either of the binary components is faint . as a result ,
samples are restricted to binaries in the solar neighborhood and are not complete down to low - mass stars . for a complete view of stellar multiplicity across a broad range of physical parameters , therefore , it is necessary to use a variety of observational techniques .
gravitational microlensing can provide a complementary method that can detect and measure binaries that are difficult to be detected by other methods .
microlensing occurs when an astronomical object is closely aligned with a background star .
the gravity of the intervening object ( lens ) causes deflection of the light from the background star ( source ) , resulting in the brightening of the source star .
if the lens is a single star , the light curve of the source star brightness is characterized by smooth rising and fall . however ,
if the lens is a binary , the light curve can be dramatically different , particularly for caustic - crossing events , which exhibit strong spikes in the light curve . among caustic - crossing binary - lens events ,
those with long time scales are of special importance because it is often possible to determine the physical parameters of lenses ( see more details in section 2 ) .
the binary separations for which caustic crossings are likely to occur are in the range of order au , for which binaries are difficult to be detected by other methods .
in addition , due to the nature of the lensing phenomenon that occurs regardless of the lens brightness , microlensing can provide an important channel to study binaries composed of low - mass stars .
furthermore , most microlensing binaries are located at distances of order kpc and thus microlensing can expand the current binary sample throughout the galaxy . in this paper , we report the detections and measurements of 2 binaries discovered from observations of long time - scale caustic - crossing binary microlensing events moa-2011-blg-090 and ogle-2011-blg-0417 . in 2 , we describe the basic physics of binary lensing and the method to determine the physical parameters of binary lenses . in 3 ,
we describe the choice of sample , observations of the events , and data reduction . in 4 , we describe the procedure of modeling the observed light curves . in 5 , we present the results from the analysis .
we discuss about the findings and conclude in 6 .
for a general lensing event , where a single star causes the brightening of a background source star , the magnification of the source star flux depends only on the projected separation between the source and the lens as @xmath1 where the separation @xmath2 is normalized in units of the angular einstein radius of the lens , @xmath3 . for a uniform change of the lens - source separation ,
the light curve of a single - lens event is characterized by a smooth and symmetric shape .
the normalized lens - source separation is related to the lensing parameters by @xmath4^{1/2 } , \label{eq2}\ ] ] where @xmath5 represents the time scale for the lens to cross the einstein radius ( einstein time scale ) , @xmath6 is the time of the closest lens - source approach , and @xmath7 is the lens - source separation at that moment . among these lensing parameters @xmath6 , @xmath5 , and @xmath7 ,
the only quantity related to the physical parameters of the lens is the einstein time scale .
however , it results from the combination of the lens mass , distance , and transverse speed of the relative lens - source motion and thus the information about the lens from the time scale is highly degenerate .
when gravitational lensing is caused by a binary , the gravitational field is asymmetric and the resulting light curves can be dramatically different from that of a single lensing event @xcite .
the most prominent feature of binary lensing that differentiates it from single lensing is a caustic .
a set of caustics form a boundary of an envelope of rays as a curve of concentrated light .
the gradient of magnification around the caustic is very large . as a result ,
the light curve of an event produced by the crossing of a source star over the caustic formed by a binary lens is characterized by sharp spikes occurring at the time of caustic crossings .
caustic - crossing binary - lens events are useful because it is often possible to measure an additional lensing parameter appearing in the expression of the einstein radius .
this is possible because the caustic - crossing part of the light curve appears to be different for events associated with source stars of different sizes @xcite . by measuring the deviation caused by this finite - source effect , it is possible to measure the source radius in units of the einstein radius , @xmath8 ( normalized source radius ) .
then , combined with the information about the source angular size , @xmath9 , the einstein radius is determined as @xmath10 . the einstein radius is related to the mass , @xmath11 , and distance to the lens , @xmath12 , by @xmath13 where @xmath14 , @xmath15 is the distance to the source , and @xmath16 represents the relative lens - source proper motion . unlike the einstein time scale ,
the einstein radius does not depend on the transverse speed of the lens - source motion and thus the physical parameters are less degenerate compared to the einstein time scale . among caustic - crossing events , those with long time scales are of special interest because it is possible to completely resolve the degeneracy of the lens parameters and thus uniquely determine the mass and distance to the lens .
this is possible because an additional lensing parameter of the lens parallax can be measured for these events .
the lens parallax is defined as the ratio of earth s orbit , i.e. 1 au , to the physical einstein radius , @xmath17 , projected on the observer plane , i.e. @xmath18 with simultaneous measurements of the einstein radius and the lens parallax , the mass and distance to the lens are uniquely determined as @xmath19 and @xmath20 respectively @xcite .
the lens parallax is measured by analyzing deviations in lensing light curves caused by the deviation of the relative lens - source motion from a rectilinear one due to the change of the observer s position induced by the orbital motion of the earth around the sun @xcite .
this deviation becomes important for long time - scale events , which endure for a significant fraction of the orbital motion of the earth .
therefore , the probability of measuring the lens parallax is high for long time - scale events .
we searched for long time - scale caustic - crossing binary events among lensing events discovered in the 2011 microlensing observation season .
we selected events to be analyzed based on the following criteria . 1 .
the overall light curve was well covered with good photometry .
at least one of caustic crossings was well resolved for the einstein radius measurement .
3 . the time scale of an event should be long enough for the lens parallax measurement . from this search
, we found 2 events of moa-2011-blg-090 and ogle-2011-blg-0417 . besides these events ,
there exist several other long time - scale caustic - crossing events , including moa-2011-blg-034 , ogle-2011-blg-0307/moa-2011-blg-241 , and moa-2011-blg-358/ogle-2011-blg-1132 .
we did not include moa-2011-blg-034 and moa-2011-blg-358/ogle-2011-blg-1132 in our analysis list because the coverage and photometry of the events are not good enough to determine the physical lens parameters by measuring subtle second - order effects in the lensing light curve .
for ogle-2011-blg-0307/moa-2011-blg-241 , the signal of the parallax effect was not strong enough to securely measure the physical parameters of the lens .
ll moa-2011-blg-090 & moa moa - ii 1.8 m , new zealand + & ogle warsaw 1.3 m , chile + & @xmath21fun ctio / smarts2 1.3 m , chile + & @xmath21fun pest 0.3 m , australia + & mindstep danish 1.54 m , chile + & robonet fts 2.0 m , australia + ogle-2011-blg-0417 & ogle warsaw 1.3 m , chile + & @xmath21fun ctio / smarts2 1.3 m , chile + & @xmath21fun auckland 0.4 m , new zealand + & @xmath21fun fco 0.36 m , new zealand + & @xmath21fun kumeu 0.36 m , new zealand + & @xmath21fun opd 0.6 m , brazil + & planet canopus 1.0 m , australia + & planet saao 1.0 m , south africa + & mindstep danish 1.54 m , chile + & robonet ftn 2.0 m , hawaii + & robonet lt 2.0 m , spain the events moa-2011-blg-090 and ogle-2011-blg-0417 were observed by the microlensing experiments that are being conducted toward galactic bulge fields by 6 different groups including moa , ogle , @xmath21fun , planet , robonet , and mindstep . among them , the moa and ogle collaborations are conducting survey observations for which the primary goal is to detect a maximum number of lensing events by monitoring a large area of sky .
the @xmath21fun , planet , robonet , and mindstep groups are conducting follow - up observations of events detected from survey observations . the events were observed by using 12 telescopes located in 3 different continents in the southern hemisphere . in table
[ table : one ] , we list the telescopes used for the observation .
reduction of the data was conducted by using photometry codes developed by the individual groups .
the ogle and moa data were reduced by photometry codes developed by @xcite and @xcite , respectively , which are based on the difference image analysis method @xcite .
the @xmath21fun data were processed using a pipeline based on the dophot software @xcite . for planet and mindstep data , a pipeline based on the pysis software @xcite is used . for robonet data ,
the dandia pipeline @xcite is used . to standardize error bars of data estimated from different observatories ,
we re - scaled them so that @xmath22 per degree of freedom becomes unity for the data set of each observatory , where @xmath22 is computed based on the best - fit model . for the data sets used for modeling , we eliminate data points with very large photometric uncertainties and those lying beyond @xmath23 from the best - fit model .
we present the light curves of events in figure [ fig : one ] and [ fig : two ] . to be noted is that the overall light curves including caustic crossings of both events are well covered by survey observations .
this demonstrates that the observational cadence of survey experiments is now high enough to characterize lensing events based on their own data .
the parts of light curves with 2455880 < hjd <
2455960 were not covered because the galactic bulge field could not be observed .
although not included in the selection criteria , both events showed a common bump to those involved with caustic crossings : at hjd@xmath242455655 for moa-2011-blg-090 and at hjd@xmath242455800 for ogle-2011-blg-0417 .
these bumps were produced during the approach of the source trajectory close to a cusp of a caustic .
@xcite pointed out that such triple - peak features help to better measure the lens parallax .
in modeling the light curve of each event , we search for a solution of lensing parameters that best characterizes the observed light curve . describing the basic feature of a binary - lens light curve
requires 6 parameters including the 3 single - lensing parameters @xmath6 , @xmath7 , and @xmath5 .
the 3 additional binary - lensing parameters include the mass ratio between the lens components , @xmath25 , the projected separation in units of the einstein radius , @xmath26 , and the angle between the source trajectory and the binary axis , @xmath27 ( trajectory angle ) .
in addition to the basic binary lensing parameters , it is required to include additional parameters to precisely describe detailed features caused by various second - order effects .
the first such effect is related to the finite size of the source star .
this finite - source effect becomes important when the source is located at a position where the gradient of magnification is very high and thus different parts of the source surface experience different amounts of magnification . for binary - lens events ,
this happens when the source approaches or crosses the caustic around which the gradient of magnification is very high . to describe the light curve variation caused by the finite - source effect ,
it is necessary to include an additional parameter of the normalized source radius , @xmath8 . for long time - scale events , such as those analyzed in this work
, it is required to additionally consider the parallax effect .
consideration of the parallax effect in modeling requires to include 2 additional parameters @xmath28 , and @xmath29 , which represent the two components of the lens parallax vector @xmath30 projected on the sky along the north and east equatorial coordinates , respectively .
the direction of the lens parallax vector corresponds to the relative lens - source relative motion in the frame of the earth at a specific time of the event .
the size of the vector corresponds to the ratio of the earth s orbit to the einstein radius projected on the observer s plane , i.e. @xmath31 $ ] , where @xmath17 is the physical size of the einstein radius .
another effect to be considered for long time - scale events is the orbital motion of the lens @xcite .
the lens orbital motion affects lensing light curves in two different ways .
first , it causes to change the binary separation and thus the magnification pattern .
second , it also causes the binary axis to rotate with respect to the source trajectory . in order to fully account for the lens orbital motion ,
4 additional parameters are needed .
the first two of these parameters are @xmath32 and @xmath33 , which represent the change rates of the projected binary separation and the trajectory angle , respectively .
the other two orbital parameters are @xmath34 and @xmath35 , where @xmath34 represents the line - of - sight separation between the binary components in units of @xmath3 and @xmath35 represents its rate of change . for a full description of the orbital lensing parameters , see the summary in the appendix of @xcite .
the deviation in a lensing light curve affected by the orbital effect is smooth and long lasting and thus is similar to the deviation induced by the parallax effect .
this implies that if the orbital effect is not considered , the measured lens parallax and the resulting lens parameters might be erroneous .
therefore , considering the orbital effect is important not only for constraining the orbital motion of the lens but also for precisely determining the physical parameters of the lens . with all these parameters , we test three different models .
in the first model , we fit the light curve with standard binary lensing parameters considering the finite - source effect ( standard model ) . in the second model
, we additionally consider the parallax effect ( parallax model ) .
finally , we take the orbital effect into consideration as well ( orbital model ) . when the source trajectory is a straight line ,
the two light curves resulting from source trajectories with positive ( @xmath36 ) and negative ( @xmath37 ) impact parameters are identical due to the symmetry of the magnification pattern with respect to the binary axis .
when either the parallax or the orbital effect is considered , on the other hand , the source trajectory deviates from a straight line and thus the light curves with @xmath36 and @xmath37 are different from each other .
we , therefore , consider both the @xmath36 and @xmath37 cases for each of the models considering the parallax and orbital effects . in modeling ,
the best - fit solution is obtained by minimizing @xmath22 in the parameter space .
we conduct this in 3 stages . in the first stage ,
grid searches are conducted over the space of a subset of parameters and the remaining parameters are searched by using a downhill approach @xcite .
we then identify local minima in the grid - parameter space by inspecting the @xmath22 distribution . in the second stage , we investigate the individual local minima by allowing the grid parameters to vary and find the exact location of each local minimum . in the final stage , we choose the best - fit solution by comparing @xmath22 values of the individual local minima .
this multiple stage procedure is needed for thorough investigation of possible degeneracy of solutions .
we choose of @xmath26 , @xmath25 , and @xmath27 as the grid parameters because they are related to the light curve features in a complex way such that a small change in the values of the parameters can lead to dramatic changes in the resulting light curve . on the other hand ,
the other parameters are more directly related to the light curve features and thus they are searched for by using a downhill approach . for the @xmath22 minimization in the downhill approach ,
we use the markov chain monte carlo ( mcmc ) method .
once a solution of the parameters is found , we estimate the uncertainties of the individual parameters based on the chain of solutions obtained from mcmc runs .
ccc @xmath38 & 0.52 & 0.71 + @xmath39 & 0.45 & 0.61 + @xmath40 & 0.37 & 0.51 + source type & fv & kiii + @xmath41 ( k ) & 6650 & 4660 + @xmath42 ( @xmath43 ) & 2 & 2 + @xmath44 ( @xmath45 ) & 4.5 & 2.5 to compute lensing magnifications affected by the finite - source effect , we use the ray - shooting method . @xcite . in this method
, rays are uniformly shot from the image plane , bent according to the lens equation , and land on the source plane .
then , a finite magnification is computed by comparing the number densities of rays on the image and source planes .
precise computation of finite magnifications by using this numerical technique requires a large number of rays and thus demands heavy computation . to minimize computation , we limit finite - magnification computation by using the ray - shooting method only when the lens is very close to caustics . in the adjacent region
, we use an analytic hexadecapole approximation @xcite . in the region with large enough distances from caustics , we use a point - source magnifications . in the finite magnification computation ,
we consider the variation of the magnification caused by the limb - darkening of the source star s surface .
we model the surface brightness profile of a source star as @xmath46,\ ] ] where @xmath47 is the linear limb - darkening coefficients , @xmath48 is the source star flux , and @xmath49 is the angle between the normal to the source star s surface and the line of sight toward the star .
the limb - darkening coefficients are set based the source type that is determined on the basis of the color and magnitude of the source . in table
[ table : two ] , we present the used limb - darkening coefficients , the corresponding source types , and the measured de - reddened color along with the assumed values of the effective temperature , @xmath41 , the surface turbulence velocity , @xmath42 , and the surface gravity , @xmath44 . for both events ,
we assume a solar metallicity .
in table [ table : three ] , we present the solutions of parameters for the tested models .
the best - fit model light curves are drawn on the top of the observed light curves in figures [ fig : one ] and [ fig : two ] . in figure
[ fig : three ] , we present the geometry of the lens systems where the source trajectory with respect to the caustic and the locations of the lens components are marked .
we note that the source trajectories are curved due to the combination of the parallax and orbital effects .
we also note that the positions of the lens components and the corresponding caustics change in time due to the orbital motion and thus we present caustics at 2 different moments that are marked in figure [ fig : three ] .
these moments correspond to those of characteristic features on the light curve such as the peak involved during a cusp approach or a caustic crossing .
to better show the differences in the fit between different models , we present the residuals of data from the best - fits of the individual models . for a close - up view of the caustic - crossing parts of the light curves , we also present enlargement of the light curve . for both events ,
the parallax and orbital effects are detected with significant statistical confidence levels .
it is found that inclusion of the second - order effects of the parallax and orbital motions improves the fits with @xmath50 and @xmath51 for moa-2011-blg-090 and ogle-2011-blg-0417 , respectively . to be noted
is that the orbital effect is considerable for ogle-2011-blg-0417 and thus the difference between the values of the lens parallax measured with ( @xmath52 ) and without ( @xmath53 ) considering the orbital effect is substantial .
since the lens parallax is directly related to the physical parameters of the lens , this implies that considering the orbital motion of the binary lens is important for the accurate measurement of the lens parallax and thus the physical parameters .
lrrrrrr @xmath54 & 5207/5164 & 4718/5162 & 4636/5158 & 4415/2627 & 2391/2625 & 1735/2621 + @xmath6 ( hjd ) & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] + @xmath7 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] + @xmath5 ( days ) & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] + @xmath56 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] + @xmath25 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] + @xmath27 ( rad ) & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] + @xmath57 ( @xmath58 ) & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] + @xmath28 & & [email protected] & [email protected] & & [email protected] & [email protected] + @xmath29 & & [email protected] & [email protected] & & [email protected] & [email protected] + @xmath59 ( @xmath60 ) & & & [email protected] & & & [email protected] + @xmath33 ( @xmath60 ) & & & [email protected] & & & [email protected] + @xmath61 & & & [email protected] & & & [email protected] + @xmath62 ( @xmath60 ) & & & [email protected] & & & [email protected] lrr @xmath63 ( @xmath64 ) & [email protected] & [email protected] + @xmath65 ( @xmath64 ) & [email protected] & [email protected] + @xmath66 ( @xmath64 ) & [email protected] & [email protected] + @xmath3 ( mas ) & [email protected] & [email protected] + @xmath21 ( mas yr@xmath67 ) & [email protected] & [email protected] + @xmath12 ( kpc ) & [email protected] & [email protected] + @xmath68 ( au ) & [email protected] & [email protected] + @xmath69 ( yr ) & [email protected] & [email protected] + @xmath70 & [email protected] & [email protected] + @xmath71 ( @xmath72 ) & [email protected] & [email protected] the finite - source effect is also clearly detected and the normalized source radii are precisely measured for both events . to obtain the einstein radius from the measured normalized source radius , @xmath8 , additional information about the source star is needed .
we obtain this information by first locating the source star on the color - magnitude diagram of stars in the field and then calibrating the source brightness and color by using the centroid of the giant clump as a reference under the assumption that the source and clump giants experience the same amount of extinction and reddening @xcite .
the measured @xmath73 colors are then translated into @xmath74 color by using the @xmath75 relations of @xcite and the angular source radius is obtained by using the @xmath74 color and the angular radius given by @xcite . in figure
[ fig : four ] , we present the color - magnitudes of field stars constructed based on ogle data and the locations of the source star .
we find that the source star is a f - type main - sequence star with a de - reddened color of @xmath76 for moa-2011-blg-090 and a k - type giant with @xmath77 for ogle-2011-blg-0417 . here
we assume that the de - reddened color and absolute magnitude of the giant clump centroid are @xmath78 and @xmath79 @xcite , respectively .
the mean distances to clump stars of @xmath247200 pc for moa-2011-blg-090 and @xmath247900 pc for ogle-2011-blg-0417 are estimated based on the galactic model of @xcite .
the measured einstein radii of the individual events are presented in table [ table : four ] .
also presented are the relative lens - source proper motions as determined by @xmath80 . with the measured lens parallax and the einstein radius , the mass and distance to the lens of each event
are determined by using the relations ( [ eq5 ] ) and ( [ eq6 ] ) .
the measured masses of the binary components are 0.43 @xmath0 and 0.39 @xmath0 for moa-2011-blg-090 and 0.57 @xmath0 and 0.17 @xmath0 for ogle-2011-blg-0417 .
it is to be noted that both lens components of moa-2011-blg-090 and one component of ogle-2011-blg-0417 are m dwarfs which are difficult to be detected by other conventional methods due to their faintness .
it is found that the lenses are located at distances @xmath81 kpc and @xmath82 kpc for moa-2011-blg-090 and ogle-2011-blg-0417 , respectively . since full keplerian motion of the binary lens
is considered in our modeling , we also determine the orbital parameters of the semi - major axis @xmath68 , period @xmath69 , eccentricity @xmath70 , and inclination @xmath71 .
we find that the binary lens components of moa-2011-blg-090 are orbiting each other with a semi - major axis of @xmath83 au and an orbital period of @xmath84 yrs . for ogle-2011-blg-0417 ,
the semi - major axis and the orbital period of the binary lens are @xmath85 au and @xmath86 yrs , respectively . in figure
[ fig : five ] , we present the distributions of the physical and orbital parameters constructed based on the mcmc chains . in table [ table : four ] , we summarize the measured parameters of the binary lenses for both events .
we note that the uncertainties of the parameters are based on the standard deviations of the mcmc distributions .
we note that the blended light of ogle-2011-blg-0417 comes very likely from the lens itself , implying that the lens can be directly observed .
based on the measured mass of 0.57 @xmath0 , the primary of the binary lens corresponds to a late k - type main sequence star with an absolute magnitude and a de - reddened color of @xmath87 and @xmath88 , respectively .
considering the distance of 0.89 kpc and assuming an extinction of @xmath89 and the color index of @xmath90 , the apparent brightness and color of the lens correspond to @xmath91 and @xmath92 , respectively .
these values match very well with the location of the blend marked on the right panel of figure [ fig : four ] , implying that the blend is very likely the lens .
the visibility of the lens signifies this event because it is possible to check the microlensing orbital solution by spectroscopic radial - velocity observation .
we reported detections and measurements of 2 binaries discovered from observations of microlensing events moa-2011-blg-090 and ogle-2011-blg-0417 .
the light curves of the events have common characteristics of long durations with caustic - crossing features , which enabled to determine the physical parameters of the lenses .
it was found that both lens components of moa-2011-blg-090 and one component of ogle-2011-blg-0417 are m dwarfs .
therefore , the discovered microlensing binaries demonstrate the usefulness of gravitational lensing in detecting and characterizing binaries composed of low - mass stars . by considering full keplerian binary motion
, we also determined the orbital parameters of the binaries .
for ogle-2011-blg-0417 , the blended light comes very likely from the lens itself and thus it would be possible to check the orbital solution from follow - up radial - velocity observation . studies of m dwarfs are important not only because they are the most abundant stars in the milky way but also they form a link between solar - type stars and brown dwarfs ; two mass regimes that might exhibit very different multiplicity characteristics . precise knowledge of multiplicity characteristics and how they change in this transitional mass region provide constraints on low - mass star and brown dwarf formation @xcite . despite the importance of m - dwarf binaries , only a few measurements of the binary fraction and distribution of low - mass stars have been made , e.g. , @xcite , and the samples are restricted to only binaries in the solar neighborhood . as a result , there are still large uncertainties about their basic physical properties as well as their formation environment . considering the rapid improvement of lensing surveys both in equipment and strategy , it is expected that the number of microlensing binaries with measured physical parameters will increase rapidly .
this will contribute to the complete view of stellar multiplicity across a wide range of binary parameters .
work by ch was supported by creative research initiative program ( 2009 - 0081561 ) of national research foundation of korea .
the ogle project has received funding from the european research council under the european community s seventh framework programme ( fp7/2007 - 2013 ) / erc grant agreement no .
the moa experiment was supported by grants jsps22403003 and jsps23340064 .
ts was supported by the grant jsps23340044 .
y. muraki acknowledges support from jsps grants jsps23540339 and jsps19340058 .
the mindstep monitoring campaign is powered by artemis ( automated terrestrial exoplanet microlensing search ; dominik et al .
2008 , an 329 , 248 ) .
mh acknowledges support by the german research foundation ( dfg ) .
dr ( boursier fria ) and j. surdej acknowledge support from the communaut franaise de belgique actions de recherche concertes acadmie universitaire wallonie - europe .
the robonet team is supported by the qatar foundation through qnrf grant nprp-09 - 476 - 1 - 78 .
cs received funding from the european union seventh framework programme ( fpt/2007 - 2013 ) under grant agreement 268421 .
this work is based in part on data collected by mindstep with the danish 1.54 m telescope at the eso la silla observatory . the danish
1.54 m telescope is operated based on a grant from the danish natural science foundation ( fnu ) .
a. gould and b.s .
gaudi acknowledge support from nsf ast-1103471 .
gaudi , a. gould , and r.w .
pogge acknowledge support from nasa grant nng04gl51 g .
work by j.c .
yee is supported by a national science foundation graduate research fellowship under grant no .
2009068160 . s. dong s research was performed under contract with the california institute of technology ( caltech ) funded by nasa through the sagan fellowship program .
research by tch was carried out under the krcf young scientist research fellowship program .
tch and cul acknowledge the support of korea astronomy and space science institute ( kasi ) grant 2012 - 1 - 410 - 02 . | despite astrophysical importance of binary star systems , detections are limited to those located in small ranges of separations , distances , and masses and thus it is necessary to use a variety of observational techniques for a complete view of stellar multiplicity across a broad range of physical parameters . in this paper , we report the detections and measurements of 2 binaries discovered from observations of microlensing events moa-2011-blg-090 and ogle-2011-blg-0417 . determinations of the binary masses are possible by simultaneously measuring the einstein radius and the lens parallax .
the measured masses of the binary components are 0.43 @xmath0 and 0.39 @xmath0 for moa-2011-blg-090 and 0.57 @xmath0 and 0.17 @xmath0 for ogle-2011-blg-0417 and thus both lens components of moa-2011-blg-090 and one component of ogle-2011-blg-0417 are m dwarfs , demonstrating the usefulness of microlensing in detecting binaries composed of low - mass components . from modeling of the light curves considering full keplerian motion of the lens
, we also measure the orbital parameters of the binaries .
the blended light of ogle-2011-blg-0417 comes very likely from the lens itself , making it possible to check the microlensing orbital solution by follow - up radial - velocity observation . for both events ,
the caustic - crossing parts of the light curves , which are critical for determining the physical lens parameters , were resolved by high - cadence survey observations and thus it is expected that the number of microlensing binaries with measured physical parameters will increase in the future . |
pulsed high - frequency ( hf ) electromagnetic ( em ) waves from transmitters on the ground are regularly used for sounding the density profile and drift velocity of the overehead ionosphere [ _ hunsucker _ , 1991 ; _ reinisch et al .
_ , 1995 , _ reinisch _ , 1996 ] .
in 1971 , it was shown theoretically by _ perkins and kaw _ [ 1971 ] that if the injected hf radio beams are strong enough , weak - turbulence parametric instabilities in the ionospheric plasma of the type predicted by _ silin _ [ 1965 ] and _ dubois and goldman _ [ 1965 ] would be excited .
ionospheric modification experiments by a high - power hf radio wave at platteville in colorado [ _ utlaut _ , 1970 ] , using ionosonde recordings and photometric measurements of artificial airglow , demonstrated the heating of electrons , the deformation in the traces on ionosonde records , the excitation of spread @xmath0 , etc .
, after the hf transmitter was turned on . the triggering of weak - turbulence parametric instabilities in the ionosphere
was first observed in 1970 in experiments on the interaction between powerful hf radio beams and the ionospheric plasma , conducted at arecibo , puerto rico , using a scatter radar diagnostic technique [ _ wong and taylor _ , 1971 ; _ carlson et al .
_ , 1972 ] .
a decade later it was found experimentally in troms that , under similar experimental conditions as in arecibo , strong , systematic , structured , wide - band secondary hf radiation escapes from the interaction region [ _ thid et al . _ , 1982 ] .
this and other observations demonstrated that complex interactions , including weak and strong em turbulence , [ _ leyser _ , 2001 ; _ thid et al . _ , 2005 ] and harmonic generation [ _ derblom et al . _ , 1989 ; _ blagoveshchenskaya et al . _ ,
1998 ] are excited in these experiments .
numerical simulations have become an important tool to understand the complex behavior of plasma turbulence .
examples include analytical and numerical studies of langmuir turbulence [ _ robinson _ , 1997 ] , and of upper - hybrid / lower - hybrid turbulence in magnetized plasmas [ _ goodman et al . _ , 1994 ; _ xi _ , 2004 ] . in this letter , we present a full - scale simulation study of the propagation of an hf em wave into the ionosphere , with ionospheric parameters typical for the high - latitude eiscat heating facility in troms , norway . to our knowledge , this is the first simulation involving realistic scale sizes of the ionosphere and the wavelength of the em waves .
our results suggest that such simulations , which are possible with today s computers , will become a powerful tool to study hf - induced ionospheric turbulence and secondary radiation on a quantitative level for direct comparison with experimental data .
we use the mks system ( si units ) in the mathematical expressions throughout the manuscript , unless otherwise stated .
we assume a vertically stratified ion number density profile @xmath1 with a constant geomagnetic field @xmath2 directed obliquely to the density gradient .
the em wave is injected vertically into the ionosphere , with spatial variations only in the @xmath3 direction .
our simple one - dimensional model neglects the em field @xmath4 falloff ( @xmath5 is the distance from the transmitter ) , the fresnel pattern created obliquely to the @xmath3 direction by the incident and reflected wave , and the the influence on the radio wave propagation due to field aligned irregularities in the ionosphere .
for the em wave , the maxwell equations give @xmath6 @xmath7 where the electron fluid velocity is obtained from the momentum equation @xmath8\ ] ] and the electron density is obtained from the poisson equation @xmath9 .
here , @xmath10 is the unit vector in the @xmath3 direction , @xmath11 is the speed of light in vacuum , @xmath12 is the magnitude of the electron charge , @xmath13 is the vacuum permittivity , and @xmath14 is the electron mass .
ms.,scaledwidth=48.0% ] the number density profile of the immobile ions , @xmath15 $ ] ( @xmath3 in kilometers ) is shown in the leftmost panel of fig .
[ fig1 ] . instead of modeling a transmitting antenna via a time - dependent boundary condition at @xmath16 km
, we assume that the em pulse has reached the altitude @xmath17 km when we start our simulation , and we give the pulse as an initial condition at time @xmath18 s. in the initial condition , we use a linearly polarized em pulse where the carrier wave has the wavelength @xmath19 ( wavenumber @xmath20 ) corresponding to a carrier frequency of @xmath21 ( @xmath22 ) .
the em pulse is amplitude modulated in the form of a gaussian pulse with a maximum amplitude of @xmath23 v / m , with the @xmath24-component of the electric field set to @xmath25\sin(0.1047\times 10^{3 } z)$ ] ( @xmath3 in kilometers ) and the @xmath26 component of the magnetic field set to @xmath27 at @xmath18 .
the other electric and magnetic field components are set to zero ; see fig . [
the spatial width of the pulse is approximately 30 km , corresponding to a temporal width of 0.1 milliseconds as the pulse propagates with the speed of light in the neutral atmosphere .
it follows from eq .
( 1 ) that @xmath28 is time - independent ; hence we do not show @xmath28 in the figures .
the geomagnetic field is set to @xmath29 tesla , corresponding to an electron cyclotron frequency of 1.4 mhz , directed downward and tilted in the @xmath30-plane with an angle of @xmath31 degrees ( @xmath32 rad ) to the @xmath3-axis , i.e. , @xmath33 . in our numerical simulation , we use @xmath34 spatial grid points to resolve the plasma for @xmath35 km .
the spatial derivatives are approximated with centered second - order difference approximations , and the time - stepping is performed with a leap - frog scheme with a time step of @xmath36 s.
the splitting of the wave is due to faraday rotation.,scaledwidth=48.0% ] ms .
b ) a closeup of the region of the turning points of the r - x and l - o modes .
we see that the wave - energy of the l - o mode is concentrated into one single half - wave envelop at @xmath37 km , while the turning point of the less localized r - x mode is at @xmath38 km.,scaledwidth=48.0% ] ms .
b ) a closeup of the region of the turning points of the r - x and l - o modes . here , the l - o mode oscillations at @xmath37 km are radiating em waves with perpendicular ( to the @xmath3 axis ) electric field components.,scaledwidth=48.0% ] in the simulation , the em pulse propagates without changing shape through the neutral atmosphere , until it reaches the ionospheric layer . at time @xmath39
ms , shown in fig .
[ fig2 ] , the em pulse has reached the lower part of the ionosphere . the initially linearly polarized em wave undergoes faraday rotation due to the different dispersion properties of the l - o and r - x modes ( we have adopted the notation `` l - o mode '' and `` r - x mode '' for the two high - frequency em modes , similarly as , e.g. , _ goertz and strangeway _ [ 1995 ] ) in the magnetized plasma , and the @xmath40 and @xmath41 components are excited . at @xmath42 ms , shown in fig .
[ fig3 - 4 ] , the l - o and r - x mode pulses are in the vicinity of their respective turning points , the turning point of the l - o mode being at a higher altitude than that of the r - x mode ; see panel a ) of fig .
[ fig3 - 4 ] .
a closeup of this region , displayed in panel b ) , shows that the first maximum of the r - x mode is at @xmath38 km , and the one of the l - o mode is at @xmath37 km .
the maximum amplitude of the r - x mode is @xmath43
v / m while that of the l - o mode is @xmath44 v / m ; the latter amplitude maximum is in agreement with those obtained by _ thid and lundborg _ , [ 1986 ] , for a similar set of parameters as used here .
the electric field components of the l - o mode , which at this stage are concentrated into a pulse with a single maximum with a width of @xmath45 m , are primarily directed along the geomagnetic field lines , and hence only the @xmath46 and @xmath47 components are excited , while the magnetic field components of the l - o mode are very small . at @xmath48 ms ,
shown in panel a ) of fig .
[ fig5 - 6 ] , both the r - x and l - o mode wave packets have widened in space , and the em wave has started turning back towards lower altitudes . in the closeup of the em wave in panel b ) of fig . [ fig5 - 6 ]
, one sees that the l - o mode oscillations at @xmath37 km are now radiating em waves with significant magnetic field components . finally , shown in fig .
[ fig7 ] at @xmath49 , the em pulse has returned to the initial location at @xmath17 km . due to the different reflection heights of the l - o and r - x modes , the leading ( lower altitude ) part of the pulse is primarily r - x mode polarized while its trailing ( higher altitude ) part is l - o mode polarized . in the center of the pulse , where we have a superposition of the r - x and l - o mode ,
the wave is almost linearly polarized with the electric field along the @xmath26 axis and the magnetic field along the @xmath24 axis .
the direction of the electric and magnetic fields here depends on the relative phase between the r - x and l - o mode .
ms.,scaledwidth=48.0% ] at @xmath50 km , near the turning point of the r - x mode , and b ) the amplitude of the electric field component @xmath46 at @xmath51 km , near the turning point of the l - o mode .
c ) a snapshot of low - amplitude electrostatic waves of wavelength @xmath52 m ( wavenumber @xmath53 ) , observed at time @xmath54 ms , and d ) dispersion curves ( lower panel ) obtained from the appleton - hartree dispersion relation with parameters @xmath55 ( 5 mhz ) , @xmath56 ( 1.4 mhz ) and @xmath57 rad .
we identify the high - frequency r - x and l - o modes , as well as the z - mode which extends to the electrostatic langmuir / upper hybrid branch for large wavenumbers ; the circles indicate the approximate locations on the dispersion curve for the electrostatic oscillations shown in panel c ) . for completeness
we also show the low - frequency electron whistler branch in panel d).,scaledwidth=50.0% ] at the altitude @xmath50 km , and b ) of @xmath46 at the altitude @xmath51 km.,scaledwidth=48.0% ] in fig .
[ fig8 - 9 ] , panel a ) , we have plotted the electric field component @xmath47 at @xmath50 km , near the turning point of the r - x mode and in panel b ) we have plotted the @xmath46 component at @xmath51 km , near the turning point of the l - o mode .
we see that the maximum amplitude of @xmath47 reaches @xmath58 v / m at @xmath59 ms , and that of @xmath46 reaches @xmath60 v / m at @xmath61 ms .
the electric field amplitude at @xmath50 km has two maxima , due to the l - o mode part of the pulse , which is reflected at the higher altitude @xmath51 km and passes twice over the altitude @xmath50 km .
we also observe weakly damped oscillations of @xmath46 at @xmath51 km for times @xmath62 ms , which decrease exponentially in time between @xmath63 ms and @xmath64 ms as @xmath65 with @xmath66 s@xmath67 .
we found from the numerical values that @xmath68 , where @xmath69 is the inverse ion density scale length at @xmath70 km , but we are not certain how general this result is .
no detectable magnetic field fluctuations are associated with these weakly damped oscillations , and we interpret them as electrostatic waves that have been produced by mode conversion of the l - o mode .
the amplitudes of the @xmath47 and @xmath40 components are also much weaker than that of the @xmath46 component for these oscillations .
a closeup of these electrostatic oscillations at @xmath54 ms is displayed in panel c ) of fig .
[ fig8 - 9 ] , where we see that they have a wavelength of approximately 33 m ( wavenumber @xmath71 ) . in panel d ) of fig .
[ fig8 - 9 ] , we have plotted the frequency @xmath72 as a function of the wavenumber @xmath73 , where @xmath74 is obtained from the appleton - hartree dispersion relation [ _ stix _ , 1992 ] @xmath75 here @xmath76^{1/2}$ ] , @xmath77 ( @xmath78 ) is the electron plasma ( cyclotron ) frequency , and @xmath79 is the angle between the geomagnetic field and the wave vector @xmath80 , which in our case is directed along the @xmath3-axis , @xmath81 .
we use @xmath82 ( corresponding to @xmath83 mhz ) , @xmath84 ( corresponding to @xmath85 mhz ) and @xmath57 rad .
the location of the electrostatic waves whose wavelength is approximately 33 m and frequency 5 mhz is indicated with circles in the diagram ; they are on the same dispersion surface as the langmuir waves and the upper hybrid waves / slow z mode waves with propagation parallel and perpendicular to the geomagnetic field lines , respectively .
the mode conversion of the l - o mode into electrostatic oscillations are relatively weak in our simulation of vertically incident em waves , and theory shows that the most efficient linear mode conversion of the l - o mode occurs at two angles of incidence in the magnetic meridian plane , given by , e.g. , eq .
( 17 ) in [ _ mjlhus _ , 1990 ] .
the nonlinear effects at the turning point of the l - o and r - x modes are investigated in fig .
[ fig10 ] which displays the frequency spectrum of the electric field component @xmath47 at the altitude @xmath86 km and of @xmath46 at the altitude @xmath51 km .
the spectrum shows the large - amplitude pump wave at 5 mhz and the relatively weak second harmonics of the pump wave at 10 mhz at both altitudes ( the slight downshift is due to numerical errors produced by the difference approximations used in space and time ) .
visible are also low - frequency oscillations ( zeroth harmonic ) due to the nonlinear down - shifting / mixing of the high - frequency wave field .
in conclusion , we have presented a full - scale numerical study of the propagation of an em wave and its linear and nonlinear interactions with an ionospheric layer .
we observe the reflection of the l - o and r - x modes at different altitudes , the mode conversion of the l - o mode into electrostatic langmuir / upper hybrid waves as well as nonlinear harmonic generation of the high - frequency waves .
second harmonic generation have been observed in ionospheric heating experiments [ _ derblom et al .
_ , 1989 ; _ blagoveshchenskaya et al .
_ , 1998 ] and may be partially explained by the cold plasma model presented here .
blagoveshchenskaya , n. f. , v. a. kornienko , m. t. rietveld , b. thid , a. brekke , i. v. moskvin , and s. nozdrachev ( 1998 ) , stimulated emissions around second harmonic of troms heater frequency observed by long - distance diagnostic hf tools .
_ geophys .
25_(6 ) , 863876 .
derblom , h. , b. thid , t. b. leyser , j. a. nordling , .
hedberg , p. stubbe , h. kopka , and m. rietveld ( 1989 ) , troms heating experiments : stimulated emission at hf pump harmonic and subharmonic frequencies , _ j. geophys .
94_(a8 ) , 1011110120 .
thid , b. , e. n. sergeev , s. m. grach , t. b. leyser , and t. d. carozzi ( 2005 ) , competition between langmuir and upper - hybrid turbulence in a high - frequency - pumped ionosphere , _
_ 95 _ , 255002 . | the time evolution of a large - amplitude electromagnetic ( em ) wave injected vertically into the overhead ionosphere is studied numerically .
the em wave has a carrier frequency of 5 mhz and is modulated as a gaussian pulse with a width of approximately 0.1 milliseconds and a vacuum amplitude of 1.5 v / m at 50 km .
this is a fair representation of a modulated radio wave transmitted from a typical high - power hf broadcast station on the ground .
the pulse is propagated through the neutral atmosphere to the critical points of the ionosphere , where the l - o and r - x modes are reflected , and back to the neutral atmosphere .
we observe mode conversion of the l - o mode to electrostatic waves , as well as harmonic generation at the turning points of both the r - x and l - o modes , where their amplitudes rise to several times the original ones .
the study has relevance for ionospheric interaction experiments in combination with ground - based and satellite or rocket observations . |
the laser driven dynamics of distant quantum systems can depend in general on the spatial and temporal envelope of the applied laser field .
for example , at a gas pressure of 1 atm . , the interparticle distance is about 100 a.u .
( 5.29 nm ) .
if such a system , e.g. composed of h atoms , is excited by a laser field with the carrier frequency @xmath5 a.u .
, corresponding to the ground - state energy of h@xmath7 at a large internuclear distance @xmath0 , the wavelength is @xmath8 a.u .
( 45.56 nm ) . if the laser field is focused within the diffraction limit onto a spot with the width @xmath9 , the gaussian spatial envelope of the field may result in quite different electric field strengths for h atoms separated by about 100 a.u . , especially at the edges of the gaussian spatial envelope .
although the h atoms are far away from each other , their electron - electron interaction should not be _ a priori _ neglected , especially upon their excitation by the laser field , because the electronic wave functions extend and vanish , strictly speaking , only at infinity .
therefore , the energy transfer among distant quantum systems , similar to that studied in , can be anticipated to occur in spatially shaped laser fields as well .
for ultrashort laser pulses , containing only few optical cycles , one must also consider the carrier envelope phase ( cep ) of the pulse @xcite . the long - range energy transfer from an excited atom to its neighbor has been recently studied by cederbaum _
et al . _ for molecular clusters @xcite and
is known as the interatomic coulombic decay ( icd ) . nowadays , icd is well established also experimentally for inner - valence excitation of many electron systems . in recent work @xcite
, icd was demonstrated experimentally for a helium dimer .
since helium atoms have no inner - valence electrons , a different type of icd is operative for this case .
it was thus concluded in @xcite that since icd in a helium dimer takes place at interatomic distances up to @xmath10 a.u .
, no overlap of the electronic wave functions is required for the process .
the present work is addressed to a quantum system composed of two h atoms with the initial internuclear separation of 100 a.u .
( 5.29 nm ) which is excited by spatially shaped laser pulses : spatially broad pulses exciting both h atoms , and spatially narrow pulses exciting only one h atom of the entire h - h system .
the relative simplicity of the h - h system under consideration ( similar to that used in @xcite ) makes it possible to treat the long - range electronic motion explicitly together with the nuclear motion such as to reveal the role played by the electron - electron interaction and by the overlap of the electronic wave functions .
an example of long - range laser - induced electron transfer ( liet ) in the one electron linear h@xmath11-h@xmath12 atom - molecule system has been treated previously within the born - oppenheimer approximation @xcite .
long - range charge and energy transfer can occur also in large molecular systems , as described recently in ref .
@xcite and references therein .
the following two types of h - h systems will be distinguished in the present work : ( i ) a ` molecular ' h - h system , representing an elongated configuration of the h@xmath7 molecule , similar to that studied recently in @xcite for long - range entanglement , and ( ii ) an ` atomic ' h - h system , representing two distant h atoms .
accordingly , the initial state of a molecular h - h system is assumed to be entangled by spin exchange and represented by the heitler - london symmetric product of atomic wave functions , while the initial state of an atomic h - h system is not entangled it is a direct - product state of atomic wave functions . in both cases
the excitation of h - h is accomplished by laser pulses with ( i ) a broad gaussian spatial envelope , such that both h atoms are excited by the laser field , with atom a being excited predominantly , and ( ii ) with a narrow spatial envelope , such that only atom a is excited by the laser field .
the paper is organized as follows .
the model of the h - h system and techniques used are described in sec .
excitation , energy transfer , and ionization of an unentangled atomic h - h system are presented in sec .
section iv is devoted to the laser - driven dynamics of an entangled molecular h - h system .
the results obtained are summarized and discussed in the concluding section v.
within the 3d four - body model of h - h excited by the temporally and spatially shaped laser field the total hamiltonian @xmath13 is divided into two parts , @xmath14 where @xmath15 represents the h - h system and @xmath16 describes the interaction of the system with the laser field .
the applied laser field is assumed to be linearly polarized along the @xmath1-axis , the nuclear and the electronic motion are restricted to the polarization direction of the laser electric field .
accordingly , two @xmath1 coordinates of electrons , @xmath2 and @xmath3 , measured with respect to the nuclear center of mass , are treated explicitly together with the internuclear distance @xmath0 .
a similar model has been used previously in @xcite for the h@xmath7 molecule , where each particle , electron or proton , is treated in 1d , i.e. , @xmath1 and @xmath0 . the total non - born - oppenheimer system hamiltonian ( employing a. u. : @xmath17 ) reads @xmath18
@xmath19 + v_{\rm e e}(z_{1},z_{2 } ) , \label{e-3}\ ] ] where @xmath20 is the proton mass , @xmath21 is the reduced electron mass , and non - diagonal mass - polarization terms are neglected . the coulomb potentials in eq .
( [ e-3 ] ) read @xmath22 @xmath23 where @xmath24 , and the regularization parameters , @xmath25 and @xmath26 , have been chosen ( similar to previous work @xcite ) such as to reproduce the ground - state ( gs ) energy of the h - h system at @xmath4 a.u .
( @xmath27 a.u . ) .
the interaction of the h - h system with the laser field is treated within the semiclassical electric dipole approximation by the hamiltonian @xmath28 where @xmath29 , @xmath30 is the vector potential , @xmath31 is the speed of light , and @xmath32 is the spatial - shape function , or envelope , of the laser field . in the general case we set @xmath33 , except for some model simulations detailed in sec .
the vector potential , @xmath30 , is chosen in the following form : @xmath34 where @xmath35 is the amplitude , @xmath36 is the pulse duration at the base , @xmath37 is the laser carrier frequency , and @xmath38 is the carrier - envelope phase ( cep ) . note that it has been shown previously @xcite that the carrier phase of the laser pulse is important only for pulses having less than 15 optical cycles . the definition of the system - field interaction by eq .
( [ e-6 ] ) via the vector potential , suggested in @xcite , assures that the electric field @xmath39 has a vanishing direct - current component , @xmath40 , and satisfies maxwell s equations in the propagation region .
it is suitable to define , on the basis of eqs .
( [ e-6 ] ) and ( [ e-7 ] ) , the local effective - field amplitudes for atoms a and b as follows : @xmath41 where @xmath42 a.u .
, @xmath43 a.u .. the respective time - dependent electric fields acting on atoms a and b read @xmath44 @xmath45 .
\label{e - etime}\ ] ] the first term in eq .
( [ e - etime ] ) corresponds to a laser pulse with a sin@xmath46-type temporal envelope , while the second , the so - called ` switching ' term , appears due to the finite pulse duration .
the 3d time - dependent schrdinger equation for 1d electrons ( @xmath2 , @xmath3 ) and 1d protons ( @xmath0 ) , @xmath47 \psi , \label{tdse}\ ] ] has been solved numerically with the propagation technique adapted from @xcite for both electron and proton quantum motion . in particular , calculations for the electron motion
have been performed by using 200-point non - equidistant grids for the hermite polynomials and corresponding weights for the numerical integration on the interval @xmath48 for the @xmath2 and @xmath3 coordinates .
for the nuclear coordinate @xmath0 , a 256-point equidistant grid has been used on the interval @xmath49 a.u . , @xmath50 a.u.@xmath51 $ ] .
the time - step of the propagation was @xmath52 a.u .
( 1 a.u.=24 asec ) .
the wave functions of the initial states have been obtained by numerical propagation of the equation of motion ( [ tdse ] ) in imaginary time without the laser field ( @xmath53 ) . upon excitation of the h - h system by the laser field , the electronic wave functions of its atomic a and b parts may overlap . in order to study the energy transfer from a to b
, the respective ` atomic ' energies , @xmath54 and @xmath55 , are defined on the basis of eqs .
( [ e-3 ] ) and ( [ e-4 ] ) such that their sum always gives the correct total energy of the entire h - h system .
the definitions of ` atomic ' energies @xmath54 and @xmath55 for the h - h systems of atomic and molecular origin are different due to the different symmetry and entanglement of the respective wave functions and will be specified in the following sections . the ionization probabilities for atoms a and b have been calculated from the time- and space - integrated outgoing fluxes separately for the positive and the negative directions of the @xmath2 and @xmath3 axes at @xmath56 a.u .. specifically , we calculated four ionization probabilities : @xmath57 and @xmath58 for atom a , and @xmath59 and @xmath60 for atom b. at the outer limits of the @xmath1-grids , absorbing boundaries have been provided by imaginary smooth optical potentials adapted from that designed in @xcite .
similar optical potentials have been also provided for the @xmath0-axis but , in practice , the wave - packet never approached the outer limits of the @xmath0-grid .
the spatial part of the initial unentangled direct - product ground - state wave function of h - h of an atomic configuration used in the imaginary time propagation is defined with the unsymmetrized heitler - london electron wave function as follows : @xmath61 where @xmath62 at @xmath42 a.u .
, @xmath63 at @xmath43 a.u . ,
and @xmath64 is a proton gaussian function centered at @xmath4 a.u .. the imaginary - time propagations have been performed with a reduced , ` non - interacting ' atoms , version of the system hamiltonian ( [ e-3 ] ) wherein the nuclear @xmath65 and electronic @xmath66 terms were omitted and only the coulombic interaction @xmath67 of each electron with its nearest proton was taken into account along with the three kinetic - energy terms of eq .
( [ e-3 ] ) .
the results obtained gave a direct - product initial state .
for the sake of comparison , we have also performed the imaginary - time propagations with the complete system hamiltonian ( [ e-3 ] ) .
the results obtained proved to be practically identical to those obtained in the case of two non - interacting h atoms .
the energy difference , for example , is less than @xmath68 a.u . , implying that the coulombic interaction @xmath66 of two h atoms in their ground states is negligible at @xmath4 a.u .. is the probability to find electron @xmath69 initially belonging to atom a at @xmath70 with any values of the other two coordinates , @xmath0 and @xmath3 ; probability @xmath71 has a similar meaning for electron @xmath72 initially belonging to atom b ; ( b ) - spatial envelopes of applied laser fields : the broad gaussian envelope of eq .
( [ e - g ] ) and the narrow envelope of eq .
( [ e - n ] ) . ] the initial unentangled direct - product state of the h - h system representing two non - interacting h atoms , a and b , is presented in fig .
1(a ) by the electron probabilities @xmath73 and @xmath71 , which are defined as follows : @xmath74 for electron @xmath69 initially belonging to atom a with proton @xmath75 , and similarly , @xmath76 for electron @xmath72 initially belonging to atom b with proton @xmath77 .
these electron probabilities give the overall probability to find an electron at a specified point of the @xmath1-axis at any position of the other electron and at any internuclear distance .
after calculation of the unentangled initial state of the atomic h - h system , its laser - driven quantum dynamics in real time has been explored with the complete system hamiltonian of eq .
( [ e-3 ] ) .
the spatial envelopes of the applied laser pulses , gaussian and narrow , defined by eqs .
( [ e - g ] ) and eq .
( [ e - n ] ) below , are presented in fig .
1(b ) to illustrate the local effective - field strengths acting on atoms a and b. two possible choices for the laser carrier frequency are : @xmath78 a.u .
( corresponding to the ground - state energy of an h atom ) , and @xmath5 a.u .
( corresponding to the ground - state energy of h - h at large @xmath0 ) .
our numerical simulations have shown that at @xmath78 a.u .
, an efficient ionization of a takes place , while the energy transfer from a to b and the ionization of b are not efficient .
in contrast , at @xmath5 a.u . , the energy transfer from a to b and the ionization of b are efficient .
this implies the existence of an optimal laser frequency suitable for the most efficient energy transfer .
this issue will be addressed in a forthcoming work ; below we present the results obtained at @xmath5 a.u .. with the laser pulse duration @xmath6 fs , as used throughout this work , the number of optical cycles during the pulse , @xmath79 , is about 33 .
therefore , the carrier phase of the laser field @xmath38 is not important @xcite and set equal to zero in our simulations below [ @xmath80 in eqs .
( [ e-7 ] ) and ( [ e - etime ] ) ] . in order to study the energy transfer from a to b
, the respective ` atomic ' energies , @xmath54 and @xmath55 , have been defined on the basis of eqs .
( [ e-3 ] ) and ( [ e-4 ] ) as follows : @xmath81 @xmath82 for atom a and similarly , @xmath83
@xmath84 for atom b. from eqs .
( [ e - ena - dp ] ) and ( [ e - enb - dp ] ) we see that the kinetic energy of nuclear motion and the potential energies of the proton - proton and the electron - electron interaction are assumed to be equally shared between atoms a and b. the kinetic energy of electron @xmath69 and the energy of coulombic interaction of both electrons , @xmath69 and @xmath72 , with proton @xmath75 are entirely assigned to atom a. similarly , the kinetic energy of electron @xmath72 and the energy of coulombic interaction of both electrons with proton @xmath77 are entirely assigned to atom b. the sum of ` atomic ' energies always gives the correct total energy of the entire h - h system .
notice , in particular , that the fact that the kinetic energy of electron @xmath85 is entirely assigned to atom a / b corresponds to the initial electron probabilities @xmath73/@xmath71 in the unentangled atomic state [ fig .
indeed , electron @xmath69 is localized in the vicinity of proton @xmath75 of atom a and electron @xmath72 is localized in the vicinity of proton @xmath77 of atom b. therefore , if e.g. the laser pulse with a narrow spatial shape [ fig . 1(b ) ] is used to excite the extended h - h system , only the ` atomic ' energy @xmath54 will increase , while @xmath55 will not be affected . the left panel of fig . 2 presents the dynamics of the atomic h - h system excited by the laser pulse with the gaussian spatial envelope @xmath86^{2}\ } , \label{e - g}\ ] ] where @xmath8 a.u .
( 45.56 nm ) and the laser field is assumed to be focused onto a spot centered at @xmath87 , where @xmath88 [ @xmath89 - 1291.5 a.u .
( -68.34 nm ) ] : cf . fig .
the laser pulse parameters are : @xmath5 a.u . , @xmath6 fs , and @xmath90 a.u .
( @xmath91 v / cm , intensity @xmath92 w / cm@xmath46 ) .
accordingly , the spatial envelope function @xmath32 in the interaction hamiltonian of eq .
( [ e-6 ] ) is set equal to @xmath94 of eq .
( [ e - g ] ) , and the effective - field amplitudes for atoms a and b defined by eq .
( [ e - eampl ] ) are : @xmath95 a.u .
( intensity @xmath96 w / cm@xmath46 ) and @xmath97 a.u . , ( @xmath98 w / cm@xmath46 ) , implying a dominant excitation of atom a , which is clearly observed in fig .
2(a ) . from fig .
2(a ) we see that the ` atomic ' energy @xmath54 is controlled by the applied laser pulse : it increases in the first half of the pulse and decreases to the end of the pulse .
in contrast , the ` atomic ' energy @xmath55 does not follow the applied laser pulse : it slowly increases up to the end of the laser pulse and even exceeds @xmath54 shortly after the end of the pulse .
a similar behavior is found for the ` atomic ' ionizations @xmath99 and @xmath100 presented in fig .
2(b ) : while the laser - induced ionization @xmath99 rises fast in the second half of the pulse , the ionization probability @xmath100 sharply rises after the end of the pulse .
such a behavior will be referred to below as a ` sequential ' ionization , implying a delayed interaction due to liet .
the time - dependent expectation values @xmath101 , @xmath102 and @xmath103 are presented in fig .
the time - dependent deviations of the expectation values @xmath102 and @xmath103 , @xmath104 @xmath105 plotted in fig .
2(c ) are scaled up by a factor of two .
we see from fig .
2(c ) that @xmath106 , implying attraction of electron @xmath69 by proton @xmath77 .
in contrast , @xmath107 , implying attraction of electron @xmath72 by proton @xmath75 .
it is also seen from figs .
2(a ) , 2(b ) and 2(c ) that both the energy transfer from a to b and the ` sequential ' ionization of b in the positive direction of the @xmath3-axis correlate with a small decrease of the spatial separation of electrons , @xmath108 , such that the electron - electron repulsion ( eer ) becomes effective .
in order to clarify these findings further , we have performed a similar numerical simulation with a narrow spatial envelope of the laser pulse acting practically only on atom a of the entire h - h system .
specifically , the following sin@xmath46-type spatial envelope of the laser pulse was used for @xmath70 and @xmath109 : @xmath110 , \ ; \ ; \ ; z_{a } \leq z \leq z_{b } , \label{e - n}\ ] ] where @xmath111 a.u .
, @xmath112 a.u . , and @xmath113 otherwise , as illustrated in fig .
accordingly , the spatial envelope function @xmath32 in eq .
( [ e-6 ] ) is set equal to @xmath114 of eq .
( [ e - n ] ) .
the amplitude of the pulse , @xmath115 a.u .
( @xmath116 w / cm@xmath46 ) , is chosen such that the ` atomic ' energy @xmath54 gained in the field with the narrow spatial envelope [ fig .
2(e ) ] is close to that gained with the gaussian spatial envelope [ fig .
note that , defined by eq .
( [ e - etime ] ) , the effective - field strength @xmath117 of the laser pulse with the narrow spatial envelope [ figs .
1(b ) and 2(h ) ] is much smaller than that of the laser pulse with the gaussian spatial envelope [ figs .
1(b ) and 2(d ) ] , but the energy transfer from a to b is by about 30% more efficient [ fig .
also note that , due to the effective - field amplitude acting on atom b in the current case of the narrow spatial envelope of the pulse is @xmath118 , the efficient energy transfer to atom b is attributed entirely to the electron - electron repulsion @xmath66 and electron - proton attraction @xmath67 in the vicinity of atom b ( @xmath119 ) .
this also applies to the ionization of atom b in the positive direction of the @xmath3-axis [ fig .
2(f ) ] , which is not only by about 30% more efficient than that induced by the laser pulse with the broad gaussian spatial envelope [ fig .
2(b ) ] , but @xmath100 is even larger than @xmath99 at @xmath120 fs [ fig .
2(f ) ] , implying that the ` sequential ' ionization @xmath100 , induced by eer , is more efficient than the laser - induced ionization @xmath99 . taking into account that in the unentangled direct - product state ( [ ini - dp ] ) , electrons @xmath69 and @xmath72 are well localized on the @xmath1-axis [ fig .
1(a ) ] , one can assume that the time - dependent expectation values @xmath102 and @xmath103 represent the positions of electrons
. a closer look at the data presented in figs .
2(c ) and 2(g ) for @xmath102 and @xmath103 shows that the two electrons of h - h do not approach each other closer than 98 a.u .
( 5.18 nm ) .
therefore , at a first glance , the energy transfer from a to b and the ` sequential ' ionization of b take place , similarly to icd , without any noticeable overlap of the respective electronic wave functions .
nevertheless , a closer look at spatial distributions of the electronic wave functions illustrated in fig .
3 reveals that eer is in fact a short - range process which takes place in the vicinity of atom b at @xmath43 a.u .. here , the coulombic attraction of electron @xmath69 by proton @xmath77 effectively localizes electron @xmath121 .
accordingly , the overlap of the electronic wave functions has a sharp narrow maximum at @xmath43 a.u .
3 ) , which results in a strong eer .
electron probabilities @xmath73 and @xmath71 presented in fig .
3 correspond to the end of the 5 fs laser pulse with the narrow spatial envelope defined by eq .
( [ e - n ] ) .
this pulse excites practically only electron @xmath69 in the vicinity of @xmath42 a.u .
, therefore the probability distribution @xmath73 at @xmath122 fs is broadened in comparison to the initial one [ fig .
1(a ) ] in both @xmath123 and @xmath124 directions .
the laser - induced extension of @xmath73 into the domain of @xmath123 gives rise to the ionization @xmath99 on the negative direction of the @xmath2-axis . at @xmath125 a.u .
, the wave function is absorbed by the imaginary optical potential .
in contrast , at @xmath124 , the laser - driven electron @xmath69 reaches the domain @xmath126 where it is attracted and accelerated by proton @xmath77 , localized at @xmath43 a.u .. due to the electron - proton coulombic attraction , the probability @xmath73 has its local maximum at @xmath127 , corresponding to liet of ref .
simultaneously , when electron @xmath69 approaches the domain of @xmath43 a.u . , where the initial probability @xmath71 of electron @xmath72 has the global maximum [ fig .
1(a ) ] , the electron - electron coulombic repulsion becomes very strong .
therefore , the probability @xmath71 is extended into the domain of @xmath128 ( fig .
3 ) , giving rise to the ionization @xmath100 in the positive direction of the @xmath3-axis .
finally , the wave function is absorbed by the imaginary optical potential at @xmath129 a.u .. note that in the current case of a narrow spatial envelope of the laser field [ fig .
1(b ) ] one can clearly distinguish between the laser - induced ionization on the negative direction of the @xmath2-axis , represented in fig .
2(f ) by @xmath57 , and the ` sequential ' eer - induced ionization on the positive direction of the @xmath3-axis , represented in fig .
2(f ) by @xmath60 .
also note that the ionization probabilities @xmath130 and @xmath131 are less than 10@xmath132 at @xmath122 fs .
( solid line ) and @xmath71 ( dashed line ) at the end of the 5 fs laser pulse with a narrow spatial envelope [ eq .
( [ e - n ] ) and fig .
1(b ) ] which excites only electron @xmath69 ( coordinate @xmath2 ) in the domain of atom a. ] ( solid line ) and @xmath71 ( dashed line ) at 2.5 fs after the end of the 5 fs laser pulse with a gaussian spatial envelope which excites both electrons , with the excitation of electron @xmath69 ( coordinate @xmath2 ) being dominant in comparison to that of electron @xmath72 ( coordinate @xmath3 ) . ] in the case of the broad gaussian spatial envelope of the laser pulse centered at @xmath133 a.u .
( -68.34 nm ) , both electrons , @xmath69 and @xmath72 , are excited by the laser field and therefore ionization in both positive and negative directions of the @xmath1-axis is induced by both the laser field and subsequently by eer .
the electron probabilities @xmath73 ( solid line ) and @xmath71 ( dashed line ) for the gaussian spatial envelope of the 5 fs laser pulse are plotted in fig . 4 at @xmath134 fs , when both ionization probabilities plotted in fig .
2(b ) approach their maximum values .
although both electrons are excited by the laser field , the excitation of electron @xmath69 in the domain of atom a is about 1.4 times stronger than the excitation of electron @xmath72 in the domain of atom b. accordingly , the local maximum of @xmath73 at @xmath43 a.u .
4 ) is almost 4 times higher than the local maximum of @xmath71 at @xmath42 a.u . and therefore the eer - induced ionization in the positive direction of the @xmath3-axis is stronger than the eer - induced ionization in the negative direction of the @xmath2-axis . in contrast ,
the laser - induced ionization in the negative direction of the @xmath2-axis is stronger than that in the positive direction of the @xmath3-axis .
indeed , it is seen from fig .
2(b ) that at the end of the 5 fs laser pulse , the ionization probability @xmath100 is more than twice smaller than @xmath99 , while @xmath135 at @xmath13610 fs .
it is seen from fig . 4 that both local maxima , @xmath73 at @xmath137 a.u . and
@xmath71 at @xmath138 a.u . , are significantly smaller than the local maximum of @xmath73 at @xmath137 a.u . produced by the laser pulse with a narrow spatial envelope
( see fig .
therefore the eer - induced ionization in the case of the broad gaussian spatial envelope of the laser pulse [ fig .
2(b ) ] is smaller than that in the case of the narrow spatial envelope of the pulse [ fig .
2(f ) ] .
it can be concluded from fig . 4 that in the general case , when both distant atoms are excited by the laser field , energy is transferred from a to b and from b to a , and eer - induced ionization occurs in both a and b parts of h - h along with the laser - induced ionization . if , for example , the laser pulse has a wide gaussian spatial envelope and/or is centered at @xmath139 , the mutual energy transfers and the eer - induced ionization probabilities are substantial for both distant atoms a and b even at a large internuclear separation .
finally we note that the appearance of sharp local maxima of @xmath73 at @xmath140 a.u . and of @xmath71 at @xmath141 a.u .
( see figs . 3 and 4 ) confirm spreading , delocalization , and non - factorization of the initially factorized and well localized on the @xmath1-axis wave function of h - h [ fig .
1(a ) ] and can therefore @xcite be treated as the emergence of long - range entanglement , or quantum non - local connection , at @xmath4 a.u .. recently this long - range entanglement attracted considerable interest both in theory and experiment .
the h - h molecular system represents an elongated configuration of the h@xmath7 molecule .
therefore , its initial electronic ground state is entangled via exchange .
the spatial part of the initial entangled ground - state wave function of h - h for a singlet electronic state is given by @xmath142 @xmath143 \psi_{\rm g}(r ) , \label{ini - enp}\ ] ] where @xmath144 are defined by eq .
( [ ini - dp ] ) and @xmath64 is a proton gaussian function centered at @xmath4 a.u .. imaginary time propagations have been performed with the complete system hamiltonian ( [ e-3 ] ) . in the entangled initial state of the h - h system , electron probabilities @xmath73 and @xmath71
are identical to each other .
for the simulations with this entangled initial state ( [ ini - enp ] ) , the ` atomic ' energies @xmath54 and @xmath55 are defined on the basis of eqs .
( [ e-3 ] ) and ( [ e-4 ] ) as follows : @xmath145 @xmath146 for atom a , and @xmath147 @xmath148 for atom b. from eqs .
( [ e - ena - enp ] ) and ( [ e - enb - enp ] ) we see that the kinetic energies of nuclear and electronic motion as well as the potential energies of the proton - proton and the electron - electron coulombic interaction are assumed to be equally shared between atoms a and b. the energy of coulombic interaction of both electrons with proton @xmath75 is assigned to atom a , while the energy of coulombic interaction of both electrons with proton @xmath77 is assigned to atom b. the sum of these ` atomic ' energies always gives the correct total energy of the entire h - h system . the choice of the electron kinetic energies in eqs .
( [ e - ena - enp ] ) and ( [ e - enb - enp ] ) corresponds to the initial electron probabilities @xmath73 and @xmath71 in the entangled molecular state
. both electrons @xmath69 and @xmath72 are localized with the 50% probability in the vicinity of proton @xmath75 of atom a and if the extended h - h system is excited e.g. by the narrowly shaped laser pulse of fig .
1(b ) , both electrons give rise to the ` atomic ' energy @xmath54 , each with 50% probability .
the quantum dynamics of h - h excited from the entangled initial state ( [ ini - enp ] ) by spatially shaped laser pulses is presented in fig .
the left panel corresponds to the broad gaussian spatial envelope centered at @xmath133 a.u .
( -68.34 nm ) and the right panel corresponds to the narrow spatial envelope centered at @xmath42 a.u .
( -2.65 nm ) , see fig .
for the sake of comparison , the laser fields used are the same as in the previous case of the unentangled direct - product initial state [ see figs .
1(b ) and 2 ] .
specifically , the gaussian spatial envelope is defined by eq .
( [ e - g ] ) , the narrow spatial envelope is defined by eq .
( [ e - n ] ) , the carrier frequency of the laser pulse with the sin@xmath46-type temporal envelope of eq .
( [ e - etime ] ) is @xmath5 a.u . ,
and the pulse duration at the base is @xmath6 fs . from the comparison of the quantum dynamics of the unentangled state [ figs .
2(a ) and 2(e ) ] to that of the entangled state [ figs . 5(a ) and 5(d ) ] the following observations are made .
\(i ) the overall energy @xmath149 transferred on a long timescale of @xmath150 fs from atom a to atom b in the entangled state is very similar to that transferred from a to b in the unentangled direct - product state : @xmath151 a.u . for the broad gaussian spatial envelope , and @xmath152 a.u .
for the narrow spatial envelope .
\(ii ) the maximum ` atomic ' energy @xmath54 gained during the laser pulse by atom a in the entangled state is smaller than that gained in the unentangled direct - product state .
in contrast , the maximum ` atomic ' energy @xmath55 gained during the pulse by atom b in the entangled state is substantially larger than that gained in the unentangled direct - product state .
\(iii ) moreover , in the case of the entangled initial state , the ` atomic ' energy @xmath55 is controlled by the laser pulse similarly to @xmath54 , even when only atom a is excited by the laser pulse with a narrow spatial envelope : the energy @xmath55 increases in the first half of the laser pulse and decreases at the end of the pulse , in contrast to the case of the unentangled direct - product initial state .
this observation implies that the changes made by the applied laser field to the entangled wave function in the domain of atom a at @xmath153 a.u .
result in simultaneous changes in atom b at @xmath154 a.u . due to the symmetry of the wave function by exchange .
such an entangled behaviour is very important for a long - range quantum communication among distant quantum systems @xcite .
\(iv ) after the end of the laser pulse , the ` atomic ' energies @xmath54 and @xmath55 demonstrate out - of - phase oscillations : slow oscillations in the case of the unentangled direct - product initial state [ figs .
2(a ) and 2(e ) ] and slow oscillations modulated with very fast ones in the case of the entangled initial state [ figs .
5(a ) and 5(d ) ] .
the amplitudes of both slow and fast oscillations of @xmath54 and @xmath55 decrease with time , which may indicate the formation of a quasi - stable configuration of the excited h - h system . ) ] by the laser pulse with a narrow spatial envelope [ eq . ( [ e - n ] ) ] acting on atom a. ( a ) - time - dependent ` atomic ' energies @xmath54 ( solid line ) and @xmath55 ( dashed line ) ; ( b ) - electron - field in - phase following ; ( c ) - the laser field acting on atom a. ] it is instructive , before proceeding , to consider the laser - driven dynamics of the entangled h - h system in the initial stage of its excitation by the laser pulse with a narrow spatial envelope ( [ e - n ] ) which excites electrons @xmath69 and @xmath72 only in the domain of atom a. in fig . 6 , the laser - driven dynamics of entangled h - h is presented on the timescale of 1 fs .
time - dependent ` atomic ' energies , @xmath54 and @xmath55 , and expectation values of electronic coordinates , @xmath102 and @xmath103 , are shown in figs .
6(a ) and 6(b ) respectively .
the time - dependent laser field is shown in fig .
it is seen from fig .
6(a ) that the ` atomic ' energies @xmath54 and @xmath55 oscillate out - of - phase with respect to each other , with @xmath54 being in - phase and @xmath55 being out - of phase with the applied laser field [ fig . 6(c ) ] . as seen from figs .
6(b ) and 6(c ) the electrons also follow the applied laser field .
note that expectation values @xmath102 and @xmath103 in the entangled h - h system are identical and not distinguishable in fig .
6(b ) . a perfect electron - field following at the laser carrier frequency being as high as 1 a.u .
is very interesting .
previously , the electron - field following on the level of expectation values of electronic coordinates have been explored only in the infrared @xcite and near - infrared @xcite domains of the laser carrier frequency .
the electron - field following at high laser frequencies is reminiscent of the well - known recollision model of corkum @xcite .
however , an important feature of fig .
6(b ) is that electrons follow the applied laser field in - phase , while according to the theoretical model used in @xcite and numerical results of @xcite electrons follow the field out - of - phase : @xmath155 decreases when electric - field strength @xmath156 increases .
a detailed study of the electron dynamics showed that , probably due to their finite albeit very small mass , the electron do not react to the first half - cycle of the applied field at @xmath157 a.u . and follows the field in - phase at @xmath158 fs . on the other hand ,
the out - of - phase electron - field following take place at @xmath159 a.u .. similar results have been obtained for the unentangled h - h system excited by the laser pulse with a narrow spatial envelope : the electron @xmath69 of atom a follows the applied field in - phase at @xmath157 a.u . and out - of - phase at @xmath159 a.u
.. these results can be explained as follows . at @xmath159 a.u .
, the electron is in the ground state , the polarizability is negative and , therefore , the electron follows the applied laser field out - of - phase .
in contrast , at @xmath157 a.u . , the electron is well above the excited state , the polarizability changes the sign , and electron follows the field in - phase .
similar behaviour has been observed in a recent work @xcite for molecular ion h@xmath12 excited at wavelength @xmath160800 nm ( @xmath161 a.u . ) . at fixed @xmath162 a.u . ,
the electron is in the ground state and follows the field out - of - phase , whereas at @xmath163 a.u .
, the laser carrier frequency @xmath161 a.u . is larger than the energy difference @xmath164 , and the electron follows the field in - phase .
coming back to the laser - driven dynamics of the entangled h - h system on the long timescale of 20 fs ( fig . 5 )
, we note that ionization of atom b starts only in the second half of the laser pulse [ see figs .
5(b ) and 5(e ) ] .
due to the symmetry of the entangled wave function , the ionization probabilities of electrons @xmath69 and @xmath69 are identical to each other both for the positive and the negative directions of the @xmath1-axes : @xmath165 and @xmath166 .
therefore , in figs .
5(b ) and 5(e ) the total ionization probabilities , @xmath167 @xmath168 are plotted . from the comparison of figs .
2(b ) and 2(f ) to figs .
5(b ) and 5(e ) one can easily see that the total time - dependent ionization probabilities @xmath169 and @xmath170 used in the case of the entangled initial state are almost equal ( but not identical ) to the time - dependent ionization probabilities @xmath57 and @xmath60 used in the case of the unentangled direct - product initial state .
one can conclude therefore that the entanglement of the initial state of h - h , including its exchange symmetry , does not change its ionization probability in comparison to the non - symmetric unentangled direct - product initial state .
the other consequence of the exchange symmetry of the entangled wave function is a very small spatial separation of electrons all over the @xmath1-grid .
the time - dependent expectation values @xmath102 and @xmath103 are almost identical , both being close to 0 on the timescale of 20 fs . if the perfect symmetry of the wave functions is even slightly changed due to the excitation of h - h by the laser field , the very small spatial separation of the electrons leads to their strong coulombic repulsion all over the @xmath1-grid .
this suggests the reason for the fast out - of - phase oscillations of ` atomic ' energies @xmath54 and @xmath55 after the end of the laser pulse [ see observation ( iv ) above and figs .
5(a ) and 5(d ) ] .
the fast out - of - phase oscillations of the ` atomic ' energies occur in the case of the entangled initial state and do not occur in the case of the unentangled direct - product initial state [ figs .
2(a ) and 2(e ) ] . in the case of the direct - product initial state ,
the minimum difference between @xmath102 and @xmath103 is about 98 a.u .
( 5.18 nm ) [ see figs .
2(c ) and 2(g ) ] , therefore eer is very weak all over the @xmath1-grid , except for the special cases of very sharp overlaps of the electronic wave functions in the vicinity of protons @xmath75 and @xmath77 ( see figs . 3 and 4 ) .
a very strong eer taking place in the case of the entangled initial state should also result in a very efficient eer - induced ` sequential ' ionization , similar to that described in sec .
iii for the unentangled direct - product initial state of h - h . on the other hand ,
it was shown above that the entanglement of the initial state of h - h does not change its ionization probability in comparison to the direct - product initial state .
therefore , a closer look at the process of eer - induced ionization in the case of the entangled initial state is required . to this end
, we performed several model simulations with the entangled initial state excited in the domain of atom a by the laser pulses with narrow spatial envelopes defined by eq .
( [ e - n ] ) for various model assumptions of the system - field interaction to be specified below .
the results obtained are presented in figs .
7 and 8 . electron probabilities @xmath73 and @xmath71 presented in figs .
7(a ) and 7(b ) correspond to the end of the 5 fs laser pulse with a narrow spatial envelope [ fig .
1(b ) ] which excites both electrons @xmath69 and @xmath72 in the vicinity of @xmath42 a.u .
( atom a ) and does not affect electrons @xmath69 and @xmath72 in the vicinity of @xmath43 a.u .
( atom b ) . apparently , electron probabilities @xmath73 and @xmath71 , presented in figs .
7(a ) and 7(b ) at @xmath122 fs , are extended in comparison to their initial ones in both @xmath171 and @xmath172 directions .
the laser - induced extension of @xmath73 and @xmath71 into domains @xmath171 gives rise to the ionization @xmath99 in the negative directions of the @xmath173-axes . at @xmath174 a.u .
, the wave function is absorbed by the imaginary optical potentials .
in contrast , at @xmath172 , the laser - driven electrons @xmath69 and @xmath72 reach the domains @xmath175 where they are attracted and accelerated by proton @xmath77 , localized at @xmath43 a.u .
, more and more efficiently . due to the electron - proton coulomb attraction , electron probabilities @xmath73 and @xmath71 increase in the vicinity of @xmath176 and thus destroy the initially perfect symmetry of the entangled wave function , as clearly seen from figs .
7(a ) and 7(b ) .
the extension of electron probabilities @xmath73 and @xmath71 into the domains of @xmath177 gives rise to the eer - induced ` sequential ' ionization @xmath100 in the positive directions of the @xmath173-axes .
the only problem to be clarified now is eer in the domain of atom b , because at a first glance one could conclude from figs .
7(a ) and 7(b ) that the electrons coming from atom a to atom b occupy the domains of @xmath177 .
in order to clarify this issue , we performed two model simulations with a 5 fs laser pulse having narrow spatial envelope , assuming that only one of the two electrons is excited by the laser field in the domain of atom a. the results obtained are presented in figs .
7(c ) and 7(d ) .
first we assume that only electron @xmath69 is excited by the laser field in the domain of atom a. to this end , we chose the spatial envelope function @xmath32 in the interaction hamiltonian of eq .
( [ e-6 ] ) as follows : @xmath178 and @xmath179 , where the narrow spatial envelope @xmath114 is given by eq .
( [ e - n ] ) and illustrated in fig .
the respective electron probabilities @xmath73 and @xmath71 at the end of the pulse ( @xmath122 fs ) are shown in fig .
secondly , we assume that only electron @xmath72 is excited by the laser field in the domain of atom a. accordingly , the spatial envelope function @xmath32 in the interaction hamiltonian of eq .
( [ e-6 ] ) is chosen as follows : @xmath180 , and @xmath181 .
the respective electron probabilities @xmath73 and @xmath71 at the end of the pulse ( @xmath122 fs ) are shown in fig .
it is seen from the results presented in figs .
7(c ) and 7(d ) that the eer - induced ionization of atom b with the entangled initial state proceeds similar to that with the unentangled direct - product initial state described in sec .
iii above .
specifically , [ see fig .
7(c ) ] the laser - driven electron @xmath69 ( coordinate @xmath2 ) comes from atom a to atom b and pushes electron @xmath72 ( coordinate @xmath3 ) out into the domain @xmath128 , where the eer - induced ionization of electron @xmath72 takes place in the positive direction of the @xmath3-axis . at the same time ,
electron @xmath69 which arrived at atom b does not affect the probability @xmath73 at @xmath182 a.u .. similarly , [ fig .
7(d ) ] , the laser - driven electron @xmath72 ( coordinate @xmath3 ) , coming from atom a to atom b , pushes electron @xmath69 ( coordinate @xmath2 ) into the domain @xmath183 , where the eer - induced ionization of electron @xmath69 takes place in the positive direction of the @xmath2-axis .
electron @xmath72 does not affect the probability @xmath71 at @xmath184 a.u .. in a realistic case , when both electrons are excited by the laser field [ figs .
7(a ) and 7(b ) ] , both processes take place simultaneously . in the domain of atom a : probability differences @xmath185 ( solid line ) and @xmath186 ( dashed line ) in the domain of atom b at the end of the 5 fs laser pulse .
the changes of the electron probabilities @xmath73 and @xmath71 are measured with respect to their initial values at @xmath187 [ see eq .
( [ e - dps ] ) ] . ] to identify local maxima of electronic wave functions , we next define the probability differences for electrons , @xmath188 and @xmath189 , with respect to the initial electronic probabilities as follows : @xmath190 @xmath191 the probability differences @xmath185 and @xmath186 plotted in fig
. 8 are calculated at the end of the 5 fs laser pulse with a narrow spatial envelope which excites in the domain of atom a only electron @xmath69 and does not affect electron @xmath72 . it can be concluded from figs .
7(c ) and 8 that the laser - driven electron @xmath69 , coming to the domain @xmath126 , is attracted and accelerated therein by proton @xmath77 and forms at @xmath122 fs a local double - peak maximum at @xmath43 a.u . , where electron @xmath72 already has its global maximum .
due to a very strong coulombic repulsion of the two electrons in the vicinity of @xmath43 a.u .
, a sharp hole appears in the global maximum of @xmath71 at the expense of the increasing electron probability @xmath71 in the domain of @xmath128 , thus giving rise to the eer - induced ionization of electron @xmath72 in the positive direction of the @xmath3-axis .
a similar process takes place when only electron @xmath72 is excited by the laser pulse with a narrow spatial envelope , giving rise to the eer - induced ionization of electron @xmath69 in the positive direction of the @xmath2-axis . in a realistic case ,
when both electrons are excited by the laser field in the domain of atom a , both described above processes take place simultaneously .
we conclude from the results presented in figs . 7 and 8 that the physical reasons of the energy transfer from atom a to atom b and the ` sequential ' ionization of atom b in the case of the entangled initial state and a laser pulse with a narrow spatial envelope , which excites only electrons of atom a , are similar to those underlying the case of the direct - product initial state : the coulombic attraction of the laser - driven electrons by proton @xmath77 , and the short - range coulombic repulsion of the two electrons in the vicinity of proton @xmath77 , where their wave functions strongly overlap .
in the present work we have studied numerically the non - born - oppenheimer quantum dynamics of two distant h atoms ( being referred to as atoms a and b ) with an arbitrary large initial internuclear separation of @xmath4 a.u .
( 5.29 nm ) , which have been excited by spatially shaped laser pulses .
we have found an efficient energy transfer from one h atom to the other and a ` sequential ' ionization of the latter , induced by a short - range eer .
the short - range eer , taking place e.g. in the vicinity of proton @xmath77 of atom b , occurs due to the preceding long - range liet from atom a to atom b enhanced by the coulombic attraction and acceleration of the laser - driven electrons by proton @xmath77 .
both unentangled direct - product atomic states and entangled molecular states have been used as the initial states of the h - h system in our numerical simulations .
the first case , unentangled atomic states , is more likely to arise in an experiment : at a gas pressure of 1 atm .
, for example , the interatomic distance is about 100 a.u .. the second case , entangled molecular states , requires an additional step : dissociation of the h@xmath7 molecule , or generation of entanglement between two individual atoms .
the field strengths of the spatially shaped laser pulses were chosen such as not to induce a strong ionization and the respective decrease of the overall norm of the wave packet of h - h . in stronger fields and with longer pulses both the energy transfer and
the ` sequential ' ionization can be more efficient than those presented above .
the parameters of the laser pulses used in the present work ( in particular the pulse duration , @xmath6 fs , and especially the laser carrier frequency , @xmath192 a.u . )
are not yet optimal , and more efforts will be required in order to find optimal spatially shaped laser fields suitable for the most efficient energy transfer and ` sequential ' ionization .
we have shown that in the case of a narrow spatial envelope of the applied laser field , when only electrons initially belonging to atom a are excited by the laser field , the physical mechanisms of the energy transfer from atom a to atom b and the ` sequential ' ionization of atom b are as follows : ( i ) the coulombic attraction of the laser - driven electrons by proton @xmath77 of atom b , resulting in the formation of narrow local maxima of the electronic wave functions in the vicinity of proton @xmath77 , and ( ii ) the short - range coulomb repulsion of the two electrons in the vicinity of proton @xmath77 , where their wave functions strongly overlap . in a more general case of a wide spatial envelope of the laser field , for example a broad gaussian ,
when both a and b atoms are excited by the field simultaneously , the same processes occur in the opposite direction as well : the energy is also transferred from atom b to atom a and the eer - induced ` sequential ' ionization also occurs in atom a. furthermore , we have shown that long - range entanglement of the initially unentangled direct - product state is established by the interaction with a spatially shaped laser pulse .
long - range entanglement attracts considerable interest these days .
for example , entanglement transfer from dissociated molecules to photons has been explored in @xcite , where the concept of transferring the quantum state of two dissociated fragments sharing internal - translational entanglement to that of two photons and vice versa has been put forward . a special kind of an entangled state of one electron and two protons in a h@xmath12 molecule adiabatically stretched to @xmath193 a.u .
( 1.06 - 5.29 nm ) has been studied recently in @xcite .
a quantum - gate mechanism based on electron spins in coupled semiconductor quantum dots , separated by 40 nm , has been considered in @xcite as an important application of long - range entanglement .
such gates provide a source of spin entanglement and can be used for quantum computers .
experimental generation of entanglement between two individual @xmath194rb atoms held in two optical tweezers , separated by 4 @xmath195 m , has been reported in @xcite .
finally we mention that quantum entanglement among extended biomolecules has been explored in @xcite ( see also references therein ) .
the results obtained in @xcite demonstrate that there exists robust quantum entanglement among chlorophyll molecules under physiological conditions for the case of a single elementary excitation .
the main characteristic features of the entangled molecular initial states in our h - h model , as compared to the unentangled atomic direct - product ones , are as follows : ( i ) the immediate response of atom b to the laser excitation of atom a , and ( ii ) the existence of fast out - of - phase oscillations of the ` atomic ' energies @xmath54 and @xmath55 after the end of the applied laser pulse .
these features should occur in a long - range quantum communication among distant quantum systems for the following reason .
the time - dependent ` atomic ' energies @xmath54 and @xmath55 represent the information encoded in the laser pulse . therefore ,
if only one atom a of the entangled a - b pair is excited by the laser pulse , the encoded information reaches atom b ( receiver ) instantaneously and , moreover , can be decoded both during and after the end of the pulse acting on atom a. we have found that in the case of a narrow spatial envelope of an applied laser pulse , which excites only electrons belonging to atom a , the ` sequential ' eer - induced ionization of atom b is more efficient than the laser - induced ionization of atom a [ see figs .
2(f ) and 6(e ) ] .
these results suggest important consequences for more complex quantum systems , e.g. those composed of three and more distant sub - systems , such as @xmath196 .
if we excite by a laser pulse with a narrow spatial envelope only sub - system @xmath197 , for example , and if ionization probabilities obey @xmath198 , we can stimulate an efficient chain of ` sequential ' ionization events with ever increasing probability in the total @xmath199 system .
the same may also apply to the energy transfer in the @xmath199 system .
to achieve such collective energy transfer and ` sequential ' ionization , a narrow spatial shaping of the laser field as reported in @xcite needs further exploration . to achieve the condition @xmath198 in the @xmath199 systems with a gaussian spatial envelope of eq.([e - g ] ) , we can always choose @xmath200 such that the effective - field amplitudes obey @xmath201 .
in this case we can anticipate that the condition @xmath198 is fulfilled at the edge of the gaussian envelope . finally , at a shorter pulse duration and/or at a smaller laser carrier frequency , such that the number of optical cycles per pulse is @xmath202 ( see ref .
@xcite ) , the role of cep - effects as explored previously in liet @xcite offers new avenues for energy and information transfer at long - range distances .
this work has been financially supported by the deutsche forschungsgemeinschaft through the sfb 652 ( g.k.p .
, o.k . ) which is gratefully acknowledged , and a.d.b .
thanks the humboldt foundation for the financial support through a humboldt research award .
t. jahnke , a. czasch , m.s .
schffler , s. schssler , a. knapp , m. kasz , j. titze , c. wimmer , k. kreidi , r.e .
grisenti , a. staudte , o. jagutzki , u. hergenhahn , h. schmidt - bcking , r. drner , phys .
* 93 * , 163401 ( 2004 ) .
t. jahnke , h. sann , t. havermeier , k. kreidi , c. stuck , m. meckel , m. schffler , n. neumann , r. wallauer , s. voss , a. czasch , o. jagutzki , a. malakzadeh , f. afaneh , t. weber , h. schmidt - bcking , r. drner , nature phys .
* 6 * , 74 ( 2010 ) .
t. havermeier , t. jahnke , k. kreidi , r. wallauer , s. voss , m. schffler , s. schssler , l. foucar , n. neumann , j. titze , h. sann , m. kuhnel , j. voigtsberger , j.h .
morilla , w. schollkopf , h. schmidt - bcking , r.e .
grisenti , r. drner , phys .
lett . * 104 * , 133401 ( 2010 ) .
v. may , o. khn , _ charge and energy transfer dynamics in molecular system , 2nd revised and enlarged edition _ , ( wiley - vch , weinheim , 2004 ) ; a. nitzan , _ chemical dynamics in condensed phases _ , ( oxford , university press , 2006 ) | the quantum dynamics of two distant h atoms excited by ultrashort and spatially shaped laser pulses is studied by the numerical solution of the non - born - oppenheimer time - dependent schrdinger equation within a three - dimensional ( 3d ) model , including the internuclear distance @xmath0 and the two @xmath1 coordinates of the electrons , @xmath2 and @xmath3 .
the two 1d hydrogen atoms , a and b , are assumed to be initially in their ground states with a large ( but otherwise arbitrary ) internuclear separation of @xmath4 a.u .
( 5.29 nm ) .
two types of a spatial envelope of a laser field linearly polarized along the @xmath1-axis are considered : ( i ) a broad gaussian envelope , such that atom a is excited by the laser field predominantly , and ( ii ) a narrow envelope , such that practically only atom a is excited by the laser field . with the laser carrier frequency @xmath5 a.u . and the pulse duration @xmath6 fs , in both cases an efficient energy transfer from atom a to atom b has been found .
the ionization of atom b achieved mostly after the end of the laser pulse is close to or even higher than that of atom a. it is shown that with a narrow spatial envelope of the laser field , the underlying mechanisms of the energy transfer from a to b and the ionization of b are the coulomb attraction of the laser driven electron by the proton of atom b and a short - range coulomb repulsion of the two electrons when their wave functions significantly overlap in the domain of atom b. in the case of a broad gaussian spatial envelope of the laser field , the opposite process also occurs , but with smaller probability : the energy is transferred from the weakly excited atom b to atom a , and the ionization of atom a is also induced by the electron - electron repulsion in the domain of atom a due to a strong overlap of the electronic wave functions . |
in differential geometry one finds many different normal forms results which share the same flavor . in the last few years we have come to realize that there is more than a shared flavor to many of these results : they are actually instances of the same general result . the result in question is a linearization result for lie groupoids , first conjectured by alan weinstein in @xcite .
the first complete proof of the linearization theorem was obtained by nguyen tien zung in @xcite . since then several clarifications and simplifications of the proof , as well as more general versions of this result , were obtained ( see @xcite ) . in these lectures
notes we give an overview of the current status of the theory .
the point of view followed here , which was greatly influenced by an ongoing collaboration with matias del hoyo @xcite , is that the linearization theorem can be thought of as an ehresmann s theorem for a submersion onto a stack .
hence , its proof should follow more or less the same steps as the proof of the classical ehresmann s theorem , which can be reduced to a simple argument using the exponential map of a metric that makes the submersion riemannian .
although i will not go at all into geometric stacks ( see the upcoming paper @xcite ) , i will adhere to the metric approach introduced in @xcite .
let us recall the kind of linearization theorems that we have in mind .
the most basic is precisely the following version of ehresmann s theorem : let @xmath0 be a * proper * surjective submersion .
then @xmath1 is locally trivial : for every @xmath2 there is a neighborhood @xmath3 , a neighborhood @xmath4 , and diffeomorphism : @xmath5_{\text{pr } } \ar[rr]^{\cong } & & \pi^{-1}(u)\subset m\ar[d]^{\pi}\\ v\ar[rr]^{\cong } & & u } \ ] ] one can also assume that there is some extra geometric structure behaving well with respect to the submersion , and then ask if one can achieve `` linearization '' of both the submersion and the extra geometric structure .
for example , if one assumes that @xmath6 is a closed 2-form such that the pullback of @xmath7 to each fiber is non - degenerate , then one can show that @xmath1 is a locally trivial symplectic fibration ( see , e.g. , @xcite ) .
we will come back to this later , for now we recall another basic linearization theorem : let @xmath8 be a foliation of @xmath9 and let @xmath10 be a compact leaf of @xmath8 with * finite holonomy*. then there exists a saturated neighborhood @xmath11 , a @xmath12-invariant neighborhood @xmath13 , and a diffeomorphism : @xmath14^{\cong } & & u\subset m}\ ] ] sending the linear foliation to @xmath15 . here
, @xmath16 denotes the holonomy cover , a @xmath12-principal bundle , and the holonomy group @xmath12 acts on the normal space @xmath17 via the linear holonomy representation . by
linear foliation " we mean the quotient of the horizontal foliation @xmath18 , @xmath19 .
notice that this result generalizes ehresmann s theorem , at least when the fibers of the submersion are connected : any leaf of the foliation by the fibers of @xmath1 has trivial holonomy so @xmath12 acts trivially on the transversal , and then reeb s theorem immediately yields ehresmann s theorem .
for this reason , maybe it is not so surprising that the two results are related .
let us turn to a third linearization result which , in general , looks to be of a different nature from the previous results .
it is a classical result from equivariant geometry often referred to as the slice theorem ( or tube theorem ) : let @xmath20 be a lie group acting in a * proper * fashion on @xmath9 . around any orbit @xmath21
the action can be linearized : there exist @xmath20-invariant neighborhoods @xmath22 and @xmath23 and a @xmath20-equivariant diffeomorphism : @xmath24^{\cong } & & u\subset m}\ ] ] here @xmath25 acts on the normal space @xmath26 via the normal isotropy representation .
if the action is locally free then the orbits form a foliation , the isotropy groups @xmath27 are finite and @xmath28 is a quotient of @xmath27 . moreover
, the action of @xmath29 on a slice descends to the linear holonomy action of @xmath28 .
the slice theorem is then a special case of the reeb stability theorem .
however , in general , the isotropy groups can have positive dimension and the two results look apparently quite different .
again , both in the case of foliations and in the case of group actions , we could consider extra geometric structures ( e.g. , a metric or a symplectic form ) and ask for linearization taking into account this extra geometric structure .
one can find such linearization theorems in the literature ( e.g. , the local normal form theorem for hamiltonian actions @xcite ) .
let us mention one such recent result from poisson geometry , due to crainic and marcut @xcite : let @xmath30 be a poisson manifold and let @xmath31 be a compact symplectic leaf . if the poisson homotopy bundle @xmath32 is a smooth * compact * manifold with vanishing second de rham cohomology group , then there is a neighborhood @xmath33 , and a poisson diffeomorphism : @xmath34 we will not discuss here the various terms appearing in the statement of this theorem , referring the reader to the original work @xcite .
however , it should be clear that this result has the same flavor as the previous ones : some compactness type assumption around a leaf / orbit leads to linearization or a normal form of the geometric structure in a neighborhood of the leaf / orbit .
although all these results have the same flavor , they do look quite different .
moreover , the proofs that one can find in the literature of these linearization results are also very distinct .
so it may come as a surprise that they are actually just special cases of a very general linearization theorem . in order to relate all these linearization theorems , and to understand the significance of the assumptions one can find in their statements
, one needs a language where all these results fit into the same geometric setup .
this language exists and it is a generalization of the usual lie theory from groups to groupoids . we will recall it in the next lecture . after that , we will be in shape to state the general linearization theorem and explain how the results stated before are special instances of it .
in this lecture we provide a quick introduction to lie groupoids and lie algebroids .
we will focus mostly on some examples which have special relevance to us .
a more detailed discussion , along with proofs , can be found in @xcite .
let us start by recalling : a * groupoid * is a small category where all arrows are invertible .
let us spell out this definition .
we have a * set of objects * @xmath9 , and a * set of arrows * @xmath35 . for each arrow
@xmath36 we can associate its source @xmath37 and its target @xmath38 , resulting in two maps @xmath39 .
we also write @xmath40 for an arrow with source @xmath41 and target @xmath42 . for any pair of composable arrows we have a * product * or composition map : @xmath43 in general ,
we will denote by @xmath44 the set of @xmath45 strings of composable arrows : @xmath46 the multiplication satisfies the _ associativity property _ : @xmath47 for each object @xmath48 there is an identity arrow @xmath49 and the _ identity property _ holds : @xmath50 it gives rise to an * identity section * @xmath51 , @xmath52 . for each arrow
@xmath36 there is an inverse arrow @xmath53 , for which the _ inverse property _ holds : @xmath54 this gives rise to the * inverse * map @xmath55 , @xmath56 .
a * morphism of groupoids * is a functor @xmath57 .
this means that we have a map @xmath58 between the sets of arrows and a map @xmath59 between the sets of objects , making the following diagram commute : @xmath60^{{\mathbf{t } } } \ar@<-0.25pc>[d]_{{\mathbf{s } } } \ar[r]^{{\ensuremath{\mathcal{f } } } } & h \ar@<0.25pc>[d]^{{\mathbf{t}}}\ar@<-0.25pc>[d]_{{\mathbf{s } } } \\
m \ar[r]_{f } & n}\ ] ] such that @xmath61 if @xmath62 are composable , and @xmath63 for all @xmath48 .
we are interested in groupoids and morphisms of groupoids in the smooth category : a * lie groupoid * is a groupoid @xmath64 whose spaces of arrows and objects are both manifolds , the structure maps @xmath65 are all smooth maps and such that @xmath66 and @xmath67 are submersions .
a * morphisms of lie groupoids * is a morphism of groupoids for which the underlying map @xmath58 is smooth .
before we give some examples of lie groupoids , let us list a few basic properties .
* the unit map @xmath51 is an embedding and the inverse @xmath55 is a diffeomorphism . *
the source fibers are embedded submanifolds of @xmath35 and * right multiplication by @xmath40 * is a diffeomorphism between @xmath66-fibers @xmath68 * the target fibers are embedded submanifolds of @xmath35 and * left multiplication by @xmath40 * is a diffeomorphism between @xmath67-fibers : @xmath69 * the * isotropy group at @xmath41 * : @xmath70 is a lie group . * the * orbit through @xmath41 * : @xmath71 is a regular immersed ( possibly disconnected ) submanifold . *
the map @xmath72 is a principal @xmath73-bundle .
the connected components of the orbits of a groupoid @xmath64 form a ( possibly singular ) foliation of @xmath9 .
the set of orbits is called the * orbit space * and denoted by @xmath74 .
the quotient topology makes the natural map @xmath75 into an open , continuous map . in general
, there is no smooth structure on @xmath74 compatible with the quotient topology , so @xmath74 is a singular space .
one may think of a groupoid as a kind of atlas for this orbit space .
there are several classes of lie groupoids which will be important for our purposes .
a lie groupoid @xmath64 is called * source @xmath76-connected * if the @xmath77-fibers @xmath78 are @xmath76-connected for every @xmath48 . when @xmath79 we say that @xmath35 is a * @xmath77-connected groupoid * , and when @xmath80 we say that @xmath35 is a * @xmath77-simply connected groupoid*. we call the groupoid * tale * if @xmath81 , which is equivalent to requiring that the source or target map be a local diffeomorphism .
the map @xmath82 is sometimes called the * anchor of the groupoid * and a groupoid is called * proper * if the anchor @xmath82 is a proper map . in particular , when the source map is proper , we call the groupoid * s - proper*. we will see later that proper and s - proper groupoids are , in some sense , analogues of compact lie groups .
perhaps the most elementary example of a lie groupoid is the * unit groupoid * @xmath83 of any manifold @xmath9 : for each object @xmath48 , there is exactly one arrow , namely the identity arrow . more generally , given an open cover @xmath84 of @xmath9 one constructs a * cover groupoid
* @xmath85 : we picture an arrow as @xmath86 if @xmath87 .
the structure maps should be obvious . in both these examples ,
the orbit space coincides with the original manifold @xmath9 and the isotropy groups are all trivial .
these are all examples of proper , tale , lie groupoids . if the open cover is not second countable , then the spaces of arrows and objects are not second countable manifolds . in lie
groupoid theory one sometimes allows manifolds which are not second countable .
however , we will always assume manifolds to be second countable .
another groupoid that one can associate to a manifold is the * pair groupoid * @xmath88 : an ordered pair @xmath89 determines an arrow @xmath90 .
this is an example of * transitive groupoid * , i.e. , a groupoid with only one orbit .
more generally , if @xmath91 is a principal @xmath20-bundle , then @xmath20 acts on the pair groupoid @xmath92 by groupoid automorphisms and the quotient @xmath93 is a transitive groupoid called the * gauge groupoid*. note that for any @xmath48 the isotropy group @xmath73 is isomorphic to @xmath20 and the principal @xmath73-bundle @xmath94 is isomorphic to the original principal bundle .
conversely , it is easy to see that any transitive groupoid @xmath64 is isomorphic to the gauge groupoid of any of the principal @xmath73-bundles @xmath94 .
it is easy to check that that the gauge groupoid associated with a principal @xmath20-bundle @xmath95 is proper if and only if @xmath20 is a compact lie group .
moreover , it is s - proper ( respectively , source @xmath76-connected ) if and only if @xmath96 is compact ( respectively , @xmath76-connected ) . to any surjective submersion @xmath0
one can associate the * submersion groupoid * @xmath97 .
this is a subgroupoid of the pair groupoid @xmath88 , but it fails to be transitive if @xmath98 has more than one point : the orbits are the fibers of @xmath0 , so the orbit space is precisely @xmath98 .
the isotropy groups are all trivial .
the submersion groupoid is always proper and it is s - proper if and only if @xmath1 is a proper map .
let @xmath8 be a foliation of a manifold @xmath9 .
we can associate to it the * fundamental groupoid * @xmath99 , whose arrows correspond to foliated homotopy classes of paths ( relative to the end - points ) . given such an arrow @xmath100 $ ] , the source and target maps are @xmath101)=\gamma(0)$ ] and @xmath102)=\gamma(1)$ ] , while multiplication corresponds to concatenation of paths .
one can show that @xmath103 is indeed a manifold , but it may fail to be hausdorff .
in fact , this groupoid is hausdorff precisely when @xmath8 has no vanishing cycles . in lie groupoid theory ,
one often allows the total space of a groupoid to be non - hausdorff , while @xmath9 and the source / target fibers are always assumed to be hausdorff . on the other hand
, we will always assume manifolds to be second countable .
another groupoid one can associate to a manifold is the * holonomy groupoid * @xmath104 , whose arrows correspond to holonomy classes of paths .
again , one can show that @xmath105 is a manifold , but it may fail to be hausdorff . in general ,
the fundamental groupoid and the holonomy groupoid are distinct , but there is an obvious groupoid morphism @xmath106 , which to a homotopy class of a path associates the holonomy class of the path ( recall that the holonomy only depends on the homotopy class of the path ) .
this map is a local diffeomorphism .
it should be clear that the leaves of these groupoids coincide with the leaves of @xmath8 and that the isotropy groups of @xmath103 ( respectively , @xmath105 ) coincide with the fundamental groups ( respectively , holonomy groups ) of the leaves .
it is easy to check also that @xmath103 is always @xmath77-connected .
one can show that @xmath103 ( respectively , @xmath105 ) is s - proper if and only if the leaves of @xmath8 are compact and have finite fundamental group ( respectively , finite holonomy group ) .
in general , it is not so easy to give a characterization in terms of @xmath8 of when these groupoids are proper
. let @xmath107 be a smooth action of a lie group @xmath20 on a manifold @xmath9 .
the associated * action groupoid * @xmath108 has arrows the pairs @xmath109 , source / target maps given by @xmath110 and @xmath111 , and composition : @xmath112 the isotropy groups of this action are the stabilizers @xmath27 and the orbits coincide with the orbits of the action .
the action groupoid is proper ( respectively , s - proper ) precisely when the action is proper ( respectively , @xmath20 is compact ) .
moreover , this groupoid is source @xmath76-connected if and only if @xmath20 is @xmath76-connected .
lie groupoids , just like lie groups , have associated infinitesimal objects known as _ lie algebroids _ :
a * lie algebroid * is a vector bundle @xmath113 together with a lie bracket @xmath114_a:\gamma(a)\times\gamma(a)\to \gamma(a)$ ] on the space of sections and a bundle map @xmath115 , such that the following leibniz identity holds : @xmath116_a = f[{\alpha},{\beta}]_a+\rho_a(x)(f ) y,\ ] ] for all @xmath117 and @xmath118 .
given a lie groupd @xmath64 the associated lie algebroid is obtained as follows .
one lets @xmath119 be the vector bundle whose fibers consists of the tangent spaces to the s - fibers along the identity section
. then sections of @xmath120 can be identified with right - invariant vector fields on the lie groupoid , so the usual lie bracket of vector fields induces a lie bracket on the sections .
the anchor is obtained by restricting the differential of the target , i.e. , @xmath121 .
there is a lie theory for lie groupoids / algebroids analogous to the usual lie theory for lie groups / algebras .
there is however one big difference : lie s third theorem fails and there are examples of lie algebroids which are not associated with a lie groupoid ( @xcite ) .
we shall not give any more details about this correspondence since we will be working almost exclusively at the level of lie groupoids .
we refer the reader to @xcite for a detailed discussion of lie theory in the context of lie groupoids and algebroids .
for a lie groupoid @xmath64 a submanifold @xmath122 is called * saturated * if it is a union of orbits . our aim now is to state the linearization theorem which , under appropriate assumptions , gives a normal form for the lie groupoid in a neighborhood of a saturated submanifold .
first we will describe this local normal form , which depends on some standard constructions in lie groupoid theory .
let @xmath64 be a lie groupoid . the * tangent lie groupoid * @xmath123 is obtained by applying the tangent functor : hence , the spaces of arrows and objects are the tangent bundles to @xmath35 and to @xmath9 , the source and target maps are the differentials @xmath124 , the multiplication is the differential @xmath125 , etc .
assume now that @xmath31 is a saturated submanifold .
an important special case to keep in mind is when @xmath126 consists of a single orbit of @xmath35
. then we can restrict the groupoid @xmath35 to @xmath126 : @xmath127 obtaining a lie subgroupoid @xmath128 of @xmath64 .
if we apply the tangent functor , we obtain a lie subgroupoid @xmath129 of @xmath123 .
there is a short exact sequence of lie groupoids : @xmath130 & tg_s \ar@<0.25pc>[d ] \ar@<-0.25pc>[d ] \ar[r ] & tg \ar@<0.25pc>[d ] \ar@<-0.25pc>[d ] \ar[r ] & \nu(g_s)\ar@<0.25pc>[d ] \ar@<-0.25pc>[d ] \ar[r ] & 1\\ 0\ar[r ] & ts \ar[r ] & tm \ar[r ] & \nu(s)\ar[r ] & 0}\ ] ] there is an alternative description of the groupoid @xmath131 which sheds some light on its nature . for each arrow @xmath132 in @xmath133 one can define the linear transformation : @xmath134 \mapsto [ { \mathrm d}_g{\mathbf{t}}(\tilde v)],\ ] ] where @xmath135 is such that @xmath136 .
one checks that this map is independent of the choice of lifting @xmath137 .
moreover , for any identity arrow one has @xmath138id@xmath139 and for any pair of composable arrows @xmath140 one finds : @xmath141 this means that @xmath128 acts linearly on the normal bundle @xmath142 , and one can form the * action groupoid * : @xmath143 the space of arrows of this groupoid is the fiber product @xmath144 , with source and target maps : @xmath145 and @xmath146 .
the product of two composable arrows @xmath147 and @xmath148 is then given by : @xmath149 one then checks that : the map @xmath150 , @xmath151)$ ] , is an isomorphism of lie groupoids .
finally , let us observe that the restricted groupoid @xmath128 sits ( as the zero section ) inside the lie groupoid @xmath131 as a lie subgroupoid .
this justifies introducing the following definition : for a saturated submanifold @xmath126 of a lie groupoid @xmath64 the local * linear model * around @xmath126 is the groupoid @xmath131 . for the submersion groupoid @xmath152 associated with a submersion @xmath0 ,
a fiber @xmath153 is a saturated submanifold ( actually , an orbit ) and we find that @xmath154 is the pair groupoid .
notice that the normal bundle @xmath155 is naturally isomorphic to the trivial bundle @xmath156 .
it follows that the local linear model is the groupoid : @xmath157 which is the direct product of the pair groupoid @xmath158 and the identity groupoid @xmath159 .
more importantly , we can view this groupoid as the submersion groupoid of the projection @xmath160 .
let @xmath8 be a foliation of @xmath9 and let @xmath161 be a leaf .
the normal bundle @xmath162 has a natural flat @xmath8-connection , which can be described as follows .
given a vector field @xmath163 let us write @xmath164 for the corresponding section of the normal bundle .
then , if @xmath165 is a foliated vector field , one sets : @xmath166}.\ ] ] one checks that this definition is independent of the choice of representative , and defines a connection @xmath167 , called the _
bott connection_. the jacobi identity shows that this connection is flat .
given a path @xmath168 in some leaf @xmath169 of @xmath8 , parallel transport along @xmath170 defines a linear map @xmath171 since the connection is flat , it is clear that this linear map only depends on the homotopy class @xmath100 $ ] .
one obtains a linear action of the fundamental groupoid @xmath103 on @xmath162 .
the restriction of @xmath103 to the leaf @xmath169 is the fundamental groupoid of the leaf @xmath169 and we obtain the linear model for @xmath103 along the leaf @xmath169 as the action groupoid : @xmath172 the fundamental groupoid @xmath173 is isomorphic to the gauge goupoid of the universal covering space @xmath174 , viewed as a principal @xmath175-bundle .
moreover , @xmath176 is isomorphic to the associated bundle @xmath177 .
it follows that the linear model coincides with the fundamental groupoid of the linear foliation of @xmath178 . the linear map @xmath179 only depends on the holonomy class of @xmath180 , since this maps coincides with the linearization of the holonomy action along @xmath180 .
for this reason , there is a similar description of the linear model of the holonomy groupoid @xmath105 along the leaf @xmath169 as an action groupoid : @xmath181 also , the holonomy groupoid is isomorphic to the gauge goupoid of the holonomy cover @xmath182 , viewed as a principal @xmath183-bundle and we can also describe the normal bundle @xmath176 as the associated bundle @xmath184 .
the underlying foliation of this linear model is still the linear foliation of @xmath176 .
however , the linear model now depends on the germ of the foliation around @xmath169 , i.e. , on the non - linear holonomy . unlike the case of the fundamental groupoid ,
the knowledge of the bott connection is not enough to build this linear model .
let @xmath20 be a lie group that acts on a manifold @xmath9 and let @xmath27 be the isotropy group of some @xmath48 . for each @xmath185 , the map @xmath186 , @xmath187
, fixes @xmath41 and maps the orbit @xmath188 to itself .
hence , @xmath189 induces a linear action of @xmath27 on the normal space @xmath190 , called the _
normal isotropy representation_. using this representation , it is not hard to check that we have a vector bundle isomorphism : @xmath191^{\simeq}\ar[dr ] & & k\times_{k_x}\nu_x({\ensuremath{\mathcal{o}}}_x)\ar[dl]\\ & { \ensuremath{\mathcal{o}}}_x}\ ] ] where the action @xmath192 is given by @xmath193 .
moreover , one has an action of @xmath20 on @xmath194 , which under this isomorphism corresponds to the action : @xmath195=[(kk',v)].\ ] ] now consider the action lie groupoid @xmath196 .
one checks that the local linear model around the orbit @xmath188 is just the action lie groupoid @xmath197 , which under the isomorphism above corresponds to the action groupoid : @xmath198 as we have already mentioned above , the linearization theorem states that , under appropriate conditions , the groupoid is locally isomorphic around a saturated submanifold to its local model . in order to make precise the expression `` locally isomorphic '' we introduce the following definition : let @xmath64 be a lie groupoid and @xmath31 a saturated submanifold .
a * groupoid neighborhood * of @xmath128 is a pair of open sets @xmath199 and @xmath200 such that @xmath201 is a subgroupoid of @xmath64 . a groupoid neighborhood @xmath201 is said to be * full * if @xmath202 .
our first version of the linearization theorem reads as follows : [ thm : weak : linear ] let @xmath64 be a lie groupoid with a 2-metric @xmath203 .
then @xmath35 is * weakly linearizable * around any saturated submanifold @xmath31 : there are groupoid neighborhoods @xmath201 of @xmath128 in @xmath64 and @xmath204 of @xmath205 in the local model @xmath131 , and an isomorphism of lie groupoids : @xmath206 which is the identity on @xmath133 .
a 2-metric is a special riemannian metric in the space of composable arrows @xmath207 .
we will discuss them in detail in the next lecture .
for now we remark that for a _ proper _ groupoid 2-metrics exist and , moreover , every groupoid neighborhood contains a full groupoid neighborhood . hence : let @xmath64 be a proper lie groupoid .
then @xmath35 is * linearizable * around any saturated submanifold @xmath31 : there exist open neighborhoods @xmath33 and @xmath208 and an isomorphism of lie groupoids @xmath209 which is the identity on @xmath133 .
this corollary does not yet yield the various linearization results stated in lecture 1 .
the reason is that the assumptions do not guarantee the existence of a saturated neighborhood @xmath33 .
this can be realized for s - proper groupoids , since for such groupoids every neighborhood @xmath210 of a saturated embedded submanifold @xmath126 , contains a saturated neighborhood of @xmath126 .
let @xmath64 be an s - proper lie groupoid .
then @xmath35 is * invariantly linearizable * around any saturated submanifold @xmath31 : there exist saturated open neighborhoods @xmath33 and @xmath208 and an isomorphism of lie groupoids @xmath209 which is the identity on @xmath133 . for a proper submersion @xmath0 the associated submersion groupoid @xmath97 is s - proper . using the description of the local model that we gave before
, it follows that for any fiber @xmath211 there is a saturated open neighborhood @xmath212 where the submersion if locally isomorphic to the trivial submersion @xmath213 , for some open neighborhood @xmath214 .
hence , we recover the classical ehresmann s theorem .
let @xmath8 be a foliation of @xmath9 whose leaves are compact with finite holonomy .
then the holonomy groupoid @xmath104 is s - proper . using the description of the local model given before
, it follows that for any leaf @xmath169 there is a saturated neighborhood @xmath215 where the canonical foliation is isomorphic to the linear foliation of @xmath216 , for some @xmath217-invariant , neighborhood @xmath218 .
hence , we recover the local reeb stability theorem .
let @xmath219 be an action of a _ compact _ lie group .
then the the action groupoid is @xmath77-proper and we obtain invariant linearization . from the description of the local model , it follows that for any orbit @xmath188 there is a saturated open neighborhood @xmath220 and a @xmath20-equivariant isomorphism @xmath221 , where @xmath222 is a @xmath27-invariant neighborhood .
hence , we recover the slice theorem for actions of compact groups . for a general proper action ,
the results above only give weak linearization , which does not allow to deduce the slice theorem .
however , due to the particular structure of the action groupoid , every orbit has a saturated neighborhood and one has a uniform bound for the injectivity radius of the 2-metric .
this gives invariant linearization and leads to the slice theorem for _ any _ proper action .
let @xmath30 be a poisson manifold .
the cotangent bundle @xmath223 has a natural lie algebroid structure with anchor @xmath224 given by contraction by @xmath1 and lie bracket on sections ( i.e. , 1-forms ) : @xmath225:={\mathcal{l}}_{\rho({\alpha})}{\beta}-{\mathcal{l}}_{\rho({\beta})}{\alpha}-{\mathrm d}\pi({\alpha},{\beta}).\ ] ] in general , this lie algebroid fails to be integrable .
however , under the assumptions of the local normal form theorem stated in lecture 1 , this groupoid is integrable , and then its source 1-connected integration is an s - proper lie groupoid whose orbits are the symplectic leaves .
this groupoid can then be linearized around a symplectic leaf , but this linearization _ does not _ yet yield the local canonical form for the poisson structure .
it turns out that the source 1-connected integration is a symplectic lie groupoid , i.e. , there is a symplectic structure on its space of arrows which is compatible with multiplication .
one can apply a moser type trick to further bring the symplectic structure on the local normal form to a canonical form , which then yields the canonical form of the poisson structure .
the details of this approach , which differ from the original proof of the canonical form due to crainic and marcut , can be found in @xcite .
let us recall that a submersion @xmath226 is called a * riemannian submersion * if the fibers are equidistant .
the base @xmath98 gets an induced metric @xmath227 for which the linear maps @xmath228 , @xmath48 , are all isometries . more generally , a ( possibly singular ) foliation in a riemannian manifold @xmath9 is called a * riemannian foliation * if the leaves are equidistant .
this is equivalent to the following property : any geodesic which is perpendicular to one leaf at some point stays perpendicular to all leaves that it intersects .
a simple proof of ereshman s theorem can be obtained by choosing a metric on the total space of the submersion @xmath0 , that makes it into a riemannian submersion .
a partition of unit argument shows that this is always possible .
then the linearization map is just the exponential map of the normal bundle of a fiber , which maps onto a saturated neighborhood of the fiber , provided the submersion is proper .
we will see that a proof similar in spirit also works for the general linearization theorem ( theorem [ thm : weak : linear ] ) .
let @xmath64 be a lie groupoid . like any category
, @xmath35 has a * simplicial model * : @xmath229\ar@<0.30pc>[r ] \ar[r ] \ar@<-0.30pc>[r]\ar@<-0.60pc>[r ] & { g^{(n ) } } \ar@<0.45pc>[r]\ar@<0.15pc>[r]\ar@<-0.15pc>[r]\ar@<-0.45pc>[r ] & \cdots \ar@<0.45pc>[r]\ar@<0.15pc>[r]\ar@<-0.15pc>[r]\ar@<-0.45pc>[r]&{g^{(2 ) } } \ar@<0.30pc>[r ] \ar[r ] \ar@<-0.30pc>[r ] & { g^{(1 ) } } \ar@<0.2pc>[r]\ar@<-0.2pc>[r ] & { g^{(0)}}\ [ \ar[r ] & { g^{(0)}}/{g^{(1)}}]}\ ] ] where , for each @xmath230 , the * face maps * @xmath231 , @xmath232 and the * degeneracy maps * @xmath233 , @xmath234 , are defined as follows : for a @xmath45-string of composable arrows the @xmath235-th face map associates the @xmath236-string of composable arrows obtained by omitting the @xmath235-object : @xmath237_{g_1 } \cdot \ar@{ .. }[r]&\cdot & \ar@/_/[l]_{g_n } \cdot } \right)= \xymatrix{\cdot & \ar@/_/[l]_{g_1 } \cdot \ar@{ .. }[r ] & & \ar@/_/[l]_{g_ig_{i+1 } } \cdot \ar@{ .. }[r]&\cdot & \ar@/_/[l]_{g_n } \cdot } \ ] ] while the @xmath235-th degeneracy map inserts into a @xmath45-string of composable arrows an identity at the @xmath235-th entry : @xmath238_{g_1 } \cdot \ar@{ .. }[r]&\cdot & \ar@/_/[l]_{g_n } \cdot } \right)= \xymatrix{\cdot & \ar@/_/[l]_{g_1 } \cdot \ar@{ .. }[r]&\cdot\ar@(ur , ul)[]_{1_{x_i } } \ar@{ .. }[r]&\cdot & \ar@/_/[l]_{g_n } \cdot } \ ] ] for a lie groupoid there is , additionally , a natural action of @xmath239 on @xmath44 : for a string of n - composable arrows @xmath240 choose @xmath241 arrows @xmath242 , all with the same source , so that : @xmath243_{h_0\ \ } ^{h_1\ \ \ } \ar@/_/[dll]^{h_2 } \ar@/_/[dl]^{\ \
\ \cdots } \ar@/_/[dr]^{h_{n-1 } } \ar@/^/[drr]^{h_n}\\ \cdot & \ar@/_/[l]^{g_1 } \cdot & \ar@/_/[l]^{g_2}\cdot \ar@{ .. }[rr ] & & \cdot & \ar@/_/[l]^{g_n } \cdot}\ ] ] then the @xmath239-action on the arrows @xmath242 by permutation gives a well defined @xmath239-action on @xmath44 .
notice that this action permutes the face maps @xmath244 , since there are maps @xmath245 such that : @xmath246 actually , it is enough in this definition to ask that one of the face maps is a riemannian submersion : the assumption that the action of @xmath239 is by isometries implies that if one face map is a riemannian submersion then all the face maps are riemannian submersions . for any @xmath249 ,
the metrics induced on @xmath250 by the different face maps @xmath231 coincide , giving a well defined metric @xmath251 , which is a @xmath236-metric .
obviously , one can repeat this process , so that a @xmath45-metric on @xmath44 determines a @xmath76-metric on @xmath252 for all @xmath253 .
when @xmath254 we adopt the convention that @xmath255 is a metric on @xmath256 which makes the orbit space a riemannian foliation and is invariant under the action of @xmath35 on the normal space to the orbits , i.e. , such that each arrow @xmath257 acts by isometries on @xmath258 .
such 0-metrics were studied by pflaum , posthuma and tang in @xcite , in the case of proper groupoids .
one can think of such metrics as determining a metric on the ( possibly singular ) orbit space @xmath74 .
indeed , it is proved in @xcite that the distance on the orbit space determined by @xmath255 enjoys some nice properties .
a 1-metric is just a metric @xmath259 on the space of arrows @xmath260 for which the source and target maps are riemannian submersions and inversion is an isometry .
these metrics were studied by gallego _
_ in @xcite .
a 1-metric induces a 0-metric @xmath255 on @xmath256 , for which the orbit foliation is riemannian .
a @xmath261-metric is a metric @xmath203 in the space of composable arrows @xmath207 , which is invariant under the @xmath262-action generated by the involution @xmath263 and the 3-cycle @xmath264 , and for which multiplication is a riemannian submersion .
hence , a 2-metric on @xmath207 induces a 1-metric on @xmath265 .
these metrics were introduced in @xcite .
notice that the first 3 stages of the nerve @xmath266^{\pi_1 } \ar[r]|m \ar@<-0.5pc>[r]_{\pi_2 } & g \ar@<0.25pc>[r]^s \ar@<-0.25pc>[r]_t & m}\ ] ] completely determine the remaining @xmath44 , for @xmath267 .
hence , one should expect that @xmath45-metrics , for @xmath267 , are determined by their @xmath261-metrics .
in fact , one has the following properties , whose proof can be found in @xcite : * there is at most one 3-metric inducing a given 2-metric and every 3-metric has a unique extension to an @xmath45-metric for every @xmath267 .
* there are examples of groupoids which admit an @xmath45-metric , but do not admit a @xmath268-metric , for @xmath269 .
* uniqueness fails in low degrees : one can have , e.g. , two different 2-metrics on @xmath207 inducing the same 1-metric on @xmath265 . the geometric realization of the nerve of a groupoid @xmath64 is usually denoted by @xmath270 , can be seen as the classifying space of principal @xmath35-bundles ( see @xcite ) .
two morita equivalent groupoids @xmath271 and @xmath272 ( see @xcite or @xcite ) , give rise to homotopy equivalent spaces @xmath273 and @xmath274 . an alternative point of view , is to think of @xmath270 as a geometric stack with atlas @xmath64 , and two atlas represent the same stack if they are morita equivalent groupoids .
the following result shows that one may think of a @xmath45-metric as a metric in @xmath270 : in fact , it is shown in @xcite that it is possible to `` transport '' a @xmath45-metric via a morita equivalence .
this constructions depends on some choices , but the _ transversal component _ of the @xmath45-metric is preserved .
we refer to those references for a proof and more details . since a @xmath45-metric determines a @xmath76-metric , for all @xmath253 , a necessary condition for a groupoid @xmath64 to admit a @xmath45-metric is that the orbit foliation can be made riemannian .
this places already some strong restrictions on the class of groupoids admitting a @xmath45-metric .
so one may wonder when can one find such metrics .
one important result in this direction is that any proper groupoid admits such a metric : the proof uses the following trick , called in @xcite the * gauge trick*. for each @xmath45 , consider the manifold @xmath278}\subset g^n$ ] of @xmath45-tuples of arrows with the same source , and the map @xmath279}\to{g^{(n)}},\quad ( h_0,h_1\dots , h_{n})\longmapsto ( h_0h_1^{-1},h_1h_2^{-1},\dots , h_{n-1}h_{n}^{-1}).\ ] ] the fibers of @xmath280 coincide with the orbits of the right - multiplication action , @xmath281}{\curvearrowleft}g\\ ( h_0,\dots , h_{n})\cdot k = ( h_0k,\dots , h_{n}k ) .
\end{matrix } \qquad \qquad \begin{matrix } \xymatrix@r=10pt{\bullet & & \\ \bullet & \bullet \ar[lu]_{h_0 } \ar[l]\ar[ld]^{h_{n } } & \bullet \ar[l]^k \\ \bullet } \end{matrix}\ ] ] and this action is free and proper , hence defining a principal @xmath35-bundle .
the strategy is to define a metric on @xmath282}$ ] in a such way that @xmath280 becomes a riemannian submersion , and that the resulting metric on @xmath44 is a @xmath45-metric .
the group @xmath239 acts on the manifold @xmath282}$ ] by permuting its coordinates , and this action covers the action @xmath283 , so the map @xmath280 is @xmath239 equivariant . on the other hand , there are @xmath241 left groupoid actions @xmath284}$ ] , each consisting in left multiplication on a given coordinate .
@xmath285^k & \\ & \bullet & \bullet \ar[lu]_{h_0 } \ar[l ] \ar[ld]^{h_{n } } \\ & \bullet } \end{matrix } \qquad \begin{matrix } \xymatrix@r=10pt { & \bullet & \\\bullet & \bullet \ar[l]|k & \bullet \ar[lu]_{h_0 } \ar[l ] \ar[ld]^{h_{n } } \\ & \bullet } \end{matrix } \qquad\dots\qquad \begin{matrix }
\xymatrix@r=10pt { & \bullet & \\ & \bullet & \bullet \ar[lu]_{h_0 } \ar[l ] \ar[ld]^{h_{n } } \\\bullet & \bullet \ar[l]_k & } \end{matrix}\ ] ] these left actions commute with the above right action and cover @xmath241-principal actions @xmath286 , with projection the face maps @xmath231 .
now for a proper groupoid one can use averaging to construct a metric on @xmath35 which is invariant under the left action of @xmath35 on itself by left translations .
the product metric on @xmath282}$ ] is invariant both under the @xmath241 left @xmath35-actions above and the @xmath239-action . in general
, it will not be invariant under the right @xmath35-action @xmath284}\to { g^{(n)}}$ ] , but we can average it to obtain a new metric which is invariant under all actions .
it follows that the resulting metric on @xmath44 is an @xmath45-metric .
the maps @xmath287}\to{g^{(n)}}$ ] , @xmath288 that appear in this proof form the simplicial model for the universal principal @xmath35-bundle @xmath289 .
this bundle plays a key role in many different constructions associated with the groupoid ( see @xcite ) .
the unit groupoid @xmath83 obviously admits @xmath45-metrics .
the pair groupoid @xmath88 and , more generally , the submersion groupoid @xmath97 associated with a submersion @xmath290 , are proper so also admit @xmath45-metrics .
for example , if @xmath291 is a metric on @xmath9 which makes @xmath1 a riemannian submersion , then one can take @xmath292 , @xmath293 , @xmath294 , etc .
let @xmath8 be a foliation of @xmath9 .
then @xmath105 and @xmath103 admit a @xmath45-metric if and only if @xmath8 can be made into a riemannian foliation .
we already know that if these groupoids admit a @xmath45-metric , then the underlying foliation , i.e. @xmath8 , must be riemannian .
conversely , if @xmath8 is riemannian then @xmath105 is a proper groupoid ( see , e.g. , @xcite ) , so it carries a @xmath45-metric .
since @xmath103 is a covering of @xmath105 , it also admits a @xmath45-metric . not that , in general , @xmath103 does not need to be proper ( e.g. , if the fundamental group of some leaf is not finite ) .
any lie group admits a @xmath45-metric .
more generally , the action lie groupoid associated with any isometric action of a lie group on a riemannian manifold admits a @xmath45-metric .
this can be shown using a gauge trick , similar to the one used in the proof of existence of a @xmath45-metric for a proper groupoid sketched above ( see @xcite ) .
note that an isometric action does not need not be proper ( e.g. , if some isotropy group is not compact ) .
[ thm : main:2 ] let @xmath64 be a lie groupoid endowed with a 2-metric @xmath295 , and let @xmath31 be a saturated embedded submanifold . then the exponential map defines a weak linearization of @xmath35 around @xmath126 .
one choses a neighborhood @xmath208 where the exponential map of @xmath255 is a diffeomorphism onto its image , and set @xmath296 one shows that @xmath204 is a groupoid neighborhood of @xmath128 in the linear model @xmath131 and that we have a commutative diagram : @xmath297^{\exp_{{\eta^{(2)}}}}\ar@<0.5pc>[d ] \ar[d ] \ar@<-0.5pc>[d ] & & { g^{(2 ) } } \ar@<0.5pc>[d ] \ar[d ] \ar@<-0.5pc>[d ] \\ \widetilde{v } \ar[rr]^{\exp_{{\eta^{(1 ) } } } } \ar@<0.25pc>[d ] \ar@<-0.25pc>[d ] & & g\ar@<0.25pc>[d ] \ar@<-0.25pc>[d]\\ v \ar[rr]^{\exp_{{\eta^{(0 ) } } } } & & m}\ ] ] it follows that the exponential maps of @xmath259 and @xmath255 give the desired weak linearization , i.e. , a groupoid isomorphism @xmath298 i would specially like to thank matias del hoyo , for much of the work reported in these notes rests upon our ongoing collaboration @xcite .
he and my two phd students , daan michiels and joel villatoro , made several very usual comments on a preliminary version of these notes , that helped improving them .
i would also like to thank marius crainic , ioan marcut , david martinez - torres and ivan struchiner , for many fruitful discussions which helped shape my view on the linearization problem .
finally , thanks go also to the organizers of the xxxiii wgmp , for the invitation to deliver these lectures and for a wonderful stay in bialowieza . | in these lectures i discuss the linearization theorem for lie groupoids , and its relation to the various classical linearization theorems for submersions , foliations and group actions .
in particular , i explain in some detail the recent metric approach to this problem proposed in @xcite . |
the mott - hubbard model of interacting bosons on a lattice has been used to describe superfluid mott - insulator transitions in a variety of systems , e.g. , josephson arrays and granular superconductors @xcite .
the recent suggestion @xcite to experimentally observe this transition in a system of cold bosonic atoms in an optical lattice and its successful experimental demonstration @xcite has rekindled the interest in the mott - insulator transition and triggered a great deal of theoretical @xcite and experimental @xcite activity . the possibility to directly manipulate and test the many - body behavior of a system of
trapped bosonic atoms in an optical lattice @xcite is very attractive .
possible applications include the use of a mott state of bosonic atoms in an optical lattice as a starting point to create controlled multiparticle entanglement as an essential ingredient for quantum computation @xcite the mott - insulator quantum phase transition is driven by the interplay of the repulsive interaction of bosons on the same lattice site and the kinetic energy .
hence the ratio of the onsite energy and the bandwidth forms the key parameter in the system . in optical lattices , this parameter can be easily controlled and varied by several orders of magnitude , enabling detailed studies of the quantum phase transition .
probing the system by taking absorption pictures to image the expansion patterns after a reasonable expansion time yields information about the momentum distribution of the state .
this procedure was used to experimentally confirm the mott transition in an optical lattice @xcite . the essential physics of cold bosonic atoms in an optical lattice is captured by a bosonic mott - hubbard model describing the competition between hopping and on - site interaction .
a number of approximation schemes have been used to study this model analytically @xcite as well as numerically , using approaches like the gutzwiller mean - field ansatz @xcite , density - matrix renormalization group ( dmrg ) @xcite , exact diagonalization ( ed)@xcite and quantum monte carlo ( qmc ) @xcite . in this article , we study the short - range correlations , not included by the gutzwiller ansatz , by using perturbation theory . the main purpose is to find corrections to the short - range behavior of the one - particle density matrix , which is directly relevant to experimentally observed expansion patterns .
these patterns are important for determining the location of the insulator - superfluid transition .
we note that in the insulating state our perturbative approach is identical to the one used in @xcite ( see also @xcite ) , although there the goal was different , viz . , studying corrections to the phase diagram .
the remainder of the article is organized as follows : in section [ modsec ] , we will introduce the model and its mean - field solution . the general perturbative approach is briefly outlined in section [ secpa ] , while details may be found in the appendix .
numerical results are presented and discussed in section [ secnr ] , first for local observables ( [ secslso ] ) and then for the density matrix ( [ secrho ] ) .
implications for expansion patterns both for bulk systems and a harmonic confining potential are discussed in section [ harmotrap ] .
the cold bosonic gas in the optical lattice can be described by a mott - hubbard model @xcite @xmath0 here , @xmath1 is the total number of lattice sites , @xmath2 ( @xmath3 ) creates ( annihilates ) a boson on site @xmath4 , @xmath5 , @xmath6 is the on - site repulsion describing the interaction between bosons on the same lattice site , and @xmath7 denotes the chemical potential .
the kinetic term includes only hopping between nearest - neighbor sites , this is denoted by the summation index @xmath8 ; @xmath9 is the hopping matrix element that we will assume to be lattice - site independent .
finally , @xmath10 describes an external on - site potential that is commonly present in experiments .
the gutzwiller ( gw ) approach is based on an ansatz for the many - body ground state that factorizes into single lattice - site wavefunctions @xmath11 the gutzwiller wavefunction represents the ground state of the following mean - field version of the mott - hubbard hamiltonian , eq .
( [ motthubb ] ) : @xmath12 here @xmath13 is the mean - field potential on the @xmath4-th lattice site , which is self - consistently defined as the expectation value of @xmath3 in terms of the gutzwiller wavefunction , @xmath14 @xcite .
using the gutzwiller ansatz to obtain an approximate variational solution for the mott - hubbard hamiltonian ( [ motthubb ] ) corresponds , however , to restricting the hilbert space to the subset of product states .
consequently , even in higher dimensions , this ansatz fails to describe the correct behavior of short - range correlations between different lattice sites , which are important for experimentally measurable observables , such as expansion patterns ( momentum distributions ) . nevertheless , in the thermodynamic limit and higher dimensions , the gutzwiller wavefunction provides a good approximation in the limits of @xmath15 and @xmath16 ( i.e. , deep in the mott insulator ( mi ) and superfluid ( sf ) phases ) .
to get a satisfactory description of the short - range correlations we will now derive perturbative corrections to the gutzwiller mean - field result .
our aim is to start from the gutzwiller approximation and improve it by perturbatively including the short - range correlations between lattice sites .
we re - express the mott - hubbard hamiltonian ( [ motthubb ] ) by adding the appropriate perturbation to the mean - field hamiltonian , eq .
( [ mf ] ) : @xmath17 with @xmath18 as the mean - field hamiltonian represents a sum of single lattice - site hamiltonians , the excited states @xmath19 and the excitation spectrum @xmath20 can be obtained numerically for each lattice site @xmath4 separately . hence we can write the excitations of @xmath21 as product states of single lattice - site excitations , @xmath22 and the excitation spectrum as a sum over the single lattice - site excitation energies , @xmath23 where @xmath24 describes the set of quantum numbers characterizing the given many - body energy eigenstate . having obtained the mean - field solution from the gutzwiller ansatz , we can now proceed to improve our wavefunction performing rayleigh - schrdinger perturbation theory @xcite in @xmath25 : @xmath26 where @xmath27 and @xmath28 are the @xmath29-th order corrections to the wavefunction and grand - canonical energy @xmath30 , respectively . knowing the excited wavefunctions , eq .
( [ es ] ) , and the excitation spectrum , eq .
( [ esp ] ) , perturbative corrections to observables can be calculated explicitly .
the resulting expressions up to second order in @xmath25 , which we use in the following , are derived and illustrated diagrammatically in appendix [ appendixa ] .
we start by computing the gutzwiller wavefunction numerically using a conjugate - gradient descent method .
propagation steps of the time - dependent gutzwiller equations @xcite in imaginary time are performed to test whether the minimum found by the conjugate - gradient descent method is indeed the ground state .
afterwards , we calculate eigenenergies and -vectors of each site , with respect to the mean - field hamiltonian . as explained above , this forms the basis for perturbation theory , which we use to calculate the corrections to the observables up to second order .
the results obtained from perturbation theory show important modifications to single - site observables as well as to the correlation function .
these observables are composed of operators acting only on single lattice sites .
hence , they describe local properties and will be less sensitive to the correlations between lattice sites .
thus , values for single - site observables obtained from the gutzwiller wavefunction already provide a good approximation in most cases ( unless fluctuations are concerned ) .
( diamonds ) , @xmath31 ( triangles ) , and @xmath32 ( squares ) dimensions .
the dashed - dotted lines in fig .
[ slso]a - c are the gutzwiller results .
( a ) number fluctuations @xmath33 calculated for a commensurate filling of one boson per lattice site .
the dashed line shows the result from the exact diagonalization for 7 lattice sites and @xmath34 bosons .
( b ) order parameter @xmath35 and ( c ) compressibility @xmath36 both computed at a fixed chemical potential @xmath37 .
[ slso],width=302 ] the leading corrections to the mean values of single lattice - site observables ( slso ) are of second order in @xmath25 .
the results for the order parameter @xmath38 , the compressibility @xmath36 and the number fluctuations @xmath39 are shown in fig .
[ slso ] , where the ration @xmath40 has been scaled by the dimension , to keep the same mean - field transition point .
the solid lines in fig .
[ slso ] show the results from perturbation theory , as compared to the gutzwiller result ( shown as the dashed - dotted line ) .
all three quantities show a vanishing perturbative correction both for small and large @xmath40 , as the gutzwiller wavefunction becomes a good approximation in these regimes ( for lattice dimensions @xmath41 )
. as expected , deviations from the mean - field picture are strongest near the mi - sf transition , where higher - order corrections will become more and more important .
the order parameter @xmath42 shown in fig .
[ slso]b gets suppressed in the sf .
perturbative corrections to the gutzwiller result are particularly large in 1d ( where the order parameter vanishes in reality ) , but get smaller with increasing dimension .
this is not surprising , as gutzwiller is a mean - field approach and hence a better approximation for higher - dimensional systems .
the critical value @xmath43 is not modified within the present perturbative approach .
figure [ slso]c shows the results for the compressibility @xmath44 the results of perturbation theory ( pt ) show a decrease of the compressibility , pointing to an increasing stiffness of the sf phase induced by the short - range interaction . finally , we computed the local particle number fluctuations @xmath45 .
exact diagonalization calculations have shown that the number fluctuations are changing smoothly at the mi - sf transition @xcite , whereas the gutzwiller result predicts vanishing fluctuations in the mi phase , @xmath46 ( see dashed - dotted line in fig . [
our perturbative results reproduce the non - vanishing part of @xmath39 in the mi regime and agree well with our result obtained from exact diagonalization of a one - dimensional system of @xmath47 lattice sites .
significant deviations from the exact 1d result are seen in the mi - sf transition regions , starting from the mean - field critical value , @xmath48 , up to values of the order of the critical values of @xmath49 usually obtained from dmrg @xcite and qmc @xcite calculations .
nevertheless , based on the good agreement of our perturbative result with the exact diagonalization for a large region in the mi , we conclude that the number fluctuations are mainly produced by next - neighbor particle - hole fluctuations included in perturbation theory .
the single - particle density matrix @xmath50 is of particular importance as it describes the correlation between the different lattice sites .
the correlation function @xmath51 shows off - diagonal long - range order in the sf state ( in dimensions @xmath41 ) , in contrast to the mi phase where @xmath51 decays exponentially .
the experimental observation @xcite of the mi transition relies on the different behavior of the density matrix in the mi and sf regimes , which can be visualized by taking absorption pictures of the freely expanding atomic cloud . assuming that the expansion time is long enough and that the gas is dilute enough ( such that atom - atom interactions can be neglected during the expansion ) , the shape of the cloud reflects the initial momentum distribution @xmath52 , which is directly given by the fourier - transform of the density matrix @xmath51 : @xmath53 as a function of site distance @xmath54 .
results for a homogeneous lattice of 7 sites with @xmath34 bosons have been obtained from exact diagonalization ( ed ) , from second - order perturbation theory ( pt ) , and from the gutzwiller mean - field ansatz ( gw ) .
( a ) @xmath55 ( mi regime ) , ( b ) @xmath56 , ( c ) @xmath57 , ( d ) @xmath58 ( deep sf regime ) .
all calculations use periodic boundary conditions .
the mean - field value @xmath43 for the mi transition in the commensurate case with one boson per lattice site is @xmath48 in 1d ( see fig .
[ slso ] ) .
( the mf value differs strongly from @xmath59 derived from qmc calculations @xcite or @xmath60 for dmrg calculations @xcite ) .
[ vgled],width=302 ] here , @xmath61 is the fourier transform of the wannier functions @xmath62 describing the wavefunction of a single lattice site .
the presence of the factor @xmath61 in eq .
( [ fourier ] ) provides a cutoff at high momenta .
the mean - field results for @xmath51 only describe the different long - range behavior in the mi and sf . for a homogeneous lattice
the correlation function calculated from the gutzwiller wavefunction gives @xmath63 for the diagonal elements and then drops instantly to @xmath64 for all off - diagonal elements @xmath65 .
short - range correlations are not reproduced by the gutzwiller approach .
this deviation is particularly severe in the mi , where the mean - field result predicts a completely flat momentum distribution , whereas the short - range correlations ( i.e. the exponential decay of @xmath51 ) yield smooth bumps in the expansion pattern . these can be distinguished from the @xmath66-peaks of the sf only after a sufficiently long expansion time .
applying perturbation theory to the gw wavefunction improves the structureless gw correlation function . in fig .
[ vgled ] , we have compared the results of gw mean - field theory , of pt , and of exact diagonalization ( ed ) .
the diagonalization has been carried out for a small 1d lattice , where it is easily feasible .
although there is no long - range order in the sf phase for the 1d case , where the density matrix exhibits a power - law decay towards zero , it is still reasonable to compare the short - range correlations .
indeed , we find a nice agreement between perturbation theory and exact results , not only for the mi ( see fig .
[ vgled]a ) , but also for the short - range behavior in the sf .
this agreement is made possible by the fact that @xmath51 decays only slowly and higher - order corrections show only negligible corrections for small lattices .
an example for the sf case is shown in fig .
[ vgled]d .
however , there is still a considerable difference for @xmath67 in fig .
[ vgled]d .
this is not surprising as we do pt up to second order .
hence correlations over a distance of three and more lattice sites are only corrected by the global mean - field correction for the infinite lattice ( see eq .
( [ cpsione ] ) , eq .
( [ cpsitwo ] ) , and fig .
[ 2ndcpsi ] in appendix [ appendixa ] ) .
we expect better agreement for larger lattice sizes as finite - size effects , arising in small lattices , are still considerable for @xmath68 sites used in our ed calculations . finally , for intermediate values of @xmath40 , shown in fig .
[ vgled]b , c , we observe a faster drop in the off - diagonal correlations , such that higher - order contributions in the pt become more important .
in any dimension , the perturbation @xmath25 is no longer small at the tip of the mi - sf transition lobe ( fig . [ vgled]c ) , and the pt breaks down .
nevertheless , comparing with exact diagonalization results , fig .
[ vgled]b shows still good agreement with ed , in contrast to fig .
[ vgled]c , which shows clear disagreement .
even though the parameters @xmath57 chosen for fig .
[ vgled]c are close to the mi - sf transition for 1d - lattices ( as predicted by dmrg @xcite and qmc @xcite calculations ) , pt reproduces the correct slope for the off - diagonal decay and lacks only the wrong offset from the mean field .
thus , even for this case , pt represents a qualitative improvement on the gutzwiller result .
the results for both approximations , gutzwiller and pt , are expected to become better in higher dimensions ( with the perturbative corrections diminishing in size ) . in second - order pt as a function of @xmath69 and @xmath40 .
the order parameter @xmath70 vanishes inside the mott - insulating lobes , whose mean - field boundaries are given by the white dashed line .
plots ( a )
( f ) display the resulting momentum distribution without the wannier form factor , @xmath71 , calculated for a 2d lattice with @xmath72 lattice sites .
( a)-(c ) are the gutzwiller mean - field results , and ( d)-(f ) are calculated using second - order pt .
the inset in ( c),(f ) shows a cut of one peak taken along @xmath73 ; dashed line for gw and solid line for the pt result .
arrows indicate the position of the respective plots in the @xmath74 phase diagram .
the parameters used are : @xmath75 , @xmath76 for ( a ) and ( d ) ; @xmath77 , @xmath78 for ( b ) and ( e ) ; and @xmath79 , @xmath80 for ( c ) and ( f ) .
the gray - scales of plots belonging to the same parameter set are identical .
expansion patterns ( a),(b),(d),and ( e ) are normalized to the peak maximum ; ( c ) and ( f ) are normalized to @xmath81 of the peak maximum.[expans ] , width=302 ] as discussed before , the perturbative enhancement of the description of short - range correlations is expected to lead to strong consequences for the momentum distributions . as an example we discuss a set of momentum distributions @xmath71 for a homogeneous 2d lattice , fig .
[ expans]a - f , and compare the gutzwiller results to those improved by pt .
the improved pt versions , figs .
[ expans]d - f , show much finer structures than the mean - field results , fig .
[ expans]a - c .
pt predicts broad peaks in the mi regions down to very small values of @xmath40 , fig .
[ expans]e , whereas the gutzwiller result without pt shows a structureless flat distribution for the whole mi region , fig .
[ expans]a , b . naturally , the modifications of @xmath71 are strongest near the phase transition , figs .
[ expans]a and [ expans]d .
going towards larger values @xmath40 into the sf phase , pt gives rise to a suppression of the peaks ( inset in fig .
[ expans]c , f )
. this suppression can be larger than @xmath82 of the original peak height and stems from the corrections to the mean field .
additionally , fig .
[ expans]f shows broad peaks induced by the inclusion of short - range correlations .
however , for large lattices , these broad peaks are small compared to the ( finite - size broadened ) sf @xmath66-peaks . on a plane through the trap center for a 3d lattice with @xmath83 lattice sites .
( a ) @xmath84 , @xmath85 , and @xmath86 .
( b ) @xmath75 , @xmath87 , and @xmath88 .
( c ) @xmath84 , @xmath87 , and @xmath86.[spd],width=302 ] -sites in the presence of a harmonic potential .
( a ) momentum distribution without the wannier form factor , @xmath71 , calculated along the @xmath89 direction .
filled symbols are the pt results and open symbols are the gw results .
the graphs for the mi surrounded by a sf shell ( diamonds ) are rescaled by a factor @xmath90 .
the results denoted by circles correspond to the occupation distribution of fig .
[ spd]a , diamonds to fig .
[ spd]c , and squares to fig .
( b ) difference between pt and gw results .
[ expans3d],title="fig:",width=264 ] -sites in the presence of a harmonic potential . ( a ) momentum distribution without the wannier form factor , @xmath71 , calculated along the @xmath89 direction .
filled symbols are the pt results and open symbols are the gw results .
the graphs for the mi surrounded by a sf shell ( diamonds ) are rescaled by a factor @xmath90 .
the results denoted by circles correspond to the occupation distribution of fig .
[ spd]a , diamonds to fig .
[ spd]c , and squares to fig .
( b ) difference between pt and gw results .
[ expans3d],title="fig:",width=264 ] in contrast to what was assumed in the last section , optical lattices used in experiments are not homogeneous . magnetic or optical trapping potentials are used @xcite to confine the atomic gas to a finite volume . the inhomogeneity caused by the trapping potential leads to slowly varying on - site energies , @xmath10 , in the mott - hubbard model eq .
( [ motthubb ] ) , that can be interpreted as a spatially varying chemical potential @xmath91 .
consequently the lattice is in general not in a pure mi or sf phase , but shows alternating shells of sf and mi regions .
an example for a sf region surrounded by a mi shell is shown in fig .
[ spd]b . considering the slowly varying on - site energy as
a spatially varying chemical potential gives a qualitative understanding of the shell structures .
the spatial variation of the chemical potential corresponds to a path parallel to the @xmath69 axis in the @xmath69-@xmath40-diagram .
starting with the potential minimum , in the trap center , and then moving off the center , decreases the effective local chemical potential . whenever a mi - sf ( sf - mi ) phase boundary is hit along the path in the @xmath69-@xmath40-diagram , a change from a mi to a sf ( sf to mi )
shell appears .
the inclusion of short - range correlations gives rise to considerable modifications also in the presence of an inhomogeneous trapping potential .
examples , calculated for a 3d - lattice , are shown in fig .
[ expans3d ] , where an underlying harmonic potential @xmath92 was chosen .
the three situations considered here correspond to a case with a large mi fraction ( fig . [ spd]a ) , a sf island surrounded by a mi shell ( fig .
[ spd]b ) , and a mi island surrounded by a sf phase ( fig .
[ spd]c ) . calculating the expansion patterns for these situations , fig .
[ expans3d ] , shows that the perturbative corrections arising from the short - range correlations lead to substantially different behavior in the different cases . for the almost complete mi state we get a correction to all wavevectors @xmath93 , with a fast drop at values close to the peak center @xmath94 , which leads to a peak broadening in the expansion picture ( see circles in fig . [ expans3d ] ) .
particularly large changes were found for the case of a sf island surrounded by a mi phase , fig .
[ expans3d ] ( squares ) .
again , corrections arise for all wavevectors , but , in contrast to the almost homogeneous case , the largest increase is now found for @xmath94 , with changes of about @xmath82 of the peak - maximum . finally , the reversed situation , a mi island surrounded by a sf phase ( diamonds in fig . [ expans3d ] ) does not show an increase of its maximum peak height but a considerable reduction ( over 5% of the peak maximum ) .
this is not surprising , as the majority of the lattice sites are now contributing to the sf phase , and a peak reduction was also observed for the bulk sf phase .
we would now like to compare our approach with the results obtained by kashurnikov _
@xcite , who used qmc calculations to calculate expansion patterns for a small 3d lattice with harmonic confinement .
there is good qualitative agreement in all cases with high superfluid fraction ( compare for example fig .
[ kash]a(b ) and fig .
2b(c ) in ref .
@xcite ) . however , the features of these expansion patterns are already well reproduced using the gutzwiller mean - field ansatz alone , in particular , the satellite peak , which was discussed as a signature of the mi - sf shell structure in @xcite .
corrections arising from pt show a suppression of the sf peak , as was discussed above . however , there are also considerable discrepancies to the qmc results for situations with a large mi fraction , even after implementing second - order pt . in these cases ( fig .
[ kash]c(d ) and fig .
2d(e ) in ref .
@xcite ) , the influence of the mi phase on the expansion picture broadens the peak and leads to a homogeneous background . the discrepancies to the qmc
are clearly visible in fig .
[ kash]c showing no peak broadening and a sattelite peak in contrast to the qmc results ( fig.2d in @xcite ) .
including the short range correlations perturbatively corrects the expansion pattern in the right direction , giving rise to a suppression of the sf - peak . considering the expansion pattern with the clearest mi features ( fig [ kash]d and fig .
2e in ref .
@xcite ) , we obtain the correct peak broadening from gw / pt calculation , but a larger ratio of mi background to sf peak .
concerning these discrepancies , we note that the expansion pattern is highly sensitive to the value of the mean field .
even small deviations can lead to a change in the sf - peak height sufficient to mask the flat distribution of the mi phase .
we checked , however , that the observed discrepancy is not due to a lack in accuracy of our numerical calculations .
we therefore believe that the discrepancies between qmc and gw / pt for situations with a large mi fraction can be attributed to the insufficiency of gw and low - order pt in describing the long - range correlations in this inhomogeneous situation .
in addition , we note that the lattice employed in @xcite is comparatively small for the given harmonic confinement potential ( with no complete shell of empty sites at the perimeter , see insets of fig .
[ kash ] ) , i.e. , the choice of boundary conditions ( periodic in the case of our numerical calculations ) may have non - negligible effects on the outer lattice sites . ) for a system of @xmath95 lattice sites considered in @xcite .
all plots are without the wannier form factor , @xmath96 .
the insets show the occupation number @xmath97 for a cut along the @xmath98-direction .
( a ) @xmath99 , @xmath100 , @xmath101 .
( b ) @xmath102 , @xmath103 , @xmath104 .
( c ) @xmath102 , @xmath105 , @xmath106 .
( d ) @xmath102 , @xmath107 , @xmath108 .
( note however , that the hamiltonian used in @xcite differs from eq .
( [ motthubb ] ) and hence the parameters are converted to the corresponding quantities used in our definitions . ) [ kash],width=321 ]
we have used perturbation theory to incorporate the effects of short - range correlations on top of the mean field solution of the bosonic mott hubbard model .
we derived corrections to local quantities , as well as to observable expansion patterns .
we numerically calculated the corrections to the mf - result , using pt up to second order , thus including correlations between next and next - nearest lattice sites .
modifications to the particle number fluctuations @xmath39 , arising from the pt , gave rise to the expected smooth transition of @xmath39 at the mi - sf transition .
moreover , comparing the pt results to the ed results for @xmath39 in 1d lattices showed good agreement for small values of @xmath40 .
of particular importance are the corrections to the correlation function , and thus to the expansion patterns .
we compared the correlation function obtained from pt with calculations from ed for small 1d - lattices and found good agreement . studying
the expansion patterns showed that the inclusion of the short - range correlations to the mf - ansatz gives rise to distinct modifications .
a broad peak can be seen in the pt results for the mi regime , visible even down to small values of @xmath40 . comparing pt and mf expansion patterns obtained for parameters in the sf region displayed a considerable suppression of the sf peak in the pt results .
additionally , on approaching the sf - mi transition from the sf side , broad peaks underlying the sf peaks were found in the pt - expansion patterns . including a harmonic confinement potential leads to situations where sf and mi regions coexist .
hence the perturbative corrections to the expansion pattern become more complex .
we studied lattices with different underlying harmonic traps giving rise to different constellations of sf - mi regions , finding modifications of up to @xmath82 of the peak maximum in the expansion patterns .
we would like to thank krishnendu sengupta for discussions .
our work was supported by the swiss nsf and the nccr nanoscience , as well as a dfg research fellowship ( f.m . ) .
schematic diagram illustrating the term @xmath109 appearing in the first - order correction @xmath110 of the density matrix , where @xmath111 is the perturbation connecting sites @xmath4 and @xmath112 .
the diagram shows the lattice sites in the horizontal direction .
the different steps needed to obtain the matrix element are shown vertically .
open circles denote the ground state ( gs ) of the given lattice site , while filled circles are excited states ( es ) of this site , for the mean - field hamiltonian @xmath21.,width=264 ]
we use standard stationary perturbation theory @xcite to calculate the corrections to the mean - field results induced by the perturbation @xmath113 where we introduce the new operators @xmath114 the expectation value @xmath115 is taken with respect to the mf wavefunction @xmath116 . defining the operator for the energy denominator as @xmath117 the expectation value of an observable @xmath118 including all corrections up to
second order is @xmath119 the first line in eq .
( [ perta ] ) is the mf - result @xmath120 , followed by two contributions which are the first - order corrections .
lines two and three in eq .
( [ perta ] ) are the second - order corrections to the mean value . as an example we will discuss the corrections to the density matrix @xmath121 .
the first - order correction to the density matrix is : @xmath122 for two different lattice sites @xmath123 , we find @xmath124 since the gutzwiller ground state is a product state . as a consequence ,
the contributions @xmath125 and @xmath126 vanish .
the only remaining contributions to @xmath127 stem from @xmath128 and @xmath129 . the graph shows a decomposition of the matrix element , with each row showing the wavefunction at an intermediate step in the evaluation of the matrix element .
as we deal with product states @xmath130 we represent the wavefunction by a row of circles , where each circle denotes the state @xmath131 of a particular lattice site @xmath4 .
open circles in fig .
[ diagramm ] denote a lattice site in its ground state ( gs ) , filled circles refer to an excited state ( es ) of this particular lattice site , with respect to the local mean - field hamiltonian .
note that , in general , this can be an arbitrarily highly excited state ( although higher contributions are suppressed by the energy denominator , and a cutoff is used in practice ) .
starting with a row of open circles , denoting the gs , @xmath116 , each following row corresponds to the state after the action of @xmath25 or @xmath132 , as indicated on the left side of the graph . as all matrix elements in eq .
( [ perta ] ) can be expressed in terms of a gs expectation value @xmath115 and a sequence of @xmath25 and @xmath132 operators , the first and last row must always be a line of open circles .
let us consider for instance the second term in eq .
( [ firstorder ] ) @xmath133 reading the graph in fig .
[ diagramm ] from top to bottom corresponds to reading the matrix element from right to left .
starting with the gs , @xmath134 , the first row consists of open circles .
the second row shows the state after the action of the perturbation @xmath25 .
acting with @xmath135 to the right onto the gs results in a state latexmath:[\[v_{ij } |g_0\rangle = \sum_{\alpha,\beta } f_{\alpha,\beta } @xmath112 are neighboring lattice sites and @xmath137 denotes the state with lattice site @xmath4 ( @xmath112 ) in the exited state @xmath138 ( @xmath139 ) and all other sites in their gs .
thus the second row shows the lattice sites @xmath4 and @xmath140 in an excited state ( filled circle ) , as the perturbation , eq .
( [ pert ] ) , allows only next neighbor interactions .
finally , the action of @xmath132 has to bring the excited states back to the gs , in order to get a non - vanishing contribution .
therefore , in first order pt , only next neighbor corrections to the correlation function arise , as the final row must represent the ground state @xmath141 again .
the graph representing the remaining first term in eq .
( [ firstorder ] ) is obtained by rotating the graph in fig .
[ diagramm ] by @xmath142 . and @xmath143 to the second - order correction of the density matrix .
( a ) terms representing contributions of the form eq .
( [ cpsione ] ) .
( b ) contributions of the form eq .
( [ cpsitwo ] ) .
all notations are the same as in fig .
[ diagramm ] .
[ 2ndcpsi],width=302 ] rewriting the second - order corrections to the density matrix , @xmath144 , in terms of the operators @xmath145 and @xmath146 gives eq .
( [ atoc ] ) , but with @xmath147 replaced by @xmath148 .
in contrast to the first - order corrections , now the terms proportional to @xmath149 and @xmath150 also give non - vanishing contributions .
using eq .
( [ perta ] ) we obtain : @xmath151 here , the primed sums run over all neighbors @xmath93 to site @xmath112 .
the corresponding subset of graphs for @xmath152 and @xmath153 are given by fig .
[ 2ndcpsi]a and fig .
[ 2ndcpsi]b for eq .
( [ cpsione ] ) and eq .
( [ cpsitwo ] ) respectively .
note that all terms of eq .
( [ cpsione ] ) and eq .
( [ cpsitwo ] ) give a correction to all matrix elements of the density matrix independent of the distance between the lattice sites .
we can understand these contributions as a modification to the mf value of the density matrix .
graphs showing second - order corrections arising from @xmath154 . diagrams ( a ) and ( b ) are coming from direct - neighbor contributions as given by eq .
( [ 2ndccone ] ) and eq .
( [ 2ndcctwo ] ) respectively . diagrams ( c ) and ( d ) are next - nearest - neighbor contributions : ( c ) corresponds to eq .
( [ 2ndccthree ] ) and eq .
( [ 2ndccfour ] ) ; ( d ) corresponds to eq .
( [ 2ndccfive]).,width=264 ] * lattice site @xmath4 and @xmath112 being direct neighbors .
in this case we get @xmath155 corrections for eq .
( [ 2ndccone ] ) and eq .
( [ 2ndcctwo ] ) are shown in fig . [ 2ndcc]a and fig .
[ 2ndcc]b , respectively .
* configurations corresponding to two lattice sites @xmath156 connected by two successive hopping steps via site @xmath93 .
this gives rise to six contributions : + @xmath157 + an example for the contributions arising from eq .
( [ 2ndccthree ] ) and eq .
( [ 2ndccfour ] ) is shown in fig .
[ 2ndcc]c . the last term , eq .
( [ 2ndccfive ] ) , has the representation shown in fig .
[ 2ndcc]d . note that for lattices with dimensions @xmath41 , the sites @xmath4,@xmath112 and @xmath93 need not necessarily form a straight line but can form a chevron . | we study the mott - insulator transition of bosonic atoms in optical lattices .
using perturbation theory , we analyze the deviations from the mean - field gutzwiller ansatz , which become appreciable for intermediate values of the ratio between hopping amplitude and interaction energy .
we discuss corrections to number fluctuations , order parameter , and compressibility .
in particular , we improve the description of the short - range correlations in the one - particle density matrix .
these corrections are important for experimentally observed expansion patterns , both for bulk lattices and in a confining trap potential . |
the paper is concerned with the spectral and computational analysis of effective preconditioners for linear systems arising from finite element approximations to the elliptic convection - diffusion problem @xmath0 with @xmath1 domain of @xmath2 .
we consider a model setting in which the structured finite element partition is made by equi - lateral triangles .
the interest of such a partition relies on the observation that automatic grid generators tend to construct equi - lateral triangles when the mesh is fine enough .
the analysis is performed having in mind two popular preconditioned krylov methods .
more precisely , we analyze the performances of the preconditioned conjugate gradient ( pcg ) method in the case of the diffusion problem and of the preconditioned generalized minimal residual ( pgmres ) in the case of the convection - diffusion problem .
we define the preconditioner as a combination of a basic ( projected ) toeplitz matrix times diagonal structures .
the diagonal part takes into account the variable coefficients in the operator of ( [ eq : modello ] ) , and especially the diffusion coefficient @xmath3 , while the ( projected ) toeplitz part derives from a special approximation of ( [ eq : modello ] ) when setting the diffusion coefficient to @xmath4 and the convective velocity field to @xmath5 . under such assumptions ,
if the problem is coercive , and the diffusive and convective coefficients are regular enough , then the proposed preconditioned matrix sequences have a strong clustering at unity , the preconditioning matrix sequence and the original matrix sequence are spectrally equivalent , and the eigenvector matrices have a mild conditioning .
the obtained results allow to show the optimality of the related preconditioned krylov methods .
it is important to stress that interest of such a study relies on the observation that automatic grid generators tend to construct equi - lateral triangles when the mesh is fine enough .
numerical tests , both on the model setting and in the non - structured case , show the effectiveness of the proposal and the correctness of the theoretical findings .
the outline of the paper is as follows . in section [ sez : fem ]
we report a brief description of the fe approximation of convection - diffusion equations and the preconditioner definition .
section [ sez : clustering ] is devoted to the spectral analysis of the underlying preconditioned matrix sequences , in the case of structured uniform meshes . in section [ sez : numerical_tests ] , after a preliminary discussion on complexity issues , selected numerical tests illustrate the convergence properties stated in the former section and their extension under weakened assumption or in the case of unstructured meshes . a final section [ sez : conclusions ] deals with perspectives and future works .
problem ( [ eq : modello ] ) can be stated in variational form as follows : @xmath6 where @xmath7 is the space of square integrable functions , with square integrable weak derivatives vanishing on @xmath8 .
we assume that @xmath1 is a polygonal domain and we make the following hypotheses on the coefficients : @xmath9 the previous assumptions guarantee existence and uniqueness for problem ( [ eq : formulazione_variazionale ] ) and hence the existence and uniqueness of the ( weak ) solution for problem ( [ eq : modello ] ) . for the sake of simplicity , we restrict ourselves to linear finite element approximation of problem ( [ eq : formulazione_variazionale ] ) . to this end , let @xmath10 be a usual finite element partition of @xmath11 into triangles , with @xmath12 and @xmath13 .
let @xmath14 be the space of linear finite elements , i.e. @xmath15 the finite element approximation of problem ( [ eq : formulazione_variazionale ] ) reads : @xmath16 for each internal node @xmath17 of the mesh @xmath18 , let @xmath19 be such that @xmath20 , and @xmath21 if @xmath22 .
then , the collection of all @xmath23 s is a base for @xmath24
. we will denote by @xmath25 the number of the internal nodes of @xmath18 , which corresponds to the dimension of @xmath24 .
then , we write @xmath26 as @xmath27 and the variational equation ( [ eq : formulazione_variazionale_fe ] ) becomes an algebraic linear system : @xmath28 according to these notations and definitions , the algebraic equations in ( [ eq : modello_discreto ] ) can be rewritten in matrix form as the linear system @xmath29 where @xmath30 and @xmath31 represent the approximation of the diffusive term and approximation of the convective term , respectively .
more precisely , we have @xmath32 where suitable quadrature formula are considered in the case of non constant coefficient functions @xmath33 and @xmath34 .
as well known , the main drawback in the linear system resolution is due to the asymptotical ill - conditioning ( i.e. very large for large dimensions ) , so that preconditioning is highly recommended .
hereafter , we refer to a preconditioning strategy previously analyzed in the case of fd / fe approximations of the diffusion problem @xcite and recently applied to fd / fe approximations @xcite of ( [ eq : modello ] ) with respect to the preconditioned hermitian and skew - hermitian splitting ( phss ) method @xcite .
more precisely , the preconditioning matrix sequence @xmath35 is defined as @xmath36 where @xmath37 , i.e. , the suitable scaled main diagonal of @xmath38 and clearly @xmath38 equals @xmath30 .
the computational aspects of this preconditioning strategy with respect to krylov methods will be discussed later in section [ sez : complexity_issues ] . here , preliminarily we want to stress as the preconditioner is tuned only with respect to the diffusion matrix @xmath30 : in other words , we are implicity assuming that the convection phenomenon is not dominant , and no stabilization is required in order to avoid spurious oscillations into the solution .
moreover , the spectral analysis is performed in the non - hermitian case by referring to the hermitian and skew - hermitian ( hss ) decomposition of @xmath39 ( that can be performed on any single elementary matrix related to @xmath18 by considering the standard assembling procedure ) . according to the definition
, the hss decomposition is given by @xmath40 where @xmath41 since by definition , the diffusion term @xmath30 is a hermitian matrix and does not contribute to the skew - hermitian part of @xmath39 .
notice also that @xmath42 if @xmath43 .
more in general , the matrix @xmath44 is symmetric and positive definite whenever @xmath45 . indeed , without the condition @xmath46
, the matrix @xmath47 does not have a definite sign in general : in fact , a specific analysis of the involved constants is required in order to guarantee the nonnegativity of the term @xmath47 . moreover
, the lemma below allows to obtain further information regarding such a spectral assumption , where @xmath48 indicate both the usual vector norms and the induced matrix norms . _
@xcite _ [ lemma : normae ] let @xmath49 be the matrix sequence defined according to _ ( [ eq : def_e_nb])_. under the assumptions in _ ( [ eq : ipotesi_coefficienti ] ) _ , then we find @xmath50 with @xmath51 absolute positive constant only depending on @xmath52 and @xmath1 .
the claim holds both in the case in which the matrix elements in _ ( [ eq : def_psi ] ) _ are evaluated exactly and whenever a quadrature formula with error @xmath53 is considered for approximating the involved integrals .
hereafter , we will denote by @xmath54 , @xmath55 , the matrix sequence associated to a family of meshes @xmath56 , with decreasing finesse parameter @xmath57 .
as customary , the whole preconditioning analysis will refer to a matrix sequence instead to a single matrix , since the goal is to quantify the difficulty of the linear system resolution in relation to the accuracy of the chosen approximation scheme .
the aim of this section is to analyze the spectral properties of the preconditioned matrix sequences @xmath58 in the case of some special domains @xmath1 partitioned with structured uniform meshes , so that spectral tools derived from toeplitz theory @xcite can be successfully applied .
the applicative interest of the considered type of domains will be motivated in the short .
indeed , let @xmath59 be a @xmath60variate lebesgue integrable function defined over @xmath61 ^ 2 $ ] . by referring to the fourier coefficients of this function @xmath59 , called generating function ,
@xmath62 with @xmath63 , one can build the sequence of toeplitz matrices @xmath64 .
the matrix @xmath65 is said to be the toeplitz matrix of order @xmath66 , @xmath67 , generated by @xmath59 .
now , the spectral properties of the matrix sequence @xmath64 are completely understood and characterized in terms of the underlying generating functions ( see , e.g. , @xcite for more details ) . in the following , with respect to the latter claim
, we will refer to the notion of equivalent generating functions .
in such a respect , we claim that two nonnegative function @xmath59 and @xmath68 defined over a domain @xmath69 are equivalent if and only if @xmath70 and @xmath71 , where @xmath72 means that there exists a pure positive constant @xmath73 such that @xmath74 almost everywhere on @xmath69 .
+ ( a ) ( b ) the main motivation of this paper lies in the analysis of the template case reported in figure [ fig : esagono_esagoni].a , where we consider a partition into equilateral triangles .
it is self - evident that such a problem represents just an academic example .
however , it is a fact that a professional mesh generator will locally produce a partitioning , which is `` asymptotically '' similar to the one reported in the figure . our goal is to prove the pcg optimality with respect to the diffusion problem approximation , i.e. , the number of iterations for reaching the solution , within a fixed accuracy , can be bounded from above by a constant independent of the dimension @xmath55 .
some additional results about pgmres convergence properties in the case of the convection - diffusion problem approximation are also reported , both in the case of the hexagonal domain @xmath1 with a structured uniform mesh as in figure [ fig : esagono_esagoni].a and in the case @xmath75 with a structured uniform mesh as in figure [ fig : esagono_esagoni].b , so extending previous results proved in @xcite with respect to the phss method @xcite .
the spectral analysis makes reference to the following definition . _
@xcite _ [ def : cluster ] let @xmath76 be a sequence of hermitian matrices of increasing dimensions @xmath77 .
the sequence @xmath76 is clustered at @xmath78 in the eigenvalue sense if for any @xmath79 , @xmath80 .
the sequence @xmath76 is properly ( or strongly ) clustered at @xmath78 if for any @xmath81 the number of the eigenvalues of @xmath82 not belonging to @xmath83 can be bounded by a pure constant eventually depending on @xmath84 , but not on @xmath77 . in the case of non - hermitian matrices ,
the same definition can be extended by considering the complex disk @xmath85 , instead of the interval @xmath83 .
we start by considering the simpler diffusion problem , i.e. , the analysis concerns the hermitian matrix sequences @xmath86 , @xmath87 .
due to the very special choice of the domain @xmath1 , the matrices @xmath87 arising in the case @xmath88 with structured meshes as in figure [ fig : esagono_esagoni].a result to be a proper projection of toeplitz matrices generated by the function @xmath89 , @xmath90 ^ 2 $ ] , related to a bigger parallelogram shaped domain @xmath91 containing @xmath1 ( see figure [ fig : esagono_esagoni_parallelogrammi ] ) , i.e. , @xmath92 , with @xmath93 number of the internal nodes . here ,
@xmath94 , @xmath95 is a proper projection matrix simply cutting those rows ( columns ) referring to nodes belonging to @xmath91 , but not to @xmath1 .
thus , the matrix is full rank , i.e. , @xmath96 , and @xmath97 .
in the same way , we can also consider the toeplitz matrix with the same generating function arising when considering a smaller parallelogram shaped domain @xmath98 contained in @xmath1 , i.e. , @xmath99 , with @xmath100 , @xmath101 proper projection matrix and such that @xmath102 , and @xmath103 .
these embedding arguments allow to bound , according to the min - max principle ( see @xcite for more details on the use of this proof technique ) , the minimal eigenvalue of the matrices @xmath104 as follows @xmath105 a first technical step in our spectral analysis concerns relationships between this generating function @xmath106 and the more classical generating function @xmath107 arising in the case of fe approximations on a square @xmath75 with friedrichs - keller meshes ( see figure [ fig : esagono_esagoni].b ) , or standard fd discretizations .
it can be easily observed that these two function are equivalent , in the sense of the previously reported definition , since we have @xmath108 ^ 2.\ ] ] thus , because @xmath109 is a matrix - valued linear positive operator ( lpo ) for every @xmath77 @xcite , the toeplitz matrix sequences generated by this equivalent functions result to be spectrally equivalent , i.e. , for any @xmath77 @xmath110 furthermore , due to the strict positivity of every @xmath109 ( see @xcite for the precise definition ) , since @xmath111 is not identically zero , we have that @xmath112 is positive definite , and since @xmath113 is not identically zero , we find that @xmath114 is also positive definite . here , we are referring to the standard ordering relation between hermitian matrices , i.e. , the notation @xmath115 , with @xmath116 and @xmath117 hermitian matrices , means that @xmath118 is nonnegative definite .
an interesting remark pertains to the fact that the function @xmath106 is the most natural one from the fe point of view , since no contribution are lost owing to the gradient orthogonality as instead in the case related to @xmath59 ; nevertheless its relationships with the function @xmath59 can be fully exploited in performing the spectral analysis .
more precisely , from ( [ eq : lmin_proiezione ] ) and ( [ eq : rel_toeplitz ] ) and taking into account ( see e.g. @xcite ) that @xmath119 we deduce @xmath120 following the very same reasoning , we also find that @xmath121 since , by the embedding argument , both @xmath93 and @xmath122 are asymptotic to @xmath77 , it follows that @xmath123 . finally ,
following the same analysis for the maximal eigenvalue , we find @xmath124 where , by @xcite , we know that @xmath125 and hence the spectral condition number of @xmath104 grows as @xmath126 i.e. @xmath127 where the constant hidden in the previous relation is mild and can be easily estimated .
it is worth stressing that the same matrix @xmath128 can also be considered in a more general setting .
in fact , the matrix sequence @xmath86 can also be defined as @xmath129 , since again each internal node in @xmath1 is a vertex of the same constant number of triangles .
therefore , by referring to projection arguments , the spectral analysis can be equivalently performed both on the matrix sequence @xmath86 and @xmath130 .
no matter about this choice , we make use of a second technical step , which is based on standard taylor s expansions .
[ lemma : taylor_expansion_esagoni ] let @xmath131 , with @xmath132 and @xmath1 hexagonal domain partitioned as in figure [ fig : esagono_esagoni].a .
for any @xmath78 such that @xmath133 there exists a proper @xmath134 such that the taylor s expansions centered at a proper @xmath134 have the form @xmath135 where @xmath136 and @xmath137 are constants independent of @xmath57 .
+ -1.0 cm in the cases at hand , the validity of this claim is just a direct check : the key point relies in the symmetry properties induced by the structured uniform nature of the considered meshes , both in the case of @xmath1 and @xmath91 .
hereafter , we give some detail with respect to the case of the hexagonal domain partitioned as in figure [ fig : esagono_esagoni].a , see @xcite for the proof in the case @xmath75 .
thus , let s consider the taylor s expansion centered at a proper @xmath134 .
by calling @xmath138 and similarly denoting the derivatives of the diffusion coefficient , it holds that @xmath139 with @xmath140^t \|$ ] and @xmath141 denoting the barycenter of the triangle @xmath142 . by referring to the assembling procedure
, we will choose as center of the related taylor s expansion in the case of figure [ fig : simmetrie].b-[fig : simmetrie].d the point marked by @xmath143 .
the symmetric position of the involved triangle barycenters ( marked by @xmath144 ) , with respect to @xmath143 allows to claim ( [ eq : exp.a ] ) .
the same holds true for ( [ eq : exp.b ] ) by considering each pertinent point @xmath143 in figure [ fig : simmetrie].a and , analogously , for ( [ eq : exp.c ] ) .
[ lemma : taylor_expansion ] _
let s considering the following notation of scaled matrix @xmath145 of a given matrix @xmath116
. under the assumptions of _ lemma [ lemma : taylor_expansion_esagoni ] _ the following representation holds true @xmath146 where @xmath147 and @xmath148 are uniformly bounded in spectral norm and have the same pattern as @xmath149 .
we are now ready to prove the pcg optimality .
[ teo : clu+se_a_esagoni ] let @xmath86 and @xmath35 be the hermitian positive definite matrix sequences defined according to _ ( [ eq : def_a ] ) _ and _ ( [ eq : def_p ] ) _ in the case of the hexagonal domain partitioned as in figure [ fig : esagono_esagoni].a . under the assumptions in _ ( [ eq : ipotesi_coefficienti ] ) _ , the sequence @xmath150 is properly clustered at @xmath4 . moreover , for any @xmath77 all the eigenvalues of @xmath151 belong to an interval @xmath152 $ ] well separated from zero _ [ spectral equivalence property]_. since @xmath153 , with @xmath94 such that @xmath96 , and @xmath97 , we have @xmath154^{-1 } a_n(\a ) \\ % & = & [ \pi \ , d_n^{\frac{1}{2}}(\a ) \ , \pi^t \ , \pi a_n(1 ) \ , \pi^t \ ,
\pi \ , d_n^{\frac{1}{2}}(\a ) \ , \pi^t ] ^{-1 } \pi \ , a_n(\a ) \ ,
\pi^t \\ % & = & [ \pi \ , \pi^t \ , \pi \ , d_n^{\frac{1}{2}}(\a ) a_n(1 ) d_n^{\frac{1}{2}}(\a ) \ , \pi^t \ ,
\pi \ , \pi^t ] ^{-1 } \pi \ , a_n(\a ) \ ,
\pi^t \\ % & = & [ \pi \ , d_n^{\frac{1}{2}}(\a ) a_n(1 ) d_n^{\frac{1}{2}}(\a ) \ , \pi^t ] ^{-1 } \pi \ , a_n(\a ) \ ,
\pi^t \\ % & = & [ \pi p_n^{-1}(\a ) \ , \pi^t ] ^{-1 } \pi \ , a_n(\a ) \ , \pi^t.\end{aligned}\ ] ] since @xmath128 is full rank , it is evident that the spectral behavior of @xmath155^{-1}\pi a_n(\a ) \pi^t\ ] ] is in principle better than the one of @xmath156 , to which we can address the spectral analysis .
thus , the proof technique refers to @xcite and the very key step is given by the relations outlined in ( [ eq : lmin_toeplitz_bottom ] ) and ( [ eq : lmin_toeplitz_top ] ) .
it is worth stressing that the previous claims can also be proved by considering the sequence @xmath150 , instead of @xmath157 , simply by directly referring to the quoted asymptotical expansions .
the interest may concern the analysis of the same uniform mesh on more general domains .
moreover , the same spectral properties has been proved in the case of uniform structured meshes as in figure [ fig : esagono_esagoni].b in @xcite .
the natural extension of the claim in theorem [ teo : clu+se_a_esagoni ] , in the case of the matrix sequence @xmath158 with @xmath159 , can be proved under the additional assumptions of lemma [ lemma : normae ] , in perfect agreement with theorem 5.3 in @xcite .
[ teo : clu+se_rea_esagoni ] let @xmath158 and @xmath35 be the hermitian positive definite matrix sequences defined according to _ ( [ eq : def_rea ] ) _ and _ ( [ eq : def_p ] ) _ in the case of the hexagonal domain partitioned as in figure [ fig : esagono_esagoni].a . under the assumptions in _ ( [ eq : ipotesi_coefficienti ] ) _ , the sequence @xmath160 is properly clustered at @xmath4 . moreover , for any @xmath77 all the eigenvalues of @xmath161 belong to an interval @xmath152 $ ] well separated from zero _ [ spectral equivalence property]_. the claim holds both in the case in which the matrix elements in _ ( [ eq : def_theta ] ) _ and _ ( [ eq : def_psi ] ) _ are evaluated exactly and whenever a quadrature formula with error @xmath53 is considered to approximate the involved integrals .
the proof can be done verbatim as in @xcite , since the key point is proved by referring to relations in ( [ eq : lmin_toeplitz_bottom ] ) and ( [ eq : lmin_toeplitz_top ] ) .
[ teo : clu+sb_ima ] let @xmath162 and @xmath35 be the hermitian matrix sequences defined according to _ ( [ eq : def_ima ] ) _ and _ ( [ eq : def_p ] ) _ in the case of the hexagonal domain partitioned as in figure [ fig : esagono_esagoni].a . under the assumptions in _ ( [ eq : ipotesi_coefficienti ] ) _ , the sequence @xmath163 is spectrally bounded and properly clustered at @xmath5 with respect to the eigenvalues .
the claim holds both in the case in which the matrix elements in _ ( [ eq : def_psi ] ) _ are evaluated exactly and whenever a quadrature formula with error @xmath53 is considered to approximate the involved integrals .
this result has been proved in @xcite with respect to friedrichs - keller triangulations , but it can easily be extended to the case of matrix sequences arising in the case of a fe partitioning as in figure [ fig : esagono_esagoni].a by using the same arguments . on the basis of these two
splitted spectral results , we can easily obtain the spectral description of the whole preconditioned matrix sequence @xmath164 , according to the theorem below .
the proof technique refers to an analogous result proved in @xcite with respect to fd discretizations of convection - diffusion equations .
[ teo : clua ] let @xmath165 and @xmath35 be the matrix sequences defined according to _ ( [ eq : def_a ] ) _ and _ ( [ eq : def_p ] ) _ , both in the case of the hexagonal domain @xmath1 with a structured uniform mesh as in figure [ fig : esagono_esagoni].a and in the case @xmath75 with a structured uniform mesh as in figure [ fig : esagono_esagoni].b under the assumptions in _
( [ eq : ipotesi_coefficienti ] ) _ , the sequence @xmath164 is properly clustered at @xmath166 with respect to the eigenvalues . in addition , these eigenvalues all belong to a uniformly bounded rectangle with positive real part , well separated from zero .
the claim holds both in the case in which the matrix elements in _ ( [ eq : def_theta ] ) _ and _ ( [ eq : def_psi ] ) _ are evaluated exactly and whenever a quadrature formula with error @xmath53 is considered to approximate the involved integrals .
a localization result for the eigenvalues of the sequence @xmath167 can be stated simply by referring to the properties of the field of values of the preconditioned matrix : in fact , for any @xmath17 @xmath168 where @xmath169 since for any @xmath17 @xmath170 the claimed localization result is obtained as a consequence of those previously proved in theorems [ teo : clu+se_rea_esagoni ] and [ teo : clu+sb_ima ] .
moreover , the same localization results together with the claimed clustering properties allow to apply theorem 4.3 in @xcite and the proof is concluded .
the subsequent result , giving estimates on the condition number of the eigenvector matrix of the preconditioned structure , is of interest in the study of the convergence speed of the gmres .
in particular a logarithmic growth in number of iteration is associated to a polynomial bound in the spectral condition number : such a bound is precisely given below .
[ lemmma : eigenvectors ] let @xmath3 and @xmath52 constant functions and let @xmath165 and @xmath35 be the matrix sequences defined as in theorem [ teo : clua ] @xmath171 can be diagonalized as @xmath172 where the matrix of the eigenvectors @xmath173 can be chosen such that @xmath174 . under the quoted assumption it holds that @xmath175 and @xmath31 is skew - symmetric .
thus , we have @xmath176 with @xmath177 skew - symmetric matrix . therefore , the matrix @xmath178 is a normal normal , so that @xmath179 , where @xmath180 is diagonal and @xmath181 is unitary . consequently ,
@xmath182 with @xmath183 eigenvector matrix . due to the fact that @xmath181 is unitary
, we have @xmath184 and finally @xmath185 by invoking the key relation ( [ eq : cond - an1 ] ) .
the section is divided into two parts . in the first we briefly discuss the complexity features of our preconditioning proposals . in the second part we report and critically discuss few numerical experiments in which the meshes are both structured and unstructured , while the regularity features required by the theoretical analysis are somehow relaxed .
first of all , we report some remarks about the computational costs of the proposed iterative procedure .
the main idea is that such a technique is easily applicable .
in fact , a krylov method is considered , so simply requiring a matrix vector routine for sparse matrices and a solver for the chosen preconditioner . + therefore , since the preconditioner is defined as @xmath186 , where @xmath187 , the solution of the linear system in ( [ eq : modello_discreto ] ) with matrix @xmath39 is reduced to computations involving diagonals and the matrix @xmath188 ( @xmath189 and @xmath188 for the diffusion problem , respectively ) .
as well known , whenever the domain is partitioned by considering a uniform structured mesh this latter task can be efficiently performed by means of fast poisson solvers , among which we can list those based on the cyclic reduction idea ( see e.g. @xcite ) and several specialized algebraic or geometric multigrid methods ( see e.g. @xcite ) . in addition , the latter can be efficiently considered also in more general mesh settings .
the underlying idea is that the main effort in devising efficient algorithms must be devoted only to this simpler problem with constant coefficient , instead of the general one .
now , the effectiveness of the proposed method is measured by referring to the optimality definition below . _
@xcite _ [ def : opt ] let @xmath190 be a given sequence of linear systems of increasing dimensions .
an iterative method is _ optimal _ if 1 .
the arithmetic cost of each iteration is at most proportional to the complexity of a matrix - vector product with matrix @xmath191 , 2 .
the number of iterations for reaching the solution within a fixed accuracy can be bounded from above by a constant independent of @xmath192 .
in such a respect theorem [ teo : clu+se_a_esagoni ] proves the optimality of the pcg method : the iterations number for reaching the solution within a fixed accuracy can be bounded from above by a constant independent of the dimension @xmath55 and the arithmetic cost of each iteration is at most proportional to the complexity of a matrix - vector product with matrix @xmath38 .
moreover , theorem [ teo : clua ] proves that all the eigenvalues of the preconditioned matrix belong in a complex rectangle @xmath193 , \
\mathrm{im}(z ) \in [ -\hat{d},\hat{d}]\}$ ] , with @xmath194 , @xmath195 independent of the dimension @xmath55 .
it is worth stressing that the existence of a proper eigenvalue cluster and the aforementioned localization results in the preconditioned spectrum can be very important for fast convergence of preconditioned gmres iterations ( see , e.g. , @xcite ) .
finally , we want to give notice that the pcg / pgmres numerical performances do not get worse in the case of unstructured meshes , despite the lack of a rigorous proof . before analyzing in detail selected numerical results , we wish to give technical information on the performed numerical experiments .
we apply the pcg or the pgmres method , in the symmetric and non - symmetric case , respectively , with the preconditioning strategy described in section [ sez : fem ] , to fe approximations of the problem ( [ eq : modello ] ) .
whenever required , the involved integrals have been approximated by means of the barycentric quadrature rule ( the approximation by means of the nodal quadrature rule gives rise to similar results , indeed both are exact when applied to linear functions ) .
the domains of integration @xmath1 are those reported in figure [ fig : esagono_esagoni ] and we assume dirichlet boundary conditions .
all the reported numerical experiments are performed in matlab , by employing the available ` pcg ` and ` gmres ` library functions ; the iterative solvers start with zero initial guess and the stopping criterion @xmath196 is considered .
the case of unstructured meshes is also discussed and compared , together with various types of regularity in the diffusion coefficient .
in fact , we consider the case of a coefficient function satisfying the assumptions ( [ eq : ipotesi_coefficienti ] ) .
more precisely , the second columns in table [ tab : it - es123esagono_triangoliequilateri_m5 ] report the number of iterations required to achieve the convergence for increasing values of the coefficient matrix size @xmath55 when considering the fe approximation with structured uniform meshes as in figure [ fig : esagono_esagoni ] and with template function @xmath197 , @xmath198^t$].the subsequent meshes are obtained by means of a progressive refinement procedure , consisting in adding new nodes corresponding to the middle point of each edge .
the numerical experiments plainly confirm the previous theoretical analysis in section [ sez : clustering ] : the convergence behavior does not depend on the coefficient matrix dimension @xmath55 . as anticipated , despite the lack of corresponding theoretical results , we want to test the convergence behavior also in the case in which the regularity assumption on @xmath199 in theorems [ teo : clu+se_a_esagoni ] and [ teo : clua ] are not satisfied .
the analysis is motivated by favorable known numerical results in the case of fd approximations ( see , for instance , @xcite ) or fe approximation with only the diffusion term @xcite .
more precisely , we consider as template the @xmath200 function @xmath201 , the @xmath202 function @xmath203 , with @xmath204 or @xmath205 . the number of required iterations is listed in the remaining columns in table [ tab : it - es123esagono_triangoliequilateri_m5 ] .
notice that also in these cases with a @xmath200 or @xmath202 diffusion function , the iteration count does not depend on the coefficient matrix dimension @xmath55 . the same results are reported in table [ tab : it - es123quadrato_triangolirettangoli_m1 ] with respect to structured uniform meshes on the domain @xmath75 .
furthermore , we want to test our proposal in the case of some unstructured meshes generated by triangle @xcite with a progressive refinement procedure .
the first meshes in the considered mesh sequences are reported in figures [ fig : mesh_m6newnew ] and [ fig : mesh_m4 ] , respectively .
tables [ tab : it - es123esagono_unstructured_m6newnew ] and [ tab : it - es123quadrato_unstructured_m4 ] report the number of required iterations in the case of the previous template functions .
negligible differences in the iteration trends are observed for increasing dimensions @xmath77 .
all these remarks are in perfect agreement with the outliers analysis of the matrices @xmath206 , with respect to a cluster at @xmath207 ( see some examples in figure [ fig : outlier_analysis ] ) . to conclude the section we take into consideration the more realistic setting in which the meshes are generated by a specialized automatic procedure .
according to this point of view , we have applied a spectral approximation of the matrix sequence @xmath208 in terms of product of low - cost matrix structures , i.e. , @xmath104 and @xmath209 , that carry the `` structural '' content of the variational problem and the specific `` informative '' content contained in the diffusion coefficient function @xmath210 , respectively .
however , the matrix @xmath104 can be still not easy to handle .
therefore , we go beyond in such idea by defining a preconditioner @xmath211 in which the matrix @xmath104 related to the unstructured mesh is replaced by a suitable projection of the toeplitz matrix @xmath212 , @xmath89 , @xmath90 ^ 2 $ ] ( see section [ sez : diffusion_equation ] ) .
the numerical results are reported in table [ tab : it_perturbed_mesh ] . as expected ,
the number of iterations is a constant with respect to the dimension when the preconditioner @xmath213 is applied .
nevertheless , for @xmath77 large enough , the same seems to be true also for the preconditioner @xmath211 .
indeed , for increasing dimensions @xmath77 , the unstructured partitioning is more and more similar to the one given by equilateral triangles sketched in fig .
[ fig : esagono_esagoni].a .
a theoretical ground supporting these observation is still missing and would be worth in our opinion to be studied and developed .
@xmath77 & @xmath214 & @xmath215 & @xmath216 + 37 & 3 & 4 & 5 + 169 & 3 & 4 & 5 + 721 & 3 & 4 & 4 + 2977 & 3 & 4 & 4 + 12097 & 3 & 4 & 4 + 48769 & 3 & 4 & 4 + @xmath77 & @xmath214 & @xmath215 & @xmath216 + 37 & 4 & 5 & 5 + 169 & 4 & 5 & 5 + 721 & 4 & 5 & 5 + 2977 & 4 & 5 & 5 + 12097 & 4 & 5 & 5 + 48769 & 4 & 5 & 5 + @xmath77 & @xmath214 & @xmath215 & @xmath216 + 81 & 3 & 4 & 4 + 361 & 3 & 4 & 5 + 1521 & 3 & 4 & 5 + 6241 & 3 & 4 & 5 + 25281 & 3 & 4 & 5 + 128881 & 3 & 4 & 5 + @xmath77 & @xmath214 & @xmath215 & @xmath216 + 81 & 4 & 5 & 5 + 3611 & 4 & 5 & 5 + 1521 & 4 & 5 & 5 + 6241 & 4 & 5 & 5 + 25281 & 4 & 5 & 5 + 128881 & 4 & 5 & 5 + @xmath77 & @xmath214 & @xmath215 & @xmath216 + 28 & 5 & 5 & 5 + 73 & 4 & 4 & 5 + 265 & 4 & 4 & 5 + 1175 & 4 & 4 & 5 + 4732 & 4 & 4 & 5 + 19288 & 4 & 4 & 5 + 76110 & 4 & 4 & 4 + @xmath77 & @xmath214 & @xmath215 & @xmath216 + 28 & 4 & 5 & 5 + 73 & 4 & 5 & 5 + 265 & 4 & 5 & 5 + 1175 & 4 & 5 & 5 + 4732 & 4 & 5 & 5 + 19288 & 4 & 5 & 5 + 76110 & 4 & 5 & 5 + @xmath77 & @xmath214 & @xmath215 & @xmath216 + 24 & 4 & 5 & 5 + 109 & 4 & 5 & 5 + 465 & 4 & 5 & 5 + 1921 & 4 & 5 & 5 + 7809 & 4 & 5 & 5 + 31489 & 4 & 5 & 5 + @xmath77 & @xmath214 & @xmath215 & @xmath216 + 24 & 4 & 5 & 5 + 109 & 4 & 5 & 5 + 465 & 4 & 5 & 5 + 1921 & 4 & 5 & 5 + 7809 & 4 & 5 & 5 + 31489 & 4 & 5 & 5 + [ tab : it - es123esagono_perturbed ] @xmath77 & @xmath217 & @xmath218 + 37 & 4 & 8 + 169 & 4 & 10 + 721 & 4 & 11 + 2977 & 4 & 11 + 12095 & 4 & 13 + 48769 & 4 & 16 + @xmath77 & @xmath217 & @xmath218 + 37 & 4 & 8 + 169 & 4 & 9 + 721 & 4 & 9 + 2977 & 4 & 10 + 12095 & 4 & 11 + 48769 & 4 & 13 + -0.3 cm + -4.5 cm
-0.5 cm -3.5 cm -0.5 cm -0.5 cm + -0.5 cm -0.5 cm + -0.5 cm -0.5 cm
as emphasized in the introduction it is clear that the problem in figure [ fig : esagono_esagoni].a is just an academic example , due to the perfect structure made by equi - lateral triangles .
however , it is a fact that a professional mesh generator will produce a partitioning , which `` asymptotically '' , that is a for a mesh fine enough , tends to the one in figure [ fig : esagono_esagoni].a .
the latter fact has a practical important counterpart , since the academic preconditioner @xmath219 is optimal for the real case with nonconstant coefficients and with the partitioning in b ) .
a theoretical ground supporting these observations is still missing and would be worth in our opinion to be studied and developed in three directions : a ) giving a formal notion of convergence of a partitioning to a structured one , b ) proving spectral and convergence results in the case of an asymptotically structured partitioning , c ) extending the spectral analysis in the case of weak regularity assumptions .
other possible developments include the case of higher order finite element spaces : it would be intriguing to find an expression of the underlying toeplitz symbol as a function of the different parameters in the considered finite element space ( degrees of freedom , polynomial degree , geometry of the mesh ) , and this could be done uniformly in @xmath220 dimension , i.e. for equation ( [ eq : modello ] ) with @xmath221 , @xmath222 .
10 o. axelsson , m. neytcheva , _ the algebraic multilevel iteration methods theory and applications .
_ in _ proceedings of the second international colloquium on numerical analysis _
( plovdiv , 1993 ) , 1323 , vsp , 1994 . z. bai , g.h . golub , m.k .
ng , _ hermitian and skew - hermitian splitting methods for non - hermitian positive definite linear systems .
_ siam j. matrix anal .
( 2003 ) , no .
3 , 603626 .
d. bertaccini , g.h .
golub , s. serra capizzano , c. tablino possio , _ preconditioned hss methods for the solution of non - hermitian positive definite linear systems and applications to the discrete convection - diffusion equation .
99 ( 2005 ) , no .
3 , 441484 .
d. bertaccini , g.h .
golub , s. serra - capizzano , _ spectral analysis of a preconditioned iterative method for the convection - diffusion equation .
_ siam j. matrix anal .
( 2007 ) , no .
1 , 260278 .
d. bertaccini and michael k. ng , band - toeplitz preconditioned gmres iterations for time - dependent pdes , bit , 40 ( 2003 ) , pp . 901914 .
a. bttcher , b. silbermann , _ introduction to large truncated toeplitz matrices .
_ springer - verlag , new york , 1998 .
b. buzbee , f. dorr , j. george , g.h .
golub , _ the direct solutions of the discrete poisson equation on irregular regions .
_ siam j. numer
( 1971 ) , 722736
. f. dorr , _ the direct solution of the discrete poisson equation on a rectangle .
_ siam rev . 12 ( 1970 ) , 248263 .
w. hackbusch , _ multigrid methods and applications . _ springer verlag , berlin , germany , ( 1985 ) .
a. russo , c. tablino possio , _ preconditioned hss method for finite element approximations of convection - diffusion equations_. siam j. matrix anal .
( 2009 ) , no .
3 , 9971018 .
s. serra capizzano , _ on the extreme eigenvalues of hermitian ( block)toeplitz matrices . _ linear algebra appl .
270 ( 1998),109129 . s. serra capizzano , _ the rate of convergence of toeplitz based pcg methods for second order nonlinear boundary value problems .
81 ( 1999 ) , no .
3 , 461495 .
s. serra capizzano , _ some theorems on linear positive operators and functionals and their applications . _ computers and mathematics with applications 39 , no .
7 - 8 ( 2000 ) , 139167 .
s. serra capizzano , _ convergence analysis of two grid methods for elliptic toeplitz and pdes matrix sequences .
92 - 3 ( 2002 ) , 433 - 465 . s. serra capizzano , c. tablino possio , _ spectral and structural analysis of high precision finite difference matrices for elliptic operators .
_ linear algebra appl . 293
( 1999 ) , no . 1 - 3 , 85131 .
s. serra capizzano , c. tablino possio , _ high - order finite difference schemes and toeplitz based preconditioners for elliptic problems .
_ electron .
11 ( 2000 ) , 5584 .
s. serra capizzano , c. tablino possio , _ finite element matrix sequences : the case of rectangular domains .
algorithms 28 ( 2001 ) , no . 1 - 4 , 309327
. s. serra capizzano , c. tablino possio , _ preconditioning strategies for 2d finite difference matrix sequences . _
anal . 16 ( 2003 ) , 129 .
s. serra capizzano , c. tablino possio , _ superlinear preconditioners for finite differences linear systems .
_ siam j. matrix anal .
appl . 25 ( 2003 ) , no
. 1 , 152164 . j.r .
shewchuk , _ a two - dimensional quality mesh generator and delaunay triangulator .
( version 1.6 ) _ , www.cs.cmu.edu/ quake
/ triangle.html p. swarztrauber,_the method of cyclic reduction , fourier analysis and the facr algorithm for the discrete solution of poisson s equation on a rectangle . _ siam rev .
19 ( 1977 ) , 490501 .
u. trottenberg , c.w .
oosterlee , and a. schller , multigrid , academic press , london , 2001 .
tyrtyshnikov , _ a unifying approach to some old and new theorems on distribution and clustering . _ linear algebra appl
. 232 ( 1996 ) , 143 . | the paper is devoted to the spectral analysis of effective preconditioners for linear systems obtained via a finite element approximation to diffusion - dominated convection - diffusion equations .
we consider a model setting in which the structured finite element partition is made by equi - lateral triangles . under such assumptions ,
if the problem is coercive , and the diffusive and convective coefficients are regular enough , then the proposed preconditioned matrix sequences exhibit a strong clustering at unity , the preconditioning matrix sequence and the original matrix sequence are spectrally equivalent , and the eigenvector matrices have a mild conditioning .
the obtained results allow to show the optimality of the related preconditioned krylov methods .
the interest of such a study relies on the observation that automatic grid generators tend to construct equi - lateral triangles when the mesh is fine enough .
numerical tests , both on the model setting and in the non - structured case , show the effectiveness of the proposal and the correctness of the theoretical findings . matrix sequences , clustering , preconditioning , non - hermitian matrix , krylov methods , finite element approximations 65f10 , 65n22 , 15a18 , 15a12 , 47b65 |
the design and characterization of strategies for controlling quantum dynamics is vital to a broad spectrum of applications within contemporary physics and engineering .
these range from traditional coherent - control settings like high - resolution nuclear @xcite and molecular spectroscopy @xcite , to a variety of tasks motivated by the rapidly growing field of quantum information science @xcite .
in particular , the ability to counteract decoherence effects that unavoidably arise in the dynamics of a real - world quantum system coupled to its surrounding environment is a prerequisite for scalable realizations of quantum information processing ( qip ) , as actively pursued through a variety of proposed device technologies @xcite .
active decoupling techniques offer a conceptually simple yet powerful control - theoretic setting for quantum - dynamical engineering of both closed - system ( unitary ) and open - system ( non - unitary ) evolutions .
inspired by the idea of _ coherent averaging _ of interactions by means of tailored pulse sequences in nuclear magnetic resonance ( nmr ) spectroscopy @xcite , decoupling protocols consist of repetitive sequences of control operations ( typically drawn from a finite repertoire ) , whose net effect is to coherently modify the natural target dynamics to a desired one . in practice , a critical decoupling task is the selective removal of unwanted couplings between subsystems of a fully or partially controllable composite quantum system .
historically , a prototype example is the elimination of unwanted phase evolution in interacting spin systems via trains of @xmath0-pulses ( the so - called hahn - echo and carr - purcell sequences @xcite ) . for open quantum systems
, this line of reasoning motivates the question of whether removing the coupling between the system of interest and its environment may be feasible by a control action restricted to the former only .
such a question was addressed in @xcite for the paradigmatic case of a single qubit coupled to a bosonic reservoir , establishing the possibility of decoherence suppression in the limit of rapid spin flipping via the echo sequence mentioned above .
the study of dynamical decoupling as a general strategy for quantum coherent and error control has since then attracted a growing interest from the point of view of both model - independent decoupling design and optimization , and the application to specific physical systems .
representative contributions include the extension to arbitrary finite - dimensional systems via dynamical - algebraic @xcite , geometric @xcite , and linear - algebraic @xcite formulations ; the construction of fault - tolerant eulerian @xcite and concatenated decoupling protocols @xcite , as well as efficient combinatorial schemes @xcite ; the connection with quantum zeno physics @xcite ; proposed applications to the compensation of specific decoherence mechanisms ( notably , magnetic state decoherence @xcite and 1/@xmath1 noise @xcite ) and/or the removal of unwanted evolution within trapped - ion @xcite and solid - state quantum computing architectures @xcite .
these theoretical advances have been paralleled by steady experimental progress .
beginning with a proof - of - principle demonstration of decoherence suppression in a single - photon polarization interferometer @xcite , dynamical decoupling techniques have been implemented alone and in conjunction with quantum error correction within liquid - state nmr qip @xcite , and have inspired charge - based @xcite and flux - based @xcite echo experiments in superconducting qubits .
recently , dynamic decoherence control of a solid - state nuclear quadrupole qubit has been reported @xcite .
all the formulations of dynamical decoupling mentioned so far share the feature of involving purely _ deterministic _ control actions . in the simplest
setting , these are arbitrarily strong , effectively instantaneous rotations ( so - called _ bang - bang controls _ ) chosen from a discrete group @xmath2 .
decoupling according to @xmath2 is then accomplished by sequentially cycling the control propagator through _ all _ the elements of @xmath2 .
if @xmath3 denotes the separation between consecutive control operations , this translates into a minimal averaging time scale @xmath4 , of length proportional to the size @xmath5 of @xmath6 .
the exploration of decoupling schemes incorporating _
stochastic _ control actions was only recently undertaken .
a general control - theoretic framework was introduced by viola and knill in @xcite ( see also @xcite ) , based on the idea of seeking faster convergence ( with respect to an appropriately defined metric ) by _ randomly sampling _ rather than systematically implementing control operations from @xmath2 .
based on general lower bounds for pure - state error probabilities , the analysis of @xcite indicated that random schemes could outperform their cyclic counterpart in situations where a large number of elementary control operations is required or , even for small control groups , when the interactions to be removed vary themselves in time over time scales long compared to @xmath3 but short compared to @xmath7 .
furthermore , it also suggested that advantageous features of pure cyclic and random methods could be enhanced by appropriately merging protocols within a hybrid design .
the usefulness of randomization in the context of actively suppressing _ coherent _ errors due to residual static interactions was meanwhile independently demonstrated by the so - called _ pauli random error correction method _ ( parec ) , followed by the more recent _ embedded dynamical decoupling method _ both due to kern and coworkers @xcite .
both protocols may be conceptually understood as following from randomization over the pauli group @xmath8 , used alone or , respectively , in conjunction with a second set of deterministic control operations .
our goal in this work is twofold : first , to develop a quantitative understanding of _ typical _ randomized control performance for both _ coherent and decoherent phase errors _ , beginning from the simplest scenario of a single qubit already investigated in detail in the deterministic case @xcite ; second , to clarify the _ physical _ picture underlying random control , by devoting , in particular , special attention to elucidate the control action and requirements in rotating frames associated with different dynamical representations .
the fact that the controlled dynamics remains exactly solvable in the bang - bang ( bb ) limit makes the single - qubit pure - dephasing setting an ideal test - bed for these purposes . from a general standpoint ,
since spin - flip decoupling corresponds to averaging over the smallest ( nontrivial ) group @xmath9 , with @xmath10 @xcite , this system is not yet expected to show the full advantage of the random approach .
remarkably , however , control scenarios can still be identified , where randomized protocols indeed represent the most suitable choice .
the content of the paper is organized as follows . after laying out the relevant system and control settings in sect .
ii , we begin the comparison between cyclic and randomized protocols by studying the task of phase refocusing in a qubit evolving unitarily in sect .
control of decoherence from purely dephasing semiclassical and quantum environments is investigated in the main part of the paper , sects .
iv and v. we focus on the relevant situations of decoherence due to random telegraph noise and to a fully quantum bosonic bath , respectively . both exact analytical and numerical results for the controlled decoherence process are presented in the latter case .
we summarize our results and discuss their significance from the broader perspective of constructively exploiting randomness in physical systems in sect .
vi , by also pointing to some directions for future research .
additional technical considerations are included in a separate appendix .
our target system @xmath11 is a single qubit , living on a state space @xmath12 .
the influence of the surrounding environment may be formally accounted for by two main modifications to the isolated qubit dynamics .
first , @xmath11 may couple to effectively _ classical _ degrees of freedom , whose net effect may be modeled through a deterministic or random time - dependent modification of the system parameters . additionally , @xmath11 may couple to a _
quantum _ environment @xmath13 , that is , a second quantum system defined on a state space @xmath14 with which @xmath11 may become entangled in the course of the evolution . for the present purposes , @xmath13 will be schematized as a bosonic reservoir consisting of independent harmonic modes .
let @xmath15 denote the identity operator on @xmath16 , respectively . throughout the paper
, we will consider different dynamical scenarios , corresponding to special cases of the following total drift hamiltonian on @xmath17 : @xmath18 where @xmath19 here , we set @xmath20 , and @xmath21 ( @xmath22 ) , @xmath23 and @xmath24 denote pauli spin matrices , and canonical creation and annihilation bosonic operators of the @xmath25th environmental mode with frequency @xmath26 , respectively . @xmath27 and @xmath28 are real and complex functions that account for an effectively time - dependent frequency of the system and its coupling to the @xmath25th reservoir mode , respectively .
we shall write @xmath29 for an appropriate choice of central values @xmath30 , @xmath31 and modulation functions @xmath32 , @xmath33 , respectively .
the adimensional parameter @xmath34 is introduced for notational convenience , allowing to include ( @xmath35 ) or not ( @xmath36 ) the coupling to @xmath13 as desired .
physically , because @xmath37 and @xmath38 commute at all times , the above hamiltonian describes a _ purely decohering _ coupling between @xmath11 and @xmath13 , which does not entail energy exchange . while in general dissipation might also occur , focusing on pure decoherence is typically justified for sufficiently short time scales @xcite and , as we shall see , has the advantage of making exact solutions available as benchmarks .
control is introduced by adjoining a classical controller acting on @xmath11 , that is by adding a time - dependent term to the above target hamiltonian , @xmath39 in our case , @xmath40 will be designed so as to implement appropriate sequences of bb pulses .
this may be accomplished by starting from a rotating radiofrequency field ( or , upon invoking the rotating wave approximation , by a linearly - polarized oscillating field ) , described by the following amplitude- and phase - modulated hamiltonian : @xmath41 \sigma_x + \sin[\omega t + \varphi_j ( t ) ] \sigma_y \hspace*{-.5mm}\big ) \,,\ ] ] with @xmath42\:.\ ] ] here , @xmath43 denotes the heaviside step function ( defined as @xmath44 for @xmath45 and @xmath46 for @xmath47 ) , @xmath48 and @xmath49 are positive parameters , and @xmath50 denotes the instants at which the pulses are applied . if the carrier frequency is tuned on resonance with the central frequency , @xmath51 , and the phase @xmath52 for each @xmath53 , the above hamiltonian schematizes a train of _ identical _ control pulses of amplitude @xmath48 and duration @xmath49 in the physical frame . under the bb requirement of impulsive switching ( @xmath54 ) with unbounded strength ( @xmath55 ) ,
it is legitimate to neglect @xmath56 ( including possible off - resonant effects ) within each pulse , effectively leading to qubit rotations about the @xmath57-axis . in particular
, a @xmath0 rotation corresponds to @xmath58 ( see also appendix a ) . in what follows ,
we shall focus on using trains of bb @xmath0-pulses to effectively achieve a net evolution characterized by the identity operator ( the so - called _ no - op _ gate ) .
this requires averaging unwanted ( coherent or decoherent ) @xmath59 evolution generated by either @xmath37 or @xmath38 or both , by subjecting the system to repeated spin - flips . in group - theoretic terms
such protocols have , as mentioned , a transparent interpretation as implementing an average over the group @xmath60 , represented on @xmath61 as @xmath62 @xcite .
the quantum operation effecting such group averaging is the projector @xmath63 on the space of operators commuting with @xmath64 , leading to @xmath65 essentially , in _ cyclic _ decoupling schemes based on @xmath2 the above symmetrization is accomplished through a _ time _ average of the effective hamiltonian determining the evolution over a cycle , @xmath7 ; in _ random _ schemes , it emerges from an _ ensemble _ average over different control histories , taken with respect to the uniform probability measure over @xmath66 @xcite .
neither deterministic nor stochastic sequences of @xmath0 pulses achieve an exact implementation of @xmath67 for a fully generic hamiltonian as in eqs .
( [ drift0])-([drift ] ) , except in the ideal limit of arbitrarily fast control where the separation between pulses approaches zero .
therefore , it makes sense to compare the performance attainable by different control sequences for realistic control rates . in this paper
we shall focus on the following options .
@xmath68 * asymmetric cyclic protocol ( a ) * , fig .
[ fig : scheme](a ) .
this is the protocol used in @xcite , corresponding to repeated spin - echoes .
cyclicity is ensured by subjecting the system to an even number of equally spaced @xmath0-pulses , applied at @xmath69 , @xmath70 , in the limit @xmath54 .
the elementary cycle consists of two pulses : the first one , applied after the system evolved freely for an interval @xmath3 , reverses the qubit original state and the second one , applied a time @xmath3 later , restores its original state .
@xmath68 * symmetric cyclic protocol ( s ) * , fig .
[ fig : scheme](b ) .
this protocols , which is directly inspired to the carr - purcell sequence of nmr , is obtained from ( a ) by rearranging the two @xmath0-pulses within each cycle in such a way that the control propagator is symmetric with respect to the middle point .
the first pulse is applied at @xmath71 and the next ones at @xmath72 , with @xmath73 .
both the a- and s - protocols have a cycle time @xmath10 , and lead to the same averaging in the limit @xmath74 . for finite @xmath75 , however , the symmetry of the s - protocol guarantees the cancellation of lowest - order corrections @xmath76 , resulting in superior averaging @xcite . @xmath68 * long symmetric cyclic protocol ( ls ) * , fig .
[ fig : scheme](c ) .
this is basically an s - protocol with a doubled control interval , @xmath77 .
equivalently , note that this scheme corresponds to alternating a @xmath0-pulse with the identity after every @xmath3 .
the cycle time becomes @xmath78 . for this amount of time , twice as many pulses would be used by protocols ( a , s ) .
still , in certain cases , the ls - protocol performs better than the a - protocol ( see sect .
v.d ) , which motivates its separate consideration here .
@xmath68 * naive random protocol ( r ) * , fig .
[ fig : scheme](d ) .
random decoupling is no longer cyclic , meaning that the control propagator does _ not _ necessarily effect a closed path ( see also @xcite for a discussion of acyclic deterministic schemes ) .
the simplest random protocol in our setting corresponds to having , at each time @xmath79 , an equal probability of rotating or not the qubit that is , at every @xmath50 the control action has a 50% chance of being a @xmath0-pulse and a 50% chance of being the identity . in order not to single out the first control slot , it is convenient to explicitly allow the value @xmath80 ( equivalently , to consider a fictitious pulse @xmath81 in the a- , s- and ls - protocols ) . for pure phase errors
as considered , such a protocol may be interpreted as a simplified parec scheme @xcite . while we will mostly focus on this naive choice in our discussion here
, several variants of this protocol may be interesting in principle , including unbalanced pulse probabilities and/or correlations between control operations .
@xmath68 * hybrid protocol ( h ) * , fig .
[ fig : scheme](e ) .
interesting control scenarios arise by combining deterministic and random design .
the simplest option , which we call `` hybrid '' protocol here , consists of alternating , after every @xmath3 , a @xmath0-pulse with a random pulse , instead of the identity as in the ls - protocol . for our system , in the embedded decoupling language of @xcite , this may be thought of as nesting the a- and r - protocols . in group - theoretic terms
, the h - protocol may be understood as _
randomization over cycles _ @xcite .
a complete asymmetric cycle may be constructed in two ways , say @xmath82 and @xmath83 .
cycle @xmath82 corresponds to traversing @xmath2 in the order @xmath84 that is , free evolution for @xmath3 ; first pulse ; free evolution for @xmath75 ; second pulse the cycle being completed right after the second pulse .
cycle @xmath83 corresponds to the reverse group path , @xmath85.thus , we have : pulse ; free evolution for @xmath3 ; second pulse ; and another free evolution for @xmath3 the system should be observed at this moment before any other pulse .
the h - protocol consists of uniformly picking at random one of the two cycles at every instant @xmath86 , where @xmath87
a single - qubit evolving according to unitary dynamics @xmath88 in eq .
( [ drift ] ) ) provides a pedagogical yet illustrative setting to study dynamical control .
since the goal here is to refocus the underlying phase evolution , the analysis of this system provides a transparent picture for the differences associated with deterministic and random pulses .
it also simplifies the comprehension of the results for the more interesting case of a single - qubit interacting with a decohering semiclassical or quantum environment , where the control purpose becomes twofold : phase refocusing and decoherence suppression .
we begin by considering the standard case of a time - independent target dynamics , @xmath89 for all @xmath90 .
for all the control protocols illustrated above , the system evolves freely between pulses , with the propagator @xmath91 whereas , during a pulse , it is only affected by the control hamiltonian .
the propagator for an instantaneous pulse applied at time @xmath92 will be indicated by @xmath93 .
let @xmath94 denote the qubit density operator in the computational basis @xmath95 , with @xmath96 and @xmath97 . the relevant phase information is contained in the off - diagonal matrix element @xmath98 .
if @xmath99 is the initial qubit state , the time evolution after @xmath100 control intervals under either deterministic or randomized protocols , @xmath101 is dictated by a propagator of the form @xmath102\ , du \right\ } = p_m u_0(t_m , t_{m-1})p_{m-1 } u_0(t_{m-1},t_{m-2 } ) \ldots
p_1 u_0(t_1,t_0)p_0 \nonumber \\ & & = \underbrace{(p_m p_{m-1 } \ldots p_1 p_0 ) } ( p_{m-1 } \ldots p_1 p_0)^{\dagger } u_0(t_m , t_{m-1 } ) ( p_{m-1 } \ldots p_1 p_0)\ldots u_0(t_2,t_1)(p_1 p_0 ) p_0^{\dagger } u_0(t_1,t_0 ) p_0\ : , \label{u_single } \\ & & \hspace{1.5 cm } u_c(t_m ) \nonumber\end{aligned}\ ] ] where @xmath103 indicates , as usual , time ordering .
recall the basic idea of deterministic phase refocusing .
for the a - protocol , @xmath81 and @xmath104 , @xmath105 [ see appendix a ] .
exact averaging is then ensured after a single control cycle , thanks to the property @xmath106 thus , the total phase that the qubit would accumulate in the absence of control is fully compensated , provided that @xmath107 complete cycles are effected ( that is , an even number @xmath108 of spin flips is applied ) .
the overall evolution implements a stroboscopic no - op gate , @xmath109 , @xmath110 , as desired @xcite .
notice that the identity operator is also recovered with the s- and ls - protocols after their corresponding cycle is completed . in preparation to the randomized protocols ( r , h ) , it is instructive to look at the system dynamics in a different frame .
in particular , a formulation which is inspired by nmr @xcite is the so - called _ toggling - frame _ or _ logical - frame _ picture , which corresponds to a time - dependent interaction representation with respect to the applied control hamiltonian .
let @xmath111 denote the control propagator associated to @xmath40 .
then the transformed state is defined as @xmath112 with tilde indicating henceforth logical - frame quantities . at the initial time
@xmath113 , the two frames coincide and @xmath114 . the evolution operator in the logical frame is immediately obtained from eqs .
( [ rho_single ] ) and ( [ rho_logical ] ) , @xmath115 with @xmath116\ , du \right\ } \:.\ ] ] that is , the control field is _ explicitly removed _ from the effective logical hamiltonian .
because , for bb multipulse control , @xmath117 the expression for the logical frame propagator may simply be read off eq .
( [ u_single ] ) , yielding @xmath118 in terms of the composite rotations @xmath119 for cyclic protocols , @xmath120 ( @xmath100 even ) that is , the logical and physical frames overlap stroboscopically in time .
thus , @xmath121 and _ phase refocusing in the logical frame is equivalent to phase refocusing in the physical frame . _ now consider the evolution under the randomized protocols .
the first pulse occurs at @xmath113 , so after a time interval @xmath122 has elapsed , @xmath123 pulses have been applied . since the final goal is to compare random with cyclic controls
, we shall take @xmath100 _ even _ henceforth . at time @xmath124
, population inversion may have happened in general in the physical frame .
this makes it both convenient and natural to consider the logical frame , where inversion does _ not _ happen , as the _ primary _ frame for control design .
the evolution operator in this frame may be expressed , using eq .
( [ u_logical ] ) , as @xmath125 where @xmath126 is a bernoulli random variable which accounts for the history of spin flips up to @xmath50 in a given realization . for each @xmath127 , if a spin flip occurs at time @xmath128 , then @xmath129 and @xmath130 , otherwise @xmath131 and @xmath132 .
equivalently , @xmath133 will take the values @xmath134 or @xmath135 with equal probability , depending on whether the composite pulse @xmath136 is the identity or a @xmath0-pulse .
let @xmath25 be an index labelling different control realizations .
for a fixed @xmath25 , the qubit coherence in the logical frame is given by @xmath137 this expression provides the starting point for analyzing control performance . for the a - protocol , the only possible realization has @xmath138 and leads to the trivial result @xmath139 . for the r - protocol ,
realizations corresponding to different strings of @xmath140 s filling up @xmath100 places give , in general , different phases and an ensemble average should be considered . if the statistical ensemble is large enough , the _ average performance _ may be approximated by the _ expected performance _ , which is obtained by averaging over _ all _ possible control realizations and will be denoted by @xmath141 .
the calculation of the expectation value is straightforward in the _ unbiased _ setting considered here . since , for each realization , @xmath142 or @xmath135 independently of the value of its predecessor @xmath143 , notice that , among the @xmath144 realizations existing in the logical frame , there are pairs , say corresponding to labels @xmath25 and @xmath145 , with opposite phases .
they have @xmath146 and @xmath147 , while @xmath148 with @xmath149 .
taking such a pairing into account , the number of independent realizations to consider is @xmath150 .
we then find @xmath151 \ , , \label{e_log_cos}\ ] ] where @xmath152 and @xmath153 .
such a pairing may be iterated .
suppose we next consider realizations with @xmath154 and @xmath155 , while @xmath148 with @xmath156 .
this gives @xmath157 \,,\ ] ] where @xmath158 and @xmath159 , and so on .
the successive pairing of realizations finally leads to the following simple expression the following expression is found : @xmath160^m \ , .
\label{e_logical}\ ] ] several remarks are in order . under random pulses , the phase accumulated during the interval @xmath161 is , on average , completely removed , regardless of the @xmath3 value . an important distinction with respect to the deterministic controls , however , is that now the different phase factors carried by each stochastic evolution may interfere among themselves , causing the ensemble average to introduce an effective _ phase damping_. in general , let us write the ensemble expectation in the form @xmath162 for real functions @xmath163 @xcite . here ,
@xmath164 , whereas @xmath165 $ ] .
complete dephasing occurs when @xmath166 , with @xmath167 odd , while for @xmath168 , with @xmath169 , @xmath170 . whenever exact knowledge of the frequency @xmath30 and precise control over the time interval @xmath3 are available , the r - protocol can be made to achieve exact averaging , like the a - protocol , under the additional _ synchronization condition _ that @xmath171 in situations where such a synchronization is not easily accessible
, one may still look for a general condition under which the r - protocol avoids ensemble dephasing . taking a taylor expansion of eq .
( [ e_logical ] ) yields @xmath172 in principle , this requirement may be fulfilled by making @xmath173 and/or @xmath3 sufficiently small .
interestingly , the condition of eq .
( [ first_scale ] ) is directly related to the bound obtained in theorem 1 of @xcite for the worst - case pure - state error probability , defined by @xmath174 where the latter term is the usual input - output state fidelity @xcite . in the limit where @xmath175 , where @xmath176 , @xmath177 , theorem 1 implies @xmath178 on the other hand , using eq .
( [ e_logical ] ) we obtain @xmath179 , the above expression gives @xmath180 which makes the connection with eq .
( [ first_scale ] ) manifest .
it remains to discuss the performance of the h - protocol .
the freedom of not always effecting a spin flip after every @xmath3 , which is one of the appealing features of the r - protocol , is still partially present here . on the other hand ,
since a spin flip does occur at every @xmath128 with @xmath181 odd , which leads to @xmath182 for @xmath53 odd , any realization of this protocol completely refocuses the qubit [ see eq .
( [ singlereal ] ) ] , so @xmath164 and @xmath183 . accordingly , _ in the logical frame , the h - protocol is optimal _ , combining the absence of phase damping of cyclic schemes with the flexibility of random pulses . in practice
, we deal with the average performance of a statistical ensemble of size @xmath184 .
to evaluate the sample size that guarantees a desired margin of error @xmath185 @xcite , we invoke the central limit theorem . because different realizations are independent , the latter ensures that the average performance is distributed normally with a mean value equal to the expected performance and standard deviation given by @xmath186 , where @xmath187 is the standard deviation for all realizations .
thus , if we want , with probability @xmath188 , that the average performance differs from the expected performance by no more than @xmath185 , the sample size must be at least as large as @xmath189 where @xmath190 is the value of the standard normal variable which has a probability @xmath191 of being exceeded . taking a taylor expansion of eqs .
( [ singlereal ] ) and ( [ e_logical ] ) , we can show that @xmath192 thus , the number of realizations required to ensure a specified degree of precision decreases as @xmath3 .
it is interesting to observe that the ensemble average may be interpreted as effecting a quantum operation , @xmath193 with @xmath194 and random coefficients @xmath195 which may be derived from eq .
( [ u_logical_chi ] ) . finally , it is important to compare the average coherence element in the logical and physical frames .
dephasing is a more delicate issue in the schrdinger picture , because spin population is not necessarily conserved and @xmath196 may be related to @xmath197 or to @xmath198 , depending on how many @xmath0-pulses occur .
if , after an interval @xmath122 , an even number of spin - flips have happened , we recover @xmath199 as in the cyclic case , but an odd number of flips leads instead to @xmath200 . by recalling eq .
( [ correspondence ] ) , for randomized schemes we find @xmath201 where @xmath202 for the h - protocol .
thus , the agreement between the expected results in the two frames depends on the initial qubit state .
results are identical if @xmath197 is real , but differ otherwise .
the worst scenario occurs if @xmath197 is purely imaginary , as the average in the physical frame vanishes .
this reflects the fact that the net evolution may be represented by a quantum operation which flips the state of the qubit with 50% probability , and leaves it alone otherwise . clearly ,
knowledge of the control history allows the system to be deterministically returned in the physical frame for any realization , if desired .
that is , having a classical register that records the total number of spin flips may be used to select realizations that guarantee a good performance of random pulses also in the physical frame for any initial state .
for example , if only realizations with an even number of spin - flips are selected , the results in both frames are equal , @xmath203 , as desired . _ to summarize : _ in the logical frame , refocusing the unwanted phase evolution is possible with any of the protocols we considered .
the r - protocol , however , introduces an average ensemble dephasing , which may only be prevented by precisely tuning @xmath204 , with @xmath169 , or by assuring that @xmath205 $ ] .
this implies the appearance of a time scale requirement which is not present when dealing with deterministic controls , nor with the h - protocol . in the physical frame ,
state - independent conclusions regarding the system behavior may be drawn conditionally to specific subsets of control realizations .
overall , the h - protocol emerges as an alternative of intermediate performance , which partially combines advantages from determinism and randomness .
we now consider the more interesting case where the qubit frequency is time dependent , @xmath206 , @xmath207 being a deterministic ( but potentially unknown ) function .
this could result , for example , from uncontrolled drifts in the experimental apparatus .
while all protocols become essentially equivalent in the limit @xmath208 , searching for the best protocol becomes meaningful in practical situations where pulsing rates are necessarily finite . under these conditions ,
the deterministic protocols described so far will no longer be able , in general , to completely refocus the qubit .
this would require a very specific sequence of spin flips for each particular function @xmath209 , which would be hard to construct under limited knowledge about the latter .
on the other hand , the average over random realizations does remove the phase accumulated for _ any _ function @xmath209 , making randomized protocols ideal choices for phase refocusing . as a drawback ,
however , ensemble dephasing may be introduced .
thus , the selection of a given protocol will be ultimately dictated by the resulting tradeoffs .
the propagator in the logical frame now reads @xmath210 \:,\ ] ] which reduces to eq .
( [ u_logical_chi ] ) when @xmath211 .
some assumptions on both the amplitude and frequency behavior of @xmath212 are needed in order to draw some general qualitative conclusions .
first , if @xmath213 , the analysis developed in the previous section will still approximately hold . in the spirit of regarding @xmath30 as a central frequency
, we will also discard the limit @xmath214 , and restrict our analysis to cases where @xmath215 . if @xmath212 is dominated by frequency components which are very fast compared to @xmath216 , the effect of @xmath212 may effectively self - average out over a time interval of the order or longer than @xmath217 . in the opposite limit , where the time dependence of @xmath212 is significantly slower than @xmath30 ,
deterministic controls are expected to be most efficient in refocusing the qubit , improving steadily as @xmath3 decreases . in intermediate situations , however , the deterministic performance may become unexpectedly poor for certain , in principle , unknown values of @xmath3 .
these features may be illustrated with a simple periodic dependence .
suppose , for example , that @xmath218 , and @xmath219 .
for a fixed time interval @xmath220 , a significant reduction of the accumulated phase is already possible with few deterministic pulses .
however , care must be taken to avoid unintended `` resonances '' between the natural and the induced sign change .
for the a - protocol , this effect is worst at @xmath221 , in which case the control pulses exactly occur at the moment the function changes sign itself , hence precluding any cancellation of @xmath212 . with the r - protocol ,
ensemble dephasing becomes the downside to face .
the ensemble average now becomes @xmath222 \bigg\}\,.\ ] ] in the absence of time dependence , phase damping is minimized as long as eq .
( [ first_scale ] ) holds . under the above assumptions on @xmath212 , the condition remains essentially unchanged , in agreement with the fact that the accuracy of random averaging only depends on @xmath223 @xcite .
refocusing is also totally achieved with the h - protocol . however , unlike in the case of the r - protocol , the ensemble average no longer depends on the time independent part of the hamiltonian , but only on the function @xmath212 , making the identification of precise requirements on @xmath3 harder in the absence of detailed information on the latter .
we have @xmath224 \bigg\}\,.\ ] ] fig .
[ fig : fixedm ] illustrates the points discussed so far .
the sinusoidal example is considered , and we contrast the two aspects to be examined : the top panels show the phase magnitude @xmath225 , which is optimally eliminated with random pulses , while the bottom ones give the dephasing rate @xmath226 , which is inexistent for deterministic controls .
the interval between pulses is fixed , @xmath227 , and the protocols are compared for two arbitrary , but relatively close values of the oscillation frequency rate : @xmath228 and @xmath229 .
the deterministic control is very sensitive to slight changes of the drift and at certain instants may behave worse than if pulses were completely avoided . similarly , the h - protocol , even though more effective than the r - protocol in this example , also suffers from uncertainties related to @xmath212 . on the contrary ,
deviations in the performance of the r - protocol are practically unnoticeable , making it _ more robust against variations in the system parameters_. , and under the a- [ ( blue ) stars ] , r- [ ( black ) circles ] , and h- [ ( purple ) plus ] protocols in the logical frame , for @xmath230 and @xmath227 .
left panels : @xmath228 ; right panels : @xmath229 .
average taken over @xmath231 realizations . in this and all simulations that will follow , we set @xmath232 . ] as a further illustrative example , we consider in fig .
[ fig : w_sin ] the following time dependence for the qubit : @xmath233 d(t).\ ] ] the left panels have , as before , @xmath234 , while for the right panels @xmath235 a fixed time @xmath236 is now divided into an increasing number @xmath100 of intervals @xmath3 . here , selecting the most appropriate protocol depends on our priorities concerning refocusing and preservation of coherence .
we may , however , as the right upper panel indicates , encounter _ adversarial _ situations where the time dependence of the qubit frequency is such that not acting on the system is comparatively better than using the a - protocol .
clearly , depending on the underlying time dependence and the pulse separation , such poor performances are also expected to occur with other deterministic protocols .
in addition , notice that , consistent with its hybrid nature , the h - protocol may perform worse for values of @xmath3 where the deterministic control becomes inefficient ( compare right upper and lower panels ) . in similar situations , from the point of view of its enhanced stability , _
the r - protocol turns out to be the method of choice_. , and under the a- [ ( blue ) stars ] , r- [ ( black ) circles ] , and h- [ ( purple ) plus ] protocols in the logical frame . the time interval considered is @xmath236 ; @xmath237 ; and @xmath238 , where @xmath239 .
the drift in the right panels includes @xmath240 .
average taken over all possible realizations . ]
_ to summarize : _ an isolated qubit with time - dependent parameters provides the simplest setting where advantages of randomization begin to be apparent , in terms of enhanced stability against parameter variations . on average ,
phase is fully compensated , and ensemble dephasing may be kept very small for sufficiently fast control .
similar features will appear for a qubit interacting with a time - varying classical or quantum environment , as we shall see in sects .
iv.c and v.e .
qubit coherence is limited by the unavoidable influence of noise sources . within a semiclassical treatment , which provides an accurate description of decoherence dynamics whenever
back - action effects from the system into the environment can be neglected , noise is modeled in terms of a classical stochastic process , effectively resulting in randomly time - dependent systems . typically , external noise sources , which in a fully quantum description are well modeled by a _ continuum _ of harmonic modes ( see sect .
v ) , are represented by a gaussian process . here , we focus on _ localized _ noise sources , which may be intrinsic to the physical device realizing the qubit notably , localized traps or background charges , leading to a quantum _ discrete _ environment . in this case , non - gaussian features become important , and are more accurately represented in terms of noise resulting from a single or a collection of classical bistable fluctuators leading to so - called random telegraph noise ( rtn ) or @xmath241-noise , respectively . beside being widely encountered in a variety of different physical phenomena @xcite ,
such noise mechanisms play a dominant role in superconducting josephson - junction based implementations of quantum computers @xcite .
recently , it has been shown that rtn and @xmath241-noise may be significantly reduced by applying cyclic sequences of bb pulses @xcite .
we now extend the analysis to randomized control . as it turns out , random decoupling is indeed viable and sometimes more stable than purely deterministic protocols . while a detailed analysis of randomized control of genuine @xmath241 noise would be interesting on its own , we begin here with the case of a single fluctuator .
this provides an accurate approximation for mesoscopic devices where noise is dominated by a few fluctuators spatially close to the system @xcite .
let the time - dependent hamiltonian describing the noisy qubit be given by eqs .
( [ drift0])-([drift ] ) , where @xmath36 and @xmath242 characterizes the stochastic process , randomly switching between two values @xmath243 , @xmath244 .
we shall in fact consider a _
semi_random telegraph noise , that is , we assume that the fluctuator initial state is always @xmath245 .
the switching rate from @xmath243 to @xmath246 is denoted by @xmath247 , with @xmath248 .
we shall also assume for simplicity that @xmath249 , corresponding to a symmetrical process .
the number of switching events @xmath250 in a given time interval @xmath90 is poisson - distributed as @xmath251 semiclassically , dephasing results from the ensemble average over different noise realizations .
this leads to the decay of the average of the coherence element , @xmath252 here , the average over rtn realizations is represented by @xmath253 and should be distinguished from the average over control realizations , which , as before , is denoted by @xmath254 .
the dephasing factor @xmath255 and the phase @xmath256 have distinctive properties depending on the ratio @xmath257 , where @xmath258 ( @xmath259 ) corresponds to a fast ( slow ) fluctuator . given the initial condition for the fluctuator @xmath260 , where @xmath261 stands for the fluctuator
initially in state @xmath262 , @xmath263 may be calculated as @xcite , @xmath264 where @xmath265 and @xmath266 is the equilibrium population difference .
note that for a symmetrical telegraph process , the only difference between the results for a fluctuator initially in state @xmath245 or @xmath267 is a sign in the above phase @xmath185 .
the decoherence rate for a slow fluctuator is much more significant than for a fast fluctuator .
this has been discussed in detail elsewhere @xcite , and has been reproduced for later comparison with the controlled case in fig .
[ fig : tele_nopulse ] , where several values of @xmath268 are considered .
a fast fluctuator behaves equivalently to an appropriate environment of harmonic oscillators and noise effects are smaller for smaller values of @xmath268 , whereas for a slow fluctuator the decoherence function saturates and becomes @xmath269 . from a symmetrical bistable fluctuator .
several values of @xmath270 are considered , resulting from changing the coupling strength @xmath271 at fixed switching rate @xmath272 a.u.,width=230 ] here we compare the reduction of rtn under the action of the a- , h- , and r - protocols . in order to isolate the effects of the noise ,
it is convenient to first carry out the analysis in the interaction picture which removes the free dynamics @xmath273 .
the density operator becomes @xmath274 where @xmath275 and the superscript @xmath276 will refer to the interaction picture henceforth .
the free propagator between pulses is now @xmath277 with @xmath278 ; while at @xmath50 , we have [ see appendix a ] @xmath279 \exp \left[-i\lambda _ j \frac{\pi}{2 } \sigma _ x\right ] \exp \left [ -i\frac{\omega_0 t_j}{2}\sigma _ z \right]\ : . \label{pclas_ip}\ ] ] a second canonical transformation into the logical frame is also considered , so that ( as before ) realizations with an even or an odd number of total spin flips are treated on an equal footing .
we will refer to the combination of the two transformations as the logical - ip frame .
similarly to eq .
( [ correspondence ] ) , the interaction and the logical - ip frame propagators are related as @xmath280 this leads to the following propagators at @xmath173 , @xmath281 where @xmath282 our goal is to compute the ratio @xmath283 where @xmath25 labels , as before , different control realizations .
note that interchanging the order of the averages does _ not _ modify the results if all pulse realizations are considered and the number of rtn realizations is large enough . with @xmath284 switch realizations
no significant variations were found by interchanging the averages .
the decoherence rate @xmath285 for the three selected protocols is shown in fig .
[ fig : tele_dephas ] , where a time @xmath286 was fixed and divided into an increasing number @xmath100 of intervals @xmath3 .
the left panels are obtained for three slow fluctuators , @xmath287 , and the right panels for @xmath288 .
these are the six different noise regimes considered in ref .
@xcite , where the a - protocol was studied .
the authors concluded that once @xmath289 , @xmath290 scales with @xmath291 , while for @xmath292 , bb pulses are still capable of partially reducing noise due to a fast fluctuator , but are mostly inefficient against slow fluctuators .
here , we verified that among all possible realizations of pulses separated by the same interval @xmath3 , the realization corresponding to the a - protocol yields the largest value of @xmath285 , whereas absence of pulses gives , as expected , the smallest value .
this justifies why , in terms of average performance for finite @xmath3 , we have in decreasing order : a- , h- , and r - protocol ; while for @xmath293 , different protocols are expected to become equivalent . .
( green ) solid lines represent the analytical results from @xcite in the absence of control .
( blue ) stars : a - protocol ; ( black ) circles : r - protocol ; and ( purple ) plus : h - protocol .
averages are taken over @xmath284 rtn realizations and all possible @xmath294 pulse realizations .
the time interval considered is @xmath286 , thus @xmath295 .
, width=326 ] .
( green ) solid line : analytical results in the absence of control pulses ; ( blue ) stars : a - protocol ; both r- and h - protocols have phase equal to zero .
the time interval considered is @xmath286 , so @xmath296 .
averages computed as in fig .
[ fig : tele_dephas ] .
, width=326,height=288 ] in terms of refocusing the unwanted phase evolution , the above randomized protocols are optimal , since @xmath297 , while the phase magnitude @xmath298 for the a - protocol is eventually compensated as @xmath100 increases .
this is shown in fig .
[ fig : tele_phase ] . notice also that the absolute phase is very small for fast fluctuators . instead of fixing a time @xmath299 , an alternative picture of the performance of different protocols
may also be obtained by fixing the number of intervals @xmath100 , as in figs .
[ fig : fixed ] ( left panels ) . as expected , larger @xmath100 leads to coherence preservation for longer times .
still another option is to fix the interval between pulses @xmath3 , as in fig .
[ fig : fixed ] ( right panel ) .
as before , the a - protocol shows the best performance , followed by the h- and r - protocols . .
left panels : @xmath300 ( top ) , @xmath301 ( bottom ) .
right panel : @xmath302 .
( green ) solid line : analytical results in the absence of control .
( blue ) stars : a - protocol ; ( black ) circles : r - protocol ; and ( purple ) plus : h - protocol .
averages are taken over @xmath303 rtn realizations and @xmath231 pulse realizations.,width=326,height=187 ] if the interaction picture is not taken into account , complete refocusing is again guaranteed , on average , when either the r- or h - protocols are used .
however , for the r - protocol , the qubit frequency plays a delicate role in the resulting dephasing process .
we now have @xmath304 which may be further simplified as follows . among the @xmath144 pulse realizations existing in the logical frame ,
there are pairs , say corresponding to labels @xmath25 and @xmath145 , where @xmath146 and @xmath147 , while @xmath148 with @xmath149 , which leads to @xmath305 . besides , since we are considering a semirandom telegraph noise , a pulse at @xmath113 is equivalent to switching the fluctuator from the initial state @xmath245 to @xmath267 , whose net effect is simply a change in the sign of the phase @xmath306 .
therefore , we may write @xmath307 e^{-\gamma^{(k)}(t_m , t_0 ) } \right\ } \ : , \nonumber\end{aligned}\ ] ] where @xmath152 and @xmath153 . in the physical frame ,
we find correspondingly @xmath308 e^{-\gamma^{(k)}(t_m , t_0 ) } \right\}\ : .\nonumber\end{aligned}\ ] ] contrary to the result obtained in the absence of noise , eq . ( [ e_logical ] ) , the additional realization - dependent phase shift @xmath309 now remains . while , on average , this phase is removed in the limit where @xmath74 , for finite control rates @xmath309 may destructively interfere with the phase gained from the free evolution , potentially increasing the coherence loss .
identifying specific values of @xmath75 where such harmful interferences may happen for the given rtn process is not possible , which makes the results for the r - protocol with finite @xmath3 unpredictable in this case .
while the above feature is a clear disadvantage , it is avoided by the h - protocol . for each realization ,
the phase accumulated with the free evolution is completely canceled , so the result in the logical - ip frame is equal to that in the logical frame : @xmath310 . if access to a classical register that records the total number of spin flips is also available
, this equivalence between frames may be further extended to the physical frame . additionally , as already found in sect .
iiib , randomized protocols tend to offer superior stability .
let us illustrate the above statement through an example where the noisy dynamics of the system is slightly perturbed .
suppose that , moving back to the interaction picture , the noise process is now @xmath311 where @xmath312 physically , @xmath313 describes a sequence of six instantaneous switches , equally separated by the interval @xmath314 , restarting again at every instant @xmath315 , @xmath25 being an odd number .
this process may be viewed as bursts of switches of duration @xmath316 followed by an interval @xmath317 of dormancy . the resulting behavior for @xmath318 in the logical - ip frame
is depicted in fig .
[ fig : tele_devil ] . subjected to a disturbance as given in eq .
( [ dist ] ) . left panels : fixed @xmath319 , so @xmath295 .
right panels : fixed @xmath320 .
( green ) solid line : results in the absence of control pulses ; ( blue ) stars : a - protocol ; ( black ) circles : r - protocol ; and ( purple ) plus : h - protocol .
the average phase for both r- and h - protocols is zero .
left panels : averages are taken over @xmath284 rtn realizations and all possible @xmath294 pulse realizations .
right panels : averages are taken over @xmath303 rtn realizations and @xmath231 pulse realizations.,width=326 ] with deterministic control , the rate of noise suppression quickly improves as the separation between pulses shrinks ( left upper panel ) , until a certain value , @xmath321 , where it suddenly shows a significant recoil , becoming almost as bad as simply not acting on the system at all .
equivalently , by fixing @xmath322 , the performance of the a - protocol becomes very poor for @xmath323 ( right upper panel ) . in practice , detailed knowledge of the system dynamics
might be unavailable , making it impossible to predict which values of @xmath3 might be adverse .
randomized schemes , on the other hand , are by their own nature more stable against such interferences . as seen from the figure , the r - protocol shows a slower , but also more consistent improvement as @xmath3 decreases , and might therefore be safer in such conditions . notice also that , in terms of coherence preservation and stability , the h - protocol shows ( as intuitively expected ) an intermediate performance between the a- and r - protocols .
_ to summarize : _ in the logical - ip frame , the effects of the rtn can be reduced not only under deterministic pulses , but also with a randomized control , though a comparatively shorter pulse separation is needed in the latter case . in the logical and physical frames , the r - protocol , besides showing the poorest performance among the three considered schemes , may also lead to dangerous interferences between the qubit frequency and the phase gained from the free evolution . such problem , however , does not exist for the h - protocol .
the benefits of randomization are most clear when limited knowledge about the system dynamics is available and deterministic control sequences may be inefficient in avoiding unwanted `` resonances '' . combining protocols , where we gain stability from randomness , but also avoid the free phase evolution , is desirable especially when working in the physical frame . in this sense ,
the h - protocol emerges as a promising compromise .
we now analyze the case of a genuine quantum reservoir , where decoherence arises from the entanglement between the qubit and the environment .
the relevant hamiltonian is given by eqs .
( [ drift0])-([drift ] ) with @xmath324 and @xmath35 . in the semiclassical limit
, the effects of the interaction with the bosonic degrees of freedom may be interpreted in terms of an external noise source whose fluctuations correspond to a gaussian random process .
a detailed analysis of deterministic decoherence suppression for this model was carried out in @xcite ( see also @xcite for an early treatment of the driven spin - boson model in a nonresonant monochromatic field and @xcite for related discussions of dynamically modified relaxation rates ) . here ,
we discuss how randomized decoupling performs . as in sect .
iv , we first focus on understanding the controlled dynamics in a frame that explicitly removes both the control field and the free evolution due to @xmath325 . let us recall some known results related to the uncontrolled dynamics .
we have @xcite @xmath326\right\}\ : , \label{u_ip}\ ] ] where @xmath327 under the standard assumptions that the qubit and the environment are initially uncorrelated , @xmath328 and that the environment is in thermal equilibrium at temperature @xmath329 [ the boltzmann constant is set = 1 ] @xmath330 the trace over the environment degrees of freedom may be performed analytically , leading to the following expression for the qubit coherence : @xmath331\ } \nonumber \\ & & = \rho_{01}(t_0)\exp [ -\gamma ( t , t_0 ) ] \ : .
\label{decay}\end{aligned}\ ] ] here , @xmath332 is the harmonic displacement operator of the @xmath25th bath mode , and the decoherence function @xmath255 is explicitly given by @xmath333 in the continuum limit , substituting @xmath334 by the spectral density @xmath335 , one finds @xmath336}{\omega^2 } \ : .
\label{g_nopulse}\ ] ] for frequencies less than an ultraviolet cutoff @xmath337 , @xmath335 may be assumed to have a power - law behavior , @xmath338 the parameter @xmath339 quantifies the overall system - bath interaction strength and @xmath340 classifies different environment behaviors : @xmath341 corresponds to the ohmic case , @xmath342 to the super - ohmic and @xmath343 to the sub - ohmic case .
remarkably , the dynamics remains exactly solvable in the presence of randomized bb kicks .
we focus first on the r - protocol viewed in the logical - ip frame . between pulses
the evolution is characterized by eq .
( [ u_ip ] ) , while at @xmath50 eq .
( [ pclas_ip ] ) applies . using eq .
( [ relation ] ) , the propagator in the logical - ip frame , apart from an irrelevant overall phase factor , may be finally written as @xmath344\hspace*{-.5 mm } \right\},\ ] ] where @xmath345 under the uncorrelated initial conditions specified above and thermal equilibrium conditions , the qubit reduced density matrix is exactly computed as @xmath346\big\ } \nonumber \\ & & = \rho_{01}(t_0 ) e^{-\gamma_{r } ( t_m , t_0)}\ : . \label{decoherence}\end{aligned}\ ] ] because @xmath133 in eq .
( [ func ] ) can be @xmath347 at random , each element in the sum corresponds to a vector in the complex plane with a different orientation at every step @xmath3 .
thus , the displacement operator above may be suggestively interpreted as a random walk in the complex plane .
the decoherence function @xmath348 is now given by @xmath349 which , in the continuum limit , becomes @xmath350 \ : .
\label{main_result}\end{aligned}\ ] ] the decoherence behavior under the a - protocol is obtained by letting @xmath351 .
we then recover the result of deterministically controlled decoherence @xcite , which may be further simplified as @xcite , @xcite , @xmath352}{\omega^2 } \tan ^2 \left ( \frac{\omega \delta t}{2 } \right)\ : .
\label{determ}\end{aligned}\ ] ] before proceeding with a numerical comparison between eqs .
( [ main_result])-([determ ] ) , some insight may be gained from an analytical lower bound for the average @xmath353)$ ] . according to jensen s inequality , @xmath354 for any convex function @xmath1 . using this and the fact that @xmath355
we have the following lower bound @xmath356 \big ) \geq \exp [ - \mathbb{e } \big ( \gamma_{r } ( t_m , t_0 ) \big ) ] \:,\hspace*{10 mm } \nonumber \\
\mathbb{e } \big ( \gamma_{r } ( t_m , t_0 ) \big ) = 4 m \int_0^{\infty } d\omega i(\omega ) \coth \left ( \frac{\omega}{2t}\right ) \frac{1-\cos ( \omega \delta t)}{\omega^2 } \ : .
\end{aligned}\ ] ] in fig .
[ fig : bound ] we compare the coherence decay corresponding to the absence of control ( [ g_nopulse ] ) , to the a - protocol ( [ determ ] ) and to the lower bound ( [ lower_bound ] ) .
two limiting cases of high and low temperature , @xmath357 and @xmath358 , are considered .
the high temperature limit corresponds to an effectively classical bath , where the properties of the environment are dominated by thermal fluctuations . in the absence of control
, decoherence is very fast on the time scale determined by the bath correlation time @xmath359 , hence coherence preservation requires very short intervals between pulses .
the a - protocol shows the best performance .
the actual randomized performance may , however , be significantly better than the lower bound in this temperature regime , though they never surpass the deterministic case [ see next subsection ] . in the case of low temperature , or fully quantum bath , decoherence is much slower and a richer interplay between thermal and vacuum fluctuations occurs .
larger values of @xmath3 may then be analyzed before total coherence loss takes place .
the interesting phenomenon of decoherence acceleration @xcite , which may happen when @xmath360 , may now be observed . for short @xmath3 ,
the a - protocol is again more efficient , though not significantly better than the lower bound . for large @xmath3 , pulses induce destructive interference and the a - protocol performs even worse than the lower bound .
in such situation the best option is simply not to act on the system . , @xmath361 , @xmath362 .
left panel : @xmath363 and @xmath364 .
right panels : @xmath365 , top : @xmath364 and bottom : @xmath366 .
( green ) solid line : no control ; ( blue ) stars : a - protocol ; ( red ) dashed line : lower bound , eq .
( [ lower_bound ] ) .
, width=331 ] when @xmath367 , some general insight may be gained by comparing appropriate limits of the lower bound and the deterministic decoherence function .
first , by taylor - expanding up to second - order in @xmath3 we have @xmath368 whereas @xmath369 \ } \nonumber \:.\end{aligned}\ ] ] therefore , in the limit of very short @xmath3 , the lower bound approaches the ideal situation of total suppression of decoherence linearly in @xmath3 , while for the a - protocol this occurs quadratically .
this analysis may be further extended by studying the two limits of eq .
( [ new_lower_bound ] ) with respect to temperature . considering the spectral density of eq .
( [ spectral ] ) , we have @xmath370 thus , a sufficient condition under which random control avoids decoherence is @xmath371 this should be compared with the general bound given in theorem 2 of @xcite .
we will see in sect .
v.d that , in the physical frame , the qubit frequency @xmath30 also plays an important role . a similar analysis for the a - protocol may be effected using eq .
( [ new_determ ] ) . for @xmath372 ,
the decoherence function decays quadratically in time , giving @xmath373 while for @xmath374 the decoherence function becomes independent of @xmath122 , and we get @xmath375 these should be compared with the error bound of theorem 3 in @xcite . based on these considerations , random pulses
may hardly be expected to outperform deterministic controls in the limits discussed above .
still , it remains interesting to quantitatively see what the actual performance is for intermediate @xmath3 and/or hybrid schemes , for instance with respect to acceleration .
moreover , further changes may be expected when some time dependence exists in the system parameters , for instance in the coupling strength to the environment [ see sect .
v.e ] . based on the exact result of eq .
( [ main_result ] ) , we now present a comparison of the average decoherence suppression achievable by the protocols described in sect .
a fixed time @xmath299 divided into an increasing number @xmath100 of intervals @xmath3 is considered .
[ fig : high ] compares the average @xmath376)$ ] in the limit of high temperature , @xmath377 , for a system evolving under the a- and r - protocols . for the fixed times chosen ,
@xmath378 ( upper panel ) and @xmath379 ( lower panel ) , the coherence element has already practically disappeared and can not be seen in the figure , while the a - protocol is able to recover it even for very few cycles .
the values of @xmath380 $ ] for different realizations are widely spread between the worst case corresponding to all @xmath140s=0 and the efficient realizations involving several spin flips . as a consequence
, the average converges to 1 slowly and has a large standard deviation @xcite .
notice , however , that it is significantly better than the lower bound . .
upper panel : @xmath378 ; lower panel : @xmath379 .
( blue ) stars : a - protocol ; ( black ) circles : average over @xmath231 realizations and respective standard deviations for the r - protocol @xcite ; ( black ) squares : expectation value ( taken over all @xmath294 realizations ) for the r - protocol ; ( red ) dashed line : lower bound .
, width=268 ] the results in the low - temperature limit , @xmath365 , are shown in fig .
[ fig : low ] . here , thanks to the fact that decoherence is overall slower , longer evolution times may be chosen : @xmath381 ( upper panel ) and @xmath382 ( lower panel ) . for the latter choice , in particular ,
when @xmath383 , decoherence enhancement occurs and , interestingly , the results for the a - protocol are _ worse _ than those for the r - protocol .
however , it takes a much smaller @xmath3 for the r - pulses to finally cross the line that separates enhancement from decoherence reduction .
notice also that the values of @xmath384 $ ] for different realizations are not so spread and the standard deviations are narrower than in the high temperature limit .
in addition , the average over realizations is very close to the lower bound , to the point that they can not be distinguished in the figure . .
upper panel : @xmath379 ; lower panel : @xmath382 .
( green ) solid line : no control ; ( blue ) stars : a - protocol ; ( black ) circles : average over @xmath231 realizations and respective standard deviations for the r - protocol ; ( black ) squares : expectation value for the r - protocol.,width=268 ] we now extend our comparison to the three remaining protocols of fig . [
fig : scheme ] , see fig .
[ fig : sym ] .
we choose a high temperature bath with @xmath379 ( upper panel ) [ low temperature , in this case , leads to similar results ] , and a low temperature bath with @xmath382 ( lower panel ) .
the s - protocol shows the best performance , which is evident in the upper panel , but hardly perceptible in the lower one . due to the different rearrangement of the time interval between pulses for this protocol
, it does not correspond to any of the @xmath294 realizations of random pulses as considered here and represent a special scheme separated from the others .
the performance of the ls - protocol , which has half the number of @xmath0-pulses used in the a - protocol , turns out to be better in all cases of a high temperature bath , but worse for a fully quantum bath with large @xmath385 .
this explains why the h - protocol , which combines symmetrization and randomness , also performs better than the a- in the high temperature limit . , and @xmath381 .
lower panel : low temperature ohmic bath , @xmath365 , and fixed time @xmath382 .
( green ) solid line : no control ; ( blue ) stars : a - protocol ; ( purple ) plus : h - protocol ; ( orange ) diamonds : ls - scheme ; and ( black ) dot - dashed line : s - protocol .
average for the h - protocol is taken over @xmath231 realizations .
, width=268 ] _ to summarize : _ in terms of performance , we have in decreasing order : s- , ls- , h- , a- , r - protocols for high temperature ; and s- , a- , h- , ls- , r - protocols for low temperature once the number of pulses are sufficient to start slowing down decoherence .
different protocols become again , as expected , essentially equivalent in the limit @xmath386 . for finite pulse separations , in the considered case of a time - independent hamiltonian ,
it is always possible to identify a deterministic protocol showing the best performance . however ,
if a balance is sought between good performance and protocols minimizing the required number of pulses , then the h - protocol again emerges as an interesting compromise .
note , in particular , that the latter outperforms the standard a - protocol in some parameter regimes .
we now investigate under which conditions decoherence suppression is attainable in the physical frame , when the system is subjected to randomized control . because , in this frame , the qubit natural frequency plays an important role , random decoupling also depends on how small @xmath3 can be made with respect to @xmath216 .
the reduced density matrix is obtained following the same steps described so far , but in order to retain the effects of the system hamiltonian , the transformation into the interaction picture is now done with respect to the environment hamiltonian only hence the superscript @xmath387 . upon tracing over the environment degrees of freedom ,
we are left with the reduced density operator in the schrdinger picture .
the unitary operator between pulses is @xmath388 \bigg\}\ : , \label{u_ipbath } \nonumber\end{aligned}\ ] ] while at @xmath50 it is given by @xmath389\ : .
\label{p_ipbath}\ ] ] by additionally moving to the logical frame we get @xmath390 \\ & & \hspace{0.5 cm } \exp \bigg\ { \frac{\sigma_z}{2 } \otimes \sum_k \left [ b^{\dagger}_k e^{i \omega_k t_0 } \eta ^r_k ( m,\delta t ) - { \rm { \rm h.c . } } \right ] \bigg\ } \ : .
\nonumber\end{aligned}\ ] ] tracing over the environment and taking the expectation over control realizations leads to the coherence element in the logical frame : @xmath391 e^{-\gamma_{r}^{(k ) } ( t_m , t_0)}\:,\ ] ] where @xmath392 is given by eq .
( [ large_chi ] ) .
thus , in addition to the decoherence described as before by eq .
( [ main_result ] ) , we now have ensemble dephasing due to the fact that each realization carries a different phase factor proportional to @xmath30 . the results for the ratios @xmath393 in the logical and logical - ip frames for the system ,
respectively , are summarized in fig .
[ fig : ipbath ] , where @xmath3 is fixed and the system is observed at different times . both a high temperature and a low temperature scenario are considered .
the phase for each realization in the logical frame is mostly irrelevant when @xmath394 .
the outcomes of the average over all realizations in both frames are then comparable , independently of the bath temperature .
the situation changes dramatically when the spin - flip energy becomes large , the worst scenario corresponding to @xmath395 , with @xmath25 odd . here , because @xmath392 is an even number , each realization makes a positive or a negative contribution to the average , which may therefore be very much reduced .
such destructive quantum interference is strongly dependent on the bath temperature . and
@xmath396 in the logical and logical - ip frames , respectively .
a fixed time interval @xmath397 is taken .
upper panel : high temperature ohmic bath , @xmath398 , lower panel : low temperature ohmic bath , @xmath365 .
( green ) solid line : no control ; ( blue ) stars : a - protocol ; ( purple ) plus : h - protocol ; ( black ) squares : r - protocol in the logical - ip frame ; ( red ) up triangles : r - protocol in the logical frame with small frequency @xmath399 ; ( red ) down triangles : r - protocol in the logical frame with large frequency @xmath400 .
average performed over all realizations.,width=259,height=297 ] among all random pulse realizations , the most effective at suppressing decoherence are those belonging to the smaller ensemble of the h - protocol .
none of them carries a phase , so they always make large positive contributions to the total ensemble average . in a high temperature bath , the realizations that can make negative contributions have often tiny values of @xmath401 , which explains why even in the extreme case of @xmath402 the r - protocol can still lead to some decoherence reduction . in a low temperature bath , on the other hand , decoherence is slower and for the time considered here , the values of @xmath401 for all realizations are very close , which justifies their cancellation when @xmath403 . in the physical frame ,
the average for the density matrix depends on the initial state of the system as @xmath404 e^{-\gamma_{r}^{(k ) } ( t_m , t_0)}\:.\end{aligned}\ ] ] as already discussed in sect .
iii , the problem associated with population inversion may be avoided if a classical register is used to record the actual number of spin flips . _ to summarize : _ two conditions need to be satisfied for the r - protocol to become efficient in reducing decoherence : @xmath405 and also @xmath406 .
notice , however , that when randomness and determinism are combined in a more elaborated protocol , such as the h - protocol , no destructive interference due to @xmath30 occurs .
in addition , the hybrid scheme is still capable to outperform the a - protocol in appropriate regimes . as a final example
, imagine that the coupling parameters @xmath28 between the system and the environment are time dependent and let us for simplicity work again in the logical - ip frame .
the total hamiltonian is given by eqs .
( [ drift0])-([drift ] ) with @xmath324 and @xmath35 .
two illustrative situations are considered : @xmath28 changes sign after certain time intervals , or it periodically oscillates in time .
suppose that @xmath407 , where @xmath408 describes instantaneous sign changes of the coupling parameter after every interval @xmath409 .
for a high temperature bath and a fixed time @xmath410 , fig .
[ fig : devil_quantum ] shows that the results for the a - protocol exhibit a drastic drop when @xmath411 and @xmath412 .
this is due to the fact that some of sign changes happen very close to or coincide with some of @xmath0-pulses of the deterministic sequence , canceling their effect .
in contrast , the occurrence of spin flips in randomized schemes is irregular , so that the latter are more protected against such `` resonances '' and steadily recover coherence as @xmath3 decreases , even though at a slower pace . , with alternating couplings .
( blue ) stars : a - protocol ; ( black ) circles : r - protocol ; ( purple ) plus : h - protocol .
the interval considered is @xmath410 .
averages taken over all possible realizations . ,
width=259 ] note that when dealing with the s- or ls - protocols , the same sort of recoil should be expected for different time dependences and different values of @xmath3 .
assume that the coupling parameter is given by @xmath413 , where @xmath414 and @xmath415 is small .
this function has two superposed periodic behaviors , one with a long period and the other fast oscillating .
the fast oscillations are shown in the left upper panel of fig .
[ fig : time_bath ] . , for @xmath416 and @xmath417 .
upper right panel : decoherence rate in the absence of control .
lower panel : decoherence rate for a fixed time interval @xmath379 . a high temperature ohmic bath , @xmath418 , is considered .
( green ) solid line : no control ; ( blue ) stars : a - protocol ; ( black ) circles : r - protocol ; ( purple ) plus : h - protocol .
averages taken over @xmath231 realizations .
standard deviations for the r - protocol are shown .
, width=297 ] we consider a high temperature bath , @xmath418 .
the right upper panel shows the qubit decoherence in the absence of pulses .
the oscillations in the decay rate are related to the oscillations in the interaction strength between the system and the bath . in the lower panel ,
we fix a time @xmath419 and compare the decoherence rate for the cases of absence of control , a- , h- and r - protocols . when @xmath420 the result for the a - protocol suddenly becomes even worse than not acting on the system .
random pulses , on the contrary , do not show any significant recoil .
the reason for the inefficiency of the a - protocol when @xmath421 becomes evident from the left upper panel of fig .
[ fig : time_bath ] . vertical dashed lines indicate where the pulses occur .
they mostly coincide with the instants where @xmath212 also changes sign . for the ls - protocol
, similar unfavorable circumstances happen for different values of @xmath3 and similar behaviors should be expected for other deterministic protocols and functions @xmath28 . _ to summarize : _ the above examples again reinforce the idea of enhanced stability of randomized controls , and suggest that randomization might represent a safer alternative in reducing decoherence when limited knowledge about the system - bath interaction is available .
a quantitative comparison between deterministic and randomized control for the most elementary target system , consisting of a single ( isolated or open ) qubit , was developed in different frames .
the main conclusions emerging from this study may be summarized as follows .
first , it is always possible to identify conditions under which purely random or hybrid schemes succeed at achieving the desired level of dynamical control .
frame considerations play an important role in specifying such conditions , satisfactory performance in a given frame being ultimately determined by a hierarchy of time scales associated with all the dynamical components in the relevant hamiltonian .
while all protocols become essentially equivalent in the limit of arbitrarily fast control , the behavior for finite pulse separation is rich and rather sensitive to the details of the underlying dynamics . as a drawback of pure random design , ensemble average tends to introduce , in general , additional phase damping , which may be however circumvented by combining determinism and randomness within a hybrid design .
second , for time - independent control settings in this simple system , it was always possible to identify a deterministic protocol with best performance .
while deterministic schemes ensuring accurate averaging of a known interaction always exist in principle @xcite , such a conclusion remains to be verified under more general circumstances , in particular access to a _ restricted _ set of control operations .
the hybrid protocol proved superior to the pure random schemes , as well as to standard asymmetric schemes in certain situations .
third , for time - varying systems , randomized protocols typically allow for enhanced stability against parameter variations , which may severely hamper the performance of deterministic schemes .
pure random design tends to perform better than hybrid in this respect , both choices , however , improving over purely cyclic controls under appropriate conditions .
overall , hybrid design emerges as a preferred strategy for merging advantageous features from different protocols , thereby allowing to better compromise between conflicting needs . from a conceptual standpoint , it is intriguing to realize that complete suppression of decoherence remains possible , in principle , by purposefully introducing a probabilistic component in the underlying control , and perhaps surprising to identify cases where this leads to improved efficiency over pure deterministic methods . in a broader context
, however , it is worth mentioning that the philosophy of recognizing a beneficial role of randomness in physical processes has a long history . within nmr , the stochastic averaging of intermolecular interactions in gases and isotropic liquids due to random translational and re - orientational motions
may be thought of as a naturally occurring random self - decoupling process @xcite .
in spectroscopic applications of so - called stochastic nmr and stochastic magnetic - resonance imaging @xcite , spin excitation via trains of weak rf pulses randomly modulated in amplitude , phase , and/or frequency are used to enhance decoupling efficiencies over a broader frequency bandwidth than attainable otherwise . even more generally , the phenomenon of stochastic resonance @xcite is paradigmatic in terms of pointing to a constructive role of noise in the transmission of physical signals . within qip ,
strategies aimed at taking advantage of noise and/or stochasticity have been considered in contexts ranging from quantum games @xcite , to quantum walks @xcite , dissipation - assisted quantum computation @xcite , as well as specific coherent - control @xcite and quantum simulation @xcite scenarios . yet another suggestive example is offered by the work of prosen and nidari , who have shown how static perturbations characterizing faulty gates may enhance the stability of quantum algorithms @xcite .
more recently , as mentioned , both pure random @xcite and hybrid @xcite active compensation schemes for static coherent errors have been proposed . while it is important to stress that none of the above applications stem from a general _ control - theoretic framework _ as developed in @xcite , it is still rewarding to fit such different examples within a unifying perspective . our present analysis should be regarded as a first step toward a better understanding and exploitation of the possibilities afforded by randomization for coherent and decoherent error control . as such
, it should be expanded in several directions , including more realistic control systems and settings , and fault - tolerance considerations . while we plan to report on that elsewhere , it is our hope that our work will stimulate fresh perspectives on further probing the interplay between the field of coherent quantum control and the world of randomness .
l. v. warmly thanks manny knill for discussions and feedback during the early stages of this project .
the authors are indebted to an anonymous referee for valuable suggestions .
l. f. s. gratefully acknowledges support from constance and walter burke through their special projects fund in quantum information science .
the control hamiltonian is designed according to the intended modification of the target dynamics in a desired frame . throughout this work
, our goal has been to freeze the system evolution by removing any phase accumulated due to the unitary evolution , as well as avoiding nonunitary ensemble dephasing and decoherence . as clarified below
, this requires the use of _ identical @xmath0-pulses in the physical frame_. this condition may be relaxed at the expense of no longer refocusing the unitary evolution .
let the control of the system be achieved via the application of an external alternating field ( e.g. , a radiofrequency magnetic field ) , @xmath422 \sigma_x \ : , \label{hc_cos}\ ] ] where the carrier is tuned on resonance with the qubit central frequency , @xmath51 . as described in the text , @xmath423 $ ] and each pulse happens at @xmath50 , having duration @xmath424 and amplitude @xmath48 . upon invoking the rotating wave approximation ( rwa ) , hence neglecting the counter - rotating terms @xmath425
$ ] and @xmath426 $ ] in ( [ hc_cos ] ) , the control hamiltonian given in the main text is found .
the function @xmath427 characterizes the phase properties of the pulses we deal with .
we compare two relevant possibilities : the propagator corresponding to the above choices may be in general obtained by seeking a transformation which removes the time dependence of @xmath431 within each pulse .
a transformation to an _ absolute _ frame rotating with the carrier frequency , which on resonance is identical with the interaction picture , leads to @xmath432 h_c^{(j)}(t ) \exp \left[-i \omega_0 t\sigma_z /2 \right ] \nonumber \\ & = & v^{(j)}(t ) \exp \left[-i \frac{\varphi_j(t)}{2}\sigma_z\right ] \sigma_x \exp \left[i \frac{\varphi_j(t)}{2}\sigma_z \right]\ : .
\nonumber\end{aligned}\ ] ] thus , the choice @xmath429 corresponds to pulses which are translationally invariant in time in this frame . in case ( i ) , the above transformation does not remove time dependence , which would instead be accomplished by moving to a _ relative _ rotating frame via a rotation @xmath433 $ ] . from the above expression ,
the interaction - picture propagators for an instantaneous @xmath0-pulse applied at @xmath50 are found , respectively , as @xmath434 \exp \left[-i \frac{\pi}{2 } \sigma_x \right ] \exp \left[-i \frac{\omega_0 t_j}{2 } \sigma_z \right ] , \nonumber \\ { \rm ( ii ) } &
\ : & p_j^i = \exp \left[-i \frac{\pi}{2 } \sigma_x \right ] .
\nonumber \end{aligned}\ ] ] we can then return to the schrdinger picture using the relation @xmath435
p_j^i \exp [ i \omega_0 t_j \sigma_z /2]\:,\ ] ] leading to the propagators @xmath436 , \label{p_varphi } \\ { \rm ( ii ) } & \ : & p_j = \exp \left[-i \frac{\omega_0 t_j}{2 } \sigma_z \right ] \exp \left[-i \frac{\pi}{2 } \sigma_x \right ] \exp \left[i \frac{\omega_0 t_j}{2 } \sigma_z \right]\ : . \nonumber \end{aligned}\ ] ] thus , pulses with @xmath428 are confirmed to be translationally invariant in time in the physical frame , as directly clear from the dependence @xmath437 in ( [ hc_cos ] ) .
the difference between the two choices to the control purposes becomes evident by considering the a - protocol on the isolated qubit . from ( [ p_varphi ] ) , the propagators @xmath438 in the physical frame are , respectively , @xmath439 , \nonumber \end{aligned}\ ] ] which leads to the conclusion that refocusing in the physical frame may only be achieved with identical pulses , that is , if @xmath440 .
clearly , for the choice @xmath429 , the accumulated phase may only be disregarded in the frame rotating with the frequency @xmath30 .
both choices are equally useful if decoherence suppression becomes the primary objective in the open system case . | we revisit the problem of switching off unwanted phase evolution and decoherence in a single two - state quantum system in the light of recent results on random dynamical decoupling methods [ l. viola and e. knill , phys .
rev .
lett . * 94 * , 060502 ( 2005 ) ] .
a systematic comparison with standard cyclic decoupling is effected for a variety of dynamical regimes , including the case of both semiclassical and fully quantum decoherence models . in particular ,
exact analytical expressions are derived for randomized control of decoherence from a bosonic environment .
we investigate quantitatively control protocols based on purely deterministic , purely random , as well as hybrid design , and identify their relative merits and weaknesses at improving system performance .
we find that for time - independent systems , hybrid protocols tend to perform better than pure random and may improve over standard asymmetric schemes , whereas random protocols can be considerably more stable against fluctuations in the system parameters . beside shedding light on the physical requirements underlying randomized control ,
our analysis further demonstrates the potential for explicit control settings where the latter may significantly improve over conventional schemes . |
it is well known nowadays that our current universe is experiencing an accelerated expansion @xcite-@xcite .
there are two general ways to explain this accelerated expansion of the universe @xcite-@xcite .
the first way is considering that the universe is essentially filled by an exotic fluid with negative pressure , responsible of it acceleration , called the dark energy .
the second way is modifying the gravitational action from the general relativity ( gr ) without the need of the dark energy , firstly considered by 1970 @xcite .
buchdahl has generalized the einstein equations by substituting the ricci s scalar @xmath2 by an arbitrary function of @xmath2 , named @xmath7 theory of gravity .
another theories , view as alternative to the gr , also have been undertaken , still in the way to better explain the obscure content of the universe , as , @xmath8@xcite-@xcite , @xmath1@xcite-@xcite and @xmath9@xcite-@xcite , where @xmath10 , @xmath3 , @xmath11 and @xmath12 are the invariant of gauss - bonnet , the trace of energy - momentum tensor , the ricci s tensor and the energy - momentum tensor corresponding to the ordinary content of the universe . in this paper
, we focus our attention on @xmath1 theory of gravity . this theory has been considered first by harko and collaborators @xcite .
another authors also have considered this present theory and important results have been found @xcite-@xcite .
however , any one of the works developed in these papers does not solve the coincidence problem , that is , how does the universe transits from the decelerated phase to the current accelerated phase ?
the universe is accelerated for the low values of the redshift and the one that splits the two phases ( transition redshift ) , denoted @xmath13 , those the current value is @xmath14 @xcite , @xcite .
the @xmath1 theory can successfully predict the transition from the matter dominated phase to the current accelerated one and several works have described this transition @xcite .
the reconstruction of @xmath1 models describing the matter dominated and accelerated phases had been performed in @xcite .
various works also have been performed , still in the optic to better explore this transition with interesting results ( see @xcite , @xcite and @xcite ) . in this paper , we focus our attention on the epoch after the recombination where , beside the well known ordinary matter ( the dust ) , there is the self - interacting matter , called collisional matter .
the model of collisional matter has been studied in some works within others theories of gravity , leading to interesting results @xcite,@xcite , @xcite , @xcite .
this approach of considering new form of matter , instead of the cold dark matter can edify us on the choice of the models of modified gravity .
oikonomou and collaborators @xcite have studied the time evolution of the cosmological parameters during the late time acceleration of the universe with the presence of the collisional matter in the framework of modified @xmath7 gravity . in this paper , we have extended the same idea to the @xmath1 .
some @xmath1 models have been considered and the behaviours of the cosmological parameters have been performed and compared with the @xmath5cdm model .
we see within many results that depending on the input parameters according to the model under consideration , the inclusion of collisional matter may lead to a better explanation of the phase transition , comparatively to the model where just the usual ordinary matter is considered .
the paper is organized as follows : in section [ sec2 ] we describe the general formalism of @xmath1 theory of gravity . the collisional matter that self - interacts is presented in the section [ sec3 ] .
the section [ sec4 ] is devoted to the study of time evolution of the cosmological parameters where the universe is considered to be filled by the usual ordinary matter and the collisional one . here , for the study of these cosmological parameters , we focus our attention on the transition from the decelerated phase to the accelerated one . in the section [ sec5 ]
we examine the evolution of the equation of state of the dark energy where the matter content is assumed as a fluid is composed by the collisional matter and the radiation . the conclusion and perspectives
are presented in the section [ sec6 ] .
in this section we present the generality of @xmath1 theory by substituting the curvature scalar @xmath2 of the gr by a function of @xmath2 and the trace @xmath3 , and writing the action as @xcite @xmath15\,\ , , \label{1}\end{aligned}\ ] ] where @xmath2 , @xmath3 denote the curvature scalar and the trace of the energy - momentum tensor , respectively , and @xmath16 , @xmath17 being the gravitation constant .
the energy - momentum tensor associated to the matter is defined by @xmath18 let us assume that the matter lagrangian density @xmath19 only depends on the components of the metric tensor , but not on its derivatives .
thereby , one gets @xmath20 within the metric formalism , varying the action @xmath21 with respect to the metric , one obtains the following field equations @xcite @xmath22 where @xmath23 are the partial derivatives of @xmath24 with respect to @xmath25and @xmath26 respectively .
the tensor @xmath27 is determined by @xmath28 as mentioned in our introduction , we assume that the whole content of the universe is a perfect fluid
. then , by setting the matter lagrangian density to @xmath29 , the energy - momentum tensor may take the following expression ( see appendix ) @xmath30 where the four - velocity satisfies the relations @xmath31 and @xmath32 = 0 .
* note that the expression ( [ 6 ] ) is obtained with the consideration according to what the pressure does not depend on the metric tensor . within this consideration according to what the lagrangian density does not depend of the metric tensor , the contribution of the last term of @xmath33 vanishes and this equation takes the following form ( see also eq . ( 21 ) of @xcite ) @xmath34 however , in the case of the electromagnetic field or scalar field , there is actively contribution of the last term of ( [ 5 ] ) because in these cases the respective lagrangian densities @xmath35 and @xmath36 , explicitly depend on the metric .
( see also comments after eq.(18 ) and eq.(19 ) of @xcite ) .
* making use of eq ( [ 7 ] ) , the field equation ( [ 4 ] ) can be reformulated as @xmath37 where @xmath38 is the effective gravitational constant and @xmath39 , \label{9}\end{aligned}\ ] ] representing the effective energy - momentum tensor of the matter .
taking into account the covariant divergence of eq.([8 ] ) , with we use the identities @xmath40 , and @xmath41 , on gets @xmath42 .
\label{222}\end{aligned}\ ] ] after some elementary transformations , eq .
( [ 222 ] ) takes the following expression @xmath43 \partial_{\mu}\rho\bigg\}. \label{223 } \end{aligned}\ ] ] it comes from the above expression that for the @xmath44 theory of gravity , the conservation of the energy - momentum tensor is not automatic . in order
the reach this condition , it is important to constraint the second hand of this equation to zero ( see eq .
( [ 15 ] ) ) .
this condition is extremely crucial for determining the suitable expression of the trace depending part of the algebraic action . *
the reference @xcite considers the constraint of the conservation to energy - momentum tensor ( as in ( [ 15 ] ) ) and one obtains , after the integration on the right side of ( [ 223 ] ) , @xmath45}+\beta_0 $ ] , with @xmath46 real constants ; ( see ( 19 ) of @xcite ) .
* here we are interested to the spatially flat friedmann - robertson - walker ( frw ) metric @xmath47.\label{10}\end{aligned}\ ] ] thus , the ricci scalar in this background is given by @xmath48
this king of matter has been introduced firstly in @xcite in the framework of gr and @xmath7 .
the evolution of the collisional matter depends on the source that drives the whole content ( matter and energy ) of the univers .
the study of the model of collisional matter with perfect fluid has been performed widely in @xcite . in this work , we recall that our basic assumption is that the matter has a total mass - energy density , denoted by @xmath49 , which is assumed to be depending on two contributions as @xmath50 here , @xmath51 refers to the part that does not change due to the fact of characterising the usual matter content @xcite,@xcite .
concerning the term @xmath52 , it expresses the energy density part of the energy momentum tensor associated with thermodynamical content of the collisional matter .
this fluid could not be dust , but possesses a positive pressure and satisfying the following equation of state @xmath53 where @xmath54 denotes the parameter of equation of state of the collisional matter whose values is such that @xmath55 .
the potential energy density is assumed to have the form @xcite,@xcite , @xmath56 where the constants @xmath57 and @xmath58 are their current values .
accordingly , after elementary transformation the total - energy density of the universe can be written as @xmath59 through the continuity equation , the motions of the volume elements in the interior of a continuous medium can be traduced @xmath60 and the energy - momentum tensor takes the following form @xmath61 where @xmath62 is the four velocity , satisfying the relation @xmath63 .
we note here that @xmath64 because the pressure of ordinary matter is negligible . making use of the frw line element ( [ 10 ] ) , the conservation law of the equation ( [ 15 ] ) @xmath65 which , upon consideration of the equations ( [ 12 ] ) and ( [ 14 ] ) , leads to @xcite @xmath66 where @xmath67 is the current scale factor .
we can describe the collisional matter by equations ( [ 14 ] ) and ( [ 18 ] ) and use therm in the rest of the manuscript in the next section .
the value of @xmath58 is equal to @xcite @xmath68 we will use the same value @xcite,@xcite of @xmath69 for making the numerical study . according to the fact that @xmath70 , and making use of eq .
( [ 13 ] ) , one gets @xmath71 , meaning that this latter is the current value of the potential energy per unit rest - mass , associated with the infinitesimal deformations ( expansions or / and compressions ) of the fluid ( see detail in @xcite ) .
in this subsection we examine the deceleration parameter @xmath0 in @xmath1 considering a that the universe , beside the ordinary matter and dark energy , also is filled with collisional matter at late time cosmological evolution . in this optic , we rewritte the field equation ( [ 4 ] ) in the following form @xmath72 , \label{19 } \end{aligned}\ ] ] @xmath73 , \label{20}\end{aligned}\ ] ] where we have assumed @xmath74 . the parameter @xmath75 denotes the hubble parameter , the dot " , the time derivative of @xmath76 .
we propose to reformulate the right side of ( [ 19 ] ) and ( [ 20 ] ) respectively in term of effective energy density @xmath77 and pressure @xmath78 , and we write ( [ 19 ] ) and ( [ 20 ] ) as @xmath79 @xmath80 here , the effective energy density @xmath81 and effective pressure @xmath82 are respectively defined as @xmath83 , \label{23}\end{aligned}\ ] ] @xmath84 \label{24}\end{aligned}\ ] ] in this way , we have a matter fluid representation of the so - called geometrical dark energy in @xmath85 gravity with the energy density @xmath86 and pressure @xmath87 . from the conservation law ,
the effective energy density evolves as @xmath88 by making use of the equations ( [ 19 ] ) and ( [ 20 ] ) , ( [ 25 ] ) can be rewritten as @xmath89 one can rewrite the above equation using the redshift @xmath90 as follows @xmath91 as convention in this paper , we set to unity the current value of the scale factor
. we also assume the algebraic function as a sum of two independent functions , such that , @xmath92 .
the @xmath93 model is the one obtained by imposing the conservation of the energy - momentum tensor * ( [ 223])*. the dynamics and stability of this model are studied in @xcite and interesting results have been found .
this model reads @xmath94 with @xmath95 . here
the trace of the energy- momentum tensor ( [ 16 ] ) reads @xmath96 from equations ( [ 14 ] ) and ( [ 18 ] ) , the model @xmath93 and its first derivative are @xmath97}^{\alpha } , \label{28 } \end{aligned}\ ] ] @xmath98}^{\alpha-1}. \label{29 } \end{aligned}\ ] ] by using ( [ 28 ] ) and ( [ 29 ] ) , one rewrite ( [ 27 ] ) as @xmath99 where @xmath100 @xmath101 @xmath102 the equation ( [ 30 ] ) is the main tool to be used for analysing the cosmological evolution of the considered @xmath1 models .
the numerical analysis also will be performed in order to find the hubble parameter . in a first step , by using the expression @xmath103 we can perform the study of the deceleration parameter with the account of the collisional matter , this latter assumed as a function of the redshift @xmath104 . in this rubric
we will focus our attention on the transition from the decelerated phase to the accelerated one .
therefore , we will compare the results with the case where the non - collisional matter ( pressure - less ) is considered , and also compare with the @xmath105 case .
we recall that in @xcite , the corresponding deceleration parameter of the expansion in the flat friedmann cosmology for the @xmath105 model is given by @xmath106 { \bigg [ \omega_{m0}{(1+z)}^{3}+l_0 \bigg]}^{-1 } , \label{33.1}\end{aligned}\ ] ] where @xmath107 and @xmath108 @xcite .
moreover , the effective equation of state parameter @xmath109 is defined as , @xmath110 and we present its evolution versus the redshift . a comparison with the results to be found will be done with the results coming from pure @xmath1 models with non - collisional matter . in this subsection
we discuss two @xmath7 models , performed in @xcite for describing the late time cosmological evolution of our universe under the consideration that it is filled with collisional ( self - interacting ) matter .
thereby , we solve numerically the differential equation about @xmath4 . the task here is to find the models that should be in agreement with the observational data , namely the deceleration parameter @xmath0 and the parameter of effective equation of state @xmath6 .
the first model considered is a modified power - law model of the curvature scalar @xmath111 where @xmath112 , @xmath113 and @xmath114 are positive constants .
the second , is an exponential model @xmath115 where @xmath116 and @xmath117 are constant parameters .
numerical plots of the deceleration parameter @xmath0 and the effective equation of state @xmath6 are presented from ( [ 30 ] ) with collisional matter @xmath118 and non - collisional matter @xmath119 for each model . [ cols= " > , < " , ] for the figs .
@xmath120 , @xmath121 , @xmath122 and @xmath123 , curves traducing the evolution of @xmath124 , @xmath125 , @xmath4 , @xmath126 and @xmath127 , present some values strongly consistent with the cosmological observational data .
these features prove that the fact of considering the coexistence of the collisional matter and radiation do not change the well know behaviour the different parameters , and confirm that such consideration may be made .
in this paper we undertake the @xmath1 theory of gravity .
we focus our attention on the cosmological evolution of the parameter of deceleration @xmath0 and also the parameter @xmath6 of the effective equation of state .
the spacial aspect in this paper is that , besides the usual ordinary and the dark energy , new matter is assumed to participate to the matter contribution of the universe .
this matter is self - interacting and called collisional matter , but with positive pressure .
what we now generally is that the ordinary matter interacts with the dark energy , but any of them does self interacts . from the cosmological considerations , its quite reasonable to introduce such a matter , i.e , a collisional , and search for its effects on the cosmological evolution of the universe . to this end , we considered two @xmath1 models as a sum of two gravitational models @xmath128 and @xmath129 , respectively two functions depending the curvature scalar @xmath2 and the trace @xmath3 of the energy - momentum tensor .
the general form of the @xmath1 models is constrained to the conservation of the energy leading to the form ( [ 000 ] ) .
the task has been to choose a suitable expression for the function @xmath7 .
therefore , we considered , first , a generalized power - law in @xmath2 , and the second expression is an exponential form in @xmath2 .
we plot the evolutions of @xmath0 and @xmath6 for both the collisional matter , non - collisional matter and the @xmath5cdm and compare them .
our results show that , depending on the type of the model under consideration , the curve characteristic of the collisional matter approaches the @xmath5cdm one that the curve that describes the evolution of the dust . in the same way
, the curve representative of @xmath6 shows that its values are consistent with the observational , confirm the importance of considering the collisional matter in the study of the evolution of our universe .
moreover , we point out other type of matter , the so - called cardassian self - interacting matter , possessing a negative pressure and its corresponding parameter of equation of state has to be different from @xmath130 .
this means that if such a matter is included in the content of the universe , there does not have vacuum energy . in this case , the same analysis have been done , as in the case of collisional matter , comparing the curves representative of this later to those of the dust and also the @xmath5cdm .
our results show that , for both the first and the second @xmath1 models , there is quite reasonable to consider the existence of the cardassian matter since the curve characteristic of this latter approached more the @xmath5cdm that the dust for some values of the redshift , confirming also that this kind of matter may be considered as a component of the universe , because also in this , the parameter @xmath6 takes values strongly consistent with the observational data . *
acknowledgments * : the authors thank prof .
s. d. odintsov for useful comments and suggestions .
e. h. baffou thanks imsp for every kind of support during the realization of this work .
manuel e. rodrigues thanks ufpa , edital 04/2014 propesp , and cnpq , edital mcti / cnpq / universal 14/2014 , for partial financial support . m.j.s .
houndjo and a.v.kpadonou thank _ ecole normale suprieure de natitingou _ for partial financial support .
* let us assume that the lagrangian density @xmath131 is a sum of the kinetic term @xmath132 and the one linked to the internal energy namely @xmath133 , i.e , @xmath134 therefore , the energy - momentum tensor @xmath12 , reads @xmath135
\\ = t_{\mu\nu(k)}+t_{\mu\nu(u)}\ ] ] the well known expression of the kinetic energy - momentum tensor is expressed as @xmath136 .
+ let us consider @xmath133 as @xmath137 using the method of lagrange multipliers , the equations for the parameter @xmath113 and the variables @xmath138 are given by @xmath139 and @xmath140 by setting @xmath141 and @xmath142 , one gets @xmath143 and @xmath144 .
+ therefore , one can rewritten @xmath133 as @xmath145 the internal energy - momentum tensor @xmath146 can be developed in the following from @xmath147 the second term @xmath148 can be expressed as @xmath149 the pressure @xmath150 does not depend on the metric tensor @xmath151 , such that @xmath152 , and then @xmath148 reduces to @xmath153 and finally , the internal energy - momentum tensor becomes @xmath154 by using the contribution of the kinetic and the internal energy - momentum tensors and absorbing the sign @xmath155 into the total stress tensor , one gets the energy - momentum tensor of matter fluid as @xmath156 it comes from the above calculus that the expression of the energy - momentum tensor is a consequence of the fact of considering that the pressure does not depends on the metric tensor @xmath151 . * s.perlmutter et al .
[ sncp collaboration ] , astrophys .
j. 517 , 565 ( 1999 ) [ arxiv : astro - ph/9812133 ] ; a. g. riess et al .
[ supernova search team collaboration ] , astron .
j. 116 , 1009 ( 1998 ) [ arxiv : astro - ph/9805201 ] .
d. n. spergel et al .
[ wmap collaboration ] , astrophys .
148 , 175 ( 2003 ) [ arxiv : astro - ph/0302209 ] ; astrophys .
170 , 377 ( 2007 ) [ arxiv : astro - ph/0603449 ] .
e. komatsu et al .
[ wmap collaboration ] , astrophys .
180 , 330 ( 2009 ) [ arxiv:0803.0547 [ astro - ph ] ] .
e. komatsu et al .
[ wmap collaboration ] , astrophys .
192 , 18 ( 2011 ) [ arxiv:1001.4538 [ astro-ph.co ] ] .
m.tegmark et al .
[ sdss collaboration ] , phys.rev.d 69 , 103501 ( 2004 ) [ arxiv : astro - ph/0310723 ] ; u. seljak et al .
[ sdss collaboration ] , phys .
d 71 , 103515 ( 2005 ) [ arxiv : astro - ph/0407372 ] .
d. j. eisenstein et al .
[ sdss collaboration ] , astrophys .
j. 633 , 560 ( 2005 ) [ arxiv : astro - ph/0501171 ] .
b. jain and a. taylor , phys .
91 , 141302 ( 2003 ) [ arxiv : astro - ph/0306046 ] .
perlmutter , s. et al .
: astrophys .
j. 483(1997)565 ; perlmutter , s. et al . : nature 391(1998)51 ; perlmutter , s. et al .
: astrophys .
j. 517(1999)565 ; riess , a.g .
: astrophys .
j. 607(2004)665 ; riess , a.g .
j. 659(2007)98 .
s. nojiri , s. d. odintsov and m. sasaki , phys . rev .
d 71 , 123509 ( 2005 ) ; t. koivisto and d. f. mota , phys .
b 644 , 104 ( 2007 ) ; t. koivisto and d. f. mota , phys . rev .
d 75 , 023518 ( 2007 ) ; s. kawai , m. a. sakagami and j. soda , phys . lett .
b 437 , 284 ( 1998 ) ; s. nojiri and s. d. odintsov , phys . lett .
b 631 , 1 ( 2005 ) ; a. de felice and s. tsujikawa , phys .
b 675 , 1 ( 2009 ) ; t. p. sotiriou , arxiv:0710.4438 [ gr - qc ] ; k. uddin , j. e. lidsey and r. tavakol , gen .
41 , 2725 ( 2009 ) ; g. cognola , e. elizalde , s. nojiri , s. d. odintsov and s. zerbini , phys . rev .
d 73 , 084007 ( 2006 ) ; b. li , j. d. barrow and d. f. mota , phys .
d 76 , 044027 ( 2007 ) ; s. y. zhou , e. j. copeland and p. m. saffin , jcap 0907 , 009 ( 2009 ) ; a. de felice and s. tsujikawa , phys .
d 80 , 063516 ( 2009 ) .
m. j. s. houndjo , int . j. mod .
d. 21 , 1250003 ( 2012 ) .
arxiv : 1107.3887 [ astro-ph.co ] ; m. j. s. houndjo and o. f. piattella , int
d. 21 , 1250024 ( 2012 ) .
arxiv : 1111.4275 [ gr.qc ] ; d. momeni , m. jamil and r. myrzakulov , euro . phys .
j. c 72 , arxiv : 1107.5807[physics.gen - ph ] .
m. j. s. houndjo , int . j. mod
d. 21 , 1250003 ( 2012 ) .
arxiv : 1107.3887 [ astro-ph.co ] ; m. j. s. houndjo and o. f. piattella , int
d. 21 , 1250024 ( 2012 ) .
arxiv : 1111.4275 [ gr.qc ] ; d. momeni , m. jamil and r. myrzakulov , euro . phys
. j. c 72 , arxiv : 1107.5807[physics.gen - ph ] .
perivolaropoulos , l. 2007 , accelerating universe : observational sta- tus and theoretical implications , in the invisible universe : dark matter and dark energy , ed .
l. papantonopoulos , lecture notes in physics 720 , 257 ( new york : springer verlag ) . | we study the evolution of the cosmological parameters , namely , the deceleration parameter @xmath0 and the parameter of effective equation of state in a universe containing , besides the ordinary matter and dark energy , a self - interacting ( collisional matter ) , in the generalized @xmath1 theory of gravity , where @xmath2 and @xmath3 are the curvature scalar and the trace of the energy - momentum tensor , respectively .
we use the generalized frw equations , the equation of continuity and obtain a differential equation in @xmath4 , and solve it numerically for studying the evolution of the cosmological parameters .
two @xmath1 models are considered and the results with collisional matter are compared with the ones of the @xmath5cdm model , and also with the model where there exists only non - collisional matter .
the curves show that the models are acceptable because the values found for @xmath6 are consistent with the observational data . |
evolutionary algorithms are a type of general problem solvers that can be applied to many difficult optimization problems . because of their generality , these algorithms act similarly like swiss army knife @xcite that is a handy set of tools that can be used to address a variety of tasks . in general ,
a definite task can be performed better with an associated special tool . however , in the absence of this tool , the swiss army knife may be more suitable as a substitute .
for example , to cut a piece of bread the kitchen knife is more suitable , but when traveling the swiss army knife is fine . similarly , when a problem to be solved from a domain where the problem - specific knowledge is absent evolutionary algorithms can be successfully applied .
evolutionary algorithms are easy to implement and often provide adequate solutions .
an origin of these algorithms is found in the darwian principles of natural selection @xcite . in accordance with these principles
, only the fittest individuals can survive in the struggle for existence and reproduce their good characteristics into next generation . as illustrated in fig .
[ pic:1 ] , evolutionary algorithms operate with the population of solutions . at first , the solution needs to be defined within an evolutionary algorithm .
usually , this definition can not be described in the original problem context directly .
in contrast , the solution is defined by data structures that describe the original problem context indirectly and thus , determine the search space within an evolutionary search ( optimization process ) .
there exists the analogy in the nature , where the genotype encodes the phenotype , as well .
consequently , a genotype - phenotype mapping determines how the genotypic representation is mapped to the phenotypic property . in other words ,
the phenotypic property determines the solution in original problem context .
before an evolutionary process actually starts , the initial population needs to be generated .
the initial population is generated most often randomly .
a basis of an evolutionary algorithm represents an evolutionary search in which the selected solutions undergo an operation of reproduction , i.e. , a crossover and a mutation . as a result , new candidate solutions ( offsprings )
are produced that compete , according to their fitness , with old ones for a place in the next generation .
the fitness is evaluated by an evaluation function ( also called fitness function ) that defines requirements of the optimization ( minimization or maximization of the fitness function ) . in this study ,
the minimization of the fitness function is considered . as the population evolves solutions becomes fitter and fitter .
finally , the evolutionary search can be iterated until a solution with sufficient quality ( fitness ) is found or the predefined number of generations is reached @xcite .
note that some steps in fig .
[ pic:1 ] can be omitted ( e.g. , mutation , survivor selection ) .
an evolutionary search is categorized by two terms : exploration and exploitation .
the former term is connected with a discovering of the new solutions , while the later with a search in the vicinity of knowing good solutions @xcite .
both terms , however , interweave each other in the evolutionary search . the evolutionary search acts correctly when a sufficient diversity of population is present .
the population diversity can be measured differently : the number of different fitness values , the number of different genotypes , the number of different phenotypes , entropy , etc .
the higher the population diversity , the better exploration can be expected .
losing of population diversity can lead to the premature convergence .
exploration and exploitation of evolutionary algorithms are controlled by the control parameters , for instance the population size , the probability of mutation @xmath0 , the probability of crossover @xmath1 , and the tournament size . to avoid a wrong setting of these
, the control parameters can be embedded into the genotype of individuals together with problem variables and undergo through evolutionary operations .
this idea is exploited by a self - adaptation .
the performance of a self - adaptive evolutionary algorithm depends on the characteristics of population distribution that directs the evolutionary search towards appropriate regions of the search space @xcite .
@xcite , however , widened the notion of self - adaptation with a generalized concept of self - adaptation .
this concept relies on the neutral theory of molecular evolution @xcite .
regarding this theory , the most mutations on molecular level are selection neutral and therefore , can not have any impact on fitness of individual . consequently ,
the major part of evolutionary changes are not result of natural selection but result of random genetic drift that acts on neutral allele .
an neutral allele is one or more forms of a particular gene that has no impact on fitness of individual @xcite .
in contrast to natural selection , the random genetic drift is a whole stochastic process that is caused by sampling error and affects the frequency of mutated allele . on basis of this theory igel and toussaint
ascertain that the neutral genotype - phenotype mapping is not injective .
that is , more genotypes can be mapped into the same phenotype . by self - adaptation , a neutral part of genotype ( problem variables ) that determines the phenotype enables discovering the search space independent of the phenotypic variations . on the other hand ,
the rest part of genotype ( control parameters ) determines the strategy of discovering the search space and therefore , influences the exploration distribution .
although evolutionary algorithms can be applied to many real - world optimization problems their performance is still subject of the no free lunch ( nfl ) theorem @xcite . according to this theorem
any two algorithms are equivalent , when their performance is compared across all possible problems .
fortunately , the nfl theorem can be circumvented for a given problem by a hybridization that incorporates the problem specific knowledge into evolutionary algorithms . in fig .
2 some possibilities to hybridize evolutionary algorithms are illustrated . at first
, the initial population can be generated by incorporating solutions of existing algorithms or by using heuristics , local search , etc .
in addition , the local search can be applied to the population of offsprings .
actually , the evolutionary algorithm hybridized with local search is called a memetic algorithm as well @xcite .
evolutionary operators ( mutation , crossover , parent and survivor selection ) can incorporate problem - specific knowledge or apply the operators from other algorithms .
finally , a fitness function offers the most possibilities for a hybridization because it can be used as decoder that decodes the indirect represented genotype into feasible solution . by this mapping , however , the problem specific knowledge or known heuristics can be incorporated to the problem solver . in this chapter the hybrid self - adaptive evolutionary algorithm ( hsa - ea ) is presented that is hybridized with : * construction heuristic , * local search , * neutral survivor selection , and * heuristic initialization procedure .
this algorithm acts as meta - heuristic , where the down - level evolutionary algorithm is used as generator of new solutions , while for the upper - level construction of the solutions a traditional heuristic is applied .
this construction heuristic represents the hybridization of evaluation function .
each generated solution is improved by the local search heuristics .
this evolutionary algorithm supports an existence of neutral solutions , i.e. , solutions with equal values of a fitness function but different genotype representation .
such solutions can be arisen often in matured generations of evolutionary process and are subject of neutral survivor selection .
this selection operator models oneself upon a neutral theory of molecular evolution @xcite and tries to direct the evolutionary search to new , undiscovered regions of search space .
in fact , the neutral survivor selection represents hybridization of evolutionary operators , in this case , the survivor selection operator .
the hybrid self - adaptive evolutionary algorithm can be used especially for solving of the hardest combinatorial optimization problems @xcite .
the chapter is further organized as follows . in the sect .
2 the self - adaptation in evolutionary algorithms is discussed . there , the connection between neutrality and self - adaptation is explained .
3 describes hybridization elements of the self - adaptive evolutionary algorithm .
4 introduces the implementations of hybrid self - adaptive evolutionary algorithm for graph 3-coloring in details .
performances of this algorithm are substantiated with extensive collection of results .
the chapter is concluded with summarization of the performed work and announcement of the possibilities for the further work .
optimization is a dynamical process , therefore , the values of parameters that are set at initialization become worse through the run .
the necessity to adapt control parameters during the runs of evolutionary algorithms born an idea of self - adaptation @xcite , where some control parameters are embedded into genotype .
this genotype undergoes effects of variation operators .
mostly , with the notion of self - adaptation evolutionary strategies @xcite are connected that are used for solving continuous optimization problems .
typically , the problem variables in evolutionary strategies are represented as real - coded vector @xmath2 that are embedded into genotype together with control parameters ( mostly mutation parameters ) .
these parameters determine mutation strengths @xmath3 that must be greater than zero .
usually , the mutation strengths are assigned to each problem variable . in that case , the uncorrelated mutation with @xmath4 step sizes is obtained @xcite . here , the candidate solution is represented as @xmath5 .
the mutation is now specified as follows : @xmath6 @xmath7 where @xmath8 and @xmath9 denote the learning rates . to keep the mutation strengths @xmath10 greater than zero
, the following rule is used @xmath11 frequently , a crossover operator is used in the self - adaptive evolutionary strategies .
this operator from two parents forms one offsprings .
typically , a discrete and arithmetic crossover is used .
the former , from among the values of two parents @xmath12 and @xmath13 that are located on @xmath14-th position , selects the value of offspring @xmath15 randomly .
the later calculates the value of offspring @xmath15 from the values of two parents @xmath12 and @xmath13 that are located on @xmath14-th position according to the following equation : @xmath16 where parameter @xmath17 captures the values from interval @xmath18 $ ] . in the case of @xmath19 ,
the uniform arithmetic crossover is obtained .
the potential benefits of neutrality was subject of many researches in the biological science @xcite .
at the same time , the growing interest for the usage of this knowledge in evolutionary computation was raised @xcite . @xcite
dealt with the non - injectivity of genotype - phenotype mapping that is the main characteristic of this mapping .
that is , more genotypes can be mapped to the same phenotype .
@xcite pointed out that in the absence of an external control and with a constant genotype - phenotype mapping only neutral genetic variations can allow an adaptation of exploration distribution without changing the phenotypes in the population .
however , the neutral genetic variations act on the genotype of parent but does not influence on the phenotype of offspring . as a result ,
control parameters in evolutionary strategies represent a search strategy .
the change of this strategy enables a discovery of new regions of the search space .
the genotype , therefore , does not include only the information addressing its phenotype but the information about further discovering of the search space as well . in summary ,
the neutrality is not necessary redundant but it is prerequisite for self - adaptation .
this concept is called the general concept of self - adaptation as well @xcite .
evolutionary algorithms are a generic tool that can be used for solving many hard optimization problems . however , the solving of that problems showed that evolutionary algorithms are too problem - independent .
therefore , there are hybridized with several techniques and heuristics that are capable to incorporate problem - specific knowledge .
@xcite identified mostly used hybrid architectures today as follows : * hybridization between two evolutionary algorithms @xcite , * neural network assisted evolutionary algorithm @xcite , * fuzzy logic assisted evolutionary algorithm @xcite , * particle swarm optimization assisted evolutionary algorithm @xcite , * ant colony optimization assisted evolutionary algorithm @xcite , * bacterial foraging optimization assisted evolutionary algorithm @xcite , * hybridization between an evolutionary algorithm and other heuristics , like local search @xcite , tabu search @xcite , simulated annealing @xcite , hill climbing @xcite , dynamic programming @xcite , etc . in general , successfully implementation of evolutionary algorithms for solving a given problem depends on incorporated problem - specific knowledge .
as already mentioned before , all elements of evolutionary algorithms can be hybridized .
mostly , a hybridization addresses the following elements of evolutionary algorithms @xcite : * initial population , * genotype - phenotype mapping , * evaluation function , and * variation and selection operators .
first , problem - specific knowledge incorporated into heuristic procedures can be used for creating an initial population .
second , genotype - phenotype mapping is used by evolutionary algorithms , where the solutions are represented in an indirect way . in that cases ,
a constructing algorithm that maps the genotype representation into a corresponding phenotypic solution needs to be applied .
this constructor can incorporate various heuristic or other problem - specific knowledge .
third , to improve the current solutions by an evaluation function , local search heuristics can be used .
finally , problem - specific knowledge can be exploited by heuristic variation and selection operators .
the mentioned hybridizations can be used to hybridize the self - adaptive evolutionary algorithms as well . in the rest of chapter ,
we propose three kinds of hybridizations that was employed to the proposed hybrid self - adaptive evolutionary algorithms : * the construction heuristics that can be used by the genotype - phenotype mapping , * the local search heuristics that can be used by the evaluation function , and * the neutral survivor selection that incorporates the problem - specific knowledge . because the initialization of initial population is problem dependent
we omit it from our discussion .
usually , evolutionary algorithms are used for problem solving , where a lot of experience and knowledge is accumulated in various heuristic algorithms . typically , these algorithms work well on limited number of problems @xcite . on the other hand ,
evolutionary algorithms are a general method suitable to solve very different kinds of problems . in general , these algorithms are less efficient than heuristics specialized to solve the given problem . if we want to combine a benefit of both kind of algorithms then the evolutionary algorithm can be used for discovering new solutions that the heuristic exploits for building of new , probably better solutions .
construction heuristics build the solution of optimization problem incrementally , i.e. , elements are added to a solution step by step ( algorithm [ alg : const ] ) .
@xmath20 ; a local search belongs to a class of improvement heuristics @xcite . in our case ,
main characteristic of these is that the current solution is taken and improved as long as improvements are perceived .
the local search is an iterative process of discovering points in the vicinity of current solution .
if a better solution is found the current solution is replaced by it .
a neighborhood of the current solution @xmath21 is defined as a set of solutions that can be reached using an unary operator @xmath22 @xcite .
in fact , each neighbor @xmath23 in neighborhood @xmath24 can be reached from current solution @xmath21 in @xmath25 strokes .
therefore , this neighborhood is called @xmath26 neighborhood of current solution @xmath21 as well . for example , let the binary represented solution @xmath21 and @xmath27 operator on it are given . in that case , each of neighbors @xmath28 can be reached changing exactly one bit .
the neighborhood of this operator is defined as @xmath29 where @xmath30 denotes a hamming distance of two binary vectors as follows @xmath31 where operator @xmath32 means @xmath33 operation .
essentially , the hamming distance in equation [ eq : hamming ] is calculated by counting the number of different bits between vectors @xmath21 and @xmath23 .
the @xmath27 operator defines the set of feasible @xmath27 strokes while the number of feasible @xmath27 strokes determines the size of neighborhood .
@xmath34 ; @xmath35 ; @xmath36 ; as illustrated by algorithm [ alg : ls ] , the local search can be described as follows @xcite : * the initial solution is generated that becomes the current solution ( procedure @xmath37 ) . *
the current solution is transformed with @xmath26 strokes and the given solution @xmath23 is evaluated ( procedure @xmath38 ) . *
if the new solution @xmath23 is better than the current @xmath21 the current solution is replaced .
on the other hand , the current solution is kept .
* lines 2 to 7 are repeated until the set of neighbors is not empty ( procedure @xmath39 ) . in summary ,
the @xmath26 operator represents a basic element of the local search from which depends how exhaustive the neighborhood will be discovered .
therefore , the problem - specific knowledge needs to be incorporated by building of the efficient operator .
a genotype diversity is one of main prerequisites for the efficient self - adaptation .
the smaller genotypic diversity causes that the population is crowded in the search space . as a result
, the search space is exploited . on the other hand
, the larger genotypic diversity causes that the population is more distributed within the search space and therefore , the search space is explored @xcite .
explicitly , the genotype diversity of population is maintained with a proposed neutral survivor selection that is inspired by the neutral theory of molecular evolution @xcite , where the neutral mutation determines to the individual three possible destinies , as follows : * the fittest individual can survive in the struggle for existence , * the less fitter individual is eliminated by the natural selection , * individual with the same fitness undergo an operation of genetic drift , where its survivor is dependent on a chance .
each candidate solution represents a point in the search space .
if the fitness value is assigned to each feasible solution then these form a fitness landscape that consists of peeks , valleys and plateaus @xcite .
in fact , the peaks in the fitness landscape represents points with higher fitness , the valleys points with the lower fitness while plateaus denotes regions , where the solutions are neutral @xcite .
the concept of the fitness landscape plays an important role in evolutionary computation as well .
moreover , with its help behavior of evolutionary algorithms by solving the optimization problem can be understood .
if on the search space we look from a standpoint of fitness landscape then the heuristical algorithm tries to navigate through this landscape with aim to discover the highest peeks in the landscape @xcite .
however , to determine how distant one solution is from the other , some measure is needed .
which measure to use depends on a given problem . in the case of genetic algorithms , where we deal with the binary solutions , the hamming distance ( equation [ eq : hamming ] ) can be used .
when the solutions are represented as real - coded vectors an euclidian distance is more appropriate .
the euclidian distance between two vectors @xmath40 and @xmath21 is expressed as follows : @xmath41 and measures the root of quadrat differences between elements of vectors @xmath40 and @xmath21 .
the main characteristics of fitness landscapes that have a great impact on the evolutionary search are the following @xcite : * the fitness differences between neighboring points in the fitness landscape : to determine a ruggedness of the landscape , i.e. , more rugged as the landscape , more difficultly the optimal solution can be found ; * the number of peaks ( local optima ) in the landscape : the higher the number of peaks , the more difficulty the evolutionary algorithms can direct the search to the optimal solution ; * how the local optima are distributed in the search space : to determine the distribution of the peeks in the fitness landscape ; * how the topology of the basins of attraction influences on the exit from the local optima : to determine how difficult the evolutionary search that gets stuck into local optima can find the exit from it and continue with the discovering of the search space ; * existence of the neutral networks : the solutions with the equal value of fitness represent a plateaus in the fitness landscape . when the stochastic fitness function is used for evaluation of individuals the fitness landscape is changed over time .
in this way , the dynamic landscape is obtained , where the concept of fitness landscape can be applied , first of all , to analyze the neutral networks that arise , typically , in the matured generations .
to determine , how the solutions are dissipated over the search space some reference point is needed .
for this reason , the current best solution @xmath42 in the population is used .
this is added to the population of @xmath43 solutions .
an operation of the neutral survivor selection is divided into two phases . in the first phase ,
the evolutionary algorithm from the population of @xmath44 offsprings finds a set of neutral solutions @xmath45 that represents the best solutions in the population of offsprings .
if the neutral solutions are better than or equal to the reference , i.e. @xmath46 for @xmath47 , then reference solution @xmath42 is replaced with the neutral solution @xmath48 that is the most faraway from reference solution according to the equation 7 .
thereby , it is expected that the evolutionary search is directed to the new , undiscovered region of the search space . in the second phase , the updated reference solution @xmath42 is used to determine the next population of survivors .
therefore , all offsprings are ordered with regard to the ordering relation @xmath49 ( read : is better than ) as follows : @xmath50 where the ordering relation @xmath49 is defined as @xmath51 finally , for the next generation the evolutionary algorithm selects the best @xmath43 offsprings according to the equation [ eq : order ] .
these individuals capture the random positions in the next generation . likewise
the neutral theory of molecular evolution , the neutral survivor selection offers to the offsprings three possible outcomes , as follows .
the best offsprings survive .
additionally , the offspring from the set of neutral solutions that is far away of reference solution can become the new reference solution .
the less fitter offsprings are usually eliminated from the population .
all other solutions , that can be neutral as well , can survive if they are ordered on the first @xmath43 positions regarding to equation [ eq : order ] .
in this section an implementation of the hybrid self - adaptive evolutionary algorithms ( hsa - ea ) for solving combinatorial optimization problems is represented .
the implementation of this algorithm in practice consists of the following phases : * finding the best heuristic that solves the problem on a traditional way and adapting it to use by the self - adaptive evolutionary algorithm , * defining the other elements of the self - adaptive evolutionary algorithm , * defining the suitable local search heuristics , and * including the neutral survivor selection .
the main idea behind use of the construction heuristics in the hsa - ea is to exploit the knowledge accumulated in existing heuristics .
moreover , this knowledge is embedded into the evolutionary algorithm that is capable to discover the new solutions . to work simultaneously
both algorithms need to operate with the same representation of solutions .
if this is not a case a decoder can be used .
the solutions are encoded by the evolutionary algorithm as the real - coded vectors and decoded before the construction of solutions .
the whole task is performed in genotype - phenotype mapping that is illustrated in fig .
[ pic:3 ] .
the genotype - phenotype mapping consists of two phases as follows : * decoding , * constructing .
evolutionary algorithms operate in genotypic search space , where each genotype consists of real - coded problem variables and control parameters . for encoded solution
only the problem variables are taken .
this solution is further decoded by decoder into a decoded solution that is appropriate for handling of a construction heuristic .
finally , the construction heuristic constructs the solution within the original problem context , i.e. , problem solution space .
this solution is evaluated by the suitable evaluation function .
the other elements of self - adaptive evolutionary algorithm consists of : * evaluation function , * population model , * parent selection mechanism , * variation operators ( mutation and crossover ) , and * initialization procedure and termination condition .
the evaluation function depends on a given problem .
the self - adaptive evolutionary algorithm uses the population model @xmath52 , where the @xmath44 offsprings is generated from the @xmath43 parents .
however , the parents that are selected with tournament selection @xcite are replaced by the @xmath43 the best offsprings according to the appropriate population model .
the ratio @xmath53 is used for the efficient self - adaptation @xcite .
typically , the normal uncorrelated mutation with @xmath4 step sizes , discrete and arithmetic crossover are used by the hsa - ea .
normally , the probabilities of mutation and crossover are set according to the given problem .
selection of the suitable local search heuristics that improve the current solution is a crucial for the performance of the hsa - ea . on the other hand ,
the implementation of neutral survivor selection is straightforward .
finally , the scheme of the hsa - ea is represented in the algorithm [ alg : hsa - ea ] .
@xmath54 ; @xmath55 ; @xmath56 ; @xmath57 ; @xmath58 ; @xmath59 ; @xmath60 ; @xmath61 ; in the rest of the chapter we present the implementation of the hsa - ea for the graph 3-coloring .
this algorithm is hybridized with the dsatur @xcite construction heuristic that is well - known traditional heuristic for the graph 3-coloring .
graph 3-coloring can be informally defined as follows .
let assume , an undirected graph @xmath62 is given , where @xmath63 denotes a finite set of vertices and @xmath64 a finite set of unordered pairs of vertices named edges @xcite .
the vertices of graph @xmath65 have to be colored with three colors such that no one of vertices connected with an edge is not colored with the same color .
graph 3-coloring can be formalized as constraint satisfaction problem ( csp ) that is denoted as a pair @xmath66 , where @xmath67 denotes a free search space and @xmath68 a boolean function on @xmath67 .
the free search space denotes the domain of candidate solutions @xmath69 and does not contain any constraints , i.e. , each candidate solution is feasible .
the function @xmath68 divides the search space @xmath67 into feasible and unfeasible regions .
the solution of constraint satisfaction problem is found when all constraints are satisfied , i.e. , when @xmath70 . however , for the 3-coloring of graph @xmath62 the free search space @xmath67 consists of all permutations of vertices @xmath71 for @xmath72 . on the other hand , the function @xmath68 ( also feasibility condition ) is composed of constraints on vertices .
that is , for each vertex @xmath71 the corresponding constraint @xmath73 is defined as the set of constraints involving vertex @xmath74 , i.e. , edges @xmath75 for @xmath76 connecting to vertex @xmath74 .
the feasibility condition is expressed as conjunction of all constraints @xmath77 .
handling direct constraints in evolutionary algorithms is not straightforward . to overcome this problem ,
the constraint satisfaction problems are , typically , transformed into unconstrained ( also free optimization problem ) by the sense of a penalty function .
the more an infeasible solution is far away from feasible region , the higher is the penalty .
moreover , this penalty function can act as an evaluation function by the evolutionary algorithm . for graph 3-coloring
it can be expressed as @xmath78 where the function @xmath79 is defined as @xmath80 note that all constraints in solution @xmath69 are satisfied , i.e. , @xmath81 if and only if @xmath82 . in this way , the equation [ eq : fit ] represents the feasibility condition and can be used to estimate the quality of solution @xmath69 in the permutation search space .
the permutation @xmath40 determines the order in which the vertices need to be colored .
the size of the search space is huge , i.e. , @xmath83 .
as can be seen from equation [ eq : fit ] , the evaluation function depends on the number of constraint violations , i.e. , the number of uncolored vertices .
this fact causes that more solutions can have the same value of the evaluation function .
consequently , the large neutral networks can arise @xcite .
however , the neutral solutions are avoided if the slightly modified evaluation function is applied , as follows : @xmath84 where @xmath85 represents the weight .
higher than the value of weights harder the appropriate vertex is to color .
the hybrid self - adaptive evolutionary algorithm is hybridized with the dsatur @xcite construction heuristic and the local search heuristics .
in addition , the problem specific knowledge is incorporated by the initialization procedure and the neutral survivor selection . in this section
we concentrate , especially , on a description of those elements in evolutionary algorithm that incorporate the problem specific knowledge . that are : * the initialization procedure , * the genotype - phenotype mapping , * local search heuristics and * the neutral survivor selection .
the other elements of this evolutionary algorithm , as well as neutral survivor selection , are common and therefore , discussed earlier in the chapter .
+ + _ the initialization procedure _ + initially , original dsatur algorithm orders the vertices @xmath86 for @xmath72 of a given graph @xmath65 descendingly according to the vertex degrees denoted by @xmath87 that counts the number of edges that are incident with the vertex @xmath74 @xcite . to simulate behavior of the original dsatur algorithm @xcite ,
the first solution in the population is initialized as follows : @xmath88 because the genotype representation is mapped into a permutation of weights by decoder the same ordering as by original dsatur is obtained , where the solution can be found in the first step .
however , the other @xmath89 solutions in the population are initialized randomly .
+ + _ the genotype - phenotype mapping _ + as illustrated in fig .
[ pic:3 ] , the solution is represented in genotype search space as tuple @xmath90 , where problem variables @xmath13 for @xmath72 denote how hard the given vertex is to color and control parameters @xmath91 for @xmath72 mutation steps of normal mutation .
a decoder decodes the problem variables into permutation of vertices and corresponding weights .
however , all feasible permutation of vertices form the permutation search space .
the solution in this search space is represented as tuple @xmath92 , where variables @xmath74 for @xmath72 denote the permutation of vertices and variables @xmath85 corresponding weights .
the vertices are ordered into permutation so that vertex @xmath74 is predecessor of vertex @xmath93 if and only if @xmath94 .
values of weights @xmath85 are obtained by assigning the corresponding values of problem variables , i.e. @xmath95 for @xmath72 .
finally , dsatur construction heuristic maps the permutation of vertices and corresponding weights into phenotypic solution space that consists of all possible 3-colorings @xmath96 .
note that the size of this space is @xmath97 .
dsatur construction heuristic acts like original dsatur algorithm @xcite , i.e. it takes the permutation of vertices and color these as follows : * the heuristic selects a vertex with the highest saturation , and colors it with the lowest of the three colors ; * in the case of a tie , the heuristic selects a vertex with the maximal weight ; * in the case of a tie , the heuristic selects a vertex randomly .
the main difference between this heuristic and the original dsatur algorithm is in the second step where the heuristic selects the vertices according to the weights instead of degrees .
+ + _ local search heuristics _ + the current solution is improved by a sense of local search heuristics . at each evaluation of solution
the best neighbor is obtained by acting of the following original local search heuristics : * inverse , * ordering by saturation , * ordering by weights , and * swap .
the evaluation of solution is presented in algorithm [ alg : eval ] from which it can be seen that the local search procedure ( @xmath98 ) is iterated until improvements are perceived . however ,
this procedure implements all four mentioned local search heuristics .
the best neighbor is generated from the current solution by local search heuristics with @xmath25-exchanging of vertices . in the case , the best neighbor is better than the current solution the later is replaced by the former .
@xmath99 ; @xmath100 ; @xmath101 ; @xmath102 ; @xmath36 ; @xmath103 ; @xmath104 ; in the rest of the subsection , an operation of the local search heuristics is illustrated in fig .
[ pic : inverse]-[pic : swap ] by samples , where a graph with nine vertices is presented .
the graph is composed of a permutation of vertices @xmath105 , corresponding coloring @xmath106 , weights @xmath107 and saturation degrees @xmath108 .
the inverse local search heuristic finds all uncolored vertices in a solution and inverts their order . as can be shown in fig .
[ pic : inverse ] , the uncolored vertices 4 , 6 and 8 are shadowed .
the best neighbor is obtained by inverting of their order as is presented on right - hand side of this figure .
the number of vertex exchanged is dependent of the number of uncolored vertices ( @xmath26 neighborhood ) .
the ordering by saturation local search heuristic acts as follows .
the first uncolored vertex is taken at the first . to this vertex a set of adjacent vertices are selected .
then , these vertices are ordered descending with regard to the values of saturation degree .
finally , the adjacent vertex with the highest value of saturation degree in the set of adjacent vertices is swapped with the uncolored vertex . here
, the simple @xmath27 neighborhood of current solution is defined by this local search heuristic . in the example on fig .
[ pic : order ] the first uncolored vertex 4 is shadowed , while its adjacent vertices 1 , 6 and 7 are hatched .
however , the vertices 1 and 7 have the same saturation degree , therefore , the vertex 7 is selected randomly .
finally , the vertices 4 in 7 are swapped ( right - hand side of fig .
[ pic : order ] ) .
when ordering of weights , the local search heuristic takes the first uncolored vertex and determines a set of adjacent vertices including it .
this set of vertices is then ordered descending with regard to the values of weights .
this local search heuristic determines the @xmath26 neighborhood of current solution , where @xmath25 is dependent of a degree of the first uncolored vertex . as illustrated by fig .
[ pic : order2 ] , the uncolored vertex 4 is shadowed , while its adjacent vertices 1 , 6 and 7 are hatched .
the appropriate ordering of the selected set of vertices is shown in the right - hand of fig .
[ pic : order2 ] after the operation of the local search heuristic .
the swap local search heuristic finds the first uncolored vertex and descendingly orders the set of all predecessors in the solution according to the saturation degree .
then , the uncolored vertex is swapped with the vertex from the set of predecessors with the highest saturation degree .
when more vertices with the same highest saturation degree are arisen , the subset of these vertices is determined .
the vertex from this subset is then selected randomly .
therefore , the best neighbor of the current solution is determined by an exchange of two vertices ( @xmath27 neighborhood ) . as illustrated in fig .
[ pic : swap ] , the first uncolored vertex 4 is shadowed , while the vertices 0 and 4 that represent the subset of vertices with the highest saturation are hatched .
in fact , the vertex 0 is selected randomly and the vertices 0 and 4 are swapped as is presented in right - hand of fig .
[ pic : swap ] .
the goal of this subsection is twofold . at the first
, an influence of the local search heuristics on results of the hsa - ea is analyzed in details .
further , a comparison of the hsa - ea hybridized with the neutral survivor selection and the hsa - ea with the deterministic selection is made . in this context , the impact of the heuristic initialization procedure are taken into consideration as well .
characteristics of the hsa - ea used in experiments were as follows .
the normal distributed mutation was employed and applied with mutation probability of 1.0 .
the crossover was not used .
the tournament selection with size 3 selects the parents for mutation .
the population model @xmath109 was suitable for the self - adaptation because the ratio between parents and generated offspring amounted to @xmath110 as recommended by @xcite .
as termination condition , the maximum number of evaluations to solution was used .
fortunately , the average number of evaluations to solution ( @xmath111 ) that counts the number of evaluation function calls was employed as the performance measure of efficiency .
in addition , the average number of uncolored nodes ( @xmath112 ) was employed as the performance measure of solution quality .
this measure was applied when the hsa - ea does not find the solution and counts the number of uncolored vertices .
nevertheless , the success rate ( @xmath113 ) was defined as the primary performance measure and expressed as the ratio between the runs in which the solution was found and all performed runs .
the @xcite random graph generator was employed for generation of random graphs that constituted the test suite .
it is capable to generate the graphs of various types , number of vertices , edge densities and seeds of random generator . in this study we concentrated on the equi - partite type of graphs .
this type of graphs is not the most difficult to color but difficult enough for many existing algorithms @xcite .
the random graph generator divides the vertices of graph into three color sets before generating randomly . in sense of equi - partite random graph , these color sets are as close in size as possible .
all generated graphs consisted of @xmath114 vertices .
an edge density is controlled by parameter @xmath115 of the random graph generator that determines probability that two vertices @xmath74 and @xmath116 in the graph @xmath65 are connected with an edge @xmath117 @xcite . however , if @xmath115 is small the graph is not connected because the edges are sparse .
when @xmath115 is increased the number of edges raised and the graph becomes interconnected . as a result , the number of constraints that needs to be satisfied by the coloring algorithm increases until suddenly the graph becomes uncolorable .
this occurrence depends on a ratio between the number of edges and the number of vertices .
the ratio is referred to as the threshold @xcite .
that is , in the vicinity of the threshold the vertices of the random generated graph becomes hard to color or even the graph becomes uncolorable .
fortunately , the graph instances with this ratio much higher that the threshold are easy to color because these graphs are densely interconnected .
therefore , many global optima exist in the search space that can be discovered easy by many graph 3-coloring algorithms .
interestingly , for random generated graphs the threshold arises near to the value 2.35 @xcite .
for example , the equi - partite graph generated with number of vertices @xmath118 and the edge density determined by @xmath119 consists of 2,366 edges .
because the ratio @xmath120 is near to the threshold , we can suppose that this instance of graph is hard to color .
the seed @xmath121 of random graph generator determines which of the two vertices @xmath74 and @xmath116 are randomly drawn from different 3-color sets to form an edge @xmath117 but it does not affect the performance of the graph 3-coloring algorithm @xcite . in this study , the instances of random graphs with seed @xmath122 were employed . to capture a phenomenon of the threshold ,
the parameter @xmath115 by generation of the equi - partite graphs was varied from @xmath123 to @xmath124 in a step of @xmath125 . in this way ,
the test suite consisted of 15 instances of graphs , in which the hardest graph with @xmath119 was presented as well . in fact , the evolutionary algorithm was applied to each instance @xmath126 times and the average results of these runs were considered . + + _ the impact of the local search heuristics _
+ in this experiments , the impact of four implemented local search heuristics on results of the hsa - ea was taken into consideration .
results of the experiments are illustrated in the fig .
[ fig : sub_3 ] that is divided into six graphs and arranged according to the particular measures @xmath113 , @xmath111 and @xmath112 .
the graphs on the left side of the figure , i.e. 8.a , 8.c and 8.e , represent a behavior of the hsa - ea hybridized with four different local search heuristics .
this kind of the hsa - ea is referred to as original hsa - ea in the rest of chapter .
a seen by the fig .
[ fig : sub_3].a , no one of the hsa - ea versions was succeed to solve the hardest instance of graph with @xmath119 .
the best results in the vicinity of the threshold is observed by the hsa - ea hybridizing with the ordering by saturation local search heuristic ( @xmath127 by @xmath128 ) .
the overall best performance is shown by the hsa - ea using the swap local search heuristic .
although the results of this algorithm is not the best by instances the nearest to the threshold ( @xmath129 by @xmath128 ) , this local search heuristic outperforms the other by solving the remaining instances in the collection . in average , results according to the @xmath111 ( fig .
[ fig : sub_3].c ) show that the hsa - ea hybridized with the swap local search heuristic finds the solutions with the smallest number of the fitness evaluations .
however , troubles are arisen in the vicinity of the threshold , where the hsa - ea with other local search heuristics are faced with the difficulties as well . moreover , at the threshold the hsa - ea hybridizing with all the used local search heuristics reaches the limit of 300,000 allowed function evaluations . the hsa - ea hybridizing with the ordering by saturation local search heuristic demonstrates the worst results according to the @xmath112 , as presented in the fig .
[ fig : sub_3].e .
the graph instance by @xmath130 was exposed as the most critical by this algorithm ( @xmath131 ) although this is not the closest to the threshold . in average ,
when the hsa - ea was hybridized with the other local search heuristics than the ordering by saturation , all instances in the collection were solved with less than 20 uncolored vertices . in the right side of the fig .
[ fig : sub_3 ] , results of different versions of the hsa - ea are collected . the first version that is designated as @xmath132 operates with the same parameters as the original hsa - ea but without the local search heuristics .
the label @xmath133 in this figure indicates the original version of the hsa - ea .
finally , the label @xmath134 denotes the original version of the hsa - ea with the exception of initialization procedure . while all considered versions of the hsa - ea uses the heuristic initialization procedure this version of the algorithm employs the pure random initialization .
note , in the figure , results for this version of the hsa - ea were obtained after 25 runs , while for the versions of the hsa - ea with the local search heuristics the average results were obtained after performing of all four local search heuristics , i.e. after @xmath135 runs . in fig .
[ fig : sub_3].b results of different versions of the hsa - ea according to the @xmath113 are presented .
the best results by the instances the nearest to the threshold ( @xmath136 $ ] ) are observed by the original hsa - ea .
conversely , the hsa - ea with the random initialization procedure ( @xmath134 ) gained the worst results by the instances the nearest to the threshold , while these were better while the edge density was raised regarding the original hsa - ea .
the turning point represents the instance of graph with @xmath137 .
after this point is reached the best results were overtaken by the hsa - ea with the random initialization procedure ( @xmath134 ) . in contrary ,
the best results by the instances the nearest to the threshold according to the @xmath111 was observed by the hsa - ea without local search heuristics ( @xmath132 ) . here ,
the turning point regarding the performance of the hsa - ea ( @xmath137 ) was observed as well .
after this point results of the hsa - ea without local search heuristics becomes worse .
conversely , the hsa - ea with random initialization procedure that was the worst by the instances before the turning point becomes the best after this . as illustrated by fig .
[ fig : sub_3].f , all versions of the hsa - ea leaved in average less than @xmath138 uncolored vertices by the 3-coloring .
the bad result by the original hsa - ea coloring the graph with @xmath130 was caused because of the ordering by saturation local search heuristic that got stuck in the local optima .
nevertheless , note that most important measure is @xmath113 .
+ + _ the impact of the neutral survivor selection _ + in this experiments the impact of the neutral survivor selection on results of the hsa - ea was analyzed . in this context , the hsa - ea with deterministic survivor selection was developed with the following characteristic : * the equation [ eq : fit1 ] that prevents the generation of neutral solutions was used instead of the equation [ eq : fit ] . *
the deterministic survivor selection was employed instead of the neutral survivor solution .
this selection orders the solutions according to the increasing values of the fitness function . in the next generation
the first @xmath43 solutions is selected to survive . before starting with the analysis
, we need to prove the existence of neutral solution and to establish they characteristics .
therefore , a typical run of the hsa - ea with neutral survivor selection is compared with the typical run of the hsa - ea with the deterministic survivor selection .
as example , the 3-coloring of the equi - partite graph with @xmath139 was taken into consideration .
this graph is easy to solve by both versions of the hsa - ea .
characteristics of the hsa - ea by solving it are presented in fig .
[ fig : sub_4 ] . in the fig .
[ fig : sub_4].a the best and the average number of uncolored nodes that were achieved by the hsa - ea with neutral and the hsa - ea with deterministic survivor selection are presented .
the figure shows that the hsa - ea with the neutral survivor selection converge to the optimal solution very fast . to improve the number of uncolored nodes from 140 to 10 only 10,000 solutions to evaluation
were needed .
after that , the improvement stagnates ( slow progress is detected only ) until the optimal solution is found .
the closer look at the average number of uncolored nodes indicates that this value changed over every generation .
typically , the average fitness value is increased when the new best solution is found because the other solutions in the population try to adapt itself to the best solution .
this self - adaptation consists of adjusting the step sizes that from larger starting values becomes smaller and smaller over the generations until the new best solution is found .
the exploring of the search space is occurred by this adjusting of the step sizes .
conversely , the average fitness values are changed by the hsa - ea in the situations where the best values are not found as well .
the reason for that behavior is the stochastic evaluation function that can evaluate the same permutation of vertices always differently .
more interestingly , the neutral solution occurs when the average fitness values comes near to the best ( fig .
[ fig : sub_4].b ) . as illustrated by this figure ,
the most neutral solutions arise in the later generations when the population becomes matured .
in example from fig . [
fig : sub_4].b , the most neutral solutions occurred after 20,000 and 30,000 evaluations of fitness function , where almost @xmath140 of neutral solution occupied the current population . in contrary ,
the hsa - ea with deterministic survivor selection starts with the lower number of uncolored vertices ( fig .
[ fig : sub_4].c ) than the hsa - ea with neutral selection
. however , the convergence of this algorithm is slower than by its counterpart with the neutral selection . a closer look at the average fitness value uncovers that the average fitness value never come close to the best fitness value . a falling and the rising of the average fitness values
are caused by the stochastic evaluation function .
in the fig . [
fig : sub_4].d a diversity of population as produced by the hsa - ea with different survivor selections is presented .
the diversity of population is calculated as a standard deviation of the vector consisting of the mean weight values in the population .
both hsa - ea from this figure lose diversity of the initial population ( close to value 8.0 ) very fast .
the diversity falls under the value 1.0 . over the generations
this diversity is raised until it becomes stable around the value 1.0 . here , the notable differences between curves of both hsa - ea are not observed . to determine what impact the neutral survivor selection has on results of the hsa - ea
, a comparison between results of the hsa - ea with neutral survivor selection ( @xmath141 ) and the hsa - ea with deterministic survivor selection ( @xmath142 ) was done .
however , both versions of the hsa - ea run without local search heuristics . results of these are represented in the fig .
[ fig : sub_5 ] . as reference point
, the results of the original hsa - ea hybridized with the swap local search heuristic ( @xmath143 ) that obtains the overall best results are added to the figure .
the figure is divided in two graphs where the first graph ( fig .
[ fig : sub_5].a ) presents results of the hsa - ea with heuristic initialization procedure and the second graph ( fig .
[ fig : sub_5].b ) results of the hsa - ea with random initialization procedure according to the @xmath113 . as shown by the fig .
[ fig : sub_5].a the hsa - ea with neutral survivor selection ( @xmath141 ) exposes better results by the instances near to the threshold ( @xmath136 $ ] ) while the hsa - ea with deterministic survivor selection ( @xmath142 ) was slightly better by the instance of graph with @xmath144 .
interestingly , while the curve of the former regularly increases the curve of the later is sawing because it raises and falls from the instance to the instance . in contrary , from the fig .
[ fig : sub_5].b it can be seen that the hsa - ea with neutral survivor selection outperforms its counterpart with deterministic survivor selection by all instances of random graphs if the random initialization procedure is applied . in summary ,
the original hsa - ea with swap local search heuristic used as reference outperforms all observed versions of the hsa - ea . in this subsection
the characteristics of the hsa - ea were studied on the collection of equi - partite graphs , where we focused on the behavior of the algorithm in the vicinity of the threshold .
therefore , an impact of the hybridizing elements , like the initialization procedure , the local search heuristics , and neutral survivor selection , on results of the hsa - ea are compared .
the results of these comparisons in vicinity of the threshold ( @xmath145 $ ] ) are presented in table [ tab : summary ] , where these are arranged according to the applied selection ( column @xmath146 ) , the local search heuristics ( column @xmath133 ) and initialization procedure ( column @xmath134 ) . in column
@xmath113 average results of the corresponding version of the hsa - ea are presented .
additionally , the column @xmath147 denotes the averages of the hsa - ea using both kind of initialization procedure .
finally , the column @xmath148 represents the average results according to @xmath113 that are dependent on the different kind of survivor selection only .
.average results of various versions of the hsa - ea according to the sr [ cols="<,<,<,^,^,^",options="header " , ] as shown by the table [ tab : summary ] , results of the hsa - ea with deterministic survivor selection without local search heuristics and without random initialization procedure ( @xmath149 , denoted as @xmath150 ) were worse than results or its counterpart with neutral survivor selection ( @xmath151 , denoted as @xmath152 ) in average for more than 10.0% . moreover , the local search heuristics improved results of the hsa - ea with neutral survivor selection and random initialization procedure from @xmath151 ( denoted as @xmath152 ) to @xmath153 ( denoted as @xmath154 ) that amounts to almost 10.0% . finally , the heuristic initialization improved results of the hsa - ea with neutral selection and with local search heuristics from @xmath153 ( denoted as @xmath154 ) to @xmath155 ( denoted as @xmath156 ) , i.e. for 1.5% .
note that the @xmath155 represents the best result that was found during the experimentation . in summary ,
the construction heuristics has the most impact on results of the hsa - ea .
that is , the basis of the graph 3-coloring represents the self - adaptive evolutionary algorithm with corresponding construction heuristic .
however , to improve results of this base algorithm additional hybrid elements were developed . as evident , the local search heuristics
improves the base algorithm for @xmath157 , the neutral survivor selection for another @xmath157 and finally the heuristic initialization procedure additionally @xmath158 .
evolutionary algorithms are a good general problem solver but suffer from a lack of domain specific knowledge .
however , the problem specific knowledge can be added to evolutionary algorithms by hybridizing different parts of evolutionary algorithms . in this chapter , the hybridization of search and selection operators are discussed . the existing heuristic function that constructs the solution of the problem in a traditional way can be used and embedded into the evolutionary algorithm that serves as a generator of new solutions . moreover
, the generation of new solutions can be improved by local search heuristics , which are problem specific . to hybridized selection operator
a new neutral selection operator has been developed that is capable to deal with neutral solutions , i.e. , solutions that have the different genotype but expose the equal values of objective function .
the aim of this operator is to directs the evolutionary search into new undiscovered regions of the search space , while on the other hand exploits problem specific knowledge . to avoid wrong setting of parameters that control the behavior of the evolutionary algorithm , the self - adaptation is used as well .
such hybrid self - adaptive evolutionary algorithms have been applied to the the graph 3-coloring that is well - known np - complete problem .
this algorithm was applied to the collection of random graphs , where the phenomenon of a threshold was captured .
a threshold determines the instanced of random generated graphs that are hard to color .
extensive experiments shown that this hybridization greatly improves the results of the evolutionary algorithms . furthermore , the impact of the particular hybridization is analyzed in details as well . in continuation of work
the graph @xmath25-coloring will be investigated . on the other hand
, the neutral selection operator needs to be improved with tabu search that will prevent that the reference solution will be selected repeatedly .
barnett , l. 1998 . ruggedness and neutrality - the nkp family of fitness landscapes , _ in _ c. adami , r. belew , h. kitano c. taylor ( eds ) , _ alife vi : sixth international conference on articial life _ , mit press , pp . 1827 .
culberson , j. luo , f. 2006 . exploring the k - colorable landscape with iterated greedy , _ in _
d. johnson m. trick ( eds ) , _ cliques , coloring and satisfiability : second dimacs implementation challenge _ , american mathematical society , rhode island , pp .
245284 .
doerr , b. , eremeev , a. , horoba , c. , neumann , f. theile , m. 2009 .
evolutionary algorithms and dynamic programming , _ gecco 09 : proceedings of the 11th annual conference on genetic and evolutionary computation _ , acm , new york , ny , usa , pp
. 771778 .
eberhart , r. kennedy , j. 1995 . a new optimizer using particle swarm theory , _ proceedings of 6th international symposium on micro machine and human science _ , ieee service center , piscataway , nj , nagoya , pp .
3943 .
fleurent , c. ferland , j. 1994 .
genetic hybrids for the quadratic assignment problems , _ in _
p. pardalos h. wolkowicz ( eds ) , _ quadratic assignment and related problems _
, dimacs series in discrete mathematics and theoretical computer science , ams : providence , rhode island , pp . 190206 .
ganesh , k. punniyamoorthy , m. 2004 .
optimization of continuous - time production planning using hybrid genetic algorithms - simulated annealing , _ international journal of advanced manufacturing technology _ * 26 * : 148154 .
grosan , c. abraham , a. 2007 .
hybrid evolutionary algorithms : methodologies , architectures , and reviews , _ in _ c. grosan , a. abraham h. ishibuchi ( eds ) , _ hybrid evolutionary algorithms _ , springer - verlag , berlin , pp .
117 .
herrera , f. lozano , m. 1996 .
adaptation of genetic algorithm parameters based on fuzzy logic controllers , _ in _
f. herrera j. verdegay ( eds ) , _ genetic algorithms and soft computing _ , physica - verlag hd , pp . 95125 .
lee , m. takagi , h. 1993 .
dynamic control of genetic algorithms using fuzzy logic techniques , _ in _
s. forrest ( ed . ) , _ proceedings of the 5th international conference on genetic algorithms _ , morgan kaufmmann , san mateo , pp .
7683 .
liu , s .- h . ,
mernik , m. bryant , b. 2009 . to explore or to exploit : an entropy - driven approach for evolutionary algorithms , _ international journal of knowledge - based and intelligent engineering systems _ * 13 * : 185206 .
meyer - nieberg , s. beyer , h .-
self - adaptation in evolutionary algorithms , _ in _
f. lobo , c. lima z. michalewicz ( eds ) , _ parameter setting in evolutionary algorithms _ , springer - verlag , berlin , pp . | evolutionary algorithms are good general problem solver but suffer from a lack of domain specific knowledge .
however , the problem specific knowledge can be added to evolutionary algorithms by hybridizing .
interestingly , all the elements of the evolutionary algorithms can be hybridized . in this chapter
, the hybridization of the three elements of the evolutionary algorithms is discussed : the objective function , the survivor selection operator and the parameter settings . as an objective function ,
the existing heuristic function that construct the solution of the problem in traditional way is used .
however , this function is embedded into the evolutionary algorithm that serves as a generator of new solutions .
in addition , the objective function is improved by local search heuristics .
the new neutral selection operator has been developed that is capable to deal with neutral solutions , i.e. solutions that have the different representation but expose the equal values of objective function .
the aim of this operator is to directs the evolutionary search into a new undiscovered regions of the search space . to avoid of wrong setting of parameters that control the behavior of the evolutionary algorithm , the self - adaptation is used
finally , such hybrid self - adaptive evolutionary algorithm is applied to the two real - world np - hard problems : the graph 3-coloring and the optimization of markers in the clothing industry .
extensive experiments shown that these hybridization improves the results of the evolutionary algorithms a lot . furthermore , the impact of the particular hybridizations is analyzed in details as well . _ to cite paper as follows : iztok fister , marjan mernik and janez brest ( 2011 ) .
hybridization of evolutionary algorithms , evolutionary algorithms , eisuke kita ( ed . ) , isbn : 978 - 953 - 307 - 171 - 8 , intech , available from : http://www.intechopen.com/books/evolutionary-algorithms/hybridization-of-evolutionary-algorithms_ |
we consider the non - negative laplace - beltrami operator @xmath3 on the torus @xmath4 , seen as a two - dimensional riemannian manifold , with @xmath5\,$ ] .
the eigenvalues of @xmath3 are given by @xmath6 with @xmath7 , and an associated basis of eigenfunctions is given , in the standard coordinates , by @xmath8 @xmath9 @xmath10 @xmath11 to be more precise , the family consisting of all the above functions that are non - zero is an orthogonal basis of @xmath12 .
let us note that the eigenspace associated with the eigenvalue @xmath13 is spanned by all the functions in this basis such that the corresponding pair of indices @xmath14 satisfies @xmath15 .
if @xmath16 is a rational number , a large eigenvalues can have a very high multiplicity , and an associated eigenfunction can possess a very complex nodal structure ( see for instance @xcite ) .
we recall that for any eigenfunction @xmath17 of @xmath18 , we call _ nodal set _ the closed set @xmath19 and _ nodal domain _ a connected component of @xmath20
. we will prove the following statements .
[ thmeven ] if @xmath16 is irrational or @xmath21 , any non - constant eigenfunction @xmath17 of @xmath3 has an even number of nodal domains .
more precisely , we can divide the nodal domains of @xmath17 into pairs of isometric domains , @xmath17 being positive on one domain of each pair and negative on the other . [ prop3dom ]
if @xmath22 , there exists an eigenfunction of @xmath3 with three nodal domains . in @xcite ,
t. hoffmann - ostenhof asked if there exists a torus @xmath23 , with @xmath24\,$ ] , for which some eigenfunction of the laplacian has an odd number of nodal domains , at least equal to three .
proposition [ prop3dom ] answers the question positively , while theorem [ thmeven ] shows that such an eigenfunction does not exist when @xmath16 is irrational or @xmath25 .
[ [ acknowledgements ] ] acknowledgements + + + + + + + + + + + + + + + + i thank bernard helffer for introducing me to this problem and for numerous discussions and corrections .
i thank thomas hoffmann - ostenhof and susanna terracini for their advice and encouragements .
this work was partially supported by the anr ( agence nationale de la recherche ) , project optiform n@xmath26 anr-12-bs01 - 0007 - 02 , and by the erc , project compat n@xmath26 erc-2013-adg .
let us outline the method we will use to prove theorem [ thmeven ] .
let us first note that to any vector @xmath27 , we can associate a bijection @xmath28 from @xmath29 to itself .
it is defined in the following way : if @xmath30 in the standard coordinates , @xmath31 .
we will prove the following result .
[ propas ] if @xmath16 is irrational or @xmath21 , and if @xmath17 is a non - constant eigenfunction of @xmath18 on @xmath32 , there exists @xmath33 such that @xmath34 for all @xmath35 .
let us show that proposition [ propas ] implies theorem [ thmeven ] .
an eigenfunction @xmath17 being given , we define the bijection @xmath36 from @xmath37 to itself .
it is an isometry that preserves @xmath38 , and exchanges the nodal domains on which @xmath17 is positive with those on which @xmath17 is negative .
this proves theorem [ thmeven ] .
let us now turn to the proof of proposition [ propas ] .
let us first consider the case where @xmath16 is irrational , and let @xmath13 be a non - zero eigenvalue of @xmath3 . since @xmath16 is irrational there exists a unique pair of integers @xmath14 , different from @xmath39 , such that @xmath15 .
the eigenspace associated with @xmath13 is therefore spanned by the functions @xmath40 , @xmath41 , @xmath42 , and @xmath43 .
let us assume that @xmath44 and let us set @xmath45 .
it is then immediate to check that , for all @xmath46 in @xmath29 , @xmath47 when @xmath17 is any of the basis functions @xmath40 , @xmath41 , @xmath42 , and @xmath43 . as a consequence we still have @xmath47 when @xmath17 is any linear combination of the previous basis functions , that is to say any eigenfunction associated with @xmath13 .
if @xmath48 , we have @xmath49 and the same holds true with @xmath50 .
this conclude the proof of proposition [ propas ] in the irrational case .
let us now consider the case @xmath21 . as in the previous case , we will prove a statement that is slightly more precise than proposition [ propas ] : we will exhibit , for any non - zero eigenvalue @xmath13 , a vector @xmath51 such that @xmath47 for every eigenfunction @xmath17 associated with @xmath13 ( see lemma [ lemeigfunct ] ) . the difference in this case is that the equality @xmath52 can be satisfied for several pairs of integers @xmath14 . to overcome this difficulty , we will need the following simple arithmetical lemma .
this result is stated and proved in @xcite , where it is used to solve a closely related problem : proving that a non - constant eigenfunction of the laplacian on the square with a neumann or a periodic boundary condition must take the value @xmath53 on the boundary .
we nevertheless give a proof of the lemma here for the sake of completeness .
[ lemarithm ] let @xmath14 be a pair of non - negative integers , with @xmath54 , and let us write @xmath55 . if @xmath56 with @xmath57 , then @xmath58 and @xmath59 , where exactly one of the integers @xmath60 and @xmath61 is odd .
if on the other hand @xmath62 with @xmath57 , then @xmath63 and @xmath59 , where both integers @xmath60 and @xmath61 are odd . from the decomposition into prime factors , we deduce that we can write any positive integer @xmath64 as @xmath65 , with @xmath66 a non - negative and @xmath67 an odd integer .
let us first consider the case where @xmath68 or @xmath69 is zero . without loss of generality , we can assume that @xmath70 .
we write @xmath71 .
we are in the case @xmath56 with @xmath72 and @xmath73 , and we obtain the desired result by setting @xmath74 ( odd ) and @xmath75 ( even ) .
we now assume that both @xmath68 and @xmath69 are positive .
we write @xmath76 and @xmath77 with @xmath78 and @xmath79 odd integers . without loss of generality
, we can assume that @xmath80 .
we find @xmath81 .
if @xmath82 , then @xmath83 is an odd integer , and we have @xmath84 , with @xmath72 and @xmath85 . in that case , we set @xmath74 ( odd ) and @xmath86 ( even ) . if @xmath87 , we find @xmath88 .
we have furthermore @xmath89 and @xmath90 , and therefore @xmath91 .
we have @xmath62 with @xmath72 and @xmath92 , and we set @xmath74 and @xmath93 .
[ lemeigfunct ] let @xmath13 be a non - zero eigenvalue of @xmath18 on @xmath94 . a. if @xmath84 , we set @xmath95 , and we have @xmath47 for every eigenfunction @xmath17 associated with @xmath96 . b. if @xmath97 , we set @xmath98 , and we have @xmath47 for every eigenfunction @xmath17 associated with @xmath96 .
let us first consider the case where @xmath84 .
let us choose a pair of indices @xmath14 such that @xmath55 , and let us consider one of the associated basis functions given in the introduction , say @xmath99 to fix the ideas .
according to lemma [ lemarithm ] , we have @xmath100 and @xmath101 , where exactly one of the integers @xmath60 and @xmath61 is odd .
we can assume , without loss of generality , that @xmath60 is odd and @xmath61 even .
then @xmath102 @xmath103 and therefore @xmath104 we show in the same way that @xmath105 , @xmath106 , and @xmath107
. since @xmath108 depends only on @xmath96 , we have @xmath47 as soon as @xmath17 is a basis function associated with @xmath96 , and therefore , by linear combination , as soon as @xmath17 is an eigenfunction associated with @xmath96 .
the case @xmath62 can be treated in the same way , taking @xmath109 ( @xmath110 would also be suitable ) .
[ rmkeven ] it can also be shown that lemma [ lemarithm ] still holds if we replace the equation @xmath52 by @xmath111 , where @xmath112 and @xmath113 are odd integers such that @xmath114 .
this implies that the conclusion of theorem [ thmeven ] still holds if @xmath115 , with @xmath112 and @xmath113 as above .
in this section , we assume that @xmath22 . let us outline the idea we will use to construct the eigenfunction whose existence is asserted in proposition [ prop3dom ] .
it will belong to the eigenspace associated with the eigenvalue @xmath116 .
we start from the eigenfunction @xmath117 which has four rectangular nodal domains , shown in figure [ fig1a ] .
we perturb this eigenfunction by adding a small multiple of the eigenfunction @xmath118 for @xmath119 , we get the eigenfunction @xmath120 since @xmath121 is negative in the neighborhood of the critical points in @xmath122 , adding @xmath123 has the effect of opening small `` channels '' that connect the nodal domains where @xmath124 is negative . as a result , if @xmath125 is small enough , @xmath126 has three nodal domains , one where it is negative and two where it is positive ( see figure [ fig2 ] ) .
let us note that these ideas have already been used , to construct examples of eigenfunctions whose nodal set satisfies some prescribed properties , in @xcite .
in particular , the desingularization of critical points in the nodal set , that we have briefly described , is studied in details in @xcite to prove rigorously these assertions , let us consider the open domain @xmath128 in @xmath29 defined , in the standard coordinates @xmath129 , as @xmath1300,\pi[\times\left]0,\frac{\pi}{\sqrt{3}}\right[\,.\ ] ] we now define ( following @xcite ) the smooth change of coordinates @xmath131 which sends @xmath128 to @xmath132 - 1,1[\times]-1,1[$ ] . in these new
coordinates the nodal set of the function @xmath126 satisfies the equation @xmath133 a simple computation shows that this is the equation of an hyperbola . furthermore , when @xmath134
, this hyperbola has one branch in the lower left quadrant @xmath132 - 1,0[\times]-1,0[$ ] and one in the upper right quadrant @xmath1320,1[\times]0,1[\,$ ] . on the other hand
, we have the following symmetries of @xmath126 : @xmath135 and @xmath136 this allows us to recover the nodal set on the whole of @xmath32 .
it consists in two simple closed curves , each containing one branch of the previously considered hyperbola .
each of these closed curves enclose a nodal domain that is homeomorphic to a disk .
let us now consider the complement of the closure of those two region .
it is the third nodal domain .
[ rmkodd ] the construction used to prove proposition [ prop3dom ] can obviously be generalized .
we can for instance consider the eigenfunction @xmath137 assuming that @xmath138 we have in that case @xmath139 . following the same line of reasoning as in this section , we see that for @xmath125 small enough , @xmath126 has @xmath140 nodal domains . in view of remarks [ rmkeven ] and [ rmkodd ] , it would be desirable to obtain a characterization of the rational numbers @xmath141 , such that there exists an eigenfunction of @xmath3 on the torus @xmath142 with an odd number of nodal domains .
unfortunately , we have not been able to reach this goal so far . | in this note , we discuss a question posed by t. hoffmann - ostenhof ( see @xcite ) concerning the parity of the number of nodal domains for a non - constant eigenfunction of the laplacian on flat tori .
we present two results .
we first show that on the torus @xmath0 , a non - constant eigenfunction has an even number of nodal domains .
we then consider the torus @xmath1 , with @xmath2 , and construct on it an eigenfunction with three nodal domains .
[ [ keywords . ] ] keywords .
+ + + + + + + + + laplacian , torus , nodal domains . [ [ msc - classification . ] ] msc classification .
+ + + + + + + + + + + + + + + + + + + 35p99 , 35j05 . |
the diffractive scattering is a process where the colliding particles scatter at very small angles and without any color flux in the final state .
this involves a propagator carrying the vacuum quantum numbers , called pomeron and described , in the soft regime , within the regge theory .
since the first operation period in 1992 , zeus and h1 , the two experiments dedicated to the dis physics at hera , observed that a big amount @xmath8 of lepton - proton dis events had a diffractive origin opening a new area of studies in diffractive production mechanism , providing a hard scale which can be varied over a wide range and therefore it is an ideal testing for qcd models .
in particular , the diffractive production of vector mesons ( vms ) and real photons in @xmath0 collisions allows to study the transition from the soft to the hard regime in strong interactions . the hard regime ( high energy and low bjorken-@xmath9 )
is dominated by the exchange of a hard pomeron sensitive to the gluon content and well described by perturbative qcd ( pqcd ) , while in the soft regime ( low-@xmath9 ) the interaction is well described within the regge phenomenology .
indicating with @xmath10 the virtuality of the exchanged photon and with @xmath11 the square mass of the produced vm , hera data suggested a universal hard scale , @xmath12 , for the diffractive exclusive pruduction of vms and real photons , which indicates the transition from the soft to the hard regime .
a new precision measurement of the reaction @xmath14 was published by zeus @xcite .
it was found that the cross section falls steeply with the increasing of @xmath10 but , unlike it was observed for the @xmath15 electroproduction @xcite , it can not be described by a simple propagator term like @xmath16 , in particular an @xmath17 value increasing with @xmath10 appears to be favored .
figure [ q2_rho ] reports the cross section for the @xmath18 electroproduction versus @xmath10 compared with several theoretical predictions : the kwm model @xcite based on the saturation model , the fss model @xcite with and without saturation and the df model @xcite .
none of the available models gives a good description of the data over the full kinematic range of the measurement .
cross section as a function of @xmath10 measured at @xmath19 and comared in ( a ) and ( b ) with different models as described in the text .
[ q2_rho],scaledwidth=70.0% ] the soft to hard transition can be observed looking at the dependence of the vms photoproduction ( @xmath20 ) cross section from the @xmath21 centre of mass energy , @xmath13 , where the scale is provided by @xmath11 .
figure [ w_php ] collects the @xmath22 as a function of @xmath13 from the lightest vector meson , @xmath18 , to the heaviest , @xmath23 , compared to the total cross section .
dependence of the cross section for exclusive vm photoproduction together with the total photoproduction cross section .
lines are the result of a @xmath24 fit to the data at high @xmath13-energy values .
[ w_php],scaledwidth=60.0% ] the cross section rises with the energy as @xmath24 , where the @xmath25 exponent increases with the hard scale @xmath11 as expected for a transition from the soft to the hard regime .
new results on the @xmath23 photoproduction @xcite , recently published by zeus , confirmed the steeper rise of @xmath26 for higher vector meson masses . the transition from the soft to the hard regime
can also be studied varying @xmath10 .
recent results were achieved by h1 @xcite and zeus @xcite for the exclusive production of a real photon , the deeply virtual compton scattering ( dvcs ) , where the hard scale is provided only by the photon virtuality , @xmath10 .
figure [ w_dvcs ] shows the h1 ( left ) and the zeus ( right ) results .
the steep rise with @xmath13 of the cross section even at low-@xmath10 , seems to suggest that the most sensitive part to the soft scale comes from the wave function of the pruduced vm .
a similar result was obtained for the @xmath15 electroproduction @xcite . [ cols="^,^ " , ] the fit of @xmath27 was performed with all the parameters fixed excepted for the normalisation .
the model does not agree with the h1 measurements ( see fig .
[ fig : fit_new]c ) but it is compatible with the new zeus preliminary results ( see fig . [
fig : fit_new]d ) .
the @xmath28 variable was calculated in h1 by the transverse four - momenta of the scattered electron and the real photon , using the apprroximation @xmath29 , while in zeus a particular silicon microstrips spectrometer , based on the roman pots technique , was used in order to have a direct measurement of the scattered proton momentum @xmath30 and then @xmath28 being calculated from the quantity : @xmath31 , by the phormula : @xmath32 .
the really low acceptance of this spectrometer is the reason of the poor statistics of the zeus data , howewer it offers a really pure selection of diffractive events , not affected by any not - diffractive background .
the agreement with the zeus preliminary measurements encourages us to be still confident in our predictions , till they will be checked by the new hera data analyses ( now in progress with and without roman pots spectrometer ) . the slope of @xmath27 , calculated according to eq
. [ a7 ] , is depicted in fig .
[ fig : slope_q2 ] as a function of @xmath10 and @xmath13 and is compared with de hera measurements .
the local slope is predicted to slightly rise with @xmath13 but to be almost independent on @xmath10 . and
@xmath13 as calculated from eq
. [ a7 ] , compared with the hera data . _ _ ] eq .
[ a7 ] was then used to make a collection of figures showing the slope dependence for different values of @xmath13 and @xmath28 .
the @xmath28 dependence of the local slope calculated from eq .
[ a7 ] is shown in fig .
[ fig : slope]a at fixed @xmath33 for three different values of energy . .
for several values of @xmath13 , @xmath10 and @xmath28 .
_ _ ] figure [ fig : slope]b depicts the @xmath10 dependence of the local slope at @xmath34 for several values of @xmath28 .
figure [ fig : slope]c shows the enerrgy dependence of the local slope at @xmath33 for several values of @xmath28 .
* after the shut - down of the hera collider , the zeus and h1 experiments have left a huge heritage of experimental results on the diffractive production of vector mesons and real photons still waiting for a better understanding .
nevertheless , new analyses on hera data are still in rpogress . *
we show that data on dvcs from hera can be fitted with a semi - hard pomeron , not being flat .
* the pomeron trajectory could be a non - linear function ; it is nearly linear at small and moderate values of @xmath35 , leveling off at large @xmath35 .
the same logarithmic trajectory could be used in hadronic reactions , for example in analyzing , with higher precision , the future diffractive exclusive data from lhc , where details , such as the two - pion threshold @xcite , neglected here , should be taken into account . * here we used a simple minimal model , with a small number of the free parameters , which , in view of the limited number of data points , made possible the convergence of the minimization procedure .
a more detailed analyses should include also the individual contributions from the longitudinal and transverse cross sections , though their ratio @xmath36 , spin effects , reaction ratios and comparison with possible qcd predictions , especially concerning the evolution in @xmath10 ( see next item and appendix ) . *
a complete solution for the @xmath10 evolution in dvcs has not been yet found , although partial results are known from the literature . in a simplified , pragmatic approach the @xmath10 and @xmath28 dependence in a dvcs ( actually not factorizable ! ) , reduces the problem to the evolution of ordinary dis by the dglap procedure .
these attempts are summarized in appendix . *
the model and the fitting procedure presented in this paper can be extended to high - energy vector meson production .
we intend to do so in a forthcoming publication .
let us notice also that any consistent extension / interpolation of the regge approach to low energies ( resonance region ) should involve dual models , as it was shown @xcite for the pomeron component in case of @xmath37 scattering .
* acknowledgment * we would like to thank r. fiore , f. paccanoni and v. nikitin for the helpful discussions .
l. jenkovszky thanks the university of calabria , where part of this work was done , for its hospitality and support .
from eq . [ a2 ]
we get the real and imaginary parts of the dvcs amplitude at @xmath38 that will be useful for the evolution : @xmath39g(t , q_0 ^ 2)\bigl(\frac{q_0 ^ 2}{s_0x_{bj}}\bigr)^{\alpha(t ) } , \label{ima}\ ] ] the singlet combination of gpd , corresponding to the exchange in the @xmath46 channel of a @xmath47 charge conjugation , even signature reggeon is @xcite @xmath48 with @xmath49 where @xmath50 is the usual quark distribution for quark @xmath51 its evolution is well known .
the valencecombination , odd signature , is even under the change : @xmath52 the point @xmath53 where dglap @xmath54 @xcite and erbl @xmath55 @xcite regimes meet , is directly accessible to to experiment since , at leading order in @xmath56 the imaginary amplitude part of the amplitude @xmath57 for dvcs is proportional to a gpd at @xmath58 @xcite @xmath59 where @xmath60 is known from eq .
[ i m a ] .
unfortunately , the evolution equation for @xmath61 involves @xmath62 for @xmath9 in the interval @xmath63 $ ] ( for dvcs ) , to be known . an approximate solution based on conformal invariance , valid at one loop level is possible @xcite .
the off - forward matrix elements of a set of twist - two operators , whose leading - log evolution is diagonal due to conformal symmetry , are the gegenbauer polynomials . in the forward kinematics ,
the gegenbauer polynomials reduce to polynomials in @xmath64 in ref .
@xcite an explicit transformation has been constructed relating the off - forward and the forward evolutions .
usually it is very difficult to apply the transformation constructed in ref .
@xcite to a realistic problem but , in our case , the simplicity of the input at @xmath65 makes possible an explicit asymptotic solution for small @xmath66 and @xmath67 it has been shown in ref .
@xcite that , in the phenomenologically important small-@xmath68 and @xmath35 region , the off - diagonal distributions are determined unambiguously by the small-@xmath9 behavior of the diagonal parton distributions . by setting @xmath69 where the values of @xmath28 and @xmath68 do not change in the evolution to a higher scale @xmath70 , it was shown in ref .
@xcite that at small @xmath68 @xmath71{d\over{dx'}}\biggl({q(x')\over{|x'|}}\biggr ) , \label{hq}\ ] ] @xmath72 the proof of eq .
[ hq ] uses the properties of the diagonal parton distributions @xmath73 that corresponds to the symmetry relations @xmath74 @xmath75 @xmath76 where , as usual , @xmath77 stays for singletquark contribution and @xmath78 means non - singlet . at small @xmath68 , the same anomalous dimensions @xmath79 control both the diagonal and off - diagonal evolution .
consider now the prediction of @xcite for the off - diagonal distribution @xmath80 by making the physically reasonable small-@xmath9 assumption that @xmath81 where @xmath82 may depend also on @xmath83 then eq . ( 18 ) can be integrated @xmath84^{\lambda_q+1 } , \label{hqbis}\ ] ] where , for singlet quarks ( i.e. when @xmath85 ) the integral is a principal value integral and eq .
[ hqbis ] becomes integrable for any @xmath86 the distribution @xmath87 for small @xmath88 has the form @xcite @xmath89 with @xmath90 where @xmath91 is a function determined by eq .
[ hqbis ] .
similarly , if @xmath92 then @xmath93 at small @xmath94 viceversa , since the amplitude in eqs .
[ i m a ] , and [ rea ] is of regge form and is proportional to @xmath95 for small @xmath68 and fixed @xmath96 then @xmath97 and the parton distribution that appears in eq .
[ hq ] has the form @xmath98 where @xmath99 can represent @xmath100 if we start from eq .
[ sume ] .
an approximate analytic evolution of the input singlet parton distribution is well known if the input is of the regge - like form . since
only the diagonal parton distributions are involved in the calculations , one can resort to libraries that generate parton distributions but , since we are interested in the small-@xmath68 region .
equation [ hq ] then determines the @xmath10 dependence of @xmath80 and , finally , eq .
[ sume ] gives the scale dependence of @xmath101 .
the knowledge of @xmath102 fixes also , at leading order in @xmath103 since @xmath104 d. muller , arxiv : hep - ph/0605013 .
kresimir kumericki , dieter mueller , kornelija passek - kumericki , _ sum rules and dualities for generalized parton distributions : is there a holographic principle ?
_ , arxiv:0805.0152 .
f. m. dittes , d. mueller , d. robaschik , b. geyer and j. horejsi , phys .
b * 209 * ( 1988 ) 325 ; + d. mueller , d. robaschik , b. geyer , f. m. dittes and j. horejsi , fortsch .
* 42 * ( 1994 ) 101 .
h1 coll .
, a. aktas et al . , eur .
phys . j. c * 44 * ( 2005 ) 1 .
v. n. gribov and l. n. lipatov , sov .
* 15 * ( 1972 ) 48 ; l. n. lipatov , sov . j. nucl . phys . * 20 * ( 1975 ) 94 ; g. altarelli and g. parisi , nucl . phys . b*126 * ( 1977 ) 298 ; y. l. dokshitzer , dov .
jetp * 46 * ( 1977 ) 641 .
l. l. jenkovszky and v. k. magas , in the proc .
of the xxxi - th ismd , held in datong ( china ) , hep - ph/0111398 ; r. fiore , l. l. jenkovszky and v. k. magas , ismd conference in rohnert park ( california , usa ) , acta phys . pol .
b * 36 * ( 2005 ) 743 . | the exclusive diffractive production of vector mesons and real photons in @xmath0 collisions has been studied at hera in a wide kinematic range . here
we present the most recent experimental results together with a regge - type model .
we deduce the pomeranchuk trajectory ( pomeron ) by analyzing the hera data on deeply virtual compton scattering ( dvcs ) , and then discuss its basic properties , namely its apparent hardnessand its non - flatbehavior , different from the claims of some authors .
0.5 cm * exclusive diffraction and pomeron trajectory * * in _ ep _ collisions * 0.3 cm s. fazio@xmath1 and l. jenkovszky@xmath2 @xmath3 _ dipartimento di fisica , universit della calabria _ _ istituto nazionale di fisica nucleare , gruppo collegato di cosenza _ _ i-87036 arcavacata di rende , cosenza , italy _
@xmath4 _ joint institute for nuclear research , 141980 - dubna , moscow region , russia _
@xmath5 _ bogolubov institute for theoretical physics , national academy of sciences of ukraine _ _
kiev-143 , 03680 ukraine _
@xmath6 _ kfki , rmki , budapest , hungary _ 0.1 cm 0.1 cm @xmath7 |
non equilibrium phenomena are on the agenda in modern statistical physics , soft condensed matter and biophysics .
open systems driven far from equilibrium are ubiquitous .
a classical case is driven navier stokes turbulence , other cases are driven lattice gases , growing interfaces , growing fractals , etc . unlike equilibrium physics where the boltzmann - gibbs scheme applies , the ensemble is not known in non equilibrium . here
the problem is defined in terms of a numerical algorithm , a master equation , or a langevin equation . an interesting class of non equilibrium systems exhibit scale invariance .
one example is diffusion limited aggregation ( dla ) driven by the accretion of random walkers yielding a growing scale invariant fractal with dimension @xmath0 in 2d .
another case is a growing interface driven by random deposition or propagating in a random environment . here
the width of the growing front @xmath1 conforms to the dynamical scaling hypothesis @xmath2 , where @xmath3 is the size of the system and @xmath4 and @xmath5 scaling exponents ; @xmath4 characterizing the roughness and @xmath5 describing the dynamical crossover to the stationary profile @xcite . in the present paper we focus on
the kardar - parisi- zhang ( kpz ) equation which describes an intrinsic non equilibrium problem and plays the same role as the ginzburg - landau functional in equilibrium physics .
the kpz equation was introduced in 1986 in a seminal paper by kardar , parisi and zhang @xcite .
it has the form @xmath6 and purports to describe non equilibrium aspects of a growing interface ; see refs .
@xcite . here
@xmath7 is the height of an interface at position @xmath8 and time @xmath9 , the linear diffusion term @xmath10 , characterized by the diffusion coefficient @xmath11 , represents a surface tension , the nonlinear growth term @xmath12 , characterized by @xmath13 , is required to account for the lateral growth , @xmath14 is an imposed constant drift , and the random aspects , i.e. , the random deposition of material or the random character of the medium , are encoded in the noise @xmath15 .
the noise is assumed to be locally correlated in space , and time , its strength characterized by @xmath16 . despite its simple form
the kpz equation is difficult to analyze and many aspects remain poorly understood @xcite . apart from its intrinsic interest the kpz equation
is also related to fundamental issues in _ turbulence _ and _
disorder_. introducing the local slope field @xmath17 the kpz equation takes the form of a burgers equation driven by conserved noise , @xmath18 in the noiseless case setting @xmath19 and regarding @xmath20 as a velocity field , the non linear term appears as the convective term in the navier - stokes equation and eq .
( [ burgers ] ) has been used to model aspects of _ turbulence_. in the 1d case the relaxation of the velocity field takes place subject to a transient pattern formation composed of domain walls and ramps with superimposed diffusive modes . in the driven case
( [ burgers ] ) was studied earlier in the context of long time tails in hydrodynamics @xcite . on the other hand , applying the non linear cole - hopf transformation the kpz equation maps to the cole - hopf equation ( ch ) @xmath21 a linear diffusion equation driven by multiplicative noise .
the ch equation has a formal path integral solution @xcite which can be interpreted as an equilibrium system of directed polymers ( dp ) with line tension @xmath22 in a quenched random potential @xmath23 ; a model system in the theory of _ disorder _ which has been studied using replica techniques @xcite .
the kpz equation lives at a critical point and conforms to the dynamical scaling hypothesis .
for the height correlations we have @xmath24 here @xmath4 , @xmath5 , and @xmath14 are the roughness exponent , dynamic exponent , and scaling function , respectively .
to extract scaling properties the initial analysis of the kpz equation was based on the dynamic renormalization group ( drg ) method , previously applied to dynamical critical phenomena and noise - driven hydrodynamics @xcite .
an expansion in powers of @xmath13 in combination with a momentum shell integration yield to leading order in @xmath25 the drg equation @xmath26 , with beta - function @xmath27 .
here @xmath28 is the effective coupling strength and @xmath29 the logarithmic scale parameter @xcite .
the emerging drg phase diagram is depicted in fig .
[ fig1 ] . .
in @xmath30 the drg flow is towards the strong coupling fixed point @xmath31 . above the lower critical dimension @xmath32 there is an unstable kinetic transition line , separating a rough phase from a smooth phase . ] before discussing the phase diagram we note two further properties of the kpz equation .
first , subject to a galilean transformation the equation is invariant provided we add a constant slope to @xmath33 and adjust the drift @xmath14 , i.e. , @xmath34 note that the slope field @xmath20 and the diffusive field @xmath35 transform like @xmath36 and @xmath37 $ ] , respectively .
the galilean invariance implies the scaling law @xmath38 relating @xmath4 and @xmath5 @xcite ; the galilean invariance is a fundamental dynamical symmetry specific to the kpz equation , delimiting the universality class .
second , a fluctuation - dissipation theorem is operational in 1d since the stationary fokker - planck equation admits the explicit solution @xcite @xmath39 . \label{stat}\end{aligned}\ ] ] this distribution shows that the slope @xmath40 fluctuations are uncorrelated and that the height field @xmath41 performs a random walk .
consequently , from eq .
( [ corr ] ) we infer the roughness exponent @xmath42 and from the scaling law ( [ scal ] ) the dynamic exponent @xmath43 .
in other words , the scaling exponents associated with the strong coupling fixed point are exactly known in 1d , see fig .
[ fig1 ] ; moreover , results for the scaling function can be obtained by loop expansions @xcite .
the lower critical dimension is @xmath32 . below @xmath32
the drg flow is towards the strong coupling fixed point . above @xmath32
the drg equation yields an unstable kinetic phase transition line as indicated in fig .
[ fig1 ] . for @xmath13 below a critical coupling strength @xmath44 the drg flow is towards @xmath45 , corresponding to the linear edwards - wilkinson ( ew ) equation , the kpz equation for @xmath46 , yielding @xmath47 and @xmath48 , note that the scaling law is not operational for @xmath45 ; for @xmath49 the drg flow is towards a non perturbative strong coupling fixed point . a drg analysis to all order in @xmath25 yields @xmath47 and @xmath50 on the transition line and a singularity at the upper critical dimension @xmath51
. mode coupling techniques also give @xmath51 , whereas a directed polymer analysis yields @xmath52 @xcite . in order to disentangle the properties of the kpz equation , both regarding scaling and
otherwise , there is a need for alternative methods . in the present paper
we summarize a non perturbative weak noise approach which we have pursued in recent years @xcite . the working hypothesis here is to focus on the noise strength @xmath16 as the determining parameter rather that the non linearity @xmath13 and ensuing drg analysis or mapping to dp . for @xmath53
the interface relaxes subject to a transient pattern formation ; for @xmath54 the interfaces initially decays but is eventually driven into a stationary fluctuating state ; the crossover time diverging for @xmath55 . in the next section
we summarize the scheme that will allow us to access the weak noise regime in a non perturbative fashion .
the weak noise scheme takes as its starting point a generic langevin equation @xmath56 determining the problem we wish to analyze . here
@xmath57 is a multi - dimensional random variable , @xmath58 a general non linear drift , and @xmath23 additive white noise correlated with strength @xmath16 . in order to implement a weak noise approximation we consider the equivalent fokker - planck equation for the distribution @xmath59 @xmath60 interpreting @xmath61 as a momentum operator , @xmath62 as an effective wave function , and @xmath16 as an effective planck constant , eq .
( [ fp ] ) has the form of an imaginary time schrdinger equation and it is natural to introduce the well - known wkb or eikonal approximation @xmath63 . \label{wkb}\end{aligned}\ ] ] to leading order in @xmath16 the action then obeys a principle of least action @xmath64 as expressed by the hamilton - jacobi equation @xmath65 with canonical momentum @xmath66 .
the hamiltonian ( energy ) takes the the form @xmath67 , \label{ham}\end{aligned}\ ] ] yielding the coupled equations of motion @xmath68 the action associated with an orbit from @xmath69 to @xmath57 in time @xmath70 is @xmath71= \frac{1}{2}\int_{x_1,0}^{x , t}dt p(t)^2 . \label{act}\end{aligned}\ ] ] the weak noise recipe is clear . we solve the equations of motion ( [ eqm ] ) and identify an orbit from @xmath69 to @xmath57 in time @xmath70 with @xmath72 as an adjusted variable .
the orbits lie on constant @xmath73 manifolds . evaluating the action @xmath74 for a specific orbit the wkb approximation ( [ wkb ] ) yields the transition probability from @xmath69 to @xmath57 in time @xmath70 . assuming @xmath75 for small @xmath57 the phase space has the generic structure depicted in fig .
[ fig2 ] . to the final @xmath57 in transition time
@xmath70 lies on the energy manifold @xmath76 . in the long time
limit the orbit migrates to the zero - energy manifold composed of a transient submanifold for @xmath77 , corresponding to the noiseless case , and a stationary manifold for @xmath78 , corresponding to the noisy case .
the submanifolds intersect in the saddle point ( sp ) , determining the markovian behavior . ]
the present variationally based weak noise scheme , dates back to onsager . in more recent formulations it corresponds to the saddlepoint contribution ( optimal path ) in the functional martin - siggia - rose scheme ,
the weak noise scheme applies directly to the kpz equation in the cole - hopf formulation ( [ cheq ] ) . extending the scheme in order to incorporate multiplicative noise ,
@xcite , the wkb scheme yields the equations of motion , action , and distribution @xmath79+k_0 ^ 2w^2p,~~ \frac{\partial p}{\partial t}=-\nu[\nabla^2 w - k^2w]-k_0 ^ 2pw^2 , \label{eqmo } \\ & & s(w , t)=(k_0 ^ 2/2)\int^{w , t}d \vec r dt(wp)^2,~~p(w ,
t)\propto\exp[-s(w , t)/\delta],\end{aligned}\ ] ] with parameters @xmath80 and @xmath81 . on the transient and stationary manifolds @xmath77 and @xmath82 the equations of motion reduce in the static case to the diffusion and non linear schrdinger equations @xmath83 admitting localized spherically symmetric solutions @xmath84 and @xmath85 for large @xmath86 , respectively . in terms of the height field @xmath87 and slope field @xmath17 we obtain the fundamental localized static growth modes for large @xmath86 @xmath88 the growth modes are characterized by the amplitude or charge @xmath89 , determined by the imposed drift @xmath14 in the kpz equation . the growth mode with positive charge @xmath90 , @xmath91 lives on the transient noiseless @xmath77 manifold and corresponds to a cone or dip in @xmath33 and a constant positive monopole in @xmath20 .
the mode carries zero action , zero energy @xmath92=0 , and zero momentum @xmath93 ; the distribution @xmath94 associated with the mode is of @xmath95 . the growth mode with negative charge @xmath96 , @xmath97 is associated with the stationary noisy manifold @xmath82 and corresponds to an inverted cone or tip in @xmath33 and a constant negative monopole in @xmath20 .
this modes carries a finite action @xmath98 yielding the distribution @xmath99 $ ] . in the 1d case
the growth modes correspond to right hand and left hand domain walls , @xmath100 .
the right hand domain wall is the viscosity broadened shock wave in the deterministic burgers equation ; the left hand domain wall is noise induced , in the present scheme characterized by a finite @xmath72 , see refs .
@xcite . in fig .
[ fig3 ] we have depicted the static growth modes in the height and slope fields . applying the galilean transformation ( [ gal ] ) we can boost the static modes and obtain the propagating growth modes @xmath101 the moving localized growth modes are the fundamental elementary excitations incorporating the non linear aspects of the kpz equation . in a quantum field theory context
the growth modes correspond to instantons or solitons .
by means of the propagating localized growth modes ( [ gm ] ) we construct a global solution of the field equations ( [ eqmo ] ) .
the galilean invariance ( [ gal ] ) determines the matching of the modes and we obtain the dilute network solution @xmath102 here @xmath103 is the assignment of charges and @xmath104 the initial positions .
this network construction correspond to a multi - instanton solution in quantum field theory .
as time evolves the modes propagate and the velocities adjust to constant values given by the self - consistent equation @xmath105 superimposed on the propagating network of growth modes is a spectrum of linear extended diffusive modes with dispersion @xmath106 , following from a linear analysis of the field equations ( [ eqmo ] ) . for large @xmath8 the slope field @xmath107 .
in order to ensure the boundary condition of a flat interface at large distances we impose the neutrality condition @xmath108 ; note that this boundary condition still allows for a local offset of @xmath33 and the propagation of facets or steps . during time evolution
the dynamical network propagates across the system . imposing periodic or bouncing boundary conditions increments
are added to @xmath33 and the interface grows , see also ref .
@xcite . from an analysis of eqs .
( [ net1 ] ) and ( [ net2 ] ) it follows that the growth modes with positive charge exerts an attraction of the other modes , whereas the negatively charged growth modes repel the other modes .
although our analysis so far only applies to a dilute network , it follows tentatively from a numerical simulations of eqs .
( [ net1 ] ) and ( [ net2 ] ) that the modes form dipoles and that the long time stable network configuration is composed of a gas of propagating dipoles , i.e. the pairing of monopole growth modes with opposite charges . in terms of the slope field
a dipole mode with charges @xmath109 and @xmath110 has the form @xmath111 $ ] , @xmath112 .
the mode propagate with velocity @xmath113 and carries energy , momentum and action , @xmath114 , @xmath115 , and @xmath116 . in terms of the height field @xmath117
$ ] the mode in depicted in fig .
asymptotically the height field is flat corresponding to a vanishing slope field .
the dipole mode corresponds to a propagating local defect or deformation . in 1d
the dipole mode corresponds in the slope field to a matched right hand and left hand domain wall propagating across the system , see fig .
the scaling issues for the kpz equation remain unsettled except in 1d where the fluctuation - dissipation theorem ( [ stat ] ) and the scaling law ( [ scal ] ) give access to the strong coupling fixed point with scaling exponents @xmath42 and @xmath118 , see fig .
[ fig1 ] . in higher d
there has been many attempts to access the strong coupling fixed point both on the basis of drg , mode coupling , directed polymers ( dp ) , and numerically ; however , the strong coupling features remain elusive @xcite .
the present weak noise method is not a scaling approach but rather a many body description of a growing interface .
nevertheless , the method allows a discussion of some of the scaling features . since the scheme is consistently galilean invariant the scaling law ( [ scal ] )
is automatically obeyed .
the roughness exponent @xmath4 is associated with the static correlations @xmath119 and requires the static distribution @xmath120 , @xmath121 $ ] , @xmath122 .
the stationary distribution is associated with the zero - energy stationary manifold which in general is difficult to identify for a system with many degrees of freedom . in 1d in terms of the slope field @xmath40 and associated noise field @xmath72 the hamiltonian takes the form @xmath123 $ ] and it follows that the stationary zero - energy manifold is given by @xmath124 , yielding a hamiltonian density as a total differential .
correspondingly , the stationary action @xmath125 and the distribution @xmath126 $ ] in accordance with ( [ stat ] ) , implying @xmath42 . since galilean invariance is built in the exponent @xmath43 follows automatically .
however , we can also infer @xmath43 from an independent argument based on the dispersion law for the low lying gapless excitations .
the elementary excitations are the right hand and left hand domain walls .
a composite quasi particle or dipole mode satisfying the boundary condition of vanishing slope can be constructed by paring two domain walls .
the dipole mode propagates with energy @xmath127 and momentum @xmath128 , where @xmath129 is the dipole amplitude . using the analogy between the stochastic formulation and quantum mechanics , i.e. , the canonical quantization of the weak noise scheme with @xmath16 as an effective planck constant , and using the spectral representation @xmath130
, @xmath131 is a form factor , we note that the bottom of the gapless dipole energy spectrum @xmath132 implies @xmath133 , where @xmath43 .
in higher @xmath134 we have not been able to identify the stationary zero - energy manifold and the exponent @xmath4 . also , in order to determine @xmath5 , for example from the form of the time dependent correlations @xmath135 , we need both the transition probabilities and the stationary distribution , i.e. , by definition @xmath136 . in the weak noise scheme @xmath137 $ ] which requires a detailed analysis of the dynamical network and the associated action . in the dipole sector we can
, however , present some preliminary scaling results . since the propagating dipole mode according to ( [ actgm ] ) carries action @xmath138 and propagates with velocity @xmath139 the mode moves the distance @xmath140 in time @xmath70 . expressing @xmath74 in the form
@xmath141 we infer the single dipole distribution @xmath142 $ ] and , correspondingly , the dipole mean square displacement @xmath143 with hurst exponent @xmath144 ; note that the dynamic exponent @xmath145 . in the stochastic representation the dipole mode thus performs anomalous diffusion .
in @xmath146 we have @xmath147 , @xmath148 , in agreement with a formal dp result @xcite . in @xmath30
we obtain @xmath149 and the exact result @xmath43 , i.e. , the dipole modes exhaust the spectrum . in @xmath32
we have @xmath150 and @xmath47 , i.e. , ordinary diffusion .
in @xmath151 we obtain @xmath152 and @xmath153 , the dipole mean square displacement falls off logarithmically @xmath154 , however , @xmath5 diverges at variance with accepted drg and dp results . in @xmath155 we have @xmath156 and the mean square displacement is arrested .
below @xmath32 ( the lower critical dimension ) the dipole modes superdiffuse , above @xmath32 we have subdiffusion .
these scaling results only refer to the dipole sector .
the last issue is the much discussed upper critical dimension for the kpz equation @xcite .
the weak noise approach allows a non scaling argument for the existence of a critical dimension . above @xmath155 the negative growth mode as a bound state solution to the non linear schrdinger equation ( [ nse ] ) ceases to exist
this implies that the dynamical network representation of a growing interface ceases to be valid .
this result follows from a numerical analysis of the non linear schrdinger equations ( [ nse ] ) but can also be inferred by an algebraic proof based on derrick s theorem .
first , introducing @xmath157 , @xmath158 , and @xmath159 we infer from ( [ nse ] ) the identity @xmath160 .
second , deducing ( [ nse ] ) from a variational principle @xmath161 with @xmath162 and performing a constrained minimization @xmath163 , @xmath164 , @xmath165 , @xmath166 and @xmath167 we have the second identity @xmath168 . finally , requiring @xmath169 the identities imply @xmath170 , q.e.d .
in this paper we have presented a short review of a recently developed asymptotic weak noise approach to the kardar - parisi - zhang equation .
the scheme provides a many body description of a growing interface in terms of a dynamical network of growth modes .
the growth modes are the elementary building blocks and their propagation accounts for the kinetic growth .
kinetic transitions are determined by an associated dynamical action , replacing the customary free energy landscape .
superimposed on the network is a gas of diffusive modes . in 1d
the dispersion laws delimit the universality classes : in the kpz case the gapless domain wall modes yield @xmath43 , the diffusive modes being subdominant ; in the ew case the domain walls are absent and the gapless diffusive modes yield @xmath47 . in higher d the scaling results based on the weak noise method are still subject to scrutiny . finally , we mention that the weak noise method has also been applied to the noise - driven ginzburg - landau equation , a finite - time - singularity model , and dna bubble dynamics @xcite
. 99 a. -l .
barabasi and h. e. stanley , _ fractal concepts in surface growth _ ( cambridge university press , 1995 ) ; j. krug and h. spohn , _ solids far from equilibrium ; kinetic roughening of growing surfaces : fractal concepts in surface growth _ ( cambridge university press , 1992 ) ; j. krug , adv .
46 * , 139 ( 1997 ) m. kardar , g. parisi and y. c. zhang , phys .
lett . * 56 * , 889 ( 1986 ) ; e. medina , t. hwa , m. kardar and y. c. zhang , phys .
a * 39 * , 3053 ( 1989 ) t. halpin - healy and y. c. zhang , phys .
254 * , 215 ( 1995 ) d. forster , d. r. nelson and m. j. stephen , phys .
* 36 * , 867 ( 1976 ) ; phys .
a bf 16 , 732 ( 1977 ) m. kardar and y. c. zhang , phys .
58 * , 2087 ( 1987 ) ; m. kardar , nucl .
b bf 290 , 582 ( 1987 ) h. c. fogedby , phys .
lett . * 94*,195702 ( 2005 ) ; phys .
e * 73 * , 031104 ( 2006 ) ; phys .
e * 68 * , 026132 ( 2003 ) ; phys .
e59 , 5065 ( 1999 ) ; phys .
e57 , 49431 ( 1998 ) ; phys .
80 , 1126 ( 1998 ) l. onsager and s. machlup , phys
. rev . * 91 * , 1505 ( 1953 ) , ibid 1512 ; p c. martin , e. d. siggia and h. a. rose , phys .
rev . a * 8 * , 423 ( 1973 ) ; r. baussch , h. k. janssen and h. wagner , z. phys .
b * 24 * , 113 ( 1976 ) d. a. huse , c. l. henley and d. s. fisher , phys .
lett . * 55 * , 2924 ( 1985 ) ; e. frey and u. c. tuber , phys .
e * 50 * , 1024 ( 1994 ) ; e. frey , u. c. tuber and t. hwa , phys .
e * 53 * , 4424 ( 1996 ) ; k. j. wiese , j. stat . phys . * 93 * , 143 ( 1998 ) ; f. colaiori and m. a. moore , phys .
lett . * 86 * , 3946 ( 2001 ) ; m. lssig , phys . rev . lett . *
80 * , 2366 ( 1998 ) ; nucl .
b * 448 * , 559 ( 1995 ) ; m. lssig and h. kinzelbach , phys .
lett . * 78 * , 903 ( 1997 ) ; p. le doussal and k.j .
wiese , phys .
e * 72 * , 035101 ( 2005 ) h. c. fogedby , j. hertz and a. svane , europhys .
lett . * 62 * , 795 ( 2003 ) ; h. c. fogedby and v. poutkaradze , phys .
e * 66 * , 021103 ( 2002 ) ; hans c. fogedby and ralf metzler , phys .
lett . * 98 * , 070601 ( 2007 ) . | we review a recent asymptotic weak noise approach to the kardar - parisi - zhang equation for the kinetic growth of an interface in higher dimensions .
the weak noise approach provides a many body picture of a growing interface in terms of a network of localized growth modes . scaling in 1d
is associated with a gapless domain wall mode .
the method also provides an independent argument for the existence of an upper critical dimension . |
corrections to the @xmath0 decay are usually described in the framework of an effective theory , ) we discarded terms proportional to @xmath1 since they give only small contributions to the branching ratio that start at next - to - leading - order ( nlo ) .
similar nnlo corrections can therefore be safely neglected . ]
@xmath2 here , @xmath3 are renormalization scale dependent effective couplings , the so - called wilson coefficients , which encode the heavy gauge boson and the heavy top quark effects .
the @xmath4-quark scale contributions , on the other hand , are seen as matrix elements of flavor changing operators , @xmath5 o_{3,5 } & = ( \bar s_l\gamma_{3,5 } b_l){\sum}_q(\bar q\gamma_{3,5}^\prime q)\ , , & o_{4,6 } & = ( \bar s_l\gamma_{3,5 } t^a b_l){\sum}_q(\bar q\gamma_{3,5}^\prime t^a q)\,,\nonumber\\[1 mm ] o_7 & = \frac{\alpha_{\rm em}}{4\pi}\,m_b(\bar s_l\sigma^{\mu\nu}b_r)f_{\mu\nu}\ , , & o_8 & = \frac{\alpha_{s}}{4\pi}\,m_b(\bar s_l\sigma^{\mu\nu } t^a b_r)g^a_{\mu\nu}\,,\end{aligned}\ ] ] where @xmath6 , @xmath7 , @xmath8 and @xmath9 . using ( [ eq::lagrangian ] ) , the differential decay rate for @xmath0 can be written as follows , @xmath10 in the equation displayed above , @xmath11 denotes the momentum of the photon and the @xmath12 , which describe the hadronic dynamics , are given by the imaginary part of forward scattering amplitudes , @xmath13 since the mass of the @xmath4-quark is much larger than the binding energy of the @xmath14-meson , which is of the order of @xmath15 , we can perform an operator product expansion ( ope ) of the time ordered products .
doing so , one finds that the leading term is the partonic decay rate which gives the dominant contribution , while the non - leading terms , the so - called power - corrections , are suppressed by powers of @xmath16 and give non - vanishing contributions starting from @xmath17 . ) and its ope hold only for @xmath18 . in all other cases
the @xmath12 contain contributions in which the photon couples to light quarks ( @xmath19 ) , and this leads to non - perturbative effects different from that mentioned above ( see section [ sec::non - perturbative ] ) .
furthermore , for @xmath20 , contributions that contain no photon in the final state should be excluded , as required by the experimental setup . ] in what follows we describe the state - of - the - art of perturbative and non - perturbative corrections in the @xmath0 decay .
the calculation of the perturbative corrections can be divided into three steps . in the first step one
has to evaluate the effective couplings @xmath3 at the high - energy scale @xmath21 by requiring equality of the standard model and the effective theory green functions . defining @xmath22 ,
the effective couplings can be expanded as follows , @xmath23 at nnlo accuracy one has to determine the coefficients @xmath24 . for @xmath25=@xmath26 it required performing a three - loop calculation @xcite whereas for the remaining cases @xmath25=@xmath27
a two - loop calculation was sufficient @xcite .
the second step involves the calculation of the anomalous dimension matrix @xmath28 which describes the mixing of the operators under renormalization .
its knowledge is necessary to solve the effective theory renormalization group equations for the effective couplings , @xmath29 and to evolve the latter down to the low - energy scale @xmath30 . performing a perturbative expansion in the strong coupling constant
, the anomalous dimension matrix takes the following form , @xmath31 at nnlo one has to determine @xmath32 which is actually a @xmath33 matrix , @xmath34 0_{2\times 6 } & c_{2\times2}^{(2 ) } \end{pmatrix}\,.\ ] ] the block matrices @xmath35 and @xmath36 describing the self - mixing of the four - quark operators and the self - mixing of the dipole operators at three loops , respectively , have been calculated in @xcite .
the block matrix @xmath14 describing the mixing of the four - quark operators into the dipole operators at four loops has been determined in @xcite . after this calculation
the first two steps of the perturbative calculation were completed , that is the effective couplings at the low - energy scale @xmath30 with resummed logarithms are now known at nnlo accuracy .. ] in the last step one has to calculate on - shell amplitudes of the operators at the low - energy scale .
this is the most difficult part of the nnlo enterprise and it is still under investigation . in order to see what has been done so far , and what still has to be done , we write the decay rate for the partonic decay @xmath37 as follows , @xmath38 where @xmath39 can again be expanded in terms of @xmath40 , @xmath41 at nnlo one has to determine the coefficients of @xmath42 which , however , has only been done in a complete manner for @xmath25=@xmath43=@xmath44 @xcite .
once we neglect on - shell amplitudes that are proportional to the small wilson coefficients of the four - quark penguin operators @xmath45-@xmath46 , the remaining cases to be considered are @xmath47=@xmath48 , and @xmath49 .
the large-@xmath50 corrections are known in all these cases except for @xmath51 and @xmath52 @xcite .
in addition , effects of the charm and bottom quark masses on the gluon lines are known in all the cases @xcite .
the other beyond - large-@xmath50 corrections have been found only in the limit @xmath53 , except for the @xmath54 and @xmath49 cases @xcite .
this limit has been used to interpolate the unknown beyond - large-@xmath50 corrections at @xmath55 to the measured value of @xmath56 @xcite .
the result for the branching ratio , for @xmath57 , is given by @xcite @xmath58 the theoretical uncertainty of this nnlo estimate is at the same level as the uncertainty of the current world average reported by hfag @xcite , @xmath59 which is furthermore expected to come down to the 5% level at the end of the b - factory era . here
a remark concerning the overall normalization of the theoretical prediction is in order . to reduce parametric uncertainties stemming from the ckm angles as well as from the @xmath60- and @xmath4-quark masses , the partonic decay rate given in ( [ eq::gamma - partonic ] )
is usually normalized using a combination of the @xmath61 and @xmath62 decay rates which is reflected by the appearance of the semileptonic phase - space factor @xmath63 in the analytical expressions @xcite .
is already known at nnlo accuracy @xcite . ] unfortunately , the determination of @xmath64 and @xmath36 from the fit from the measured spectrum of the @xmath61 decay in the 1s scheme @xcite differs from that in the kinetic scheme @xcite . and
@xmath36 from @xcite were adopted . ] using the values for @xmath64 and @xmath36 of the latter determination results in a higher central value for the @xmath0 decay rate @xcite , @xmath65 the difference of @xmath64 and @xmath36 in the 1s and kinetic scheme is likely to be due to different input data , differences in the fit method , and treatment of theory errors . in this respect , supplementing the fit , for example , by the determination of the @xmath60- and @xmath4-quark masses from sumrules @xcite could possibly be helpful to reduce the discrepancy of @xmath36 in both schemes .
we should also remark that not all of the aforementioned contributions to the function @xmath66 entered the analysis of @xcite .
these are the massive fermionic corrections presented in @xcite and the large-@xmath50 contributions for @xmath47@xmath67@xmath68 from @xcite . into the chromomagnetic dipole operator @xmath69 @xcite
was not included in @xcite . ]
these contributions will be included in a future update together with so far unknown contributions , which is , for example , the complete knowledge of @xmath70 @xcite .
also the complete calculation of @xmath71 for @xmath72 is underway @xcite .
especially the latter will prove useful in the reduction of the uncertainty stemming from the interpolation in @xmath64 .
apart from the nnlo corrections also tree - level diagrams with the @xmath73-quark analogues of @xmath74 and the four - quark operators @xmath75 have been neglected so far @xcite .
the numerical effect of all of these contributions on the branching ratio is or is expected to remain within the uncertainty of the nnlo estimate given in ( [ eq::nnlo - estimate ] ) .
finally , we should mention that there are also cutoff - enhanced corrections which matter close to the endpoint @xcite .
however , as demonstrated in @xcite , the resummation of the cutoff - enhanced logarithms overestimates the effect of the @xmath76-terms for @xmath77
. therefore , the prediction for the branching ratio given in ( [ eq::nnlo - estimate ] ) should , at present , be considered as more reliable .
since the perturbative calculations in @xmath0 are now performed at nnlo the non - perturbative corrections become more important . in general well under control are the power - corrections stemming from the ope of the time ordered products contained in ( [ eq::wij ] ) . for the self - interference of the electromagnetic operator , that is for @xmath25=@xmath43=@xmath44 , they are known at @xmath17 @xcite and @xmath78 @xcite . for all other combinations of @xmath79 and @xmath80 the time ordered products include contributions in which the photon couples to light quarks , and that causes the breakdown of the ope . in this case
non - perturbative collinear effects @xcite as well as power - corrections at @xmath81 show up @xcite .
the combined effect of all of the aforementioned non - perturbative corrections is of around 3% in the branching ratio . besides
, non - perturbative effects appearing at @xmath82 show up when the photon couples to light quarks .
their size is not known at present and hence a 5% uncertainty related to all the unknown non - perturbative effects has been included in ( [ eq::nnlo - estimate ] ) .
the size of this uncertainty is supported by the estimate of the @xmath82-corrections in the interference of the electro- and chromomagnetic dipole operators performed in @xcite . as pointed out in references @xcite , the magnitude of the effect considered in @xcite
could be probed by an improved measurement @xcite of the isospin asymmetry @xmath83 finally , we note that a non - perturbative uncertainty appears also when extrapolating the three different measurements performed at cleo , babar and belle down to the common lower cut @xmath57 in the photon energy .
it is accounted for in the error of the world average given in ( [ eq::hfag ] ) .
at present the uncertainties in the branching ratio of @xmath0 are on the same level on both the theoretical and experimental side .
thanks to the ongoing calculations of the perturbative corrections the uncertainty stemming from this part will further reduce . however , to reach the 5% level or even less on the theoretical side a better understanding of the non - perturbative power - corrections at @xmath82 is required . _
acknowledgements : _ i am grateful to h. czyz , m. krawczyk , and m. misiak for the invitation to this workshop .
i also would like to thank p. gambino , and especially m. misiak , for comments on the final version of this manuscript .
this work was supported by miur under contract 2004021808 - 009 and by a european community s marie - curie research training network under contract mrtn - ct-2006 - 035505 ` tools and precision calculations for physics discoveries at colliders ' .
k. melnikov and a. mitov , phys .
b * 620 * ( 2005 ) 69 ; i. blokland , a. czarnecki , m. misiak , m. slusarczyk and f. tkachov , phys .
d * 72 * ( 2005 ) 033014 ; h. m. asatrian , a. hovhannisyan , v. poghosyan , t. ewerth , c. greub and t. hurth , nucl .
b * 749 * ( 2006 ) 325 ; h. m. asatrian , t. ewerth , a. ferroglia , p. gambino and c. greub , nucl . phys .
b * 762 * ( 2007 ) 212 .
s. chen _ et al .
_ [ cleo collab . ] , phys . rev . lett .
* 87 * ( 2001 ) 251807 ; b. aubert _ et al . _
[ babar collab . ] , phys .
d * 77 * ( 2008 ) 051103 ; k. abe _ et al .
_ [ belle collab . ] , aip conf . proc .
* 1078 * ( 2009 ) 342 .
t. van ritbergen , phys .
b * 454 * ( 1999 ) 353 ; k. melnikov , phys .
b * 666 * ( 2008 ) 336 ; a. pak and a. czarnecki , phys . rev .
* 100 * ( 2008 ) 241807 and phys
. rev .
d*78 * ( 2008 ) 114015 ; m. dowling , a. pak and a. czarnecki , phys .
d * 78 * ( 2008 ) 074029 .
m. b. voloshin , phys .
b * 397 * ( 1997 ) 275 ; z. ligeti , l. randall and m. b. wise , phys .
b * 402 * ( 1997 ) 178 ; a. k. grant , a. g. morgan , s. nussinov and r. d. peccei , phys .
d * 56 * ( 1997 ) 3151 ; g. buchalla , g. isidori and s. j. rey , nucl .
b * 511 * ( 1998 ) 594 . | we briefly summarize the current status of perturbative calculations at next - to - next - to - leading - order ( nnlo ) accuracy in the @xmath0 decay rate as well as that of non - perturbative power - corrections . |
fesi is a non - magnetic @xcite , narrow - gap semiconductor @xcite at low temperatures .
its magnetic susceptibility @xmath4 increases with temperature and passes through a maximum at @xmath5 k @xcite .
fesi becomes metallic above 300 k@xcite .
the substitution of co for fe ( about 10% co ) yields a magnet with a helical spin order @xcite .
local density functional@xcite band structure calculations @xcite give the correct value of the semiconducting gap ( about 0.1 ev ) , but can not explain the large magnitude of @xmath4 . according to infrared and optical measurements @xcite , the gap of 50 mev is gradually filled with increasing temperature , with new spectral weight which can not be explained within the conventional band structure picture . in connection with a temperature - induced local moment ,
a model based on unified spin - fluctuation theory was proposed in ref .
which explains @xmath4 using a model density of states ( dos ) . in spite of the large number of publications devoted to measurements of x - ray photoelectron spectra ( xps ) and ultraviolet photoelectron spectra ( ups ) of fesi@xcite , most measurements were performed using polycrystalline samples which do not allow precise measurements on clean surfaces free of contamination . in this paper
, we present a full set of precise spectral measurements , data , including xps of the valence band and x - ray emission valence spectra for both components of fesi , which were obtained on the same single crystal , and providing experimental information about the distribution of total and partial dos in the valence band .
the already published information on the calculated electronic structure of fesi presented in ref.@xcite reveals only total dos and fe @xmath6 , si @xmath7 partial dos distributions in ref . , and fe @xmath6 , si @xmath8 dos for the cscl - type structure ( which is rather different from that known for the bulk fesi@xcite ) in ref . .
because of this we performed a new set of band structure calculations of fesi by two independent methods
linearized muffin - tin orbitals ( lmto ) and linearized augmented plane wave ( lapw ) which give more detailed information about the total and the fe @xmath6 , fe @xmath9 , si @xmath7 , si @xmath6 and si @xmath8 partial dos distributions .
the fe @xmath1 ( @xmath10@xmath11 transition ) x - ray emission spectrum was measured on the rsm-500 type x - ray vacuum spectrometer with a diffraction grating ( @xmath12 lines / mm and @xmath13 m ) and electron excitation .
the spectra were recorded in the second order of reflection by a secondary electron multiplier with a csi photocathode .
the energy resolution was about 0.350.40 ev .
the x - ray tube was operated at @xmath14 kev , @xmath15 ma . the si @xmath2 ( @xmath16@xmath8 transition ) x - ray emission spectrum was measured using a fluorescent johan - type vacuum spectrometer with a position - sensitive detector @xcite .
the pd @xmath17-x - ray radiation from a special sealed x - ray tube was used for the excitation of the fluorescent si@xmath2 spectra .
a quartz @xmath18 single crystal curved to @xmath19 mm served as a crystal - analyzer .
the spectra were measured with an energy resolution of approximately 0.20.3 ev .
the x - ray tube was operated at @xmath20 kev , @xmath21 ma . the si @xmath3 ( @xmath10@xmath22 transition ) x - ray emission spectra of fesi were taken from ref . , and the fe @xmath0 ( @xmath16@xmath9 transition ) x
- ray emission spectrum was reproduced from ref . .
the xps valence band spectra of fesi were measured using a perkin - elmer esca spectrometer ( phi 5600 ci , monochromatized al @xmath23 radiation ) .
the fesi single crystal was cleaved in high vacuum prior to the xps measurements .
the xps spectra were calibrated based on the au @xmath24-spectra of au metal ( @xmath25=84.0 ev ) .
x - ray emission spectra have been brought to the scale of binding energies with respect to the fermi level using the binding energies of the relevant initial ( core level ) states of the x - ray transitions as measured by the xps technique .
corresponding binding energies are @xmath26 ev , @xmath27 ev . the values of @xmath28 ev and @xmath29 ev were taken for comparison of fe @xmath1 and fe @xmath0 , si @xmath3 and si @xmath2 x - ray emission spectra of fesi .
the measured xps and x - ray emission spectra are shown in fig .
electronic structure calculations have been performed for the cubic fesi structure ( 8 atoms / cell , space group @xmath30 ) as determined in ref . and discussed in detail in ref . .
we have used the cubic lattice constant of @xmath31 , with fe atoms occupying the @xmath32 and equivalent positions of the @xmath33 structure , while si atoms occupy the @xmath34 and equivalent positions . in the calculations using the tight - binding lmto method @xcite
, we used space - filling atomic spheres of equal size on fe and si sites ( @xmath35 ) and no empty spheres were introduced .
the exchange - correlation potential as proposed by von barth and hedin @xcite was used .
dos calculated by the tetrahedron method over 470 * k*-points in the irreducible part of the brillouin zone are shown in fig .
[ spec ] , and compared with the experimental spectra .
our calculated electron bands are very close to those obtained by fu _
et al._@xcite using the augmented spherical wave method , and our calculated dos agree well with those of ref . .
we found a direct gap of 0.14 ev at the @xmath36 point of the brillouin zone and an indirect gap of 0.050.08 ev , in agreement with the resistivity measurement data reported in ref . . some small deviations from the bands calculated in ref .
using the lapw method without any shape approximation imposed on the potential seem to have quite negligible effect on the densities of states , in the present context of making a comparison with x - ray spectra using the lmto method .
we also carried out an independent lapw@xcite calculation , in which we used the local density approximation @xcite form for exchange and correlation given by ceperley and alder @xcite , as parameterized by perdew and zunger@xcite . in the expansion of the charge density and potential inside the muffin - tin spheres ( @xmath37 ) , lattice harmonics up to angular momentum of @xmath38
have been used , while the interstitial region was described by a plane wave expansion .
core orbitals were treated self - consistently , retaining , however , only the spherically - symmetric part of their density .
scalar relativistic effects ( neglecting spin - orbit coupling ) were included for the valence states , and full relativistic effects were included for the core states .
a large basis set consisting of both real - space orbitals ( inside the muffin - tin regions ) and plane waves was used .
the total number of basis functions used was @xmath39 570 per unit cell of 8 atoms .
the brillouin zone summations were performed using 24 * k*-vectors in the irreducible section of the brillouin zone in the self - consistency loop .
the densities of states calculated using the lapw method , with the tetrahedron integration over 176 * k*-points in the irreducible part of the brillouin zone , are shown in figs .
[ dospart ] and [ dostot ] .
note that the total dos calculated in both lmto and lapw methods per unit cell ( i.e. 4 formula units ) are in good absolute agreement , whereas the partial dos are somehow smaller in the lapw calculation ( fig .
[ dospart ] because they are attributed to non - overlapping muffin - tin spheres rather than space - filling atomic spheres as in lmto .
the expanded part of the total densities of states plot in the vicinity of the gap is shown at the inset in fig .
[ dostot ] .
two sharp peaks formed mostly by fe @xmath40 states are separated by the indirect gap of about 0.065 ev .
there seems to be generally good agreement between the several lda - based band structure calculation schemes in reproducing the bands , total densities of states , and bandwidths in fesi .
below , we concentrate on discussing particular features of partial densities of states of the constituents as revealed in the x - ray emission spectra and the present calculations .
the x - ray photoelectron valence band spectrum and x - ray emission valence band spectra of fe ( fe @xmath1 , fe @xmath0 ) and si ( si @xmath3 , si @xmath2 ) are presented in fig .
as is known , the xps valence band spectrum gives an information about the total dos distribution ( accurate up to the weight function depending on the atomic photoemission cross - sections , see ref . ) , whereas the fe @xmath1 , fe @xmath0 , si @xmath3 , si @xmath2 x - ray emission spectra ( in accordance to dipole selection rules ) give information about partial fe @xmath11 , fe @xmath9 , si @xmath22 and si @xmath8 densities of states , respectively .
these spectra are compared in fig .
[ spec ] with the results of the lmto band structure calculations by aligning the calculated and experimental positions of the fermi level .
we note that the experimental xps valence band spectrum reproduces the total dos distribution of fesi quite well .
especially , it is very important that in our experiments we have found the splitting of the main peak of xps spectrum in the range 02 ev which corresponds to the fe @xmath6 band splitting .
this distinct splitting was not found before in xps@xcite and ups@xcite measurements on fesi , and its absence was attributed in ref . to a hole lifetime broadening which increases with binding energies .
the similar splitting of the xps valence band spectrum of fesi was found in ref . only in the measurements done at low temperatures ( @xmath41=120 k ) . in our opinion , this discrepancy between fine structure in the present and previous xps ( ups ) valence band spectra is due to the high quality of the fesi single crystal which was used for the present measurements .
the low - energy si @xmath8 and si @xmath7 subbands are reflected in the xps valence band spectrum as low - intensity ( due to low values of respective photoinization cross - sections @xcite ) features located at binding energies of approximately 4.5 and 9 ev , respectively . the experimental fe @xmath1 x - ray emission spectrum reproduces fairly well the position of the center of gravity of calculated fe @xmath6 dos distribution in the valence band of fesi , but not the splitting of the fe @xmath6 band .
this is due to the large total distortion of the fe @xmath1 x - ray emission spectrum which includes the instrumental distortion ( about 0.4 ev ) and the width of the inner ( core ) fe @xmath10 level ( about 0.8 - 1.0 ev ) which is determined by the lifetime of the core - level vacancy under an x - ray transition .
the experimental fe @xmath0 spectrum shows two maxima located at binding energies of approximately 3.5 and 9.0 ev that is in accordance with the theoretical distribution of the fe @xmath9 partial dos .
it is also seen from both the theoretical and experimental spectra that fe @xmath9 states are hybridized mostly with si @xmath8 and si @xmath6 states .
the experimental si @xmath3 x - ray emission spectra corresponds to the @xmath10@xmath22 transition .
usually it is considered that the si @xmath6 states do not take part in the chemical bonding in @xmath6 transition metal silicides . as was emphasized in ref . , the high - energy subband of si@xmath3 x - ray emission spectra of transition metal silicides fesi , mnsi and nisi , as well as that of disilicides fesi@xmath42 , mnsi@xmath42 and nisi@xmath42 , can not be explained without an assumption that si @xmath6 states contribute to the chemical bonding .
subsequently , the same conclusion was drawn for pt silicides ( pt@xmath42si , ptsi ) in ref . based on an analysis of the si @xmath3 x - ray emission spectra of these compounds .
this conclusion is again confirmed now for fesi by two independent sets of band structure calculations ( see figs . [ spec ] and [ dospart ] ) .
one can see from fig .
[ spec ] that the si @xmath7 dos is concentrated in the range of binding energies about 6 - 13 ev where the most intensive low - energy subband of the si @xmath3 x - ray emission spectrum is located .
the center of gravity of the si @xmath6 states distribution corresponds to a binding energy about 2 ev where the high - energy maximum of the si @xmath3 spectrum is situated .
the si @xmath6 states are strongly hybridized with fe3@xmath40 states , and the same splitting of the si @xmath6 band ( about 2 ev ) as that in the fe @xmath6 subband is observed in the band structure calculations .
the energy position and fine structure of the si @xmath2 x - ray emission spectrum is in a good accord with the si3 @xmath43 partial dos distribution ( see fig . [ spec ] ) .
the results of high - energy spectroscopy measurements of fesi single crystals , including the xps valence band spectrum and the x - ray emission valence band spectra of both constituents , are presented .
they are compared with two independent _ ab initio _ band structure calculations of fesi , performed using the lmto and lapw methods , and good agreement between the experimental and theoretical spectra is found .
the authors are grateful to castor fu for sending his data on the calculated density of states of fesi .
avp , stu and mn appreciate the financial support of the deutsche forschungsgemeinschaft ( sfb 225 , graduate college ) .
this work was supported by russian foundation for fundamental research ( grant no.94 - 03 - 08040 ) and nato international scientific exchange program ( project htech lg 940861 ) .
zwl , bmk , and zps thank the support by the university research funds of the university of california at davis . on leave from institute of metal physics , russian academy of sciences , yekaterinburg , russia .
h. watanabe , h. yamamoto , and k. ito , j. phys.soc.jpn .
* 18 * , 995 ( 1963 ) .
v. jaccarino , g. k. wertheim , j. h. wernick , l. r. walker , and s. arajs , phys .
rev . * 160 * , 476 ( 1967 ) .
v. i. kaidanov , v. a. tselishchev , i. k. iesalniek , l. d. dudkin , b. k. voronov , and n. n. trusova , sov . phys
. semicond .
* 2 * , 382 ( 1968 ) .
r. wolfe , j. h. wernick , and s. e. haszko , phys .
* 19 * , 449 ( 1965 ) .
j. beille , j. voiron , and m. roth , solid state commun .
* 47 * , 399 ( 1983 ) .
p. hohenberg and w. kohn , phys
. rev . * 136 * , b864 ( 1964 ) ; w. kohn and l. j. sham , phys . rev . * 140 * , a1133 ( 1965 ) .
l. f. mattheiss and d. r. hamann , phys .
b * 47 * , 13114 ( 1993 ) .
castor fu , m. p. c. m. krijn , and s. doniach , phys .
b * 49 * , 2219 ( 1994 ) .
r. girlanda , e. piparo , and a. balzarotti , j. appl . phys . * 76 * , 2837 ( 1994 )
. z. schlesinger , z. fisk , h .-
zhang , m. b. maple , j. f. di tusa , and g. aeppli , phys .
lett . * 71 * , 1748 ( 1993 ) .
y. takahashi and t. moriya , j. phys .
. jpn . * 46 * , 1451 ( 1979 ) .
w. speier , e. v.leuken , j. c. fuggle , d. d. sarma , l. kumar , b. dauth , and k. h. j. buschow , phys .
b * 39 * , 6008 ( 1989 ) .
h. von knel , k. a. mder , e. mller , n. onda , and h. sirringhaus , phys .
b * 45 * , 13807 ( 1992 ) .
a. chainani , t. yokoya , t. morimoto , t. takahashi , s. yoshii , and m. kasaya , phys .
b * 50 * , 8915 ( 1994 ) . s .-
oh , j. w. allen , and j. m. lawrence , phys .
b * 35 * , 2267 ( 1987 ) .
a. kaklizaki , h. sugawara , i. nagakura , y. ishikawa , t. komatsubara , and t. ishii , j. phys .
. jpn . * 51 * , 2597 ( 1982 ) .
t. saitoh , a. sekiyama , t. mizokawa , a. fujimori , k. ito , h. nakamura , and m. shiga ( submitted to phys . rev .
lett . , august 29 , 1994 ) .
v. e. dolgih , v. m. cherkashenko , e. z. kurmaev , d. a. goganov , e. k. ovchinnikov , and yu . m. yarmoshenko , nucl . instrum .
methods * 224 * , 117 ( 1984 ) .
e. z. kurmaev , v. v. fedorenko , s. n. shamin , g. wiech , and y. kim , phys .
scr.*t41 * , 288 ( 1992 ) .
k. m. kolobova , ph.d.thesis , institute of metal physics , sverdlovsk ( ussr ) , 1970 .
o. k. andersen and o. jepsen , phys .
53 * , 2571 ( 1984 ) ; + o. k. andersen , z. pawlowska , and o. jepsen , phys .
b * 34 * , 5253 ( 1986 ) .
u. von barth and l. hedin , j. phys .
c * 5 * , 1629 ( 1972 ) .
castor fu , private communication . o. k. andersen , phys .
b * 12 * , 3060 ( 1975 ) ; e. wimmer , h. krakauer , m. weinert , and a. j. freeman , _ ibid .
_ * 24 * , 864 ( 1981 ) ; d. r. hamann , phys .
lett . * 42 * , 662 ( 1979 ) ; s .- h .
wei and h. krakauer , _ ibid .
_ * 55 * , 1200 ( 1985 ) ; s .- h .
wei , h. krakauer , and m. weinert , phys .
b * 32 * , 7792 ( 1985 ) . d. j. singh , _
planewaves , pseudopotentials , and the lapw method _ , ( kluwer , boston , 1994 ) . d. m. ceperley and b. j. alder , phys .
* 45 * , 566 ( 1980 ) .
j. p. perdew and a. zunger , phys .
b * 23 * , 5048 ( 1981 ) .
j. j. yeh and i. lindau , at .
. data tables * 32 * , 1 ( 1985 ) .
s. yamauchi , m. hirai , m. kusaka , m. iwami , h.nakamura , h. ohshima , and t. hattori , jpn .
. phys . * 33 * , l1012 ( 1994 ) . | the full set of high - energy spectroscopy measurements including x - ray photoelectron valence band spectra and soft x - ray emission valence band spectra of both components of fesi ( fe @xmath0 , fe @xmath1 , si @xmath2 and si @xmath3 ) are performed and compared with the results of _ ab - initio _ band structure calculations using the linearized muffin - tin orbital method and linearized augmented plane wave method . |
let @xmath1 be a @xmath7-dimensional riemannian model ( @xmath8 ) , namely a manifold admitting a pole @xmath9 and whose metric is given , in spherical coordinates around @xmath9 , by @xmath10 where @xmath11 denotes the canonical metric on the unit sphere @xmath12 and * @xmath13 is a @xmath14 nonnegative function on @xmath15 , positive on @xmath16 such that @xmath17 and @xmath18 for all @xmath19 . these conditions on @xmath13
ensure that the manifold is smooth and the metric at the pole @xmath9 is given by the euclidean metric ( * ? ? ?
* chapter 1 , 3.4 ) .
then , by construction , @xmath20 is the geodesic distance between a point @xmath21 whose coordinates are @xmath22 and @xmath9 .
let @xmath23 denote the laplace - beltrami operator on @xmath6 .
our paper concerns least energy solutions to the equation @xmath24 as a prototype of the nonlinearity think to @xmath25 , where @xmath26 is a suitable measurable function and @xmath27 if @xmath28 ( @xmath29 if @xmath30 ) , but most of the results stated in the paper hold for more general @xmath31 , see section [ partial sym ] . nonlinear elliptic equations like on manifolds with negative sectional curvatures have been the subject of intensive research in the past few years .
many papers are settled on the simplest example of manifold with negative curvature : the hyperbolic space @xmath32 , corresponding to @xmath33 in .
see @xcite and references therein , where @xmath34 is chosen . in this case
, a great attention has been devoted to the study of radial solutions ( non necessarily in the energy class ) either in @xmath32 @xcite or in the more general riemannian model @xcite .
see also @xcite where fully nonlinear elliptic equations have been recently studied in the same setting .
it becomes then a natural and interesting subject of investigation the study of symmetry properties of solutions to . in the hyperbolic setting ,
radial symmetry of solutions has been proved in @xcite for power - type nonlinearities and for positive solutions in the energy class .
see also @xcite .
the results in @xcite hold for quite general nonlinearities @xmath35 and non - energy solutions are also dealt .
furthermore , their extension to general manifolds is also discussed . in the wake of the seminal paper @xcite , the proofs of the just mentioned results rely on the moving plane method and
strongly exploit the structure of the space under consideration .
hence , their extension to general manifolds seems quite difficult to be reached . in @xcite this topic is addressed by requiring two kinds of assumptions : either group action properties , which generalize what happens in @xmath36 and @xmath32 , or suitable foliation conditions .
coming back to our riemannian model , the results in @xcite only apply if @xmath37 or @xmath38 ( @xmath39 ) , namely to the euclidean and hyperbolic cases , see open problem [ o : symmetry ] .
it is therefore appropriate to investigate whether , at least , some partial symmetry holds . in the present paper , under quite general assumptions on @xmath13 and @xmath31 , we prove that ground states to are foliated schwarz symmetric with respect to some point ( see theorem [ main ] ) . in particular , they are either radial symmetric or axially symmetric .
the same can be said for corresponding dirichlet boundary value problems ( see theorem [ main dir ] ) .
we refer to @xcite for related results about dirichlet problems on riemannian models .
we observe that our symmetry result admits nonlinearities of the type @xmath40 with no monotonicity condition with respect to @xmath5 .
as far as we are aware , this case was not covered by previous works , not even in the hyperbolic space @xmath3 .
we mention the paper @xcite where symmetry was proved for the solutions to a dirichlet problem posed on manifolds conformally equivalent to @xmath41 and for nonlinearities @xmath40 decreasing with respect to @xmath5 .
our result guarantees that , when they exist , least energy solutions to are foliated schwarz symmetric .
the problem of existence of least energy solutions to with subcritical growth can be easily handled if radial symmetry is a - priori assumed ( see @xcite ) . in this perspective , for instance , compactness is gained in @xcite by requiring suitable symmetry properties of solutions .
if no extra constrain is assumed , the loss of compactness may represent a serious obstacle to show existence . when @xmath34 and @xmath42 , existence of least energy solutions has been independently proved in @xcite and in @xcite .
both the proofs exploit peculiar properties of @xmath32 and can be hardly extended to a more general setting . an important contribution in this direction
is given in @xcite where existence is proved for power - type nonlinearities when the equation is posed on a weakly homogenous space .
we show in subsection [ ex ] that , under the weakly homogeneity assumption , our riemannian model reduces either to @xmath41 or @xmath32 .
nevertheless , a thorough analysis of the peculiar structure of allows us to obtain some compactness and finally to prove in theorem [ compact 1 ] existence of least energy solutions to for suitable families of @xmath31 and for quite general @xmath13 .
it is worth noticing that theorem [ compact 1 ] applies to riemannian models with unbounded sectional geometry ( see remark [ r : example ] ) . the paper is organized as follows . in section
[ setting ] we fix the notation and describe our geometric setting
. section [ partial sym ] contains the main theorems : in subsection [ partial - symmetry ] we state the partial symmetry results and in subsection [ ex ] we state the existence results .
all the proofs are given in sections 4 - 7 .
the following table summarizes most of the notation we shall use in the paper . * for any @xmath43 and @xmath44 we denote by @xmath45 the scalar product on @xmath46 associated with the metric @xmath47 . * for any @xmath43 and @xmath48 we denote by @xmath49 the norm of the vector @xmath50 . *
@xmath51 denotes the volume measure in @xmath1 . *
@xmath52 denotes the riemannian gradient in @xmath1 . * for any @xmath53 we define the banach space @xmath54 endowed with the corresponding @xmath55-norm . * for any @xmath56 and any measurable function @xmath57 we define the banach space @xmath58 endowed with the corresponding weighted @xmath55-norm .
* we denote by @xmath59 the classical sobolev space in @xmath6 , i.e. @xmath60 endowed with the usual @xmath61-norm .
let @xmath1 be an @xmath7-dimensional riemannian model whose metric is defined by formula with a function @xmath13 satisfying condition ( h ) .
let @xmath62 be the area of the @xmath7-dimensional unit sphere .
then @xmath63 represent , respectively , the area of the geodesic sphere @xmath64 and the volume of the geodesic ball @xmath65 , where @xmath65 denotes the geodesic ball centered at @xmath9 of radius @xmath5 , i.e. @xmath66 the riemannian laplacian of a scalar function @xmath31 on @xmath6 is given , in the above coordinates , by @xmath67+\frac1{(\psi(r))^2}\delta_{{\mathbb s}^{n-1}}f(r,\theta),\ ] ] where @xmath68 is the riemannian laplacian on the unit sphere @xmath12 .
in particular , for _ radial functions , namely functions depending only on @xmath5 , one has @xmath69^\prime = f^{\prime\prime}(r)+(n-1)\frac{\psi^\prime(r)}{\psi(r)}f^\prime(r),\ ] ] where from now on a prime will denote , for radial functions , derivative w.r.t . @xmath5 . by standard arguments
we deduce that the bottom of the @xmath70 spectrum of @xmath71 in @xmath6 admits the following variational characterization : @xmath72 notice that if @xmath73 , then @xmath74 ( see ( * ? ? ?
* lemma 4.1 ) ) .
let @xmath75 be a orthonormal frame on @xmath1 , where @xmath76 correspond to the spherical coordinates and @xmath77 corresponds to the radial coordinate .
the curvature operator , the ricci curvature and the scalar curvature can be computed in terms of @xmath13 ( see ( * ? ? ?
* chapter 3 , 2.3 ) , @xcite ) .
the curvature operator is given by @xmath78 @xmath79 the ricci curvature is given by @xmath80 where @xmath81 , and the scalar curvature is @xmath82 hence , if @xmath83 then the ricci curvature of @xmath1 is bounded from below , i.e. there exists a real number @xmath84 such that @xmath85 .
assume furthermore that @xmath86 then all the sectional curvatures are nonpositive and , by hadamard theorem ( * ? ? ?
* theorem 3.1 ) , the injectivity radius of @xmath1 is @xmath87 . finally , if @xmath88 , and hold , by ( * ? ? ?
* theorem 2.21 ) the following sobolev embedding holds : @xmath89 where @xmath90 denotes the critical sobolev exponent given by @xmath91 _
throughout this section @xmath6 denotes a @xmath7-dimensional riemannian model , @xmath2 , with the function @xmath13 in satisfying ( h ) .
consider the equation @xmath92 where @xmath93 and @xmath94 is a function satisfying the following conditions : * @xmath95 for any @xmath96 and , denoting by @xmath97 the derivative with respect to @xmath98 , we have @xmath99 where @xmath100 , @xmath101 is a positive constant , @xmath102 is such that @xmath103 and @xmath59 is continuously embedded into @xmath104 .
notice that @xmath26 denotes the radial function on @xmath6 defined by @xmath105 ; * @xmath106 for any @xmath96 and @xmath107 .
many situations are allowed by assumptions ( h ) on @xmath13 and @xmath108 on @xmath31 .
for example , if we further assume that @xmath13 satisfies - , then thanks to we may choose @xmath109 .
on the other hand , if we do not require any further restriction on @xmath13 , we can always find a suitable function @xmath31 satisfying @xmath108 , as one can see from remark [ r : example ] ( where examples of manifolds with unbounded negative sectional curvatures are given ) . thanks to @xmath108 , we may define the action functional associated with @xmath110 where @xmath111 , and the nehari manifold @xmath112 we say that @xmath113 is a _ least energy _ or _ ground state _ solution to if it achieves the following infimum @xmath114 by @xmath108 the set @xmath115 is a natural constraint in the sense that every constraint stationary point of the functional @xmath116 is a `` free stationary point '' of @xmath116 itself , i.e. @xmath117 we also observe that assumption @xmath118 yields @xmath119 for every @xmath120 so that by @xmath121 and if it is achieved by some function @xmath113 then @xmath122 , see the proof of lemma [ positivity ] . [
d : foliated ] a continuous function @xmath123 is foliated schwarz symmetric with respect to some @xmath124 if the value of @xmath125 at @xmath126 only depends on @xmath5 and @xmath127 , where @xmath128 denotes the standard scalar product in @xmath129 .
we state our main symmetry result .
[ main ] let @xmath1 be the riemannian model defined by with @xmath13 satisfying assumption @xmath88 .
let @xmath130 be a @xmath131 function satisfying assumptions @xmath108 .
then , any least energy solution @xmath125 to is foliated schwarz symmetric with respect to some @xmath124 and strictly of one sign in @xmath6 . moreover , either @xmath125 is radial or @xmath125 is strictly decreasing with respect to @xmath132 $ ] for any @xmath96 when @xmath133 in @xmath6 ( respectively strictly increasing with respect to @xmath132 $ ] for any @xmath96 when @xmath134 in @xmath6 ) . theorem [ main ] nothing says about existence of least energy solutions .
nevertheless , when they exist , least energy solutions @xmath125 to are axially symmetric with respect to the axis @xmath135 and @xmath136 .
when @xmath137 and @xmath35 , it is known that @xmath125 does not depend on @xmath138 , namely @xmath125 is radial , see @xcite and @xcite . for general manifolds ,
a first step in this direction is made by proposition [ separa ] below .
[ separa ] under the same assumptions of theorem [ main ] , let @xmath139 be a nonradial least energy solution to .
if the map @xmath140 is strictly increasing and locally lipschitz then @xmath125 does not admit a representation of the type @xmath141 where @xmath142 and @xmath143\rightarrow { \mathbb r}$ ] . the same proof of theorem [ main ] yields foliated schwarz symmetry of least energy solutions to the dirichlet problem @xmath144 where @xmath145 is fixed .
indeed , by simply replacing @xmath146 in the proof of theorem [ main ] with @xmath147 where @xmath148 let @xmath1 be the riemannian model defined by with @xmath13 satisfying assumption @xmath88 .
let @xmath149 be a @xmath131 function satisfying assumptions @xmath108 . then , any least energy solution @xmath125 to is foliated schwarz symmetric with respect to some @xmath124 and strictly of one sign in @xmath150 .
moreover , either @xmath125 is radial or @xmath125 is strictly decreasing with respect to @xmath132 $ ] for any @xmath151 when @xmath133 in @xmath150 ( respectively strictly increasing with respect to @xmath132 $ ] for any @xmath151 when @xmath134 in @xmath150 ) .
let @xmath152 denote the set of functions @xmath153 which are radially symmetric .
we define @xmath154 then , @xmath155 . when @xmath156 , least energy solutions ( if they exist ) must be nonradial and by theorem [ main ] they are axially symmetric .
the same hold for @xmath157 and @xmath158 if @xmath42 and @xmath159 , it has been recently proved in @xcite that @xmath160 as @xmath161 . in this case , theorem [ main dir ] gives a sharp information .
[ o : symmetry ] when the nonlinearity depends on @xmath5 , our symmetry result can be considered optimal in the sense explained above .
one may ask if something more can be said about symmetry of least energy solutions of when @xmath31 only depends on @xmath125 . in @xcite , radial symmetry about some point
is obtained for positive solutions of when @xmath162 and @xmath31 is a suitable nonlinearity depending only on @xmath125 . here , by radial symmetry of a function @xmath123 with respect to a point @xmath163 , which is not necessarily the pole of the riemannian model , we mean that @xmath125 is constant on geodesic balls centered at @xmath164 .
we leave as an open problem to show if radial symmetry about some point occurs also in more general riemannian models , at least for ground state solutions .
note that the assumptions on @xmath1 introduced in ( * ? ? ?
* section 4 ) are too restrictive in the setting of our riemannian models .
indeed , they would imply that @xmath6 has constant scalar curvature and then @xmath37 or @xmath165 , i.e. @xmath1 is either the euclidean space or an hyperbolic space whose scalar curvature is @xmath166 ( see also proposition [ p : constant - k ] ) .
in subsection [ partial - symmetry ] we discussed partial symmetry of least energy solutions for the equation under the assumptions @xmath108 . here
we address the problem of existence of least energy solutions of . as already mentioned in the introduction , one way to gain existence of least energy solutions
is to assume that the manifold @xmath6 satisfies the so - called weakly homogeneity condition , see @xcite . for completeness
we recall here its definition .
[ d : weakly - hom ] let @xmath1 be a @xmath7-dimensional complete riemannian manifold .
we say that @xmath6 is weakly homogeneous if there exists a group @xmath167 of isometries of @xmath6 and @xmath168 such that for every @xmath169 , there exists @xmath170 such that @xmath171 . here
@xmath172 represents the geodesic distance in @xmath6 .
assume that @xmath1 is a riemannian complete weakly homogeneous manifold ( not necessarily a riemannian model in the sense of ) with bounded geometry .
proceeding as in @xcite , where the pure power equation is dealt , we prove existence of least energy solutions for the following class of equations @xmath173 with @xmath174 and @xmath175 satisfying suitable assumptions . similarly to what we did for we may define the nehari manifold associated with by @xmath176 and the least energy solutions of as the solutions @xmath113 of satisfying @xmath177 where @xmath178\ , dv_g-\int_m h(v)\ , dv_g\ ] ] and @xmath179 for any @xmath180 . [
p : existence ] let @xmath1 be a riemannian complete weakly homogeneous manifold of dimension @xmath2 with positive injectivity radius and bounded curvature .
let @xmath181 satisfy @xmath182 with @xmath183 as in and @xmath184 where @xmath185 denotes the geodesic distance in @xmath1 . furthermore ,
suppose that @xmath186 satisfies @xmath187 with @xmath188 , @xmath189 and that there exists @xmath190 such that @xmath191 then admits a least energy solution in the sense of .
notice that the geometric assumptions of proposition [ p : existence ] ensure the validity of the sobolev embedding ( * ? ? ?
* theorem 2.21 ) .
we observe that condition , is the ambrosetti - rabinowitz condition .
we refer to @xcite and references therein for possible alternative assumptions on @xmath175 in the euclidean setting .
but in the present paper we are mainly interested in the study of over the riemannian model
. one may ask what happens in this case if the weakly homogeneity condition is assumed .
an answer to this question is given by the following [ p : constant - k ] let @xmath1 be a @xmath7-dimensional riemannian model defined by with @xmath2 and @xmath13 satisfying assumption @xmath88 and suppose that @xmath1 is weakly homogeneous
. then the scalar curvature is constant .
moreover if we denote by @xmath84 this constant then only two alternatives may occur : * @xmath192 and @xmath1 is a hyperbolic space , namely @xmath193 * @xmath194 and @xmath1 is the euclidean space , namely @xmath37 . for results concerning manifolds with constant curvature see @xcite .
it is clear from proposition [ p : constant - k ] that weakly homogeneity becomes a too restrictive condition if @xmath1 is the riemannian model .
therefore , we look for some alternative conditions on @xmath1 and on the nonlinearity in which guarantee compactness of the embeddings of @xmath59 into suitable weighted lebesgue spaces .
more precisely , we assume that @xmath195 is such that * @xmath95 for any @xmath96 and , denoting by @xmath97 the derivative with respect to @xmath98 , we have @xmath196 where @xmath188 , @xmath101 is a positive constant , @xmath197 with @xmath183 as in , and @xmath103 is a nonnegative function ; * there exists @xmath190 such that @xmath198 where @xmath199 is as in @xmath200 , @xmath201 and @xmath202 .
we are ready to state the main result of this subsection where we restrict ourselves to @xmath28 .
[ compact 1 ] let @xmath1 be the @xmath7-dimensional riemannian model defined by with @xmath203 and @xmath13 satisfying assumption @xmath88 .
let @xmath204 satisfy @xmath205 .
letting @xmath206 and @xmath26 be as in @xmath200 , we assume that * @xmath207 for every @xmath208 ; * the functions @xmath209 and @xmath210 are nondecreasing in @xmath211 for some @xmath212 ; * @xmath213 as @xmath214 . then admits a least energy solution in the sense of .
[ r : example ] it is worth noticing that theorem [ compact 1 ] does not work for @xmath37 ( ( ii ) and ( iii ) would yield a contradiction ) while it works if @xmath215 , @xmath216 , @xmath217 , @xmath218 , @xmath219 for @xmath220 , with @xmath221 and @xmath222 .
namely , unbounded negative sectional curvatures are allowed . for any of the above models , as an explicit example of @xmath31 ,
take @xmath223 with @xmath224 , where @xmath225 and @xmath197 .
let @xmath125 be the least energy solution found in theorem [ compact 1 ] .
we observe that if in theorem [ compact 1 ] we add the assumption @xmath226 for any @xmath96 and @xmath227 , then @xmath125 satisfies all the sign and symmetry properties stated in theorem [ main ] .
next we want to focus our attention on the condition @xmath228 required in @xmath200 .
if this condition is dropped , then existence of nontrivial constant sign solutions may fail as shown in the following [ p : non - ex ] let @xmath1 be the @xmath7-dimensional riemannian model defined by with @xmath203 and @xmath13 satisfying assumption @xmath88 .
let @xmath31 be in the form @xmath229 with @xmath230 and suppose it satisfies @xmath200 and @xmath231 .
furthermore , suppose that ( i)-(iii ) in theorem [ compact 1 ] hold true with @xmath26 as in @xmath200 .
then does not admit any nontrivial constant sign solution @xmath153 . in proposition
[ p : non - ex ] we assumed that @xmath230 so that one may ask what happens when @xmath232 .
a partial answer can be found in ( * ? ? ?
* theorem 1.1 ) where nonexistence of positive @xmath233-solutions is proved when @xmath1 coincides with the hyperbolic space @xmath3 and @xmath234 .
we conclude this section with the following open problem : [ o : existence ] in order to guarantee existence of least energy solutions we made two different kinds of assumptions : either the weakly homogeneity and the bounded geometry of @xmath6 in proposition [ p : existence ] or the compactness of the embedding @xmath235 in theorem [ compact 1 ] .
both the results do not cover the case in which @xmath6 is a riemannian model with nonconstant curvature and @xmath236 , @xmath188 in .
as far as we are aware , the existence of a least energy solution for the equation @xmath237 is still an open problem when @xmath6 is a general manifold different from @xmath238 or @xmath3 .
note that @xmath125 is a least energy solution of if and only if @xmath125 is a minimizer of @xmath239 we know that admits a positive radial solution which is a minimizer for @xmath240 where @xmath241 denotes the space of radial functions in @xmath59 , see ( * ? ? ?
* theorem 2.5 ) .
we ask whether the two constants @xmath242 and @xmath243 coincide
. this question is strictly related with open problem [ o : symmetry ] .
indeed , once proved the existence of a minimizer @xmath125 for @xmath242 , then one way to establish the validity of the identity @xmath244 is to check whether or not @xmath125 is a radially symmetric function .
we first prove some preliminary lemmas which we shall apply in the proof of the symmetry result .
[ positivity ] suppose that @xmath13 satisfies @xmath88 and @xmath31 satisfies @xmath108 .
then , any least energy solution to belongs to @xmath245 and it is either strictly positive or strictly negative in @xmath6
. let @xmath125 be a least energy solution of .
standard brezis - kato type estimates @xcite combined with elliptic regularity estimates yield @xmath246 for any @xmath247 .
then , assumption @xmath248 combined with classical schauder estimates yields @xmath249 . in order to prove that @xmath125 does not change sign
we first show that @xmath250 .
indeed by @xmath118 and a simple integration by parts we infer @xmath251 for any @xmath96 and @xmath107 , and hence since @xmath113 we obtain @xmath252\ , dv_g>0 \ , .\ ] ] to show that @xmath125 is positive or negative , let @xmath253 and @xmath254 denote the positive and negative part of @xmath125 .
suppose that @xmath255 are not identically equal to zero in @xmath6 .
testing with @xmath256 and @xmath257 , one gets @xmath258 .
then , @xmath259 a contradiction with @xmath260 .
hence , we have @xmath261 or @xmath262 .
an application of the strong maximum principle yields the strict inequalities , see @xcite . for every @xmath263 ,
let @xmath264 and @xmath265 . for @xmath263
, we define @xmath266 by @xmath267 . for simplicity , we also put @xmath268 and @xmath269 for @xmath270 when the underlying @xmath271 is understood . for a measurable function @xmath272 , we define the _ polarization _ @xmath273 of @xmath274 relative to @xmath275 by @xmath276 [ strict - alt ]
suppose that @xmath13 satisfies @xmath88 and @xmath31 satisfies @xmath108 .
let @xmath125 be a ( positive ) least energy solution to and @xmath277 . then , one of the following is true : * @xmath278 in @xmath279 , + * @xmath280 in @xmath279 , + * @xmath281 in @xmath6 .
let us define @xmath282 and @xmath283 .
similarly to ( * ? ? ?
* proposition 2.3 ) , by we have @xmath284 \\[7pt ] \nabla_g u(p_h(x ) ) & \qquad \text{a.e . on }
( h\cap m^- ) \cup [ ( m\setminus h)\cap m^+ ] \ , .
\end{cases}\ ] ] by , and the change of variable @xmath285 we obtain @xmath286 in the same way we also obtain @xmath287 combining and we deduce that if @xmath288 is a minimizer of @xmath116 on @xmath115 , then also @xmath289 is a minimizer of @xmath116 on @xmath115 .
therefore , it follows that @xmath290 is a ( positive ) least energy solution of , see lemma [ l : relation - ps ] for more more details .
following @xcite , we consider @xmath291 by lemma [ positivity ] we have that @xmath292 .
furthermore , from @xmath248 and the fact that @xmath293 is an isometry in @xmath6 , we deduce @xmath294+[f(r , u_h)-f(r , u \circ
p_h)]\geq 0\ \ \ { \rm on}\ h \ , .\end{aligned}\ ] ] for every @xmath295 , let @xmath296 be a ball such that @xmath297
. by the strong maximum principle for elliptic operators , see @xcite , we deduce that either @xmath298 in @xmath299 or @xmath300 in @xmath301 and hence we conclude that either @xmath302 on @xmath275 or @xmath303 on @xmath304 . for any @xmath263 we now define @xmath305 we will prove theorem [ main ] with the help of the following characterization that can be deduced by ( * ? ? ? * proposition 2.4 ) .
[ sobol2 ] let @xmath263 .
a continuous function @xmath272 is foliated schwarz symmetric with respect to @xmath306 if and only if @xmath307 for every @xmath308 .
* proof of theorem [ main ] . *
it is not restrictive assuming that @xmath125 is a positive least energy solution to since otherwise one may define @xmath309 and obtain a positive least energy solution of @xmath310 where @xmath311 defined for any @xmath96 and @xmath227 satisfies assumptions @xmath108 .
take @xmath312 such that @xmath313 then , for any @xmath314 there exists @xmath315 such that @xmath316 and hence by lemma [ strict - alt ] we deduce that only case ( i ) or ( iii ) in the same lemma may occur ; in both situations this yields @xmath317 hence , @xmath125 is foliated schwarz symmetric with respect to @xmath306 by lemma [ sobol2 ] .
therefore we can write @xmath136 , where @xmath318 .
it remains to prove that @xmath319 we follow the argument in @xcite .
we already know that no half - space @xmath320 satisfies property ( ii ) of lemma [ strict - alt ] . moreover , if property ( i ) of this lemma holds for all half - spaces @xmath320 , then @xmath321 is strictly decreasing in @xmath322 for every @xmath96 .
it remains to consider the case where property ( iii ) of lemma [ strict - alt ] holds for some @xmath323 .
let @xmath324 be the angle formed by @xmath306 and the hyperplane @xmath325 .
. then @xmath327 .
moreover , ( iii ) implies that @xmath328 for @xmath96 .
since @xmath125 is nonincreasing in the angle @xmath322 , we conclude that @xmath329 for all @xmath330 . from lemma [ strict - alt ]
we then deduce that ( iii ) holds for all @xmath320 for which the angle between @xmath306 and @xmath275 is less then @xmath331 .
then , by the same argument as before , @xmath329 for all @xmath332
. arguing successively , in a finite number of steps we obtain @xmath329 for all @xmath333 .
this shows that @xmath125 is radial and completes the proof of .
* proof of proposition [ separa ] . *
[ s : separabili ] we exploit the idea of ( * ? ? ?
* remark 6.3 ) .
it is not restrictive to assume @xmath133 . by contradiction ,
let @xmath335 .
since we are also assuming that @xmath125 is not radial , then by theorem [ main ] we deduce that @xmath175 is strictly decreasing in @xmath336 $ ] .
few computations yield @xmath337 if @xmath338 , then @xmath339 and hence for any @xmath340 there exists @xmath341 such that @xmath342 where @xmath343 .
namely , there exists @xmath344 such that @xmath345 for any @xmath346 .
this gives a contradiction as @xmath347 since @xmath125 is a classical solution of in view of lemma [ positivity ] .
@xmath334 we first state some basic facts .
let @xmath348 be measurable locally bounded functions . for @xmath349 ,
we denote by @xmath350 the set of all measurable functions @xmath125 on @xmath238 such that @xmath351 and by @xmath352 the set of all functions @xmath353 such that @xmath354 .
then , we associate to sobolev and @xmath55 spaces on the model manifold @xmath6 suitable weighted sobolev and @xmath55 spaces on @xmath238 .
[ norme ] set @xmath206 and suppose that @xmath207 for all @xmath96 . for every @xmath28 , the following hold : * @xmath355 ; * @xmath356 the equality ( i ) is an immediate consequence of the formula for the riemannian measure in polar coordinates . to prove ( ii ) notice that @xmath357\,\psi^{n-1}\ , dr\,d\theta\\ & \geq \int_{0}^{+\infty}\int_{{\mathbb s}^{n-1 } } [ u_r^2 + \
, \frac{1}{r^{2}}|\nabla_{{\mathbb s}^{n-1 } } u|^2]\,r^{n-1}\ , \big(\frac{\psi}{r}\big)^{n-3 } dr\,d\theta\\ & = \|\nabla u\|^2 _ { l^2(\mathbb r^n ; \phi^{n-3 } ) } \ , .
\end{aligned}\ ] ] we will exploit the following [ embedding ] let @xmath28 , @xmath100 and @xmath348 be locally bounded radial weights on @xmath238 , i.e. measurable and a.e .
positive functions in @xmath41 .
suppose there exists @xmath145 , a positive measurable function @xmath358 defined on @xmath359 , @xmath360 and @xmath361 such that * @xmath362 ; * @xmath363 assume furthermore that there exist two positive measurable functions @xmath364 defined on @xmath359 such that @xmath365 if @xmath366 then the embedding @xmath367 is continuous and if @xmath368 , the embedding is also compact .
the proof follows by slightly modifying the proof of ( * ? ? ?
* theorem 18.7 ) . by combining lemma [ norme ] with proposition [ embedding ]
we get [ our embedding ] let @xmath28 , @xmath100 and @xmath1 be the @xmath7-dimensional riemannian model defined by with @xmath13 satisfying assumption @xmath88 .
let @xmath103 be a nonnegative function such that ( i)-(iii ) of theorem hold .
then , the embedding @xmath235 is continuous and if @xmath368 , the embedding is also compact . by lemma [ norme ] ( ii ) we have that @xmath369 .
on the other hand , by applying proposition [ embedding ] with @xmath370 and @xmath371 we deduce that the embedding @xmath372 is continuous and if @xmath368 , it is also compact .
this , combined with lemma [ norme ] ( i ) , proves the statement .
let us consider the nehari manifold @xmath115 defined in and let us define the functional @xmath373 by @xmath374 where @xmath375 denotes the frchet derivative of @xmath116 .
in this way we have @xmath376 and thanks to condition @xmath118 we also have @xmath377 for any @xmath113 . in the sequel we use the notation @xmath378 in the next lemma
we prove some properties of the nehari manifold exploiting in the proofs the results contained in @xcite .
[ l : nehari-1 ] suppose that all the assumptions of theorem [ compact 1 ] are satisfied . then we have * the functional @xmath116 introduced in is well defined on @xmath59 and it is of class @xmath131 over @xmath59 ; * the set @xmath115 defined in is nonempty and for any @xmath379 there exists a unique @xmath380 such that @xmath381 ; * if for any @xmath379 we define @xmath382 for any @xmath383 , then @xmath384 and it satisfies @xmath385 for any @xmath386 , @xmath387 and @xmath388 for any @xmath389 ; * the following estimates hold true : @xmath390 for any @xmath391 compact in @xmath392 with @xmath392 as in ; * the map @xmath393 is a homeomorphism from @xmath392 onto @xmath115 such that @xmath394 for any @xmath395 . * ( i ) * by @xmath200 we have that latexmath:[\[\label{eq : est - f - f } for any @xmath96 and @xmath227 . combining these two estimates with lemma [ our embedding ]
the proof of ( i ) follows in a standard way . *
( ii)-(iii ) * consider an arbitrary nontrivial function @xmath397 and consider the function @xmath382 defined for any @xmath383 . by , lemma [ our embedding ] and
the fact that @xmath228 we obtain @xmath398 for some suitable constants @xmath399 .
this shows that @xmath116 is strictly positive in a neighborhood of @xmath400 and , in turn , also the function @xmath401 is positive in a right neighborhood of @xmath402 . on the other hand by @xmath231 and integration
it follows @xmath403 where @xmath404 is fixed arbitrarily and @xmath405 .
therefore , for @xmath406 we have @xmath407 as @xmath408 being @xmath190 and @xmath409 as @xmath408 in view of and since , by ( i ) and ( iii ) in theorem [ compact 1 ] , @xmath26 is globally bounded .
summarizing @xmath384 , @xmath410 , @xmath401 is positive in a right neighborhood of @xmath402 and eventually negative as @xmath408 and hence there exists @xmath380 such that @xmath387 .
this implies @xmath381 thus proving that @xmath411 .
it remains to prove that @xmath412 is the unique stationary point of @xmath401 in @xmath413 . to this purpose ,
for any @xmath406 we define the function @xmath414 for any @xmath383 in such a way that stationary points of @xmath401 coincide with zero points of @xmath415 .
by @xmath118 we have @xmath416\ , dv_g<0 \qquad \text{for any } u\in \mathcal n \ , .\end{aligned}\ ] ] this implies that if @xmath379 and @xmath383 is such that @xmath417 then @xmath418 and @xmath419 thus proving that @xmath415 admits a unique positive zero point .
this completes the proof of ( ii ) and ( iii ) . *
( iv ) * the first two inequalities in ( iv ) follows from and ( iii ) . in order to prove the third inequality let us proceed by contradiction by supposing that there exists a sequence @xmath420 such that @xmath421 and @xmath422 as @xmath423 .
for simplicity in the rest of the proof we write @xmath424 in place of @xmath425 . proceeding as in , as @xmath423
we obtain @xmath426 since @xmath427 & \int_{\{|t_k v_k|<\sigma\ } } h(r , t_k v_k)\ , dv_g = o(t_k^2 ) \qquad \text{as } k\to + \infty \ , , \quad \text{and } \\[8pt ] & \liminf_{k\to + \infty } \int_{\{|t_k v_k|\ge \sigma\ } } \upsilon(r)|v_k|^\mu dv_g \ge \int_m \upsilon(r)|v|^\mu dv_g>0\end{aligned}\ ] ] where we used the strong convergence @xmath428 in @xmath59 and the fatou lemma .
the proof is complete since contradicts the fact that @xmath429 is positive on @xmath115 .
* ( v ) * it follows immediately applying proposition 8 in @xcite .
indeed assumptions ( @xmath430)-(@xmath431 ) in @xcite are an immediate consequence of ( iii ) and ( iv ) .
next we define @xmath432 by putting @xmath433 for any @xmath395 .
we say that a sequence @xmath434 is a palais - smale sequence for @xmath435 if @xmath436 is bounded and @xmath437 as @xmath423 , i.e. @xmath438 where @xmath439 denotes the tangent space to @xmath392 at @xmath440 . in the next result
we show that there is a strict relationship between palais - smale sequences for @xmath435 and @xmath116 .
[ l : relation - ps ] suppose that all the assumptions of theorem [ compact 1 ] are satisfied .
then we have * if @xmath441 is palais - smale sequence for @xmath435 then @xmath442 is a palais - smale sequence for @xmath116 ; * if @xmath443 is a critical point for @xmath435 then @xmath444 is a nontrivial critical point for @xmath116 ; * if @xmath113 is a constrained critical point for @xmath116 then @xmath125 is a critical point for @xmath116 .
the proof of ( i)-(ii ) follows immediately applying corollary 10 in @xcite .
indeed assumptions ( @xmath430)-(@xmath431 ) in @xcite are an immediate consequence of ( iii ) and ( iv ) in lemma [ l : nehari-1 ] . in order to prove ( iii ) we observe that if @xmath113 is a constrained critical point for @xmath116 then by the lagrange multiplier method there exists @xmath445 such that @xmath446 .
in particular @xmath447 and by we know that @xmath448 thus proving that @xmath194 and , in turn , that @xmath449 . then we prove that the functional @xmath116 satisfies the palais - smale condition .
[ l : ps - validity ] suppose that all the assumptions of theorem [ compact 1 ] are satisfied .
then the functional @xmath116 satisfies the palais - smale condition .
let @xmath450 be a palais - smale sequence for @xmath116 .
we first prove that @xmath450 is bounded in @xmath59 .
since @xmath451 is bounded and @xmath452 is @xmath453 as @xmath423 , then by @xmath454 we obtain @xmath455\ , dv_g \ge c_2 \|v_k\|_{h^1}^2\end{aligned}\ ] ] for some suitable positive constants @xmath456 .
this proves boundedness of @xmath450 .
then , up to a subsequence , we may assume that there exists @xmath397 such that @xmath457 weakly in @xmath59 .
hence , by lemma [ our embedding ] @xmath428 in @xmath104 .
by @xmath454 we deduce that @xmath458 and by the strong convergence @xmath428 in @xmath104 we obtain @xmath459 by and the fact that @xmath450 is a weakly convergent palais - smale sequence , we infer @xmath460 since @xmath228 then the map @xmath461 is an equivalent norm in @xmath59 and hence , together with the weak convergence @xmath457 , yields @xmath462 strongly in @xmath59 .
this completes the proof of the lemma .
* proof of theorem [ compact 1 ] .
* let @xmath463 be a minimizing sequence for @xmath116 on @xmath115 and for any @xmath464 define @xmath465 .
then @xmath466 is a minimizing sequence for @xmath435 on @xmath392 . by a classical argument based on the ekeland variational principle
we deduce that there exists a sequence @xmath434 such that @xmath467 , @xmath468 and @xmath437 as @xmath423 ( see for example ( * ? ? ?
* corollary a3 ) for more details ) .
this means that @xmath450 is a palais - smale sequence for @xmath435 on @xmath392 and hence by lemma [ l : relation - ps ] ( i ) it follows that @xmath469 is a palais - smale sequence for @xmath116 where we put @xmath470 for any @xmath464 . by lemma [ l : ps - validity ]
we deduce that up to a subsequence @xmath471 strongly converges in @xmath59 to a function @xmath153 .
this proves that @xmath113 , @xmath125 is a minimizer for @xmath116 on @xmath115 and by lemma [ l : relation - ps ] ( iii ) @xmath125 is a nontrivial critical point for @xmath116 .
@xmath334 * proof of proposition [ p : non - ex ] .
* we follow closely the argument used in the proof of theorem 1.1 in @xcite . for completeness
we give here the details .
first of all using a standard truncation argument one can show that the space @xmath472 of @xmath473 functions with compact support is dense in @xmath59 .
then we show that @xmath474 with @xmath475 clearly the map @xmath476 is nonincreasing and @xmath477 for any @xmath145 and moreover by density of @xmath472 in @xmath59 it follows that for any @xmath478 there exists @xmath479 such that @xmath480 taking @xmath145 large enough such that @xmath481 we obtain @xmath482 . using the same argument introduced in the proof of theorem [ main ] it is not restrictive assuming that admits a positive solution @xmath153 . for any @xmath145 denote by @xmath483 a positive eigenfunction of @xmath71 corresponding to the first eigenvalue @xmath484 .
then by @xmath231 and integration by parts we obtain @xmath485 \int_{b_r } u\varphi_{1,r } \ , dv_g\end{aligned}\ ] ] which yields @xmath486 being @xmath487 . passing to the limit as @xmath488
we obtain @xmath489 , a contradiction . @xmath334
since @xmath490 the hilbert space @xmath59 may be endowed with the following equivalent norm @xmath491 we also introduce the functional @xmath492 and then corresponding nehari manifold @xmath493 we also put @xmath494 .
we observe that the assumptions - are completely similar to @xmath205 of theorem [ compact 1 ] so that one may follow closely the first part of the proof of theorem [ compact 1 ] and prove the existence of a bounded palais - smale sequence @xmath469 for @xmath116 such that @xmath495 . up to subsequences
we may suppose that @xmath471 is weakly convergent in @xmath59 .
* suppose that @xmath496 with @xmath497 .
then in a standard way one can show that @xmath125 is a nontrivial solution of .
this means that @xmath113 .
it remains to show that @xmath125 is a least energy solution of . since @xmath471 is a bounded palais - smale sequence , by and the fatou lemma we have @xmath498 \ , dv_g \\ & \ge \frac 12 \int_m [ h(u)u-2h(u ) ] \
, dv_g=\phi(u)+\frac 12 \langle \phi'(u),u \rangle=\phi(u ) \ , .\end{aligned}\ ] ] this proves that @xmath125 is a least energy solution .
* case 2 . *
suppose that @xmath499 . by the continuous embedding @xmath500 we also have @xmath499 in @xmath501 with @xmath502 as in .
we claim that @xmath503 in @xmath501 . by
we deduce that for any @xmath478 there exists @xmath504 such that @xmath505 in @xmath501 then by with @xmath506 we obtain @xmath507 thus proving that @xmath508 strongly in @xmath59 . this is absurd since @xmath115 is bounded away from zero as one can show proceeding as in the proof of lemma [ l : nehari-1 ] ( iv ) . in the geometric assumptions of this proposition , by (
* lemma 2.26 ) there exists a uniformly locally finite covering of @xmath6 by balls of a fixed radius
. then we can apply ( * ? ? ?
* lemma 2.1.2 ) to deduce that there exist @xmath509 and @xmath510 such that @xmath511 where @xmath512 denotes the geodesic ball of radius @xmath513 centered at @xmath514 .
the sequence @xmath515 is necessarily unbounded since otherwise we would have @xmath516 .
hence it is not restrictive assuming , up to subsequences , that @xmath517 where @xmath9 is a point of @xmath6 fixed arbitrarily . since @xmath6 is weakly homogeneous there exists a sequence of isometries @xmath518 such that @xmath519 for some @xmath168 .
then , for any @xmath464 , we define @xmath520 so that @xmath521 is bounded in @xmath59 and @xmath522 .
this shows that @xmath523 in @xmath59 up to subsequences .
we first complete the proof of the theorem when @xmath174 is a constant function . in this way
@xmath521 is still a palais - smale sequence for @xmath116 such that @xmath524 and its weak limit @xmath525 is nontrivial .
we found a nontrivial critical point of @xmath116 which is also a least energy solution as one can prove by proceeding as in case 1 .
it remains to consider the general case when @xmath174 is not necessarily constant .
by we deduce that @xmath525 is a nontrivial critical point for the functional @xmath526 and in particular @xmath527 and @xmath528 . therefore by and fatou lemma @xmath529 \,dv_g \\ \notag & = \lim_{k\to + \infty } \frac 12 \int_m\left[h(\widehat{u}_k)\widehat{u}_k-2h(\widehat{u}_k)\right]\,dv_g\geq \frac 12 \int_m\left[h(\widehat u)\widehat u-2h(\widehat u)\right]\,dv_g\\ \notag & = \phi_{\infty}(\widehat u)\geq c_{\infty } \ , .\end{aligned}\ ] ] on the other hand , by and we have that @xmath146 and @xmath530 admits the following minimax characterization @xmath531 see lemma [ l : nehari-1 ] ( iii ) for more details .
but @xmath532 for any @xmath397 and hence by we obtain @xmath533 which combined with gives @xmath534 .
as shown above in the case @xmath174 constant , the functional @xmath526 admits a least energy solution @xmath535 . proceeding as in lemma [ positivity ] on
can show that @xmath536 does not change sign and moreover it is either strictly positive or strictly negative as a consequence of and the strong maximum principle for variational solutions , see ( * ? ? ?
* section 8.7 ) .
therefore , if @xmath174 is not constant , i.e. @xmath537 on set of positive measure , we have @xmath538 this implies @xmath539 a contradiction . @xmath334 we first prove that the scalar curvature @xmath541 is necessarily constant in @xmath6 .
let @xmath168 and @xmath167 be as in definition [ d : weakly - hom ] .
let @xmath542 be fixed arbitrarily and let @xmath543 be such that @xmath544 .
then there exists @xmath170 such that @xmath545 .
moreover @xmath546 and @xmath547
. indeed if we assume by contradiction that @xmath548 then we would have @xmath549 and this is absurd .
let @xmath550 and denote by @xmath551 the distance of @xmath21 from the pole . since @xmath541 is constant on @xmath552 for any @xmath145 and since @xmath553 is an isometry then @xmath541 is constant on @xmath554 for any @xmath145 .
we denote by @xmath555 this constant value .
let @xmath556 be fixed arbitrarily and consider the ball @xmath557 .
let @xmath558 , @xmath559 .
we observe that @xmath560 and hence @xmath561 ; but @xmath562 so that @xmath563 and in particular @xmath564 . since @xmath541 is constant on @xmath554 and on @xmath565 and @xmath566 then @xmath567 .
we have proved that @xmath567 for any @xmath558 and for any @xmath556 .
moreover we also have @xmath567 for any @xmath568 and for any @xmath556 thanks to the rotational symmetry of @xmath1 with respect to @xmath9 .
in particular @xmath569 for any @xmath570 and hence @xmath571 for any @xmath572 .
but we recall that @xmath573 and hence one may find @xmath21 such that @xmath551 is arbitrarily large simply choosing @xmath574 large enough . in this way one proves that @xmath541 is a constant @xmath84 over all @xmath6 .
let @xmath577 be two solutions of .
define @xmath578 for @xmath579 so that @xmath580 then we define @xmath581 and we obtain @xmath582-\frac{\beta}2w \\ w(0)=0 \ , , \quad w'(0)=0 \ , .
\end{cases}\ ] ] multiplying by @xmath583 and integrating we obtain @xmath584t^n-\tfrac{\beta}2t^nw \right\}dt \ , .\end{aligned}\ ] ] since @xmath585 and since the function of two real variables @xmath586 is lipschitzian in a neighborhood of @xmath587 , we deduce that for any @xmath478 there exist @xmath344 and @xmath509 such that for any @xmath588 we have @xmath589 } |w(t)|+c \int_0^r |w'(t)|dt\,.\end{aligned}\ ] ] since @xmath590 we obtain latexmath:[\[\begin{aligned } up to shrinking @xmath358 if necessary we may assume that @xmath592 for any @xmath588 in order to obtain @xmath593 then by it follows @xmath594 this implies that the map @xmath595 is nonincreasing in @xmath596 and hence if we choose @xmath597 , for any @xmath588 we have @xmath598 this proves that @xmath599 in @xmath596 and hence , in turn , we also have @xmath600 ; since @xmath601 then @xmath298 in @xmath596 and by the definition of @xmath602 we finally obtain @xmath603 in @xmath596 . the fact that @xmath604 and @xmath605 coincide over all @xmath413 follows immediately from classical uniqueness for cauchy problems
we claim that @xmath606 , otherwise if @xmath607 , by uniqueness , we would have @xmath608 where we put @xmath609 , in contradiction with @xmath88 .
the two alternatives in proposition [ p : constant - k ] follow as well . | we consider least energy solutions to the nonlinear equation @xmath0 posed on a class of riemannian models @xmath1 of dimension @xmath2 which include the classical hyperbolic space @xmath3 as well as manifolds with unbounded sectional geometry .
partial symmetry and existence of least energy solutions is proved for quite general nonlinearities @xmath4 , where @xmath5 denotes the geodesic distance from the pole of @xmath6 . |
the main new concept in this paper is a quantification of the classically differential notion of transversality .
this is achieved by extending persistence from filtrations of homology groups to zigzag modules of well groups .
[ [ motivation . ] ] motivation .
+ + + + + + + + + + + in hind - sight , we place the starting point for the work described in this paper at the difference between qualitative and quantitative statements and their relevance in the sciences ; see eg . the discussion in thom s book ( * ? ? ?
* chapters 1.3 and 13.8 ) .
it appears the conscious mind thinks in qualitative terms , delegating the quantitative details to the unconscious , if possible . in the sciences ,
quantitative statements are a requirement for testing a hypothesis .
without such a test , the hypothesis is not falsifiable and , by popular philosophical interpretation , not scientific @xcite .
the particular field discussed in @xcite is the mathematical study of singularities of smooth mappings , which is dominated by qualitative statements .
we refer to the seminal papers by whitney @xcite and the book by arnold @xcite for introductions .
a unifying concept in this field is the transversality of an intersection between two spaces .
its origins go far back in history and appear among others in the work of poincar about a century ago .
it took a good development toward its present form under pontryagin , whitney , and thom ; see eg .
@xcite . in his review of zeeman s book @xcite ,
smale criticizes the unscientific aspects of the work promoted in the then popular area of catastrophe theory , thus significantly contributing to the discussion of qualitative versus quantitative statements and to the fate of that field . at the same time , smale points to positive aspects and stresses the importance of the concept of transversality in the study of singularities . in a nutshell ,
an intersection is transverse if it forms a non - zero angle and is therefore stable under infinitesimal perturbations ; see section [ sec2 ] for a formal definition .
[ [ results . ] ] results .
+ + + + + + + + we view our work as a measure theoretic extension of the essentially differential concept of transversality .
we extend by relaxing the requirements on the perturbations from smooth mappings between manifolds to continuous mappings between topological spaces . at the same time
, we are more tolerant to changes in the intersection . to rationalize this tolerance , we measure intersections using real numbers as opposed to @xmath0 and @xmath1 to indicate existence .
the measurements are made using the concept of persistent homology ; see @xcite for the original paper .
however , we have need for modifications and use the extension of persistence from filtrations to zigzag modules as proposed in @xcite .
an important property of persistence , as originally defined for filtrations , is the stability of its diagrams ; see @xcite for the original proof .
there is no comparably general result known for zigzag modules .
our main result is a step in this direction .
specifically , we view the following as the main contributions of this paper : 1 . the introduction of well groups that capture the tolerance of intersections to perturbations in a given space of allowable perturbations ; 2 .
the proof that the diagram defined by the well groups is stable ; 3 .
the application of these results to fixed points and periodic orbits of continuous mappings .
in addition , our results have ramifications in the study of the set of critical values , the apparent contour of a smooth mapping .
specifically , the stability of the diagrams mentioned above results in a stability result for the apparent contour of a smooth mapping from an orientable @xmath2-manifold to the plane @xcite .
the need for these stable diagrams was indeed what triggered the development described in this paper .
[ [ outline . ] ] outline .
+ + + + + + + + section [ sec2 ] provides the relevant background .
section [ sec3 ] explains how we measure robustness using well groups and zigzag modules .
section [ sec4 ] proves our main result , the stability of the diagrams defined by the modules .
section [ sec5 ] discusses applications .
section [ sec6 ] concludes the paper .
we need the algebraic concept of persistent homology to extend the differential notion of transversality as explained in the introduction . in this section , we give a formal definition of transversality , referring to @xcite for general background in differential topology .
we also introduce homology and persistent homology , referring to @xcite for general background in classic algebraic topology and to @xcite for a text in computational topology .
[ [ transversality . ] ] transversality .
+ + + + + + + + + + + + + + + let @xmath3 be manifolds , @xmath4 a smooth mapping , and @xmath5 a smoothly embedded submanifold of the range .
we assume the manifolds have finite dimension and no boundary , writing @xmath6 , @xmath7 , and @xmath8 .
given a point @xmath9 and a smooth curve @xmath10 with @xmath11 , we call @xmath12 the _ tangent vector _ of @xmath13 at @xmath14 . varying the curve while maintaining that it passes through @xmath14 , we get a set of tangent vectors called the _ tangent space _ of @xmath15 at @xmath14 , denoted as @xmath16 . composing the curves with the mapping , @xmath17 , we get a subset of all smooth curves passing through @xmath18 .
the _ derivative _ of @xmath19 at @xmath14 is @xmath20 defined by mapping the tangent vector of @xmath13 at @xmath14 to the tangent vector of @xmath21 at @xmath22 .
the derivative is a linear map and its image is a subspace of @xmath23 .
the dimensions of the tangent spaces are @xmath24 and @xmath25 , which implies that the dimension of the image of the derivative is @xmath26 .
we are interested in properties of @xmath19 that are stable under perturbations .
we call a property _ infinitesimally stable _ if for every smooth homotopy , @xmath27 \to { { \ifmmode{{\mathbb y}}\else{\mbox{\({\mathbb y}\)}}\fi}}$ ] with @xmath28 , there is a real number @xmath29 such that @xmath30 possesses the same property for all @xmath31 , where @xmath32 for all @xmath9
. an important example of such a property is the following .
the mapping @xmath19 is _ transverse _ to @xmath33 , denoted as @xmath34{$\top$}\makebox[0pt]{$\cap$}\;\;}{{\ifmmode{{\mathbb a}}\else{\mbox{\({\mathbb a}\)}}\fi}}$ ] , if for each @xmath9 with @xmath35 , the image of the derivative of @xmath19 at @xmath14 together with the tangent space of @xmath33 at @xmath36 spans the tangent space of @xmath37 at @xmath38 .
more formally , @xmath34{$\top$}\makebox[0pt]{$\cap$}\;\;}{{\ifmmode{{\mathbb a}}\else{\mbox{\({\mathbb a}\)}}\fi}}$ ] if @xmath39 . it is plausible but also true that transversality is an infinitesimally stable property .
[ [ product - spaces . ] ] product spaces .
+ + + + + + + + + + + + + + + it is convenient to recast transversality in terms of intersections of subspaces of @xmath40 , a manifold of dimension @xmath41 .
consider the graphs of @xmath19 and of its restriction to the preimage of @xmath33 , @xmath42 the intersection of interest is between @xmath43 and @xmath44 , two manifolds of dimensions @xmath45 and @xmath46 embedded in @xmath40 .
this intersection is the graph of @xmath47 , which is homeomorphic to the preimage of @xmath33 .
see figure [ fig : crossings ] for an example in which @xmath48 and @xmath49 . here
, @xmath50 and transversality requires that whenever the curve , @xmath43 , intersects the line , @xmath44 , it crosses at a non - zero angle .
this is the case in figure [ fig : crossings ] which implies that having a cardinality four preimage of @xmath38 is an infinitesimally stable property of @xmath19 . nevertheless , the left two intersection points are clearly more stable than the right two intersection points , but we will need some algebra to give precise meaning to this statement . [ [ homology . ] ] homology .
+ + + + + + + + + the algebraic language of homology is a means to define and count holes in a topological space .
it is a functor that maps a space to an abelian group and a continuous map between spaces to a homomorphism between the corresponding groups .
there is such a functor for each dimension , @xmath51 .
it is convenient to combine the homology groups of all dimensions into a single algebraic structure .
writing @xmath52 for the @xmath51-dimensional homology group of the topological space @xmath15 , we form a graded group by taking direct sums , @xmath53 to simplify language and notation , we will suppress dimensions and refer to @xmath54 as the _ homology group _ of @xmath15 .
its elements are formally written as polynomials , @xmath55 , where @xmath56 is a @xmath51-dimensional homology class and only finitely many of the classes are non - zero . as usual , adding two polynomials is done componentwise .
the groups @xmath52 depend on a choice of coefficient group .
the theory of persistence introduced below requires we use field coefficients .
an example is modulo two arithmetic in which the field is @xmath57 .
the @xmath51-dimensional group is then a vector space , @xmath58 , and its rank , the dimension of the vector space , is the _ @xmath51-th betti number _ , @xmath59 .
similarly , @xmath54 is a vector space of dimension @xmath60 .
we say @xmath15 and @xmath37 _ have the same homology _ if there is an isomorphism between @xmath54 and @xmath61 whose restrictions to the components are isomorphisms .
equivalently , @xmath62 for all non - negative integers @xmath51 .
[ [ persistent - homology . ] ] persistent homology .
+ + + + + + + + + + + + + + + + + + + + now suppose we have a finite sequence of nested spaces , @xmath63 .
writing @xmath64 for the homology group of the @xmath65-th space , we get a sequence of vector spaces connected from left to right by homomorphisms induced by inclusion : @xmath66 we call this sequence a _
filtration_. to study the evolution of the homology classes as we progress from left to right in the filtration , we let @xmath67 be the composition of the maps between @xmath68 and @xmath69 , for @xmath70 .
we say a class @xmath71 is _ born _ at @xmath68 if it does not belong to the image of @xmath72 .
furthermore , this class @xmath73 _ dies entering _ @xmath69 if @xmath74 does not belong to the image of @xmath75 but @xmath76 does belong to the image of @xmath77 .
we call the images of the maps @xmath67 the _ persistent homology groups _ of the filtration and record the evolution of the homology classes in the _ persistence diagram _ of the filtration , denoted as @xmath78 .
this is a multiset of points in the extended plane , @xmath79 .
marking an increase in rank on the horizontal , birth axis and a drop in rank on the vertical , death axis , each point represents the birth and the death of a generator and records where these events happen ; see figure [ fig : diagram - h ] . for technical reasons that will become clear shortly ,
we add infinitely many copies of each point on the diagonal to the diagram . given an index , @xmath65
, we can read off the rank of @xmath80 by counting the points in the half - open upper - left quadrant , @xmath81 \times ( i , \infty]$ ] , anchored at the point @xmath82 on the diagonal .
more generally , the rank of the image of @xmath67 equals the number of points in the upper - left quadrant anchored at @xmath83 .
[ [ stability . ] ] stability .
+ + + + + + + + + + consider now the case in which the spaces in the sequence are sublevel sets of a real valued function @xmath84 , that is , there are values @xmath85 such that @xmath86 $ ] for each @xmath65
. a _ homological critical value _ of @xmath87 is a value @xmath88 such that for every sufficiently small @xmath89 , the homomorphism from @xmath90)$ ] to @xmath91)$ ] induced by inclusion is not an isomorphism .
we suppose @xmath87 is _ tame _ by which we mean that each sublevel set has finite rank homology and there are only finitely many homological critical values , denoted as @xmath92 .
we can represent the evolution of the homology classes by the finite filtration consisting of the groups @xmath93)$ ] , for @xmath94 , and by the persistence diagram of that filtration , @xmath95 .
letting @xmath96 be another tame function , we get another filtraton , @xmath97 , and another persistence diagram , @xmath98 . the _ bottleneck distance _ between the two is the infimum , over all bijections , @xmath99 , of the longest length edge in the matching , @xmath100 an important result is the stability of the persistence diagram under perturbations of the function .
let @xmath87 and @xmath101 be tame , real - valued functions on @xmath15 .
then the bottleneck distance between their persistence diagrams is bounded from above by @xmath102 . here , @xmath103 , as usual .
the original form of this result is slightly stronger as it restricts itself to dimension preserving bijections .
the theorem implies that the bottleneck distance between the diagrams defined by @xmath87 and @xmath101 goes to zero as the difference between the two functions approaches zero .
the main new concept in this section is the well group defined by a mapping @xmath104 , a subspace @xmath5 , and a parameter @xmath88 .
it encodes the persistent homology of the preimage of the subspace . [
[ admissible - mappings . ] ] admissible mappings . + + + + + + + + + + + + + + + + + + + + in this paper , we limit the class of mappings to those with manageable properties . while our goal is a statement of our results in a context that is sufficiently broad to support interesting applications , we are aware of the technical burden that comes with generality .
we hope that the following class of mappings gives a happy median between the conflicting goals of generality and transparency .
let @xmath15 and @xmath37 be topological spaces and @xmath33 a subspace of @xmath37 .
a continuous mapping @xmath4 is _ admissible _ if @xmath105 has a finite rank homology group .
requiring that the preimage of @xmath33 has finite rank homology is strictly weaker than demanding tameness of the well function defined next .
[ [ perturbations - and - well - groups . ] ] perturbations and well groups .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we are interested in how we explore the neighborhood of @xmath106 as we perturb @xmath19 . for this purpose
, we introduce a subspace @xmath107 of @xmath108 , the space of continuous mappings from @xmath15 to @xmath37 .
for example , @xmath109 may be the space of continuous mappings @xmath110 homotopic to @xmath19 ; that is , there exists a continuous mapping @xmath111 \to { { \ifmmode{{\mathbb y}}\else{\mbox{\({\mathbb y}\)}}\fi}}$ ] with @xmath112 and @xmath113 for all @xmath9 .
we assume a metric on @xmath109 , writing @xmath114 for the distance between two mappings @xmath115 .
for example , we could construct one by assuming a metric on @xmath37 , lifting it to define the distance between mappings in @xmath108 , and taking @xmath114 to be the infimum length of all curves of mappings that connect @xmath19 and @xmath110 within @xmath109 .
we call @xmath110 an
_ @xmath88-perturbation _ of @xmath19 if @xmath116 .
we use these definitions to introduce the _ well function _
@xmath117 of @xmath19 and @xmath33 by setting @xmath118 to the infimum value of @xmath88 for which there is an @xmath88-perturbation @xmath119 such that @xmath120 .
the _ level set _ of @xmath121 at a value @xmath88 is the preimage of that value : @xmath122 .
the _ sublevel set _ for the same value @xmath88 is the union of the level sets at values at most @xmath88 : @xmath123 $ ] .
note that @xmath124 is contained in @xmath125 for every @xmath88-perturbation @xmath119 .
this inclusion induces a homomorphism between the corresponding homology groups : @xmath126 where we simplify notation by writing @xmath127 for @xmath128)$ ] .
the image of this map , denoted by @xmath129 , is a subgroup of @xmath127 .
the intersection of subgroups is again a subgroup .
given a metric space of perturbations , @xmath107 , the _ well group _ of @xmath125 is the subgroup @xmath130 obtained by intersecting the images , @xmath129 , over all @xmath88-perturbations @xmath110 of @xmath19 in @xmath109 .
a different space of perturbations gives different well groups and therefore a different interpretation of their meaning .
[ [ example . ] ] example .
+ + + + + + + + to illustrate the definitions , let us consider again the example in figure [ fig : crossings ] of the mapping from the real line to itself .
the preimage of @xmath131 is a set of four points separated by three critical points of @xmath19 . from left to right , the values of @xmath19 at these critical points are @xmath132 , @xmath133 , @xmath134 .
correspondingly , the well function , @xmath135 , has three homological critical values , namely @xmath136 . setting @xmath107 to the set of all continuous mappings @xmath137 and measuring distance by @xmath138
, we have a well group @xmath139 for each radius @xmath140 .
table [ tbl : ranks ] shows the ranks of @xmath127 and @xmath139 for values of @xmath88 in the interior of the four intervals delimited by the homological critical values .
.the ranks of the homology and well groups defined for the mapping @xmath19 and the submanifold @xmath131 in figure . [ cols="^,^,^,^,^",options="header " , ] starting with @xmath141 , we have four points , each forming a component represented by a class in the homology group and in the well group of the sublevel set of @xmath142 .
therefore , both groups are the same and have rank four , see the first column in table [ tbl : ranks ] .
growing @xmath88 turns the points into intervals but leaves the groups the same until @xmath88 passes @xmath143 , the smallest of the three homological critical values .
the two right intervals merge into one , so the rank of the homology group drops to three .
we can find an @xmath88-perturbation , @xmath144 , whose level set at @xmath38 consists of the left two points of @xmath145 but the right two points have disappeared . indeed , the level set of every @xmath88-perturbation , for @xmath146 , has a non - empty intersection with the first two but can have empty intersection with the merged interval on the right .
hence , the left two intervals have a representation in the well group , the merged interval does not , and the rank of the well group is two ; see the second column in table [ tbl : ranks ] .
the next change happens when @xmath88 passes @xmath147 .
the middle interval merges with the merged interval on the right .
the rank of the homology group drops to two , while the rank of the well group remains unchanged at two ; see the third column in table [ tbl : ranks ] .
finally , when @xmath88 passes @xmath148 , the remaining two intervals merge into one , so the rank of the homology group drops to one .
we can find an @xmath88-perturbation , @xmath149 , whose level set at @xmath38 is empty , so the rank of the well group drops to zero ; see the last column in table [ tbl : ranks ] .
[ [ terminal - critical - values . ] ] terminal critical values .
+ + + + + + + + + + + + + + + + + + + + + + + + + recall that we assume the mapping @xmath4 is admissible .
the initial homology group , @xmath150 , has therefore finite rank , and because @xmath151 , the initial well group has finite rank . for convenience ,
we permit negative parameter values by stipulating @xmath152 and @xmath153 whenever @xmath154 .
imagine we grow the sublevel set by gradually increasing @xmath88 to infinity .
since the admissibility of @xmath19 does not imply the tameness of the well function , this leaves open the possibility that @xmath155 has infinitely many homological critical values .
we call a radius , @xmath88 , a _ terminal critical value _ of @xmath155 if for every sufficiently small @xmath29 the homomorphism from @xmath156 to @xmath157 applied to @xmath158 does not give @xmath159 .
in contrast to the homological critical values , there can only be a finite number of terminal critical values . to see this
, we note that the set of perturbations that define the well groups grows with increasing @xmath88 .
it follows that the well groups can not increase in rank . to state
this relationship between well groups more formally , we write @xmath160 for the homomorphism induced by inclusion . for each choice of radii @xmath161 , the image of the well group at @xmath88 contains the well group at @xmath162 , that is , @xmath163 .
the only way the well group can change is by lowering its rank .
since we start with a finite rank , there can only be finitely many terminal critical values , which we denote as @xmath164 . to this sequence , we add @xmath165 on the left and @xmath166 on the right .
we choose an interleaved sequence @xmath167 and index the homology and the well groups accordingly , writing @xmath168 and @xmath169 , for all @xmath65 . to these sequences ,
we add @xmath170 on the left and @xmath171 on the right .
furthermore , we write @xmath172 instead of @xmath173 for all feasible choices of @xmath70 . [
[ well - module . ] ] well module .
+ + + + + + + + + + + + in contrast to the homology groups , the well groups of the sublevel sets do not form a filtration .
instead , they form a special kind of zigzag module . by definition of terminal critical values , the rank of @xmath174 exceeds the rank of @xmath175 .
the rank of the image , @xmath176 , is somewhere between these two ranks .
we call a difference between @xmath174 and its image a _ conventional death _ , in which a class maps to zero , and a difference between the image and @xmath175 an _ unconventional death _ , in which the image of a class lies outside the next well group .
we capture both cases by inserting a new group between the contiguous well groups ; see figure [ fig : well ] . to this end
, we consider the restriction of @xmath177 to @xmath174 and in particular its kernel , @xmath178 , which we refer to as the _ vanishing subgroup _ of @xmath174 . using this subgroup , we construct @xmath179 . the forward map , @xmath180 , is defined by mapping a class @xmath181 to @xmath182 .
it is clearly surjective .
the backward map , @xmath183 , is defined by mapping a class @xmath184 to @xmath182 , where @xmath181 belongs to @xmath185 .
this map is clearly injective . instead of a filtration in which all maps go from left to
right , we get a sequence in which the maps alternate between going forward and backward . as indicated below , every other group in the sequence is a subgroup of the corresponding homology group : @xmath186
we call this sequence the _ well module _ of @xmath155 , denoted as @xmath187 .
we remark that @xmath187 is a special case of a zigzag module as introduced in @xcite .
it is special because all forward maps are surjective and all backward maps are injective .
equivalently , there are no births other than at the start , when we go from @xmath188 to @xmath189 .
[ [ left - filtration . ] ] left filtration .
+ + + + + + + + + + + + + + + + perhaps surprisingly , the evolution of the homology classes can still be fully described by pairing births with deaths , just like for a filtration . to shed light on this construction , we follow @xcite and turn a zigzag module into a filtration . in our case
, all births happen at @xmath189 , so this transformation is easier than for general zigzag modules .
write @xmath190 for the restriction of @xmath191 to the initial well group . by the shrinking wellness lemma ,
the image of this map contains the @xmath65-th well group , that is , @xmath192 .
we consider the preimages of the well groups in @xmath189 together with the preimages of their vanishing subgroups , @xmath193 and @xmath194 ; see figure [ fig : left - filtration ] .
we note that @xmath195 and @xmath196 . in words ,
the first quotient represents the homology classes that die a conventional death , going from @xmath174 to @xmath175 , and the second quotient represents the homology classes that die an unconventional death in the same transition .
as illustrated in figure [ fig : left - filtration ] , the preimages form a nested sequence of subgroups of @xmath189 .
together with the inclusion maps , this gives the _ left filtration _ of the zigzag module , @xmath197 we can recover the well groups with @xmath198 .
recall that @xmath199 , which implies @xmath200 .
it follows that the middle two groups in the left filtration , @xmath201 and @xmath202 , are indeed equal .
[ [ compatible - bases . ] ] compatible bases .
+ + + + + + + + + + + + + + + + + a useful property of the left filtration is the existence of compatible bases of all its groups . by this we mean a basis of @xmath189 that contains a basis for each @xmath203 and each @xmath204 .
specifically , we rewrite @xmath189 as a direct sum of kernels of forward maps and cokernels of backward maps : @xmath205 reading this decomposition from left to right , we encounter the @xmath203 and the @xmath204 in the sequence they occur in the left filtration . choosing a basis for each kernel and each cokernel , we thus get compatible bases for all groups in the left filtration .
we call this the _ left filtration basis _ of @xmath189 .
it is unique up to choosing bases for the kernels and cokernels .
consider now a homology class @xmath73 in @xmath189 and its representation as a sum of basis vectors .
we write @xmath206 for the projection of @xmath73 to the preimage of the kernel of the @xmath65-th forward map , which is obtained by removing all vectors that do not belong to the basis of @xmath207 .
similarly , we write @xmath208 for the projection of @xmath73 to the preimage of @xmath209 . letting @xmath210
be the minimum index such that @xmath211 for all @xmath212 , we say that @xmath73 _ falls ill _ at @xmath213 .
[ [ well - diagrams . ] ] well diagrams .
+ + + + + + + + + + + + + + constructing the birth - death pairs that describe the well module is now easy .
all classes are born at @xmath189 , however , to distinguish the changes in the well group from those in the homology group , we say all the classes _ get well _ at @xmath189 .
they fall ill later , and once they fall ill , they do not get well any more .
the drop in rank from @xmath214 to @xmath174 is @xmath215 .
we thus have @xmath216 copies of the point @xmath217 is the diagram .
there is no information in the first coordinates , which are all zero .
we thus define the _ well diagram _ as the multiset of points @xmath218 with multiplicities @xmath216 , denoting it as @xmath219 . for technical reasons that will become obvious in the next section , we add infinitely many copies of @xmath0 to this diagram .
hence , each point in @xmath219 is either @xmath0 , a positive real number , or @xmath220 , and the diagram itself is a multiset of points on the extended line , @xmath221 .
it has infinitely many points at @xmath0 and a finite number of non - zero points . as suggested by the heading of this section , we think of each point in the diagram as a measure for how resilient a homology class of @xmath105 is against perturbations in @xmath109 .
at each well group @xmath174 , an entire set of homology classes falls ill , and we call @xmath218 the _ robustness _ of each class @xmath73 in this set , denoting it as @xmath222 .
we are interested in relating the distance between mappings to the distance between their well diagrams , both defined using a common perturbation space @xmath109 . after quantifying these distances , we connect parallel well modules to form new modules , and we finally prove that the well diagram is stable . [ [ distance - between - functions . ] ] distance between functions .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath223 be two admissible mappings between topological spaces , @xmath5 a subspace , and @xmath109 a metric space of perturbations that contains both @xmath19 and @xmath224 .
using @xmath33 , we get the two well functions : @xmath225 .
the distance between them is the supremum difference between corresponding values : @xmath226 the distance between the two well functions is related to the distance , @xmath227 , between the two mappings @xmath19 and @xmath224 in @xmath109 .
specifically , the distance between the well functions can not exceed the distance between the mappings . assuming the above notation
, we have @xmath228 . proof .
fix a point @xmath9 and let @xmath229 . by construction of @xmath121
, there exists an @xmath88-perturbation @xmath110 of @xmath19 in @xmath109 with @xmath230 . by the triangle inequality , @xmath231 .
hence , @xmath232 .
thus , @xmath233 . [
[ distance - between - diagrams . ] ] distance between diagrams .
+ + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath234 be the homology group and @xmath235 the well group of @xmath236 $ ] .
as for @xmath19 , we insert quotients between contiguous well groups and connect them with forward and backward maps to form a well module , denoted as @xmath237 .
the corresponding well diagram , @xmath238 , is again a multiset of points in @xmath239 , consisting of infinitely many copies of @xmath0 and finitely many non - zero points .
recall that the bottleneck distance between the diagrams of @xmath19 and @xmath224 is the length of the longest edge in the minimizing matching .
because our diagrams are one - dimensional , the bottleneck distance is easy to compute . to describe the algorithm
, we order the positive points in both diagrams , getting @xmath240 where we add zeros to make sure we have two sequences of the same length . the _ inversion - free matching _
pairs @xmath218 with @xmath241 for all @xmath65 .
we prove that this matching gives the bottleneck distance . assuming the above notation
, the bottleneck distance between @xmath219 and @xmath238 is equal to @xmath242 .
proof . for a given matching , we consider the vector of absolute differences , which we sort largest first
. comparing two such vectors lexicographically , we now prove that the inversion - free matching gives the minimum vector .
this implies the claimed equality , @xmath243 to prove minimality , we consider a matching that has at least one inversion , that is , pairs @xmath244 and @xmath245 with @xmath246 and @xmath247 .
if @xmath248 or @xmath249 then switching to the pairs @xmath250 and @xmath251 preserves the sorted vector of absolute differences .
otherwise , the new vector is lexicographically smaller than the old vector .
indeed , the minimum of the four points is @xmath218 or @xmath252 and the maximum is @xmath253 or @xmath254 . if the minimum and the maximum are from opposite diagrams then they delimit the largest of the four absolute differences , and this largest difference belongs to the old vector .
otherwise , both absolute differences shrink when we switch the pairs . repeatedly removing inversions
as described eventually leads to the inversion - free matching , which shows that it minimizes the vector and its largest entry is the bottleneck distance . [
[ bridges . ] ] bridges .
+ + + + + + + + the main tool in the proof of stability is the concept of a short bridge between parallel filtrations .
the length of these bridges relates to the distance between the functions defining the filtrations .
let @xmath255 . by the distance lemma
, we have @xmath256 , which implies that the sublevel set of @xmath257 for radius @xmath88 is contained in the sublevel set of @xmath155 for radius @xmath258 .
hence , there is a homomorphism @xmath259 , which we call the _ bridge _ from @xmath260 to @xmath261 at radius @xmath88 .
we use the bridge to connect the initial segment of @xmath260 to the terminal segment of @xmath261 .
the endpoints of the bridge satisfy the property expressed in the shrinking wellness lemma .
let @xmath262 be the bridge at @xmath88 , where @xmath255 and @xmath263 .
then @xmath264 . proof .
let @xmath73 be a homology group in @xmath265 .
by definition of the well group , @xmath73 belongs to the image of @xmath266 in @xmath267 for every @xmath268-perturbation @xmath110 of @xmath19 in @xmath109 .
this includes all @xmath88-perturbations of @xmath224 .
it follows that the preimage of @xmath73 in @xmath234 belongs to the well group , that is , @xmath269 .
everything we said about bridges is of course symmetric in @xmath261 and @xmath260 . in other words , @xmath270 \subseteq g_{{\ifmmode{{\mathbb a}}\else{\mbox{\({\mathbb a}\)}}\fi}}^{-1 } [ 0,r+{{\ifmmode{\varepsilon}\else{\mbox{\(\varepsilon\)}}\fi}}]$ ] and there is a bridge from @xmath127 to @xmath271 for every @xmath140 . [
[ new - modules . ] ] new modules .
+ + + + + + + + + + + + we use the bridge lemma to construct new zigzag modules from the well modules of @xmath19 and @xmath224 .
specifically , we use @xmath272 to connect the initial segment of @xmath237 , from @xmath273 to @xmath274 , to the terminal segment of @xmath187 , from @xmath265 to @xmath275 . to complete the module
, we insert @xmath276 between @xmath274 and @xmath265 . the forward map , from @xmath274 to @xmath277 , is surjective , and the backward map , from @xmath265 to @xmath277 is injective ; see figure [ fig : bridge ] .
the new zigzag module is thus of the same type as the well modules implying it has a left filtration basis that gives rise to a family of compatible bases for the groups in the left filtration . a particular construction starts with the filtrations @xmath278 and @xmath279 and adds @xmath280 . following the bridge from @xmath260 to @xmath261 at @xmath0
, we get a new filtration and a new zigzag module , denoting the latter as @xmath281 ; see figure [ fig : newmodules ] .
the decomposition of @xmath282 by the left filtration of @xmath281 is similar to the decomposition of @xmath283 by the left filtration of @xmath187 ; see figure [ fig : left - filtration ] .
letting @xmath65 be the index such that @xmath284 , we have @xmath285 .
the classes in @xmath286 and in @xmath287 die before we reach @xmath288 .
the remaining classes form @xmath289 .
correspondingly , there are homology classes in @xmath290 that die before we reach @xmath288 , namely the ones in the kernel of the forward map , from @xmath290 to @xmath291 , and in the preimage of the cokernel of the backward map , from @xmath292 to @xmath291 .
the remaining classes form @xmath293 .
the two quotient groups , @xmath292 and @xmath294 , are decomposed in parallel so that choosing a basis for @xmath292 gives one for @xmath294
. this will be useful shortly .
[ [ main - result . ] ] main result
. + + + + + + + + + + + + we are now ready to state and prove the stability of the well diagram .
let @xmath19 and @xmath224 be admissible mappings between topological spaces @xmath15 and @xmath37 , let @xmath33 be a subspace of @xmath37 , and let @xmath109 a metric space of perturbations containing both @xmath19 and @xmath224 .
then @xmath295 for the well modules @xmath187 and @xmath237 of @xmath19 and @xmath224 .
we construct a bijection from @xmath219 to @xmath238 such that the difference between matched points is at most @xmath255 .
specifically , we match each point @xmath296 in @xmath219 with a copy of @xmath0 in @xmath238 , and we use the parallel bases of @xmath292 and @xmath294 for the rest , where @xmath281 is the zigzag module obtained by adding the bridge from @xmath260 to @xmath261 at radius @xmath0 , as described above .
let @xmath73 belong to the left filtration basis of @xmath283 such that its image belongs to the basis of @xmath292 .
let @xmath88 be the value at which @xmath73 falls ill and note that @xmath297 .
let @xmath298 belong to the left filtration basis of @xmath299 such that the images of @xmath73 and @xmath298 in @xmath300 coincide .
we now construct yet another zigzag module , by adding a first bridge from @xmath301 to @xmath156 and a second bridge from @xmath157 back to @xmath302 , where @xmath29 is sufficiently small such that there is no death in the interval @xmath303 $ ] , except possibly at @xmath88 .
we denote the resulting module by @xmath304 ; see figure [ fig : newmodules ] .
we note that all maps between groups are induced by inclusions so that the diagram formed by the filtrations and the bridges between them commutes . by construction ,
the image of @xmath298 in @xmath156 is non - zero and belongs to @xmath158 . in contrast , the image of @xmath298 in @xmath157 is either zero or lies outside @xmath159 .
applying the bridge lemma going backward along the first bridge , we note that the image of @xmath305 in @xmath301 is non - zero and belongs to @xmath306 . applying the bridge lemma going forward along the second bridge
, we note that the image of @xmath298 in @xmath302 is either zero or lies outside @xmath307 .
since we can choose @xmath29 as small as we like , this implies that @xmath298 falls ill somewhere in the interval @xmath308 $ ] . in the matching , this radius is paired with @xmath88 , the radius at which @xmath73 falls ill in @xmath187 .
the absolute difference between the two radii is at most @xmath309 , as required .
in this section , we use the stability of the transversality measure to derive stability results for fixed points , periodic orbits , and apparent contours .
all three problems can be recast in terms of intersections between topological spaces and are therefore amenable to the tools developed in this paper . [
[ fixed - points . ] ] fixed points .
+ + + + + + + + + + + + + a _ fixed point _ of a continuous mapping from a topological space to itself is a point that is its own image . assuming this space is the @xmath45-dimensional euclidean space and @xmath310 is the mapping , we introduce a mapping @xmath311 defined by @xmath312 .
a fixed point of @xmath310 is a root of @xmath19 , that is , @xmath313 .
writing @xmath314 and @xmath315 , the origin of @xmath316 , we get the setting studied in this paper .
each fixed point @xmath14 of @xmath310 corresponds to a class in the @xmath0-dimensional homology group of @xmath317 . using the methods of this paper ,
we assign a non - negative robustness measure , @xmath318 , to @xmath14 .
it gives the magnitude of the perturbation necessary to remove this fixed point .
this does not mean that a perturbation of smaller magnitude has a fixed point at precisely the same location but rather that it has one or more fixed points in lieu of @xmath14 .
letting @xmath318 be the maximum robustness of all the fixed points of @xmath19 , every perturbation of magnitude less than @xmath318 has at least one fixed point .
this implication suffices to give a new proof of a classic topological result on fixed points ; see @xcite .
let @xmath319 be the closed unit ball in @xmath316 .
every continuous mapping @xmath320 has a fixed point .
extend @xmath310 to a mapping from @xmath316 to @xmath316 by defining @xmath321 equal to its value at @xmath322 whenever @xmath323 .
let @xmath311 be defined by @xmath312 and let @xmath324 be the identity , defined by @xmath325 .
we may assume that @xmath19 is admissible , else the homology group of @xmath317 has infinite rank and @xmath19 has infinitely many roots .
the other mapping , @xmath224 , is clearly admissible , with a single root at @xmath326 .
letting @xmath109 be the space of all continuous mappings from @xmath316 to @xmath316 , and measuring the distance between @xmath19 and @xmath224 by taking the supremum of the euclidean distance between corresponding image points , we get @xmath327 which is at most @xmath1 .
the well diagram of the identity consists of a single , non - zero point at plus infinity .
the stability theorem for well diagrams implies that the well diagram of @xmath19 also has a point at plus infinity .
but this implies that @xmath19 has a root and , equivalently , that @xmath310 has a fixed point .
the above reduction of fixed points to an intersection setting uses the difference between two points , an operation not available if the mapping @xmath328 is defined on a general metric space . in this case , we can use the correspondence between the fixed points of @xmath310 and the intersection points between the graph of @xmath310 and the diagonal in @xmath329 . to apply the results of this paper , we set @xmath330 , @xmath331 , and @xmath332 .
furthermore , we define the distance between two points @xmath333 and @xmath334 in @xmath329 equal to @xmath335 we restrict @xmath109 to those mappings @xmath336 that arise as graphs of a continuous mapping from @xmath337 to itself .
it is not difficult to see that this setting gives the same robustness values for the case @xmath338 discussed above .
[ [ periodic - orbits . ] ] periodic orbits .
+ + + + + + + + + + + + + + + + we generalize the above setting by allowing for fixed points of iterations of the mapping .
letting @xmath337 be a metric space and @xmath339 a mapping , we write @xmath340 for the @xmath210-fold composition of @xmath19 with itself .
a sequence @xmath341 is an _
order-@xmath210 periodic orbit _ of @xmath19 if @xmath342 .
it is straightforward to see the following relationship between @xmath19 and its @xmath210-fold composite .
a point @xmath343 is a fixed point of @xmath344 iff @xmath345 is an order-@xmath210 periodic orbit of @xmath19 .
we can therefore use the methods of this paper to measure the robustness of @xmath14 , that is , to determine how much @xmath344 needs to be perturbed to remove the fixed point .
however , it is more interesting to measure how much @xmath19 needs to be perturbed to remove the periodic orbit .
this is different because not every mapping can be written as the @xmath210-fold composite of another mapping .
this motivates us to introduce @xmath109 as the space of perturbations of @xmath344 that are @xmath210-fold composites of perturbations @xmath110 of @xmath19 . using this space @xmath109 , we intersect the images of the homomorphisms induced by the @xmath346 . with this setup
, we construct the well diagram of @xmath344 and interpret the resulting values as the robustness of the order-@xmath210 periodic orbits of @xmath19 .
[ [ apparent - contours . ] ] apparent contours .
+ + + + + + + + + + + + + + + + + + as mentioned in the introduction , @xcite reduces the stability of the contour of a mapping to the stability of well diagrams , the main result of this paper .
we briefly review the reduction .
let @xmath337 be a compact , orientable @xmath2-manifold and @xmath347 a smooth mapping .
the derivative of @xmath19 at a point @xmath14 is a linear map from the tangent space to @xmath348 .
the point @xmath14 is _ critical _ if the derivative at @xmath14 is not surjective , and the _ apparent contour _ of @xmath19 is the set of images of critical points . beyond smoothness of @xmath19 , we assume that the well functions it defines are admissible . specifically , for each @xmath349 , the function @xmath350 is defined by mapping every point @xmath14 to @xmath351 and we assume that @xmath352 consists of a finite number of points .
to study the apparent contour , we consider the entire @xmath2-parameter family of well functions .
fixing a value @xmath349 , the sublevel sets of @xmath142 form a filtration of homology groups and a zigzag module of well groups .
each point in the preimage of @xmath38 falls ill at a particular radius interpreted as the robustness of that point .
the main result of this paper implies that this measure is stable , that is , @xmath353 , where @xmath187 and @xmath237 are the well modules defined by the mappings @xmath354 and the value @xmath349 .
as shown in @xcite , this implies that the apparent contours of @xmath19 and of @xmath224 are close .
the sense in which they are close is interesting in its own right and we refer to that paper for details .
the main contribution of this paper is the definition of a robustness measure for the homology of the intersection between topological spaces , and a proof that this measure is stable .
while robustness and persistence are related , there are also differences between these notions . for example
, robustness adapts to a given metric space of perturbations , and this extra degree of freedom is sometimes essential , such as for a meaningful analysis of periodic orbits . the results in this paper
raise a number of questions and invite extensions in several directions .
* fixed points of mappings play an important role in game theory @xcite .
can the results of this paper be used to gain insights into the nature of fixed points as they arise in different games ?
what are contexts in which the robustness of a fixed point is relevant to the understanding of the dynamics of a game ?
* the three applications sketched in section [ sec5 ] barely scratch the surface of the possible . an interesting direction for further research
are mappings from lower to higher dimensions .
for example , the boundary of a computer - aided design model is the image of a mapping from a @xmath2-manifold to @xmath355 .
can our results be used to detect and remove accidental self - intersections , a problem of significant economic importance @xcite . * except for a few special settings , we have no algorithms for computing well diagrams .
the main obstacle is the possibly infinite set of perturbations that appears in the definition of well groups .
however , since the groups that arise for admissible mappings are finite , only a finite number of perturbations are relevant .
can we approach the algorithmic question from this direction ?
the stability of the apparent contour of an orientable @xmath2-manifold . in _ topological methods in data
analysis and visualization : theory , algorithms , and applications _ ,
eds . v. pascucci and j. tierny , springer - verlag , to appear . | by definition , transverse intersections are stable under infinitesimal perturbations . using persistent homology , we extend this notion to a measure . given a space of perturbations ,
we assign to each homology class of the intersection its robustness , the magnitude of a perturbation in this space necessary to kill it , and prove that robustness is stable . among the applications of this result
is a stable notion of robustness for fixed points of continuous mappings and a statement of stability for contours of smooth mappings . |
the most fundamental difference between classical and quantum information is that while classical information can be copied perfectly , quantum information can not
. in particular , it follows from the no - cloning theorem @xcite that one can not create a perfect duplicate of an arbitrary qubit .
nevertheless , buzek and hillery and other authors @xcite have shown that imperfect copies can be made by a universal quantum cloning machine ( uqcm ) , the outputs of which are identical .
the price which must be paid is that there is a difference between the original input and the copies , because of residual entanglement between the machine and copies .
however , not only similarity is lost during the cloning process .
perhaps even more import than the no - cloning feature of quantum mechanics is entanglement , first noted by einstein - podolsky - rosen ( epr)and schrdinger @xcite . for decades , quantum entanglement has been the focus of much work in the foundations of quantum mechanics . in particular
, it is associated with quantum nonseparability , the violation of bell s inequalities , and the so - called epr paradox . beyond this fundamental aspect , creating and manipulating of entangled states are essential for quantum information applications . among these applications
are quantum teleportation @xcite , quantum dense coding @xcite , quantum error correction @xcite , and quantum computational speedups @xcite , quantum cryptography @xcite , and quantum positioning and clock synchronization @xcite .
hence , quantum entanglement has been viewed as an essential resource for quantum information processing .
all of these applications depend upon the strength of quantum entanglement .
if the cloning process applied to entangled subsystems is to be anything more than a basic curiosity , and is to find a practical application in the field of quantum information theory , it should be possible to obtain not only maximally accurate copies of the original state , but also copies which preserve entangling characteristic of the original state when the input state is an entangled state .
masiak and knight @xcite have shown that copies of entangled pair of qubits can be generated by using the uqcm . in this paper
, we will investigate the degree to which entanglement survives when a three qubit entangled state is copied by using local and non - local processes , respectively .
we will show how these copying processes degrade quantum entanglement due to a residual correlation between the copied output and the copying machine .
we will also show that entanglement is rapidly destroyed by the copying process .
we consider quantum copying of a non - local state of a three - particle system consisting of a three qubits and assume that the three qubits are initially prepared in a entangled pure state expressed by @xmath0 there are two approaches to realize quantum copying of the above entangled state .
the first one is that each of the three original qubits is copied separately by using three different local copying machines .
the second method is that the entangled state of the three qubits is treated as a state in a large hilbert space and copied as a whole .
the former is called as local cloning while the latter is regarded as non - local cloning .
we will measure quantum entanglement in a three - particle system by making use of the entanglement tensor approach which is first proposed in ref . @xcite and further developed for a three - particle system by present authors @xcite .
entanglement measures of a three - qubit system involve both an inter - three - qubit entanglement measure denoted by @xmath1 and an inter - two - qubit entanglement measure labelled by @xmath2 .
based on entanglement tensors @xmath3 and @xmath4 introduced in ref .
@xcite , entanglement measures @xmath1 and @xmath2 can be defined as follows @xmath5 here the entanglement tensors on the rhs of eqs .
( [ 2 ] ) and ( [ 3 ] ) are defined by @xmath6 where @xmath7 , @xmath8 , and @xmath9 are the correlation tensors and coherence vectors defined in ref.@xcite , respectively .
these measures are invariant under local unitary transformations of the subsystems and their values can change between @xmath10 ( unentangled states ) and @xmath11 ( maximum entangled states ) .
@xmath1 quantifies the three - qubit entanglement .
the larger @xmath1 is , the stronger the three - qubit entanglement is .
and @xmath12 quantifies the entanglement between any two qubits @xmath13 in the three - qubit system .
the larger @xmath12 is , the stronger the entanglement between two qubit @xmath13 is . making use of eqs .
( 2 - 8 ) in ref.@xcite we can obtain the nonzero components of the coherence vector and correlation tensor @xmath14 from eqs .
( [ 4 ] ) and ( [ 5 ] ) one can find the nonzero components of the entanglement tensor @xmath15 while all other components of the entanglement tensor vanish .
then from eqs .
( [ 2 ] ) and ( [ 3])we find the values of entanglement measures @xmath16 and @xmath17 as follows @xmath18 which indicate that the three - qubit ghz state @xmath19 is the maximally entangled three - qubit state with the maximal value of the inter - three - qubit entanglement @xmath20 .
it is well known that there are two approaches to realize quantum copying of the entangled state given by eq .
( [ 1 ] ) .
the first one is that each of the three original qubits is copied separately by using three different local copying machines .
the second method is that the entangled state of the three qubits is treated as a state in a large hilbert space and copied as a whole .
the former is called as local cloning while the latter is regarded as non - local cloning . in what follows
we will investigate entanglement variation and copying fidelity in the two types of quantum cloning processes .
* a. local cloning * a scheme which will achieve local cloning is described by the following process @xcite : three distant parties share an entangled three qubit state @xmath21 .
each of them performs some local transformations on their own qubit using distant quantum copying machines .
we assume that the three additional qubits , employed in the copying process , are initially unentangled .
three copiers separately make copies of the qubits by the following local unitary transformations @xcite @xmath22 where @xmath23 .
the system labelled by @xmath24 is the original ( input ) qubit , while the other system @xmath25 represents the target qubit onto which the information is copied .
the states of the copying machine are labelled by @xmath26 .
the state space of the copying machine is two dimensional and we assume that it is always in the same state @xmath27 initially .
the result of the cloning process is an output state which is no longer a pure state , but is a mixed state described by the following density matrix @xmath28 we now calculate entanglement for the output state of local quantum copying described by above density operator . making use of eqs .
( 2 - 8 ) in ref.@xcite we can obtain the nonzero coherence vectors and correlation tensors @xmath29 from eqs .
( [ 4 ] ) and ( [ 5 ] ) one can find the following entanglement tensors @xmath30 while all other components of the entanglement tensor vanish .
then from eqs .
( [ 2 ] ) and ( [ 3])we find the values of @xmath16 and @xmath17 entanglements as follows @xmath31 which indicate that both the inter - three - qubit entanglement and inter - two - qubit entanglement in a three qubit system characterized by @xmath1 and @xmath17 , respectively , can be broadcasted via local quantum cloning , since the amount of @xmath16 and @xmath17 entanglements are nonzero , i.e. , @xmath32 , @xmath33 except @xmath34 or @xmath35 .
in particular , when @xmath36 , we have @xmath37 and @xmath38 which recover the result in ref .
@xcite for quantum copying of the ghz state . comparing remaining @xmath1 and @xmath17 entanglements after local cloning given by eqs .
( [ 25 ] ) and ( [ 26 ] ) with the original @xmath1 and @xmath17 entanglements before local cloning given by eqs .
( [ e14 ] ) and ( [ e15 ] ) , we can find that both @xmath1 and @xmath17 entanglements are reduced in the local cloning process .
the remaining @xmath1 entanglement after local cloning is only 8.76 percent of the original @xmath1 entanglement , and the remaining @xmath2 entanglement after local cloning is 19.8 percent of the original @xmath2 entanglement .
therefore , the @xmath2 entanglement is preserved better than the @xmath1 entanglement in the local cloning process .
entanglement of the three - qubit pure state @xmath39 ( dashed curve ) and remaining @xmath1 entanglement after the first step of local ( dotted curve ) and non - local ( solid curve ) cloning.,width=364,height=249 ] * b. non - local cloning * the entangled state given in eq .
( [ 3 ] ) @xmath39 can also be cloned non - locally @xcite . in this case
the entangled state of the three - qubits is treated as a state in a larger hilbert space and cloned as a whole .
the non - local quantum copying machine @xcite is an @xmath40 dimensional quantum system , and we shall let @xmath41 ( @xmath42 ) be a set of orthonormal basis of the copying machine hilbert space .
this copier is initially prepared in a particular state @xmath43 .
the action of the cloning transformation can be specified by a unitary transformation acting on the basis vectors of the tensor product space of the original quantum system @xmath44 , the copier , and an additional @xmath40-dimensional system which is to become the copy ( which is initially prepared in an arbitrary state @xmath45 ) .
the corresponding transformation @xmath46 is given by @xmath47 where @xmath48 , @xmath49 , and @xmath50 .
the final state of the three - qubit copies at the output of the cloning machine is given by the following density operator @xmath51 now we check whether entanglement is broadcasted .
through lengthy calculations , we obtain the nonzero components of the coherent vector and correlation tensor @xmath52 which leads to the following non - vanishing components of entanglement tensor @xmath53 substituting eqs.([22]-[25 ] ) into eqs .
( [ 2 ] ) and ( [ 3 ] ) , we find the values @xmath16 and @xmath17 to be @xmath54 from which we see that both @xmath1 and @xmath2 entanglements can be broadcasted via the non - local process . obviously , when @xmath36 , we have @xmath55 and @xmath56 which recover the result in ref .
@xcite for quantum copying of the ghz state .
entanglement of the three - qubit pure state @xmath39 ( dashed curve ) and remaining @xmath2 entanglement after the first step of local ( dotted curve ) and non - local ( solid curve ) cloning.,width=364,height=249 ] we now numerically discuss entanglement variation in the local and non - local copying processes with respect to the values of @xmath57 . in fig.1 and fig.2
we plot the @xmath1 entanglement and @xmath2 entanglement in the three - qubit input state @xmath39 and the three - qubit output states obtained as the result of local and non - local copying of the state @xmath39 , respectively . from fig.1 and fig.2
we can see that both local and non - local copy can broadcast entanglement in the three - qubit system .
and we find that non - local cloning is much more efficient than the local copying process for broadcasting entanglement . in the local cloning case , the amount of both @xmath1 entanglement and @xmath2 entanglement of the copied state is less than that of the original state for all values of @xmath57 . in the non - local cloning case ,
the amount of @xmath1 entanglement of the copied state is always less than that of the original state for all values of @xmath57 while the amount of @xmath2 entanglement of the copied state is less than that of the original state for most range of @xmath57 , i.e. , @xmath58 . in other words , the remaining @xmath2 entanglement after the non - local cloning is amplified versus the initial state when @xmath59 or @xmath60 .
this inter - two - particle entanglement amplification in a three - particle system is originated from inter - two - particle , and particle - copier interaction .
however , inter - three - qubit entanglement can not be amplified for both local and nonlocal cloning processes .
hence this does not violate the well known fact that copying a quantum system does not increase the information obtainable about the originals . finally , it is interesting to compare the fidelity @xmath61 of the output density operator after local copying relative to @xmath39 with the fidelity @xmath62 of the output density operator after non - local copying relative to @xmath39 .
the fidelity of a density matrix @xmath63 relative to @xmath39 is defined by @xmath64 . from eqs .
( [ 1 ] ) , ( [ e6 ] ) , and ( [ 17 ] ) we get that @xmath65 from which we find that the fidelity @xmath62 of the output state after non - local copying is independent of the values of @xmath57 .
this means that whether the initial state is like , the fidelity of the output is definite .
however , in the local copying process , the fidelity @xmath61 depends upon the initial state . from eq .
( [ 28 ] ) , we see that when @xmath66 or @xmath11 , the output state is the closest to the original state . when @xmath67 , the output state is the farthest to the original state . in fig.3
we plot the fidelity with respect to @xmath57 for local and nonlocal cloning cases , respectively . from fig .
3 it can be seen that the fidelity in the non - local cloning process is always greater than that in the local cloning process , i.e. , @xmath68 .
this means that the output state after non - local copying is closer to the original state than the output state after local copying .
we can see from the previous section that both local and non - local copying are not rather efficient process from the point of view of preserving entanglement . even in the case
when the maximally inseparable state , i.e. ghz state where @xmath36 , is the input state of the cloning process , only a small amount of entanglement survives the cloning .
it is an interesting question to ask what will happen when the output state of the copier is used as an input state in the next step of a sequence of cloning process . in particular
, we are interested in discovering how fast the entanglement decreases when the copying is iterated .
we restrict ourself in the case of non - local cloning , because this scheme is much more effective as we saw above , and results obtained in this case can be treated as an upper bound for all other schemes .
the final state of the copying process is a mixed state , described by the density matrix , eq .
( [ 17 ] ) .
this density matrix can not be used directly as input data in computations , because the non - local copying scheme works straightforwardly only when an input state is initially in a pure , potentially entangled state .
the density matrix should be first converted to a form which allows us to perform the second cloning .
it turns out that a simple diagonalization of the density matrix is sufficient . in this new diagonal basis ,
the density matrix is given by the mixture of projection operators @xmath69 the weights @xmath70 in the decomposition are the eigenvalues of the density matrix @xmath71 and the vectors @xmath72 are the normalized eigenvectors of @xmath71 .
each vector @xmath72 can be cloned separately .
the mixture of the resultant density matrices taken with the weights @xmath70 is the result of the second cloning process . as a specific example of iterating copying , in what follows we consider repetition of non - local quantum cloning for the ghz state given by eq .
( [ 1 ] ) through taking @xmath36 .
after the first non - local cloning , from eq .
( [ 17 ] ) we express the density operator of the out state of the original three input qubits as the form of the state ( [ 30 ] with the coefficients @xmath73 and eigenvectors of the out - state density operator given by @xmath74 .entanglement @xmath16 and @xmath75 of clones of the ghz state as a function of the number of cloning steps . [
cols="^,^,^,^,^,^,^,^",options="header " , ] the state described by the density operator ( [ 30 ] ) is the input state of the second copying .
we now separately copy the eigenstates of the density operator ( [ 30 ] ) , i.e. , @xmath76 , via non - local process .
after the second copying , we arrive at the output state expressed by the following density operator @xmath77 which leads to the following entanglement tensors @xmath78 hence after finishing the second copying , entanglement of the three - qubit system becomes @xmath79 from above calculations we can see that entanglement in the three - qubit system is further reduced after the second copying . repeating the above procedure we can investigate changes in entanglement @xmath1 and @xmath2 at each stage of iterating copy process . in table 1
we show the entanglement measures @xmath16 and @xmath17 of the clones of the maximally entangled ghz state for the case of the first five stages of the cloning . from the this table , we see that the entanglement decreases extremely rapidly , and after just six iterations the copy of the ghz state , the resultant entanglement goes to zero .
in conclusion , we have investigated remaining entanglement in quantum copying and iterating quantum copying of a three - qubit system by using the entanglement tensor approach .
specifically , we have studied the remaining inter - two - qubit and inter - three - qubit entanglements in the local and non - local cloning processes , respectively .
we have shown that entanglement of the three - qubit pure state can be locally or non - locally copied with the help of local quantum copiers or non - local quantum copiers , respectively , and that the amount of entanglements of the resultant copiers is reduced . it has been found that it is more efficient to preserve the entanglement by using non - local methods .
the amount of entanglement decreases rapidly with the increase of the number of copying times in the iterating quantum copying .
the amount of both @xmath1 and @xmath2 entanglements approach zero after six - times iterating quantum copying .
it means that even qubits , which are copies of copies of copies of copies of copies of copies of the ghz state are already in local state .
they do not have any nonclassical correlations and are useless as a resource in cryptographic conferencing or in multi - partite generalizations of super - dense coding @xcite .
this reduction is due to a residual entanglement between the copies and the quantum copying machine .
we have also shown that the inter - two - qubit entanglement is preserved better than the inter - three - qubit entanglement in the local cloning process while non - local cloning is much more efficient than the local cloning for broadcasting entanglement , and output state via non - local cloning exhibits the fidelity better than local cloning .
w. k. wootters and w. h. zurek , nature 299 ( 1982 ) 802 .
d. mozyrsky , v. privman , and m. hillery , phys .
lett . a 226 ( 1997 ) 253 .
v. buzek and m. hillery , phys .
a 54 ( 1996 ) 1844 .
n. gisin and s. massar , phys .
79 ( 1997 ) 2153 .
d. bruss , a. ekert , and c.macchiavello , phys .
81 ( 1998)2598 .
a. einstein , b. podolsky and n. rosen , phys .
47 ( 1935 ) 777 ; e. schrdinger , naturwissenschaften , 23 ( 1935 ) 807 ; 823 ; 844 . c. h. bennett , g. brassard , c. crpeau , r. jozsa , a. peres , and w. k. wootters , phys .
( 1993 ) 1895 ; d. bouwmeester , j .- w .
pan , k. mattle , m.eibl , h. weinfurter , a. zeilinger , nature 390 ( 1997 ) 575 . c. h. bennett and s. j. wiesner , phys .
69 ( 1992 ) 2881 .
w. shor , phys .
a 52 ( 1995 ) 2493 .
d. deutsch , proc .
london , a 425 ( 1989 ) 73 .
t. jennewein , c. simon , g. weihs , h. weinfurter , and a. zeilinger , phys .
84 ( 2000 ) 4729 .
v. giovannetti , s. lioyd , and l. maccone , nature 412 ( 2001)417 ; v. giovannetti , s. lloyd , and l. maccone , phys .
a 65 ( 2002 ) 022309 .
p. masiak and p. l. klight , quant - ph/9808043 .
v. buzek , v. vedral , m. b. plenio , p. l. knight , and m. hillery , phys . rev . a 55 ( 1997 ) 3327 .
j. schlienz and g. mahler , phys .
a 52 ( 1995 ) 4396 .
y . tong and l .- m .
kuang , chin .
17 ( 2000 ) 469 .
v. buzek and m. hillery , phys.rev . a 54 ( 1996 ) 1844 ; d. bru , d.p .
divincenzo , a. ekert , c.a .
fuchs , c. macchiavello , and j.a .
smolin , phys .
a 57 ( 1998 ) 2368 .
v. buzek and m. hillery , quant - ph/9801009 .
r. f. werner , phys .
a58 ( 1998 ) 1827 .
s. bose , v. vedral and p. l. klight , phys .
a. 57 ( 1998 ) 822 . | we study the degree to which quantum entanglement survives when a three - qubit entangled state is copied by using local and non - local processes , respectively , and investigate iterating quantum copying for the three - qubit system
. there may exist inter - three - qubit entanglement and inter - two - qubit entanglement for the three - qubit system .
we show that both local and non - local copying processes degrade quantum entanglement in the three - particle system due to a residual correlation between the copied output and the copying machine .
we also show that the inter - two - qubit entanglement is preserved better than the inter - three - qubit entanglement in the local cloning process .
we find that non - local cloning is much more efficient than the local copying for broadcasting entanglement , and output state via non - local cloning exhibits the fidelity better than local cloning . |
there is still no fundamental physical theory of the very early universe which addresses issues that arise from the regime of transplanckian physics .
one of these issues relates to the origin of the cosmological perturbation spectrum . in an expanding universe ,
the physical momentum gets blue - shifted back in time , therefore the observed low values of the momentum today that contribute to the cmbr spectrum may have originated from values larger than the planck mass @xmath4 in the early universe .
this is similar to the problems that arise in trying to explain the origin of hawking radiation in black hole physics . in a series of papers @xcite
, it was demonstrated that the hawking radiation remains unaffected by modifications of the ultra high energy regime , expressed through the modification of the usual linear dispersion relation at energies larger than a certain ultraviolet scale @xmath5 . following a similar procedure , in the case of an expanding friedmann - lemaitre - robertson - walker ( flrw ) spacetime , martin and brandenberger in ref .
@xcite ( see also @xcite ) showed that standard predictions of inflation are indeed sensitive to trans - planckian physics : different dispersion relations lead to different results for the cmbr spectrum .
it is the lack of a fundamental theory , valid at all energies , that makes the model building of the transplanckian regime very interesting .
the main issue is how much are the known observables affected by the unknown theory .
the apparently _ ad hoc _ modification of the dispersion relation at high energies is contrained by the criterion that its low energy predictions do no conflict the observables .
specifically , in this paper we address two questions : a ) can the transplanckian regime contribute to the dark energy of the universe , and b ) how sensitive is the cmbr spectrum to energies higher than the planck scale @xmath4 , where our current physics theory is known to break down .
we choose a family of dispersion relations for the frequency of the wavefunctions that modifies the behaviour of the field at the ultrahigh energies of the transplanckian regime .
the dispersion relation has the following features : it is smooth , nearly linear for energies less than the planck scale , reaches a maximum , and attenuates to zero at ultrahigh momenta thereby producing ultralow frequencies at very short distances .
we name the @xmath3 that part of the dispersion graph of very short distances that contains the range of ultralow frequencies less or equal to the current hubble constant @xmath2 .
it follows that the @xmath3 modes are still currently frozen .
we calculate the energy of the @xmath3 modes in order to address the former question ( a ) and show that although the @xmath3 does not contribute significantly to the cmbr spectrum , it has a dominant contribution to the dark energy of the universe @xcite .
the energy density of the @xmath3 modes is of the same order today as the matter energy density .
the second question ( b ) is motivated by the problem that in most inflationary models the present large scale structure of the universe is extrapolated from a regime of ultra - high energies ( known as the transplanckian regime ) originating from before the last 60 e - foldings of the exponential expansion . in refs .
@xcite the authors have demonstrated that the problem of calculating the spectrum of perturbations with a time - dependent dispersive frequency can be reduced to the familiar topic of particle creation on a time - dependent background @xcite .
we will use their observation in what follows .
they also conjecture that the observed power spectrum can always be recovered only by using a smooth dispersion relation , which ensures an adiabatic time - evolution of the modes . by taking the frequency dispersion relations to be the general class of epstein functions @xcite ,
we check and lend strong support to their conjecture . we present the exact solutions to the mode equation for the scalar field with a `` time - dependent mass '' , and the resulting cmbr spectrum below .
we show that the major contribution to the cmbr spectrum comes from the long wavelength modes when they re - enter the horizon .
the spectrum is nearly insensitive to the very short wavelength modes inside the hubble horizon .
the paper is organized as follows : in section 2 , we present the set - up and formalism of our analysis .
the family of dispersion functions , exact solutions to the mode equations of motion and the resulting cmbr spectrum ( from the bogoliubov method ) are reported in section 3 . in section 4
, we calculate the contribution of the _ tail _ modes to the dark energy of the universe today . in this work ,
we have neglected the backreaction effects of particle production .
this assumption is fully justified from the calculation of the energy for the transplanckian modes , in section 4 .
due to the _ dispersed _ ultralow frequency of these modes , the energy contained in that transplanckian regime is very small ( @xmath6 ) , thus the backreaction effect is reasonably negligible @xcite .
we present our conclusions in section 5 .
let us start with the generalized friedmann - lemaitre - robertson - walker ( flrw ) line - element which , in the presence of scalar and tensor perturbations , takes the form @xcite @xmath7 d x^i d x^j \right\ } \
, , \label{frw}\end{aligned}\ ] ] where @xmath8 is the conformal time and @xmath9 the scale factor .
the dimensionless quantity @xmath10 is the comoving wavevector , related to the physical vector @xmath11 by @xmath12 as usual . the functions @xmath13 and @xmath14 represent the scalar sector of perturbations while @xmath15 represents the gravitational waves .
@xmath16 and @xmath17 are the eigenfunction and eigentensor , respectively , of the laplace operator on the flat spacelike hypersurfaces . for simplicity , we will take a scale factor @xmath9 given by a power law , @xmath18 , where @xmath19 and @xmath20 .
the initial power spectrum of the perturbations can be computed once we solve the time - dependent equations in the scalar and tensor sector .
the mode equations for both sectors reduce @xcite to a klein - gordon equation of the form @xmath21 \mu_n=0 \ , , \label{kg}\ ] ] where the prime denotes derivative with respect to conformal time .
therefore , studying perturbations in a flrw background is equivalent to solving the mode equations for a scalar field @xmath22 related ( through bardeen variables @xcite ) to the perturbation field in the expanding background .
the above equation represents a linear dispersion relation for the frequency @xmath23 , @xmath24 the dispersion relation of eq .
( [ ldis ] ) holds for values of momentum smaller than the planck scale .
there is no reason to believe that it remains linear at ultra - high energies larger than @xmath4 . yet , nonlinear dispersion relations are quite likely to occur from the lagrangian of some effective theory obtained by the yet unknown fundamental theory .
nonlinear dispersion relations , similar to the ones we consider in this work , are known to arise in effective theories of : nonlocal condensed matter or particle physics models arising from non - canonical kinetic terms @xcite ; from the dissipative behavior of a quantum system immersed into an environment after coarse - graining @xcite ; or from effective theories with phase transitions , time - dependent mass squared terms or effective potentials @xcite .
perhaps , trans - planckian models motivated by superstring theory @xcite or a two - stage inflationary model @xcite are plausible . in the latter case
, one could easily envision for example a scenario with the first stage of inflation occurring at energy scales above the planck mass followed by a nonthermal phase transition @xcite .
the preheating @xcite from the nonthermal phase transition then leads to the second stage of inflation below planck energies .
in the former case , the motivation comes from the common belief that the superstring theory is the one that describes or at least is valid at energies of the transplanckian period .
taking this idea one step further , we incorporate the concept of superstring duality ( which applies at transplanckian regimes ) in our analysis by choosing a particular family of dispersion relations that exhibits _ dual _ behavior and winding radius @xmath25 , the frequency mode spectrum is dual in the sense that @xmath26 and @xmath25 are related as @xmath27 .
this means that each normal mode with a frequency @xmath28 , where @xmath29 is an integer , has its dual winding mode with decreasing energy that goes like @xmath30 @xcite . ] , i.e. appearance of ultra - low mode frequencies both at low and high momenta @xmath0 . despite the above comments and possible approaches
, we should stress that any modeling of planck scale physics by analogy with the already familiar systems is pure speculation .
we lack the fundamental theory that may naturally motivate or reproduce such dispersion behaviour
. nevertheless , it would be instructive to derive these dispersion relations from particle physics and string theory , as a step towards understanding the physical nature of the model . in what follows ,
we replace the linear relation @xmath31 with a nonlinear dispersion relation @xmath32 .
the family of dispersion functions @xmath33 for our model is introduced in section 3 .
these functions have the following features : they are linear for low momenta up to the plank scale @xmath34 , taken to be the cuttoff scale , but beyond the cutoff they smoothly turn down and asymptotically approach zero whereby producing ultra - low frequencies at very short distances . therefore , in eq . ( [ kg ] ) ,
@xmath35 should be replaced by : @xmath36 ^ 2 \ , .
\label{neff}\ ] ] we will also consider the general case of non - minimal or conformal coupling to gravity by keeping some arbitrary , unspecified , coupling constant @xmath37 . then , the equation for the scalar and tensor perturbations , that we need to solve , takes the form @xmath38 \mu_n = 0 \ , .
\label{modeneff}\ ] ] for future reference , we define the generalised comoving frequency as @xmath39 the dynamics of the scale factor is determined by the evolution of the background inflaton field @xmath40 , with potential @xmath41 , and the friedmann equation . in conformal time , these equations are : @xmath42 \phi^{\prime \prime } + 2 \frac{a^\prime}{a } \phi^{\prime } + a^2 \frac{\partial v(\phi)}{\partial \phi } & = & 0 \ , .
\label{phia2}\end{aligned}\ ] ] most of the contribution in the perturbation spectrum comes from long - wavelength modes , since at late times they are non relativistic and act like a classical homogeneous field with an amplitude @xmath43 given by : @xmath44 these are produced at early stages of inflation , thus they are very sensitive to the initial conditions .
the correct vacuum state is given by the solution to eq .
( [ modeneff ] ) as clarified below .
but if the stage of inflation is very long , and the hubble parameter @xmath45 is not changing considerably , then a minkowski - like vacuum is a good first order approximation for the initial vacuum state ( @xmath46 ) , with : @xmath47 however , if @xmath45 was greater at the early stages of inflation , before the last 60 e - foldings , and/or the inflationary stage is short , then one must solve the wave equation and find the solution that minimizes the energy @xcite .
this is the correct vacuum state of the system .
otherwise , if one a priori chooses the minkowski vacuum to be the initial vacuum state describing the system , the resulting value of @xmath48 is considerably underestimated as shown rigorously by felder et al .
@xcite . as already shown in ref .
@xcite , eq .
( [ modeneff ] ) represents particle production in a time - dependent background .
we will follow the method of bogoliubov transformation to determine the spectrum .
the correct initial condition for the vacuum state is the solution to the equation that minimizes the energy .
hence , if the time - dependent background goes asymptotically flat at late times , then in that limit the wavefunction should behave as a plane wave : @xmath49 as it is well known on this scenario , in general at late times one has a squeezed state due to the curved background that mixes positive and negative frequencies .
the evolution of the mode function @xmath50 at late times fixes the bogoliubov coefficients @xmath51 and @xmath52 , @xmath53 with the normalization condition : @xmath54 in the above expressions , @xmath55 and @xmath56 denote the asymptotic values of @xmath57 when @xmath58 .
the spectrum of particles per mode is then calculated with the conventional bogoliubov method @xcite .
the number of particles created @xmath29 and their energy density @xmath59 are calculated by the following expressions @xmath60 of the particles produced has a ( nearly ) thermal distribution , we can conclude that the cmbr in our problem is that of a ( nearly ) black body spectrum .
we introduce the class of epstein functions as the family of dispersion relations in section 3 , and derive the cmbr spectrum from the exact solutions to the evolution equations .
meanwhile , for the special features of our choice of dispersion relation , the modes at very high momenta but of ultra low frequencies @xmath61 are frozen for as long as the hubble expansion rate @xmath45 of the universe dominates over their frequency .
we refer to that as the @xmath3 of the dispersion graph . in fig .
( [ fig2 ] ) , for the dispersed @xmath62 vs. @xmath0 , the tail corresponds to all the modes beyond the point @xmath63 , where @xmath63 is defined by the condition @xmath64 , where @xmath2 is the hubble rate today .
it then follows that the tail modes are still frozen at present .
we calculate the total energy of the particles by using eq .
( [ energy ] ) , as well as the frozen energy of the @xmath3 .
thus the energy of the tail is a contribution to the dark energy of the universe : up to present it has the equation of state of a cosmological constant term .
however , through the friedmann equation , @xmath45 is a decreasing function of time because until now it has been dominated by the energy density of matter and radiation .
therefore , whenever @xmath45 drops below the frequency @xmath23 of an ultralow frequency mode , this mode becomes dynamic by picking a kinetic term and redshifts away very quickly .
hence , when the dominant contribution to the evolution equation for @xmath45 comes from the @xmath3 energy , the behaviour of those modes with equations of motion coupled to the friedmann equation becomes very complex .
it is hard to calculate at which rate @xmath45 drops in this situation .
if eventually , @xmath45 drops all the way to zero , all the modes in the tail must have decayed .
their equation of state , when @xmath45 becomes zero , is that of radiation . the reason can be traced back at their origin in transplanckian regime .
it is well known that scalar perturbations produced during inflation do not contribute to the total energy .
thus the origin of this modes is in the tensor perturbations . in their physical nature
they correspond to gravitational waves of very short distance but ultralow energy .
we calculate their energy today in section 4 .
we would like to elaborate on yet another possibility , which has not been mentioned before in the literature , that can give rise to a similar dispersion relation : very short or very large distance physics may have a curvature different from the flrw element of eq .
( [ frw ] ) , e.g. a different scale factor .
this becomes clearer when recognizing the strong relation between the time - dependent dispersion relation and the curvature given by the time derivatives of the scale factor @xmath9 .
the basic argument is the observation that the modulated frequency @xmath65 in the wave equation contains the contribution of these two terms , as given in eq .
( [ omegan ] ) .
therefore , while keeping the generalised frequency invariant , changing the first term in @xmath65 can be viewed or attributed to changes in the second term , such that : @xmath66 where @xmath67 is the new scale factor at very short distances . even in the conformal case @xmath68 , when the term @xmath69 drops out of eq .
( [ modeneff ] ) , the time dependent frequency @xmath70 can mimic a term proportional to @xmath71 at @xmath72 . thus , any modulation of the dispersion relation is equivalent to a change in the behavior of the time - dependence of the background ( a.k.a . ,
the scale factor / curvature ) .
in other words , we could have introduced a different curvature at very short ( or large ) distances from the start instead of a dispersed frequency .
we will consider the class of inflationary scenarios that through eqs .
( [ phia1 ] ) and ( [ phia2 ] ) has a power law solution for the scale factor @xmath9 in conformal time , @xmath73 , with @xmath74 , and the following epstein function @xcite for the dispersion relation : @xmath75 where @xmath76 , with @xmath77 .
this is the most general expression for this family of functions . for our purposes
, we will constrain some of the parameters of the epstein family in order to satisfy the features required for the dispersion relation as follows . first , imposing the requirement of superstring duality , in order to have ultralow frequencies for very high momenta , we demand that the dispersion functions go asymptotically to zero .
that fixes @xmath78 on the other hand , the condition of a nearly linear dispersion relation for @xmath79 requires that @xmath80 still we will have a whole family of functions parametrised by the constant @xmath81 , as can be seen in fig .
[ fig1 ] . with the change of variables @xmath82 ,
the scalar wave equation ( [ modeneff ] ) for the mode @xmath50 becomes : @xmath83 \mu_n = 0 \ , , \label{modev}\ ] ] with : @xmath84 where : @xmath85 in the case of conformal coupling to gravity , @xmath68 , eq .
( [ modev ] ) is exactly solvable in terms of hypergeometric functions @xcite .
this is a well studied case in the context of particle creation in a curved background @xcite .
even if we are not in the case of conformal coupling , the contribution @xmath86 is going to be negligible at early times ( @xmath87 ) ; at late times , it can be absorbed in the dispersion relation eq .
( [ disp ] ) redefining the constants @xmath88 .
as explained in section 2 , the correct initial condition is the vacuum state solution that minimizes the energy . in the case where @xmath89 , this vacuum state behaves as a plane wave in the asymptotic limit @xmath87 , with @xmath90 .
however , when @xmath91 as in our case , the correct behavior of the mode function in the remote past is given by the solution of eq .
( [ modeneff ] ) in the limit @xmath87 .
the exact solution which matches this asymptotic behavior is then given by : @xmath92 \,,\ ] ] where @xmath93 is a normalization constant , and @xmath94 at late times the solution becomes a squeezed state by mixing of positive and negative frequencies : @xmath95 with @xmath96 being the bogoliubov coefficient equal to the particle creation number per mode @xmath29 , and @xmath97 . using the linear transformation properties of hypergeometric functions @xcite , we find that in eq .
( [ disp ] ) , it is obtained @xcite : @xmath98 with @xmath99 .
also in this case the spectrum of the fluctuations is nearly thermal , with the parameter @xmath100 controlling the deviation from thermality . ] @xmath101 if @xmath102 is a real number ( @xmath103 ) , then we obtain : @xmath104 it is clear from eqs .
( [ betak ] ) and ( [ betak2 ] ) that the spectrum of created particles is nearly thermal to high accuracy ) and a very small energy contained in these modes . clearly , the particle number per mode being small is consistent with the result of the exponentially suppressed , near - thermal bogoliubov coefficient .
see section 4 for the energy .
] , @xmath105 thus , we can immediately conclude that the cmbr spectrum is that of a ( nearly ) black body spectrum .
that means the spectrum is ( nearly ) scale invariant , i.e. , the spectral index is @xmath106 .
this is consistent with previous results obtained in the literature @xcite , when using a smooth dispersion relation and the correct choice of the initial vacuum state , as discussed above . in refs . @xcite and @xcite , dispersion relations , that were originally applied to black hole physics @xcite ,
were used in the context of cosmology .
new models of dispersion relations were proposed by the authors of refs . @xcite and @xcite .
our proposal for a 1-parameter class of models has a significantly different feature from the above , namely : the appearence of ultra - low frequency modes in the transplanckian regime .
the implications of such a behavior for high momenta on the production of dark energy are discussed in the next section .
in section 2 we defined the @xmath3 as the range of those modes in the frequency dispersion class ( originating from the transplanckian regime ) , whose frequency is less or at most equal to the present hubble rate , @xmath2 ( see fig .
it then follows that they have not decayed and redshifted away but are still frozen @xmath107 .
since @xmath45 has been a decreasing function of time , many modes , even those in the ultralow frequency range , have become dynamic and redshifted away one by one , everytime the above condition is broken , i.e. when the expansion rate @xmath45 dropped below their frequency . clearly , the other modes have long decayed into radiation and the tail modes are the only modes still frozen .
they contain vacuum energy of very short distance , hence of very low energy .
the last mode in the tail would decay when and if @xmath108 .
when the tail modes become dynamic by acquiring a kinetic term ( when @xmath109 ) , they decay away as gravitational waves ( explained in section 2 ) .
the @xmath3 starts from some value @xmath63 which must be found by solving the equation @xmath110 the range of the modes defining the @xmath3 is then for @xmath111 .
their time - dependent behaviour when they decay depends on the evolution of @xmath45 and is complicated because they contribute to the expansion rate for @xmath45 .
thus , their equations of motion are coupled to the friedmann equation
. however we can calculate their contribution to the dark energy today , when they are still frozen , thereby mimicking a cosmological constant .
we calculate numerically ( using mathematica ) the range of the modes in the tail from eq .
( [ omegah ] ) and use this value for the limit of integration in the tail energy given by eq .
( [ energy ] ) .
below we report these results for the case of a scale factor with @xmath112 but other values of @xmath113 were also considered numerically and they produce an even smaller dark energy due to the extra suppression in the integral coming from the bogoliubov coefficient @xmath114 .
( [ omegah ] ) is a messy transcendental equation but the solution to that equation is crucial to the dark energy since the value @xmath63 is the limit of the energy integral .
that is why we solved eq .
( [ omegah ] ) numerically and replaced it in eq .
( [ energy ] ) for the energy , using different representative values of the parameter @xmath81 .
the energy for the tail is given by : @xmath115 while the expression for the total energy is : @xmath116 the numerical calculation of the tail energy produced the following result : for random different values of the free parameters , the dark energy of the tail is @xmath117 , times less than the total energy _ during inflation _ ,
i.e. @xmath118 at planck time .
the prefactor @xmath119 , which depends weakly on the parameter of the dispersion family @xmath81 , is a small number between 1 to 9 , which clearly can contribute at the most by 1 order of magnitude .
_ this is an amazing result ! _
it can readily be checked by plugging in the dispersion expression for @xmath120 , eq .
( [ omega ] ) , in the integral expression of eq .
( [ entail ] ) for the tail energy , then using as the limit of integration the value @xmath63 found by the condition in eq .
( [ omegah ] ) . this result can be understood qualitatively by noticing that the behavior of the frequency for the `` tail '' modes is nearly an exponential decay ( see eq .
( [ omega ] ) ) , and as such dominates over the other terms in the energy integrand of eq .
( [ entail ] ) : @xmath121 hence , due to the decaying exponential , the main contribution to the energy integral in eq .
( [ entail ] ) comes from the highest value of this exponentially decaying frequency , which is the value of the integrand at the tail starting point , @xmath122 , i.e. , @xmath123 due to the physical requirement that the tail modes must have always been frozen , the tail starting frequency @xmath124 is then proportional to the current value of hubble rate @xmath2 ( eq . ( [ omegah ] ) ) .
we suspect this result is generic for any scenario that features _ ultralow frequencies _ which exponentially decay to zero at very high momenta for two reasons .
first , all the modes with an ultralow frequency @xmath125 will be frozen and thus produce dark energy .
secondly , their contribution to the energy may be small because of the following . due to this kind of dispersion in the high momentum regime , the phase space available for the ultra - low frequency modes with @xmath126
gets drastically reduced when compared to the phase space factor in the case of a non - dispersive transplanckian regime , controlling in this way these modes contribution to the energy density .
the result for the tail energy also means that the tail energy dominates today s expansion of the universe .
thus , at present , we can not tell a priori the evolution of these modes and the time when they may become dynamic .
only the solution to the equation for the modes coupled ( strongly at present ) to the friedmann equation would answer the question as to whether @xmath45 will continue to decrease whith time .
if that were the case , then these tail modes would also eventually become dynamic and decay .
however , we calculated the equation of state for the limiting case when @xmath108 . in this case , all the modes in the tail are dynamic .
the calculation of the energy density of the tail in the dynamic case from eq .
( [ energy ] ) confirms that the tail decays in the form of radiation , as expected since their physical nature is that of gravity waves of very short distances ( but ultralow energy ) , originating from tensor perturbations during inflation .
the opposite case is also a possible outcome to the coupled equations .
it s possible that the frozen modes of the tail will prevent @xmath45 from dropping further below , in which case these modes will never decay .
we have not solved these coupled equation yet , therefore we are just speculating on the two possible outcomes of that calculation .
the solution is left for future work . at present , these modes , originating from the transplanckian regime , are behaving as dark energy of the same magnitude as the current total energy in the universe .
this idea is then a leap forward to this longstanding and challenging problem of dark energy , for at least two reasons : first , inspired by superstring duality , it is very plausible to speak of scenarios with ultralow frequencies and very high momenta .
the tail modes , that are frozen at present , provide a good candidate for the dark energy as our calculations show .
secondly , although smooth dispersion functions that model the transplanckian regime do not affect the cmbr spectrum , this regime still leaves its imprints in the contribution to the energy of the universe .
this is a rich and currently underexplored area to consider with respect to the cosmological constant mystery .
in this work we investigated two phenomenological aspects of transplanckian physics : the issue of dark energy production , and the sensitivity of the observed cmbr spectrum to the transplanckian regime . for this purpose , a family of dispersion relations is introduced that modulate the high frequencies of the inflationary perturbation modes at large values of the momenta @xmath0 for the transplanckian regime .
the smooth dispersion relations are chosen such that the frequency graph attenuates to zero at very high @xmath0 , thereby producing ultralow frequencies corresponding to very short distances , but it is nearly linear for low values of @xmath0 up to the cuttoff scale @xmath127 .
we present the exact solutions to the mode equations and calculate the spectrum through the method of bogoliubov coefficients .
the resulting cmbr spectrum is shown to be ( nearly ) that of a _
black body_. this calculation lends strong support to the conjecture that smooth dispersion relations which ensure an adiabatic time - evolution of the modes produce a nearly scale invariant spectrum .
further , we elaborate on the issue of the _ initial conditions _ to which the spectrum is highly sensitive and show that there is _ no ambiguity _ in the correct choice of the initial vacuum state .
the only initial vacuum is the _ adiabatic vacuum _ obtained by the solution to the mode equation . on the other hand
, we showed that the assumption of neglecting the possible back reaction effects of the tail modes on the inflaton field is reasonable and is justified by the result of the tail energy calculation of section 4 .
also the bogoliubov coefficient obtained , eq .
( [ bogol ] ) , is exponentially suppressed , so back - reaction does not become significant .
we would also like to stress that due to the dispersion class of functions chosen , defined in the whole range of momenta from zero to infinity , the total energy contribution of the modes produced is _ finite _ , without the need of applying any renormalization / subtraction scheme . in a sense
, the regularization - renormalization procedure is encoded in the class of dispersion we postulate . the most exciting result of this work is the generation of dark energy in the observed amount for the present universe @xcite .
this has its origin at the transplanckian regime , due to the presence of the dispersed _
tail modes _ with ultralow frequencies equal or less than the current hubble constant .
the evolution of these modes is given by their equation of motion , and it depends on the hubble rate through the friction - like term @xmath128 . on the other hand ,
the evolution of the hubble rate , given by the friedmann equation , contains contribution from the energy of these modes .
but currently , the hubble constant dominates over their frequency in the mode equation of motion .
it then follows that the tail modes , up to present , are still frozen and have been behaving like a cosmological constant term .
therefore their energy is dark energy .
we have calculated numerically the energy of the @xmath3 during the inflationary stage for different values of the dispersion parameter @xmath129 .
the calculation showed that the tail energy was @xmath130 ( times a prefactor @xmath119 which weakly depends on @xmath81 and , for random values of the parameter , takes values between 1 to 9 ) orders less than the total energy during inflation .
this result is true for the whole class of dispersion relations .
we chose different random values of the 1-parameter dispersion family @xmath81 , and the numerical calculation shows that @xmath81 influences the energy at the most by less than an order of magnitude .
we _ did not need to do any tuning _ of the parameters and used the _ planck scale as the fundamental scale of the theory_. clearly , at present the tail energy dominates in the friedmann equation , if the ratio of its energy to the total energy ( as found by the calculation ) was @xmath131 during inflation .
the @xmath3 thus provides a _ strong candidate _ for explaining the dark energy of the universe @xcite .
we suspect that the above result of producing such an extremely small number for the @xmath3 energy without any fine - tuning ( and by using @xmath132 as the only fundamental scale of the theory ) , is generic for any dispersion graph with a @xmath3 .
the family of dispersion relations that feature a @xmath3 , corresponding to vacuum modes of very short distances , was motivated by superstring duality @xcite
. however , in section 2 we made the observation that _ introducing a dispersion relation is equivalent to introducing changes in the curvature of the universe _ , at very short or very large distances while keeping the generalized frequency @xmath57 of eq .
( [ modeneff ] ) invariant .
it is quite possible that the dispersion relation for the @xmath3 modes results from a different curvature of the universe at very short distances .
this is an important link and we use it in a sequential paper @xcite to demonstrate that the sn1a data can not be reinterpreted away by changing the large scale curvature of the universe .
although it is counterintuitive , since large distance would correspond to low energy theories , we show in @xcite that any changes in large scale curvature would disagree with the observed cmbr spectrum
. it would be interesting to know what happens to the tail and the hubble rate in the future .
after all , a model is useful in so far as it can make future predictions .
although conceptually it is straightforward to find out the answer , given by the solution to the coupled equations ( [ modeneff ] ) , ( [ phia1 ] ) and ( [ phia2 ] ) , technically it appears difficult to predict the future evolution of the tail modes .
the technical difficulty lies in the fact that at present , the @xmath3 equation of motion is strongly coupled to the friedmann equation for @xmath45 since the @xmath3 energy dominates .
the hubble rate would continue to decrease _ only if _ these modes decay , but
these modes can decay _ only when _ the hubble rate decreases below their frequency .
it may be possible that the frozen tail will sustain a constant hubble rate which in turn will not allow the further decay of the tail modes .
it is also possible that @xmath45 will continue to decrease in which case the tail modes will become dynamic and redshift away in the future .
it is only the solution to the mode equations coupled to the friedmann equation , that will provide the answer on whether the hubble constant and tail will decay in the future or remain at their current value .
we do not have this solution yet , and the work is left for future investigation .
in addition , the equation of state , @xmath133 , is an @xmath134 that will provide a test to the model @xcite , especially with the new data coming in the near future from the snap @xcite and sdss @xcite missions .
however , we know that currently these @xmath3 modes are frozen vacuum modes of ultralow energy but very short distance , thus their energy behaves like a cosmological constant energy .
we also know they become dynamic and acquire a kinetic term only when the hubble rate drops below the frequency . and , if they decay , the product is radiation of gravity waves at very short distances since their physical origin is from the tensor perturbations during inflation ( it is well known that scalar perturbations do not contribute to the energy ) .
the condition for the decay of the last mode in the tail is fulfilled when @xmath45 has dropped to zero .
many of our results , e.g. the dispersion family and the exact solutions together with the bogoliubov coefficients , could be applied to the black hole physics .
the issue of transplanckian physics was originally raised in the black hole context with respect to the sensitivity of the hawking radiation to the blueshifted , superplanckian energy wavepackets . following a phenomenological approach ,
a few dispersive models were introduced @xcite-@xcite in order to introduce a bound on the blueshifted energies and check the sensitivity of the black hole spectrum .
we have introduced a new , different family of dispersive models , that also gives rise to a thermal spectrum .
the analytical results of our class of dispersion models can be applied to the black hole physics and reproduce the same thermal hawking spectrum .
there are many subtleties involved due to the different symmetries of the two scenarios , but these issues are beyond the purpose and scope of this paper .
it is left for future work . however ,
if the hawking radiation for this class of dispersions is again thermal , it lends strong support to unruh s conjecture that black hole radiation is insensitive to physics in the far ultraviolet ( trans - planckian ) regime , being predominantly an infrared effect . on the other hand , our class of models departs from the previous ones in the asymptotic behavior at very high momenta , with the presence of an infinite `` tail '' of ultralow frequency modes .
the @xmath3 feature and energy results , applied to a black hole case , may raise interesting issues , in particular with respect to the black hole s information loss paradox .
* acknowledgment : * we are very grateful to a. riotto for many beneficial comments and for pointing out the connection with superstring duality . we also want to thank r. barbieri , r. rattazzi , l. pilo , s. carroll , c. t. hill for helpful discussions .
p.k . would like to acknowledge financial support by ec under the tmr contract no .
hprn - ct-2000 - 00148 .
l. parker , phys . rev . * 183 * , 1057 ( 1969 ) ; r. u. sexl and h. k. urbantke , phys . rev . * 179 * , 1247 ( 1969 ) ; y. zeldovich and a. starobinsky , zh .
. fiz . * 61 * , 2161 ( 1971 ) [ sov .
jetp * 34 * , 1159 ( 1971 ) ] ; ya .
zeldovich , pisma zh . eksp .
* 12 * , 443 ( 1970 ) [ sov .
jetp lett .
* 12 * , 302 ( 1970 ) ] ; b. l. hu , phys . rev .
* d9 * , 3263 ( 1974 ) ; s. w. hawking , nature * 248 * , 30 ( 1974 ) ; comm .
* 43 * , 199 ( 1975 ) ; w. g. unruh , phys .
* d14 * , 870 ( 1976 ) ; b. k. berger , phys . rev . *
d12 * , 368 ( 1975 ) ; l. mersini , int . j. mod
a13 * , 2123 ( 1998 ) .
w. g. unruh , phys .
* 46 * , 1351 ( 1981 ) ; d. a. lowe et al .
, phys . rev .
* d52 * , 6997 ( 1995 ) ; m. visser , class .
grav . * 15 * , 1767 ( 1998 ) ; t. a. jacobson and d. mattingly , gr - qc/0007031 ; hep - ph/0009051 .
b. l. hu and y. zhang , `` coarse - graining , scaling and inflation '' , univ .
maryland preprint 90 - 186 ( 1990 ) ; b. l. hu , in `` relativity and gravitation : classical and quantum '' , proc .
silarg vii , cocoyoc , mexico , 1990 , eds .
j. c. d olivo et al .
( world scientific , singapore 1991 ) .
a. albrecht and p. steinhardt , phys .
lett . * 48 * , 1220 ( 1982 ) ; l. a. kofman et al .
, phys . rev . * b157 * , 361 ( 1985 ) ; a. albrecht , in proc .
`` the santa fe tasi 87 '' , eds .
r. slansky and g. west , world scientific ( 1998 ) ; n. b. kopnin and g. e. volovik , phys . rev .
* b57 * , 8525 ( 1998 ) ; t. a. jacobson and g. e. volovik , phys . rev .
* d58 * , 064021 ( 1998 ) ; g. l. kane , phys
. rep . * 307 * , 197 ( 1998 ) ; a. a. starobinsky , jetp lett . * 68 * , 757 ( 1998 ) ; nature * 331 * , 673 ( 1998 ) ; m. s. turner , astro - ph/9912211 ; p. j. steinhardt et al .
, astro - ph/0006373 ; d. huterer and m. s. turner , astro - ph/0012510 . | as yet , there is no underlying fundamental theory for the transplanckian regime .
there is a need to address the issue of how the observables in our present universe are affected by processes that may have occurred at superplanckian energies ( referred to as the _ transplanckian regime _ ) . specifically , we focus on the impact the transplanckian regime has on two observables , namely : dark energy and the cmbr spectrum .
we model the transplanckian regime by introducing a 1-parameter family of smooth non - linear dispersion relations which modify the frequencies at very short distances .
a particular feature of the family of dispersion functions chosen is the production of ultralow frequencies at very high momenta @xmath0 ( for @xmath1 ) .
we name the range of the ultralow energy modes ( of very short distances ) that have frequencies equal or less than the current hubble rate @xmath2 as the _ tail _ modes .
these modes are still frozen today due to the expansion of the universe .
we calculate their energy today and show that the @xmath3 provides a strong candidate for the _ dark energy _ of the universe . during inflation ,
their energy is about 122 - 123 orders of magnitude smaller than the total energy , for any random value of the free parameter in the family of dispersion relations . for this family of dispersions ,
we present the exact solutions and show that : the cmbr spectrum is that of a ( nearly ) black body , and that the adiabatic vacuum is the only choice for the initial conditions .
6.0 in 9.0 in # 1#2#3 # 1#2#3 # 1#2#3 # 1#2#3 # 1#2#3 # 1#2#3 # 1#2#3 hep - ph/0101210 + sns - ph/01 - 01 + + laura mersini , mar bastero - gil and panagiota kanti + + _ scuola normale superiore and infn , piazza dei cavalieri 7 , + i-56126 pisa , italy _
+ [ cols= " < " , ] |
recently , spinor bose - einstein condensates ( becs ) trapped in optical potentials have received much attention in both experimental stenger , anderson , liu and theoretical studies @xcite .
spinor becs have internal degrees of freedom due to the hyperfine spin of the atoms which liberate a rich variety of phenomena such as spin domains @xcite and textures @xcite
. when the potential valley is so deep that the individual sites are mutually independent , spinor becs at each lattice site behave like spin magnets and can interact with each other through both the light - induced and the static , magnetic dipole - dipole interactions .
these site - to - site dipolar interactions can cause the ferromagnetic phase transition @xcite leading to a macroscopic magnetization of the condensate array and the spin - wave like excitation @xcite analogous to the spin - wave in a ferromagnetic spin chain . for the real spin chain ,
the site - to - site interaction is caused mainly by the exchange interaction , while the dipole - dipole interaction is negligibly small .
for the spinor becs in the optical lattice , the exchange interaction is absent .
the individual spin magnets are coupled by the magnetic and the light - induced dipole - dipole interactions @xcite which are no longer negligible due to the large number of atoms @xmath0 at each lattice site , typically of the order of 1000 or more .
therefore , the spinor becs in an optical lattice offer a totally new environment to study spin dynamics in periodic structures .
the magnetic soliton excited by the interaction between the spin waves @xcite is an important and interesting phenomenon in spinor becs . in this paper , we demonstrate that the magnetic soliton and elastic soliton collision are admitted for spinor becs in a one - dimensional optical lattice and are controllable by adjusting of the light - induced and the magnetic dipole - dipole interactions . the heisenberg model of spin - spin interactions is considered as the starting point for understanding many complex magnetic structures in solids . in particular , it explains the existence of ferromagnetism and antiferromagnetism at temperatures below the curie temperature .
the magnetic soliton @xcite , which describes localized magnetization , is an important excitation in the heisenberg spin chain tjon , li , ablowitz , huang .
the haldane gap @xcite of antiferromagnets has been reported in integer heisenberg spin chain . by means of the neutron inelastic scattering @xcite and electron spin resonance asano00 ,
the magnetic soliton has already been probed experimentally in quasi - one dimensional magnetic systems .
solitons can travel over long distances with neither attenuation nor change of shape , since the dispersion is compensated by nonlinear effects .
the study of soliton has been conducted in as diverse fields as particle physics , molecular biology , geology , oceanography , astrophysics , and nonlinear optics .
perhaps the most prominent application of solitons is in high - rate telecommunications with optical fibers .
however , the generation of controllable solitons is an extremely difficult task due to the complexity of the conventional magnetic materials .
the spinor becs seems an ideal system to serve as a new test ground for studying the nonlinear excitations of spin waves both theoretically and experimentally .
the outline of this paper is organized as follows : in sec .
ii the landau - lifshitz equation of spinor bec in an optical lattice is derived in detail .
next , we obtain the one - soliton solution of spinor bec in an optical lattice .
the result shows that the time - oscillation of the amplitude and the size of soliton can be controlled by adjusting of the light - induced dipole - dipole interaction .
we also present that the magnetization varies with time periodically . in sec .
vi , the general two - soliton solution for spinor bec in an optical lattice is investigated .
analysis reveals that elastic soliton collision occurs and there is a phase exchange during collision .
finally , our concluding remarks are given in sec .
the dynamics of spinor becs trapped in an optical lattice * * * * is primarily governed by three types of two - body interactions : spin - dependent collision characterized by the @xmath1-wave scattering length , magnetic dipole - dipole interaction ( of the order of bohr magneton @xmath2 ) , and light - induced dipole - dipole interaction adjusted by the laser frequency in experiment .
our starting point is the hamiltonian describing an @xmath3 spinor condensate at zero temperature trapped in an optical lattice , which is subject to the magnetic and the light - induced dipole - dipole interactions and is coupled to an external magnetic field via the magnetic dipole hamiltonian @xmath4 @xcite , @xmath5\hat{\psi}_{\alpha } ( % \mathbf{r } ) \notag \\ & & + \sum_{\alpha , \beta , \upsilon , \tau } \int d\mathbf{r}d\mathbf{r}^{\prime } % \hat{\psi}_{\alpha } ^{\dagger } ( \mathbf{r})\hat{\psi}_{\beta } ^{\dagger } ( % \mathbf{r}^{\prime } ) \left [ { } \right .
u_{\alpha \upsilon \beta \tau } ^{coll}(% \mathbf{r , r}^{\prime } ) + u_{\alpha \upsilon \beta \tau } ^{d - d}(\mathbf{r , r}% ^{\prime } ) \left . { } \right ] \hat{\psi}_{\tau } ( \mathbf{r}^{\prime } ) \hat{\psi% } _ { \upsilon } ( \mathbf{r})+h_{b } , \label{hamilton}\end{aligned}\]]where @xmath6 is the field annihilation operator for an atom in the hyperfine state @xmath7 , @xmath8 is the lattice potential , the indices @xmath9 which run through the values @xmath10 denote the zeeman sublevels of the ground state .
the parameter @xmath11 describes the two - body ground - state collisions and @xmath12 includes the magnetic dipole - dipole interaction and the light - induced dipole - dipole interaction .
when the optical lattice potential is deep enough there is no spatial overlap between the condensates at different lattice sites .
we can then expand the atomic field operator as @xmath13 @xmath14 , where @xmath15 labels the lattice sites , @xmath16 is the condensate wave function for the @xmath15th microtrap and the operators @xmath17 satisfy the bosonic commutation relations @xmath18=\delta _ { \alpha \beta } \delta _ { nl}$ ] .
it is assumed that all zeeman components share the same spatial wave function .
if the condensates at each lattice site contain the same * * * * number of atoms * * * * @xmath0 , the ground - state wave functions for different sites have the same form @xmath19 .
in this paper we consider a one - dimensional optical lattice along the @xmath20-direction , which we also choose as the quantization axis . in the absence of spatial overlap between individual condensates , and neglecting unimportant constants , we can construct the effective spin hamiltonian @xcite as @xmath21 , \label{hamilton2}\ ] ] where @xmath22 , @xmath23 and @xmath24 represent the magnetic and the light - induced dipole - dipole interaction respectively .
the direction of the magnetic field @xmath25 is along the one - dimensional optical lattice and @xmath26 is the gyromagnetic ratio .
the spin operators are defined as @xmath27 , where @xmath28 is the vector operator for the hyperfine spin of an atom , with components represented by @xmath29 matrices in the @xmath30 subspace . the first term in eq .
( [ hamilton2 ] ) * * * * is resulted * * * * from the spin - dependent interatomic collisions at a given site , with @xmath31 , where @xmath32 characterizes the spin - dependent @xmath1-wave collisions .
the second and the third terms describe the site to site spin coupling induced by the static magnetic field dipolar interaction and the light - induced dipole - dipole interaction . for @xmath33 , the transfer of transverse excitation from site to site is allowed , resulting in the distortion of the ground - state spin structure .
this distortion can propagate and hence generate spin waves along the atomic spin chain . for an optical lattice created by blue - detuned laser beams ,
the atoms are trapped in the dark - field nodes of the lattice and the light - induced dipole - dipole interaction is very small @xcite .
however , this small light - induced dipole - dipole interaction induces the amplitude and size of the soliton varying with time periodically as we will show in the following section . from hamiltonian ( [ hamilton2 ] ) we can derive the heisenberg equation of motion at @xmath34th site for the spin excitations .
when the optical lattice is infinitely long and the spin excitations are in the long - wavelength limit , i.e. , the continuum limit , @xmath35 , we obtain the landau - lifshitz equation of a spinor becs in an optical lattice as follows : @xmath36+\frac{\gamma bs^{y}}{% \hbar } , \notag \\
\frac{\partial s^{y}}{\partial t } & = & \frac{2\lambda } { \hbar } [ a^{2}(s^{z}% \frac{\partial ^{2}}{\partial z^{2}}s^{x}-s^{x}\frac{\partial ^{2}}{\partial z^{2}}s^{z})+4\frac{\lambda ^{ld}}{\lambda } s^{z}s^{x}]-\frac{\gamma bs^{x}}{% \hbar } , \notag \\
\frac{\partial s^{z}}{\partial t } & = & \frac{2\lambda } { \hbar } [ a^{2}(s^{x}% \frac{\partial ^{2}}{\partial z^{2}}s^{y}-s^{y}\frac{\partial ^{2}}{\partial z^{2}}s^{x } ) ] , \label{magnet}\end{aligned}\]]where we assume that all nearest - neighbor interactions are the same , namely @xmath37 , which is a good approximation in the one - dimensional optical lattice for the large lattice constant @xcite . * * * * in a rotating frame around @xmath20-axis with angular frequency @xmath38 the spin vector @xmath39 is related with the original one by the transformation @xmath40thus the landau - lifshitz equation ( [ magnet ] ) in the rotating coordinate system can be written as @xmath41where @xmath42 , and the superscript ( @xmath43 ) is omitted for the sake of pithiness .
the dimensionless time @xmath44 and coordinate @xmath20 in eq .
( [ magnet2r ] ) are scaled in unit @xmath45 and @xmath46 respectively , where @xmath46 denotes the lattice constant . also , the terms including the external magnetic field in eq .
( [ magnet ] ) have been eliminated with the help of the transformation .
the equation ( [ magnet2r ] ) has a form of the landau - lifshitz ( ll ) type which is similar to the ll equation for a spin chain with an easy plane anisotropy @xcite . by introducing a particular parameter huang huang showed that the jost solutions can be generated and the lax equations are satisfied , and moreover constructed darboux transformation matrices .
an explicit expression of the one - soliton solution in terms of elementary functions of @xmath20 and @xmath44 was reported . here using the similar method in ref .
@xcite we obtain both the one - and two - soliton solutions ( for detail see the appendix ) denoted by @xmath47 with @xmath48 of eq .
( [ magnet2r ] ) * * * * in the following form : @xmath49 where the parameters in the solution are defined by @xmath50with @xmath51 , @xmath52 , @xmath53 , @xmath54 , @xmath55 , @xmath56 , @xmath57 , and @xmath58 .
the one - soliton solution , namely @xmath59 , is simply * * * * that @xmath60the parameter @xmath61 denotes the velocity of envelope motion of the magnetic soliton . the real constants @xmath62 and @xmath63 represent the center position and the initial phase respectively .
the parameter @xmath64 is eigenvalue with @xmath65 , @xmath66 being the real and imaginary parts . the one - soliton solution ( [ onesoliton ] ) describes a spin precession characterized by four real parameters : velocity @xmath67 , phase @xmath68 , the center coordinate of the solitary wave @xmath62 and initial phase @xmath63 . from the one - soliton solution we obtain the properties of the soliton : ( i ) both the amplitude and the size of the soliton vary with time periodically , as shown in figure 1 , in which we demonstrate graphically the dynamics of soliton with the parameters chosen as @xmath69 , @xmath70 , @xmath71 , @xmath72 and @xmath73 . this property * * * * is resulted from the term @xmath74 in eq .
( [ magnet2r ] ) which is determined by the light - induced dipole - dipole interaction .
this significant observation from the one - soliton solution shows that the time - oscillation of the amplitude and the size of soliton can be controlled in practical experiment by adjusting of the light - induced dipole - dipole interaction .
( ii ) the magnetization defined by * * * * @xmath75 varies with time periodically as shown in figure 2 .
these properties are similar to that of the heisenberg spin chain with an easy plane anisotropy where the dipolar coupling is typically several orders of magnitude weaker than the exchange coupling and thus would correspond to curie temperatures much below the observed values . hence the contribution of the dipolar coupling to the spin wave can be neglected in practice . however , for the spinor bec in the optical lattice the exchange interaction is absent and the individual spin magnets are coupled by the magnetic and the light - induced dipole - dipole interactions . due to the large number of atoms @xmath0 at each lattice site , these site to site interactions , despite the large distance between sites , explain the natural existence of magnetic soliton which agrees with the results in refs .
@xcite . to see closely the physical significance of one - soliton solution , it is helpful to show the parameter - dependence of euler angles of the magnetization vector which in a spherical coordinate is @xmath76 from eqs .
( [ solitonr ] ) and ( [ onesoliton ] ) we find @xmath77 + \chi _ { 3,1}\sin ^{2}\phi _ { 1}}% , \notag \\ \varphi & = & \frac{\pi } { 2}+\arctan \left ( \eta _ { 1}\tan \phi _ { 1}\right ) + \arctan \left ( \tanh \theta _ { 1}\right ) , \label{polar1r}\end{aligned}\ ] ] where @xmath78 and the maximal amplitude @xmath79 .
when @xmath80 , the phase @xmath81 can be rewritten as @xmath82 where the wave number @xmath83 and the frequency of magnetization precession @xmath84 are related by the dispersion law @xmath85 we also see that the position of minimum of energy spectrum @xmath86 is located at @xmath87 .
if amplitude @xmath88 approaches zero , namely @xmath89 , the parameter @xmath90 diverges and eq .
( [ polar1r ] ) takes the asymptotic form @xmath91 indicating a small linear solution of magnon . in this case
the dispersion law reduces to @xmath92 .
the magnetic soliton collision in spinor becs is an interesting phenomenon in spin dynamics .
here in terms of a darboux transformation we first of all give the two - soliton solution ( for detail see the appendix ) of * * * * eq .
( [ magnet2r ] ) @xmath93where @xmath94 , @xmath95 and @xmath96 @xmath97 are defined in eq .
( [ solitonr ] ) and @xmath98 @xmath99 take form as follows : @xmath100 , \notag \\
r_{5 } & = & \frac{1}{\lambda _ { 2}}(\chi _ { 2,2}\eta _ { 2}\sin \phi _ { 2}\sinh \theta _ { 2}-\chi _ { 1,2}\cosh \theta _ { 2}\cos \phi _ { 2 } ) , \notag \\ r_{6 } & = & \frac{-1}{\lambda _ { 2}}\allowbreak ( \chi _ { 2,2}\eta _ { 2}\sin \phi _ { 2}\cos \phi _
{ 2}+\chi _ { 1,2}\cosh \theta _ { 2}\sinh \theta _ { 2 } ) , \label{twosoliton2r}\end{aligned}\]]where @xmath101 , @xmath102 , @xmath103 , @xmath104 , @xmath105 and @xmath106 @xmath107 are defined in eq .
( [ para1r ] ) .
the solution ( [ twosoliton1r ] ) describes a general elastic scattering process of two solitary waves with different center velocities @xmath61 and @xmath108 , different phases @xmath68 and @xmath102 . before collision , they move towards each other , one with velocity @xmath61 and shape variation frequency @xmath84 and the other with @xmath109 and @xmath103 . in order to understand the nature of two - soliton interaction , we analyze the asymptotic behavior of two - soliton solution ( [ twosoliton1r ] ) .
asymptotically , the two - soliton waves ( [ twosoliton1r ] ) can be written as a combination of two one - soliton waves ( [ onesoliton ] ) with different amplitudes and phases .
the asymptotic form of two - soliton solution in limits @xmath110 and @xmath111 is similar to that of the one - soliton solution ( [ onesoliton ] ) . in order to analyze the asymptotic behavior of two - soliton solutions ( [ twosoliton1r ] ) we * * * * show first of all the asymptotic behavior of @xmath94 , @xmath95 , @xmath112 , and @xmath98(@xmath113 ) in the corresponding limits @xmath114 from eqs .
( [ solitonr ] ) and ( [ twosoliton2r ] ) @xmath115without loss of generality , we assume that @xmath116 @xmath117 and @xmath118 which corresponds to a head - on collision of the solitons . for the above parametric choice , the variables @xmath119(@xmath48 ) for the two - soliton behave asymptotically as ( i ) @xmath120 , @xmath121 , as @xmath114 ; and ( ii ) @xmath122 , @xmath123 , as @xmath124 .
this leads to the following asymptotic forms for the two - soliton solution .
( for the other choices of @xmath125 and @xmath126 , similar analysis can be performed straightforwardly ) .
\(i ) before collision , namely the case of limit @xmath110 .
\(a ) soliton 1 ( @xmath127 , @xmath128 ) .
@xmath129 where @xmath130 $ ] and the parameters @xmath131 and @xmath132 are defined in eq .
( [ polar1r ] ) .
\(b ) soliton 2 ( @xmath133 , @xmath134 ) .
@xmath135 \(ii ) after collision , namely the case of limit @xmath111 .
\(a ) soliton 1 ( @xmath127 , @xmath136 ) .
@xmath137 \(b ) soliton 2 ( @xmath138 , @xmath139 ) .
@xmath140 here for the expressions of solitons before and after collision , @xmath141 , @xmath142 and @xmath143 @xmath97 are defined in eq .
( [ solitonr ] ) .
analysis reveals that there is no amplitude exchange among three components @xmath144 , @xmath145 and @xmath146 for soliton 1 and soliton 2 during collision .
however , from eqs .
( [ asym1a ] ) and ( [ asym1b ] ) one can see that there is a phase exchange @xmath147 between two components @xmath145 and @xmath146 for soliton 1 during collision .
this elastic collision between two magnetic solitons in the optical lattice is different from that of coupled nonlinear schrdinger equations @xcite .
it shows that the information held in each soliton will almost not be disturbed by each other in soliton propagation .
these properties may have potential application in future quantum communication .
it should be noted that the inelastic collision may appear if the influence of higher - order terms in eq .
( hamilton2 ) is considered .
magnetic soliton dynamics of spinor becs in an optical lattice is studied in terms of a modified landau - lifshitz equation which is derived from the effective hamiltonian of a pseudospin chain .
the soliton solutions are obtained analytically and the elastic collision of two solitons is demonstrated .
the significant observation is that time - oscillation of the soliton amplitude and size can be controlled by adjusting of the light - induced dipole - dipole interactions .
it should be interesting to discuss how to create the magnetic soliton and how to detect such magnetic soliton in experiment . in the previous work miesner using landau - zener rf - sweeps at high fields ( 30 g ) a condensate was prepared in the hyperfine state @xmath148 , i.e. the the ground state of the spinor becs .
then the atoms of the ground state can be excited to the hyperfine state @xmath149 by laser light experimentally .
therefore the excited state of the spinor becs , i.e. the magnetic soliton can be created . as can be seen from fig .
1 , the spatial - temporal spin variations in the soliton state are significant .
this makes it possible to take a direct detection of the magnetic soliton of spinor becs . by counting the difference numbers of the population between the spin
@xmath150 and @xmath151 zeeman sublevel , the average of spin component @xmath152 is measured directly . while transverse components can be measured by use of a short magnetic pulse to rotate the transverse spin component to the longitudinal direction . any optical or magnetic method which can excite the internal transitions between the atomic zeeman sublevels
can be used for this purpose . in current experiments in optical lattices ,
the lattice number is in the range of @xmath153-@xmath154 , and each lattice site can accommodate a few thousand atoms .
this leads to a requirement for the frequency measurement precision of about 10 - 100 khz .
this is achievable with current techniques .
we can also see that the detection of the magnetic soliton of the spinor becs is different from that of the heisenberg spin chain .
the magnetic soliton of spinor becs in an optical lattice is mainly caused by the magnetic and the light - induced dipole - dipole interactions between different lattice sites .
since these long - range interactions are highly controllable the spinor becs in optical lattice which is an exceedingly clean system can serve as a test ground to study the static and dynamic aspects of soliton excitations .
this work was supported by the nsf of china under grant nos .
10475053 , 60490280 , 90406017 and provincial overseas scholar foundation of shanxi .
the corresponding lax equations for the eq .
( [ magnet2r ] ) are written as where @xmath156here @xmath157(@xmath158 ) is pauli matrix , the parameters @xmath159 and @xmath160 satisfy the relation @xmath161 .
( [ magnet2r ] ) can be recovered from the compatibility condition @xmath162=0 $ ] .
we introduce an auxiliary parameter @xmath163 such that @xmath164 * * * * and the complex parameter is defined by * * * * @xmath165 . we then have @xmath173where the overbar denotes complex conjugate
. suppose * * * * that @xmath174 is a simple pole of @xmath175 then @xmath176 is also a pole of @xmath170 .
if @xmath170 has only these two simple poles we have @xmath177where @xmath178 , @xmath179 , and @xmath180 are @xmath181 matrix independent of * * * * @xmath163 , the terms @xmath182 * * * * and * * * * @xmath183 * * * * are * * * * residues at * * * * @xmath174 * * * * and * * * * @xmath184 , respectively . from eq ( [ lax5 ] ) , * * * * we have @xmath185from eqs .
( [ lax2 ] ) and ( [ lax2c ] ) we see that @xmath186and hence we have @xmath187since @xmath188 , it has no poles .
then we obtain @xmath189which shows the degeneracy of @xmath179 .
one can write @xmath190 where the superscript @xmath191 means transpose . substituting this expression into ( [ darboux5 ] )
we obtain @xmath192where @xmath193to determine @xmath194 and @xmath195 we substitute ( [ lax3 ] ) into ( [ lax1 ] ) and take the limit @xmath196 and then obtain @xmath197because of the degeneracy of @xmath198 the second factor of the right - hand sides of eq .
( [ darboux8 ] ) , namely , ( @xmath199 must appear in the left - hand side in its original form and , hence , it is independent of @xmath20 and @xmath44 .
we simply let @xmath200here @xmath201 is a constant . hence , the darboux matrices @xmath202 have been determined recursively , except for @xmath203 by a simple algebraic procedure , * * * * it is seen that @xmath204 is always non - vanishing regardless of the values @xmath20 and @xmath205 this shows the regularity of @xmath179 and then @xmath206 . in the limit as @xmath207 , from eq .
( [ lax2a ] ) we have @xmath208and then from eq .
( [ darboux8 ] ) we obtain @xmath209 d_{n}^{\dag } \left ( 1\right ) , n=1,2,3 , .... \label{solution1}\]]considering the eq .
( [ darboux3 ] ) and eq .
( [ darboux4 ] ) we get @xmath210which shows that the matrix @xmath178 is diagonal with the help of the eq .
( [ darboux2 ] ) and @xmath211then we can write @xmath212 which is real and characterizes the rotation - angle of spin in the @xmath213-plane .
it is necessary to mention that @xmath214 may he dependent on @xmath20 and @xmath44 . to determine @xmath214
one must examine the lax equations carefully .
since @xmath215 denotes a rotation around the @xmath20-axis , it does not affect the value of @xmath216 . substituting ( [ darboux1a ] ) into ( [ darboux8 ] ) and taking the limits as @xmath217 and @xmath218 respectively ,
we obtain @xmath219comparing these two equations , we derive @xmath220the eq .
( [ darboux9 ] ) gives @xmath221where @xmath222here the parameters @xmath119 and @xmath223 are defined in eq .
( para1r ) . setting @xmath224 and substituting the eqs .
( [ darboux1a ] ) , ( darboux6 ) , ( [ darboux10 ] ) , and ( [ darboux11 ] ) into ( [ solution1 ] ) we can obtain the one - soliton solution ( [ onesoliton ] ) . setting @xmath225 and with the similar procedure
the expression of the two - soliton solution ( twosoliton1r ) is derived .
b. p. anderson and m. a. kasevich , science * 282 * , 1686 ( 1998 ) ; d. m. stamper - kurn , m. r. andrews , a. p. chikkatur , s. inouye , h .- j .
miesner , j. stenger , and w. ketterle , phys .
lett . * 80 * , 2027 ( 1998 ) .
z. d. li , l. li , w. m. liu , j. q. liang and t. ziman , phys .
e * 68 * , 036102 ( 2003 ) ; z. d. li , j. q. liang , l. li , w. m. liu , phys .
e * 69 * , 066611 ( 2004 ) ; z. d. li , l. li , j. q. liang , chin .
lett * 20 * , 39 ( 2003 ) ; chin .
lett * 21 * , 443 ( 2004 ) ; q. y. li , z. w. xie , l. li , z. d. li , and j. q. liang , annals of physics * 312 * , 128 ( 2004 ) .
the dynamics of a ferromagnet can be described by the landau - lifshitz equation @xmath226 , where @xmath227 is spin vector , and @xmath228 describes the magnetic anisotropy , namely , the easy axis case with @xmath229 and the easy - plane case with @xmath230 . for detail see the ref .
liu , b. wu , x. zhou , d.k .
campbell , s.t .
chui , and q. niu , phys .
b * 65 * , 172416 ( 2002 ) .
figure 1 the amplitude and size of soliton in eq .
( [ onesoliton ] ) vary periodically with time , where @xmath69 , @xmath70 , @xmath72 , @xmath73 , @xmath71 .
the unit for time @xmath44 is @xmath231 and @xmath46 for space @xmath20 .
figure 2 the magnetization @xmath232 ( the integral @xmath233 ) vary with time periodically , where @xmath69 , @xmath234 , @xmath72 , @xmath235 , @xmath71 .
the unit for time @xmath44 is @xmath45 and @xmath46 for space @xmath20**. * * | we study the magnetic soliton dynamics of spinor bose - einstein condensates in an optical lattice which results in an effective hamiltonian of anisotropic pseudospin chain . a modified landau - lifshitz equation is derived and exact magnetic soliton solutions are obtained analytically .
our results show that the time - oscillation of the soliton size can be controlled in practical experiment by adjusting of * * * * the light - induced dipole - dipole interaction .
moreover , the elastic collision of two solitons is investigated . |
the action evaluation , including the lagrange multipliers , is implemented directly in the discrete setting , which gives discrete variants of eq .
( [ eq.hamilton_rho_eom ] ) , along with the boundary conditions . in the numerical implementation ,
time and space are discretized , and @xmath4 is kept at points @xmath145 in 1d , and @xmath146 in 2d .
we start by describing the method in 2d , and then discuss the simplifications which occur in 1d . the action , eq .
( [ eq : action ] ) , is discretized as @xmath147 where @xmath148 is the value of @xmath37 associated with the time interval @xmath149 $ ] .
this allows for the time resolution to vary . for each @xmath150 separately , @xmath148 is evaluated as @xmath151 where @xmath152 .
@xmath153 corresponds to the bond connecting @xmath154 to @xmath155 ( and similarly for other half - integer indices ) .
let @xmath156 .
then @xmath153 is given by@xmath157 with @xmath158 , @xmath159 , and a similar expression for @xmath160 .
the currents @xmath161 are constrained to satisfy a discretized version of the continuity equation , eq .
( [ eq : conserve ] ) , @xmath162 where @xmath163 , and @xmath164 are the ( constant ) spacings in the @xmath165- and @xmath166-directions . to minimize the currents subject to eq .
( [ eq : kirchoff ] ) , we define @xmath167 .
differentiating @xmath168 with respect to the currents gives@xmath169 with a similar expression for @xmath170 .
this is a discrete variant of @xmath71 @xmath72 . substituting eq .
( [ eq : appendix_j ] ) into eq .
( [ eq : kirchoff ] ) , one obtains a linear set of equations for the @xmath69-variables , which corresponds to eq .
( [ eq.hamilton_rho_eom ] ) .
these are solved to find the @xmath69-variables .
note that on boundary sites eq .
( [ eq : kirchoff ] ) involves only three currents ( or two at corners of the lattice ) , which is equivalent to setting @xmath171 for @xmath172 outside the lattice .
this corresponds to the boundary conditions @xmath173 in the continuum .
given the @xmath69 values , the final expression for @xmath37 is obtained by combining eqs .
( [ eq : appendix_del_s]),([eq : appendix_sx ] ) and ( [ eq : appendix_j ] ) , and reads@xmath174{c}\sigma\left ( \rho_{i+1/2,j}\right ) \left ( \frac{\hat{\rho}_{i+1,j}-\hat{\rho}_{i , j}}{\delta x}\right ) ^{2}\\ + \sigma\left ( \rho_{i , j+1/2}\right ) \left ( \frac{\hat{\rho}_{i , j+1}-\hat{\rho}_{i , j}}{\delta
y}\right ) ^{2}\end{array } \right ] \ , \ ] ] which serves as the discrete analog of @xmath65 = \frac{1}{2}\int dtd\mathbf{x}\sigma\left ( \rho\right ) \left ( \mathbf{\nabla}\hat{\rho}\right ) ^{2}$ ] .
this concludes the evaluation of the action @xmath37 for a given @xmath4 .
this procedure is used as a building block in the optimization algorithm , where @xmath37 is evaluated for different histories @xmath6 , see the main text . in 1d
the above scheme is somewhat simplified .
of course , only terms in the @xmath165 direction appear .
the continuity eq .
( [ eq : kirchoff ] ) is now @xmath175 , so @xmath176 , where @xmath177 is independent of the position @xmath178 ( but may depend on time ) . summing over eq .
( [ eq : appendix_j ] ) , and using @xmath179 we find @xmath180 where @xmath177 is fixed by requiring that the boundary condition @xmath181 holds . as an additional tool to improve accuracy ,
it is possible to interpolate @xmath6 onto a finer grid in @xmath182 before evaluating the action .
this simple step improves accuracy and stability at low resolutions . in the example
presented below , we use this technique to double the time resolution .
the algorithm was tested in 2d against the model @xmath183 and @xmath77 .
this model is a particular case of the open boundary zero range process @xcite , and its large deviation is given by@xmath46 = \int d^{2}x\left ( \rho_{f}\ln\frac{\rho_{f}}{\bar{\rho}}+\bar{\rho}-\rho_{f}\right ) \ .\ ] ] @xmath184 $ ] was calculated for @xmath85 in fig .
[ fig : ni_2d_low_rez](a ) . fig .
[ fig : ni_2d_low_rez](b ) shows a comparison of the numerical method with the exact result at a relatively low resolution , with @xmath185 divisions in each space dimension and @xmath186 divisions in time , starting from @xmath187 .
the profiles were interpolated onto a grid with twice the time resolution before the action evaluation .
the relative error in @xmath184 $ ] was @xmath188 . | we study rare events in systems of diffusive fields driven out of equilibrium by the boundaries .
we present a numerical technique and use it to calculate the probabilities of rare events in one and two dimensions . using this technique ,
we show that the probability density of a slowly varying configuration can be captured with a small number of long wave - length modes . for a configuration which varies rapidly in space
this description can be complemented by a local equilibrium assumption . in many cases
the typical size of fluctuations in a physical system with @xmath0 degrees of freedom is of order @xmath1 .
larger fluctuations are rare , and their probability scales as @xmath2 \sim\exp\left ( -n\phi\left [ \rho\right ] \right ) \,$ ] , where @xmath3 is an intensive function of the state @xmath4 .
the function @xmath5 $ ] is known as the _ large - deviation _
_ function _ ( ldf ) and is of fundamental interest in statistical mechanics . in equilibrium systems , @xmath3 is equal to the free - energy density .
away from equilibrium , a simple expression for @xmath3 is in general not known , and it may be affected by details of the system s dynamics . besides its fundamental interest for non - equilibrium physics ,
the function @xmath3 is important in various applications , e.g. for calculating escape rates from metastable states , with applications ranging from chemistry and population dynamics to cosmology @xcite . in a non - equilibrium steady - state , to compute the probability of a rare - event , one must calculate the dynamics leading up to that event @xcite .
this is in general a difficult task , even more so for spatially extended systems , where only a handful of analytical solutions exist @xcite .
if a general understanding is to emerge , additional methods beyond exact solutions need to be considered .
indeed , recent years have seen a considerable effort to develop numerical techniques to calculate the ldf in a variety of systems @xcite . in this letter
, we study the ldf in bulk - conserving diffusive systems , which are driven out of equilibrium by the boundaries .
these describe a broad range of transport phenomena , including electronic systems , ionic conductors , and heat conduction @xcite .
we show that the ldf in such systems can be efficiently evaluated numerically for a _ general _ interacting system in one and two dimensions , giving us access to previously unavailable information .
this is done by searching for the most probable history @xmath6 of the conserved density function @xmath4 leading to a rare state @xmath7 .
importantly , using the numerical technique we show that for many non - trivial cases , the ldf of a slowly varying configuration @xmath7 can be calculated by considering only histories @xmath6 which are slowly varying in space , i.e. which are given by the sum of only a few long wave - length modes .
this implies that the long wave - length structure of the ldf can be understood using an _ effective finite - dimensional theory _
, instead of the full infinite dimensional one .
in addition , we find that a local equilibrium assumption can capture much of the short wave length structure .
this could suggest a simple framework to treat the ldf in these systems .
in bulk conserving diffusive systems , the conserved density @xmath8 , representing e.g. charge or energy density , is related to the current @xmath9 by@xmath10 where the current is given by @xmath11 @xmath12 is a density - dependent diffusivity function , while @xmath13 controls the size of the white noise @xmath14 , which in @xmath15 dimensions satisfies @xmath16 and @xmath17 .
the prefactor @xmath18 in the noise variance results from the fact that we have scaled distances by @xmath19 the system size , and time by @xmath20 . after this
rescaling the noise is small as a consequence of the coarse graining . @xmath21 and @xmath22 are related via a fluctuation - dissipation relation ( nyquist noise in electronic systems ) , which for particle systems reads @xmath23 where @xmath24 is the compressibility @xcite . here
we study a system on a domain @xmath25 connected to reservoirs which fix the density at the boundary @xmath26 , @xmath27 if the boundary density is not constant a current is induced through the system , driving it out of equilibrium .
the average , or most probable density profile for the system @xmath28 , is obtained by solving @xmath29 = 0 $ ] , with @xmath30 at the boundaries . in equilibrium
( i.e. when @xmath31 is constant ) , the steady - state probability of any other density profile @xmath32 is easy to obtain : the large deviation functional @xmath5 $ ] is then given by the free - energy , which is local in @xmath4 .
by contrast , the steady - state probability distribution away from equilibrium is notoriously hard to compute . in general
, it is known that despite the local nature of the dynamics , @xmath5 $ ] is non - local in @xmath4 , leading to generic long - range correlations @xcite .
analytical results for @xmath5 $ ] are known only for a few models of interacting systems , corresponding to specific choices of @xmath21 and @xmath22 , and almost exclusively in one dimension ( 1d ) .
the only known example in higher dimensions is the zero - range process , which admits a trivial product measure @xcite .
hence , despite the central role that @xmath5 $ ] plays in the understanding of non - equilibrium phenomena , little is known about its properties . to compute the large deviation for the model described above , we first note that the probability of a noise realization @xmath14 is gaussian , @xmath33 .
using this expression together with eq .
( [ eq : j_def ] ) , the probability of a history @xmath34 during time @xmath35 is @xmath36 , where the action @xmath37 is given by @xmath38 ^{2}}{2\sigma\left ( \rho\left ( \mathbf{x},t\right ) \right ) } \ .\label{eq : action}\ ] ] as the noise is small , the system spends most of the time close to @xmath39 , the unique fixed point of the diffusion equation at zero noise . to calculate the steady - state probability of a rare event , @xmath40 , we consider trajectories @xmath41 starting from @xmath28 in the distant past @xmath42 and ending at @xmath43 at @xmath44 . for large @xmath0 , its probability @xmath45 \right\ } $ ]
is given by @xcite@xmath46 = \inf_{\rho,\mathbf{j}}s\ , \label{eq : phi_min_s}\ ] ] where the infimum is over histories satisfying eq .
( [ eq : conserve ] ) , with initial and final conditions @xmath47 , @xmath48 , and the boundary conditions , eq .
( [ eq : boundaries ] ) .
we now describe a numerical method which utilizes this formulation to evaluate the large deviation @xmath49 $ ] . to calculate numerically the large deviation , we directly minimize the action , eq .
( [ eq : action ] ) .
we present a simple algorithm , which efficiently finds minima of the action for such systems .
the algorithm is based on starting with a problem where the solution is known and gradually modifying it while maintaining a minimizing solution . specifically , we start with a problem where the initial and final states are identical , @xmath50 . in the exact solution @xmath51 ; here we take a sufficiently early time @xmath52 , before any significant evolution has begun , and check for convergence .
eq .
( [ eq : action ] ) has a unique minimum to this problem , @xmath53 and @xmath54 , for which the action vanishes , hence @xmath55 = 0 $ ] .
we now gradually change the final condition : define a series of gradually changing profiles @xmath56 , @xmath57 , with @xmath58 , and @xmath59 .
we call the series @xmath60 the _ final - state trajectory_. the solution for the problem with @xmath61 is known . in the next iteration we solve the same problem as above , only with final conditions @xmath62 , using as an initial guess @xmath41 as obtained from the previous iteration .
this procedure is iterated until we reach the final condition @xmath63 .
standard algorithms can be used for the minimization at each step ; we have experimented with different algorithms , and the final results did not depend on this choice , but computational efficiency did . since @xmath37 is a sum of squares , non - linear least - squares methods are applicable and
were found to be efficient . within each iteration , we minimize the action @xmath64
= \inf_{\rho}\tilde{s}\left [ \rho\right ] $ ] , where @xmath65 \equiv\inf_{\mathbf{j}}s\left [ \rho , \mathbf{j}\right ] $ ] . to evaluate @xmath65 $ ]
we take into account the constraint eq .
( [ eq : conserve ] ) , by introducing a lagrange multiplier @xmath66 , and optimizing @xmath67 with respect to @xmath68 and @xmath69 .
@xmath70 gives @xmath71 @xmath72 , which together with eq .
( [ eq : conserve ] ) reads @xmath73 with boundary conditions @xmath74 @xcite .
this is a linear equation for @xmath66 in terms of @xmath8 , which can easily be solved numerically . substituting the expression for @xmath68 into eq .
( [ eq : action ] ) , we find that @xmath65 = \frac{1}{2}\int dtd\mathbf{x}\sigma\left ( \rho\right ) \left ( \mathbf{\nabla}\hat{\rho}\right ) ^{2}$ ] . in practice
this is carried out numerically by discretizing the equations , as is described in detail in the appendix .
for reference below , we note that on the minimal path @xmath75 also holds , which yields an equation of motion for @xmath69:@xmath76 note that the @xmath69 field is the momentum conjugate to @xmath4 in a hamiltonian formulation of the problem @xcite . in the context of boundary driven - diffusive systems numerical techniques
were used to study current large deviations ( and similar quantities ) @xcite .
algorithms have also been devised to calculate generating functions in related systems @xcite .
however , both of these quantities do not yield direct information on the probability density at a specific state . * * * * our algorithm directly minimizes the action , as do @xcite .
a key feature of our algorithm is the gradual change of the final state , which makes it both stable and insensitive to the choice of optimization algorithm .
the algorithm is easy to implement . as a further advantage ,
the algorithm is easily modified to use only a small number of modes , as explored below .
the numerics were tested against the known 1d and two dimensional ( 2d ) models for which analytical expressions for large deviation and the trajectory minimizing the action exist . as a first demonstration , we consider the simple symmetric exclusion process ( ssep ) in 1d @xcite .
the model describes a lattice gas with hard - core exclusion , and in the continuum limit leads to @xmath77 , @xmath78 with @xmath79 .
we take the domain @xmath80 , and the boundary conditions are @xmath81 . when @xmath82 the system is driven out of equilibrium .
the most probable state is @xmath83 .
fig .
[ fig : ssep_numerics ] shows an example of a path @xmath84 minimizing the action for the ssep , with a given final state @xmath85 . in this case
it is known @xcite that there is a unique local minimizer for the action , and indeed we find a single solution , independent of the different final state trajectories tested . shown is the trajectory from the initial state @xmath28 to the final state @xmath85 at different times , compared with the numerical solution , showing close agreement .
the inset shows the contribution to the large deviation ( the action integral , eq . ( [ eq : action ] ) ) , integrated up to time @xmath86 .
the numerics where carried out with @xmath87 space divisions and @xmath88 time divisions . in order to capture the time evolution more exactly , the size of the time intervals @xmath89 was taken to be a geometric series , where the last division is @xmath90 times smaller than the first .
the initial time was @xmath91 .
as is clear from the inset , contributions from earlier time are negligible .
the relative error in the large deviation is @xmath92 . by taking @xmath93 the error
is reduced to @xmath94 .
[ ptb ] ssep_numerics.eps we now show that the dynamics of the large deviations of slowly varying configurations can be _ captured by a small number of variables _ , which describe the long wave - length behavior .
this provides an effective large - scale description of @xmath3 in terms of a small number of degrees of freedom .
to this end we define a family of functions @xmath95 which span the function space , ordered such that @xmath96 is the slowest varying in space , followed by @xmath97 , etc .. we then consider an approximation in which the configurations leading to @xmath85 are restricted to be linear combinations of a finite number of the slowest - varying @xmath98,@xmath99 for @xmath100 this recovers the exact extremal solutions .
for finite @xmath101 , when @xmath85 is itself of the form of eq .
( [ eq.finite_sum ] ) the solutions give upper bounds for the exact value of @xmath102 $ ] , since the minimization is only on a subset all histories @xmath6 .
below we are interested in how well they approximate the exact solutions . a natural choice of the functions @xmath103 are the normal modes of the linearized hamilton evolution , eqs .
( [ eq.hamilton_rho_eom ] ) and ( [ eq.hamilton_rho_hat_eom ] ) , linearized around @xmath104 and @xmath105 @xcite .
these are obtained by substituting @xmath106 , @xmath107 and keeping only linear terms in @xmath108 and @xmath109 @xmath110 with boundary conditions @xmath111 .
the solution @xmath6 of these equations is the optimal trajectory of eq .
( [ eq : action ] ) for small fluctuations of @xmath4 around @xmath28 .
these equations admit two types of solutions .
in one type , @xmath112 vanish identically .
these solutions correspond to the zero - noise evolution , and do not satisfy the initial condition at @xmath113 ( except in the trivial case @xmath114 ) .
the other set of solutions is obtained by first solving eq .
( [ eq : modes_eom2 ] ) , which is an eigenvalue problem for @xmath112 , independently of @xmath98 .
the resulting @xmath115 are then substituted into eq .
( [ eq : modes_eom1 ] ) , and @xmath108 is solved for . as a convention
, we take all @xmath98 functions to be normalized with @xmath116 , and in 1d have a positive slope at @xmath117 . as a first example we return to the profile @xmath85 given in fig .
[ fig : ssep_numerics ] , which is of the form @xmath118 , where @xmath119 are the two lowest eigenvalues , @xmath120 , and @xmath121 , @xmath122 .
the modes @xmath98 , for @xmath123 are shown in fig .
[ ssep_1d_modes_err_evolution_v5](a ) .
we now minimize the action with histories constrained to be of the form of eq .
( [ eq.finite_sum ] ) , with @xmath124 .
fig .
[ ssep_1d_modes_err_evolution_v5](b ) shows the histories for @xmath125 ( dashed lines ) . even for @xmath125 ,
the large deviation is obtained to within 2% , see fig .
[ ssep_1d_modes_err_evolution_v5](c ) , suggesting that even at this level the system s behavior can be captured by an effective model with only _ two degrees of freedom_. such a small error is striking considering the highly non - linear nature of the problem ( as @xmath126 is far from @xmath28 ) , which is generically expected to mix higher modes in significant amounts . by comparison , a local equilibrium approximation ( where a space - dependent chemical potential is set to reproduce @xmath28 ) gives a relative error of 16% , whereas extending the linearized dynamics to the full evolution @xcite gives an error in the large deviation of 67% ( mostly due to the long - range gaussian corrections ) .
this highlights both the importance of non - linearities in this problem , and the success of the truncated approximation .
the mode approximation can also be used as a high precision numerical method . as shown in fig .
[ ssep_1d_modes_err_evolution_v5](c ) , for @xmath127 , the relative error is reduced to @xmath128 , well below the error obtained by a straightforward discretization of space .
indeed similar approaches have been used as numerical tools to improve accuracy in @xcite .
[ ptb ] ssep_1d_modes_err_evolution_v5.eps the low mode approximation is also useful in cases where an exact analytical solution does not exist .
for example , our method allows us to study two - dimensional ( 2d ) systems . in 2d analytic solutions
are only known for a set of models which exhibit no long - range correlations .
we used one such model , with @xmath77 and @xmath129 , corresponding to a model of non - interacting particles , as a benchmark for our method , and found agreement between the numerics and the exact solution ( see the appendix ) . in what follows we present results for an interacting system , the ssep in 2d , for which an exact expression for the large deviation is not known .
this exhibits the real power of the numerics . to show the generality of the low mode approximation
, we take a somewhat arbitrary choice of boundary conditions , @xmath130 , on the square domain @xmath131 ^{2}$ ] , see fig .
[ fig:2d_ssep_combined ] . the most probable density profile @xmath28 is shown in fig .
[ fig:2d_ssep_combined](a ) , and we present results for the profile @xmath85 shown in fig .
[ fig:2d_ssep_combined](b ) , which , as in the 1d discussion , is of the form @xmath132 , where @xmath133 are the lowest modes in this system , and @xmath134 .
similar results were obtained for other profiles .
fig .
[ fig:2d_ssep_combined](c ) shows the growth of the modes for @xmath135 .
the first two modes give the exact large deviation to within @xmath136 , as estimated by @xmath137 -\phi_{14}\left [ \rho_{f}\right ] \right ) /\phi_{14}\left [ \rho_{f}\right ] $ ] , see fig . [
fig:2d_ssep_combined](d ) .
once more , as in 1d , this means that the evolution is well described by a two - parameter space @xmath138 , despite the non - linear nature of the problem .
interestingly , a local equilibrium approximation gives a relatively low error of 1.2% .
[ ptb ] 2d_ssep_combined.eps we have shown so far that the large deviation function is well - approximated using only a few modes , provided that @xmath85 itself is a slowly - varying function of space .
here we discuss how to extend these results to cases where @xmath85 is not necessarily slowly varying , but has some high - mode content . recalling that the bulk behavior is governed by equilibrium dynamics
, one might expect that small , high - mode perturbations around the low mode behavior would be captured by a local equilibrium theory . in particular , define a local free - energy density @xmath139 . in equilibrium , when all boundary densities are equal , this is precisely the free - energy density .
we then expect @xmath46 -\phi\left [ \rho_{f , m}\right ] \simeq\int\left [ f_{eq,\bar{\rho}}\left ( \rho_{f}\right ) -f_{eq,\bar{\rho}}\left ( \rho _ { f , m}\right ) \right ] dx\ , \label{eq : local_eq}\ ] ] where @xmath140 is @xmath85 projected to the subspace spanned by @xmath141 . in other words , the error due to truncation of the high modes is approximately accounted for by a local equilibrium theory .
we now show that this is indeed the case in an example on the kipnis
marchioro presutti ( kmp ) model for heat transfer @xcite , whose continuum limit @xcite gives @xmath77 and @xmath142 ( similar results are obtained for other models ) . in fig .
[ fig : local_eq ] , we take @xmath85 of the form @xmath143 . for this profile , the lhs of eq .
( [ eq : local_eq ] ) for @xmath125 equals -0.03667 and the rhs equals -0.03623 . thus eq .
( [ eq : local_eq ] ) is satisfied with relative accuracy of 1.2% .
hence , for profiles with high mode content the effective low mode description can be corrected for using eq .
( [ eq : local_eq ] ) .
[ ptb ] local_eq.eps in summary , our findings suggest that the evolution leading to a smooth rare event is smooth : the continuous diffusion @xmath144 , makes an enduring high frequency perturbation highly unlikely .
indeed we show that even when the evolution leading to the rare event is restricted to profiles involving just a few modes , good quantitative agreement with the exact large deviation may be found .
this means that the large deviation is well - described in a space involving just a few variables .
finally , we note that in general the action may have more than a single local minimum , an effect which is well - known in finite - dimensional systems @xcite .
while this does not happen in the well - known models used above in our comparisons , it does in fact exist in other models , and can be studied using the numerical method described here by considering various different final - state trajectories .
for example , when two solutions are present , different choices of final - state trajectories will lead to the different local minima @xcite .
a detailed account of this issue and its physical implications is beyond the scope of the present work , and will be disscussed elsewhere @xcite .
99 p. g. bolhuis , d. chandler , c. dellago , p. l. geissler , annu .
rev .
phys .
chem .
53 , 291 - 318 ( 2002 ) a. kamenev and b. meerson , phys .
rev .
e. 77 , 061107 ( 2008 ) s. coleman , phys .
rev .
d 15 , 29292936 ( 1977 ) f. moss and p. v. e. mcclintock , ed . , noise in nonlinear dynamical systems , cambridge university press , cambridge ( 1989 ) d. g. luchinsky , p. v. e. mcclintock , and m. i. dykman , rep
. prog .
phys .
, 61(8):889 - 997 ( 1998 ) r. s. maier and d. l. stein , phys .
rev .
e , 48(2):931 - 938 ( 1993 ) r. graham and t. tl , phys . rev .
lett .
52 , 912 ( 1984 ) m.i .
dykman , m.m .
millonas and v.n .
smelyanskiy , phys .
lett .
a , 195 ( 1994 ) , 53 h. touchette , r. j. harris , large deviation approach to nonequilibrium systems , in r. klages , w. just , c. jarzynski ( eds ) , nonequilibrium statistical physics of small systems : fluctuation relations and beyond , wiley - vch , weinheim ( 2012 ) m. i. freidlinand a. d. wentzell , random perturbations of dynamical systems , springer - verlag ( 1984 ) b. derrida , j. stat .
mech .
p07023 ( 2007 ) c. giardin , j. kurchan , and l. peliti , phys .
rev .
lett .
96 , 120603 ( 2006 ) v. lecomte and j. tailleur , j. stat .
mech .
p03004 ( 2007 ) j. tailleur and v. lecomte , aip conf . proc .
1091 , 212 - 219 ( 2008 ) c. giardina , j. kurchan , v. lecomte , j. tailleur , j. stat .
mech .
45 4 ( 2011 ) d. passerone and m. parrinello , phys .
rev .
lett .
87 , 108302 ( 2001 ) .
w. e , w. ren and e. vanden - eijnden , commun .
pure appl . math .
57 , 637 ( 2004 ) .
x. wan , j. comp .
phys . 230
( 2011 ) 86698682 .
f. bouchet , j. laurie and o. zaboronski , j. phys .
conf .
ser . 318
( 2011 ) 022041 ya .
m. blanter and m. bttiker , physics reports , 336 , 1 - 2 ( 2000 ) , w. dieterich , p. fulde and i. peschel , adv . in phys .
29 , 527 - 605 ( 1980 ) c. kipnis , c. marchioro and e. presutti , j. stat .
phys .
27 65 ( 1982 ) c. kipnis and c. landim , scaling limits of interacting particle systems ( springer , new york , 1999 ) l. bertini , a. de sole , d. gabrielli , g. jona - lasinio , and c. landim , phys .
rev .
lett .
87 , 040601 ( 2001 ) l. bertini , a. de sole , d. gabrielli , g. jona - lasinio , and c. landim , j. stat .
phys .
107 , p07014 ( 2002 ) a. n. jordan , e. v. sukhorukov and s. pilgram , j. math .
phys .
45 4386 - 4417 ( 2004 ) j. tailleur , j. kurchan and v. lecomte , j. phys . a : math
. theor .
41 505001 ( 2008 ) p. i. hurtado and p. l. garrido , phys .
rev .
lett .
107 , 180601 ( 2011 ) p. i. hurtado et .
al . , proc .
natl .
acad .
sci .
usa .
108(19 ) 77047709 ( 2011 ) v. elgart and a. kamenev , phys . rev .
e 70 , 041106 ( 2004 ) g. bunin , y. kafri , d. podolsky , in preparation .
every history which locally minimizes the action will be found by a final state trajectory which follows this history .
h. spohn , j. phys .
a : math . gen
. 16 4275 ( 1983 ) b. derrida , j. l. lebowitz , and e. r. speer , j. stat .
phys .
107 775 - 810 ( 2002 ) other choices of modes , such as simple plane waves were also tested , and also give good results , but with slower convergence .
l. bertini , d. gabrielli and j. lebowitz , j. stat .
phys .
121 843 ( 2005 ) |
the new availability of wide - field images from schmidt telescopes in the 1940 s meant that astronomers no longer had to make educated guesses about where to look to find new and interesting phenomena but were now spoilt for choice .
the advent of synoptic surveys presents more extreme opportunities ; as an illustration , consider the sdss which over the course of 5 years represents a factor of a million increase in information over previous surveys ; however , the ( large sky synoptic telescope , tyson ( 2002 ) ) will amass a sdss every 3 nights . although overviews of synoptic surveys are riddled with cliches concerning undiscovered countries and uncharted waters , the exploration of the temporal domain results in data sets that are not just more voluminous than before , but far richer and more complex ( paczynski 2001 ; djorgovski et al .
this presents challenges to all aspects of astronomy : data gathering , distribution , reduction , analysis , storage , archiving , dissemination and mining .
vo technologies are being designed precisely to meet these types of challenges , but to use them requires changes in survey design philosophies .
the is a major new survey being undertaken by caltech , yale , jpl and indiana university employing the world s largest astronomical camera and the recently refurbished oschin schmidt telescope at palomar to observe a third of the sky ( @xmath1 sq . deg . between @xmath2 ) a minimum of 8 times in 7 passbands to nominally twice the depth of sdss .
the quest camera consists of 112 ccds arranged in four filter strips .
each ccd has 2400 @xmath5 600 13@xmath6 m @xmath5 13@xmath6 m pixels , giving a total of 161 @xmath5 10@xmath7 pixels . at the prime focus of the oschin schmidt ,
quest covers a sky area of 4.6@xmath8 3.6@xmath3 ( the effective area is @xmath9 sq .
deg ) and in a night can survey @xmath10 sq .
two filter sets are used : johnson @xmath11 and gunn @xmath12 , with a doubling of gunn @xmath13 to afford extra depth .
the data rate is 2.45mb / s and with a monthly average of 10 nights observing , quest produces @xmath4 tb of data / month .
some of the immediate science goals are searching for high redshift quasars , strong gravitational lensing , supernovae and grbs , and near - earth asteroids and trans - neptunian objects . obviously once there is a sufficient body of repeat observations , searching for new types of variable object and phenomena will play a dominant part ; in particular , a rapid response mechanism to transients ( see section 4 ) is planned .
as this survey is one of the first of the new breed of synoptic surveys , it is being used as a testbed for the vo technologies which will enable astronomers to exploit such surveys to the full .
there are currently four areas of attention : different groups want to process the raw data in different ways to optimize the detection of specific types of object .
access requirements to the data are also either near real - time or delayed .
data distribution must be secure , fault tolerant ( error checking , multiply redundant ) and accountable ( transaction logging ) .
the nature of the data is extremely well suited to parallelization , either on a multi - processor machine or in a more general distributed computing environment , e.g. an advanced highly cpu - intensive pipeline would be a suitable grid - level application .
the identification of variable objects poses many problems : * associating different observations under different conditions ( e.g. seeing ) with the same identification ; * handling objects which only appear once ( e.g. supernovae ) * handling moving objects ( e.g. asteroids ) * optimally characterizing the variability of an object ( periodic / aperiodic ) * determining the best sampling strategy to maximize the range of temporal baselines covered other federated data sets will be employed in the data analysis to assist identification , e.g. sdss , dposs , 2mass .
the deployment of quest as a federated data set needs to support both interactive and batch mode access .
access to data products also needs to be transparent to the access rights of different users : quest survey team , collaborators and the general astronomy community .
to illustrate how quest will make use of vo technologies in an integrated fashion , consider one of the pipeline systems under construction ( see fig . 1 for a cartoon depiction ) :
this will produce real time ( within four minutes of the data being taken ) alerts of transient events ( e.g. supernovae ) .
the specific processes which need to mesh are : * _ distribution : _ every 140s , 112 @xmath5 3.1 mb raw fits files are produced at palomar and streamed to caltech ( at 10mb / s ) where the cit data broker distributes the data to other sites , the raw image archive and the reduction pipeline * _ processing : _ the cit fast pipeline computes a real - time flat and extracts objects - each field produces @xmath1410000 objects * _ analysis : _ variable and transient objects are detected by comparing the latest observations with the fiducial sky ( composed from all quest observations and possibly other data sets ) in the master archive . they are processed to determine whether they might be asteroids and checked against lists of known variables . source classification is attempted using other federated data archives . * _ dissemination : _ the alert decision engine decides whether an alert should be issued based on decision algorithms and all available data and posts results to the website .
palomar - quest is the prototype vo - integrated synoptic sky survey and marks the beginning of an exciting new era in astronomy : the characterization of the variable optical sky .
djorgovski , s. g. et al .
2001 , in virtual observatories of the future , ed .
r. j. brunner , s. g. djorgovski & a. s. szalay ( asp conf .
225 ) , 52 paczynski , b. 2000 , , 112 , 1281 tyson , j. a. 2002 , spie , 4836 , 10 | the advent of wide - area multicolour synoptic sky surveys is leading to data sets unprecedented in size , complexity and data throughput . vo technology offers a way to exploit these to the full but requires changes in design philosophy .
the palomar - quest survey is a major new survey being undertaken by caltech , yale , jpl and indiana university to repeatedly observe @xmath0 of the sky ( @xmath1 sq
. deg . between @xmath2 ) in seven passbands . utilising the 48-inch oschin schmidt telescope at the palomar observatory with the 112-ccd quest camera covering the full 4@xmath3 x 4@xmath3 field of view
, it will generate @xmath4 tb of data per month . in this paper
, we review the design of quest as a vo resource , a federated data set and an exemplar of vo standards . |
because of their large masses , charm quarks are produced very early and propagate through the quark - gluon plasma formed in relativistic heavy ion collisions .
any modifications of charm quark spectrum thus carry information on the properties of the quark - gluon plasma .
although charmed hadrons are at present not directly observable in central nucleus - nucleus collisions at the relativistic heavy ion collider ( rhic ) , experimental data on the transverse momentum spectrum of electrons from their decays have already provided useful information on the interaction of charm quarks in the quark - gluon plasma .
for example , the transverse momentum spectrum of these electrons is found to be insensitive to the charm final - state interactions as results from both the pythia model and the blastwave hydrodynamic model are consistent with the experimental data @xcite . on the other hand ,
the large elliptic flow of these electrons is consistent with the prediction of the coalescence model @xcite which assumes that charm and light quarks are in thermal equilibrium and have same elliptic flow . in the present talk
, we discuss in the framework of the ampt model @xcite the mechanism for the generation of charm quark elliptic flow and the dependence of its value on the charm scattering cross section in the quark - gluon plasma @xcite .
the ampt model has four components : initial conditions , parton cascade , hadronization , and hadron cascade . for studying charm elliptic flow
, we use the version with string melting , in which hadrons that are generated from the hijing model @xcite are converted to partons according to their valence structures with a formation time that is determined by the transverse momentum of the parent hadron , in order to simulate the evolution of the energy stored in initial strings and the effects of particle production from the coherent color field .
the space - time evolution of resulting partonic system is modeled by the zpc model @xcite , which includes elastic scatterings between partons with a cross section given by the leading pqcd and regulated by a screening mass that is taken as a parameter to adjust the magnitude of the cross section .
after the partonic system freezes out , closest quarks and anti - quarks are recombined into hadrons with their subsequent evolution simulated by the art model @xcite .
using parton cross sections @xmath0-@xmath1 mb , the ampt model with string melting can give a good description of measured low transverse momentum particle spectrum @xcite , elliptic flow @xcite , higher - order anisotropic flows @xcite , and the pion interferometry @xcite .
charmed particles are rare particles even at rhic as only about two pairs are produced in the mid - rapidity region of central au+au collisions at available top energies .
to simulate charm particles efficiently , we use the perturbative method @xcite by introducing an enhancement factor for their production from initial hard scattering , so that each charmed particle carries a probability given by the inverse of corresponding enhancement factor , and neglects the effects due to charmed particle scattering on un - charmed particles . using the power law parametrization of d meson spectrum measured in d+au collisions by the star collaboration @xcite
, we first generate the transverse momentum distribution of d mesons between rapidity of -2 and + 2 with their distribution in the transverse plane according to the positions of initial nucleon - nucleon collisions .
the initial phase - space distribution of charm quarks is then obtained by dissociating d mesons into their valence quarks after a formation time given by inverse of the d meson transverse momentum . the charm scattering cross section with other partons in the quark - gluon plasma
is taken to be the same as the cross section for collisions between light quarks .
both cross sections of 3 mb , which is similar to that given by the perturbative qcd , and 10 mb that is needed for describing observables related to light quarks , are used in present study . at the freeze - out of partons ,
charm quarks are combined with light quarks into d mesons according to the coordinate - space coalescence model used in the ampt model . for the scattering of charmed mesons with other hadrons ,
their cross sections are simply taken to be the same as the charm - parton cross section as their effect on the charmed meson spectrum is insignificant .
from their transverse momentum spectra at mid - rapidity , charm quarks are found to approach thermal equilibrium when their scattering cross sections increase .
this result is reminiscent of the transition from the charm spectrum obtained from the pythia to that of the blastwave hydrodynamic model . as in ref.@xcite ,
the spectrum of electrons from the decay of final charmed mesons is not very sensitive to charm final - state interactions and is compatible with experimental data @xcite for both cross sections . for elliptic flows ,
results from the ampt model are shown in fig .
[ fig : v2pt2 ] .
it is seen from panels ( a ) and ( b ) that the magnitude of charm quark elliptic flow increases with increasing parton scattering cross section . for both parton cross sections ,
there is , however , a strong mass ordering of quark elliptic flows with charm quark elliptic flow saturating to about the same value at a larger transverse momentum compared with that of light quarks .
the elliptic flows of d mesons and their decay electrons are shown in panels ( c ) and ( d ) .
both are seen to follow essentially that of charm quarks as the momentum of a charmed meson is largely given by that of charm quark when light quarks have only bare masses as in the ampt model
. a larger charmed meson elliptic flow of about 10% at @xmath2 gev / c is obtained if an isotropic instead of screened coulomb cross section is used for charm quark scattering as suggested recently in ref.@xcite that charm quark scattering is dominated by the formation of charmed meson resonances in the quark - gluon plasma . to explain the large charm elliptic flow of more than 10% in the available data @xcite thus requires effects beyond the perturbative qcd in the quark - gluon plasma .
we have not included in the present study electrons from the decay of mesons consisting of bottom quark .
their contribution is expected to become important at @xmath3 gev / c . also , the effect of radiative energy loss on charm quark elliptic flow , which becomes non - negligible as the charm quark transverse momentum increases , is not considered . for charm quarks with high transverse momentum
, charm meson production will be mainly from fragmentation instead of recombination .
these effects need to be included for a more complete study of heavy quark elliptic flow at high transverse momentum .
our results are , nevertheless , consistent with other recent studies of charm collective flow @xcite , i.e. , charm elliptic flow is sensitive to charm final - state interactions and a large elliptic flow seen in available data requires a charm scattering cross section larger than that given by the perturbative qcd .
j. adams _ et al . _ [ star collaboration ] , phys . rev
. lett . 94 ( 2005 ) 062301 .
a. tai [ star collaboration ] , j. phys .
g 30 ( 2004 ) s809 .
l. ruan [ star collaboration ] , j. phys .
g 30 ( 2004 ) s1197 .
adler _ et al .
_ [ phenix collaboration ] , phys .
lett . 94 ( 2005 ) 082301 . h. van hees and r. rapp , phys .
c 71 ( 2005 ) 034907 .
adler [ phenix collaboration ] , phys .
c 72 ( 2005 ) 024901
. f. laue [ star collaboration ] , j. phys .
g 31 ( 2005 ) s27 .
l.w . chen and c.m .
ko , j. phys .
g 31 ( 2005 ) s49 .
d. molnar , j. phys .
g 31 ( 2005 ) s421 .
bratkovskaya , w. cassing , h. stocker and n. xu , phys .
c 71 ( 2005 ) 044901 .
moore and d. teaney , phys .
c 71 ( 2005 ) 064904 .
m. djordjevic , m. gyulassy , r. vogt and s. wicks , arxiv : nucl - th/0507019 . | using the perturbative method for the simulation of charmed particles , the dynamical origin of charm quark elliptic flow is studied in the framework of a multi - phase transport ( ampt ) model . besides the expected ordering relative to that of light quarks according to quark masses , charm quark elliptic flow is seen to be sensitive to the parton scattering cross section . to describe the observed large elliptic flow of electrons from the decay of charmed mesons , a charm quark elastic scattering cross section much larger than that estimated from the perturbative qcd is required . |
the first indication of the existence of dark matter has already been found in the 1930s @xcite .
by now astrophysicists have strong evidence @xcite-@xcite to believe that a large fraction ( more than 80% ) of the matter in the universe is dark ( i.e. , interacts at most very weakly with electromagnetic radiation ) .
the dominant component of this cosmological dark matter must be due to some yet to be discovered , non baryonic particles . weakly interacting massive particles ( wimps ) @xmath0 are one of the leading candidates for dark matter .
wimps are stable particles which arise in several extensions of the standard model of electroweak interactions .
typically they are presumed to have masses between 10 gev and a few tev and interact with ordinary matter only weakly ( for reviews , see @xcite ) . currently , the most promising method to detect many different wimp candidates is the direct detection of the recoil energy deposited in a low background laboratory detector by elastic scattering of ambient wimps on the nuclei in a detector @xcite-@xcite .
the event rate of direct wimp detection depends strongly on the velocity distribution of the incident particles . usually and for simplicity
, the local velocity distribution of wimps is assumed to be a shifted maxwell distribution , as would arise if the milky way halo is isothermal @xcite-@xcite . however , our halo is certainly not a precisely isothermal sphere .
possibilities that have been considered in the literature include axisymmetric halo models @xcite , the so called secondary infall model of halo formation @xcite , and a possible bulk rotations of the halo of our galaxy @xcite .
if the halo of our galaxy consists of wimps , about @xmath1 wimps should pass through every square centimeter of the earth s ( and our ! ) surface per second ( for @xmath2 ) . however , the cross section of wimps on ordinary materials is very low and makes these interactions quite rare @xcite .
for example , in typical susy models with neutralino wimp , the event rate is about @xmath3 event / kg / day and the energy deposited in the detector by a single interaction is about @xmath4 kev .
typical event rates due to cosmic rays and ambient radioactivity are much larger .
the annual modulation of the event rate due to the orbital motion of the earth around the sun has been suggested as a way to discriminate signal from background @xcite .
actually , the dama collaboration has claimed that they have observed this annual modulation of the event rate @xcite .
note , however , that the annual modulation of the signal is expected to amount only to a few percent ; this method can therefore only be used once more than one hundred signal events have been accumulated . in the meantime , a more promising approach is to reduce the background by vetoing events that do not look like nuclear recoil .
this method is , e.g. , being used by the cdms @xcite , cresst @xcite and edelweiss @xcite collaborations .
the presently best null result , from cdms @xcite , contradicts the dama claim for standard halo models .
gev , and fast wimps with large scattering cross section @xcite .
another possible way out is to postulate that the detected events are actually inelastic , leading to the production of a second particle that is almost , but not exactly , degenerate with the wimp @xcite . ]
so far most theoretical analyses of direct wimp detection have predicted the detection rate for a given ( class of ) wimp(s ) , based on a specific model of the galactic halo .
the goal of this paper is to invert this process .
that is , we wish to study , as model independently as possible , what future direct detection experiments can teach us about the wimp halo . in other words
, we want to start the ( theoretical ) exploration of `` wimp astronomy '' . in this first study
we use a time averaged recoil spectrum , and assume that no directional information exists .
we can thus only hope to construct the ( time - averaged ) one dimensional velocity distribution @xmath5 , where @xmath6 is the absolute value of the wimp velocity in the earth rest frame .
note that our ansatz is quite different from that of the recent paper @xcite , which assumes a wimp velocity distribution and then analyses with which precision the wimp mass can be determined from the direct detection experiment .
the remainder of this article is organized as follows . in sec .
2 we show how to find the velocity distribution of wimps from the functional form of the recoil spectrum ; our assumption here is that this functional form has been determined by fitting the data of some ( future ) experiment(s ) .
we then derive formulae for moments of the velocity distribution function , such as the mean velocity and the velocity dispersion of wimps , which can be compared with model predictions .
we also discuss some simple halo models . in sec .
3 we will develop a method that allows to reconstruct the wimp velocity distribution function directly from recorded signal events .
this allows statistically meaningful tests of predicted distribution functions .
we will also show how to calculate the moments of the velocity distribution directly from these data . in sec .
4 we conclude our work and discuss some further projects . some technical details of our calculations are given in the appendices .
in this section we first show how to reconstruct ( moments of ) the wimp velocity distribution , and then discuss some simple model distributions .
the differential rate for elastic wimp nucleus scattering is given by @xcite : [ eqn201 ] = dv . here
@xmath7 is the direct detection event rate , i.e. , the number of events per unit time and unit mass of detector material , @xmath8 is the energy deposited in the detector , @xmath9 is the elastic nuclear form factor , and @xmath5 is the one dimensional velocity distribution function of the wimps impinging on the detector .
the constant coefficient @xmath10 is defined as [ eqn202 ] , where @xmath11 is the wimp density near the earth and @xmath12 is the total cross section ignoring the form fact suppression .
the reduced mass @xmath13 is defined by [ eqn203 ] m_r , where @xmath14 is the wimp mass and @xmath15 that of the target nucleus .
finally , @xmath16 is the minimal incoming velocity of incident wimps that can deposit the energy @xmath8 in the detector : [ eqn204 ] v_min = , where we define [ eqn205 ] . in eqs.([eqn201])([eqn205 ] ) we have assumed that the detector essentially only consists of nuclei of a single isotope . if the detector contains several different nuclei ( e.g. nai as in the dama detector ) , the right hand side ( rhs ) of eq.([eqn201 ] ) has to be replaced by a sum of terms , each term describing the contribution of one isotope . for simplicity , in the remainder of this article we will focus on mono
isotopic detectors . in this section
we wish to invert eq.([eqn201 ] ) , i.e. , we want to find an expression for the one dimensional velocity distribution function @xmath5 for given ( as yet only hypothetical ) measured recoil spectrum @xmath17 . to that end
, we first define = , i.e. , @xmath18 is the primitive of @xmath19 .
eq.([eqn201 ] ) can then be rewritten as [ eqn211 ] = dv = f_1(v ) - f_1(v_min ) . since wimps in today s universe
are quite slow , @xmath5 must vanish as @xmath6 approaches infinity , f_1(v ) 0 . hence |_v
this means that @xmath20 approaches a finite value .
differentiating both sides of eq.([eqn211 ] ) with respect to @xmath16 and using eq.([eqn204 ] ) , we find & = & - \ { _ q = v_min^2/^2 } + + & = & \{-2 q } _ q = v_min^2/^2 .
since this expression holds for arbitrary @xmath16 , we can write down the following result directly : [ en1 ] = = \{-2 q } .
the rhs of eq.([en1 ] ) depends on the as yet unknown constant @xmath10 .
recall , however , that @xmath21 is the _
normalized _ velocity distribution , i.e. , it is defined to satisfy [ eqn214 ] f_1(v ) ~ dv = 1 .
therefore , the normalized one dimensional velocity distribution function can be written as [ eqn212 ] f_1(v ) = , with normalization constant @xmath22 ( see appendix [ appn ] ) [ eqn213 ] = ^-1 .
note that the integral on the rhs of eq.([eqn213 ] ) starts at @xmath23 . in the next step
we wish to compute _ moments _ of the distribution function @xmath21 : [ en2 ] v^n = _
v_min(q_thre)^v^n f_1(v ) ~ dv . in eq.([en2 ] ) we have introduced a threshold energy @xmath24 .
this is needed experimentally , since at very low recoil energies , the signal is swamped by electronic noise .
moreover , we will later meet expressions that ( formally ) diverge as @xmath25 .
@xmath26 is calculated as in eq.([eqn204 ] ) . inserting eq.([eqn212 ] ) into eq.([en2 ] ) and integrating by parts , we find ( see appendix [ appn ] ) [ eqn215 ] v^n = , with [ en3 ] i_n(q_thre ) = _ q_thre^q^(n-1)/2 dq . physically , @xmath27 is the average wimp velocity , while @xmath28 gives the velocity dispersion . in the statistical sense
is given by @xmath29 .
] we emphasize that eqs.([eqn215 ] ) and ( [ en3 ] ) can be evaluated directly once the recoil spectrum is known ; knowledge of the functional form of @xmath5 is not required .
note that the first term in eq.([eqn215 ] ) vanishes for @xmath30 if @xmath31 . in the same limit ,
@xmath32 by virtue of eq.([eqn213 ] ) . on the other hand , as written in eqs.([eqn212 ] ) and ( [ eqn213 ] ) , the velocity distribution integrated over the experimentally accessible range of wimp velocities gives a value smaller than unity . using only quantities that can be measured in the presence of a nonvanishing energy threshold @xmath24
, we can replace eq.([eqn213 ] ) by [ en4 ] ( q_thre ) = ^-1 .
using @xmath33 in eq.([eqn212 ] ) ensures that the velocity distribution integrated over @xmath34 gives unity .
we emphasize again that our final results in eqs.([eqn212 ] ) and ( [ eqn215 ] ) are independent of the as yet unknown quantity @xmath10 defined in eq.([eqn202 ] ) .
they do , however , depend on the wimp mass @xmath14 through the coefficient @xmath35 defined in eq.([eqn205 ] ) .
this mass can be extracted from a single recoil spectrum only if one makes some assumptions about the velocity distribution @xmath5 . in the kind of model
independent analysis pursued here , @xmath14 has to be determined by requiring that the recoil spectra using two ( or more ) different target nuclei lead to the same distribution @xmath5 through eq.([eqn212 ] ) .
note that this can be done independent of the detailed particle physics model , which determine the value of @xmath12 for the two targets .
however , one will need to know both form factors , which strongly depend on whether spin dependent or spin independent interactions dominate @xcite . within a given particle physics model ,
the best determination of @xmath14 should eventually come from experiments at high energy colliders . the simplest semi realistic model halo is a maxwellian halo .
the normalized one dimensional velocity distribution function in the rest frame of our galaxy is then @xcite [ eqn221 ] f_1,(v ) = e^-v^2/v_0 ^ 2 , where @xmath36 is the orbital speed of the sun around the galactic center , which characterizes the velocity of all virialized objects in the solar vicinity . on the other hand ,
when we take into account the orbital motion of the solar system around the galaxy , as well as the orbit of the earth around the sun , the distribution function must be modified to @xcite [ eqn222 ] f_1,(v , v_e ) = . here
[ eqn223 ] v_e(t ) = v_0 , where @xmath37 june 2nd is the date on which the velocity of the earth relative to the wimp halo is maximal @xcite .
eq.([eqn223 ] ) includes the effect of the rotation of the earth around the sun ( second term ) , but does not allow for the possibility that the halo itself might rotate . substituting these two expressions into eq.([eqn201 ] )
, one easily obtains the corresponding recoil energy spectra that wimp direct detection would measure @xcite : [ eqn224a ] _
= e^-^2 q / v_0 ^ 2 , and [ eqn225a ] _ = . here
@xmath38 is the error function , defined as ( x ) = _
0^x e^-t^2 dt .
hence [ eqn224b ] _
e^-^2 q / v_0 ^ 2 , and [ eqn225b ] _ - . for future reference
we note that the ( unrealistic ) `` reduced '' spectrum ( i.e. , the recoil spectrum divided by the squared nuclear form factor ) in eq.([eqn224b ] ) is exactly exponential ; this remains approximately true for the potentially quasi
realistic spectrum in eq.([eqn225b ] ) as well . in order to test our formulae , we calculated @xmath39 and @xmath40 , first from the normalized distribution functions in eqs.([eqn221 ] ) and ( [ eqn222 ] ) , [ eqn226a ] _ = v_0 , [ eqn226b ] _ = v_0 ^ 2 , and [ eqn227a ]
_ = e^-v_e^2/v_0 ^ 2 + ( + v_e ) , [ eqn227b ] _
= v_0 ^ 2+v_e^2 .
then we used the spectra of eqs.([eqn224b ] ) and ( [ eqn225b ] ) in eqs.([eqn212 ] ) , ( [ eqn213 ] ) and ( [ eqn215 ] ) , taking @xmath41 .
they reproduced the normalized velocity distribution functions in eqs.([eqn221 ] ) and ( [ eqn222 ] ) , as well as the results in eqs.([eqn226a ] ) to ( [ eqn227b ] ) .
in the previous section we have derived formulae for the normalized one dimensional velocity distribution function of wimps , @xmath5 , and for its moments @xmath42 , from the recoil energy spectrum , @xmath17 . in order to use these expressions
, one would need a functional form for @xmath17 . in practice
this might result from a fit to experimental data .
note that our expression for @xmath5 in eq.([eqn212 ] ) requires knowledge not only of @xmath17 , but also of its derivative with respect to @xmath8 , i.e. , we need to know both the spectrum and its slope .
this will complicate the error analysis considerably , if @xmath17 is the result of a fit . in this section
we therefore go one step further , and derive expressions that allow to reconstruct @xmath5 and its moments _ directly from the data_. the assumption we have to make is that the sample to be analyzed only contains signal events , i.e. , is free of background .
this should be possible in principle : the most copious backgrounds ( from radioactive @xmath43 and @xmath44 decays ) can be discriminated on an event by event basis in many modern wimp detectors .
the remaining background is dominated by elastic scattering of fast neutrons . while these look just like signal events , this background can at least in principle
be made arbitrarily small by careful shielding and the use of a muon veto system ( to veto cosmic ray induced neutrons ) . having a sample of pure signal events , we can proceed with a complete statistical analysis of the precision with which we can reconstruct @xmath5 and its moments . in the absence of a true experimental sample of this kind
, we had to resort to monte carlo experiments .
to that end we wrote an unweighted event generator .
to do so , we had to specify the form factor @xmath9 appearing in eq.([eqn201 ] ) ; this is the topic of the first subsection .
we then proceed to analyze the reconstruction of @xmath5 and its moments in two subsequent subsections .
we start by presenting the two most commonly used parameterizations of the squared nuclear form factor @xmath45 appearing in eq.([eqn201 ] ) .
we focus on spin independent scattering , which usually dominates the event rate ; moreover , spin dependent form factors are still only poorly understood .
the simplest form factor is the exponential one , first introduced by ahlen _
_ @xcite and freese _ et al .
_ @xcite : [ eqn301]f_ex^2(q ) = e^-q / q_0 , where @xmath8 is the recoil energy transferred from the incident wimp to the target nucleus , q_0 = is the nuclear coherence energy and r_0 = fm is the radius of the nucleus .
the exponential form factor implies that the radial density profile of the nucleus has a gaussian form .
this gaussian density profile is simple , but not very realistic .
engel has therefore suggested a more accurate form factor @xcite , inspired by the woods
saxon nuclear density profile , [ eqn302]f_ws^2(q ) = ^2 e^-(q s)^2 .
here @xmath46 is a spherical bessel function , q = is the transferred 3momentum , and r_1 = with r_a 1.2 ~ a^1/3 fm , s 1 fm . in our simulation
we used the more accurate woods
saxon form factor in eq.([eqn302 ] ) .
since we assume a detector without directional sensitivity , a single event is uniquely characterized by the measured recoil energy @xmath8 .
existing experiments such as cresst @xcite and cdms @xcite can determine the recoil energy quite accurately .
we will see shortly that the statistical errors on the reconstructed velocity distribution @xmath5 will be quite sizable even for next generation experiments , given existing bounds on the scattering rate .
we can therefore to good approximation ignore the error of @xmath8 in our analysis . in the following
we do not distinguish between the recoil spectrum @xmath17 and the actual differential counting rate @xmath47 .
since @xmath17 is usually given per unit detector weight and unit time , the two quantities differ by a multiplicative constant .
this constant cancels in eq.([eqn212 ] ) , since it will also appear in the normalization constant ( [ eqn213 ] ) .
we divide the total energy range into @xmath48 bins with central points @xmath49 and widths @xmath50 . in each bin
, @xmath51 events will be recorded .
our simulated data set can therefore be described by [ e3_0 ] q_n- , i = 1 , 2 , , n_n , n = 1 , 2 , , b. the standard estimate for @xmath17 at @xmath52 is then and for the ideal recoil spectrum . ]
[ e3_1 ] r_n _ q = q_n = , n = 1 , 2 , , b. the squared statistical error on @xmath17 is accordingly [ e3_2 ] ^2 ( r_n ) = . as noted earlier , we also need the _ slope _ of the recoil spectrum to reconstruct the velocity distribution ; see eq.([eqn212 ] ) .
a rather crude estimator of this slope is [ e3_3 ] s_1,n _ q = q_n = , where @xmath53 and @xmath54 are the numbers of events in bin @xmath55 which have measured recoil energy @xmath8 larger and smaller than @xmath49 , respectively .
this estimator is rather crude , since it only uses the information in which half of its bin a given event falls .
it is clear intuitively that an estimator that makes use of the exact @xmath56value of each event should be better .
this can e.g. be obtained from the average @xmath56value in a given bin , defined as [ e3_4 ] q_n = _
i=1^n_n q_n , i .
expanding @xmath57 around @xmath58 , keeping terms up to linear order , gives [ e3_5 ] _ q q_n _ q = q_n + ( q - q_n ) = r_n + ( q - q_n ) s_n .
eq.([e3_5 ] ) allows to calculate @xmath59 as expectation value of @xmath8 in the @xmath60th bin : [ e3_6 ] q_n - q_n = = .
an improved estimator of the slope is thus [ e3_7 ] s_2,n = .
a simple calculation shows that the estimator ( [ e3_7 ] ) indeed has a smaller statistical error than the one in eq.([e3_3 ] ) .
the definition ( [ e3_3 ] ) immediately implies [ e3_8 ] ^2(s_1,n ) = = , where we have used eqs.([e3_1 ] ) and ( [ e3_2 ] ) . in order to calculate the statistical error of the estimator @xmath61
, we first have to compute [ e3_9 ] _
n = = . treating the number of events and the average @xmath56value in a given bin as independent variables and using . ]
^2 = , this yields [ e3_10 ] ^2(s_2,n ) = ( ) ^2 ^2(r_n ) + ( ) ^2 ^2 = .
this is smaller than the error ( [ e3_8 ] ) , by a factor 3/4 .
an important observation is that the statistical error of both estimators of the slope of the recoil spectrum scale like the bin width to the power @xmath62 .
this can intuitively be understood from the argument that the variation of @xmath17 , which we are trying to determine , will be larger for larger bins .
one would therefore naively conclude that the errors of the estimated slopes would be minimized by choosing very large bins .
however , both eq.([e3_3 ] ) and eq.([e3_7 ] ) reproduce the actual slope at @xmath58 only if the taylor expansion ( [ e3_5 ] ) holds ; in the presence of higher powers of @xmath63 neither of these estimates exactly reproduces the true slope at @xmath58 .
not surprisingly , the influence of these higher powers , which induce uncontrolled systematic errors , will _ increase _ quickly with increasing bin width @xmath64 .
we had seen at the end of sec . 2 that the predicted recoil spectrum resembles a falling exponential .
this is confirmed in fig .
[ fig301 ] , which shows the predicted recoil spectrum of a 100 gev wimp on @xmath65ge , using the woods
saxon form factor ( [ eqn302 ] ) and the `` shifted maxwellian '' velocity distribution of eq.([eqn222 ] ) ; as usual , we cut the velocity distribution off at a velocity @xmath66 , here taken to be 700 km / s , since wimps with @xmath67 can escape the gravitational well of our galaxy .
this figure also shows the result of a simulated experiment , where the exposure time and cross section are chosen such that the expected number of events is 5,000 ; these have been collected in seven bins in recoil energy . while an approximately exponential function can be approximated by a linear ansatz , as in eq.([e3_5 ] ) , only over a narrow range of @xmath8 , i.e. , for small bin widths , the _ logarithm _ of this function can be approximated by a linear ansatz for much wider bins .
this corresponds to the ansatz [ eqn311 ] _ n _ q q_n _
n e^(q - q_n ) e^(q - q_s , n ) .
here @xmath68 is the recoil spectrum at the point @xmath52 , _
n_q = q_n , while @xmath69 is the _ logarithmic slope _ of the recoil spectrum in the @xmath60th bin .
our next task is to find estimators for @xmath68 and @xmath69 using ( simulated ) data .
note that for @xmath70 , @xmath68 can not be estimated from the number of events @xmath51 in the @xmath60th bin alone .
instead , one has [ eqn312 ] n_n = _ n dq = b_n _ n , where we have introduced the dimensionless quantities [ e3_11 ] x_n . hence ,
_ n= [ eqn313 ] depends on @xmath69 . on the other hand , the second , equivalent expression in eq.([eqn311 ] ) still uses the quantities @xmath71 as normalization .
evidently , they describe the spectrum @xmath17 at the shifted points @xmath72 .
equivalence of the two expressions in eq.([eqn311 ] ) implies [ e3_12 ] q_s , n = q_n + ( ) . the second expression in eq.([eqn311 ] ) thus has the advantage that the prefactor @xmath73 can be computed more easily than @xmath68 ; on the other hand , while the @xmath49 are simply the midpoints of the @xmath60th bin , and can thus be chosen at will , the @xmath72 are derived quantities ; they depend on the logarithmic slopes @xmath69 , which we havent determined yet .
to do so , we again use the average @xmath56value in the @xmath60th bin . from eq.([eqn311 ] ) we find : [ eqn314 ] q_n - q_n = = x_n - . unfortunately this expression can not be solved analytically for @xmath69 .
it is , however , straightforward to find @xmath69 numerically once @xmath59 is known .
alternatively , we can make use of the second moment of the recoil spectrum in the @xmath60th bin , defined as [ eqn315 ] _
n = = ^2 . multiplying both sides of eq.([eqn314 ] ) with @xmath74 and adding to eq.([eqn315 ] ) , we can calculate the logarithmic slopes as [ eqn316 ] = .
note that @xmath75 determined either from eq.([eqn314 ] ) or from eq.([eqn316 ] ) is independent of the normalization @xmath73 or @xmath68 . in the following
we will estimate the logarithmic slopes from eq.([eqn314 ] ) , since it simplifies the error analysis somewhat ; note that the statistical errors of @xmath59 and @xmath76 are correlated . using standard error propagation
, we have [ e3_13 ] ^2(k_n ) = ^2 ^-2 .
the error on the average energy transfer can be estimated directly from the data , using [ e3_14 ] ^2 = , where now @xmath76 is estimated from the data , analogously to the experimental definition of @xmath59 in eq.([e3_4 ] ) .
the second factor in eq.([e3_13 ] ) can be calculated straightforwardly from eq.([eqn314 ] ) : [ e3_15 ] ^-1 = , where we have defined the auxiliary function [ e3_15a ] f(x ) = 1 - ( ) ^2 . for given input values
@xmath77 and @xmath64 , eqs.([e3_13 ] ) and ( [ e3_15 ] ) also allow to calculate the expected statistical error of the estimated @xmath69 , using eq.([eqn315 ] ) to calculate the expected error of @xmath78 .
the result is shown in fig .
[ fig302 ] . by normalizing the bin width to the inverse slope , and the expected error to its value for a given bin width ,
the result becomes independent of @xmath73 , and can in fact be used for all combinations of @xmath69 and @xmath64 .
we observe that for small bins , the expected error again scales like @xmath79 , just as the expected errors ( [ e3_8 ] ) and ( [ e3_10 ] ) of our two estimators for the linear slope .
if the bin width is significantly larger than the absolute value of the inverse of the logarithmic slope , the error decreases even faster with increasing bin width . )
one might assume that the statistical error of the estimated values of the @xmath69 could be minimized by estimating them from the average values of @xmath80 $ ] , for some fixed value of @xmath81 .
for sufficiently small @xmath82 this in fact amounts to using the average value of @xmath8 , as described in the text .
increasing @xmath82 leads to slightly _ larger _ expected statistical errors . ]
this again argues in favor of using large bins .
however , we again have to consider systematic errors .
after all , it is quite unlikely that the ( as yet unknown ) recoil spectrum @xmath57 exactly satisfies our ansatz ( [ eqn311 ] ) over an extended range of @xmath8 .
rather , this ansatz should be considered as the first terms in a taylor expansion of the logarithm of @xmath57 . in this case the next order term , which contributes @xmath83 in the exponent , will already modify @xmath59 .
since we estimate @xmath69 from the numerical value of @xmath59 using eq.([eqn314 ] ) , which is exact only for @xmath84 , any non zero @xmath85 will introduce some systematic error in our estimate of @xmath69 . fortunately much of this error can be absorbed by a simple trick . according to ( [ eqn311 ] ) the logarithmic slope is constant over the entire bin , i.e.
, we could use @xmath69 extracted from eq.([eqn314 ] ) as estimate of the logarithmic slope at any point @xmath8 between @xmath86 and @xmath87 .
once @xmath88 the true logarithmic slope will in fact vary with @xmath8 .
however , one may hope that the expectation value of our estimator still reproduces the true logarithmic slope at _ some _ value of @xmath8 within the @xmath60th bin .
this is illustrated by fig .
[ fig303 ] , which shows various evaluations of the logarithmic slope within one bin as function of the quadratic coefficient @xmath89 .
the true logarithmic slope at the center of the bin is , of course , still given by @xmath90 , independent of the correction @xmath89 . as argued above
, the expectation value of our estimator , shown by the dashed ( red ) line , does depend on @xmath89 .
note , however , that our estimator comes quite close to the true logarithmic slope at the shifted value @xmath91 defined in eq.([e3_12 ] ) , which is shown by the dot dashed ( blue ) line ; this is true for both signs of @xmath89 .
we therefore conclude that we can minimize the leading systematic error by interpreting our estimator of @xmath69 as logarithmic slope of the recoil spectrum , not at the center of the bin @xmath49 , but at the shifted point @xmath72 .
note that @xmath72 itself depends on @xmath69 ; this , however , does not introduce any additional error , if we simply interpret eq.([e3_12 ] ) as an admittedly somewhat complicated prescription for the determination of the @xmath56values where we wish to estimate the logarithmic slope of the recoil spectrum . using large bins
has a second , obvious disadvantage : the number of bins scales inversely with their size , i.e. , by using large bins we d be able to estimate @xmath21 only at a small number of velocities .
this can be alleviated by using overlapping bins , or
equivalently by combining several relatively small bins into overlapping `` windows '' .
this means that a given data point @xmath92 may well contribute to several different windows , and hence to the measurement of @xmath21 at several values of @xmath6 .
this can increase the total amount of information about @xmath21 since the only information we use about the data points in a given window is encoded in the average recoil energy in this window .
this averaging effectively destroys information . by letting a given data point
contribute to several overlapping windows , this loss of information can be reduced .
a final disadvantage of using large bins or windows is that it would lead to a quite large minimal value of @xmath6 where @xmath21 can be reconstructed , simply because the central value @xmath93 , and also the shifted point @xmath94 , of a large first bin would be quite large .
this can be again be alleviated by first collecting our data in relatively small bins , and then combining varying numbers of bins into overlapping windows . in particular , the first window would be identical with the first bin .
a final consideration concerns the size of the bins . choosing fixed bin sizes , and
therefore also mostly fixed window sizes , would lead to errors on the estimated logarithmic slopes , and hence also on the estimates of @xmath21 , that increase quickly with increasing @xmath8 or @xmath6 .
this is due to the essentially exponential form of the recoil spectrum , which would lead to a quickly falling number of events in equal sized bins .
we found that we get roughly equal errors in all bins if we instead take linearly increasing bins .
these considerations motivate the following set up for our mock experimental analysis .
we start by binning the data , as in eq.([e3_0 ] ) , where the bin widths satisfy [ e3_16 ] b_n = b_1 + ( n-1 ) ; here the increment @xmath95 satisfies [ e3_17 ] = , @xmath48 being the total number of bins , and @xmath96 being the ( kinematical or instrumental ) extrema of the recoil energy .
we then collect up to @xmath97 bins into a window , with smaller windows at the borders of the range of @xmath8 . in the following we use latin indices @xmath98
to label bins , and greek indices @xmath99 to label windows ; later on we will use latin indices @xmath100 to label all events in the sample .
for @xmath101 the @xmath102th window simply consists of the first @xmath103 bins ; for @xmath104 , the @xmath102th window consists of bins @xmath105 ; and for @xmath106 , the @xmath102th window consists of last @xmath107 bins .
this can also be described by introducing the indices @xmath108 and @xmath109 which label the first and last bin contributing to the @xmath102th window , with [ e3_18 ] i_- = \ { ll 1 , & n_w + - n_w + 1 , & n_w .
, i_+ = \ { ll , & b + b , & b . , ( 1 b + n_w - 1 ) . the center of the @xmath110th bin is called @xmath111 , as before .
the total number of windows defined through eq.([e3_18 ] ) is evidently @xmath112 .
the basic observables needed for the reconstruction of @xmath21 are then the number of events @xmath113 in the @xmath110th bin , as well as the average @xmath114 defined as in eq.([e3_4 ] ) . from these
one easily calculates the number of events per window , [ e3_19 ] n_= _ i = i_-^i_+ n_i as well as the averages [ e3_20 ] q_= _ i = i_-^i_+ n_i q_i .
one drawback of the use of overlapping windows in the analysis is that the observables defined in eqs.([e3_19 ] ) and ( [ e3_20 ] ) are all correlated ( for @xmath115 ) .
the slope in a given window will again be calculated as in eq.([eqn314 ] ) , with `` bin '' quantities replaced by `` window '' quantities .
we thus need the covariance matrix for the @xmath116 , where @xmath117 is the midpoint of the @xmath102th window ; it follows directly from the definition ( [ e3_20 ] ) : [ e3_21 ] cov(q_- q _
, q_- q _ ) = _
i = i_-^i_+ , where @xmath118 is defined as in eq.([e3_14 ] ) . in eq.([e3_21 ] ) we have assumed @xmath119 ; the covariance matrix is , of course , symmetric .
moreover , the sum is understood to vanish if the two windows @xmath120 do not overlap , i.e. , if @xmath121 .
the ansatz ( [ eqn311 ] ) is now assumed to hold over an entire window .
we again estimate the prefactor as [ e3_22 ] r_= , @xmath122 being the width of the @xmath102th window .
this implies [ e3_23 ] cov(r _ , r _ ) = _
i = i_-^i_+ n_i , where we have again taken @xmath119 .
finally , the mixed covariance matrix is given by [ e3_24 ] cov(r _ , q_- q _ ) = _ i = i_-^i_+ n_i ( q_i - q _ ) .
matrix is not symmetric under the exchange of @xmath103 and @xmath123 . in the definition of @xmath124 and @xmath125 we therefore have to distinguish two cases : @xmath126 as before , the sum in eq.([e3_24 ] )
is understood to vanish if @xmath127 .
the covariance matrices involving our estimators of the logarithmic slopes @xmath128 , derived from eq.([eqn314 ] ) with @xmath129 everywhere , can be calculated in terms of the covariance matrices in eqs.([e3_21 ] ) and ( [ e3_24 ] ) : [ e3_26 ] cov(k _ , k _ ) = ( q_- q _ , q_- q _ ) , where @xmath130 is as in eq.([e3_11 ] ) with @xmath129 , and the function @xmath131 has been defined in eq.([e3_15a ] ) ; and [ e3_26a ] cov(r _ , k _ ) = ( r _ , _ - q _ ) .
we are now ready to put all pieces together to compute the reconstructed velocity distribution and its statistical error . inserting the ansatz ( [ eqn311 ] ) with the substitution @xmath129 into eq.([eqn212 ] ) , one finds the reconstructed velocity distribution [ e3_27 ] f_1,r(v _ ) = n . here , @xmath132 is given by eq.([e3_12 ] ) with @xmath129 , and [ e3_28 ] v_= , see eq.([eqn204 ] ) .
finally , the normalization @xmath133 defined in eq.([eqn213 ] ) can be estimated directly from the data : [ e3_29 ] n^-1 = _ a where the sum runs over all events in the sample . since neighboring windows overlap ,
the estimates of @xmath21 at adjacent values of @xmath134 are correlated .
this is described by the covariance matrix @xmath135 the covariance matrices involving the normalized counting rates @xmath136 and logarithmic slopes @xmath128 have been given in eqs.([e3_23 ] ) , ( [ e3_26 ] ) and ( [ e3_26a ] ) . in principle
eq.([e3_30 ] ) should also include contributions involving the statistical error of our estimator ( [ e3_29 ] ) for @xmath133 .
however , we find this error , and its correlations with the errors of the @xmath136 and @xmath128 , to be negligible compared to the errors included in eq.([e3_30 ] ) .
we are finally in a position to present some numerical results .
we first validate our results by presenting @xmath137 distributions , defined via [ e3_31 ] ^2_f _ , c _ . here
@xmath138 is our estimate ( [ e3_27 ] ) of the velocity distribution , @xmath21 is the true ( input ) distribution , and @xmath139 is the inverse of the covariance matrix of eq.([e3_30 ] ) .
we expect @xmath140 to be ( roughly ) distributed according to the standard @xmath141 distribution when the results of sufficiently many simulated experiments , with sufficiently many events per experiment , are analyzed . figs .
[ fig304 ] show @xmath137 distributions for 5,000 simulated experiments , with on average 500 ( top ) and 5,000 ( bottom ) events per experiments .
note that the actual number of events in a given simulated experiment varies according to the poisson distribution ; otherwise one would introduce an artificial correlation between the normalized counting rates @xmath142 in different bins .
moreover , the number of bins has been fixed a priori in these analyses .
the last bin is typically empty , and has therefore been ignored in the analysis .
this also reduces the number of windows used in the analysis by one , i.e. , the upper ( lower ) frame shows results for @xmath143 ( @xmath144 ) .
+ the two histograms in each figure differ by the number of terms that have been included in the estimate of the covariance matrix for @xmath138 .
the solid ( blue ) histograms have been obtained by only including the second term in eq.([e3_30 ] ) , while the dashed ( red ) histograms also include the other terms , which are due to the statistical errors on the rescaled event numbers @xmath136 .
we note that including these terms on average leads to a slight overestimate of the true error of @xmath138 , i.e. , the average of @xmath140 is somewhat smaller than unity .
this is partly due to the fact that we have ignored the error on the normalization @xmath133 , which is correlated quite strongly with the errors on the @xmath136 .
the lower figure demonstrates that for an average of 5,000 events per experiment the distribution of @xmath140 values becomes quite similar to the well known @xmath141 distribution , shown by the smooth curve .
at least two effects contribute to the difference .
first , we heavily relied on gaussian error propagation in our estimate ( [ e3_30 ] ) of the covariance matrix of the reconstructed velocity distribution .
this is essentially a taylor expansion , including only the first non trivial term .
it therefore becomes exact only in the limit of small errors , i.e. , for large numbers of events in a given window .
since the recoil spectrum is falling essentially exponentially , this condition is practically always violated at least in the last bin(s ) , see fig .
[ fig301 ] .
we discard windows containing less than 3 events , but it is clear that this at best alleviates the problem . in the case at hand , this evidently results in an overestimate of the true error . secondly , our estimator ( [ e3_27 ] ) for the velocity distribution relies on the estimate of the logarithmic slopes @xmath128 , which in turn is based on eq.([eqn314 ] ) . as illustrated in fig .
[ fig303 ] this estimate of @xmath128 in general has some systematic error , which would tend to increase @xmath140 .
however , this figure also led us to expect small systematic errors if @xmath128 is interpreted as estimator of the logarithmic slope at the shifted points @xmath132 . indeed , as stated above , the total expression ( [ e3_30 ] )
somewhat overestimates the true error even in the lower frame of figs .
[ fig304 ] , which assumes a large number of events but uses a rather small number of bins , which thus have to be quite large . had we instead interpreted @xmath128 as estimator of the slope at @xmath117 , the average @xmath140 would be about 2.9 , indicating that the systematic error would have dominated .
[ fig304 ] also show an excess of simulated experiments with rather large values of @xmath140 if the covariance matrix for @xmath138 is estimated based on the errors on the @xmath128 only .
this is true also for the upper frame , even though in this case the average value of @xmath140 is only about 0.93 . to be conservative , from now on
we therefore take the full eq.([e3_30 ] ) as our estimator of the covariance matrix of @xmath138 , leading to average @xmath145 for the upper ( lower ) frame of figs .
[ fig304 ] . in figs .
[ fig305 ] we show results for the reconstructed velocity distribution , for `` typical '' simulated experiments with 500 ( top ) and 5,000 ( bottom ) events . in the top frame
we choose @xmath146 bins , the first bin having a width @xmath147 kev , and combine up to three bins into a window .
since the last bin is in fact empty , this leaves us with @xmath143 windows , i.e. , we can determine @xmath21 for six discrete values of the wimp velocity @xmath6 ; recall that these `` measurements '' of @xmath21 are correlated , as indicated by the horizontal bars in the figure . in the lower frame
we choose @xmath148 bins with @xmath149 kev , and combine up to four bins into one window .
the bins are thus significantly smaller than in the upper frame . as a result ,
the last two bins are now ( almost ) empty , leaving us with @xmath150 windows . figs . [ fig305 ] indicate that one will need at least a few hundred events for a meaningful direct reconstruction of @xmath21 . recall that @xmath21 is normalized to unity .
the overall magnitude of @xmath21 is therefore essentially fixed by the range of observed @xmath56values ; only the _ shape _ of this distribution then remains to be determined .
one measure of the information content of the reconstructed @xmath138 is therefore the confidence level with which one can exclude a constant @xmath21 . + in fig . [ fig306 ]
we show one minus this confidence level , i.e. , the probability that a reconstructed velocity distribution is compatible with a constant .
this has been estimated by defining a @xmath140 variable as in eq.([e3_31 ] ) for the hypothesis @xmath151 const . , and integrating the theoretical @xmath141 distribution over the range @xmath152 . here
the constant has been chosen as @xmath153 , where @xmath154 have been calculated as in eq.([eqn204 ] ) using the largest and smallest recoil energy , respectively , that has been measured in a given experiment . since this probability differs quite widely from one ( simulated ) experiment to the next , we show both the mean and the median probability .
we see that we ll need at least 200 events if we want to reject the hypothesis of a constant @xmath21 at the 90% c. l. ( on average ) .
the confidence level then increases very quickly as additional events are added ; by the time 1,000 events have been accumulated , we can be quite sure that a constant @xmath21 can be excluded with high confidence .
.this table illustrates how the binning , and in particular the combination of bins into `` windows '' , affects the information that can be gleaned from the reconstructed wimp velocity distribution .
the first four columns show the average number of events in a given experiment , the number of bins , the size of the first bin in kev , and the number of bins per window .
the remaining four columns show the mean and median probability that the reconstructed @xmath21 is compatible with a constant , the mean of the quantity @xmath155 defined in eq.([e3_32 ] ) , and the average @xmath140 of eq.([e3_31 ] ) . [
cols="^,^,^,^,^,^,^,^",options="header " , ] this confidence level , as well as more general measures of the information that can be extracted from a given experiment , depend on the choices of @xmath156 and , in particular , @xmath97 .
this is illustrated in table 1 , which shows results for different combinations of @xmath157 and @xmath97 for 500 ( first five rows ) and 5,000 ( last six rows ) expected events per experiment . here
the mean and median probabilities are the same as in fig .
[ fig306 ] .
in addition we show the mean of the quantity @xmath155 , defined as [ e3_32 ] = _ , c _ f_1,r(v _ ) f_1,r(v _ ) .
formally @xmath155 determines the significance with which the hypothesis @xmath158 can be rejected .
since @xmath21 is normalized , this hypothesis is unphysical .
nevertheless @xmath155 can be regarded as a measure of the information content of a set of reconstructed @xmath159 ; in the absence of correlations , it becomes the sum over the inverse squares of the _ relative _ errors .
note that , in contrast to @xmath140 , @xmath155 does not have a @xmath160 factor in front ; after all , by adding more windows we also add more values @xmath134 at which @xmath21 is determined , which can increase the information content .
the first four rows , as well as the last four rows , show the effect of varying @xmath97 , the maximal number of bins that are collected in a window .
we see that there is an optimal choice for this quantity .
reducing @xmath97 leads to loss of information , as indicated by greatly increased values for the probability that @xmath138 is compatible with @xmath21 being constant as well as reduced values of @xmath155 . on the other hand , making the windows too large introduces too large systematic uncertainties in the estimates of the logarithmic slopes @xmath128 , which in turn leads to too large average values of @xmath140 .
this is illustrated by rows four and seven , which have large windows due to our choice of a large @xmath97 ( row 4 ) or a large @xmath161 ( row 7 ) .
the table also shows that the choice of @xmath161 has some impact on the confidence level with which the hypothesis of a constant @xmath21 can be rejected .
we saw in figs .
[ fig305 ] that our input @xmath21 has a broad maximum at @xmath162 km / s .
rejection of the hypothesis of a constant @xmath21 is therefore optimized by maximizing the information about the outer reaches of @xmath21 . getting accurate information about @xmath21 at large velocities
is very difficult ; this would need a large number of events at large @xmath8 , where the counting rate is very small .
this leaves the region of small wimp velocity . by choosing a large first bin ,
one greatly reduces the error on @xmath138 in this first bin , which is also the first window ; this was illustrated in fig .
[ fig302 ] .
in fact , for 500 events and @xmath163 one can formally maximize the confidence level with which a constant @xmath21 can be rejected by considering only two bins , and making the first bin very large ; this is shown in the fifth row .
note that this leads to an average @xmath140 well below unity , indicating that in spite of the large bins , systematic errors are still insignificant .
however , we note that in this case our assumption that the error on @xmath133 is negligible is clearly not justified , since @xmath133 receives almost its entire contribution from the large first bin . by including the error on @xmath164 but ignoring the ( strongly correlated ! ) error on @xmath133 we clearly over
estimate the total statistical error in this case .
recall also that a large first bin leads to a large value for the smallest velocity , @xmath165 , where @xmath21 is determined .
our `` figure of merit '' @xmath155 is less dependent on the details of binning , although , as stated earlier , it does strongly benefit from combining several bins into windows .
we also note that the optimal achievable @xmath155 is essentially proportional to the number of events in the sample .
this is expected , since @xmath155 is something like an inverse squared relative error .
we saw in the previous subsection that a direct reconstruction of the wimp velocity distribution @xmath21 will only be possible once several hundred elastic nuclear recoil events have been collected .
this is a tall order , given that not a single such event has so far been detected ( barring the possible dama observation ) .
the basic reason for the large required event sample is that , @xmath21 being a normalized distribution , only information on the _ shape _ of @xmath21 is meaningful . in order to obtain such shape information via direct reconstruction
, we have to separate the events into several bins or windows .
moreover , each window should contain sufficiently many events to allow an estimate of the _ slope _ of the recoil spectrum in this window . on the other hand , at the end of sec .
2 we also gave expressions for the _ moments _ of @xmath21 . with the exception of the moment with @xmath166 , these are entirely inclusive quantities , i.e. , each moment is sensitive to the entire data set ; no binning is required , nor do we need to determine any slope ( with one possible minor exception ; see below ) .
it thus seems reasonable to expect that one can obtain meaningful information about these moments with fewer events .
an independent motivation for the determination of these moments is that they are sensitive also to @xmath21 at large values of the wimp velocity @xmath6 .
we saw above that direct reconstruction of @xmath21 at large @xmath6 is very difficult , due to the small number of events expected in this region .
moreover , a delta function like contribution to @xmath21 at the highest velocity , @xmath167 is very difficult to detect using direct reconstruction ; such a contribution is expected in `` late infall '' models of galaxy formation @xcite .
the experimental implementation of eq.([eqn215 ] ) is quite straightforward . for @xmath168 ,
the normalization @xmath133 has already been given in eq.([e3_29 ] ) .
the case of non vanishing threshold energy @xmath24 can be treated straightforwardly , using eq.([en4 ] ) . to that end
we need to estimate the recoil spectrum at the threshold energy .
one possibility would be to choose an artificially high value of @xmath24 , and simply count the events in a bin centered on @xmath24 .
however , in this case the events with @xmath169 would be left out of the determination of the moments .
we therefore prefer to keep @xmath24 as small as experimentally possible , and to estimate the counting rate at threshold using the ansatz ( [ eqn311 ] ) .
since we need the recoil spectrum only at this single point , we only have to determine the quantities @xmath164 and @xmath170 parameterizing @xmath17 in the first bin ; this can be done as described in the previous subsection , without the need to distinguish between bins and `` windows '' . introducing the shorthand notation [ e3_32a ] r_thre _ q = q_thre
, the resulting error can be written as [ e3_33 ] ^2 ( r_thre ) = r_thre^2 .
the squared errors for @xmath164 and @xmath170 are simply the corresponding diagonal entries of the respective covariance matrices given in eqs.([e3_23 ] ) and ( [ e3_26 ] ) .
finally , the definition ( [ e3_12 ] ) of @xmath94 implies [ e3_34 ] q_s,1 + k_1 = q_1 - + x_1 , where @xmath171 as before and @xmath93 is the central @xmath56value in the first bin .
it should be noted that the first term in eq.([eqn215 ] ) is negligible for all @xmath172 if @xmath173 kev ; however , even for this low threshold energy it contributes significantly to the normalization constant @xmath133 , as described by eq.([en4 ] ) .
of course , the first term in eq.([eqn215 ] ) always dominates for @xmath174 .
this is not surprising , since the very starting point of our analysis , eq.(1 ) , already shows that the counting rate at @xmath24 is proportional to the `` minus first '' moment of the velocity distribution .
the integral appearing in eq.([eqn215 ] ) can be estimated through the sum [ e3_35 ] i_n = _ a , see eq.([e3_29 ] ) .
since all @xmath175 are determined from the same data , they are correlated , with [ e3_36 ] cov(i_n , i_m ) = _ a .
this can e.g. be seen by writing eq.([e3_35 ] ) as a sum over narrow bins , such that the recoil spectrum within each bin can be approximated by a constant .
each term in the sum would then have to be multiplied with the number of events in this bin ; eq.([e3_36 ] ) then follows from standard error propagation .
note that , when re converted into an integral , the expression for @xmath176 will diverge logarithmically for @xmath31 ; equivalently , the numerical estimate of this entry can become very large if the sample contains events with very small @xmath56values .
our numerical results presented below have therefore been obtained with @xmath177 kev ; many existing experiments in fact require significantly larger energy transfers in their definition of a wimp signal . with this correction , the reconstructed @xmath187
indeed closely reproduce the input values after averaging over sufficiently many experiments , even if the number of events in a given experiment is small .
however , the numerical analysis revealed a number of additional problems .
these can be understood from the observation that the @xmath175 in eq.([e3_35 ] ) , and even more the entries in their covariance matrix ( [ e3_36 ] ) , receive significant contributions from large @xmath8 values , where the counting rate itself is already very small .
this is illustrated in fig .
[ fig307 ] .
the @xmath188axis shows the quantity [ e3_45 ] r(q_min ) _ q_min^q_max dq , divided by the total counting rate @xmath189 . here
, @xmath190 is the kinematic maximum of @xmath8 for given input parameters , and @xmath191 is varied freely between 0 and @xmath190 .
the @xmath192axis shows analogously the contributions to some @xmath175 ( lower curves ) and to the corresponding diagonal elements of the covariance matrix , i.e. , the squared errors ( upper curves ) , that come from the region @xmath193 . in the latter case
we have converted the sums in eq.([e3_36 ] ) back into integrals .
the figure shows that the region of @xmath56values that contributes 99% of the counting rate only contributes about 92% to @xmath194 , 73% to @xmath195 and 51% to @xmath196 ; for the given input parameters , this corresponds to the region @xmath197 kev .
even worse , this region only contributes about 35% to @xmath198 and 5% to @xmath199 !
this implies that an experiment collecting only a small number of events will typically underestimate @xmath187 and , especially , its error ; the problem will become worse with increasing @xmath55 . on the other hand , as mentioned above , when averaged over sufficiently many experiments , our estimates for the @xmath175 do reproduce the true ( input ) values .
this implies that occasionally an experiment will greatly _ over_estimate the @xmath175 , the problem again getting worse for larger @xmath55 .
our numerical analysis also shows that , after averaging over ( very ) many experiments , eq.([e3_36 ] ) reproduces the mean square deviation between our estimated @xmath175 and the true ( input ) value .
nevertheless , we just saw that in most cases this error is being underestimated . in order to be conservative , we therefore added `` the error on the error '' to the diagonal entries of the covariance matrix ; the off diagonal entries are then scaled up such that the correlation matrix remains unaltered .
the squared `` error on the error '' is defined as [ e3_46 ] ^2(cov(i_n , i_n ) ) = _ a . with this modification
, the average @xmath141 , again averaged over many experiments , is in the vicinity of unity at least for the first few moments . ,
where @xmath200 are the values of our estimators based on eq.([e3_35 ] ) and @xmath175 are the true ( input ) values , in general does not imply that @xmath201 . adding the `` error of the error '' to the covariance matrix brings the average of this ratio closer to unity . ]
another problem is that the errors of the @xmath175 are very highly correlated .
this can also be understood from fig .
[ fig307 ] : a single event at high @xmath8 will contribute greatly to all moments with sufficiently large @xmath55 .
numerically we find correlations of more than 98% between @xmath202 and @xmath203 for all @xmath204 ; the correlation between @xmath27 and @xmath205 still amounts to more than 87% .
this implies that the higher moments unfortunately add only little to the available information .
worse , attempting to include high moments in a @xmath141 fit often leads to numerical instabilities ; recall that a covariance matrix containing 100% correlated entries become singular , i.e. , can no longer be inverted . in practice
only the moments with @xmath206 therefore seem to be useful .
we are now ready to present some representative numerical results .
[ fig308 ] shows the first 10 moments reconstructed with 100 events , using our standard input parameters ( see fig .
[ fig301 ] ) .
the estimated values of the moments have been divided by the true values .
we see that in this `` typical '' example the high moments are indeed underestimated .
we also see that the estimated relative errors at first increase with increasing @xmath55 ; this reproduces correctly the trend of the actual deviation of the estimated moments from the exact values .
however , even after adding `` the error on the error '' , we find that the relative errors start to decline again for @xmath207 .
this effect is probably entirely spurious ; recall that the errors are likely to be even more underestimated than the moments themselves .
nevertheless we find it encouraging that already with 100 events a couple of moments can be determined with errors of about 15% .
[ fig309 ] show a example of the information that might be gleaned from analyses of reconstructed moments of the wimp velocity distribution . here
we attempt to constrain a possible `` late infall '' component in @xmath21 @xcite , defined by the ansatz [ e3_47 ] f_1(v ; v_esc , n_l ) = n_s f_1,(v ) ( v_esc - v ) + n_l ( v - v_esc ) . here
@xmath208 is the standard `` shifted gaussian '' distribution ( [ eqn222 ] ) .
as before , we have multiplied it with a cut off at @xmath66 .
in addition , we introduce a contribution of wimps with fixed velocity , which we set equal to @xmath66 ; these wimps are just falling into our galaxy .
is significantly larger than @xmath209 , this smearing should not matter very much .
see ref .
@xcite for a discussion of the effect of late wimp infall on the recoil spectrum . ]
@xmath210 and @xmath66 are our free parameters ; @xmath211 is then chosen such that the total @xmath21 is normalized to unity . we then plot contours of @xmath212 , defined as the deviation of @xmath141 from its minimal value , where @xmath141 is defined as [ e3_48 ] ^2 = _ m , n _ mn ; here @xmath213 are the reconstructed moments in our mock experiment , @xmath214 are the predictions for these quantities based on eq.([e3_47 ] ) , and @xmath215 is the inverse of the covariance matrix ( [ e3_41 ] ) . figs .
[ fig309 ] show a strong degeneracy in the fit .
if the galactic escape velocity @xmath66 is kept fixed at its input value of 700 km / s , a quite significant upper bound on the normalization @xmath210 of the late infall component could be derived already from our simulated experiment with 25 events . however ,
if @xmath66 is kept free , no significant upper bound can be derived even from the simulated experiment with 100 events .
note that the two experiments have been simulated independently , i.e. , the 25 events used for the analysis in the upper frame are not part of the 100 events used in the lower frame .
the simulated experiment with 100 events was a bit `` unlucky '' in that the input values lie just outside the @xmath216 contour .
as a result , the upper bound on @xmath210 for fixed @xmath217 km / s actually comes out a little worse in this case than in the experiment with only 25 events .
this is in spite of the fact that the larger data sample allowed us to include one more moment in the fit .
note that , according to the definition ( [ e3_47 ] ) , all wimps in our galactic neighborhood have velocity @xmath218 .
this implies that a lower bound on @xmath66 can be derived from the highest observed @xmath56value , @xmath219 , see eq.([eqn204 ] ) .
unfortunately for our standard set of input parameters , this method on average only yields lower bounds of about 400 ( 460 ) km / s for experiments with 25 ( 100 ) events .
even the experiment with 100 events would then still allow 60% of all wimps to originate from a late infall component ; this is to be compared with the theoretical expectation @xmath220 .
finally , we note that for @xmath221 , it will be essentially impossible to derive a meaningful upper bound on @xmath66 from these experiments : because the original `` shifted gaussian '' velocity distribution is already very small at our input value @xmath217 km / s , increasing @xmath66 has essentially no effect on the measured recoil spectrum .
in this paper , we have developed methods that allow to extract information on the wimp velocity distribution from the recoil energy spectrum @xmath17 measured in elastic wimp nucleus scattering experiments . in the long term
this information can be used to test or constrain models of the dark halo of our galaxy ; this information would complement the information on the density distribution of wimps , which can be derived e.g. from measurements of the galactic rotation curve . to this end , in sec .
2 we derived expressions that allow to directly reconstruct the normalized one dimensional velocity distribution function of wimps , @xmath5 , given an expression ( e.g. , a fit to data ) for the recoil spectrum .
we have also derived formulae for the moments of @xmath21 .
all these expressions are independent of the as yet unknown wimp density near the earth as well as of the wimp nucleus cross section ; the only information about the nature of the wimp that is needed is its mass . furthermore , in sec . 3 we have developed methods that allow to apply our expressions directly to data , without the need to fit the recoil spectrum to a functional form .
we found that a good variable that allows direct reconstruction of @xmath21 is the average recoil energy in a given bin ( or `` window '' ; see sec .
this average energy is sensitive to the slope of the recoil spectrum , which is the quantity we need to reconstruct @xmath21 .
we carefully analyzed the statistical errors .
unfortunately we found that several hundred events will be needed for this method to be able to extract meaningful information on @xmath21 .
this is partly due to the fact that @xmath5 is normalized , i.e. , only the _ shape _ of this distribution contains meaningful information , and partly because this shape depends on the _ slope _ of the recoil spectrum , which is intrinsically difficult to determine .
we therefore turned to an analysis of the moments of @xmath21 .
we found that reliably estimating higher moments , and in particular estimating their errors , is difficult .
the main reason is that these higher moments get large contributions from very rare events with large recoil energies .
nevertheless we found that , based only on the first two or three moments , some non trivial information can already be extracted from @xmath222 events .
the main emphasis of this exploratory study was on the basic expressions as well as on their implementation in actual experiments .
the models for @xmath21 we tried to constrain in our applications ( a constant in sec .
3.2 , a `` late
infall '' component with fixed velocity in the earth rest frame in sec .
3.3 ) are not physical ; nevertheless they illustrate the difficulties one will have in extracting information from these experiments that are of interest for the modeling of the galactic wimp distribution .
our analysis is based on several simplifying assumptions .
first , we ignored all experimental systematic uncertainties , as well as the uncertainty on the determination of the recoil energy @xmath8 .
this is probably quite a good approximation , given that we found that we ll have to live with quite large statistical uncertainties in the foreseeable future ; recall that not a single wimp event has as yet been unambiguously recorded .
we also assumed that our detector consists of a single isotope .
this is quite realistic for the current semiconductor ( si or ge ) detectors .
the cresst detector @xcite contains three different nuclei .
however , by simultaneously measuring heat and light , one might be able to tell on an event by event basis which kind of nucleus has been struck . in this case , our methods can be applied straightforwardly to the three separate sub spectra .
our analyses treat each recorded event as signal , i.e. , we ignore backgrounds altogether . at least after introducing a lower cut @xmath24 on the recoil energy , this may in fact not be unrealistic for modern detectors , which contain cosmic ray veto and neutron shielding systems .
background subtraction should be relatively straightforward when fitting some function to the data , which would allow to use the expressions of sec .
2 . it should also be feasible in the method described in sec . 3.2 , if its effect on the average @xmath56values in the bins can be determined ; in particular , an approximately flat ( @xmath56independent ) background would not change the slope of the recoil spectrum . subtracting the background in the determination of the moments as described in sec .
3.3 might be ( even ) more difficult .
as noted earlier , we need to know the wimp mass @xmath223 .
this is true for any reconstruction method based on data taken with a mono isotopic detector . in this case
one can always `` reconstruct '' @xmath5 , for any ( assumed ) value of @xmath223 .
fortunately in well motivated wimp models , @xmath223 can be determined with high accuracy from future collider data . even in this case
one will want to check experimentally that the wimps seen in dark matter detection experiments are in fact the same ones produced at colliders .
this can be done by using the methods developed here on two different data sets , obtained with different detector materials , and demanding consistent results for ( the moments of ) @xmath21 .
the feasibility of such an analysis is currently under investigation . in our analysis
we ignored the annual modulation of the wimp flux .
again , given the large statistical errors expected in the foreseeable future , this is a reasonable first approximation .
nevertheless , eventually one will have to extend the formulae and methods developed here to allow for an annual modulation .
this is fairly straightforward if the background is again negligible . on the other hand
, new methods may be needed to extract information on @xmath21 in a situation where the total counting rate is dominated by background events ; this is most likely the case for the dama data @xcite , even if they indeed contain a signal , which remains highly controversial . in summary , we have begun to explore what direct dark matter detection experiments can teach us about the velocity distribution of dark matter particles in our galactic neighborhood .
our analyses show that this will require substantial data samples .
we hope this will encourage our experimental colleagues to plan future experiments well beyond the stage of `` merely '' detecting dark matter .
we thank holger drees for illuminating discussions on stochastics .
this work was partially supported by the marie curie training research network `` universenet '' under contract no . mrtn - ct-2006 - 035863 , as well as by the european network of theoretical astroparticle physics entapp ilias / n6 under contract no .
rii3-ct-2004 - 506222 .
since v = , [ eqna01 ] we have dv = dq , [ eqna02 ] from eq.([eqn212 ] ) and according to the normalization condition in eq.([eqn214 ] ) , we have , f_1(v ) ~ dv & = & \{-2 q } dq + & = & dq + & = & + & = & dq + & = & 1 , [ eqna03 ] where we have used the conditions |_q 0 and |_q 0 . eq.([eqn213 ] ) immediately followed from eq.([eqna03 ] ) . using eqs.([eqna01 ] ) , ( [ eqna02 ] ) and integration by parts , we can also find the moments of @xmath21 , defined with a lower cut off @xmath24 on the energy transfer , as follows : & = & _ v_min(q_thre)^v^n f_1(v ) ~ dv + & = & _ q_thre^^n dq + & = & _ q_thre^ q^(n+1)/2 dq + & = & ^n+1 \ { _ q = q_thre + _ q_thre^q^(n-1)/2 dq } .
this reproduces eq.([eqn215 ] ) in sec . 2 .
starting point is the observation that we wish to compute the ratio of two integrals , [ eb1 ] = . in the second step
the integrals have been discretized , i.e. , replaced by sums over bins @xmath224 with @xmath225 events per bin .
we now write the @xmath225 as sum of average value @xmath226 and fluctuation @xmath227 : [ eb2 ] = . introducing the notation g_a
= _ i second order in the @xmath227 , we have : @xmath228 \nonumber \\ \&\simeq\ & \frac { \bar g_1 } { \bar g_2 } + \frac { 1}{\bar g_2 } \abrac{\sum_i \delta n_i g_1(x_i ) } - \frac { \bar g_1 } { \bar g_2 ^ 2 } \abrac{\sum_i \delta n_i g_2(x_i ) } \nonumber \\
\&~\ & ~~~~~~~~ - \frac{1}{\bar g_2 ^ 2 } \abrac{\sum_i \delta n_i g_1(x_i ) } \abigg{\sum_j \delta n_j g_2(x_j ) } + \frac { \bar g_1 } { \bar g_2 ^ 3 } \left ( \sum_i \delta n_i g_2(x_i ) \right)^2 \ , .\end{aligned}\ ] ] we now consider the average over many experiments . of course , @xmath227 averages to zero , but the product @xmath229 averages to @xmath230 , i.e. , it is non
zero for @xmath231 . hence : [ eb4 ] - + .
the sums appearing in the two correction terms also appear in the definition of the covariance matrix between @xmath232 and @xmath233 .
note that we wish to compute the first term on the right hand side , since in our case the estimators for @xmath232 and @xmath233 indeed average to the correct values .
this then leads to the final result [ eb5 ] - = ( g_1,g_2 ) - ( g_2,g_2 ) . applying this result to eqs.([eqn215 ] ) and ( [ en4 ] ) then immediately leads to eq.([e3_44 ] ) .
v. c. rubin and w. k. ford , _ astrophys .
j. _ * 159 * , 379 ( 1970 ) ; s. m. faber and j. s. gallagher , _ annu . rev
astrophys . _
* 17 * , 135 ( 1979 ) ; v. c. rubin , w. k. ford , and n. thonnard , _ astrophys .
j. _ * 238 * , 471 ( 1980 ) ; k. g. begeman , a. h. broeils , and r. h. sanders , _ mon . not .
r. astron .
* 249 * , 523 ( 1991 ) ; r. p. olling and m. r. merrifield , _ mon . not .
r. astron .
soc . _ * 311 * , 361 ( 2000 ) .
j. f. navarro , c. s. frenk , and s. d. m. white , _ astrophys .
j. _ * 462 * , 563 ( 1996 ) ; a. v. kravtsov _
_ , _ astrophys
. j. _ * 502 * , 48 ( 1998 ) ; b. moore _
_ , _ mon . not .
r. astron .
soc . _ * 310 * , 1147 ( 1999 ) . | weakly interacting massive particles ( wimps ) are one of the leading candidates for dark matter . currently , the most promising method to detect many different wimp candidates is the direct detection of the recoil energy deposited in a low background laboratory detector due to elastic wimp nucleus scattering . so far
the usual procedure has been to predict the event rate of direct detection of wimps based on some model(s ) of the galactic halo .
the aim of our work is to invert this process . that is , we study what future direct detection experiment can teach us about the wimp halo . as the first step
we consider a time averaged recoil spectrum , assuming that no directional information exists .
we develop a method to construct the ( time averaged ) one dimensional velocity distribution function from this spectrum .
moments of this function , such as the mean velocity and velocity dispersion of wimps , can also be obtained directly from the recoil spectrum .
the only input needed in addition to a measured recoil spectrum is the mass of the wimp ; no information about the scattering cross section or wimp density is required .
march 2007 * reconstructing the velocity distribution of wimps + from direct dark matter detection data * + and chung - lin shan + |
the fundamental paradigm of solar system astronomy prior to the time of copernicus was that the earth was at the center of the solar system .
also , celestial bodies were assumed to move along perfect circles .
this led to the system of deferents and epicycles .
one prime motivation for the use of epicycles was to account for retrograde motion .
copernicus great book _ on the revolutions of the heavenly spheres _
( 1543 ) asserted that the sun was physically and truly at the center of the solar system , and that this provided a much simpler explanation for the retrograde motion of the planets .
however , copernicus retained circular motion . also , he retained the notion of epicycles because they were needed to account for variations of distance of the planets from the sun @xcite . in ancient greek astronomy ,
the _ direction _ towards the moon , sun , or a planet was more important than the implied distance to it .
ptolemy s model of the motion of the moon implied that its distance from the earth varied nearly a factor of two .
naked eye observations by this author have demonstrated that without a telescope one can show that the angular size of the moon varies in a regular fashion , implying that moon s distance varies in a regular fashion @xcite .
the implied eccentricity of the moon s orbit was @xmath10.04 .
( the true eccentricity of the moon s orbit is 0.055 , but its orbit is anything but a simple ellipse , owing to the gravitational force of the sun . )
the point here is that epicycles in ptolemaic astronomy were a geometrical device to explain retrograde motion ( in the case of the planets ) or to determine the direction towards the moon . in copernicus model
, the use of an epicycle implies a realistic , physical variation of distance . in
1609 johannes kepler published the original versions of his first two laws of planetary motion : 1 ) the orbit of a planet is an ellipse , with the sun at one focus ; and 2 ) what we now call the law of areas , that the radius vector of a planet sweeps out equal areas in equal times .
the second law can be stated as follows : @xmath2 where @xmath3 is the distance between a planet and the sun , d@xmath4 is an angular increment in radians , and @xmath5 is a constant unique to each planet .
newton s breakthroughs in mathematics and mechanics led to the realization that kepler s first law needed correction .
the very center of the sun is not at the focus of a planetary orbit .
a planet orbits the _ center of mass _ of the planet - sun system , and the sun orbits that center of mass too @xcite .
this idea , of course , has led to the discovery of many extra - solar planets via the radial velocity method . in the autumn of 2015
mars was nicely situated in the constellation leo before sunrise .
we began a sequence of observations of mars using a simple cross staff ( fig .
[ f1 ] ) .
say the full width of the cross staff is @xmath6 , and suppose at a linear distance @xmath7 down the ruler the angular separation of two celestial objects exactly matches the width of the cross staff .
then the angular separation of the two objects will be @xmath8 if an observer can measure the angular separation of a planet and two stars of known celestial coordinates , there are two solutions for the position of the planet , one on each side of the great circle arc joining the two stars . if the planet is close to being on the great circle arc between the two stars , perhaps no solution results , given errors of measurement .
if the positions of a planet and the two stars form a spherical triangle with reasonably equal sides , this is ideal , and the planetary position can be determined as accurately as possible .
because we are using a handheld naked eye instrument , it is advised to use three to five reference stars .
we assume a system of accurate stellar coordinates of bright stars along the zodiac .
we adopt the j2000.0 coordinates of such stars from the simbad database .
understanding johannes kepler s efforts to discover the elliptical nature of mars orbit requires serious effort . a good place to start is an article by @xcite .
to make a long story shorter , kepler believed that an ovoid would fit the ( then ) most accurate data available ( obtained by tycho and his assistants ) .
kepler found anomalies amounting to 8 between the measured ecliptic longitudes and his model .
since he believed that tycho s data were good to @xmath02 or better , he concluded that there was a problem with the _
model_. this led him to conclude that his _ approximation _ to the ovoid ( namely , an ellipse ) was the true orbital shape .
we wondered if it were possible to demonstrate from simple naked eye observations that the orbit of mars is indeed an ellipse . or , requiring less rigor , are the positions of mars consistent with an elliptical orbit ? here we present results based on nearly a year of observations .
a full blown orbital determination for the planet mars is beyond the scope of the present paper .
that would involve simultaneously solving for all six orbital elements .
we are only trying to show that a data set obtained with simple equipment can be fit with an ellipse of eccentricity @xmath1 0.093 .
other values of the eccentricity can be shown to give ecliptic longitudes that differ from the observational data by 1.5 deg , far larger systematic differences than the internal random errors of the observations .
in table [ data ] we give various data relating to mars . for each julian date
we give the `` true '' right ascension ( @xmath9 ) and declination ( @xmath10 ) of the planet , obtained using an algorithm by @xcite .
these coordinates are accurate to @xmath01 .
note that these coordinates will correspond to the equinox of date in the years 2015 or 2016 . to convert these coordinates to ecliptic latitude ( @xmath11 ) and longitude ( @xmath12 )
we need the following formulas from spherical trigonometry @xcite : @xmath13 @xmath14 @xmath15 @xmath16 is the obliquity of the ecliptic , 23 26 21.406 for the year 2000 . using the atan2 function in fortran or python with arguments sin(@xmath12 ) and cos(@xmath12 ) , we obtain the ecliptic longitude in the correct quadrant . table [ data ] also gives the observed right ascension , declination , ecliptic longitude , and ecliptic latitude of mars derived from the cross staff measurements , along with the number of reference stars used and the value for each date of the sun s ecliptic longitude .
the values of the sun s longitude were calculated using the second method of @xcite .
this method uses the sun s mean anomaly , but apparently does _ not _ use the earth s orbital eccentricity . on occasion
we desired one more sufficiently bright reference star and used the _ derived _ position of saturn as that reference position .
consider two celestial objects with equatorial coordinates ( @xmath17 , @xmath18 ) and ( @xmath19 , @xmath20 ) . the angular separation ( @xmath4 ) between two objects is : @xmath21 next consider a spherical quadrilateral that is bounded by starting and ending right ascensions , and starting and ending declinations .
the quadrilateral is divided into a grid , given a nominal increment in each coordinate of 0.01 deg .
we used a computer program of our devising that uses the coordinates of two reference stars and the measured angular distance of a planet from each of these stars to determine the coordinates of the planet .
if the coordinates of the stars are j2000.0 coordinates , then the derived right ascension and declination of the planet also have j2000.0 coordinates .
the derived ecliptic coordinates of mars are shown in figure [ f2 ] .
the solid line in the plot shows the locus of `` true '' positions from @xcite .
the earliest observations are of lesser quality . on these occasions
we only used two reference stars , and was getting used to a new observing procedure . on almost all subsequent occasions we used more than two reference stars .
two of these first three observations are easily shown to be outliers .
our first three observations will thus be excluded from subsequent analysis .
another thing to note about figure [ f2 ] is that mars was north of the ecliptic until jd @xmath1 2,457,504 ( april 25 , 2016 ) and then was located south of the ecliptic .
in other words , the orbit of mars is inclined to the ecliptic .
the modern value of the orbital inclination is 1.850 deg ( * ? ? ?
* on p. 295 ) . for our purposes here
we will assume that mars orbit is coplanar with the ecliptic .
given that most positions of mars listed in table [ data ] were derived from angular separations with respect to three to five reference stars , almost all derived right ascensions and declinations have easy - to - calculate internal random errors .
these are sometimes as small as @xmath00.01 deg ( which we do not really believe ) . on one occasion ( jd 2457399.9840 )
the internal random error for right ascension was @xmath00.23 deg and the internal random error of declination was @xmath00.47 deg .
typical internal random errors for right ascension and declination are @xmath22 deg . since we have `` true '' positions of mars from @xcite we can make a direct estimate of the accuracy of our observations .
the easiest way to do this is to precess the ecliptic longitudes from column 4 in table [ data ] to equinox j2000 by subtracting 50.25 arc seconds per year times the number of years from jd 2,457,543.5 ( january 0.0 , 2000 ) to the date of observation .
the ecliptic latitudes require no precession correction .
the subsequent differences of ecliptic longitude and latitude ( `` observed '' _ minus _ `` true '' ) give mean differences of @xmath23 deg in ecliptic longitude and @xmath24 deg in ecliptic latitude .
the standard deviations of the distributions of differences are : @xmath250.10 deg and @xmath260.17 deg .
the square root of the sum of squares of those errors is @xmath00.20 deg , or @xmath012 .
this is the accuracy of the position of a bright object ( such as a planet ) obtainable with our simple cross staff .
in book 5 , chapter 19 , of _ on the revolutions of the heavenly spheres _
@xcite derived the perigee , mean , and apogee distances of mars .
he obtained the values 1.374 , 1.520 , and 1.649 au , respectively .
thus , copernicus knew the amount by which mars distance from the sun varies , and his mean distance is very close to the modern value of the semi - major axis size of mars orbit ( 1.5237 au ) .
further discussion of copernicus model can be found in appendix [ sec : comparison ] .
let us consider the elliptical orbit of mars .
the equation of an ellipse is : @xmath27 where r is the mars - sun distance , @xmath28 is the semi - major axis of the ellipse , @xmath29 is the eccentricity , and angle @xmath4 = 0 when mars is at perihelion .
the velocity along the orbit ( * ? ? ?
2.36 ) is @xmath30 since the mass of mars is @xmath1 3.23 @xmath31 10@xmath32 m@xmath33 @xcite , for our purposes here we shall ignore it .
we wish to calculate the position of mars at increments of one earth day starting at the moment of its perihelion . at perihelion
r = r@xmath34 = @xmath35 . using the known semi - major axis size and eccentricity of mars orbit
, we can calculate the maximum velocity at perihelion with equation 8 .
on perihelion day traveling at velocity @xmath36 mars moves 0.6349 degrees along its orbit as viewed from the sun .
this allows us to calculate the constant @xmath5 for mars using equation 1 .
then , by alternating use of equations 7 and 1 we can calculate @xmath3 and @xmath4 for mars each day along its orbit .
the x - y coordinates are obtained simply : @xmath37 cos(@xmath4 ) and @xmath38 sin(@xmath4 ) .
for the earth let us adopt a circular orbit .
( the eccentricity of the earth s orbit is actually about 0.0167 . ) on average , the earth moves 360.0/365.2422 = 0.985647 degrees per day with respect to the sun .
this gives us another set of x - y coordinates in the same coordinate system , with the sun at the origin and the + x - axis in the direction of mars perihelion .
consider figure [ f3 ] .
according to the solar systems dynamic group , horizons on - line ephemeris system at jet propulsion laboratory , the previous perihelion of mars occurred on julian date 2,457,003.8524 ( december 12 , 2014 , at 08:27:29 ut ) .
the next perihelion occurs on julian date 2,457,691.0507 ( october 29 , 2016 , at 13:13:04 ut ) . for the moment we will take the perihelion dates as given . they are not directly observable , but the date of opposition of mars essentially is .
mars ecliptic longitude differs from that of the sun by 180 deg near the mid - time of its retrograde motion . from our data we find opposition to have occurred on may 22 , 2016 , at 2 hours ut . according to the _ astronomical almanac _ for 2016 opposition occurred on may 22nd at 11 hours ut .
the agreement is reasonably good .
mars latest opposition occurred 526.9 days after the perihelion of december 12 , 2014 . call it 527 days .
the x - y coordinates of the day - by - day position of mars in our coordinate system give @xmath4 = 265.545 deg on the day of opposition .
since the earth moves 0.985647 deg per day along its orbit , the 269th pair of earth coordinates gives an angle @xmath4 most closely matching that of mars ( 265.139 vs. 265.545 deg , in fact ) .
given the index @xmath39 of the earth s coordinates , the corresponding index of mars coordinates for the same day is equal to @xmath39 + ( 527 @xmath40 269 ) .
we wrote a simple program in python that calculates the x - y coordinates of mars for each day starting at perihelion , and x - y coordinates of the earth . with an appropriate offset of the indices of the two sets of coordinates
we can obtain the direction toward mars _ from the earth _ for any given date .
then we can find an arithmetic offset between days - since - mars - perihelion and julian date , and we can find another arithmetic offset to transform the angles obtained to ecliptic longitude .
this is just an example of shifting a template in two coordinates to minimize the sum of squares of deviations from the data to the offset template .
this produces a goodness of fit parameter .
the square root of the goodness of fit parameter divided by the number of data points minus 2 gives the rms scatter of the fit of the template to the data .
varying the model parameters allows us to search for an even better fit to the data .
our program lets us a generate a template of ecliptic longitude values vs. time . using the official values of mars perihelion date , orbital eccentricity , and orbit size given in table [ summary ]
, we can adjust the x and y axis values to minimize a goodness of fit parameter and make a plot that fits the data well enough to eye . the goodness of fit parameter is the sum of squares of differences between a template and the data , in other words , like @xmath41 minimization , but with equal weights for all the points .
we wondered if a better fit to our data might be obtained with different fit parameters .
in other words , we start by deriving the date of perihelion , using our data . in the top panel of figure [ f4 ]
we show the goodness of fit parameter for a range of dates with respect to october 29.5 . on the basis of our data ,
we find the best fit is obtained for a perihelion date of october 25.4 . the goodness of fit parameter doubles , compared to the value at the minimum , for perihelion dates october 21.5 and october 29.3 .
our perihelion date is therefore october 25.4 @xmath0 3.9 , 2016 ( ut ) .
this is the equivalent of december 8.5 , 2014 ( the date of the previous perihelion ) . in what follows
that is our effective perihelion date , allowing us to work _ forward _ in time toward the subsequent perihelion on october 25.4 , 2016 .
how robust is the eccentricity ? in the middle panel of figure [ f4 ] we show the goodness of fit parameter for a range of eccentricities ranging from 0.043 to 0.133 .
the minimum of the sum of squares of residuals occurs for @xmath29 = 0.086 , with an uncertainty of @xmath00.010 . finally , we tried a number of values of the orbital semi - major axis size , ranging from 1.50 to 1.55 .
the goodness of fit parameter is minimized for @xmath28 = 1.526 au . using our determination of the time of perihelion , the eccentricity , and a slight adjustment to the mean distance from the sun
, we obtain a sum of squares of residuals of 1.111 .
there were 42 data points used for the analysis , and 2 constraints ( adjusting time to truncated julian date and adjusting angle to ecliptic longitude ) .
the best fit to our data gives an rms residual of @xmath42 for the fit of the ecliptic longitudes to the best model .
this is only slightly better than the average difference between the data and the `` true '' values ( @xmath012 ) . ignoring the eccentricity of the earth s orbit was not a serious mistake . in figure
[ f5 ] we plot the observed ecliptic longitudes of mars ( for equinox j2000 ) vs. the julian date .
low order polynomial fits to the two stationary points indicate that retrograde motion lasted for 74.1 days .
the range of ecliptic longitude between the stationary points was 15.23 @xmath0 0.42 deg .
we also show three fits of the data for values of the orbital eccentricity of 0.053 , 0.086 , and 0.123 . even if this figure is displayed in color , it is difficult to see that the middle locus is better than the other two . in figure [ f6 ]
we show the differentials ( data _ minus _ model ) .
figure [ f6 ] clearly shows that the eccentricity can not be as small as 0.053 or as large as 0.123 . in both cases
the model differs from the data by 1.0 to 1.5 degrees 190 days before opposition and 70 days after opposition .
given that the individual positions of mars are good to @xmath00.2 deg on average , values of the eccentricity of 0.053 or 0.123 are strongly rejected by the data .
in table [ summary ] we summarize the values derived from our simple dataset based on naked eye observations using a handheld cross staff . for comparison
we also give the official ( `` true '' ) values based on a much more sophisticated orbital determination using the most modern methods .
admittedly , in science we almost never have the `` true '' values for comparison .
typically , we can only calculate our internal random errors and estimate sources of systematic error . our value of the orbital eccentricity of mars ( 0.086 @xmath0 0.010 ) compares reasonably well the official modern value ( 0.0934 ) .
while our data are not accurate enough to _ prove _ that mars orbit is an ellipse , if we fit the data with an ellipse , the eccentricity must be near 0.09 .
thus , a dataset based on naked eye observations can be shown to be in strong agreement with kepler s first law .
we made use of the simbad database , operated at cds , strasbourg , france .
in figure [ copernicus ] we show copernicus model for the orbit of mars . he retained the notion of perfectly circular orbits and therefore required an epicycle to account for the known variation of the distance of the planet from the sun . here
we use a primary circle of radius 1.520 au and one epicycle .
perihelion is obtained at position @xmath43 , when r = 1.374 au .
the epicycle turns around twice for every rotation of the primary circle .
the combination gives us a nearly elliptical orbit .
in figure [ comparison ] we compare copernicus two circle solution for mars with the keplerian ellipse having @xmath28 = 1.52366 and @xmath29 = 0.0934 .
the copernican solution differs from the keplerian ellipse by as much as 0.027 au .
gingerich , o. 1989 , in _ the general history of astronomy , vol .
2 , planetary astronomy from the renaissance to the rise of astrophysics , part a : tycho brahe to newton _ , r. taton and c. wilson , eds . , cambridge : cambridge univ . press , pp . 54
- 78 lrrrrrrrccc 313.9757 & 166.2509 & 7.3301 & 164.508 & 1.335 & 166.36 & 7.67 & 164.48 & 1.69 & 2 & 204.771 + 322.9604 & 171.3656 & 5.2083 & 170.019 & 1.365 & 171.98 & 4.29 & 170.94 & 0.76 & 2 & 213.712 + 329.9688 & 175.3095 & 3.5410 & 174.291 & 1.387 & 177.03 & 3.69 & 175.81 & 2.21 & 2 & 220.717 + 338.9674 & 180.3206 & 1.4016 & 179.737 & 1.413 & 180.31 & 1.68 & 179.62 & 1.67 & 2 & 229.748 + 344.9931 & 183.6448 & @xmath400.0205 & 183.353 & 1.430 & 183.35 & 0.10 & 183.03 & 1.42 & 4 & 235.817 + + 346.9708 & 184.7339 & @xmath400.4856 & 184.538 & 1.435 & 184.58 & @xmath400.26 & 184.31 & 1.59 & 2 & 237.812 + 359.9819 & 191.8332 & @xmath403.4881 & 192.243 & 1.467 & 191.65 & @xmath403.57 & 192.11 & 1.32 & 4 & 250.977 + 364.9833 & 194.5444 & @xmath404.6133 & 195.173 & 1.477 & 194.12 & @xmath404.34 & 194.68 & 1.57 & 3 & 256.053 + 370.9813 & 197.7750 & @xmath405.9320 & 198.651 & 1.488 & 197.45 & @xmath405.95 & 198.36 & 1.35 & 4 & 262.149 + 373.9674 & 199.3779 & @xmath406.5758 & 200.370 & 1.492 & 199.19 & @xmath406.50 & 200.17 & 1.49 & 4 & 265.187 + + 380.9542 & 203.1074 & @xmath408.0433 & 204.351 & 1.501 & 203.01 & @xmath408.09 & 204.28 & 1.42 & 4 & 272.301 + 387.9889 & 206.8359 & @xmath409.4620 & 208.301 & 1.507 & 206.51 & @xmath409.44 & 207.99 & 1.41 & 4 & 279.469 + 391.9813 & 208.9367 & @xmath4010.2376 & 210.512 & 1.509 & 208.73 & @xmath4010.12 & 210.28 & 1.55 & 4 & 283.538 + 399.9840 & 213.1084 & @xmath4011.7237 & 214.870 & 1.508 & 212.89 & @xmath4011.56 & 214.61 & 1.59 & 4 & 291.694 + 406.9861 & 216.7061 & @xmath4012.9424 & 218.591 & 1.503 & 216.45 & @xmath4012.77 & 218.30 & 1.59 & 4 & 298.986 + + 410.9840 & 218.7332 & @xmath4013.6022 & 220.672 & 1.497 & 218.60 & @xmath4013.08 & 220.39 & 1.95 & 5 & 302.894 + 415.9944 & 221.2445 & @xmath4014.3918 & 223.235 & 1.488 & 220.94 & @xmath4014.19 & 222.89 & 1.59 & 4 & 307.990 + 425.9861 & 226.1101 & @xmath4015.8320 & 228.149 & 1.460 & 225.91 & @xmath4015.74 & 227.94 & 1.50 & 5 & 318.128 + 429.9715 & 227.9932 & @xmath4016.3574 & 230.034 & 1.445 & 227.85 & @xmath4016.34 & 229.90 & 1.43 & 5 & 322.164 + 435.9937 & 230.7561 & @xmath4017.0962 & 232.782 & 1.417 & 230.26 & @xmath4017.05 & 232.31 & 1.34 & 5 & 328.250 + + 443.9931 & 234.2464 & @xmath4017.9769 & 236.225 & 1.369 & 233.84 & @xmath4017.98 & 235.85 & 1.28 & 5 & 336.312 + 449.9799 & 236.6977 & @xmath4018.5632 & 238.626 & 1.322 & 236.20 & @xmath4018.77 & 238.21 & 1.02 & 5 & 342.326 + 451.9677 & 237.4743 & @xmath4018.7441 & 239.383 & 1.305 & 236.99 & @xmath4019.08 & 239.01 & 0.88 & 5 & 344.318 + 467.9639 & 242.9168 & @xmath4019.9740 & 244.661 & 1.114 & 242.59 & @xmath4020.01 & 244.37 & 1.02 & 5 & 0.280 + 472.9521 & 244.2518 & @xmath4020.2824 & 245.949 & 1.033 & 243.94 & @xmath4020.39 & 245.68 & 0.88 & 5 & 5.228 + + 480.9486 & 245.9309 & @xmath4020.7106 & 247.572 & 0.874 & 245.74 & @xmath4020.76 & 247.40 & 0.80 & 5 & 13.132 + 499.9549 & 247.1279 & @xmath4021.4551 & 248.792 & 0.316 & 246.72 & @xmath4021.55 & 248.43 & 0.16 & 4 & 31.773 + 512.9493 & 245.2332 & @xmath4021.7269 & 247.097 & @xmath400.235 & 244.72 & @xmath4021.72 & 246.63 & @xmath400.31 & 5 & 44.407 + 515.9406 & 244.4834 & @xmath4021.7551 & 246.416 & @xmath400.381 & 244.07 & @xmath4021.62 & 246.01 & @xmath400.31 & 5 & 47.304 + 538.6354 & 236.5451 & @xmath4021.4892 & 239.126 & @xmath401.564 & 236.29 & @xmath4021.59 & 238.92 & @xmath401.72 & 5 & 69.160 + + 547.6240 & 233.4845 & @xmath4021.2476 & 236.295 & @xmath401.979 & 233.11 & @xmath4021.32 & 235.97 & @xmath402.13 & 3 & 77.768 + 555.6153 & 231.4404 & @xmath4021.0834 & 234.404 & @xmath402.282 & 231.00 & @xmath4021.20 & 234.03 & @xmath402.50 & 5 & 85.406 + 564.6441 & 230.1675 & @xmath4021.0421 & 233.243 & @xmath402.541 & 229.98 & @xmath4020.87 & 233.03 & @xmath402.42 & 5 & 94.022 + 569.6226 & 229.9735 & @xmath4021.1049 & 233.084 & @xmath402.648 & 229.67 & @xmath4021.42 & 232.89 & @xmath403.03 & 5 & 98.770 + 576.6399 & 230.3070 & @xmath4021.3033 & 233.436 & @xmath402.760 & 230.12 & @xmath4021.13 & 233.22 & @xmath402.64 & 5 & 105.460 + + 582.6149 & 231.1228 & @xmath4021.5684 & 234.238 & @xmath402.825 & 230.94 & @xmath4021.37 & 234.02 & @xmath402.68 & 5 & 111.158 + 591.6299 & 233.1991 & @xmath4022.1097 & 236.242 & @xmath402.880 & 232.79 & @xmath4022.22 & 235.90 & @xmath403.08 & 5 & 119.759 + 597.6517 & 235.0988 & @xmath4022.5434 & 238.054 & @xmath402.891 & 234.89 & @xmath4022.38 & 237.83 & @xmath402.78 & 4 & 125.512 + 602.6056 & 236.9372 & @xmath4022.9283 & 239.795 & @xmath402.889 & 236.56 & @xmath4023.13 & 239.50 & @xmath403.16 & 5 & 130.249 + 624.6229 & 247.6334 & @xmath4024.6631 & 249.743 & @xmath402.783 & 247.18 & @xmath4024.45 & 249.30 & @xmath402.64 & 5 & 151.392 + + 633.5979 & 252.9637 & @xmath4025.2414 & 254.615 & @xmath402.707 & 252.55 & @xmath4025.30 & 254.25 & @xmath402.81 & 5 & 160.063 + 639.5740 & 256.7638 & @xmath4025.5420 & 258.065 & @xmath402.649 & 256.41 & @xmath4025.57 & 257.75 & @xmath402.71 & 5 & 165.857 + 646.5729 & 261.4329 & @xmath4025.7851 & 262.284 & @xmath402.575 & 261.01 & @xmath4025.79 & 261.90 & @xmath402.60 & 5 & 172.665 + 653.5681 & 266.2981 & @xmath4025.8906 & 266.667 & @xmath402.494 & 266.05 & @xmath4025.93 & 266.45 & @xmath402.54 & 5 & 179.495 + 660.6153 & 271.3719 & @xmath4025.8399 & 271.236 & @xmath402.407 & 270.84 & @xmath4026.06 & 270.76 & @xmath402.62 & 5 & 186.403 + lll opposition date & may 22.1 & may 22.46 + perihelion date & october 25.4 @xmath0 3.9 & october 29.5507 + orbital eccentricity & 0.086 @xmath0 0.010 & 0.09341233 + semi - major axis size & 1.526 & 1.52366231 + | over the course of eleven months we determined the position of mars on 45 occasions using a handheld cross staff and two to five bright reference stars of known right ascension and declination on each occasion . on average the observed positions are within 12 of the true positions . given that we took data prior to the start of retrograde motion and well past the end of retrograde motion , we can easily derive the date of opposition to the sun .
we were able to derive the date of perihelion , the orbital eccentricity , and the semi - major axis size of mars orbit .
we obtain a value of the eccentricity of 0.086 @xmath0 0.010 , which is to be compared to the modern value of 0.0934 .
values as low as 0.053 or as high as 0.123 can be rejected at a high confidence level .
a simple dataset can be obtained with cardboard and a ruler that demonstrates the elliptical shape of mars orbit .
= 1 |
cosmological models with scalar field matter have been much studied in the context of inflation and , more recently , in the context of the late - time acceleration that is indicated by current astronomical observations ( see @xcite for a recent review ) .
one theoretical motivation for these studies is that scalar fields arise naturally from the compactification of higher - dimensional theories , such as string or m - theory .
however , the type of scalar field potential obtained in these compactifications is sufficiently restrictive that until recently it was considered to be difficult to get accelerating cosmologies in this way , although the existence of an accelerating phase in a hyperbolic ( @xmath13 ) universe obtained by compactification had been noted @xcite , and non - perturbative effects in m - theory have since been shown to allow unstable de sitter vacua @xcite . in an earlier paper , we pointed out that compactification on a compact hyperbolic manifold with a time - dependent volume modulus naturally leads to a flat ( @xmath14 ) expanding universe that undergoes a transient period of accelerating expansion @xcite .
numerous subsequent studies have shown that such cosmological solutions are typical to all compactifications that involve compact hyperbolic spaces or non - vanishing @xmath15-form field strengths ( flux ) @xcite , and this was additionally confirmed in a systematic study @xcite .
furthermore , the transient acceleration in these models is easily understood @xcite in terms of the positive scalar field potential that both hyperbolic and flux compactifications produce in the effective , lower - dimensional , action .
this perspective also makes clear the generic nature of transient acceleration . for
any realistic application one would want the lower - dimensional spacetime to be four - dimensional , but for theoretical studies it is useful to consider a general @xmath16-dimensional spacetime . assuming that we have gravity , described by a metric @xmath17 , coupled to @xmath18 scalar fields @xmath19 taking values in a riemannian target space with metric @xmath20 and with potential energy @xmath1
, the effective action must take the form @xmath21 where @xmath22 is the ( spacetime ) ricci scalar .
we are interested in solutions of the field equations of the action ( [ action ] ) for which the line element has the friedmann - lematre - robertson - walker ( flrw ) form for a homogeneous and isotropic spacetime . in standard
coordinates , @xmath23 where the function @xmath24 is the scale factor , and @xmath25 represents the @xmath26-dimensional spatial sections of constant curvature @xmath27 .
we normalise @xmath27 such that it may take values @xmath28 for a riemann tensor @xmath29 with metric @xmath30 on @xmath31 .
the scalar fields are taken to depend only on time , which is the only choice compatible with the symmetries of flrw spacetimes .
the universe is expanding if @xmath32 and accelerating if @xmath33 .
we need only discuss expanding cosmologies because contracting cosmologies are obtained by time - reversal . in some simple cases ,
the target space has a one - dimensional factor parametrised by a dilaton field @xmath34 , and the potential takes the form @xmath35 for some ` dilaton coupling ' constant @xmath36 and constant @xmath37 .
this model is of special interest , in part because of its amenability to analysis .
the special case for which the dilaton is the _ only _ scalar field was analysed many years ago ( for @xmath38 ) using the observation that , for an appropriate choice of time variable , cosmological solutions correspond to trajectories in the ` phase - plane ' parametrised by the first time - derivatives of the dilaton and the scale factor @xcite .
this method ( which has recently been extended to potentials that are the sum of two exponentials @xcite ) allows a simple visualisation of all possible cosmological trajectories .
moreover , all trajectories for flat cosmologies can be found explicitly @xcite ( see also @xcite ) , and a related method allows a visualisation of their global nature @xcite . it was noted in @xcite that there is both a ` critical ' and a ` hypercritical ' value of the dilaton coupling constant @xmath36 , at which the set of trajectories undergoes a qualitative change . in spacetime dimension
@xmath16 , these values are @xcite @xmath39 below the critical value ( @xmath40 ) there exists a late - time attractor universe undergoing accelerating expansion , whereas only transient acceleration is possible above it .
the hypercritical coupling ( @xmath41 ) separates ` steep ' exponential potentials ( which arise in flux compactifications ) and ` gentle ' exponential potentials ( which arise from hyperbolic compactification , in which case @xmath42 so the potential is still too steep to allow eternal acceleration ) .
one aim of this paper is to generalise this type of analysis to the multi - scalar case . for @xmath14 ,
this has been done already for what could be called a ` multi - dilaton ' model @xcite , and for the multi - scalar model with exponential potential ( [ eq.exppot ] ) @xcite . here
we consider _ all _ cosmological trajectories ( arbitrary @xmath27 ) for an exponential potential of either sign , and for any spacetime dimension @xmath16 .
in particular , we find exact solutions for all flat cosmologies when @xmath43 , following the method used in @xcite for @xmath44 , and the exact phase - plane trajectories for all @xmath27 when @xmath45 .
a more ambitious aim of this paper is to determine what can be said about cosmologies with more general scalar potentials . what kind of model - independent behaviour can one expect , and how generic is the phenomenon of transient acceleration ?
exponential potentials are simple partly because of the power - law attractor cosmologies that they permit , but such simple solutions do not occur for other potentials so other methods are needed . in this paper
, we develop an alternative method of visualising cosmological solutions that applies to _ any _ potential , and we illustrate it by an application to exponential potentials .
as we explain , exponential potentials serve as ` reference potentials ' in determining the late - time behaviour of cosmologies arising in a large class of models with other potentials .
our starting point for the new formalism is the observation @xcite that flat cosmological solutions of gravity coupled to @xmath18 scalar fields with @xmath5 can be viewed as null geodesics in an ` augmented ' target space of dimension @xmath46 with a metric of lorentzian signature ( see @xcite for related results )
. trajectories in this space corresponding to non - flat cosmologies are neither null nor geodesic .
however , we will see that they are projections of null geodesics in a ` doubly - augmented ' target space with a metric of signature @xmath47 for @xmath13 and signature @xmath48 for @xmath49 .
it turns out that the extension of these results to @xmath50 is very simple .
flat cosmologies are again geodesics in the augmented target space but with respect to a conformally rescaled metric ; the geodesic is timelike if @xmath44 and spacelike if @xmath43 .
null geodesics are unaffected by the conformal rescaling and hence continue to correspond to ( @xmath14 ) cosmologies for @xmath5 .
the analysis of acceleration is also simple in this framework : within the lightcone is an ` acceleration cone ' , and if a geodesic enters its acceleration cone then the corresponding universe will accelerate . in the case of a flat target space with an exponential potential , the fixed point
geodesics are just straight lines , and the universe is accelerating if the straight line lies within the acceleration cone .
the further extension to @xmath51 cosmologies is achieved exactly as in the @xmath5 case .
cosmological trajectories are projections of geodesics in the doubly - augmented target space with respect to a conformally rescaled metric of signature @xmath47 for @xmath13 and @xmath48 for @xmath49 .
the geodesic is timelike if @xmath44 , null if @xmath5 and spacelike if @xmath43 .
the plan of this paper is as follows .
section [ sec.mscalar ] reviews , in our conventions , the equations of motion in multi - scalar cosmologies , and we derive a useful alternative criterion for the acceleration of the scale factor . in section [ sec.fixedpoints ] , we consider possible fixed points , concluding that these occur only for exponential potentials , and we determine the nature of these fixed points for arbitrary @xmath16 , @xmath27 and sign of the potential . in section [ sec.phase ]
we discuss the phase - plane trajectories and present some exact @xmath43 solutions . in section [ sec.manifold ]
we develop the interpretation of multi - scalar cosmologies as geodesic motion . in section [ sec.applications ]
we introduce the notion of the acceleration cone , and we show how the cosmologies arising in exponential - potential models fit into the new framework ; other potentials of current interest are also considered .
we summarise in section [ sec.discussion ] .
we begin by briefly reviewing the equations of motion for multi - scalar cosmologies that follow from the action ( [ action ] ) .
these equations will be given in two different forms , distinguished by the choice of time coordinate .
the second form will be useful to our analysis in the next section of possible ` fixed points ' , which correspond to power - law cosmologies in the first form . to simplify the equations we define @xmath52 where @xmath53 is the levi - civit connection for the target space metric @xmath54 , and we introduce the following ` characteristic functions ' of the potential v : @xmath55 these are the components of a 1-form on the target space dual to a vector field @xmath56 . in a vector notation
, the scalar field equation can now be written as @xmath57 where @xmath58 is the hubble function .
the friedmann constraint is @xmath59 = \left(d-1\right)\alpha_c^{-2}k,\ ] ] where the norm @xmath60 ( and hence the inner product ) is the one induced by the target space metric : @xmath61 .
differentiating the friedmann constraint , and using the scalar field equation ( [ eq.sca ] ) , we deduce the ` acceleration equation ' @xmath62 is the ` critical ' constant given in ( [ exponents ] ) .
clearly , acceleration is possible only if @xmath44 .
note too that @xmath63 can not vanish unless either @xmath43 or @xmath64 ( or @xmath5 , @xmath14 and @xmath65 ) so only in these cases can an expanding universe recollapse . for the following section , it will prove useful to rewrite the above equations in terms of a new time coordinate @xmath66 , defined as a function of @xmath67 by the relation @xmath68 note that we have allowed for the possibility that @xmath6 , although @xmath5 is excluded .
we will also set @xmath69 in terms of the variable @xmath70 , the condition @xmath71 for expansion is @xmath72 , while the condition @xmath73 for acceleration is @xmath74 where the overdot denotes differentiation with respect to @xmath75 . in the new time
coordinate , the scalar field equation becomes @xmath76 and the friedmann constraint becomes @xmath77 = \left(d-1\right)\alpha_c^{-2 } k\,.\ ] ] the acceleration equation is now @xmath78 and the condition ( [ accel1 ] ) for acceleration is therefore equivalent to @xmath79 , and hence valid for all @xmath80 .
as expected , it can be satisfied only if @xmath44 .
we now wish to determine whether the system of equations ( [ newscalar ] , [ accelerationeq ] ) subject to the constraint ( [ newfried ] ) , admits any fixed point solutions for which @xmath81 we shall see that these conditions are consistent only for exponential potentials , for which the equations ( [ newscalar ] , [ accelerationeq ] ) become those of an autonomous dynamical system , and we determine the fixed points and their type for all @xmath16 , @xmath27 and for either sign of the potential . for this analysis ,
it is convenient to introduce the quantity @xmath82 recall that @xmath83 is the ` hypercritical ' constant given in ( [ exponents ] ) . given @xmath84 , equation ( [ newscalar ] ) yields @xmath85 where @xmath86 solves the quadratic equation @xmath87 given @xmath88 , equation ( [ accelerationeq ] ) yields @xmath89 \nonumber\\ & = & \left(d-2\right)c^2a^2 + { 1\over2}\left(d-1\right ) \left(d-2\right)\left[\dot\beta^2 - ca^2 \dot\beta \right ] \end{aligned}\ ] ] where we have used ( [ cterms ] ) to get to the second line .
using this equation to eliminate the @xmath90 term from ( [ quadc ] ) we arrive at the quadratic equation @xmath91 there are therefore two types of fixed point : * @xmath92 . using this in ( [ quadc ] ) we deduce that @xmath93 and that @xmath94 this requires @xmath95 and @xmath96 it then follows that @xmath97 this is just the friedmann constraint for @xmath14 , so this type of fixed point occurs only for @xmath14 .
* @xmath98 . using this in ( [ quadc ] )
we deduce that @xmath99 it follows that this type of fixed point can occur only for @xmath44 , and that @xmath100 using this in the friedmann constraint we deduce that @xmath101 we may exclude the possibility that @xmath102 because this yields a @xmath14 fixed point of the type already considered , so @xmath51 and we need @xmath64 for @xmath40 and @xmath103 for @xmath104 . note that @xmath105 for this type of fixed point , so the fixed point cosmology is neither accelerating nor decelerating .
the above results are in complete analogy with those of @xcite for the one - scalar case . to see why
, it should first be appreciated that the fixed point conditions have been derived on the assumption that @xmath106 is constant , and that @xmath107 is covariantly constant .
it then follows from ( [ cterms ] , [ quadc ] ) that @xmath56 must be covariantly constant too .
however , for any space that admits a non - zero covariantly constant vector , there exist coordinates in which this vector is _ constant _ , not just covariantly constant . for coordinates
in which @xmath56 is constant the potential takes the form @xmath108 which is of the form ( [ eq.exppot ] ) with @xmath109 .
moreover , for any target space on which this potential is globally defined , one can find new coordinates such that @xmath110 where @xmath111 is a metric on a ` reduced ' target space with @xmath112 coordinates @xmath113 . in these target space
coordinates , the scalar field equations are @xmath114 and the friedmann constraint is @xmath115 - \left(d-1\right)\dot\beta^2 = k\left(d-1\right)e^{2\left(a\varphi-\beta\right ) } |v_0|^{-1}\ , .\ ] ] these imply the acceleration equation for @xmath116 . if we suppose that the reduced target space is flat then the scalar field equations , together with the acceleration equation , define an autonomous dynamical system : @xmath117 - \dot\beta^2 + a\dot\beta \dot\varphi\ , , \nonumber\\ \ddot\chi & = & a\dot\varphi\ , \dot\chi - \left(d-1\right)\dot\beta\ , \dot\chi\ , .
\end{aligned}\ ] ] note that we may consistently set @xmath118 to recover the equations of the one - scalar model . as there are no fixed points with @xmath119 , the fixed points of the multi - scalar model are the same as those of the one - scalar model .
if the reduced target space is not flat then a connection term must be added to the last equation in ( [ autonomous ] ) .
as this introduces dependence on @xmath113 , the equations no longer define an autonomous dynamical system in just @xmath46 variables .
it is not clear how much difference this makes to the results : note that the connection term has no effect on the stability of fixed points with @xmath118 because it is quadratic in @xmath120 .
however , we will suppose for the analysis to follow that the reduced target space _ is _ flat . we may then assume with no essential loss of generality that it is also one - dimensional ( i.e. @xmath121 ) so we have an autonomous dynamical system for just three variables .
we linearise about a fixed point with @xmath122 and @xmath123 by writing @xmath124 the linearised equations for @xmath125 take the form @xmath126 , where @xmath127 is the @xmath128 matrix @xmath129 the eigenvalues of this matrix determine the nature of the fixed point .
we consider the two types of fixed point in turn : * @xmath14 . in this case
@xmath130 the eigenvalues of @xmath127 are @xmath131 for an expanding universe we must take the top sign ( corresponding , for @xmath44 , to the upper branch of the @xmath14 hyperboloid ) .
then we have a stable node ( all eigenvalues real and negative ) for @xmath40 .
for @xmath132 we have a saddle , with one real - negative and two real - positive eigenvalues ; the instability due to the positive eigenvalues is what leads to the ` flatness problem ' of standard big - bang cosmology .
+ for @xmath43 we may have @xmath133 , in which case the fixed point is an unstable node ( all eigenvalues real and positive ) .
all @xmath13 trajectories start at this unstable fixed point and end at the stable fixed point on the other branch of the @xmath14 hyperboloid ; as @xmath134 at this other fixed point , all @xmath13 universes recollapse to a big crunch singularity .
however , for @xmath135 there is a long quasi - static period on some of these trajectories .
* @xmath51 . in this case
@xmath136 the eigenvalues of @xmath127 are @xmath137 , \qquad \mp 2a/\alpha_c\,,\ ] ] where @xmath138 for one eigenvalue and @xmath139 for another .
take the top sign again . for @xmath140
all eigenvalues are real but one is positive and the others negative , so we have a saddle point . for @xmath42
all eigenvalues are real and negative provided _ either _ @xmath141 _ or _ ( if @xmath142 ) @xmath143 otherwise ( i.e. @xmath142 and @xmath144 ) we have one real negative eigenvalue and two complex eigenvalues with negative real parts , and hence a stable ` spiral - node ' ( a spiral in the @xmath145 plane and a node in the @xmath146 direction ) . note that @xmath147 where the equality occurs for @xmath148 .
thus , for any @xmath16 the @xmath51 fixed point is an unstable saddle if @xmath40 , and is stable for @xmath42 but can be either a node or a spiral - node . for @xmath141
it is always a node , but for @xmath148 it becomes a spiral - node at @xmath41 and for @xmath149 it becomes a spiral - node at @xmath150 . in the supergravity context
we are restricted to @xmath151 , but there is no scalar potential possible in @xmath152 and the only possible potential in @xmath153 is a positive exponential ( with @xmath154 so there is no fixed - point solution ) .
we now have all the information needed to determine the qualitative behaviour of all phase - space trajectories .
we begin with the one - scalar model , obtained by setting @xmath155 .
the trajectories for @xmath44 are sketched in @xcite but , for the reader s convenience , we give a brief summary in words .
all @xmath14 trajectories begin , after a big bang singularity , with a period of kinetic energy domination with effectively vanishing potential .
for @xmath133 they end up approaching a similar late - time phase , but for @xmath156 they approach the @xmath14 fixed - point solution , which is accelerating if @xmath40 and decelerating if @xmath42 . for @xmath40 , all @xmath51 trajectories sufficiently close to a @xmath14 trajectory also approach the @xmath14 fixed point ; in fact , all @xmath13 trajectories have this property .
the behaviour of a @xmath49 trajectory is determined by its relation to the separatrix of the unstable @xmath49 fixed point ( which is the einstein static universe for @xmath157 ) ; those on one side approach the @xmath14 fixed point while those on the other side represent universes that recollapse to a big crunch .
for @xmath42 _ all _ @xmath49 universes recollapse to a big crunch whereas all @xmath13 universes now approach a zero - acceleration milne universe ( the @xmath14 fixed point now being a saddle point ) . for @xmath144
the @xmath13 fixed point is an attractor spiral so that all @xmath13 trajectories spiral around a point of zero - acceleration ; as observed in @xcite , this leads to eternal oscillation between acceleration and deceleration .
for @xmath43 the @xmath14 trajectories were studied in @xcite but there seems to be no complete study of all trajectories for this case .
we will therefore present some results of the one - scalar model for @xmath43 before moving on to discuss modifications that arise in the multi - scalar case .
it is well known that anti - de sitter space is a solution of einstein s equations with a negative cosmological constant .
this space must correspond to one of the cosmological trajectories in a model with @xmath45 ( which implies that @xmath157 ) .
set @xmath158 to get the equations @xmath159\ ] ] where @xmath160 these equations imply that @xmath161 = cx^{2/(d-1)}\ ] ] for constant @xmath162 .
the curves with @xmath163 are the @xmath49 trajectories and those with @xmath164 are the @xmath13 trajectories .
the two branches of the @xmath165 curve , a hyperbola , are the @xmath14 trajectories .
the ads solution corresponds to @xmath166 , and hence @xmath167 ; for this special case we have @xmath168\ ] ] which has the solution @xmath169 , and hence @xmath170 the friedmann constraint implies that @xmath13 , as expected , and @xmath171,\ ] ] and hence @xmath172 this is anti - de sitter space .
in contrast to de sitter space , which can be realised ( for @xmath173 ) as a flrw universe for _ any _ @xmath27 ( in particular as the @xmath14 fixed - point solution ) , anti - de sitter space is realised as an flrw cosmology only for @xmath13 , and it does not correspond to a fixed point cosmology ( at least in the sense of this paper ) .
we now turn to the @xmath14 trajectories for arbitrary @xmath36 . in this case
we have to solve @xmath174 subject to the constraint @xmath175 the constraint is solved by setting @xmath176 and the equation of motion becomes @xmath177.\ ] ] this equation is immediately integrated if @xmath41 .
having found @xmath178 we integrate ( [ xydef ] ) to deduce that @xmath179^{1\over 2\left(d-1\right)}\,,\ ] ] for integration constant @xmath180 ( we have set to zero the other integration constant as it can be absorbed into @xmath37 ) .
if @xmath93 then we introduce the new variable @xmath181 the two branches of the @xmath14 hyperbola correspond to positive and negative @xmath182 ; we take @xmath183 .
we now consider separately the cases @xmath133 and @xmath156 : * @xmath156 . in this case
the equation for @xmath182 is @xmath184,\ ] ] where we recall that @xmath185 is the function of @xmath36 given in ( [ omegadef ] ) .
integration of ( [ xydef ] ) now leads to @xmath186^{{\alpha_c^2\over 2\alpha_h\omega^2}}\ , .
\end{aligned}\ ] ] * @xmath154 . in this case
the equation for @xmath182 is @xmath187 there is a fixed point at @xmath188 , so we must distinguish between @xmath189 and @xmath190 .
one finds that @xmath191 for @xmath189 integration of ( [ xydef ] ) yields @xmath192^{\alpha_c^2\over 2\alpha_h\omega^2}\ , .
\end{aligned}\ ] ] for @xmath190 integration of ( [ xydef ] ) yields @xmath193^{\alpha_c^2\over 2\alpha_h\omega^2}\ , .
\end{aligned}\ ] ] in the multi - scalar case , the ` spectator ' fields @xmath113 have an effect on the scale factor due to the @xmath194 term in the acceleration equation , and in the friedmann constraint ( [ friedchi ] ) .
phase space trajectories with non - zero @xmath120 will depend on the reduced target space metric but , as long as there are no new fixed points , one may expect the qualitative behaviour to be independent of this metric .
let us therefore continue to suppose that the target space ( and hence the reduced target space ) is flat . in this case , all @xmath14 trajectories lie in a hyperboloid in phase space that separates the @xmath13 and @xmath49 trajectories .
this is true for either sign of @xmath195 but we will now suppose that @xmath44 ; in this case , we see from ( [ eq.accpot ] ) that acceleration occurs whenever the phase - plane trajectory enters the region with hyperspherical boundary @xmath196 .
this was observed in @xcite in the context of an analysis of @xmath14 trajectories , but it is also valid for non - zero @xmath27 .
the only exponential models with eternally accelerating cosmological solutions are those for which a fixed point lies inside or on the sphere of acceleration , as happens when @xmath140 .
for all other values of @xmath197 , any acceleration can be at most transient , and only a subset of the trajectories undergo even transient acceleration ; these are the trajectories that pass through the sphere of acceleration .
the more scalar fields there are , the more freedom there is to avoid the sphere of acceleration , so transient acceleration in multi - scalar models is less generic than it is in the single - scalar model . to make this statement
quantitative would require an understanding , along the lines of @xcite , of what the appropriate measure might be on the space of trajectories .
the most significant fact about a @xmath14 fixed point is whether it lies inside or outside the sphere of acceleration .
figure [ fig.k0psdia ] illustrates a few examples in a model with two scalar fields and a positive exponential , scalar potential .
we now develop a method to describe the generic evolution of multi - scalar cosmologies .
it will be convenient to define a new independent variable @xmath198 by @xmath199 although we will still use @xmath70 where this is more convenient .
the friedmann constraint can now be written as @xmath200 and the scalar field equation as @xmath201 these two equations imply the acceleration equation @xmath202 = 0\,.\ ] ] using the friedmann constraint to eliminate @xmath203 from the scalar and acceleration equations we have @xmath204 & = & k\,\alpha_c^{-2}\left(d-1\right)e^{-2\beta}\ , \mathbf{a}\ , , \nonumber\\ \partial_t^2\gamma + \alpha_h \left|\partial_t{{\bm{\phi}}}\right|^2 & = & k\,\alpha_c^{-2}\alpha_h e^{-2\beta}\ , .
\label{accgamma}\end{aligned}\ ] ] before considering the general case ( arbitrary @xmath27 and arbitrary @xmath195 ) , we first discuss the special cases of @xmath5 ( but arbitrary @xmath27 ) and @xmath14 ( but arbitrary @xmath195 ) .
we may consider the @xmath205 variables @xmath206 to be maps from the cosmological trajectory to an ` augmented ' target space . in this notation ,
the friedmann constraint for @xmath5 can be written as @xmath207 where @xmath208 is a _ lorentz - signature _ metric on the augmented target space . using the friedmann constraint to eliminate the @xmath209 term in the second equation in ( [ accgamma ] ) yields @xmath210 if this is taken together with the scalar field equation then the two combine into the single ( albeit coordinate - dependent ) equation @xmath211 for @xmath14 this is the equation for a geodesic in a non - affine parametrisation . in terms of the the new time variable @xmath212 defined by @xmath213 we have @xmath214 for @xmath14
this is the equation of an affinely parametrised geodesic .
although the cosmological trajectory in the augmented target space is not a geodesic for @xmath51 , it can be viewed as the projection of a geodesic in a ` doubly - augmented ' target space of dimension @xmath215 that is foliated by hypersurfaces isometric to the augmented target space .
let @xmath216 be the @xmath215 coordinates .
we take the metric to be @xmath217 where @xmath218
as we now have one more variable we also need another equation .
we take this to be the ` projection equation ' @xmath219 with this choice , the friedmann constraint becomes @xmath220 and the combined scalar field and acceleration equations ( [ knonzero ] ) are equivalent to the single geodesic equation @xmath221 thus , cosmological trajectories for @xmath5 are null geodesics in the doubly - augmented target space .
note that the signature of this space is lorentzian if @xmath103 but non - lorentzian ( with two ` times ' ) if @xmath64 .
because @xmath222 is not constant , the motion is not restricted to a single hypersurface of constant @xmath222 and this accounts for the fact that the projection of the motion onto one such hypersurface is not geodesic .
this argument fails for @xmath14 , for which the projected motion _ is _ geodesic , because the metric on the doubly - augmented target space is degenerate when @xmath14 . the friedmann constraint for @xmath14 , but arbitrary @xmath195 , can be written as @xmath223 thus , the cosmological trajectory in this space is timelike for @xmath44 , null for @xmath5 and spacelike for @xmath43 .
as we have seen , it is a null _ geodesic _ when @xmath5 .
when @xmath50 the cosmological trajectories are no longer geodesic with respect to the metric @xmath224 , but they _ are _ geodesics with respect to the conformally - rescaled metric @xmath225 where the conformal factor is @xmath226 in fact , one finds that the equations ( [ basiceqs ] ) are equivalent , for @xmath14 , to the equation @xmath227 where @xmath228 is the levi - civit connection for @xmath229 , and when @xmath5 as @xmath230 is undefined in this case . ]
@xmath231 this is the equation for a geodesic in a non - affine parametrisation ; noting that @xmath232 we deduce that the geodesics are affinely parametrised by a new time - coordinate @xmath233 for which @xmath234 in other words , @xmath235 the friedmann constraint now takes the form @xmath236 notice that @xmath233 differs from @xmath212 as defined ( for @xmath5 ) by ( [ tprime ] ) .
in fact , @xmath237 this difference occurs because the affine parameter of a null geodesic is affected by a conformal rescaling of the metric , and the metric we are now considering is the conformally rescaled one @xmath238 ; clearly a null curve that is geodesic with respect to @xmath239 is also geodesic with respect to any other conformally equivalent metric , such as @xmath238 , but the affine parametrisation will differ in general .
in contrast , when @xmath50 the cosmological trajectories in the augmented target space are geodesics with respect to a unique metric ( up to homothety ) in the class of metrics that are conformally equivalent to @xmath239 , and this metric is @xmath238 . the steps that led to ( [ affinegeo ] ) for @xmath14 lead , for general @xmath27 , to the equation @xmath240
although this is not the equation of a geodesic when @xmath51 , it is the projection of a geodesic in the doubly - augmented target space with respect to the conformally rescaled metric @xmath241 where the conformal factor @xmath242 is as given in ( [ conformalfac ] ) , and @xmath243 is as given in ( [ doubleaug ] ) . from ( [ phistar ] ) and ( [ ttprime ] ) we deduce that @xmath244 given this , one then finds that ( [ general ] ) is equivalent to the geodesic equation @xmath245 and that the friedmann constraint is @xmath246 thus
, the general cosmological trajectory is a geodesic in the doubly - augmented target space , one that is timelike for @xmath44 , null for @xmath5 and spacelike for @xmath43 .
we now consider some applications of the formalism just developed . as we are particularly interested in accelerating cosmologies
we first consider the implications of the acceleration condition .
we will then see how to interpret the fixed point cosmologies that arise for exponential potentials .
this turns out to be useful when considering the asymptotic behaviour of more general potentials .
we first show how the condition for acceleration ( [ eq.accpot ] ) acquires a geometrical meaning in the new framework .
we will now assume that @xmath44 , as acceleration can not otherwise occur , and before proceeding we note that the friedmann constraint ( [ newfried ] ) can be written as @xmath247 we now introduce the following ` acceleration ' metric on the augmented target space , @xmath248 and the corresponding conformally rescaled metric @xmath249 where the conformal factor is the same as before .
noting that @xmath250 one finds that @xmath251 using the friedmann constraint in the form ( [ friedform ] ) to eliminate the @xmath252 term , we deduce that @xmath253 + { k\over 2\alpha_c^{2}v}\ , e^{-2\beta}\,.\ ] ] recalling that acceleration occurs when @xmath254 , we see that the condition for acceleration is equivalent , for @xmath14 , to @xmath255 geometrically , this states that a universe corresponding to a timelike geodesic trajectory is accelerating when its tangent vector lies within a subcone of the lightcone defined by the acceleration metric on the augmented target space .
as one might expect , an analogous result holds for @xmath51 in terms of the doubly - augmented target space . in this case , the acceleration metric is @xmath256 .
\end{aligned}\ ] ] using the projection equation ( [ phistarhat ] ) one finds that @xmath257 , \ ] ] where @xmath258 is the conformally rescaled acceleration metric .
the condition for acceleration is therefore equivalent to @xmath259 this states that a universe is accelerating when the tangent to its geodesic trajectory in the doubly - augmented target space lies with an acceleration subcone of the lightcone .
we now discuss how the cosmologies arising for an exponential potential of the form ( [ exppot2 ] ) fit into the new geometrical framework . for reasons given earlier we assume in this case that the target space is flat , in which case the augmented target space metric @xmath239 is also flat , and hence @xmath238 is conformally flat , with conformal factor @xmath260.\ ] ] our first task is to determine the trajectories in the augmented target space that are associated with the fixed point solutions .
consider first the @xmath14 fixed point .
this solution has the property that the linear function @xmath261 is time - independent : @xmath262 . in other words , @xmath263 is a vector tangent to the hyperplanes of constant @xmath264 . for @xmath265
there is only one such direction , which is determined by the gradient of @xmath242 .
thus , @xmath266 for @xmath265 .
in fact , this remains true for arbitrary @xmath18 as a direct calculation shows : @xmath267 a fixed point solution for @xmath51 has the property that the linear function @xmath268 is time - independent . as surfaces of constant @xmath269 are hyperplanes , a @xmath51 fixed point trajectory is _ also _ a straight line , as a direct computation confirms : @xmath270 thus , fixed point solutions are ( particular ) straight - line trajectories in the augmented target space .
in addition , as follows from our earlier results , the @xmath14 straight - line trajectories are geodesics with respect to the metric @xmath271 , whereas this is not true for @xmath51 .
this difference can be understood as follows : as we have assumed a flat target space , the metric @xmath239 is flat and _ all _ straight lines are geodesics with respect to it .
however , the relevant metric is @xmath238 , and the only geodesic of @xmath239 that is also a geodesic of @xmath238 is the line of steepest descent of the function @xmath242 ; as we have seen , this is precisely the direction of the @xmath14 fixed point trajectory .
note that the line determined by ( [ direction.nonzerok ] ) is a generator of the acceleration cone ( as expected from the zero - acceleration property of the @xmath51 fixed point ) .
more generally , if a straight line trajectory lies within the acceleration cone then it corresponds to an eternally accelerating universe , and if it lies outside the acceleration cone ( @xmath42 ) then it corresponds to an eternally decelerating universe . in the latter case
it might also lie outside the ` lightcone ' ( @xmath133 ) , in which case the geodesic corresponds to the @xmath14 fixed point for @xmath43 .
note that a change in the value of the constant vector @xmath56 effects a lorentz transformation of the @xmath14 straight - line trajectory in the augmented target space .
such a transformation can take any timelike line into any other timelike line , and any spacelike line into any other spacelike line , but it can not take a spacelike line to a timelike one or vice versa ( as expected from the fact that this requires a change of sign of the potential ) . also , it can not take a timelike or spacelike line into a null line , a fact that is consistent with the absence of a fixed - point solution for @xmath41 .
generic geodesics , corresponding to generic @xmath14 cosmologies , are not straight lines in the extended target space , but they still have a simple description . the regime where the solutions are dominated by kinetic energy is given by geodesics with tangent vectors having large @xmath272 .
this means that the generic geodesics start out in null directions , as could be anticipated from the fact that @xmath14 cosmologies are null geodesics when @xmath5 .
let us now follow the subsequent evolution for the @xmath44 case . as the potential becomes more important , the geodesics bend into the timelike cone , ultimately approaching the timelike straight line corresponding to the fixed point solution if @xmath156 . on the other hand ,
if @xmath273 , then the geodesic ultimately approaches a null straight - line geodesic .
for the single - scalar case , this behaviour is shown in figure [ fig.k0geodesics ] .
cosmologies accelerate precisely when their corresponding geodesics bend their tangent vectors into the acceleration cone . for @xmath51 ,
cosmological trajectories are no longer geodesics in the augmented target space but they are projections of geodesics in the doubly - augmented target space . the vector tangent to the geodesic corresponding to a @xmath51 fixed point is @xmath274 where we have used ( [ phistarhat ] ) to get the last entry . note that this is _ not _ a constant @xmath215-vector ( even allowing for the fact that @xmath275 is constant at the fixed point ) so the fixed point solution is not a straight line in the doubly - augmented target space , as was to be expected because its metric is neither flat nor conformally flat . however , using ( [ fixedcurve ] ) , and then ( [ keq ] ) , one finds that @xmath276 = 0\,.\ ] ] this again confirms the zero acceleration of the @xmath51 fixed point cosmologies . the asymptotic ( late - time ) behaviour of a large class of cosmologies in models with rather general potentials can be determined by comparison with the asymptotic behaviour of cosmologies in models with exponential potentials .
to demonstrate this , we discuss the case of flat cosmologies with positive potentials . in this case
, the scalar fields must approach @xmath277 at late times , where some components of the constant @xmath18-vector @xmath278 may be infinite .
the characteristic functions of the potential in this limit are also constant : @xmath279 the absolute value @xmath280 determines the late - time behaviour , by comparison with the critical and hypercritical exponents : if @xmath281 , then there will be late - time acceleration , with the trajectory approaching an approximate accelerating attractor ; obvious examples are models for which @xmath195 has a strictly positive lower bound , so most models of interest will be those for which @xmath195 either tends to zero or has a minimum at zero . for @xmath282 , there is late - time deceleration corresponding to the existence of an approximate late - time decelerating attractor .
the asymptotic geodesics in both these cases are timelike . in contrast , for @xmath283 the timelike geodesic asymptotes a null ( and hence decelerating ) trajectory . to illustrate the above observations , we consider the following single - scalar examples . * inverse power law potentials : @xmath284 these potentials have been much studied in quintessence models ( see e.g. @xcite ) .
they also arise from dynamical breaking of supersymmetry @xcite . with @xmath285 ,
one finds @xmath286 .
this implies late - time acceleration , _ even though the potential tends to zero at late - times_. * that any other value of @xmath287 may occur is illustrated by the potential @xmath288 for which @xmath289 if we suppose that @xmath290 at late times . for such cosmologies the asymptotic behaviour is the same as it would be in the exponential potential model with dilaton coupling constant @xmath36 .
* it may happen that at late times the scalar fields become trapped near a minimum of the potential .
near such a minimum , which we may assume to occur at the origin of field space , the potential takes the form @xmath291 for integer @xmath292 .
for positive @xmath293 one finds @xmath294 which tends to zero ( i.e. @xmath286 ) as @xmath295 for any @xmath296 .
this implies late - time acceleration , due to the effective cosmological constant @xmath293 .
however , we may now consider the limit @xmath297 , in which case @xmath298 , implying that the late - time trajectory is null , and hence decelerating .
in this paper we have developed a geometrical method for the classification and visualisation of homogeneous isotropic cosmologies in multi - scalar models with an arbitrary scalar potential @xmath195 .
the method involves two steps . in the first step ,
the target space parametrised by the @xmath18 scalar fields is augmented to a larger @xmath205-dimensional space of lorentzian signature , the logarithm of the scale factor playing the role of time .
this is reminiscent of the role of the scale factor in mini - superspace models of quantum cosmology , especially if one views the scalar fields of our model as moduli - fields for extra dimensions . in this case
, the lorentzian metric on the augmented target space is the one induced from the wheeler - dewitt metric on the space of higher - dimensional metrics .
this was the point of view adopted in @xcite , where it was also observed that when @xmath5 all flat ( @xmath14 ) cosmologies are null geodesics in this metric ( an observation that goes back to work of dewitt on kasner metrics @xcite ) .
when @xmath50 , flat cosmologies again correspond to trajectories in the augmented target space but they are neither null nor geodesic
. however , as we have shown , _ all _ flat cosmologies are geodesics with respect to a _ conformally rescaled _ metric on the augmented target space , the conformal factor depending both on the scale factor and the potential .
the conformal rescaling has no effect on null geodesics , of course , so flat cosmologies are still null geodesics whenever @xmath5 , but now the trajectories of flat cosmologies are timelike geodesics when @xmath44 and spacelike geodesics when @xmath43 .
this is true not only for a potential of fixed sign but also for one that changes sign ; in this case the tangent to the geodesic becomes null just as the scalar fields take values for which @xmath5 .
the second step , which is needed for non - flat ( @xmath299 ) cosmologies , is to further enlarge the target space to a ` doubly - augmented ' target space foliated by copies of the augmented target space .
one can choose the metric on this space , of signature @xmath47 for @xmath13 and @xmath48 for @xmath300 and degenerate for @xmath14 , such that geodesics yield ( on projection ) cosmological trajectories for _ any _ @xmath27 when projected onto a given hypersurface , and the geodesics are again timelike , null or spacelike according to whether @xmath195 is positive , zero or negative .
this general construction thus includes all the previous ones as special cases .
accelerating cosmologies have a simple interpretation in this new framework . within the lightcone ( defined by the metric with respect to which the trajectories are geodesics )
there is an acceleration subcone .
a given trajectory corresponds to an accelerating universe whenever its tangent vector lies within the acceleration cone . in particular
, a flat cosmology undergoes acceleration whenever its geodesic trajectory bends into the acceleration cone within the lightcone of the augmented target space .
this can happen only if the trajectory is timelike , which it will be if the potential is positive .
we have also presented in this paper a complete treatment ( complementing many previous studies ) of cosmological trajectories , for any @xmath27 and any spacetime dimension @xmath16 , for the special case of simple exponential potentials .
these potentials are characterised by a sign ( the potential may be positive or negative ) and a ( dilaton ) coupling constant @xmath36 ( the magnitude of an @xmath18-vector coupling constant @xmath56 ) . as has long been appreciated @xcite , cosmological solutions in such models
correspond , for an appropriate choice of time parameter , to trajectories of an autonomous dynamical system in a ` phase space ' parametrised by the time - derivatives of the scale factor and scalar fields .
the qualitative features of these trajectories ( which should not be confused with trajectories in the ` augmented target space ' parametrized by the fields themselves ) are determined by the position and nature of any fixed points , which are of two types .
there is always a @xmath14 fixed point unless @xmath41 ( the ` hypercritical ' value of @xmath36 defined by this property ) but it occurs only for @xmath44 if @xmath133 and only for @xmath43 if @xmath133 .
there is also a @xmath51 fixed point if @xmath44 , which coincides with the @xmath14 fixed point when @xmath102 , where @xmath301 is the ` critical ' value of @xmath36 ( defined as the value at which fixed point cosmologies have zero acceleration ) .
these fixed point cosmologies have a very simple interpretation as trajectories in the augmented target space : they are straight lines . in the @xmath14 case
these lines are also geodesics . for @xmath43
these geodesic lines lie outside the lightcone . for @xmath44
they lie inside the lightcone , but may lie either inside or outside the acceleration cone . for models obtained by ( classical ) compactification from a higher dimensional theory without a scalar field potential , the @xmath14
geodesic line always lies outside the acceleration cone , so only transient acceleration is possible in these models . for a single scalar field
the transient acceleration is generic in the sense that it is a feature of all @xmath14 trajectories when @xmath133 and of some when @xmath156 .
in contrast , in the multi - scalar case , there are trajectories that correspond to eternally decelerating universes even when @xmath133 , so transient acceleration is less generic for more than one scalar field .
although exponential potentials are of limited phenomenological value , they are important as ` reference potentials ' in determining the asymptotic behaviour of cosmologies in models with other potentials , such as inverse power potentials . essentially , any potential that falls to zero slower than the critical exponential potential ( as do inverse power potentials ) will lead to late - time eternal acceleration .
any potential that falls to zero faster than the hypercritical exponential potential will have a late - time behaviour that is approximately that of a model with zero potential .
of course , there is no evidence for the existence of cosmological scalar fields , but there _ is _ evidence that the expansion of the universe is accelerating and hence for dark energy .
whatever produces this energy , it seems reasonable to suppose that it can be modeled by scalar fields .
if so , it is possible that future observations may be interpreted as telling us something about the potential energy of these fields . in view of our current complete ignorance of what this potential
might be , we have tried , as much as possible , to understand generic properties , and we hope that the methods developed here will be of further use in this respect .
the interpretation of solutions of gravitational theories as geodesics in a suitable metric space has a considerable history of which we were mostly unaware at the time of writing this paper .
the geodesic interpretation of @xmath14 cosmologies that we have presented is closely related to the maupertuis - jacobi principle of classical mechanics .
relatively recent work on this topic includes @xcite
. a particle mechanics formulation of the geodesic interpretation described here , including our proposed extension to @xmath51 , has recently been found , and applied to a model in which cosmological singularities correspond to horizons in the augmented target space @xcite . | for gravity coupled to @xmath0 scalar fields , with arbitrary potential @xmath1 , it is shown that all flat ( homogeneous and isotropic ) cosmologies correspond to geodesics in an @xmath2-dimensional ` augmented ' target space of lorentzian signature @xmath3 , timelike if @xmath4 , null if @xmath5 and spacelike if @xmath6 .
accelerating cosmologies correspond to timelike geodesics that lie within an ` acceleration subcone ' of the ` lightcone ' .
non - flat ( @xmath7 ) cosmologies are shown to evolve as projections of geodesic motion in a space of dimension @xmath8 , of signature @xmath9 for @xmath10 and signature @xmath11 for @xmath12 .
this formalism is illustrated by cosmological solutions of models with an exponential potential , which are comprehensively analysed ; the late - time behaviour for other potentials of current interest is deduced by comparison . |
in the course of an extensive search for very wide , common proper motion companions to nearby hipparcos stars from the nomad catalog @xcite , we came across an optically faint and red star at an angular separation of @xmath5 , position angle @xmath6 from the active and rapidly rotating g9v star hip 115147 ( bd + 78 826 , hd 220140 , v368 cep ) .
this faint star was subsequently identified with lspm j2322 + 7847 , a candidate nearby low - mass dwarf detected by @xcite from the lspm - north catalog @xcite .
the original identification of this star as one with significant proper motion traces to luyten ( 1967 ) where it was reported as a magnitude 17 object with @xmath7 mas yr@xmath8 at a position angle of @xmath9 and assigned the name lp12 - 90 . in the present paper
we present new @xmath2 photometry of hip 115147 , its known visual companion hip 115147b and lspm j2322 + 7847 , and obtain preliminary trigonometric parallax astrometry for the latter companion .
then we discuss the possible young age and origin of this interesting triple system .
following the identification of lspm j2322 + 7847 as a potential widely separated common proper motion companion to hip 115147 , @xmath2 photometry on the cousins system was obtained for the fainter star on ut 30 august 2005 using the 1.0-meter reflector at the flagstaff station .
the photometry was calibrated with an instrumental zero point term and a first order airmass term .
the calibration field was the standard field pg2213 - 006 from @xcite .
additional photometric observations were subsequently obtained on ut 16 and 17 june 2007 when the individual components of the brighter system hip 115147ab ( sep . 11 " ) were also measured .
the photometric results are presented in table 1 along with @xmath10 photometry extracted from 2mass , proper motions from nomad and parallax determinations .
the estimated uncertainties in the @xmath2 measures are @xmath11 mag in the case of lspm j2322 + 7847 and @xmath12 - 0.04 mag for the hip 115147 components where the short exposure times introduced additional error from scintillation and shutter timing .
since the august 2005 photometry indicated that lspm j2322 + 7847 was most likely an m - dwarf at approximately the same distance as hip 115147 , it was added to the trigonometric parallax program at usno s flagstaff station . through june 2007
, a total of 66 acceptable ccd observations have been accumulated on this field , covering an epoch range of 1.65 years .
the same tek2k ccd , observational procedures , and reduction algorithms have been employed as summarized in @xcite . using a total of 29 reference stars , the current preliminary solution yields @xmath13 mas .
this solution appears to be very robust , with the separate solutions for parallax in the ra and dec directions in very satisfactory agreement ( @xmath14 mas versus @xmath15 mas , respectively ) .
correction to absolute parallax was performed using usno @xmath16 photometry for the individual reference stars along with a calibrated @xmath17 versus @xmath18 relationship to derive a mean photometric parallax of @xmath19 mas for the 29 star ensemble . together
this then translates to @xmath20 mas for lspm j2322 + 7847 .
the star hip 115147 was identified with the bright x - ray source h2311 + 77 , which led @xcite to suggest rs cvn - type activity .
it was shown later on that the star is not evolved , and that it is not a short - period spectroscopic binary , justifying the currently accepted classification as a very young , naked " post - t tauri dwarf @xcite .
being one of the most powerful x - ray emitters in the solar neighborhood @xcite at @xmath21 ergs s@xmath8 @xcite , the star has the same space velocity as the local young stream ( or local association , @xcite ) .
this association limits its age to @xmath22 myr .
this stream includes isolated stars , groups and associations of diverse ages , some as young as 1 myr ( e.g. , in lupus and ophiuchus ) .
therefore , the assignment to this stream by itself does not lend a more precise estimation of age .
an x - ray luminosity of @xmath23 is typical of weak - lined tt stars in taurus - auriga - perseus , but significantly larger than that of classical tt stars ; if anything , it points at an age older than a few myr .
hip 115147 is listed as variable star v368 cep @xcite .
the slight variability allowed @xcite to determine the period of rotation of this star , 2.74 d. the fast rotation is responsible for the high degree of chromospheric and coronal activity .
the primary star is identified as the extreme ultraviolet source euve j@xmath24 with strong detections at @xmath25 band , as well as at 250 ev in x - rays @xcite .
hip 115147 is one of the 181 extreme - ultraviolet sources in the rosat wfc all - sky bright source survey identified with late - type main - sequence stars @xcite , with high signal - to noise detections in both @xmath26-@xmath27 and @xmath28-@xmath29 passbands .
an unusually high level of chromospheric activity of @xmath30 was determined by @xcite ; a spectral type k2v is also specified in the latter paper as opposed to g9v given in the simbad database .
since the rate of rotation diminishes fairly quickly with age in single stars , so does x - ray luminosity , and open clusters older than @xmath31 persei ( 50 myr ) are usually more quiescent than the youngest ones ( ic 2602 and ic 2391 ) .
the high degree of chromospheric and extreme ultraviolet activity suggests a very young age , possibly less than 20 myr .
v368 cep is more powerful in x - rays than an average k - type pleiades member by a factor of 10 , indicating an age less than 100 myr .
finally , the equivalent width of at @xmath32 is smaller than the upper limit for the pleiades by @xmath33 according to @xcite , which points at an age similar to the pleiades or older .
however , the lithium surface content is a poor estimator of age for young stars because of the large intrinsic spread of this parameter in stars of the same mass and age .
we are left with the most reliable and frequently used method of age estimation by model isochrones . using the photometric data from the literature and our own ( table 1 ) , we plot hr diagrams in the 2mass @xmath34 and @xmath3 , and @xmath35 passbands in figs .
[ jk.fig ] and [ vk.fig ] . for reference
, the theoretical isochrones at @xmath36 , @xmath37 and @xmath38 myr are drawn from @xcite . according to the data in @xcite derived from strmgren _
uvby@xmath39 _ photometry , the star hip 115147 has a markedly subsolar metallicity at [ fe / h]@xmath40 .
this determination has to be taken with caution , because the photometrically derived metallicities are sensitive to the color index @xmath41 ( b. nordstrm 2007 , priv .
comm . ) , which may be abnormally high for chromospherically active stars @xcite . a wider range of metal abundances is supplied in the evolution models of @xcite , and we present also a 25 myr isochrone for @xmath42 from their models in fig . [ jk.fig ] with a dashed line .
we find that the siess et al .
50 myr isochrone ( @xmath43 ) provides the best fit for all three alleged comoving components ( marked with crossed circles ) , including the secondary m dwarf hip 115147 b , separated by @xmath44 from the primary @xcite .
this pair is listed in the wds @xcite as wds 23194@xmath45 ( lds 2035 , discovered by luyten in 1969 ) , at separation @xmath46 and position angle @xmath47 .
the relative position did not change significantly between the first ac2000.2 measures taken in 1901 and the latest epoch 2000 in 2mass ( w. hartkopf & g. wycoff 2007 , priv .
comm . ) , confirming that the inner pair is physical .
the small deviation of the faintest component lspm j2322 + 7847 from the siess 50 myr isochrone appears to arise mostly from a sudden twist of the isochrone at the latest data point corresponding to mass @xmath48 .
we are not in a position to discuss if this twist has a certain physical meaning , but the overall match is good , and the estimated mass of the star is roughly @xmath48 .
the younger isochrones for 16 and 7 myr from @xcite lie significantly higher in fig .
[ jk.fig ] than the matching 50 myr isochrone .
the 25 myr isochrone from @xcite at a lower metallicity provides a poor fit to our stars , predicting much bluer colors , but these models may be strongly biased for young ages as was found by @xcite for solar - type stars in the alpha persei open cluster at 52 myr .
the members of the very young twa association are shown in fig .
[ jk.fig ] with open circles .
these stars are certainly much younger than hip 115147 and its companions .
they appear to be brighter and redder than the 7 myr isochrone corresponding to the probable age of this group ( see also * ? ? ?
* ) , but this deviation may be the result of the near - infrared @xmath49 and @xmath34 excess commonly observed in classical t tauri stars and attributed to a hot inner rim in their dusty accretion disks @xcite . since our stars under investigation have no accretion disks and their metal abundance may be lower , we need to find a young star with similar parameters .
it was pointed out by @xcite that a few young stars currently in the solar neighborhood , that traveled from the vicinity of the ophiuchus star forming region , share the moderately poor metal abundance of hip 115147 as determined from the _
uvby@xmath39 _ photometry , probably affected by the high degree of chromospheric activity .
one of these stars is the extremely active dwarf pz tel , a likely member of the @xmath39 pic associations estimated to be 12 - 20 myr old .
this star is indicated with an open square in fig .
[ jk.fig ] .
its position matches the 16 myr , @xmath43 isochrone quite well .
pz tel is significantly brighter than hip 115147 in @xmath3 having approximately the same color ; thus , the latter star and its companions are probably older than 16 myr .
it is verified in fig .
[ vk.fig ] that this difference in @xmath50 between the two stars is not originating in a @xmath3-band excess , since both hip 115147 and lspm j2322 + 7847 match best the 52 myr , @xmath43 isochrone in a @xmath50 versus @xmath51 diagram as well . the latter star is brighter in @xmath3 than the empirically determined main sequence for field dwarfs from @xcite by 1 mag . if this star were an unrelated old m6 dwarf , its distance would be only 12 pc .
our trigonometric parallax ( table 1 ) yields a distance @xmath52 pc .
a number of stars with outstanding signs of activity and young age scattered in the solar vicinity can be traced back to their places of origin in , or close to , the ob associations in the sco - cen complex @xcite .
the star hip 115147 has come from the vicinity of the molecular cloud ldn 1709 in the ophiuchus star forming region , flanking the sco - cen complex at @xmath53 @xcite .
the closest approach to the estimated center of that association is 10.7 pc , while the separation today is 170 pc .
the closest approach 16 myr ago saw the star flying by at a relative velocity of 10.8 km s@xmath8 .
if hip 115147 was born in the ophiuchus association , its probable age should be 16 myr , considerably younger than the previous isochrone analysis and the lithium abundance suggest .
because of the high departing velocity , it is more likely that this close fly - by is a chance occurrence .
finally , we note that the difference in proper motion between hip 115147 and lspm j2322 + 7847 , which appears to be statistically significant , may be caused by a systematic error of 5 to 10 mas yr@xmath8 in the nomad proper motions for very faint stars .
the formal errors specified in nomad may be strongly underestimated for this optically dim star , whose position at the mean epoch 1979 is based on schmidt plates . the proper motion difference , if it is true , implies that the two stars are not gravitationally bound as a truly multiple system .
this is commonly observed among very young stars in the solar neighborhood ; for example , the common proper motion pair at mic and au mic ( approximately 10 myr old ) are separated by at least 0.23 pc and have proper motions different by @xmath54% , but their physical association is not in doubt . at the suggested age ( 20 - 50 myr ) , the hip 115147 companions are likely to stay close to each other as a moving group .
more accurate absolute proper motions for both companions and spectroscopic radial velocities are required to clarify the dynamical status of the system .
the currently available evidence is consistent with lspm j2322 + 7847 being one of the youngest , later - type stars in the near solar neighborhood .
we thank t. tilleman and c. dahn for taking some of the parallax images .
the research described in this paper was in part carried out at the jet propulsion laboratory , california institute of technology , under a contract with the national aeronautics and space administration .
this research has made use of the simbad database , operated at cds , strasbourg , france ; and data products from the 2mass , which is a joint project of the university of massachusetts and the infrared processing and analysis center , california technology institute , funded by nasa and the nsf .
ra j2000 & 23 19 26.632 & 23 19 24.53 & 23 22 53.873 + dec . j2000 & + 79 00 12.67 & + 79 00 03.8 & + 78 47 38.81 + @xmath55 & @xmath1 & & @xmath0 + proper motion & @xmath56 & @xmath57 & @xmath58 + @xmath59 & 8.60 & 13.75 & 17.99 + @xmath35 & 7.73 & 12.24 & 16.16 + @xmath60 & 7.17 & 11.04 & 14.61 + @xmath61 & 6.76 & 9.54 & 12.54 + @xmath34 & 5.90 & 8.04 & 10.42 + @xmath62 & 5.51 & 7.39 & 9.84 + @xmath3 & 5.40 & 7.20 & 9.52 + | we report a late m - type , common proper motion companion to a nearby young visual binary hip 115147 ( v368 cep ) , separated by 963 arcseconds from the primary k0 dwarf .
this optically dim star has been identified as a candidate high proper motion , nearby dwarf lspm j2322 + 7847 by lpine in 2005 .
the wide companion is one of the latest post - t tauri low mass stars found within 20 pc .
we obtain a trigonometric parallax of @xmath0 mas , in good agreement with the hipparcos parallax of the primary star ( @xmath1 mas ) . our @xmath2 photometric data and near - infrared data from 2mass are consistent with lspm j2322 + 7847 being brighter by 1 magnitude in @xmath3 than field m dwarfs at @xmath4 , which indicates its pre - main sequence status .
we conclude that the most likely age of the primary hip 115147 and its 11-arcsecond companion hip 115147b is 20 - 50 myr .
the primary appears to be older than its close analog pz tel ( age 12 - 20 myr ) and members of the twa association ( 7 myr ) .
[ firstpage ] |
the electronic band structure and the related exciton - polariton transitions in wurtzite semiconductors were investigated in countless experimental and theoretical studies over the past five to six decades . indeed for zno important contributions in the field were already published in the early 1960s @xcite .
nevertheless , fundamental properties such as the ordering of the valence bands ( vbs ) and the influence of stress on the exciton - polariton transitions remain controversial to the present day .
magnetic fields and external pressure constitute powerful tools to obtain detailed information about electronic band structures and optical transitions by studying the exciton - polariton fine structure .
in addition , in recent years , significant advances in computational power and theoretical algorithms have enabled electronic - structure calculations , including quasiparticle ( qp ) effects and electron - hole interaction which reproduce experimental results to a high degree of precision .
consequently , it is now possible to conduct complex calculations taking into account , for instance , the influence of stress on the electronic band structure and the exciton - polariton transitions in direct comparison to experimental results .
the long - standing disagreement of the valence band ordering in zno is one important example for the necessity to comprehend the effects of built - in strain and external stress on the electronic band structure and optical transitions in detail .
recent magneto - optical measurements of free and bound excitons provide strong evidence that the topmost a valence band has @xmath2 symmetry @xcite .
these results are in accordance with _ first - principles _ qp band - structure calculations @xcite as well as a multitude of earlier theoretical and experimental works supporting the valence band ordering originally proposed by thomas in 1960.@xcite by contrast , several publications exist which postulate @xmath3 symmetry for the a valence band ( for a summary of the relevant literature see e.g. refs . and ) .
while some of the conflicting results have been resolved@xcite , the important question remains if the presence of strain ( or the application of external stress ) that preserves the wurtzite structure of zno ( i.e. uniaxial pressure along the @xmath4 axis , biaxial stress perpendicular to the @xmath4 axis , and hydrostatic pressure ) may result in a reversal of the valence band ordering and thus could explain the different assignments in the literature .
such a reversal of the a and b vb symmetry as function of strain was discussed e.g. for the wurtzite iii - v semiconductors gan @xcite and aln@xcite .
et al . _
@xcite first predicted a crossing of the upper two vbs in gan for biaxial in - plane strain .
this prediction was supported by shikanai _
_ et al.__@xcite and more recently by fu _ _ et al.__@xcite who reported a reversal of the vb ordering for uniaxial strain values of @xmath5 and @xmath6 , respectively .
_ et al.__@xcite and fu _ _ et al.__@xcite calculated an exchange of the upper two vb characteristics in aln for uniaxial strain values of @xmath7 and @xmath8 , respectively . in the case of zno , gil _
et al . _ suggested a reversal of the a and b valence band ordering for a biaxial compressive stress of @xmath9 kbar which corresponds to a uniaxial part of the strain tensor of @xmath10 and thus concluded that the vb ordering in zno is quite sensitive to strain.@xcite however , the authors had to rely on exciton energies reported in the literature which included only one work by butkhuzi _ _ et al.__@xcite with sufficiently shifted free exciton energies to suggest a large in - plain strain and a strain - related reversal of the vb ordering
. a closer look into this work reveals several major problems which render the reported transition energies in ref .
questionable for a strain analysis : ( i ) the luminescence spectra were acquired at 77 k instead of 4 k resulting in a shift of the transition energies in accordance with the temperature dependence of the band gap @xcite , ( ii ) the luminescence was excited by a pulsed nitrogen laser with high pulse energy leading to an excitation density related shift of the observed exciton lines@xcite , and ( iii ) an inaccurate proportionality factor was used to convert the wavelength values ( nm ) into energy ( ev ) resulting in energy values which are too small by about 2.5 mev .
the combination of these effects leads to significant deviations of the exciton energies and consequently results in misleading strain values and conclusions in the experimental part of ref .
this situation provides a strong motivation to revisit the stress dependence of the valence band ordering in zno in detail .
the large research interest in this field is also reflected by several most recent publications about the strain dependence of the electronics bands in related materials.@xcite apart from the influence of stress and strain on the vb ordering , important elastic and electronic parameters such as the phonon- and exciton deformation potentials ( dp ) can be derived by e.g. raman and luminescence studies as function of applied pressure .
hydrostatic pressure was widely used to study the phase transition from the wurtzite to the rocksalt structure@xcite as well as a variety of phonon related parameters such as the grneisen parameters and the pressure dependence of the born effective charge.@xcite in addition , the phonon deformation potentials were determined by raman spectroscopy as function of uniaxial pressure.@xcite however , only few sources are available for the electronic deformation potentials . for gan and aln these
were recently studied using reflectance spectroscopy under uniaxial stress by ishii et al.@xcite in the case of zno , experimental studies and theoretical calculations were published without including the effects of spin - orbit coupling ( soc ) @xcite .
the effects of uniaxial pressure were experimentally studied in great detail by wrzesinski and frhlich using two - photon and three - photon spectroscopy but the spin - orbit splitting and the contributions of exchange interaction were not resolved and discussed separately.@xcite in order to remedy this situation , we investigate in this work the effects of stress on the band structure and exciton - polariton energies in wurtzite zno both theoretically and experimentally .
we perform a detailed theoretical analysis of the effects of external uniaxial stress , biaxial stress , and hydrostatic pressure on the a - b and a - c exciton - polariton splittings , the exchange splitting , and the longitudinal - transversal ( lt ) splitting using the quasi - cubic approximation and taking the interaction between all exciton states into account.@xcite from the results of _ first - principles _ calculations based on the hse+@xmath0 method ( see section [ abinitio ] ) , we compute the deformation potentials that describe all effects of external stress which preserve the wurtzite symmetry of zno .
the full set of the dp constants includes those describing the spin - dependent part ( allowing for the anisotropy of the spin - orbit interaction ) as well as the kinetic - energy part of the effective hamiltonian in the quasi - cubic approximation .
special attention is paid to the effects of stress on the spin - orbit interaction constant and its anisotropy .
the results of these considerations are compared with uniaxial stress - dependent photoluminescence ( pl ) measurements for zno substrates from different manufacturers as well as reported data in the literature .
the paper is structured as follows : after the description of the theoretical and experimental techniques in sections [ abinitio ] and [ setup ] , we first present the results of the _ ab - initio _ calculations for the electronic bands as function of pure uniaxial and biaxial stress and their combination ( section [ stress ] ) .
the excitonic deformation potentials including the stress dependence of the spin - orbit interaction are discussed in section [ defpot ] .
the theoretical considerations are complemented by measurements of the exciton polariton energies as function of uniaxial stress ( section [ results ] ) .
finally , the experimentally resolved exciton - polariton fine structure is used to investigate the effects of stress on the exchange interaction in section [ pol ] and the results are summarized in section [ conclusion ] .
in order to compute energies of electronic states that can be compared to experimental results , it is important to account for single - particle effects which are computed by solving a qp equation that properly accounts for the electronic self energy @xmath11 .
nowadays , hedin s @xmath12 approximation @xcite of @xmath11 is the state - of - the - art approach @xcite . due to the high computational cost of a self - consistent solution of the qp equation ,
first - order perturbation theory is used to obtain qp energies by correcting an initial electronic structure @xcite .
this so - called @xmath0 approach requires an initial electronic structure not too far from the final results .
we rely on solving a generalized kohn - sham equation @xcite , since for zno it has been shown before that the nonlocal hse hybrid functional@xcite using a parameter of @xmath13 a.u.@xmath14 ( see ref . for details ) provides an excellent starting point for @xmath0 calculations @xcite .
furthermore , in the present work the influence of the spin - orbit interaction on electronic states of strained zno is investigated .
earlier studies @xcite indicate that the soc only slightly modifies the kohn - sham states of zno ; the impact on the screened coulomb potential @xmath15 is expected to be small .
therefore , the influence of the spin - orbit interaction on the qp _ corrections _ is neglected here .
qp energies that account for the soc - related effects are calculated by adding qp shifts calculated without soc to hse eigenvalues that include spin - orbit - interaction effects @xcite .
this approach is expected to produce reasonable results because the absolute spin - orbit - induced shifts are small compared to the qp corrections for zno .
the qp electronic structures ( including soc ) of the uniaxially and biaxially strained cells are calculated using the vienna _ ab - initio _ simulation package ( vasp ) @xcite .
the o@xmath16 and o@xmath17 electrons as well as the zn@xmath18 and zn@xmath19 states are treated as valence electrons .
the projector - augmented wave method is applied to describe the wave functions in the core regions @xcite .
a plane - wave cutoff energy of @xmath20 ev is chosen to obtain converged results . in order to carry out the hse+@xmath0 calculations , @xmath21-centered @xmath22 monkhorst - pack@xcite @xmath23-point meshs along with at least 300 conduction bands
were used in this work .
we are aware that this method slightly underestimates the fundamental gaps , however , it is well suited for the description of the strain dependence of the qp electronic structure .
( color online ) illustration of the pressure apparatus used for the uniaxial pressure measurements .
the sample is placed between a movable hardened steel hemisphere and the lower end of the pressure piston ( enlarged section ) .
pressure is applied by an external pressurized he gas bottle.,width=226 ] the theoretical calculations are complemented by experimental results : pl measurements of the free exciton - polaritons as function of uniaxial pressure are performed for zno substrates from different suppliers ( crystec , tokyo denpa , and uniwafer ) .
the luminescence of the samples is excited by the 325 nm line of a hecd laser with an excitation power of 20 mw and detected using a bi - alkali photomultiplier attached to a spex 1 m double monochromator with a spectral resolution of 50 @xmath24ev .
uniaxial pressure measurements were performed in a self - built pressure apparatus ( fig .
[ uniaxial - apparatus ] ) which is placed inside a helium bath cryostat with a temperature of 4 k. the pressure on the samples was applied via a stainless - steel piston using a pneumatic construction . the primary gas pressure in the pressure transmitting chamber on top of the piston was fine tuned by a two stage pressure regulator with an error of 5 mbar .
the pressure in this chamber was used to push the piston downwards against a hardened steel hemisphere in between which the sample was placed .
the surface ratio between the pressure - transmitting chamber and the surface area of the samples determines the pressure transmission ratio .
typical values are between 500 and 1000 for samples with surface areas between 2 to 4 mm@xmath25 . assuming a precision of @xmath26 @xmath24 m for the size determination of the samples and the pressure stamp and considering the pressure accuracy in the primary gas chamber , the error of the applied pressure is estimated to @xmath27 gpa . for the pressure - dependent measurements the @xmath4 axis was oriented parallel to the direction of the uniaxial pressure ( @xmath28 ) .
the luminescence was excited and detected from the edge of the sample in @xmath29 direction with @xmath30 indicating the direction of the exciting and emitted light .
polarization - dependent measurements were performed using a linear polarizer to select the @xmath31 and @xmath32 components of the luminescence .
the strain dependence of the conduction and valence band extrema is theoretically determined by _ ab - initio _ calculations as described in section [ abinitio ] .
the lattice of unstrained zno is characterized by the equilibrium values of the lattice constants @xmath33 and @xmath34 . in the present work we use the theoretical values reported by schleife et al.@xcite for the unstrained values of the lattice constants . as perturbation we consider uniaxial stress along the @xmath4 axis , isotropic biaxial in - plane stress , and their simultaneous combination which preserve the wurtzite symmetry of zno .
the following subsections briefly explain the procedure for these three cases .
pure uniaxial stress @xmath35 is defined by vanishing forces in the plane and constant forces along the @xmath4 axis ( @xmath36 axis ) whereas pure biaxial stress @xmath37 is defined by constant forces in the plane and vanishing forces along the @xmath4 axis : @xcite @xmath38 with @xmath35 ( @xmath37 ) being positive in the case of uniaxial ( biaxial ) compression along the @xmath36 axis ( in the plane ) and negative otherwise .
using hooke s law with the condition @xmath39 or @xmath40 , one obtains for the cases of uniaxial and biaxial stress @xmath41 where @xmath42 is the poisson ratio and @xmath43 is the biaxial relaxation coefficient . within the _ ab - initio _ calculations , we vary the lattice constants @xmath44 and @xmath45 from their relaxed value @xmath33 and @xmath34 to the uniaxially or biaxially stressed value @xmath46 or @xmath47 in order to obtain the value @xmath48 or @xmath49 , corresponding to the uniaxial and biaxial stress , respectively . for each value of the uniaxial strain @xmath50 or biaxial strain @xmath51 , we then compute the qp energies ( including spin - orbit interaction , cf .
[ abinitio ] ) of the conduction band and valence band extrema .
we compute linear fits to these results so that the qp energies @xmath52 can be expressed as @xmath53 where @xmath54 and @xmath55 are the uniaxial and biaxial strain coefficients , respectively .
these strain coefficient are summarized for the conduction and valence bands with and without soc in table [ eabc ] . the uniaxial stress @xmath35 and biaxial stress @xmath37 are then given by @xmath56 with the young modulus e and biaxial modulus y. ( color online ) results of _ ab - initio _
calculations for the qp energies ( including spin - orbit coupling ) of the conduction- and valence - band states at the @xmath21 point as a function of uniaxial strain along the @xmath4 axis ( a ) and biaxial in - plane strain ( b ) .
solid lines represent linear fits to the data .
the results of these fits are listed in table [ eabc ] . ]
the computed values for the qp energies of the conduction band and valence bands as function of uniaxial and biaxial strain are visualized in fig .
[ theory ] together with the linear fits to these results .
the upper a and b bands ( without strain ) in fig . [ theory ] have @xmath2 and @xmath3 symmetry , respectively , and originate from the @xmath57 band ( without considering soc ) .
this ordering indicates a negative soc in zno , in agreement with earlier _ ab - initio _ calculations@xcite and the conclusions based on magneto - optical studies by wagner _
_ et al.__@xcite and rodina _ _ et al.__@xcite one can note that the shift of the a and b valence bands as function of uniaxial and biaxial strain is very small .
consequently , the strain dependence of the energy splitting between these two bands is also small which clearly shows that the symmetry ordering of the a and b valence bands can not be reversed for any reasonable tensile or compressive strain . assuming a linear relation even at very high strain values , a uniaxial tensile strain of @xmath58 or a biaxial compressive strain of @xmath59 would be required to result in a reversal of the topmost two vbs .
apparently , this amount of strain is neither realistic for uniaxial or biaxial pressure experiments nor for strain due to strongly lattice mismatched hetero - epitaxial growth . [
cols="<,^,^,^,^,^,^,^,^,^,^ " , ] ( color online ) energy shift of free exciton - polaritons in different zno substrates as function of uniaxial pressure .
dashed gray line : experimental results reported in ref .
.,width=264 ] ( color online ) splitting between the exciton - polaritons involving hole states from the a , b , and c vbs as function of uniaxial pressure .
dashed gray line : experimental results according to ref .
.,width=245 ] the energies and stress rates of the observed transitions are not expected to be identical with the theoretical energies @xmath60 , @xmath61 , and @xmath62 and their shift rates described in secs .
[ stress ] and [ defpot ] since the contributions of the exchange interaction and the exciton - polariton coupling have to be included in the theoretical description .
this is discussed in detail in section [ pol ] .
nevertheless , the measured energies of @xmath63 and @xmath64 can be used as good approximation to obtain the band energies @xmath60 and @xmath61 and their stress rates .
based on the observation of the @xmath65 and @xmath66 transitions of the a and b excitons , the exciton binding energies at zero stress can be estimated using the hydrogen - like model described in eq . , leading to @xmath67 mev and @xmath68 mev .
the uniaxial stress rates for the binding energies of the a and b excitons are found to be equal with a value of @xmath69 mev / gpa . for the different energy splittings in the tokyo denpa sample we obtain experimental stress rates of @xmath70 mev / gpa and @xmath71 mev / gpa .
these values are only slightly smaller than the theoretical results of @xmath72 mev / gpa and @xmath73 mev / gpa for the a and b band - gap energies ( see table [ eabc ] ) .
the shift rate of the @xmath74 exciton splitting of @xmath75 mev / gpa is in excellent agreement with the calculated value of 1.4 mev / gpa as listed in table [ eabc ] .
alternatively , one can use the @xmath76 , @xmath77 , and @xmath78 energies ( cf .
[ uniaxial - shift ] ) as reference to estimate the vb splittings as function of uniaxial pressure @xmath79 and @xmath80 as shown in fig .
[ uniaxial - splitting ] . using the @xmath81 splitting to estimate the stress rates @xmath79 for the three different zno samples we obtain experimental values of 0.7 mev / gpa ( crystec ) , 1.3 mev / gpa ( tokyo denpa ) , and 1.5 mev / gpa ( uniwafer ) which are in good agreement with the theoretical prediction of 1.4 mev / gpa ( see table [ eabc ] ) .
in contrast , the @xmath82 splitting as function of uniaxial pressure yields stress rates between @xmath83 mev / gpa and @xmath84 mev / gpa for the three different samples which is significantly smaller than the theoretical result of @xmath85 mev / gpa ( see table [ eabc ] ) .
a possible reason could be that the theoretical description above does neither account for the differences in the stress rates of the binding energies of the a , b , and c excitons nor for the effects of stress on the exchange interaction and the lt splittings .
these effects are including in the theoretical considerations in the next section .
for the a and b excitons the stress rates are found to be equal , however , the data for the stress rate of the binding energy of the c exciton is missing . for the following analysis this rate
is assumed to be equal to those of the a and b excitons which seems reasonable due to the reported similarity in the case of hydrostatic pressure.@xcite the lt splitting of the a exciton - polaritons as function of uniaxial pressure is displayed in the lower panel of fig .
[ uniaxial - splitting ] ; the stress dependence is very small .
similar considerations apply for the lt splitting of the b exciton - polariton ( cf .
table [ tzero ] ) . in the case of the c exciton - polaritons ,
an experimental determination of the stress rate of the lt splitting is not possible due to the previously discussed overlap of the @xmath86 line with the a@xmath87 lines .
thus , the pressure dependence of the lt splitting of the c exciton has to be obtained from theoretical considerations .
in this section , we discuss the effect of stress on the exchange interaction and lt splitting of the exciton - polaritons by deriving the analytical expression for the stress dependent exciton and exciton - polariton energies .
we use the experimentally observed uniaxial stress rates for the exciton - polariton lines in order to compute the missing stress rates for the @xmath86 line as well as for the @xmath60 , @xmath61 , and @xmath62 exciton energies in order to extract the experimental stress rates for the energy splittings and compare them with the results of the _ ab - initio _ calculations in table [ eabc ] .
the excitonic hamiltonian of eq . does not include the effects of the electron - hole exchange interaction which contains short range and long range contributions .
the short range exchange interaction can be added as@xcite @xmath88 where @xmath89 is the characteristic energy of the isotropic short - range exchange interaction and @xmath90 and @xmath91 are the electron and hole spin operators , respectively .
the energy @xmath89 is defined according to ref . by @xmath92 , where @xmath93 is the volume of the unit cell
, @xmath94 is the 1s hydrogen - like function which describes the relative motion of electron and hole , and @xmath95 is an atomic - like exchange energy . the use of the same characteristic energy @xmath89 for the a , b , and c excitons is justified since their binding energies and their corresponding @xmath96 functions are very similar.@xcite in the case of well separated noninteracting a , b , and c excitons , the hamiltonian @xmath97 yields the energies of the @xmath98 , @xmath99 , and @xmath100 transitions as @xmath101 , @xmath102 , and @xmath103 , respectively .
the values @xmath44 and @xmath104 describe the contributions of the @xmath105 component to the basis function of the @xmath106 and @xmath107 valence bands so that @xmath108 .
the explicit expressions for them can be found in refs . .
in zno , the value of @xmath109 is very close to 1 and @xmath110 is small .
since the exchange interaction energy between the a and b excitons is comparable with the energy separation @xmath111 , this interaction needs to be taken into account .
the transverse exciton states with @xmath112 symmetry are then given by : @xmath113 for the a and c excitons of @xmath114 symmetry one obtains @xmath115 and @xmath116 , without taking an interaction between them into account .
although the oscillator strength for @xmath117 is expected to be small in zno ( proportional to @xmath110 ) , the @xmath117 transition is stronger than the @xmath118 transition in the @xmath119 spectrum due to the intraband relaxation and larger population of holes to the a vb at low temperatures .
in contrast to the a - b energy splitting , the energy separation between the a and c excitons in zno is much larger than the characteristic exchange energy ( cf .
tables [ eabc ] and [ tzeroparameter ] ) . nevertheless , it is important to include the exchange interaction between the a and c excitons for a proper description of the stress effects on the exciton - polariton fine structure since the @xmath120 energy is strongly affected by stress . including this interaction , we obtain the following expressions for the energies of the a(@xmath114 ) and c(@xmath114 ) excitons : @xmath121 in the used @xmath122 geometry ( we may assume that @xmath123 is directed e.g. along the @xmath124 direction ) , the @xmath125 exciton states ( for each a and b exciton ) are the transverse states with the energies @xmath63 and @xmath64 , respectively , while the @xmath126 states are the longitudinal excitons with energies @xmath127 and @xmath128 , respectively .
their energies can be derived from the condition @xmath129 where @xmath130 is the frequency dependent dielectric function for an electric field @xmath131 with frequencies close to the a and b excitonic resonances.@xcite longitudinal excitons can be observed for the @xmath132 polarization . due to the interaction of the transverse excitons with photons ,
a lower and an upper polariton branch are formed .
their energies can be derived from the condition @xmath133 with @xmath45 being the speed of light . as the energies of the upper transverse polariton branches coincide with the energies of the longitudinal excitons at @xmath134 ,
they are often labeled @xmath127 and @xmath128 , whereas the energies of the lower transverse polariton branch at the @xmath21 point are marked as @xmath63 and @xmath64 . in order to determine @xmath130 for frequencies close to the exciton resonances and to obtain the longitudinal - transversal ( lt ) splitting it is instructive to follow the general approach described by ivchenko.@xcite neglecting the interaction between the a and b excitons and considering them as well separated resonances , one obtains @xmath135 near the @xmath136 exciton resonance and @xmath137 close to the energy of the @xmath138 exciton . here , @xmath139 is the characteristic energy for the dipole interaction with an electric field @xmath131 , @xmath140 are the dipole matrix element between bloch functions of the @xmath112 valence band and conduction band , and @xmath141 is the background ( high frequency ) dielectric constant .
the resulting upper branch energies are given by @xmath142 and @xmath143 .
apparently , the lt splittings @xmath144 for the @xmath145 ) exciton and @xmath146 for the @xmath147 ) exciton become almost equal as the parameter @xmath109 is close to 1 for zno . the large differences for the lt splittings of the @xmath145 ) and @xmath147 ) excitons observed in the experiment
can be explained by the interaction between themselves .
accounting for the simultaneous contributions of the a(@xmath112 ) and b(@xmath112 ) excitons to the dielectric function and their short - range exchange interaction , we derive @xmath130 following the same general approach : @xmath148 [ ( \hbar\omega)^2-\mathrm{b_t}^2]}\\ \nonumber & \cdot \big[{\rm a}_{\rm t}{\rm b}_{\rm t}(e_{\rm a}+a^2e_{\rm b})\\ \nonumber & + ( \hbar\omega)^2 [ ( e_{\rm a}+a^2e_{\rm b})-(\mathrm{a_t}+\mathrm{b_t})(1+a^2 ) ] \big ] .\end{aligned}\ ] ] using the conditions @xmath149 and @xmath150 , we find approximate energies of the longitudinal excitons / upper polariton branches of the a(@xmath112 ) and b(@xmath112 ) excitons as : @xmath151 in the case of @xmath114 states , all excitons in the @xmath122 geometry are transverse
. for @xmath132 polarized light one can observe also upper transverse polariton branches of the a(@xmath114 ) and c(@xmath114 ) excitons at the energies @xmath152 and @xmath153 .
these energies can be determined from the condition @xmath154 , where @xmath155 is now the frequency dependent dielectric function for the electric field @xmath132 in the range of frequencies close to the a and c excitonic resonances , respectively .
considering again the simultaneous contributions of the a(@xmath114 ) and c(@xmath114 ) excitons and their short - range exchange interaction and following the general approach of ref .
@xcite we derive @xmath155 as : @xmath156 [ ( \hbar \omega)^2-\mathrm{c_t}^2]}\\ \nonumber & \cdot\big[\mathrm{a_{t1}}\mathrm{c_t}(b^2e_c+a^2e_{\rm a})\\ \nonumber & + ( \hbar\omega)^2 [ ( b^2e_c+a^2e_{\rm a})-(\mathrm{a_{t1}}+\mathrm{c_t } ) ] \big],\end{aligned}\ ] ] where @xmath157 is the background ( high frequency ) dielectric constant , @xmath158 is the characteristic energy for the dipole interaction with an electric field @xmath132 , and @xmath159 is the dipole matrix element between the bloch functions of the @xmath114 valence band and conduction band ) . using the conditions @xmath160 and @xmath161 as well as @xmath162
, we find approximate energies of the longitudinal excitons / upper polariton branches of the a(@xmath114 ) and c(@xmath114 ) excitons as : @xmath163 it should be noted that even without considering the interaction between a(@xmath114 ) and c(@xmath114 ) excitons , the theory predicts the lt splitting @xmath164 of the a(@xmath114 ) exciton to be much smaller than the lt splitting @xmath165 of the c(@xmath114 ) exciton in agreement with the experimental data .
however , the inclusion of this interaction is crucial for a proper description of the stress dependences because of the large stress rate for the energy separation between a and c excitons .
cccccc & & & ref . & ref .
+ & symmetry & @xmath166 & @xmath167 & @xmath166 & @xmath166 + & @xmath168 & ( ev ) & ( mev / gpa ) & ( ev ) & ( ev ) + + @xmath169 & @xmath114 & 3.37504 & 20.56 & 3.37516 & 3.3756 + @xmath170 & @xmath114 & 3.37523 & 20.60 & 3.37539 & 3.3759 + @xmath63 & @xmath112 & 3.37538 & 20.4 & 3.37599 & 3.3759 + @xmath127 & @xmath112 & 3.37735 & 21.0 & 3.37741 & 3.3778 + @xmath64 & @xmath112 & 3.3815 & 21.7 & 3.38256 & 3.3816 + @xmath128 & @xmath112 & 3.3917 & 22.3 & 3.39253 & 3.3929 + @xmath153 & @xmath114 & 3.4323 & 13.5 & 3.43264 & 3.4327 + + @xmath86 & @xmath114 & 3.4213 & 2.6 & 3.42162 & 3.4209 + @xmath60 & & 3.37501 & 20.6 & 3.37505 & 3.37555 + @xmath61 & & 3.38106 & 21.9 & 3.38112 & 3.38118 + @xmath62 & & 3.42051 & 3.0 & 3.41930 & 3.42015 + @xmath89 & & 0.00041 & @xmath171 & 0.00122 & 0.00040 + @xmath172 & & 0.00627 & 0.65 & 0.00591 & 0.00694 + @xmath173 & & 0.00838 & 5.7 & 0.00844 & 0.00897 + @xmath174 & & 0.98112 & -0.0004 & 0.97643 & 0.96960 + @xmath175 the parameter @xmath44 is dimensionless , the units of all other parameters are as specified .
+ consequently , we have now derived eight equations
, , , , and which completely determine the exciton - polariton energies @xmath169 , @xmath170 , @xmath63 , @xmath127 , @xmath64 , @xmath128 , @xmath86 , and @xmath153 as experimentally observed in the polarized pl spectra ( see tables [ tzero ] and [ tzeroparameter ] ) .
from this set of equations we can calculate all eight unknown variables .
these are the zero stress exciton energies @xmath86 , @xmath60 , @xmath61 , and @xmath62 , the exchange and dipole interaction parameters @xmath89 , @xmath173 , @xmath172 , and the wave function parameter @xmath44 . the resulting values are listed in table [ tzeroparameter ] . for comparison
, we have also calculated these parameters using the exciton - polariton energies reported by wrzesinski _
_ @xcite and fiebig _ _ et al.__@xcite ( cf .
table [ tzeroparameter ] ) .
one can see that the resulting energies and parameters are in very good agreement which justifies the described approach . in the next step ,
we use the zero stress energies and the experimentally determined uniaxial stress rates for the lines @xmath169 , @xmath170 , @xmath63 , @xmath127 , @xmath64 , @xmath128 , and @xmath153 in order to compute the stress dependence of all output parameters by solving equations .
the resulting stress rates are also listed in table [ tzeroparameter ] .
it is important to note that the computed results for the @xmath86 line are very sensitive to the input parameters .
thus , even small errors of the input stress rates or zero stress energies may result in about one order of magnitude larger errors for the position and stress rates of the @xmath86 line as compared to the input values .
the stress rate for the @xmath86 line of 2.6 mev / gpa is found to be significantly smaller than the measured value of @xmath176 mev / gpa for the @xmath153 line .
this large difference in the stress rates for @xmath153 and @xmath86 lines ( and thus the large stress rate for the lt splitting of the @xmath177 exciton of 10.9 mev / gpa ) is partly caused by the stress dependence of the dipole interaction energy @xmath165 of the @xmath177 exciton and partly due to the large stress rate for the energy splitting between a and c excitons .
compared to the c excitons , the stress rate for the lt splitting of the @xmath178 excitons is significantly smaller because of the much weaker dipole interaction energy @xmath164 .
nevertheless , the small difference in the stress rates of the @xmath117 and @xmath179 exciton lines can still be observed in fig .
3 as a broadening of the @xmath178 exciton line with increasing uniaxial stress . using the computed uniaxial stress rates @xmath60 , @xmath61 , and @xmath62 and assuming the stress rates of the a , b , and c exciton binding energies to be equal ( 2.8 mev / gpa ) , we obtain the uniaxial rates @xmath167 for the energy splittings @xmath180 , @xmath181 , @xmath182 , @xmath111 , @xmath183 , and @xmath120 listed in the eights column of table [ eabc ] as experimentally determined rates . these shifting rates are in a very good agreement with our computed rates by _ ab - initio _ calculations ( sevenths column of table [ eabc ] ) .
the procedure described above considerably improves the estimation for the a - c exciton splitting . indeed , using the rates for @xmath60 and @xmath62
, we obtain a value of about @xmath184 mev / gpa for the @xmath120 splitting in good agreement with the theoretical result of @xmath185 mev / gpa .
the remaining discrepancy could be caused e.g. by the fact that we have assumed the same stress rate for the c exciton binding energy as experimentally determined for the a and b excitons as well as because of the still large uncertainty for the stress rate of the @xmath86 exciton line .
we have presented a detailed theoretical and experimental study on the effects of stress on the band gap energies and the exciton - polariton fine structure in wurtzite zno .
the theoretical approach combined _ ab - initio _
calculations for the qp energies as function of uniaxial and biaxial stress with the @xmath186 modeling of the band structure hamiltonian and the exciton - polariton splittings .
these calculations were complemented by polarization dependent luminescence measurements of the exciton - polariton emission lines in different zno substrates as function of uniaxial stress . based on the joint experimental and theoretical approach , clear evidence is found that the ordering of the topmost a(@xmath2 ) and b(@xmath3 ) valence bands in zno remains unchanged even for large uniaxial and biaxial stress .
the zero - stress energies and the strain and stress rates for the band gap energies , valence band splittings , crystal - field splitting , and anisotropic spin - orbit coupling are computed for uniaxial , biaxial , and hydrostatic pressure . while the crystal - field spitting is found to be highly sensitive to uniaxial and biaxial stress ,
the spin - orbit coupling is almost unaffected by stress , thus explaining the stability of the symmetry ordering of the a and b valence bands in zno .
in addition , the full set of symmetry preserving deformation potentials is determined including the spin - orbit interaction and its anisotropy .
based on the combination of polarized luminescence measurements and @xmath186 modeling , the uniaxial stress rates are determined for the longitudinal - transverse ( lt ) splitting of the exciton polaritons ( including the lt splitting of the a(@xmath114 ) exciton - polariton ) and their excitonic binding energies . while the stress dependence of the lt splitting of the a and b exciton - polaritons
is found to be small for uniaxial pressure , the lt splitting of the c exciton - polariton is significant which is explained by the large stress dependences of the dipole interaction energy and the crystal - field interaction . it is shown that only through the inclusion of the stress dependence of the exchange interaction and the lt splitting in the theoretical description , it is possible to achieve an accurate description of the experimental data .
furthermore , we also obtain the zero - stress energies and uniaxial stress rates for the dipole interaction and wave function parameters .
in summary , the combination of theoretical and experimental techniques in this work proved that the valence band ordering in zno is unaffected by any realistic amount of strain due to e.g. lattice mismatch and different coefficients of thermal expansion in hetero - epitaxially grown layers or different impurity concentrations in bulk material .
our results also provide a multitude of important electronic parameters including the spin - dependent deformation potentials as well as the energies and stress rates of the crystal - field interaction , spin - orbit coupling , exchange interaction , and exciton - polariton splittings . the research presented here received funding from the european community s seventh framework program ( fp7/2007 - 2013 ) under grant agreement no .
211956 and from the deutsche forschungsgemeinschaft ( project no
. be1346/20 - 1 ) .
acknowledges support of the russian foundation for basic research ( grant no . 13 - 02 - 00888-a ) and the support received from the swiss national science foundation .
thanks the carl - zeiss - stiftung for support .
we acknowledge grants of computer time from the hlrs stuttgart and the lrz munich .
support from the dfg within sfb 787 and the cluster of excellence `` unicat '' is acknowledged .
part of this work was performed under the auspices of the u.s .
department of energy at lawrence livermore national laboratory under contract de - ac52 - 07a27344 .
77ifxundefined [ 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 10.1016/0022 - 3697(59)90064 - 2 [ * * , ( ) ] link:\doibase 10.1016/0022 - 3697(60)90105 - 0 [ * * , ( ) ] link:\doibase 10.1016/0022 - 3697(60)90104 - 9 [ * * , ( ) ] link:\doibase 10.1103/physrev.140.a1726 [ * * , ( ) ] link:\doibase 10.1103/physrev.143.512 [ * * , ( ) ] link:\doibase 10.1103/physrevb.80.205203 [ * * , ( ) ] link:\doibase 10.1103/physrevb.69.125206 [ * * , ( ) ] link:\doibase 10.1063/1.2825277 [ * * , ( ) ] link:\doibase 10.1103/physrevb.65.075207 [ * * , ( ) ] in @noop _ _ , , vol .
( , ) pp . @noop * * , ( ) link:\doibase 10.1103/physrevb.54.7678 [ * * , ( ) ] link:\doibase 10.1063/1.364074 [ * * , ( ) ] link:\doibase 10.1063/1.3174436 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1143/jjap.40.l1089 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1063/1.3447798 [ * * , ( ) ] link:\doibase 10.1063/1.3554434 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevb.2.4005 [ * * , ( ) ] link:\doibase 10.1103/physrevb.56.13087 [ * * , ( ) ] @noop ( ) link:\doibase 10.1016/s0022 - 0248(98)80143 - 7 [ * * , ( ) ] link:\doibase 10.1016/s0038 - 1098(97)10127 - 2 [ * * , ( ) ] link:\doibase 10.1103/physrev.139.a796 [ * * , ( ) ] in link:\doibase 10.1016/s0081 - 1947(08)60248 - 9 [ _ _ ] , , vol . , ( , ) pp .
link:\doibase 10.1103/revmodphys.74.601 [ * * , ( ) ] link:\doibase 10.1103/physrevb.34.5390 [ * * , ( ) ] link:\doibase 10.1103/physrev.140.a1133 [ * * , ( ) ] link:\doibase 10.1103/physrevb.53.3764 [ * * , ( ) ] link:\doibase 10.1063/1.2204597 [ * * , ( ) ] link:\doibase 10.1063/1.2404663 [ * * , ( ) ] link:\doibase 10.1063/1.2187006 [ * * , ( ) ] link:\doibase 10.1063/1.2403866 [ * * , ( ) ] link:\doibase 10.1103/physrevb.80.035112 [ * * , ( ) ] link:\doibase 10.3938/jkps.53.2811 [ * * , ( ) ] link:\doibase 10.1002/pssb.200945204 [ * * , ( ) ] link:\doibase 10.1103/physrevb.79.205205 [ * * , ( ) ] link:\doibase 10.1103/physrevb.78.155114 [ * * , ( ) ] link:\doibase 10.1103/physrevb.54.11169 [ * * , ( ) ] link:\doibase 10.1016/0927 - 0256(96)00008 - 0 [ * * , ( ) ] link:\doibase 10.1103/physrevb.74.035101 [ * * , ( ) ] link:\doibase 10.1103/physrevb.50.17953 [ * * , ( ) ] link:\doibase 10.1103/physrevb.59.1758 [ * * , ( ) ] link:\doibase 10.1103/physrevb.62.11556 [ * * , ( ) ] link:\doibase 10.1103/physrevb.13.5188 [ * * , ( ) ] link:\doibase 10.1103/physrevb.73.245212 [ * * , ( ) ] link:\doibase 10.1103/physrevb.66.115202 [ * * , ( ) ] link:\doibase 10.1016/0038 - 1098(95)00054 - 2 [ * * , ( ) ] @noop ( ) link:\doibase 10.1063/1.2177928 [ * * , ( ) ] link:\doibase 10.1103/physrevb.64.115204 [ * * , ( ) ] @noop ( ) link:\doibase 10.1103/physrevb.76.184120 [ * * , ( ) ] link:\doibase 10.1016/j.mejo.2008.07.025 [ * * , ( ) ] @noop ( ) link:\doibase 10.1002/pssb.2220750113 [ * * , ( ) ] link:\doibase 10.1063/1.3284656 [ * * , ( ) ] link:\doibase 10.1103/physrevb.88.235207 [ * * , ( ) ] link:\doibase 10.1103/physrevb.82.115207 [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.035313 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) in @noop _ _ , ( , ) pp . in @noop
_ _ , ( , ) p.
link:\doibase 10.1002/pssb.2221770114 [ * * , ( ) ] | the uniaxial stress dependence of the band structure and the exciton - polariton transitions in wurtzite zno is thoroughly studied using modern first - principles calculations based on the hse+@xmath0 approach , @xmath1 modeling using the deformation potential framework , and polarized photoluminescence measurements .
the ordering of the valence bands [ a(@xmath2 ) , b(@xmath3 ) , c(@xmath2 ) ] is found to be robust even for high uniaxial and biaxial strains .
theoretical results for the uniaxial pressure coefficients and splitting rates of the a , b , and c valence bands and their optical transitions are obtained including the effects of the spin - orbit interaction .
the excitonic deformation potentials are derived and the stress rates for hydrostatic pressure are determined based on the results for uniaxial and biaxial stress .
in addition , the theory for the stress dependence of the exchange interaction and longitudinal - transversal splitting of the exciton - polaritons is developed using the basic exciton functions of the quasi - cubic approximation and taking the interaction between all exciton states into account .
it is shown that the consideration of these effects is crucial for an accurate description of the stress dependence of the optical spectra in zno .
the theoretical results are compared to polarized photoluminescence measurements of different zno substrates as function of uniaxial pressure and experimental values reported in the literature demonstrating an excellent agreement with the computed pressure coefficients .
+ doi : 10.1103/physrevb.88.235210 |
consider the heat - type equation @xmath8 of order @xmath0 where @xmath9 if @xmath10 is even and @xmath11 if @xmath10 is odd .
let us introduce the corresponding kernel @xmath12 which is characterized by @xmath13 this kernel defines a pseudo - process @xmath14 driven by a signed measure with total mass equal to 1 ( which is not a probability measure ) according as the usual markov rules : we set for @xmath15 , @xmath16 and @xmath17 , @xmath18 and @xmath19 since we are dealing with a signed measure , it seems impossible to extend the definition of the pseudo - process over all the positive times .
we can find in the literature two possible ad - hoc constructions : one over the set of times of the form @xmath20 , @xmath21 ( depending on a fixed time @xmath22 , see @xcite and @xcite for pioneering works related to this construction ) , the other one over the set of dyadic times @xmath23 , @xmath21 .
( which do not depend on any particular time , see @xcite for this last construction ) . for @xmath24 ,
this is the most well - known brownian motion and for @xmath25 , @xmath14 is the so - called biharmonic pseudo - process . for the pseudo - process @xmath14 started at a point @xmath26 , we introduce : * the first overshooting times of a one - sided barrier @xmath4 ( or , equivalently , the first hitting time of the half - line @xmath6 ) or a two - sided barrier @xmath5 ( with the convention @xmath27 ) : @xmath28\}\mbox { for $ x\in[a , b]$};\end{aligned}\ ] ] * the last overshooting times of such barriers before a fixed time @xmath22 ( with the convention @xmath29 ) : @xmath30:x(s)\ge a\},\quad \sigma_{ab}(t ) = \sup\{s\in[0,t]:x(t)\notin ( a , b)\};\end{aligned}\ ] ] * the sojourn times in the intervals @xmath6 and @xmath7 $ ] up to a fixed time @xmath22 : @xmath31:x(s)\ge a\},\quad t_{ab}(t ) = \mathrm{measure}\{s\in[0,t]:x(s)\in[a , b]\};\ ] ] * the maximum up to time @xmath22 : @xmath32 in the foregoing rough definitions , the pseudo - distribution of the quantity @xmath33 for instance is to be understood as the limit of @xmath34 when @xmath35 .
we could introduce the alternative first hitting time of @xmath36 $ ] , the alternative sojourn time in @xmath36 $ ] and the up - to - date minimum @xmath37 . actually , the pseudo - distributions of these three quantities are obviously related to the pseudo - distributions of the foregoing ones .
we shall also consider the pseudo - process with a drift @xmath38 defined by @xmath39 where @xmath40 is a fixed real number .
for this latter , we introduce : * the first overshooting time of the threshold @xmath41 : @xmath42 if the set @xmath43 is not empty , else we set @xmath44 ; * the maximum functional up to time @xmath22 : @xmath45 the aim of this survey is to provide a list of explicit results concerning the pseudo - distributions of @xmath46 , @xmath47 , @xmath48 and @xmath49 , as well as those related to the pseudo - process with a drift . in particular ,
remarkable results hold for the pseudo - distributions of @xmath50 and @xmath51 .
we also provide some methods for deriving those of @xmath52 , @xmath53 and @xmath54 . a way consists in using the feynman - kac functional @xmath55 which is a solution to the partial differential equation @xmath56 with @xmath57 .
its laplace transform @xmath58 is a solution to the ordinary differential equation @xmath59 another way consists in using spitzer s identities which work actually when the starting point is @xmath60 and @xmath10 is even .
indeed , their validity holds thanks to the fact that the integral @xmath61 is finite , which is true only when @xmath10 is even .
additionally , spitzer s identities hinge on a symmetry property which is fulfilled only when the starting point of the pseudo - process is @xmath60 . in the case @xmath25 , see @xcite for many connections with fourth - order partial differential equations with various boundary value conditions .
let us introduce the @xmath62 roots of @xmath63 : @xmath64 and @xmath65 @xmath66 that will be used for solving the above differential equation .
the notations @xmath67 and @xmath68 stand for the cardinalities of the sets @xmath69 and @xmath70 .
we have @xmath71 for any @xmath72 .
set , for @xmath73 and @xmath74 , @xmath75 @xmath76 let us also introduce the @xmath77 roots of the complex number @xmath78 : @xmath79 and @xmath80 .
we shall need to introduce the roots @xmath81 of the polynomial @xmath82 ( where @xmath83 ) .
these last settings will be used for the pseudo - process with a drift . finally , set for any integer @xmath84 such that @xmath85 @xmath86 the functions @xmath87 satisfy @xmath88{\lambda}\,\xi}$ ] for @xmath89 and @xmath90 .
they will be useful for expressing several distributions .
the results are presented by topic and in certain topics we have chosen to exhibit them from the most particular to the most general thus following the chronology .
moreover , it is not easy sometimes to deduce the particular cases from the most general ones .
see @xcite for the chronology of the results concerning the distributions related to @xmath33 as well as for the connections with the maximum and minimum functionals of @xmath91 .
set @xmath92 for @xmath93 and @xmath94 .
the quantity @xmath95 should be understood as @xmath96 using the feynman - kac approach , it can be seen that the function @xmath97 satisfies the system @xmath98 \lambda\ , \phi(x)-1 & \mbox{for } x\in(-\infty , a ) , \end{array}\right.\ ] ] and @xmath99 this system can be explicitly solved by computing vandermonde determinants . in particular , for @xmath100 , the following formula holds : @xmath101{\lambda^{\ # k}(\lambda+\mu)^{\ # j}}}\ ] ] and this two - parameters laplace transform can be inverted ( @xcite ) .
the distribution of @xmath33 under @xmath102 is the same as that of @xmath50 under @xmath103 .
[ th1 ] the pseudo - distribution of @xmath50 is a beta law : @xmath104{s^{\ # k } ( t - s)^{\ # j } } } \mbox { for } s\in(0,t).\ ] ] if @xmath10 is even , @xmath50 obeys the famous paul lvy s arcsine law : @xmath105 in the history of pseudo - processes , this result was discovered by krylov ( @xcite ) when @xmath10 is even . for @xmath106 , orsingher obtained ( @xcite ) @xmath107{s^2(t - s ) } } & \mbox{when $ \kappa_{_{\ !
n}}=+1 $ , } \\[2ex ] \displaystyle\frac{\sqrt3}{2\pi\sqrt[3]{s(t - s)^2 } } & \mbox{when $ \kappa_{_{\ ! n}}=-1$. } \end{cases}\ ] ] using a similar method , the following simple results can be obtained ( @xcite ) .
[ th2 ] the pseudo - distribution of @xmath50 conditioned on @xmath108 is the uniform law on @xmath109 : for @xmath110 , @xmath111 the pseudo - distributions of @xmath50 conditioned on @xmath112 and @xmath113 are beta laws : for @xmath110 , @xmath114 if @xmath10 is even , for @xmath110 , @xmath115 the results of theorems [ th1 ] and [ th2 ] were found by hochberg , nikitin and orsingher ( @xcite ) in the cases @xmath116 and conjectured in the general case . set @xmath117 for @xmath118 and @xmath119 .
the quantity @xmath97 can be understood as @xmath120 a spitzer s identity yields the following relationship which holds for @xmath121 : @xmath122 & = \frac{1}{1-z}\,\exp\!\left(\,\sum_{k=1}^{\infty } \frac{1}{k } \,\mathbb{e}_0\ ! \big(e^{i\mu x(k/2^n)}\zeta^{k{1\hspace{-.2em}\mbox{\scriptsize\rm l}}_{[0,+\infty)}(x(k/2^n))}-1\big)z^k \right)\!.\end{aligned}\ ] ] with this identity at hand , it can be seen that @xmath123{\lambda+\nu}-i\mu \theta_j ) \displaystyle\mathop { \textstyle{\prod}}_{k \in k } ( \!\sqrt[n\!]{\lambda}-i \mu \theta_k ) } .\ ] ] this three - parameters laplace - fourier transform can be inverted ( @xcite ) .
the pseudo - distribution of the vector @xmath124 is given , for @xmath110 and @xmath125 , by @xmath126 and , for @xmath110 and @xmath127 , by @xmath128 where @xmath129 , @xmath130 for any integer @xmath131 ( in particular @xmath132 , @xmath133 ) , @xmath134 and @xmath135 is the mittag - leffler function @xmath136 . by choosing @xmath137 in the pseudo - distribution of @xmath124 in the foregoing theorem and next dividing the result by @xmath138
, we could retrieve the pseudo - distribution of the corresponding sojourn time of the `` pseudo - bridge '' @xmath139 displayed in theorem [ th2 ] .
set @xmath141 for @xmath93 and @xmath142 .
it can be seen that @xmath97 solves the differential equation @xmath143 \lambda\ , \phi(x , y)-\delta_y(x ) & \mbox{for } x\in(-\infty , a ) , \end{array}\right.\ ] ] with regularity conditions @xmath144 \displaystyle\forall k\in\{0,1,\ldots , n-2\},\ ; \frac{\partial^k \phi}{\partial x^k}(y^+,y)=\frac{\partial^k \phi}{\partial x^k}(y^-,y ) \mbox { and } \displaystyle\frac{\partial^{n-1}\phi}{\partial x^{n-1}}(y^+,y)- \frac{\partial^{n-1}\phi}{\partial x^{n-1}}(y^-,y)=\kappa_{_{\
! n}}. \end{array}\right.\ ] ] this system can be explicitly solved by computing vandermonde determinants and the inversion of the two - parameters laplace transform can be performed ( @xcite ) .
set @xmath145 1 .
assume that @xmath146 .
for @xmath110 , if @xmath147 $ ] , @xmath148 & = \kappa_{_{\ !
n } } { 1\hspace{-.27em}\mbox{\rm l}}_{\{\ # j=\#k+1\ } } \bigg(\sum_{j \in j}\theta_j a_j i_{\ # k}(t - s;-\theta_j x)\bigg ) \!\bigg(\sum_{j\in j } \theta_j^{\ # j } a_j i_{\#k}(s;\theta_j y ) \bigg ) \\ & \hphantom{=\;}+\kappa_{_{\ !
n } } \sum_{j , j'\in j,\,k\in k } \frac{a_j\theta_{j'}a_{j'}\theta_k^{\#k-\#j } b_k c_{j j ' k}}{\theta_j^{\#k-1}(\theta_{j'}-\theta_k ) } \int_0^s \sigma^{\frac 1 n-1 } i_{\#j-1}(s-\sigma;\theta_{j ' } y ) \,\mathrm{d}\sigma \\ & \hphantom{=\;}\times \int_0^{t - s } i_{\#k}(\tau;-\theta_j x ) f\big(t - s-\tau;\frac{\theta_j}{\theta_k}\!\sqrt[n\ ! ]
\sigma\,\big ) \,\mathrm{d}\tau\end{aligned}\ ] ] and if @xmath149 , @xmath150 & = \kappa_{_{\ ! n } } \sum_{j , j'\in j,\,k \in k } \frac{\theta_j a_j a_{j ' } b_k c_{jj'k}}{\theta_k^{\#j-1}(\theta_{j}-\theta_k ) } \int_0^s \sigma^{\frac 1n-1}i_{n-2 } ( s-\sigma;\theta_{j}y-\theta_k x ) \,\mathrm{d}\sigma \\ & \hphantom{=\;}\times \int_0^{t - s } \frac{(t - s-\tau)^{\frac 1 n-1}}{\gamma(\frac 1 n ) } f\big(\tau ;
\frac { \theta_{j ' } } { \theta_k}\!\sqrt[n\ ! ] \sigma\,\big ) \,\mathrm{d}\tau.\end{aligned}\ ] ] for @xmath151 , if @xmath147 $ ] , @xmath152 and if @xmath153 , there is an atom given by @xmath154 2 .
assume that @xmath155 .
for @xmath110 , if @xmath147 $ ] , @xmath150 & = -\sum_{j\in j,\,k , k'\in k } \frac{a_j\theta_k b_kb_{k ' } d_{j k k'}}{\theta_j^{\#k-2 } \theta_{k'}(\theta_k-\theta_j ) } \,s^{\frac 1n-1 } \int_0^{t - s } i_{n-1}(\tau;\theta_k y-\theta_j x ) f\big(t - s-\tau;\frac{\theta_{j}}{\theta_{k ' } } \!\sqrt[n\!]s\,\big)\,\mathrm{d}\tau\end{aligned}\ ] ] and if @xmath156 , @xmath157 & = { 1\hspace{-.27em}\mbox{\rm l}}_{\{\ # k=\#j+1\ } } \bigg(\sum_{k\in k } \theta_k b_k i_{\ # j}(s;-\theta_k x)\bigg ) \!\bigg(\sum_{k\in k } \theta_k^{\ # k } b_k i_{\#j}(t - s;\theta_k y)\bigg ) \\ & \hphantom{=\ ; } + \sum_{j\in j,\,k , k'\in k } \frac{\theta_j^{\#j-\#k+1}a_j b_k \theta_{k'}b_{k ' } d_{jkk'}}{\theta_k^{\#j } ( \theta_j-\theta_{k ' } ) } \int_0^s \sigma^{\frac 1n-1 } i_{\#j-1}(s -\sigma;-\theta_k x ) \,\mathrm{d}\sigma \\ & \hphantom{=\;}\times \int_0^{t - s } i_{\#k}(\tau;\theta_{k ' } y ) f\big(t - s-\tau;\frac{\theta_j}{\theta_k}\!\sqrt[n\ ! ] \sigma\,\big ) \,\mathrm{d}\tau.\end{aligned}\ ] ] for @xmath158 , if @xmath156 , @xmath159 and if @xmath147 $ ] , there is an atom given by @xmath160 the following relationship between @xmath50 , @xmath161 and @xmath162 holds : for @xmath163 , @xmath164 by integrating the joint pseudo - distribution of @xmath124 with respect to @xmath165 in the foregoing theorem , we could derive the pseudo - distribution of @xmath50 . actually , the result does not simplify so much .
see @xcite for references related to the various distributions related to @xmath161 .
the variables @xmath33 and @xmath161 are related together according to @xmath166 the quantity @xmath167 should be understood as @xmath168 thanks to this connection , it is possible to deduce that , for @xmath169 and @xmath89 , @xmath170{\lambda } \,(x - a)}\bigg)\!.\ ] ] the foregoing laplace transform can be inverted ( @xcite ) .
the pseudo - distribution of @xmath161 is given by @xmath171 where @xmath172 the coefficients @xmath173 , @xmath174 , @xmath175 , are given by @xmath176 and for @xmath177 , @xmath178 from the second displayed expression of @xmath179 , we can see that the famous reflection principle for brownian motion does not hold any longer for pseudo - processes related to an order @xmath0 . for @xmath106 ,
orsingher ( @xcite ) derived the historical result @xmath180 \displaystyle\frac{2}{\gamma(\frac13)}\int_0^t p(s;x - a ) \,\frac{\mathrm{d}s}{(t - s)^{\frac23}}-\,\frac{1}{\gamma(\frac23 ) } \int_0^t \frac{\partial p}{\partial x}(s;x - a)\,\frac{\mathrm{d}s}{(t - s)^{\frac13 } } & \mbox{when $ \kappa_{_{\ !
n}}=-1$. } \end{cases}\end{aligned}\ ] ] for @xmath25 , hochberg ( @xcite ) derived the historical result @xmath181{\lambda } \,(x - a)/\sqrt 2 } \sin \!\bigg(\frac{\sqrt[4]{\lambda } \,(x - a)}{\sqrt 2}\bigg)\ ] ] which was subsequently completed by beghin , orsingher and ragozina ( @xcite ) : @xmath182 set @xmath183 for @xmath118 , @xmath119 and @xmath94 .
the quantity @xmath95 can be understood as @xmath184 another spitzer s identity yields the following relationship which holds for @xmath185 : @xmath186 the laplace - fourier transform of the vector @xmath47 ensues : @xmath187{\lambda}-(i\mu-\nu)\theta_j ) \displaystyle\mathop{\textstyle{\prod}}_{k\in k } ( \!\sqrt[n\!]{\lambda}-i\mu\theta_k)}.\ ] ] this three - parameters transform can be progressively inverted ( @xcite ) .
for @xmath188 , @xmath189\,\mathrm{d}t = \frac{1}{\lambda}\,\chi_{_{\scriptstyle j}}(\lambda;x - z)\ , \chi_{_{\scriptstyle k}}(\lambda;z - y),\ ] ] with @xmath190{\lambda } \,\sum_{j\in j } \theta_ja_j\,e^{\theta_j\!\!\sqrt[n\!]{\lambda}\,\xi } , \quad \chi_{_{\scriptstyle k}}(\lambda;\xi)=-\!\sqrt[n\!]{\lambda } \,\sum_{k\in k } \theta_kb_k\,e^{\theta_k\!\!\sqrt[n\!]{\lambda}\,\xi}.\ ] ] this last laplace transform can also be inverted ( @xcite ) .
the joint pseudo - distribution of @xmath47 admits the representation below . for @xmath188 , @xmath191 where @xmath192 for @xmath106 , beghin , orsingher and ragozina ( @xcite ) derived the result @xmath193 \displaystyle\frac{1}{\gamma(1/3)}\int_0^t\!\!\int_0^s p(\sigma;z - y)\,q(s-\sigma;x - z)\ , \frac{\mathrm{d}s\,\mathrm{d}\sigma}{(t - s)^{2/3 } } & \mbox{when $ \kappa_{_{\ !
n}}=-1 $ , } \end{cases}\end{aligned}\ ] ] with @xmath194 and @xmath195 \displaystyle \frac{\xi}{\pi t } \bigg[\sqrt3 \int_0^{+\infty } e^{-t\lambda^3+\xi\lambda}\,\mathrm{d}\lambda + \int_0^{+\infty } e^{-t\lambda^3-\frac12\,\xi\lambda } \sin\!\bigg(\frac{\sqrt3}{2}\,\xi\lambda+\frac{\pi}{3}\bigg)\,\mathrm{d}\lambda\bigg ] & \mbox{when $ \kappa_{_{\ ! n}}=-1$. } \end{cases}\end{aligned}\ ] ] for @xmath25 , they derived @xmath196 with @xmath197 q_2(t;\xi ) & = \frac{\xi}{2\pi^2\,t } \int_0^{+\infty } e^{-t\lambda^4 } \left[\cos(\xi\lambda)+\sin(\xi\lambda)-e^{-\xi\lambda}\right]\mathrm{d}\lambda.\end{aligned}\ ] ] the laplace - fourier transform of the vector @xmath198 is given , for @xmath118 and @xmath119 , by @xmath199 this three - parameters transform can be partially inverted for giving the following result ( @xcite ) .
the laplace transform with respect to time @xmath22 of the joint pseudo - distribution of @xmath198 is given , for @xmath89 and @xmath188 , by @xmath200\,\mathrm{d}t = \chi_{_{\scriptstyle j}}^b(\lambda;x - z)\,\chi_{_{\scriptstyle k}}^b(\lambda;z - y)\ ] ] where @xmath201 the variables @xmath49 and @xmath161 are related together according as @xmath202 the pseudo - distribution of @xmath49 under @xmath102 is the same as that of @xmath203 under @xmath103 .
the iterated laplace transform of @xmath203 under @xmath103 is given , for @xmath93 , by @xmath204{\frac{\lambda+\mu}{\lambda } } \,\right)\!\!\bigg]\!.\ ] ]
in this section , @xmath10 is assumed to be an even integer .
the reader is referred to @xcite . using the definition
@xmath206 it can be seen that the laplace - fourier transform of the vector @xmath48 is related to the pseudo - distribution of the vector @xmath47 according as , for @xmath89 , @xmath119 and @xmath169 : @xmath207 from this , it can be deduced that , for @xmath169 , @xmath208{\lambda}}\,\bar{\theta}_\ell ) \,e^{\theta_j\!\!\sqrt[n\!]{\lambda}\,(x - a ) } \,e^{i\mu a}.\ ] ] in particular , @xmath209{\lambda}\,(x - a)}.\ ] ] the variables @xmath205 and @xmath161 are related together according to @xmath210 .
the two - parameters laplace - fourier transform can be inverted ( @xcite ) .
the joint pseudo - distribution of @xmath48 is given , for @xmath211 , by @xmath212 with @xmath213 .
in particular , @xmath214 the @xmath215 are the successive derivatives of the schwartz distribution @xmath216 , that is , for any test function @xmath217 , @xmath218 .
the pseudo - distribution of @xmath51 is remarkable since it means that the pseudo - process @xmath14 is formally concentrated at the site @xmath41 at time @xmath205 in a `` distributional '' sense . in the case
@xmath25 , nishioka ( @xcite ) obtained the remarkable result @xmath219 moreover , @xmath220 with @xmath221 the pseudo - distributions of the vectors @xmath223 and @xmath198 are related together according as , for @xmath89 , @xmath119 and @xmath169 , @xmath224 from which it comes that the laplace - fourier transform of the vector @xmath223 is given , for @xmath89 , @xmath119 and @xmath169 , by @xmath225 in the case @xmath25 , nakajima and sato ( @xcite ) derived the following result : @xmath226 where @xmath227 and @xmath228 are the two roots of the polynomial @xmath229 having positive imaginary part .
we can deduce the pseudo - distribution of the overshooting place @xmath230 on the set @xmath231 and the pseudo - probability of eventually hitting the interval @xmath6 ( @xcite ) .
the pseudo - distribution of the overshooting place @xmath232 is given , for @xmath169 , by @xmath233 where the @xmath234 s are given by @xmath235 and for @xmath236 , @xmath237 the pseudo - probability of eventually overshooting the level @xmath41 is given by @xmath238 for @xmath25 , the above results yield for @xmath169 , in the case where @xmath239 , @xmath240b}\big(1-e^{\!\sqrt[3]b\,(x - a)}\big)\delta_a'(z),\ ] ] and , in the case where @xmath241 , @xmath242 with @xmath243{|b|}\,\xi } \cos\!\bigg(\frac{\sqrt3}{2}\,\sqrt[3]{|b|}\,\xi+\frac{\pi}{6}\bigg)\ ! , \quad p_1^b(\xi ) = -\frac{2}{\sqrt3\,\sqrt[3]{|b|}}\,e^{\frac12\sqrt[3]{|b|}\,\xi } \sin\!\bigg(\frac{\sqrt3}{2}\,\sqrt[3]{|b|}\,\xi\bigg)\!.\ ] ]
the problem of the two - sided barrier @xmath5 is much more difficult to tackle than the single one .
the variables @xmath52 , @xmath244 and the maximum / minimum functionals are related together according as , for @xmath245 , @xmath246 the variable @xmath52 is introduced in @xcite for defining a local time for the pseudo - process @xmath91 .
set @xmath247 for @xmath93 and @xmath94 .
the function @xmath97 satisfies the system @xmath248 \lambda\ , \phi(x)-1 & \mbox{for } x\notin(a , b ) , \end{array}\right.\ ] ] and @xmath249 it seems to be difficult to solve explicitly this system . in the same way
, the problem of the joint pseudo - distribution of @xmath250 should be more complicated .
the sojourn time within a strip is used by beghin & orsingher ( @xcite ) for defining a local time at 0 for @xmath14 : @xmath251\}}\ , \mathrm{d}s$ ] .
they obtained the very simple formula @xmath252{\lambda}}\ ] ] where @xmath253 by introducing a pseudo - random walk as in @xcite and studying the similar functional to @xmath244 ( first exit time from a finite interval ) , we can derive the pseudo - distribution of @xmath54 in the case where @xmath10 is even .
we have obtained the following result ( @xcite ) .
the pseudo - distribution of @xmath254 has the form @xmath255 where @xmath256 and @xmath257 are some functions .
the pseudo - distribution of @xmath254 is given by @xmath258 where the functions @xmath259 and @xmath260 , @xmath261 , are the interpolation hermite polynomials such that @xmath262 in particular , the `` ruin pseudo - probabilities '' are given by @xmath263 by applying the pseudo - markov property , it can be easily seen that the distributions of @xmath53 and @xmath244 are related together according as , for @xmath94 and @xmath267 $ ] , @xmath268 , x(s)\in [ a , b]\}=\mathbb{e}_x\big(\mathbb{p}_{x(\sigma ) } \{\tau_{ab}\ge t-\sigma\}{1\hspace{-.27em}\mbox{\rm l}}_{\{x(\sigma)\in[a , b]\}}\big).\ ] ] this equality may be extended to an `` excursion '' between the thresholds @xmath41 and @xmath40 . indeed , by introducing @xmath269 , we have , for @xmath94 and @xmath270 , @xmath271 , x(s)\in [ a , b ] , x(\varsigma_{ab}(t))\in \mathrm{d}y\ } \\ & = \mathbb{e}_x\big(\mathbb{p}_{x(\sigma)}\{\tau_{ab}\ge\varsigma-\sigma , x(\tau_{ab})\in \mathrm{d}y\}{1\hspace{-.27em}\mbox{\rm l}}_{\{x(\sigma)\in[a , b]\}}\big ) . \\ & = \int_a^b p(\sigma;x - z)\,\mathbb{p}_{\!z}\{\tau_{ab}\ge\varsigma-\sigma , x(\tau_{ab})\in \mathrm{d}y\}\big)\,\mathrm{d}z.\end{aligned}\ ] ] acknowledgements .
this survey is the text associated with the talk i gave at the fifth iwap held in madrid in july 2010 at the session `` special stochastic processes '' chaired by k.j .
it has been a great honor for me to give this talk in the presence of k.j .
hochberg and e. orsingher who were pioneers in the domain of pseudo - processes .
cammarota , v. and lachal , a. ( 2010 ) : joint distribution of the process and its sojourn time in the positive half - line for pseudo - processes governed by high - order heat equation .
_ electron .
j. probab .
_ , _ 15 _ , 895931 .
cammarota , v. and lachal , a. ( 2010 ) : joint distribution of the process and its sojourn time in a half - line for pseudo - processes governed by higher - order heat - type equations .
_ stochastic process .
_ lachal , a. ( 2006 ) : joint law of the process and its maximum , first hitting time and place of a half - line for the pseudo - process driven by the equation @xmath273 .
_ c. r. math .
paris _ , _ 343 no . 8 _ , 525530 . nakajima , t. and sato , s. ( 1999 ) : on the joint distribution of the first hitting time and the first hitting place to the space - time wedge domain of a biharmonic pseudo process . _ tokyo j. math .
_ , _ 22 no . 2 _ , 399413 | fix an integer @xmath0 and let @xmath1 be the pseudo - process driven by the high - order heat - type equation @xmath2 .
the denomination `` pseudo - process '' means that @xmath3 is related to a signed measure ( which is not a probability measure ) with total mass equal to 1 . in this survey
, we present several explicit results and discuss some problems concerning the pseudo - distributions of various functionals of the pseudo - process @xmath3 : the first or last overshooting times of a single barrier @xmath4 or a double barrier @xmath5 by @xmath3 ; the sojourn times of @xmath3 in the intervals @xmath6 and @xmath7 $ ] up to a fixed time ; the maximum or minimum of @xmath3 up to a fixed time .
* keywords .
pseudo - process ; pseudo - distribution ; first hitting or overshooting time ; sojourn time ; up - to - date maximum . * |
the doped rare earth manganites @xmath4mno@xmath2 ( @xmath5 is a rare earth such as la and @xmath6 is a divalent element such as sr or ca ) due to their peculiar correlation between magnetism and conductivity have been extensively studied during 1950s and 60s @xcite .
the most throughly investigated was the la@xmath0sr@xmath1mno@xmath2 system .
undoped ( @xmath7=0 ) , lamno@xmath2 is an antiferromagnetic insulator . upon doping with sr ,
this perovskite oxide becomes a ferromagnetic metal ; the connection between metallicity and ferromagnetism was well explained by the double exchange hopping mechanism @xcite .
the discovery of colossal magnetoresistance phenomena in samples with sr dopant densities in the [email protected] regime @xcite brought a revival of the interest to these sytems .
the mn@xmath9 ion in the hole - undoped compound lamno@xmath2 has the high - spin @xmath10 electron configuration @xmath11 .
the @xmath12 orbitals hybridize with o2p orbitals much more weakly than the @xmath13 orbitals and can be regarded as forming the localized spin ( @xmath14=3/2 ) .
in contrast to that , @xmath13 orbitals , which have lobes directed to the neighboring oxygen atoms , hybridize strongly with o2p producing in the result rather broad bands .
the strong exchange interaction with @xmath15 subshell leads to the splitting of the @xmath13 band into unoccupied @xmath16 and half - occupied @xmath17 subbands .
the half filled @xmath17 subband is a typical example of the jahn - teller system , and , indeed , lamno@xmath2 has an orthorombic crystal structure @xcite with distorted ( elongated ) mno@xmath18 octahedra .
this cooperative jahn - teller effect is usually considered responsible for the opening of the gap in the half - filled @xmath17 band and an insulator ground state of lamno@xmath2 .
the orthorombic crystal structure of lamno@xmath2 ( @xmath19 space group ) can be described as a perovskite with two types of distortion from a cubic structure : 1)tilting ( rotation ) of the mno@xmath18 octahedra , so that mn - o - mn angles become less than 180@xmath20 , and 2)jahn - teller distortion of the octahedra , with one long mn - o bond and two short bonds . with doping by sr
the orhtorombicity ( the deviation of the values of the lattice parameters @xmath21 and @xmath22 from @xmath23 ) becomes smaller and for @xmath24 the stable crystal structure at room temperature becomes rombohedral , where only tilting of oxygen octahedra is present but all three mn - o bonds are equal . around the same value of the doping @xmath7 , la@xmath0sr@xmath1mno@xmath2 becomes metallic . for sr doping with [email protected]
, la@xmath0sr@xmath1mno@xmath2 is a ferromagnetic metal at low temperatures .
however , as the temperature increases and approaches the curie temperature @xmath25 a sudden increase of the resistivity is observed @xcite . as the temperature dependence of
the resistivity @xmath26 has typically semiconducting behavior ( @xmath27 ) above t@xmath28 , this increase in resistivity at around @xmath25 is usually desribed as a metal - insulator transition .
the colossal magnetoresistance effect happens when the temperature is close to @xmath25 : the external magnetic field leads to suppression of the resistivity increase or to shifting the metal - insulator transition to higher temperatures resulting in a very large negative value of @xmath29/\rho(0)$ ] .
the crucial point in the understanding of the effect of the colossal magnetoresistance is the nature of the metal - insulator transition with the temperature variation across @xmath25 .
it was shown @xcite that the double exchange mechanism alone is not enough to explain such transition .
_ @xcite suggested local jahn - teller distortions ( jahn - teller polarons ) as a main mechanism causing localization .
their idea is that a @xmath30 ratio ( @xmath31 is self - trapping jahn - teller polaron energy and @xmath32 is effective band width ) is close to the critical value in la@xmath0sr@xmath1mno@xmath2 and as @xmath32 decreases with destroying the ferromagnetic order above @xmath25 ( the effective hopping takes the maximum value for a parallel spin alignment ) , this ratio becomes larger than the critical value to result in polaron localization .
varma @xcite argues that lamno@xmath2 is not an insulator due to the jahn - teller distortion but is a mott insulator and the jahn - teller distortion occurs parasitically .
he explains localisation of the holes in the paramagnetic phase by spin polaron mechanism . by replacing la with other trivalent ions with smaller ionic radius ,
the mn - o - mn bond angle becomes smaller and the @xmath13 band width @xmath32 is reduced to enhance the tendency to the carrier localization and lattice distortion .
variety of dramatic phenomena have been observed in ( pr , nd , sm)(sr , ca)mno@xmath2 systems . as one of such examples , we study pr@xmath3ca@xmath3mno@xmath2 in the present paper .
this system has a very peculiar phase diagram @xcite . at low temperature
, it is an antiferromagenetic insulator with a charge ordering of mn@xmath33 and mn@xmath34 accompanied by an orbital ordering : the orbital ordering and spin ordering is of ce - type . between @xmath35 ( @xmath36 k ) and @xmath37 ( @xmath38 k )
, there is no long - range magnetic ordering but the system remains to be insulating because of the persisitence of the charge ordering . above @xmath39 , it is a paramagnetic insulator . the local spin density approximation ( lsda)@xcite was used by several groups for theoretical investigation of the electronic structure of la@xmath0sr@xmath1mno@xmath2 @xcite .
those calculations have confirmed the importance of the jahn - teller distortion for correct description by the lsda of the insulating antiferromagnetic ground state of the undoped lamno@xmath2 , because the calculations with an undistorted cubic perovskite crystal structure produce a half - filled metallic @xmath17 band . however in order to address the charge and orbital order observed in doped manganites it is necessary to go beyond lsda by including the intra d - shell coulomb interaction ( lsda+u method @xcite ) .
recently such an approach was used for treating the charge ordering in magnetite fe@xmath2o@xmath40 , where lsda gives a metallic uniform solution .
the lsda+u calculation gave a charge ordered insulator as the ground state of this compound @xcite .
in this paper we describe the results of the electronic structure calculation for pr@xmath0ca@xmath1mno@xmath2 for @xmath41 and 0.5 in the lsda+u approximation .
the problem of the charge ordering can not be treated properly by the standard lsda because of the non - vanishing self - interaction .
in contrast to the hartree - fock approximation , where the self - interaction is explicitly excluded for every orbital , in the lsda it is canceled to a good approximation in the total energy , but rather poorly in the local one - electron potentials , which are orbital - independent . the spurious self - interaction present in the lsda leads to an increase in the coulomb interaction when the distribution of the electron charge deviates from the uniform one .
this effect can be illustrated in the following way .
if we consider the system having less than one electron per site in the partially filled electronic subshell ( for example @xmath13-electrons in la@xmath0sr@xmath1mno@xmath2 ) then the potential which this electron feels at some site must not depend on the orbital occupancy of this particular site , as the electron does not interact with itself .
however , as the lsda potential is a functional of the electron density only , increasing the electron density at one site and decreasing it at another one , with a formation of the charge ordering , will lead to an increase of the potential at the first site and a decrease at the second one . as a result , in the self - consistency loops
, the charge distribution will return to the uniform density . in order to cure this deficiency ,
it is necessary to remove the self - interaction .
the lsda+u method @xcite is one of the practical methods to solve the self - interaction problem .
let us introduce the d - orbitals occupation matrix defined by : @xmath42 where @xmath43 are the elements of the green function matrix ( @xmath44 : index of atom site , @xmath45 : main and orbital quantum numbers ( for example 3@xmath46 ) , @xmath47 : index of a particular @xmath46-orbital and @xmath48 : spin projection index ) . in terms of the elements of this occupation matrix @xmath49
, the lsda+u functional @xcite is defined as follows : @xmath50=e^{lsda}[\rho ^\sigma ( { \bf r)}]+e^u[\{n^\sigma \}]-e_{dc}[\{n^\sigma \}]\ ] ] where @xmath51 is the charge density for spin-@xmath52 electrons and @xmath53 $ ] is the standard lsda ( local spin - density approximation ) functional .
@xmath54 $ ] is the hartree - fock type electron - electron interaction energy given by @xmath55=\frac 12\sum_{\{m\},\sigma } \{\langle m , m^{\prime \prime } \mid v_{ee}\mid m^{\prime } , m^{\prime \prime \prime } \rangle n_{mm^{\prime } } ^\sigma n_{m^{\prime \prime } m^{\prime \prime \prime } } ^{-\sigma } \\ \\ -(\langle m , m^{\prime \prime } \mid v_{ee}\mid m^{\prime } , m^{\prime \prime \prime } \rangle -\langle m , m^{\prime \prime } \mid v_{ee}\mid m^{\prime \prime \prime } , m^{\prime } \rangle ) n_{mm^{\prime } } ^\sigma n_{m^{\prime \prime } m^{\prime \prime \prime } } ^\sigma \ } , \end{array}\ ] ] where @xmath56 is the screened coulomb interaction among the @xmath45 electrons .
finally , the last term in eq.([u1 ] ) corrects for double counting and is given by @xmath57=\frac 12un(n-1)-\frac 12j[n^{\uparrow } ( n^{\uparrow } -1)+n^{\downarrow } ( n^{\downarrow } -1)],\ ] ] were @xmath58 and @xmath59 @xmath60 and @xmath61 are screened coulomb and exchange parameters @xcite .
in addition to the usual lsda potential , the variation of the functional ( [ u1 ] ) gives an effective single - particle potential to be used in the single - particle hamiltonian : @xmath62 @xmath63 the matrix elements of the screened coulomb interaction @xmath56 can be expressed in terms of complex spherical harmonics and effective slater integrals @xmath64@xcite as @xmath65 where @xmath66 and @xmath67 for d - electrons one needs @xmath68 and @xmath69 and these can be linked to the coulomb- and stoner parameters @xmath60 and @xmath61 obtained by the lsda - supercell procedures @xcite via @xmath70 and @xmath71 with the ratio @xmath72 being to a good accuracy a constant @xmath73 for the 3d elements @xcite .
the lsda+u approximation was realized in the framework of the linear muffin - tin orbital method in the atomic sphere approximation ( lmto - asa)@xcite .
the values of @xmath60=7.9 ev and @xmath61=0.9 ev were used as obtained from the constrained lsda supercell calculations .
as a first step the electronic structure of the undoped prmno@xmath2 was calculated .
this compound has practically the same properties as lamno@xmath2 .
the only difference is that due to the smaller ionic radius of pr ion compared with that of la ion , the tilting of the oxygen octahedra in orthorombic @xmath19 crystal structure is stronger , resulting in smaller effective @xmath74 hopping between mn atoms and narrower band width
. the crystal structure of the doped pr@xmath0ca@xmath1mno@xmath2 does not show transition to the rombohedral symmetry , as it is the case for la@xmath0sr@xmath1mno@xmath2 , preserving the orthorombic @xmath19 space group@xcite .
however the values of the lattice parameters @xmath23,@xmath21 and @xmath22 for @xmath7=0.5 are so close that it can be called a pseudocubic structure . in this structure
only the tilting of the oxygen octahedra is present , but octahedra themselves are not jahn - teller distorted as in the undoped crystal structure . in order to study effects of the orbital polarization purely due to the intrashell d - d coulomb interaction without any influence of the jahn - teller lattice distortion the calculation for the undoped prmno@xmath2
was performed with the crystal structure parameters corresponding to pr@xmath3ca@xmath3mno@xmath2 .
kugel and khomskii @xcite demonstrated for a wide range of jahn - teller magnetic compounds that the exchange mechanism ( caused by an interplay between on - site coulomb interaction @xmath60 and intersite hopping @xmath75 ) by itself , without taking janh - teller distortions into consideration , is able to give a correct picture of the spin and orbital ordering . in this case the change in the lattice structure ( the structural transition ) is a secondary effect .
one may say that the system shows the `` jahn - teller ordering '' without the `` jahn - teller interaction '' .
actually an exchange mechanism acts jointly with the jahn - teller and the quadrupole mechanisms .
it is quite difficult to determine just which mechanism will be predominant in each specific case due to complexity of the real compound .
one of the possible ways to answer this question in the frame of _ ab - initio _ band structure calculations is to use a crystal structure excluding distortions caused by the jahn - teller interaction .
such kind of approach was used for the cooperative jahn - teller system kcuf@xmath2 @xcite .
it also has the perovskite crystal structure , but there is no tilting of the cuf@xmath18 octahedra and all cu - f - cu angles are equal to 180@xmath20 .
divalent copper ion in this compound has a @xmath76 configuration ( @xmath77 ) known to be a jahn - teller ion , and , indeed , the jahn - teller distortion is observed in the form of the elongated cuf@xmath18 octahedra with long axes alternatively along @xmath23 and @xmath21 axes in the @xmath78-plane .
kugel and khomskii @xcite showed that a simple hubbard model type calculation in a mean - field approximation without taking into account the electron - phonon interaction could predict the orbital ordering ( alternating @xmath79 and @xmath80 cu3d - orbitals for electrons or @xmath81 and @xmath82 orbitals for holes ) corresponding to the symmetry of the experimentally observed lattice distortions .
the lsda+u functional ( eq.[u1 ] ) has a definite important improvement upon the lsda : explicit cancellation of the self - interaction and the proper description of the lowering of coulomb interaction energy with development of the orbital polarization .
the functional ( eq.[u1 ] ) is also rotationally invariant ( the total energy does not change with the unitary transformation of the d - orbital basis set @xmath83 ) , meaning that if the local coordinate system ( and , correspodingly d - orbital basis set @xmath83 ) was chosen inappropriately for describing the resulting orbital polarisation , the unitary transformation which diagonalizes the non - diagonal occupation matrix @xmath84 ( eq.[occ ] ) will give a proper orbital basis set . the lsda+u calculation for kcuf@xmath2 @xcite in _ cubic _ perovskite crystal structure gave results fully confirming kugel - khomskii s prediction - alternating @xmath79 and @xmath80 orbital order .
moreover , it was not just the lowest total energy configuration among many possible stationary configurations as in the model calculation @xcite , but the _ only one stable solution_. the lattice relaxation to minimize the total energy is in good agreement with experimental data .
note that the standard lsda did not show the lattice instability with respect to the jahn - teller distortion in kcuf@xmath2 .
we would like to emphasize that the orbital polarization caused by the d - d coulomb interactions practically did not change with the lattice distortion @xcite .
the use of the crystal structure parameters corresponding to pr@xmath3ca@xmath3mno@xmath2 ( with undistorted oxygen octahedra ) for prmno@xmath2 may correspond to the use of the cubic perovskite crystal structure for kcuf@xmath2 .
however , for prmno@xmath2 , we have to take into account the tilting of mno@xmath18 octahedra ( present in pr@xmath3ca@xmath3mno@xmath2 structure ) , which is essential for the proper description of the reduction in the effective @xmath74 hopping due to the bending of the mn - o - mn bond . the lsda+u calculation with the experimental a - type antiferromagnetic spin alignment resulted in a non - diagonal occupation matrix for spin density ( for the @xmath85 subspace ( @xmath86 and @xmath87 orbitals ) of one particular mn atom ) : @xmath88 the diagonalization of this matrix gives two new @xmath17 orbitals : @xmath89 with occupancy 0.71 and @xmath90 with occupancy 0.18 . for the second type of mn atom @xmath91 and @xmath92
orbitals can be obtained by transposition of @xmath7 and @xmath93 .
the resulting orbital order can be presented graphically by plotting the angle distribution of the @xmath13-electron spin density : @xmath94 as a mn@xmath9 ion has formally only one electron in partially filled @xmath17 subshell , the @xmath13-spin density must correspond to the density of this electron ( fig.1 ) .
the plot of fig.1 and the form of the @xmath91 and @xmath92 orbitals diagonalizing the spin density occupation matrix show that the orbital order obtained in our lsda+u calculations for prmno@xmath2 is of the same symmetry as the one in kcuf@xmath2 . the `` parasitical '' lattice distortion , which one could expect from such orbital ordering : elongation of mno@xmath18 octahedra in the direction of the lobes of @xmath91 orbitals ,
will reproduce the experimentally observed jahn - teller lattice distortion .
the orbital polarization is far from being 100% , the difference in the occupancy of @xmath91 and @xmath92 orbitals is only 0.53 .
the reason for this is the strong hybridization of @xmath17-orbitals with oxygen 2p - orbitals . in fig.2 , the total and partial densities of states ( dos ) for prmno@xmath2
are shown ( for @xmath13-electrons dos projected on @xmath91 and @xmath92 orbitals are also presented ) .
the electronic structure obtained in our lsda+u calculation is semiconducting with a band gap of 0.5 ev .
note that the standard lsda needs the jahn - teller distorted crystal structure to reproduce the non - metallic ground state .
the @xmath17-band is split by this gap into two subbands : the occupied one with the predominantly @xmath91 character , and the empty one with the @xmath92 character .
however there is significant admixture of @xmath92 in the occupied subband and @xmath91 in the empty one . after checking that our calculation scheme gives a proper orbital order in the undoped prmno@xmath2 , we applyed it to pr@xmath3ca@xmath3mno@xmath2 .
in order to treat the experimentally observed ce - type of antiferromagnetism it was necessary to quadruple the @xmath19-type unit cell and the supercell had 16 formula units .
the distribution of the pr and ca atoms in the lattice was chosen according to a model where all pr ions have only ca as their nearst neighbores in the lanthanide sublattice .
we did not impose any symmetry restrictions in our calculation , the integration in the reciprocal space being performed for the whole brillouin zone , and the self - consistency iteration was started from a uniform distribution of @xmath13-electrons over both @xmath13-orbitals of all mn atoms .
the result of the self - consistent calculation was a charge and orbital ordered insulator .
the @xmath13-electron spin - density - plot calculated by eq.[f1 ] is presented in fig.3 .
this is in a striking agreement with the orbital order derived in @xcite from the neutron diffraction mesuarements ( fig.4 ) .
the unexpected result is that the total number of d - electrons at all types of mn sites are nearly equal ( 4.99 and 5.01 ) , so that formally mn@xmath34 ions ( in the fig.3 the ones with symmetric in - plane density distribution ) and mn@xmath33 ions ( the ones with the density strongly anisotropic ) have nearly the same number of 3d - electrons .
however the difference in the magnetic moment values is more pronounced : 3.34@xmath95 for mn@xmath34 and 3.44 @xmath95 for mn@xmath33 .
the results of our lsda+u calculations for pr@xmath0ca@xmath1mno@xmath2 demonstrate that taking into account the coulomb interaction inside the d - shell of mn together with the intersite hopping is enough to reproduce not only the orbital order present in the undoped manganites , but also the localization and ordering of the holes in the doped materials .
naturally the lattice relaxation ( jahn - teller and polaron lattice distortions ) will significantly renormalize the total energies , gap values , charge disproportion and magnetic moments .
however it seems that the symmetry of the ordered state can be obtained without explicitly including the electron - lattice interactions .
this is in clear contrast to the standard lsda where energy band splitting and orbital polarization appear only as results of the jahn - teller crystal structure distortion .
v.i.a . and m.a.k .
acknowledge the nair guest research program .
the present work is supported partly by russian foundation for basic research ( rfbr grant 96 - 02 - 16167 ) , new energy and industrial development organization in japan and also by a grant - in - aid for scientific research in the priority area `` anomalous metallic states near the mott transition '' from the ministry of education , science and culture of japan .
99 g. h. jonker and j. h. van santen , physica ( amsterdam ) * 16 * , 337 ( 1950 ) ; j. h. van santen and g. h. jonker , physica ( amsterdam ) * 16 * , 599 ( 1950 ) . c. zener , phys
. rev . * 82*,403 ( 1951 ) ; p. w. anderson and h. hasegava , phys . rev . * 100 * , 675 ( 1955 ) ; p. g. de gennes , phys . rev . * 118 * , 141 ( 1960 ) .
r. m. kusters , j. singleton , d. a. keen , r. mcgreevy and w. hayes , physica ( amsterdam ) * 155b * , 362 ( 1989 ) ; a. urushibara , y. moritomo , t. arima , a. asamitsu , g. kido and y. tokura , phys . rev.b * 51*,14103 ( 1995 ) .
j. b. a. a. elemans , b. van laar , k .r .
van der veen and b. o. loopstra , j. solid state chem .
* 3 * , 238 ( 1971 ) .
a. j. millis , p. b. littlewood and b. shraiman , phys .
. lett . * 74 * , 5144 ( 1995 ) .
a. j. millis , b. shraiman and r. mueller , phys .
lett . * 77 * , 175 ( 1996 ) . c. m. varma , preprint .
z. jirak , s. krupicka , v. nekvasil , e. pollert , g. villeneuve and f. zounova , j. magn .
magn . mater . *
15 - 18 * , 519 ( 1980 ) ; z. jirak , s. krupicka , z. simsa , m. dlouha and z. vratislav , @xmath96 * 53 * , 153 ( 1985 ) .
y. tomioka , a. asamitsu , h. kuwahara , y. moritomo and y. tokura , phys .
rev . b*53 * , r1689 ( 1996 ) .
p. hohenberg and w. kohn , phys . rev .
* 136 * , b864 ( 1964 ) ; w. kohn and l. j. sham , _ ibid . _
* 140 * , a1133 ( 1965 ) .
n. hamada , h. sawada and k. terakura , in _ spectroscopy of mott insulators and correlated metals _ , edited by a. fujimori and y. tokura ( springer - verlag , berlin , 1995 ) , pp.95 - 105 .
w. pickett and d. singh , phys .
rev.b * 53 * , 1146 ( 1996 ) .
s. satpathy , z. popovic and ph .
vukajlovic , phys .
76 * , 960 ( 1996 ) . v. i. anisimov , j. zaanen and o. k. andersen , phys .
b * 44 * , 943 ( 1991 ) . v.
i. anisimov , i. v. solovyev , m. a. korotin , m. t. czyzyk and g. a. sawatzky , phys.rev .
b * 48 * , 16929 ( 1993 )
. a. i. lichtenstein , j. zaanen and v. i. anisimov , phys .
rev.b*52 * , r5467 , ( 1995 ) .
v. i. anisimov , i. s. elfimov , n. hamada and k. terakura , phys.rev.b * 54 * , 3847 ( 1996 ) o. gunnarsson , o. k. andersen , o. jepsen and j.zaanen , phys .
rev . b*39 * , 1708 ( 1989 ) . v.
i. anisimov and o. gunnarson , phys .
b * 43 * , 7570 ( 1991 ) .
b. r. judd , operator techniques in atomic spectroscopy , mcgrow - hill , new york,1963 .
f. m. f. de groot , j. c. fuggle , b. t. thole and g. a. sawatzky , phys .
b * 42 * , 5459 ( 1990 )
. o. k. andersen , phys .
b * 12 * , 3060 ( 1975 ) . k. i. kugel and d. i. khomskii , usp . fiz . nauk . * 136 * , 621(1982 ) [ sov .
phys . -usp . * 25 * , 231(1982 ) ] .
the lsda works much better in lamno@xmath2 and perhaps also in prmno@xmath2 than in kcuf@xmath2 . in the latter case
, the lsda does not give even a stable antiferromagnetic state .
however , in lamno@xmath2 and prmno@xmath2 , the antiferromagnetic state can be obtained by lsda .
our recent calculation ( h. sawada , et al . ) for lamno@xmath2 reproduces about 50% of the experimental jahn - teller distortion by the theoretical structural optimization for experimental lattice parameters .
2 . the total and partial densities of states ( dos ) from the result of lsda+u calculation of prmno@xmath2 .
a)total dos per unit cell ( 4 formula units ) ; b)partial @xmath13 dos of mn .
solid lines are for the @xmath91-projected dos , and dashed lines for the @xmath92-projected dos with @xmath91 and @xmath92 -orbitals diagonalizing the occupation matrix . ;
c ) and d ) partial dos for two types of oxygen orbitals . fig .
the scheme of the spin , charge and orbital order in pr@xmath3ca@xmath3mno@xmath2 deduced from the neutron difraction data @xcite .
the open circles with the lobe of the @xmath13 electron density distribution denote mn@xmath33 and the filled circles mn@xmath34 .
the arrows denote the magnetic moments . | the electronic structure of the doped manganites pr@xmath0ca@xmath1mno@xmath2 was calculated by the lsda+u method which takes into account the local coulomb interaction between d - electrons of transition metal ions .
in contrast to the standard local spin density approximation ( lsda ) no jahn - teller lattice distortions are needed to reproduce the experimentally observed orbital ordering in the undoped prmno@xmath2 .
the correct ground state : a charge and orbital ordered antiferromagnetic insulator , was obtained for pr@xmath3ca@xmath3mno@xmath2 .
the results are in good agreement with the neutron diffraction data . |
the _ spitzer space telescope _
@xcite has conducted numerous wide - area surveys , yielding a rich legacy of image data and point source catalogs that have become heavily used commodities . in terms of sky coverage ,
the largest of these are the galactic legacy infrared mid - plane survey extraordinaire programs ( glimpse , glimpse ii , glimpse 3d ; * ? ? ?
* ; * ? ? ?
* ) covering almost 400 square degrees of the milky way plane in all four mid - infrared bands with the infrared array camera ( irac ; * ? ? ?
the @xmath1 mapping of the outer galaxy ( smog ; * ? ? ? * ) , the @xmath1 legacy survey of the cygnus - x complex @xcite , and the vela - carina large program @xcite together cover another 120 square degrees . the warm @xmath1 mission program glimpse 360 @xcite imaged @xmath2511 square degrees of the outer galactic plane in the 3.6 and 4.5 @xmath0 m bands .
deep glimpse @xcite re - imaged a similar portion of the galactic plane as glimpse i / ii , but only at 3.6 and 4.5 @xmath0 m , using a longer 10.4 s integration time .
other galactic programs and numerous extragalactic legacy programs provided much deeper datasets but over relatively small fields of view .
our focus here is primarily the galactic science programs that have yielded large lists of point sources .
table [ surveys.tab ] summarizes the approximate coverages , exposure times , observational bandpasses , and total number of photometered point sources for these and other galactic _ spitzer _ survey .
there are two types of glimpse source lists : a high reliability point source `` catalog '' and a more complete point source `` archive '' .
the source lists are a result of doing photometry on each irac frame , averaging all detections made in a single band ( in - band merge ) , then merging the photometry from all wavelengths , including 2mass @xmath3 sources , at a given position on the sky ( cross - band merge ) .
the glimpse source list criteria have been developed to ensure that each source is a legitimate detection and that the fluxes reported for the irac bands are of high quality . as of this writing , these @xmath1 programs have generated nearly 158 million sources in the highly reliable _ glimpse point source catalogs _
( gpsc ) and 229 million in the more complete _ glimpse point source archives _ ( gpsa ) .
the utility of point source catalogs resulting from the _ spitzer _ mission depends , partly , on understanding their completeness ( i.e. , the ratio of true sources detected to the number of all true sources ) . the stellar magnitude at which the gpsc or gpsa is nominally complete varies with bandpass .
it also varies greatly as a function of background surface brightness and as a function of local source density where confusion with neighboring sources limits detection and photometry .
figure 2 of @xcite shows how high diffuse background levels inhibit the detection of point sources in the _ spitzer _ 8.0 @xmath0 m band , which is affected by strong , diffuse emission features arising from polycyclic aromatic hydrocarbons ( pahs ) in and around regions of star formation . the irac 5.8 @xmath0 m band , and to a much lesser extent the 3.6 @xmath0 m band ,
also encompass these features , leading to strong spatial completeness variations , especially in complex fields near the galactic plane .
it is widely recognized that sensitivity to point sources is significantly reduced in high - background fields and fields having high point source densities , but there has not yet been a general quantitative analysis of this effect . our goal in this contribution is to quantify the photometric completeness in the four _ spitzer _ irac bandpasses , considering separately the effects of 1 ) high diffuse backgrounds , and 2 ) high point source surface densities . in section 2 we describe an artificial star analysis procedure for addressing the effects of incompleteness resulting from diffuse background emission . in section 3
we provide plots and tables of completeness as a function of diffuse background level and stellar magnitude . in section 4
we address the effects of incompleteness resulting from source crowding ( i.e. , confusion ) , and we provide plots and tables of the magnitude at which 90% completeness is achieved as a function of point source surface density . all magnitudes referenced herein refer to the vega magnitude system used by the glimpse pipeline and data products .
glimpse i obtained two 1.2 s exposures of each object , while glimpse ii obtained three such 1.2 s exposures .
glimpse 3d utilized either two or three 1.2 s exposures at each location .
glimpse 360 and deep glimpse employed the high - dynamic - range ( hdr ) mode to obtain three 0.4 and 10.4 s exposures at each position .
we adopt the glimpse i / ii/3d survey strategy and its resulting gpsc and gpsa as our baseline dataset , and we perform a similar analysis for the glimpse 360/deep glimpse survey strategy . although it is unrealistic to perform a completeness characterization for every possible survey strategy and sightline , we intend that these two results be general enough that consumers of source lists generated as part of glimpse and similarly processed _ spitzer _ irac surveys , such as smog , cygnus - x , and vela - carina can infer completeness for those programs as well . this completeness analysis is pertinent only to point sources and is not appropriate for extended sources .
we selected a @xmath21.5 square degree region of the glimpse i survey between @xmath4 , @xmath5 , exhibiting a large range in diffuse background level to use as a test region .
this area contains some of the brightest mid - ir background regions in the galactic plane .
figure [ pretty ] shows a logarithmic greyscale representation of this region in the 8 @xmath0 m irac band .
this region of the galactic plane spans a large dynamic range from 20 mjy / sr in the lower left to over 5000 mjy / sr .
yellow crosses show detected 8.0 @xmath0 m point sources fainter than 12th magnitude as included in the glimpse point source catalog .
the varying density of cataloged sources from lower left to upper right dramatically illustrates the effects of high and complex background levels on completeness . the glimpse i data in this region is comprised of @xmath2600 individual 1.2-second exposures in each of the four @xmath6 bandpasses .
the irac pixel size is 1.2 at all four bandpasses . to each 256@xmath7256 pixel
basic calibrated data ( bcd ) image we added roughly 600 artificial stars at random locations using an average psf appropriate for each band , as constructed by the glimpse data processing team from an ensemble of irac point sources .
the psf was scaled to a random magnitude between @xmath8 and @xmath9 , gaussian noise ( @xmath10 ) was added to each artificial star psf on a pixel - by - pixel basis , and the artificial star was added at a ( pseudo- ) random location to the bcd image .
the center locations of the artificial stars were constrained to 1 ) lie no closer than 5 pixels from the edge of the image , and 2 ) lie no closer than 3 pixels to a real star detected during previous processing by the glimpse pipeline ( dubbed the `` actual '' list of point sources ) .
this avoids both edge effects and the effects of confusion from neighboring point sources in order to focus on effects solely related to diffuse background levels .
we set @xmath8=8 @xmath9=15 for all bands , spanning seven magnitudes in brightness to cover nearly the entire dynamic range of the @xmath11 survey in order to sample both very complete and very incomplete flux levels .
the artificial stars have the same magnitude in each band .
effectively , this means that we are simulating sources having zero color , which comprise the vast majority ( @xmath294% of sources have @xmath12 ) of all gpsc sources having measurements in all four bands .
however , it is not practical to conduct a general analysis for arbitrary spectral energy distributions .
the glimpse pipeline was run to generate a list of point sources from the actual+artificial images in the same manner as the published _
glimpse point source catalogs _ and _ glimpse point source archives_. inclusion in the highly reliable _ glimpse point source catalog _ requires that a source be detected _ twice _
( i.e. , in at least two individual irac exposures ; detection is performed on the bcd frames ) in any one irac band at s / n@xmath135 and _ at least once _ in any adjacent band at s / n@xmath135 , where the 2mass k@xmath14 band is allowed as an adjacent band for irac 3.6 @xmath0 m .
hence , it is possible for a source to be included in the gpsc but lack photometry in as many as two ( or even three ) of the four irac bands .
most commonly , non - detections occur at irac bands 5.8 and/or 8.0 @xmath0 m as a result of the reduced instrumental sensitivity and the decreasing flux on the rayleigh - jeans tail of the spectral energy distribution for the majority of astrophysical sources .
objects with rising spectral energy distributions , such as protostars , evolved stars exhibiting circumstellar excesses , or heavily extincted objects , could have detections at irac 5.8 & 8.0 @xmath0 m but not at the shorter wavelengths .
inclusion in the catalog also requires that 1 ) a source does not fall within the psf wings of a saturated source ( as defined in the glimpse data release documentation ) , 2 ) no hot or dead pixel lies within 3 pixels of source center , 3 ) the source is not within 3 pixels of a frame edge , and 4 ) no archive source lies within 2 of the source . furthermore , the sources used to identify and confirm a detection must be brighter than 0.6 mjy ( 14.2 mag ) , 0.4 mjy ( 14.1 ) , 2 mjy ( 11.9 ) , and 10.0 mjy ( 9.5 ) in the 3.6 , 4.5 , 5.8 , and 8.0 @xmath0 m bands , respectively .
however , sources fainter than these limits at a particular band may appear in the catalogs if they are identified and confirmed using _ other _ bands .
users should consult the glimpse data release documentation for a full description of source selection and rejection criteria .
inclusion in the less reliable but more complete _ glimpse point source archive _ requires that a source be detected at least once in any two bands at s / n@xmath133 and that there are no neighboring sources within 0.5 .
sources that lie within the psf wings of a saturated star _ are _ included in the archive .
this last criterion means that the completeness we estimate for the catalog is going to be suppressed relative to the archive at all magnitudes and at all background levels because of the presence of saturated stars .
the saturation limit for the short 1.2 s glimpse i / ii/3d frame times is approximately 7 , 6.5 , 4 , and 4 magnitudes ( 439 , 450 , 2930 , and 1590 mjy ) for irac bands 3.6 , 4.5 , 5.8 , and 8.0 @xmath0 m , respectively .
the saturated source areal density is a strong function of galactic position but is on the order of @xmath240 per square degree at @xmath15 , as estimated from 2mass k - band source counts .
band is used as a proxy for the irac 3.6 @xmath0 m band because the glimpse pipeline does not detect or extract saturated sources . ] therefore , the effect of saturated stars on completeness in the glimpse i / ii/3d catalogs and similar shallow survey catalogs is minimal , given that the psf wings ( @xmath1624 pixels , but an irregular region given by the shape of diffraction spikes ) comprise @xmath171% of the survey area . to characterize the completeness at each band the list of detected point sources from the `` actual+artificial '' images
was compared to the input list of artificial sources ( i.e. , the artificial list ) .
any detected source within 1.2 pixels of a source in the `` artificial '' list was considered to be a successfully recovered source .
we find that the magnitudes of recovered stars always match the magnitudes of inserted stars within the expected uncertainties , typically 0.05 mag at the bright end and @xmath20.5 mag at the faint end .
therefore , the matching process does not include a magnitude criterion .
we found it necessary to develop two definitions of completeness . in the first definition , dubbed completeness1 ,
we define completeness as the number of artificial sources present in the resulting gpsc divided by the number of input artificial sources .
this inclusive definition allows for the possibility that a given star may be included in the gpsc _ even though it may lack a detection _ at any given band , n. in the second definition , completeness2 , we require that a given source actually be detected in band n specifically , rather than being included in the catalog only by virtue of its detection at other bands .
the more restrictive completeness2 values are always lower than completeness1 , especially at 5.8 and 8.0 @xmath0 m where the deleterious effects of high and structured backgrounds on source identification are most severe .
figure [ comp1a ] ( upper panel ) plots completeness1 versus background sky brightness in mjy sr@xmath18 for the irac 3.6 @xmath0 m artificial star simulations .
the ordinate is partitioned into logarithmically spaced sky brightness bins to increase the reliability of high - surface - brightness bins which comprise only a small fraction of the test area .
linestyles indicate stellar magnitude bins from 8th to 15th magnitude , as shown by the legend .
the heavy solid curve shows the cumulative distribution of the background sky brightness .
for example , the solid curve shows that approximately 35% of the background regions in our test area are brighter than 5 mjy sr@xmath18 at 3.6 @xmath0 m .
the rapid decline in completeness at background levels greater than about 10 mjy sr@xmath18 is clear in all but the brightest magnitude ranges . furthermore , the faintest magnitude bin , 14th
15th magnitude , is never more than 25% complete even in the regions of lowest sky brightness .
we attempted to fit simple analytical formulae to these data to provide an easily integrable functional form for completeness as a function of magnitude and background level , but no simple form ( gaussian , exponential , trigonometric , polynomial ) yielded good fits across the range of magnitudes and background levels pictured in figure [ comp1a ] .
we therefore elected to simply tabulate the results and allow end users to perform their own interpolation or fitting .
table [ comp1cat.tab ] lists numerical values for completeness1 as a function of magnitude and background level for the 3.6 @xmath0 m bandpass as shown graphically in figure [ comp1a ] .
note that the uncertainties on the tabulated completeness are at the level of 1015% for the bin at the highest background levels as a result of statistical limitations , but in the other bins uncertainties are typically @xmath192% .
statistics in the lowest surface brightness bin are also uncertain at the level of a few percent as a result of the paucity of extremely dark regions in the galactic plane .
figure [ comp1a ] ( lower panel ) plots completeness2 versus background sky brightness for the irac 3.6 @xmath0 m .
table [ comp2cat.tab ] lists these results in the same manner as for completeness1 in table [ comp1cat.tab ] .
the results are nearly identical to the completeness1 case because pah emission from diffuse background sources is relatively minor at 3.6 @xmath0 m and the flux from stellar photospheres is a maximum , thereby aiding source identification and extraction .
sources in the gpsc are most commonly included by virtue of having a detection at irac 3.6 @xmath0 m .
figure [ comp2a ] plots completeness1 and completeness2 versus background sky brightness for the irac 4.5 @xmath0 m artificial star simulations .
the overall trends are similar to 3.6 @xmath0 m where both completeness1 and completeness2 decline markedly in backgrounds above about 10 mjy sr@xmath18 . only the brightest stars , @xmath2011th magnitude , are readily detected and show a slower drop in completeness .
the numerical results for completeness2 are nearly identical to completeness1 .
this results from the fact that background levels are low in the 4.5 @xmath0 m band ( no pah features ) and that sources in the gpsc are almost always included by virtue of having a detection at 4.5 @xmath0 m .
tables [ comp1cat.tab ] and [ comp2cat.tab ] record these numerical completeness data for the 4.5 @xmath0 m analysis . figure [ comp3a ] plots completeness1 and completeness2 versus background sky brightness for the irac 5.8 @xmath0 m simulations .
these histograms show dramatically decreased completeness levels for all but the brightest stars as the background levels exceed @xmath250 mjy sr@xmath18 .
the irac 5.8 @xmath0 m band contains strong pah features , and regions with specific intensities exceeding 200 mjy sr@xmath18 are not uncommon in this and other major star - forming complexes .
a comparison of figure [ comp3a ] shows that completeness2 is significantly lower than completeness1 for stars fainter than about 11th magnitude for all background levels greater than @xmath2100 mjy sr@xmath18 .
this results from sources being included in the catalog ( completeness1 ) but not having a detection at 5.8 @xmath0 m ( completeness2 ) .
tables [ comp1cat.tab ] and [ comp2cat.tab ] record these numerical completeness data for the 5.8 @xmath0 m analysis . figure [ comp4a ] plots completeness1 and completeness2 versus background sky brightness for the 8.0 @xmath0 m simulations .
while the behavior of the calculated completeness is similar to the plots for 5.8 @xmath0 m above , the difference between completeness1 and completeness2 is even more dramatic .
completeness2 shows a much more rapid decline than completeness1 , to the point where sources fainter than 12th magnitude are rarely detected in any but the darkest background regions .
the difference between the upper and lower panels of figure [ comp4a ] illustrates why the number of 8.0 @xmath0 m detections in glimpse averages @xmath2030% the number of 3.6 @xmath0 m detections for a typical low - latitude sightline . in regions with background levels
greater than 100 mjy sr@xmath18 , the lower panel shows that only 11th magnitude and brighter sources are likely to be represented in the catalog and have a measurement at 8.0 @xmath0 m .
however , on a more positive note , the upper panel shows that faint 8.0 @xmath0 m sources _ are _ likely to be represented in the catalogs by virtue of having measurements at other bands , principally at 3.6 and 4.5 @xmath0 m . in summary , figures [ comp1a ] through [ comp4a ] ,
in conjunction with tables [ comp1cat.tab ] and [ comp2cat.tab ] provide a means of estimating the likelihood of detecting sources in a particular band and being included in the glimpse point source catalogs .
these figures and tables should also be applicable to other _ spitzer _
irac observations using similar short exposure times , low point source densities , and a small number of observations of each target , such as the vela - carina project which was processed by the glimpse pipeline .
we performed the same analysis for the gpsa as for the gpsc .
figures [ comp1aa ] through [ comp4aa ] plot completeness1 and completeness2 as a function of background brightness for the archive .
tables [ comp1arc.tab ] and [ comp2arc.tab ] summarize these results numerically .
a comparison of figures [ comp1aa ] through [ comp4aa ] with the corresponding sequence in figures [ comp1a ] through [ comp4a ] reveals that more stars are recovered in the archives versus the catalogs at all bands .
the typical difference is a few percent to as much as a few tens of percent , with the most dramatic improvement being at low background levels for the faintest stars .
glimpse 360 and deep glimpse employed the irac high - dynamic - range mode using 3.6 and 4.5 @xmath0 m bands during the _ spitzer _ warm mission to obtain three 0.4 and three 10.4-second exposures at each location along a large portion of the galactic plane following the warp of the disk in the outer milky way .
we employed the same artificial star methodology described above in order to generate completeness data for the 3.6 and 4.5 @xmath0 m photometry obtained in the course of these large programs .
the selected test region covers a wide range of background levels in the region @xmath21 and @xmath22 obtained during the glimpse 360 observing program . the glimpse 360 completeness analysis adds and recovers artificial stars only on the 10.4 s exposures , since we are interested primarily in completeness for the faint end of the observed magnitude range .
the glimpse 360 and deep glimpse observing strategy is sufficiently different from the preceding glimpse programs that slightly different criteria are applied to determine inclusion in the catalogs and archives . given three hdr sequences ( an hdr sequence means one 0.4 s exposure plus one 10.4 s exposure ) , there are six possible frames at each location in each of two bands where a source might be detected . in general , m is the number of detections for a given source out of a possible n detections .
inclusion in the glimpse 360/deep glimpse catalog requires m / [email protected] in one band ( the selection band ) and m / n@xmath23 in an adjacent band ( the confirmation band ) , with s / n@xmath135 for both bands .
detection in the 2mass k@xmath24 band counts as a confirmation , but in practice , only a few percent of all glimpse 360 sources are included in the catalog by virtue of having a k - band detection in addition to a 3.6 @xmath0 m detection .
additionally , inclusion in the catalog requires that 1 ) the source be fainter than the non - linear regime for the irac array ( 10.25 and 9.6 mag at 3.6 and 4.5 @xmath0 m , respectively , for the irac 10.4 s exposure times ; the hdr 0.4 s data has brighter limits but is not considered here in our simulations ) , 2 ) that the source not lie within 2 of an archive source , 3 ) that the source not lie within the diffraction spikes of a saturated source , and 4 ) that no hot or dead pixels lie within 3 pixels of the source center .
inclusion in the archive requires the same m / n limits but sources within 0.5 of another source are allowed , as are sources within the psf wings of a saturated source .
end users should consult the glimpse 360 data delivery document for a full understanding of source flags and subtle criteria that define the catalog and the archive sources .
figure [ comp5a ] shows the completeness1 and completeness2 results for the glimpse 360/deep glimpse catalog for the 3.6 @xmath0 m band .
figure [ comp6a ] shows the completeness1 and completeness2 catalog results for 4.5 @xmath0 m .
the histogram linestyles indicate magnitude ranges from 11 18 .
tables [ g360c1.tab ] and [ g360c2.tab ] report numerical data for the glimpse 360/deep glimpse catalog simulations for the 3.6 and 4.5 @xmath0 m bands .
these figures show that a similar trend to the 3.6 and 4.5 @xmath0 m results from the glimpse i / ii/3d simulations reported above .
completeness varies as a function of both magnitude and background level . given the deeper exposures for glimpse 360/deep glimpse compared to glimpse i / ii/3d , the completeness is generally greater at
any given magnitude at any given background level .
one noteworthy feature of these histograms is the drop in completeness at all magnitudes , even at relatively faint sky background levels .
this means that the gpsc for glimpse 360/deep glimpse is almost never 100% complete , even at the brightest magnitudes and darkest sky backgrounds .
the reason for this is described subsequently .
figure [ comp7a ] shows the completeness1 and completeness2 results for the glimpse 360/deep glimpse archive for the 3.6 @xmath0 m band .
figure [ comp8a ] shows the completeness1 and completeness2 results at 4.5 @xmath0 m .
tables [ g360a1.tab ] and [ g360a2.tab ] report numerical data for the glimpse 360/deep glimpse archive simulations for the 3.6 and 4.5 @xmath0 m bands . these figures and tables show that the archive completeness1 is very similar to completeness2 .
this is expected , given that the lack of irac 5.8 and 8.0 @xmath0 m data in glimpse 360 and similar surveys essentially means that all of the sources must be detected both at 3.6 and 4.5 @xmath0 m in order to appear in the catalog or the archive . notably , glimpse 360 archive completeness in figures [ comp7a ] through [ comp8a ] is higher than catalog completeness in figures [ comp5a ] through [ comp6a ] , _ even at bright magnitudes and low background levels_. this results from the inclusion of sources in the archive that lie within the psf wings of saturated stars .
such sources are rejected from the catalog in order to ensure high reliability .
saturated stars are a much greater problem in the longer exposures used for glimpse 360/deep glimpse compared to glimpse i / ii/3d .
this means the photometry for a significant fraction of the areal coverage in glimpse 360/deep glimpse catalogs is nullified because it lies under the psf wings of saturated stars . the glimpse 360/deep glimpse archives , by contrast , are significantly more complete because they include sources that lie in the psf wings of saturated stars .
the magnitude of this effect in the glimpse 360/deep glimpse catalogs varies dramatically with galactic positions , being on the order of a few percent in the outer galaxy but rises to many tens of percent in crowded regions of the inner galaxy .
figure [ saturate180 ] and figure [ saturate030 ] show greyscale representations of the glimpse 360/deep glimpse irac 3.6 @xmath0 m mosaic near @xmath25 and @xmath26 , respectively .
red circles in the left panel of each figure mark sources included in the gpsc , while green circles in the right panels mark sources included in the gpsa .
the effect of `` dead zones '' around bright stars is dramatic in the left panel of figure [ saturate030 ] where high stellar densities cause the removal of photometric results from the gpsc for a significant fraction of the survey area .
by contrast , the effects of saturated sources in the gpsc in figure [ saturate180 ] for the @xmath25 field are rather subtle . here and in other such fields having low stellar densities
, the completeness of the catalog and archive will be more similar than for high - density fields , as only a few percent of the survey area falls under the wings of saturated stars . in summary ,
catalog incompleteness in glimpse 360/deep glimpse varies greatly with location , and users are encouraged to consider the merits of the gpsa where high levels of completeness are desired .
having empirically assessed the completeness simply as a function of one parameter , i.e. , background level , we now consider the role that the structure of the diffuse background emission plays in limiting point source source detection .
figure [ pretty ] shows that not only is the galactic plane filled with regions of high surface brightness ( i.e. , regions where poisson noise from background photons is elevated ) , but the emission is also highly structured , having a variety of filamentary morphologies on scales all the way down to the irac psf ( and probably on smaller scales that would be significant in data having higher angular resolution ) .
this structure further complicates source identification and extraction . as a test of
whether _ photon noise _ or _ structure _ from diffuse backgrounds is the more significant cause of incompleteness , we conducted the same artificial star test described in section 2.1 using the `` residual '' images generated by the glimpse pipeline after point sources have been removed .
we smoothed these residual images in the @xmath27333 test region , first with a 101-pixel box and then a 27-pixel box to eliminate `` structure '' on scales smaller than about 25 times the point spread function and 7 times the psf , respectively .
these smoothing scales represent large and moderate smoothing , respectively . to avoid edge - of - frame effects
, we smoothed a mosaic image of the residuals and then reconstructed the original bcd images . to the smoothed residual bcd images we added poisson noise , at a level appropriate for each pixel , and readnoise , according to the irac documentation : 11.8 , 12.1 , 9.1 , and 7.1 electrons per pixel for the irac
bandpasses 3.6 , 4.5 , 5.8 , and 8.0 @xmath0 m , respectively .
artificial stars were added at exactly the same positions and magnitudes as in section 2.1 before running the glimpse pipeline to extract sources and construct a catalog and archive .
the input list of artificial stars was compared to the recovered list of stars to assess the completeness .
figure [ sm27 - 1 ] shows the resulting gpsc completeness2 versus background level for 8.0 @xmath0 m , in a way exactly analogous to figure [ comp4a ] , except that now the effect of structure has been eliminated using a 27-pixel smoothing box .
a comparison of figure [ sm27 - 1 ] with figure [ comp4a ] shows that the completeness is much greater for any given magnitude or sky brightness level once the diffuse background has been smoothed .
results for other bands are similar .
figure [ sm101 - 1 ] shows gpsc completeness2 versus background level for the 8.0 @xmath0 m band , in a way exactly analogous to figure [ comp4a ] , except that now the effect of structure have been eliminated using a 101-pixel smoothing box .
comparison with figures [ comp4a ] and [ sm27 - 1 ] reveals that a much greater fraction of sources at all magnitudes and background levels are recovered and recorded in the catalog when the diffuse background structure is reduced even further .
we conclude that , at the magnitude ranges of interest to users of glimpse and similar @xmath1 irac surveys , incompleteness resulting from diffuse backgrounds is dominated by the effects of `` structure '' in the background rather than purely by photon noise .
the artificial star analysis performed in section 2.1 implicitly includes both effects . in the absence of a simple , single parameter suitable for describing the complexity of structured background emission in the galactic plane
, we retain the tables and figures , as previously presented , as a way to assess incompleteness as a function of mean background level .
point source crowding ( i.e. , confusion ) inhibits source detection and affects the reliability of photometry independently from the diffuse background effects addressed above .
point source densities in the glimpse archives range from @xmath2010 sources arcmin@xmath28 at 5.8 and 8.0 @xmath0 m in some local regions of the glimpse i / ii/3d surveys to over 150 sources arcmin@xmath28 at 3.6 and 4.5 @xmath0 m in some portions the glimpse 360/deep glimpse surveys . in the first extreme ,
source densities are so low that their psfs rarely overlap and confusion effects are negligible . in the latter extreme
, confusion is of considerable consequence .
ability to detect any given source will depend on the proximity to a neighboring source ( or sources ) and on the flux ratio between the source in question and the neighboring source(s ) .
as such , this issue involves one more free parameter than the diffuse background problem addressed above .
accordingly , we have elected to perform a more general and more empirical characterization of its effects using the glimpse data itself rather than an artificial star analysis .
figure [ lognlogs ] plots the base 10 logarithm of the point source number counts ( n ) in 0.05 mag wide bins versus magnitude ( i.e. , flux ) from the glimpse archives .
the @xmath29(n ) mag plot in each panel shows a different irac band either for the glimpse i / ii/3d 1.2 s survey ( all four bands ) or the glimpse 360/deep glimpse 10.4 s survey ( two bands ) .
line styles denote glimpse archive data from three distinct galactic longitude strips , as labeled .
only the sources having local point source densities in the lowest quartile of all sources in each longitude segment were used in order to mitigate possible effects of confusion and ensure that results are comparable to the simulations detailed in section 3 where artificial stars were added at locations devoid of known sources .
the curves exhibit a quasi - linear rise toward fainter magnitudes and then a well - defined turnover .
the bin having the maximum counts is designated as the `` turnover '' or `` peak '' magnitude bin and marks where the archive completeness becomes appreciable , as quantified below .
the turnover magnitude varies systematically with galactic longitude in the sense that regions farther from the galactic center are complete to fainter magnitudes as the source density drops .
the lower left panel shows that the turnover magnitude toward the galactic anticenter is [ 3.6]=16.6 , while at @xmath27328330 it is [ 3.6]@xmath3014.9 .
it is also clear that the 10.4 s glimpse 360/deep glimpse data go deeper than the glimpse i / ii/3d surveys .
the shapes of the curves , specifically the slopes as a function of magnitude , exhibit subtle but real differences reflecting variations in stellar populations along different sightlines . a detailed analysis of the @xmath29(n ) mag plots can be used to infer features such as the galactic `` long bar '' interior to @xmath31 and spiral arm tangencies ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , and additional analyses will be presented elsewhere .
nevertheless , figure [ lognlogs ] serves to illustrate the point that the completeness of the glimpse surveys as a function of stellar surface density may be estimated from data itself using a series of similar plots where the @xmath29(n ) mag curves are generated using subsets of archive data selected by local source density instead of longitude .
figure [ lognlogstest1 ] shows a grid of @xmath29(n ) mag curves from the glimpse i survey at @xmath27332333 .
the four columns correspond to the four irac bandpasses .
the five rows correspond sources having the specified range of diffuse background intensity within the @xmath27332333 segment , increasing from very low values ( top row ) to moderate
high values for each band ( bottom row ) , as labeled in each panel in mjy sr@xmath18 .
the black solid curve shows source counts in 0.1 mag bins for areas having local source densities less than the indicated value , in units of source arcmin@xmath28 .
these correspond to relatively low surface densities where confusion should be minimized .
blue dotted curves , available in some panels , plot the source counts in regions of high point source density .
the vertical black dashed and blue dotted lines mark the peak bin for the low - density and high - density curves , respectively . there is a clear progression of the peaks from right to left ( from fainter to brighter magnitudes ) as the mean sky level increases from top rows to the bottom rows .
the blue dotted line always lies to the left ( brighter magnitudes ) of the solid line , showing that incompleteness affects brighter sources in regions of higher source density .
vertical red lines mark the magnitude range of approximate 90% completeness inferred from the simulations reported in section 3 .
figure [ lognlogstest1 ] illustrates the excellent agreement between the vertical red lines and the vertical black lines , demonstrating that the peaks of the @xmath29(n ) mag curves are reliable indicators of the 90% completeness level . as a secondary method of inferring completeness we fit lines to the linear part of the @xmath29(n ) mag curve , extrapolated the line to the location of the peak , and then measured the ratio of the extrapolated n to the actual peak n in each plot as an estimate of the completeness .
this method generally supports the claim that the completeness at the peak of @xmath29(n ) mag curve is 7095% , but the inherently non - linear nature of the curves , especially in bright magnitude bins having few sources , precludes more precise measurements .
figure [ lognlogstest2 ] is similar to figure [ lognlogstest1 ] , but plots the @xmath29(n ) mag curves for 3.6 @xmath0 m and 4.5 @xmath0 m from the deep glimpse survey coverage of the @xmath27333region .
point source densities are much higher in these deeper 10.4 s exposures , so the dashed blue curves indicate that the peak is shifted to brighter sources , by about 0.51.0 mag relative to the deep glimpse simulations ( red vertical lines ) which are appropriate to uncrowded regions .
these results indicate that the deep glimpse survey of the inner galaxy is limited by confusion at most locations , as expected . as in figure [ lognlogstest1 ]
, the survey completeness begins to drop at brighter magnitudes as sky levels increase from top to bottom .
the agreement between the blue dotted vertical line and the red vertical lines improves in the lower panels .
this indicates that diffuse background effects become increasingly important relative to confusion as the sky brightness exceeds 12 mjy sr@xmath18 at both 3.6 @xmath0 m and 4.5 @xmath0 m .
figure [ lognlogstest3 ] shows @xmath29(n ) mag curves for 3.6 @xmath0 m and 4.5 @xmath0 m from the glimpse 360 survey coverage of the @xmath27133 region .
this outer - galaxy sightline exhibits lower source densities and suffers less from the effects of confusion , despite the same 10.4 s exposure time as in figure [ lognlogstest2 ] . as in figure
[ lognlogstest1 ] , the peak magnitude shifts toward brighter sources as sky levels increase from top to bottom .
there is good agreement between the completeness levels indicated by the simulations for glimpse 360 90% completeness magnitudes ( red vertical lines ) and the peak of the histogram ( vertical dashed line ) , except where sky brightness becomes large ( lower rows ) .
this supports the use of the peak in the @xmath29(n ) mag as an indicator of the 90% completeness level with regard to effects of confusion .
agreement between the simulated completeness levels and the peak of the @xmath29(n ) mag histograms in the preceding figures allows us to empirically infer the approximate magnitude where the glimpse archives are 90% complete with respect to point source confusion .
figure [ turnover ] plots the @xmath29(n ) mag turnover magnitude versus local point source density for each of the four irac bands and each of the two major survey modes .
symbols distinguish data from each of the five glimpse legacy surveys . at each band , only regions having sky brightness less than the indicated values ( corresponding to the lowest quartile of sky values in each band ) are used in order to minimize the impact of diffuse background on point source detection .
each data point in figure [ turnover ] represents the peak of the @xmath29(n ) mag plot for a specified range of source densities from 1020 arcmin@xmath28 to 190200 arcmin@xmath28 within a particular segment of the glimpse surveys . in some cases , the full range of source densities
is not realized in a given band for a given survey .
for example , source densities never exceed 50 arcmin@xmath28 at 8.0 @xmath0 m ( lower panel ) .
the dispersion of measurements at a given source density bin reflects both the uncertainties in measuring the turnover magnitude and real variations that stem from the uniqueness of stellar population and extinction distributions along each sightline .
the solid black curve connects the weighted mean turnover magnitudes in each source density bin .
figure [ turnover ] illustrates that the limiting magnitude is a decreasing function of source density , most convincingly at 3.6 and 4.5 @xmath0 m where stellar photospheres are brightest and source densities are highest . on average ,
the glimpse 360/deep glimpse surveys go almost two magnitudes deeper than the glimpse i / ii/3d surveys , but the slopes are similar .
the dispersion within a given bin is notably large in the deep glimpse data at 3.6 @xmath0 m . in the lower two panels
source densities almost never exceed 50 arcmin@xmath28 , and the trends are less clear .
the limiting magnitudes are nearly constant at [ 3.6]@xmath3012.4 and [ 4.5]@xmath3012.1 until the two highest density bins where the statistics become poor . the dispersion at a given source density is especially large for the 8.0 @xmath0 m panel , reflecting effects from variable levels of diffuse sky emission that can not be completely removed by limiting the analysis to the lowest quartile of sky backgrounds .
figure [ turnover ] provides the best general estimate yet available for estimating point source completeness levels in glimpse data products , as affected by confusion .
the results should be broadly applicable to regions covered by the various glimpse surveys .
users of glimpse data products requiring high levels of completeness precision should conduct their own analysis in specific regions of interest by constructing @xmath29(n ) mag curves and making use of the srcdens attribute provided as part of the glimpse catalogs and archives .
consumers of glimpse data products are advised that either diffuse background emission or point source confusion ( or both ! ) limits the completeness of the glimpse data products in various regimes .
figures [ gal_pnt ] and [ gal_bkg ] plot point source densities and diffuse background levels , respectively , as a function of longitude for each of the irac bands .
thin lines show the glimpse i / ii survey data and bold lines show the glimpse 360/deep glimpse survey data .
red / black / blue colors denote the 99.9%/median/0.1% points of the distribution , respectively , at a given longitude .
discrepancies between surveys at the same longitude are the result of different exposure times and different latitude ranges for the glimpsei / ii surveys relative to the deep glimpse survey .
figures [ gal_pnt ] and [ gal_bkg ] show that point source densities and sky levels generally increase toward the galactic center .
median point source densities above 50 sources arcmin@xmath28 at 3.6 and 4.5 @xmath0 m are common in the 1.2 s glimpse i / ii/3d surveys toward the inner galaxy , but at 5.8 and 8.0 @xmath0 m densities rarely exceed 30 sources arcmin@xmath28 .
median sky brightness levels vary between 0.1 and 2 mjy sr@xmath18 toward the inner galaxy at 3.6 and 4.5 @xmath0 m , but at 5.8 and 8.0 @xmath0 m levels between 10 and 100 mjy sr@xmath18 are common . the outer galaxy , by contrast , shows vastly reduced point source densities and diffuse sky levels .
source densities and diffuse levels also increase toward the mid - plane , but these variations are not reflected in the latitude - averaged plots .
spatial variations in these quantities can be large over very small angular scales .
for example , infrared dark clouds exhibit some of the lowest sky levels of @xmath201 mjy sr@xmath18 , and these often lie adjacent to bright - rimmed clouds and photodissociation regions with sky levels exceeding 100 mjy sr@xmath18 .
this means that a detailed description of how confusion or diffuse background affects photometry as a function of sky position would require a more complex analysis than undertaken here , but our simulations provide some insights specific to individual bandpasses . _
3.6 @xmath0 m _ figure [ turnover ] shows that , for regions with very low sky background @xmath202.6 mjy sr@xmath18 , the glimpse i / ii/3d archives are complete to between [ 3.6 ] = 14 14.9 mag , depending on local source density . by comparison ,
figure [ comp1aa ] ( lower panel ) reveals that completeness never exceeds 80% in this faintest magnitude bin , even at low sky levels , and it drops markedly in regions with sky brightness above 6 mjy sr@xmath18 .
diffuse background significantly impacts completeness in brighter magnitude bins when it exceeds 10 mjyr sr@xmath18 .
we conclude that for regions with sky levels @xmath173 mjy sr@xmath18 , the limiting magnitude is set by confusion but is otherwise dominated by diffuse emission . _
4.5 @xmath0 m _ figure [ turnover ] shows that , for regions with background levels @xmath201.6 mjy sr@xmath18 , the glimpse i / ii/3d archives are complete to [ 4.5 ] = 14 14.7 mag , for source densities from 85 to 15 arcmin@xmath28 .
figure [ comp3aa ] ( lower panel ) , by comparison , shows that the archives are , at most , 50% complete for the [ 4.5]=1415 mag bin even in the darkest regions . for sky levels exceeding about 6 mjy sr@xmath18 ,
incompleteness becomes appreciable in the [ 4.5]=1314 mag range .
we conclude that for the darkest regions and highest source densities completeness is limited by confusion but for sky levels @xmath326 mjy sr@xmath18 , diffuse background becomes the limiting factor . _
5.8 @xmath0 m _ figure [ turnover ] shows that the limiting magnitude is roughly constant at [ 5.8 ] = 12.3 across the range of observed source densities for sky brightness @xmath2018.6 mjy sr@xmath18 .
figure [ comp4aa ] ( lower panel ) indicates that the archives are already no better than 55% complete in this magnitude range , and that by 40 mjy sr@xmath18 , the [ 5.8]=1112 mag range has become substantially incomplete .
we conclude that above 40 mjy sr@xmath18 the incompleteness is dominated by diffuse emission at all magnitude ranges . in the darkest areas having @xmath2018.6 mjy sr@xmath18
the effects of confusion and diffuse background may be comparable for sources [ 5.8]= 1213 .
given the very low source densities and strong pah emission at 5.8 @xmath0 m we conclude that diffuse emission often limits the detection of point sources for typical galactic plane sightlines . _
8.0 @xmath0 m _ figure [ turnover ] shows that the limiting magnitude is approximately constant at [ 8.0]@xmath3012 across the small range of observed source densities for sky values @xmath2049.5 mjy sr@xmath18 .
figure [ comp2aa ] ( lower panel ) indicates that in the [ 8.0]=1112 range , the completeness has already dropped below 90% for sky values near 49.5 mjy sr@xmath18 . by 100 mjy sr@xmath18 stars in the 1011 mag range
become substantially incomplete .
given the very low source densities and strong pah emission at 8.0 @xmath0 m we conclude that diffuse emission is frequently the limiting agent regarding the detection of point sources . _
3.6 @xmath0 m _ figure [ turnover ] displays a range of limiting magnitudes from 16.6 at the lowest source densities ( 15 arcmin@xmath28 ) to 15.2 at densities approaching 155 arcmin@xmath28 for sky backgrounds @xmath202.6 mjy sr@xmath18 . by comparison ,
figure [ comp7a ] reveals that at these sky levels stars in the [ 3.6]=1516 bin are already less than 90% complete .
we conclude that the effects of confusion are dominant for sightlines with the highest stellar densities and lowest backgrounds , but that diffuse emission limits completeness for many sightlines and magnitude ranges .
diffuse sky background becomes a signficant / limiting factor above about 12 mjy sr@xmath18 at both 3.6 and 4.5 @xmath0 m . _
4.5 @xmath0 m _ figure [ turnover ] illustrates that limiting magnitudes vary from [ 4.5]=16.314.7 as source densities increase from 15 to 145 arcmin@xmath28 for sky backgrounds @xmath201.6 mjy sr@xmath18 . by comparison ,
figure [ comp8a ] shows that stars in the [ 4.5]=1516 mag bin are already less than 90% complete at these sky levels .
we conclude that sky background effects dominate the incompleteness unless sky levels are @xmath172 mjy sr@xmath18 and point source densities are @xmath3280 sources arcmin@xmath28 .
we have conducted an analysis of the completeness of point sources appearing in the various glimpse point source catalogs and archives to assess the impact of 1 ) high diffuse galactic backgrounds and 2 ) high point source densities . with regard to the effects of diffuse backgrounds , we find : 1 .
completeness is , as expected , a strong function of both stellar magnitude and background sky level , such that no simple characterization of the completeness is possible .
completeness drops rapidly with increasing stellar magnitude and sky brightness , especially in the irac 5.8 and 8.0 @xmath0 m bands where strong pah features produce large sky background brightnesses .
we provide figures and tabular data , suitable for interpolation , listing completeness1 and completeness2 values for each band for each of the two glimpse survey strategies for both the catalogs and archives .
3 . completeness , as defined in two slightly different ways , is generally higher in the glimpse point source archives than in the catalogs , at any given irac band .
completeness2 is always equal to or lower than completeness1 .
catalogs resulting from the glimpse 360/deep glimpse surveys may suffer significantly reduced completeness relative to the equivalent archives because sources in the psf wings of saturated stars are removed from the catalogs .
this effect is most severe in high - density regions toward the inner galaxy .
consumers of glimpse 360/deep glimpse data products should consider using the archive rather than the catalog if high completeness is desired .
glimpse i / ii/3d surveys , by virtue of their shorter exposure times , exhibit a relatively minor completeness differences between catalogs and archives from this effect .
we assess whether background `` brightness '' or background `` structure '' is the limiting factor for point source detection .
we conclude that both play a role , but that the effects of structure appear to dominate over pure photon noise in hindering the identification and extraction of point sources .
although our artificial star recovery analysis is performed on images drawn from the glimpse i / ii/3d ( multiple 1.2-second exposures at all four irac bands ) and glimpse 360 and deep glimpse ( three 10.4-second exposures at 3.6 and 4.5 @xmath0 m ) surveys , the results can be generalized to estimate completeness for other _
irac programs that used similar observing strategies and have data products ( catalogs and archives ) constructed in the same manner .
these include the smog , cygnus - x , and the vela - carina surveys . with regard to the effects of high stellar density on completeness , we conclude that only in the regions of low sky background do the effects of confusion limit the detection of point sources in the shallow glimpse i / ii/3d surveys .
the relative importance of diffuse background emission increases from 3.6 @xmath0 m to 8.0 @xmath0 m for a typical sightline and as the point source densities decrease , such as toward the outer galaxy . in glimpse
i / ii/3d , confusion is the dominant source of incompleteness only at 3.6 and 4.5 @xmath0 m and in regions of high point source density @xmath3280 sources arcmin@xmath28 , and where sky background levels are @xmath172 mjy sr@xmath18 .
irac 5.8 and 8.0 @xmath0 m are almost never limited by confusion as point source densities never exceed about 50 sources arcmin@xmath28 .
glimpse 360/deep glimpse surveys reach several magnitudes deeper and are limited by both confusion and diffuse emission , to varying degrees .
confusion is the dominant source of incompleteness at 3.6 and 4.5 @xmath0 m and regions of high point source density @xmath3280 sources arcmin@xmath28 , and where sky background levels are @xmath172 mjy sr@xmath18 .
figure [ turnover ] provides the best current estimates for confusion limits in each of the glimpse surveys and bandpasses as a function of point source surface density .
glimpse data products include attributes for quantifying local point source density and diffuse sky levels for each detected source .
finally , we caution that completeness in the glimpse data products ( and any higher - level data catalog products ) is inevitably also affected by additional selection and matching criteria that are imposed to ensure high reliability ( i.e. , robustness against false sources ) .
therefore , completeness is necessarily a multi - faceted , non - trivial quantity that is unique to a particular survey strategy and source extraction paradigm .
consumers of glimpse data requiring highly precise completeness estimates are urged to understand the details of the source detection and selection process for the data products in question .
we are grateful to matt povich , bob benjamin , and dan clemens for their assistance and advice on this analysis .
we thank the extended glimpse team and the irac instrument team members who , together , have helped deliver a genuinely `` legacy '' dataset to the astronomical community .
we thank the anonymous referee whose astute suggestions improved this manuscript .
this publication makes use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by the national aeronautics and space administration and the national science foundation .
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 under a contract with nasa , and support was provided by nasa through awards issued by jpl / caltech 1368699 and 1367334 .
source detection in the glimpse team s photometry pipeline utilizes a modified version of the daophot @xcite find routine ( see appendix 2 of the daophot ii user s manual ; http://www.astro.wisc.edu/glimpse/daophot2.pdf ) and is more fully described in the glimpse documentation on point source photometry ( http://www.astro.wisc.edu/sirtf/ glimpse_photometry_v1.0.pdf ) .
the glimpse team modified the daophot ii find routine in two ways to accommodate the high - reliability requirement of the glimpse catalogs .
the original daophot find routine determines a global `` sky '' value from the entire input image .
since the glimpse images contain highly variable diffuse structure , we first estimate the local sky level using a 5-pixel median - smoothed version of the input image .
the `` sky '' value is measured locally in a 5@xmath75 pixel box centered at each pixel location . in order to ensure that detected sources in high - background , high - structure regions are real
, we adopt the maximum sky value within the box as the effective sky value .
the maximum sky value , readout noise , and detector gain are then used to determine the local noise at each central pixel .
point source detection is then performed as a two - step process .
initial detections are conducted using a 3@xmath33 threshold .
detected sources are then fit with the instrument psf and extracted from the input image .
if the photometric s / n ratio of the extracted source falls below two , the source is discarded .
once the initial set of sources are extracted , the residual image ( input image minus the subtracted sources ) is then used as a new input image .
a new sky image is constructed from this image and secondary sources are then detected at a 5@xmath33 threshold .
the second iteration of find is performed at a higher detection level because the initial source subtraction adds noise around extracted sources , requiring a higher subsequent detection threshold to avoid finding false sources while still detecting real sources hidden in the wings of the brighter stars .
first - pass source extractions are then combined with the second - pass extractions to produce the final list which we designate as the `` full '' list of candidate sources for each bcd image .
candidate sources must still pass additional cross - band and multi - detection criteria before being included in the catalogs or archives , as described in section 2 and in glimpse data products documents referenced herein .
99 benjamin , r. , et al .
2003 , pasp 115 , 953 benjamin , r. a. 2009 , iau symposium , 254 , 319 benjamin , r. a. , churchwell , e. , babler , b. l. , et al .
2005 , , 630 , l149 benjamin , r. a. , & glimpse team 2007 , bulletin of the american astronomical society , 38 , 760 carey , s. , etal .
2008 , spitzer proposal i d # 50398 churchwell , e. , babler , b. l. , meade , m. r. , et al .
2009 , , 121 , 213 fazio , g. g. , hora , j. l. , allen , l. e. , et al .
2004 , , 154 , 10 hora , j. , et al .
2007 , spitzer proposal i d # 40184 robitaille , t. p. , et al .
2008 , , 136 , 2413 stetson , p. 1987 , pasp , 99 , 191 werner , m. w. , roellig , t. l. , low , f. j. , et al .
2004 , , 154 , 1 whitney , b. , arendt , r. , babler , b. , et al .
2008 , spitzer proposal , 60020 whitney , b. , benjamin , r. , churchwell , e. , et al .
2011 , spitzer proposal , 80074 zasowski , g. , majewski , s. r. , indebetouw , r. , et al .
2009 , , 707 , 510 lclcccrr glimpse i & 10@xmath3465 ; @xmath35 & 220 sq .
deg . & irac [ 3.6],[4.5],[5.6],[8.0 ] & @xmath36 s & churchwell et al .
( 2009 ) & 31,154,438 & 49,133,194 + glimpse ii & @xmath37 ; @xmath38 @xmath171.5 & 54 sq .
deg . & irac [ 3.6],[4.5],[5.6],[8.0 ] & @xmath39 s & churchwell et al . ( 2009 ) & 18,145,818 & 23,125,046 + glimpse 3d & @xmath40 31;@xmath41 & 120 sq .
deg . & irac [ 3.6],[4.5],[5.6],[8.0 ] & @xmath42 s & churchwell et al . (
2009 ) & 20,403,915 & 32,214,210 + glimpse 360 & @xmath43=65 76,82 102,109265 & 511 sq .
deg . & irac [ 3.6],[4.5 ] & @xmath44 s & whitney et al .
( 2008 ) & 42,602,112 & 49,378,042 + & @xmath45 & & & & & & + deep glimpse & @xmath43 = 265350 , @xmath46=-2 + 0.1 & 208 sq .
deg . & irac [ 3.6],[4.5 ] & @xmath44 s & whitney et al .
( 2011 ) & 38,279,639 & 63,522,165 + & @xmath43 = 2565 , @xmath46= 0 + 2.7 & & & & & & + smog & @xmath43=102 109 ; b= 0 3 & 21 sq . deg . & irac [ 3.6],[4.5],[5.6],[8.0 ] & @xmath47 s & carey et al .
( 2008 ) & 2,512,099 & 2,836,618 +
cygnus - x & @xmath43= 76 82 ; b = -2.3 + 4.1 & 24 sq .
& irac [ 3.6],[4.5],[5.6],[8.0 ] & @xmath44 s & hora et al .
( 2007 ) & 3,913,559 & 4,455,066 + vela - carina & @xmath43=255 295 ; b @xmath48 -1.5 + 1.5 & 80 sq .
deg . & irac [ 3.6],[4.5],[5.6],[8.0 ] & @xmath36 s & zasowski et al .
( 2009 ) & 2,000,188 & 4,547,327 + ptabularcr ptabularcr ptabularcr 1.2 1.8 & 0.98 & 1.00 & 0.99 & 1.00 & 0.98 & 0.89 & 0.26 + 1.8 2.6 & 0.99 & 1.00 & 1.00 & 0.99 & 1.00 & 0.96 & 0.23 + 2.6 3.8 & 0.99 & 0.99 & 0.99 & 0.99 & 0.98 & 0.95 & 0.24 + 3.8 5.6 & 0.99 & 0.99 & 0.99 & 0.98 & 0.96 & 0.91 & 0.22 + 5.6 8.2 & 0.99 & 0.99 & 0.98 & 0.95 & 0.92 & 0.81 & 0.18 + 8.2 12.0 & 0.97 & 0.97 & 0.96 & 0.91 & 0.81 & 0.57 & 0.10 + 12.0 17.6 & 0.96 & 0.96 & 0.92 & 0.84 & 0.64 & 0.30 & 0.03 + 17.6 25.9 & 0.95 & 0.91 & 0.84 & 0.68 & 0.43 & 0.20 & 0.02 + 25.9 37.9 & 0.87 & 0.77 & 0.70 & 0.44 & 0.29 & 0.06 & 0.00 + 37.9 55.7 & 0.68 & 0.58 & 0.29 & 0.13 & 0.02 & 0.00 & 0.02 + 55.7 81.8 & 0.64 & 0.41 & 0.20 & 0.12 & 0.00 & 0.00 & 0.00 + 81.8 120.0 & 0.31 & 0.07 & 0.00 & 0.06 & 0.00 & 0.00 & 0.00 + ptabularcr ptabularcr ptabularcr 0.8 1.1 & 0.98 & 0.99 & 1.00 & 0.99 & 1.00 & 0.93 & 0.24 + 1.1 1.6 & 0.99 & 1.00 & 1.00 & 0.99 & 0.99 & 0.96 & 0.23 + 1.6 2.4 & 0.99 & 1.00 & 0.99 & 0.99 & 0.98 & 0.95 & 0.25 + 2.4 3.4 & 0.99 & 0.99 & 1.00 & 0.99 & 0.98 & 0.93 & 0.23 + 3.4 4.8 & 0.99 & 0.99 & 0.99 & 0.98 & 0.95 & 0.90 & 0.20 + 4.8 6.9 & 0.99 & 0.99 & 0.98 & 0.94 & 0.92 & 0.80 & 0.18 + 6.9 9.9 & 0.98 & 0.97 & 0.97 & 0.92 & 0.86 & 0.63 & 0.10 + 9.9 14.2 & 0.96 & 0.96 & 0.92 & 0.87 & 0.68 & 0.34 & 0.04 + 14.2 20.4 & 0.94 & 0.92 & 0.88 & 0.74 & 0.54 & 0.26 & 0.01 + 20.4 29.2 & 0.97 & 0.87 & 0.78 & 0.64 & 0.36 & 0.12 & 0.01 + 29.2 41.9 & 0.76 & 0.68 & 0.61 & 0.32 & 0.17 & 0.02 & 0.00 + 41.9 60.0 & 0.67 & 0.59 & 0.31 & 0.20 & 0.02 & 0.00 & 0.02 + ptabularcr ptabularcr ptabularcr 9.0 12.9 & 0.99 & 1.00 & 1.00 & 0.99 & 0.99 & 0.95 & 0.25 + 12.9 18.6 & 0.99 & 0.99 & 0.99 & 0.99 & 0.97 & 0.93 & 0.25 + 18.6 26.7 & 0.99 & 0.99 & 0.99 & 0.98 & 0.96 & 0.91 & 0.22 + 26.7 38.4 & 0.99 & 0.99 & 0.99 & 0.96 & 0.94 & 0.88 & 0.20 + 38.4 55.2 & 0.98 & 0.98 & 0.97 & 0.94 & 0.93 & 0.83 & 0.18 + 55.2 79.4 & 0.98 & 0.96 & 0.95 & 0.93 & 0.89 & 0.72 & 0.14 + 79.4 114.1 & 0.97 & 0.96 & 0.97 & 0.94 & 0.83 & 0.47 & 0.06 + 114.1 164.0 & 0.96 & 0.97 & 0.94 & 0.86 & 0.66 & 0.29 & 0.03 + 164.0 235.7 & 0.94 & 0.93 & 0.93 & 0.76 & 0.49 & 0.23 & 0.02 + 235.7 338.8 & 0.93 & 0.88 & 0.68 & 0.47 & 0.27 & 0.02 & 0.00 + 338.8 487.0 & 0.75 & 0.58 & 0.33 & 0.20 & 0.02 & 0.00 & 0.02 + 487.0
700.0 & 0.77 & 0.66 & 0.16 & 0.09 & 0.00 & 0.00 & 0.00 + ptabularcr ptabularcr ptabularcr 25.0 35.2 & 0.99 & 1.00 & 1.00 & 0.98 & 0.98 & 0.92 & 0.26 + 35.2 49.5 & 0.99 & 0.99 & 0.99 & 0.99 & 0.95 & 0.92 & 0.24 + 49.5 69.6 & 0.98 & 0.99 & 0.99 & 0.98 & 0.96 & 0.91 & 0.23 + 69.6 97.9 & 0.98 & 0.99 & 0.99 & 0.97 & 0.94 & 0.89 & 0.21 + 97.9 137.7 & 0.99 & 0.99 & 0.98 & 0.96 & 0.95 & 0.87 & 0.19 + 137.7 193.6 & 0.99 & 0.98 & 0.97 & 0.95 & 0.93 & 0.84 & 0.18 + 193.6 272.4 & 0.99 & 0.98 & 0.97 & 0.96 & 0.91 & 0.71 & 0.12 + 272.4 383.2 & 0.98 & 0.98 & 0.97 & 0.94 & 0.82 & 0.45 & 0.06 + 383.2 539.0 & 0.98 & 0.98 & 0.95 & 0.86 & 0.65 & 0.29 & 0.02 + 539.0 758.1 & 0.95 & 0.97 & 0.93 & 0.71 & 0.46 & 0.19 & 0.03 + 758.11066.4 & 0.90 & 0.82 & 0.64 & 0.45 & 0.25 & 0.01 & 0.00 + 1066.41500.0 & 0.86 & 0.56 & 0.25 & 0.16 & 0.00 & 0.00 & 0.02 + ptabularcr ptabularcr ptabularcr 1.2 1.8 & 0.98 & 1.00 & 0.99 & 1.00 & 0.98 & 0.89 & 0.26 + 1.8 2.6 & 0.99 & 1.00 & 0.99 & 0.99 & 1.00 & 0.96 & 0.23 + 2.6 3.8 & 0.98 & 0.99 & 0.99 & 0.99 & 0.98 & 0.95 & 0.24 + 3.8 5.6 & 0.97 & 0.98 & 0.98 & 0.97 & 0.96 & 0.91 & 0.22 + 5.6 8.2 & 0.96 & 0.96 & 0.96 & 0.94 & 0.92 & 0.81 & 0.18 + 8.2 12.0 & 0.93 & 0.93 & 0.92 & 0.90 & 0.81 & 0.57 & 0.10 + 12.0 17.6 & 0.92 & 0.90 & 0.89 & 0.83 & 0.64 & 0.30 & 0.03 + 17.6 25.9 & 0.90 & 0.88 & 0.81 & 0.66 & 0.43 & 0.20 & 0.02 + 25.9 37.9 & 0.80 & 0.73 & 0.65 & 0.42 & 0.29 & 0.06 & 0.00 + 37.9 55.7 & 0.58 & 0.53 & 0.29 & 0.11 & 0.02 & 0.00 & 0.02 + 55.7 81.8 & 0.64 & 0.41 & 0.20 & 0.12 & 0.00 & 0.00 & 0.00 + 81.8 120.0 & 0.25 & 0.07 & 0.00 & 0.06 & 0.00 & 0.00 & 0.00 + ptabularcr ptabularcr ptabularcr 0.8 1.1 & 0.96 & 0.99 & 0.98 & 0.99 & 1.00 & 0.93 & 0.24 + 1.1 1.6 & 0.99 & 0.99 & 0.99 & 0.99 & 0.99 & 0.96 & 0.23 + 1.6 2.4 & 0.98 & 0.98 & 0.98 & 0.98 & 0.98 & 0.95 & 0.25 + 2.4 3.4 & 0.97 & 0.98 & 0.98 & 0.98 & 0.98 & 0.93 & 0.23 + 3.4 4.8 & 0.95 & 0.96 & 0.95 & 0.96 & 0.95 & 0.90 & 0.20 + 4.8 6.9 & 0.93 & 0.92 & 0.93 & 0.93 & 0.92 & 0.80 & 0.18 + 6.9 9.9 & 0.90 & 0.90 & 0.93 & 0.91 & 0.86 & 0.63 & 0.10 + 9.9 14.2 & 0.87 & 0.88 & 0.87 & 0.86 & 0.68 & 0.34 & 0.04 + 14.2 20.4 & 0.87 & 0.88 & 0.84 & 0.73 & 0.54 & 0.26 & 0.01 + 20.4 29.2 & 0.90 & 0.83 & 0.77 & 0.63 & 0.36 & 0.12 & 0.01 + 29.2 41.9 & 0.70 & 0.65 & 0.61 & 0.32 & 0.17 & 0.02 & 0.00 + 41.9 60.0 & 0.64 & 0.59 & 0.29 & 0.20 & 0.02 & 0.00 & 0.02 + ptabularcr ptabularcr ptabularcr 9.0 12.9 & 0.99 & 1.00 & 1.00 & 0.99 & 0.64 & 0.05 & 0.00 + 12.9 18.6 & 0.99 & 0.99 & 0.99 & 0.98 & 0.53 & 0.03 & 0.00 + 18.6 26.7 & 0.99 & 0.99 & 0.99 & 0.97 & 0.42 & 0.02 & 0.00 + 26.7 38.4 & 0.99 & 0.99 & 0.98 & 0.92 & 0.29 & 0.01 & 0.00 + 38.4 55.2 & 0.98 & 0.98 & 0.96 & 0.81 & 0.18 & 0.01 & 0.00 + 55.2 79.4 & 0.98 & 0.96 & 0.89 & 0.56 & 0.06 & 0.01 & 0.00 + 79.4 114.1 & 0.97 & 0.91 & 0.77 & 0.27 & 0.02 & 0.01 & 0.00 + 114.1 164.0 & 0.92 & 0.84 & 0.49 & 0.11 & 0.01 & 0.00 & 0.00 + 164.0 235.7 & 0.83 & 0.65 & 0.35 & 0.05 & 0.01 & 0.00 & 0.00 + 235.7 338.8 & 0.64 & 0.29 & 0.05 & 0.03 & 0.01 & 0.00 & 0.00 + 338.8 487.0 & 0.22 & 0.10 & 0.05 & 0.00 & 0.00 & 0.00 &
0.00 + 487.0 700.0 & 0.13 & 0.09 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 + ptabularcr ptabularcr ptabularcr 25.0 35.2 & 0.99 & 1.00 & 0.99 & 0.92 & 0.24 & 0.00 & 0.00 + 35.2 49.5 & 0.98 & 0.99 & 0.99 & 0.87 & 0.15 & 0.00 & 0.00 + 49.5 69.6 & 0.98 & 0.99 & 0.98 & 0.69 & 0.05 & 0.00 & 0.00 + 69.6 97.9 & 0.98 & 0.98 & 0.95 & 0.50 & 0.02 & 0.00 & 0.00 + 97.9 137.7 & 0.98 & 0.97 & 0.83 & 0.24 & 0.01 & 0.00 & 0.00 + 137.7 193.6 & 0.97 & 0.90 & 0.57 & 0.07 & 0.00 & 0.00 & 0.00 + 193.6 272.4 & 0.89 & 0.66 & 0.21 & 0.01 & 0.00 & 0.00 & 0.00 + 272.4 383.2 & 0.67 & 0.30 & 0.04 & 0.00 & 0.00 & 0.00 & 0.00 + 383.2 539.0 & 0.38 & 0.16 & 0.01 & 0.00 & 0.00 & 0.00 & 0.00 + 539.0 758.1 & 0.20 & 0.03 & 0.01 & 0.01 & 0.00 & 0.00 & 0.00 + 758.11066.4 & 0.04 & 0.02 & 0.00 & 0.00 & 0.00 & 0.00 &
0.00 + 1066.41500.0 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 + ptabularcr ptabularcr ptabularcr 1.2 1.8 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.96 & 0.68 + 1.8 2.6 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.72 + 2.6 3.8 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.68 + 3.8 5.6 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.98 & 0.59 + 5.6 8.2 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.92 & 0.44 + 8.2 12.0 & 1.00 & 1.00 & 1.00 & 0.99 & 0.94 & 0.72 & 0.21 + 12.0 17.6 & 1.00 & 1.00 & 0.99 & 0.96 & 0.81 & 0.44 & 0.07 + 17.6 25.9 & 0.99 & 0.98 & 0.94 & 0.78 & 0.56 & 0.30 & 0.04 + 25.9 37.9 & 0.99 & 0.93 & 0.82 & 0.56 & 0.41 & 0.10 & 0.00 + 37.9 55.7 & 0.87 & 0.84 & 0.46 & 0.24 & 0.13 & 0.01 & 0.02 + 55.7 81.8 & 0.81 & 0.59 & 0.33 & 0.14 & 0.06 & 0.00 & 0.00 + 81.8 120.0 & 0.44 & 0.33 & 0.07 & 0.06 & 0.00 & 0.00 & 0.00 + ptabularcr ptabularcr ptabularcr 0.8 1.1 & 0.99 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.75 + 1.1 1.6 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.73 + 1.6 2.4 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.69 + 2.4 3.4 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.98 & 0.64 + 3.4 4.8 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.97 & 0.55 + 4.8 6.9 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.91 & 0.44 + 6.9 9.9 & 1.00 & 1.00 & 1.00 & 0.99 & 0.95 & 0.78 & 0.24 + 9.9 14.2 & 1.00 & 1.00 & 0.99 & 0.96 & 0.84 & 0.48 & 0.09 + 14.2 20.4 & 0.99 & 0.98 & 0.98 & 0.85 & 0.68 & 0.39 & 0.04 + 20.4 29.2 & 0.99 & 0.98 & 0.88 & 0.71 & 0.51 & 0.18 & 0.02 + 29.2 41.9 & 0.95 & 0.86 & 0.72 & 0.49 & 0.26 & 0.06 & 0.00 + 41.9 60.0 & 0.88 & 0.76 & 0.53 & 0.27 & 0.13 & 0.00 & 0.02 + ptabularcr ptabularcr ptabularcr 9.0 12.9 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.98 & 0.72 + 12.9 18.6 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.98 & 0.67 + 18.6 26.7 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.97 & 0.61 + 26.7 38.4 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.95 & 0.56 + 38.4 55.2 & 1.00 & 1.00 & 1.00 & 0.99 & 0.98 & 0.92 & 0.48 + 55.2 79.4 & 1.00 & 0.99 & 0.99 & 0.98 & 0.95 & 0.83 & 0.31 + 79.4 114.1 & 1.00 & 0.99 & 0.99 & 0.97 & 0.92 & 0.61 & 0.13 + 114.1 164.0 & 0.99 & 0.99 & 0.98 & 0.92 & 0.80 & 0.45 & 0.07 + 164.0 235.7 & 0.99 & 0.95 & 0.94 & 0.82 & 0.62 & 0.31 & 0.03 + 235.7 338.8 & 0.97 & 0.94 & 0.79 & 0.60 & 0.38 & 0.05 & 0.00 + 338.8 487.0 & 0.86 & 0.71 & 0.51 & 0.33 & 0.12 & 0.03 & 0.02 + 487.0
700.0 & 0.87 & 0.83 & 0.30 & 0.12 & 0.09 & 0.00 & 0.00 + ptabularcr ptabularcr ptabularcr 25.0 35.2 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.98 & 0.72 + 35.2 49.5 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.97 & 0.66 + 49.5 69.6 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.97 & 0.64 + 69.6 97.9 & 1.00 & 1.00 & 1.00 & 0.99 & 0.99 & 0.96 & 0.59 + 97.9 137.7 & 1.00 & 1.00 & 1.00 & 0.99 & 0.98 & 0.95 & 0.54 + 137.7 193.6 & 1.00 & 1.00 & 1.00 & 0.99 & 0.98 & 0.92 & 0.45 + 193.6 272.4 & 1.00 & 0.99 & 1.00 & 0.99 & 0.97 & 0.82 & 0.28 + 272.4 383.2 & 1.00 & 1.00 & 0.99 & 0.98 & 0.93 & 0.61 & 0.12 + 383.2 539.0 & 1.00 & 0.99 & 0.97 & 0.93 & 0.78 & 0.43 & 0.05 + 539.0 758.1 & 1.00 & 0.98 & 0.96 & 0.79 & 0.58 & 0.27 & 0.05 + 758.11066.4 & 0.98 & 0.92 & 0.81 & 0.60 & 0.36 & 0.03 & 0.00 + 1066.41500.0 & 0.92 & 0.71 & 0.44 & 0.24 & 0.17 & 0.02 & 0.02 + ptabularcr ptabularcr ptabularcr 1.2 1.8 & 0.98 & 1.00 & 1.00 & 1.00 & 1.00 & 0.96 & 0.68 + 1.8 2.6 & 0.99 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.72 + 2.6 3.8 & 0.99 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.67 + 3.8 5.6 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.98 & 0.59 + 5.6 8.2 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.92 & 0.44 + 8.2 12.0 & 1.00 & 1.00 & 1.00 & 0.98 & 0.94 & 0.71 & 0.21 + 12.0 17.6 & 0.99 & 0.99 & 0.97 & 0.95 & 0.80 & 0.44 & 0.07 + 17.6 25.9 & 0.98 & 0.97 & 0.91 & 0.75 & 0.55 & 0.30 & 0.04 + 25.9 37.9 & 0.90 & 0.88 & 0.76 & 0.53 & 0.40 & 0.10 & 0.00 + 37.9 55.7 & 0.70 & 0.77 & 0.41 & 0.20 & 0.11 & 0.01 & 0.02 + 55.7 81.8 & 0.79 & 0.51 & 0.33 & 0.14 & 0.06 & 0.00 & 0.00 + 81.8 120.0 & 0.38 & 0.20 & 0.07 & 0.06 & 0.00 & 0.00 & 0.00 + ptabularcr ptabularcr ptabularcr
0.8 1.1 & 0.98 & 0.99 & 1.00 & 1.00 & 1.00 & 0.98 & 0.49 + 1.1 1.6 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.97 & 0.51 + 1.6 2.4 & 0.99 & 1.00 & 1.00 & 1.00 & 1.00 & 0.98 & 0.50 + 2.4 3.4 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.96 & 0.44 + 3.4 4.8 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.94 & 0.36 + 4.8 6.9 & 1.00 & 1.00 & 1.00 & 0.99 & 0.98 & 0.86 & 0.28 + 6.9 9.9 & 1.00 & 0.99 & 0.99 & 0.98 & 0.92 & 0.67 & 0.14 + 9.9 14.2 & 0.98 & 0.98 & 0.95 & 0.92 & 0.76 & 0.37 & 0.05 + 14.2 20.4 & 0.98 & 0.96 & 0.93 & 0.80 & 0.59 & 0.28 & 0.01 + 20.4 29.2 & 0.94 & 0.96 & 0.82 & 0.66 & 0.38 & 0.14 & 0.01 + 29.2 41.9 & 0.85 & 0.79 & 0.66 & 0.38 & 0.17 & 0.02 & 0.00 + 41.9 60.0 & 0.81 & 0.67 & 0.36 & 0.21 & 0.04 & 0.00 & 0.02 + ptabularcr ptabularcr ptabularcr 9.0 12.9 & 1.00 & 1.00 & 1.00 & 1.00 & 0.64 & 0.05 & 0.00 + 12.9 18.6 & 0.99 & 1.00 & 1.00 & 0.99 & 0.55 & 0.03 & 0.00 + 18.6 26.7 & 1.00 & 1.00 & 1.00 & 0.98 & 0.43 & 0.02 & 0.00 + 26.7 38.4 & 0.99 & 1.00 & 0.99 & 0.94 & 0.31 & 0.01 & 0.00 + 38.4 55.2 & 1.00 & 1.00 & 0.98 & 0.84 & 0.18 & 0.01 & 0.00 + 55.2 79.4 & 0.99 & 0.98 & 0.93 & 0.59 & 0.06 & 0.01 & 0.00 + 79.4 114.1 & 0.98 & 0.93 & 0.79 & 0.27 & 0.02 & 0.01 & 0.00 + 114.1 164.0 & 0.94 & 0.85 & 0.49 & 0.11 & 0.01 & 0.01 & 0.00 + 164.0 235.7 & 0.88 & 0.66 & 0.35 & 0.05 & 0.01 & 0.01 & 0.00 + 235.7 338.8 & 0.67 & 0.29 & 0.06 & 0.04 & 0.01 & 0.00 & 0.00 + 338.8 487.0 & 0.25 & 0.12 & 0.07 & 0.00 & 0.00 & 0.00 & 0.00 + 487.0 700.0 & 0.16 & 0.09 & 0.00 & 0.00 & 0.00 & 0.00 &
0.00 + ptabularcr ptabularcr ptabularcr 25.0 35.2 & 1.00 & 1.00 & 1.00 & 0.94 & 0.24 & 0.00 & 0.00 + 35.2 49.5 & 0.99 & 1.00 & 1.00 & 0.87 & 0.15 & 0.01 & 0.00 + 49.5 69.6 & 1.00 & 0.99 & 0.99 & 0.69 & 0.05 & 0.00 & 0.00 + 69.6 97.9 & 0.99 & 0.99 & 0.96 & 0.50 & 0.02 & 0.00 & 0.00 + 97.9 137.7 & 0.99 & 0.98 & 0.84 & 0.24 & 0.01 & 0.00 & 0.00 + 137.7 193.6 & 0.98 & 0.91 & 0.57 & 0.07 & 0.00 & 0.00 & 0.00 + 193.6 272.4 & 0.90 & 0.66 & 0.21 & 0.01 & 0.00 & 0.00 & 0.00 + 272.4 383.2 & 0.67 & 0.30 & 0.04 & 0.00 & 0.00 & 0.00 & 0.00 + 383.2 539.0 & 0.40 & 0.16 & 0.01 & 0.00 & 0.00 & 0.00 & 0.00 + 539.0 758.1 & 0.21 & 0.03 & 0.01 & 0.01 & 0.00 & 0.00 & 0.00 + 758.11066.4 & 0.04 & 0.02 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 + 1066.41500.0 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 + lcccccccc ptabularcr ptabularcr ptabularcr 0.12 0.21 & 0.97 & 0.97 & 0.97 & 0.97 & 0.97 & 0.85 & 0.11 + 0.21 0.37 & 0.95 & 0.94 & 0.95 & 0.95 & 0.95 & 0.80 & 0.09 + 0.37 0.66 & 0.92 & 0.93 & 0.93 & 0.92 & 0.92 & 0.70 & 0.05 + 0.66 1.17 & 0.90 & 0.90 & 0.90 & 0.90 & 0.88 & 0.56 & 0.02 + 1.17 2.06 & 0.88 & 0.88 & 0.89 & 0.87 & 0.82 & 0.36 & 0.01 + 2.06 3.63 & 0.87 & 0.89 & 0.89 & 0.84 & 0.65 & 0.14 & 0.00 + 3.63 6.41 & 0.85 & 0.82 & 0.75 & 0.63 & 0.29 & 0.02 & 0.00 + 6.41 11.32 & 0.71 & 0.62 & 0.42 & 0.24 & 0.04 & 0.00 & 0.00 + 11.32 19.99 & 0.43 & 0.31 & 0.10 & 0.03 & 0.00 & 0.00 & 0.00 + 19.99 35.29 & 0.11 & 0.03 & 0.02 & 0.00 & 0.00 & 0.00 & 0.00 + 35.29 62.31 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 &
0.00 + 62.31110.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 + ptabularcr ptabularcr ptabularcr 0.07 0.12 & 0.98 & 0.98 & 0.97 & 0.98 & 0.98 & 0.86 & 0.11 + 0.12 0.22 & 0.95 & 0.96 & 0.96 & 0.96 & 0.96 & 0.83 & 0.10 + 0.22 0.38 & 0.93 & 0.93 & 0.93 & 0.93 & 0.93 & 0.75 & 0.07 + 0.38 0.66 & 0.92 & 0.92 & 0.92 & 0.92 & 0.91 & 0.65 & 0.04 + 0.66 1.17 & 0.89 & 0.89 & 0.89 & 0.88 & 0.85 & 0.49 & 0.01 + 1.17 2.05 & 0.87 & 0.89 & 0.89 & 0.86 & 0.78 & 0.28 & 0.00 + 2.05 3.60 & 0.88 & 0.88 & 0.86 & 0.82 & 0.56 & 0.08 & 0.00 + 3.60 6.32 & 0.82 & 0.80 & 0.71 & 0.54 & 0.21 & 0.01 & 0.00 + 6.32 11.09 & 0.72 & 0.61 & 0.42 & 0.26 & 0.03 & 0.00 & 0.00 + 11.09 19.47 & 0.45 & 0.27 & 0.09 & 0.02 & 0.00 & 0.00 & 0.00 + 19.47 34.18 & 0.14 & 0.07 & 0.03 & 0.00 & 0.00 & 0.00 & 0.00 + 34.18 60.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 + lccccccc ptabularcr ptabularcr ptabularcr 0.12 0.21 & 0.96 & 0.96 & 0.95 & 0.96 & 0.96 & 0.83 & 0.10 + 0.21 0.37 & 0.93 & 0.93 & 0.93 & 0.93 & 0.93 & 0.78 & 0.08 + 0.37 0.66 & 0.91 & 0.91 & 0.91 & 0.91 & 0.90 & 0.68 & 0.04 + 0.66 1.17 & 0.89 & 0.89 & 0.89 & 0.88 & 0.86 & 0.54 & 0.02 + 1.17 2.06 & 0.87 & 0.87 & 0.88 & 0.87 & 0.80 & 0.34 & 0.01 + 2.06 3.63 & 0.86 & 0.88 & 0.88 & 0.83 & 0.62 & 0.12 & 0.00 + 3.63 6.41 & 0.84 & 0.80 & 0.73 & 0.60 & 0.26 & 0.02 & 0.00 + 6.41 11.32 & 0.69 & 0.60 & 0.41 & 0.23 & 0.03 & 0.00 & 0.00 + 11.32 19.99 & 0.42 & 0.28 & 0.08 & 0.03 & 0.00 & 0.00 & 0.00 + 19.99 35.29 & 0.11 & 0.03 & 0.02 & 0.00 & 0.00 & 0.00 & 0.00 + 35.29 62.31 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 &
0.00 + 62.31110.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 + ptabularcr ptabularcr ptabularcr 0.07 0.12 & 0.96 & 0.96 & 0.96 & 0.96 & 0.96 & 0.67 & 0.04 + 0.12 0.22 & 0.94 & 0.94 & 0.95 & 0.95 & 0.94 & 0.63 & 0.03 + 0.22 0.38 & 0.92 & 0.91 & 0.92 & 0.91 & 0.91 & 0.56 & 0.02 + 0.38 0.66 & 0.91 & 0.91 & 0.91 & 0.91 & 0.89 & 0.46 & 0.01 + 0.66 1.17 & 0.88 & 0.88 & 0.88 & 0.87 & 0.82 & 0.31 & 0.00 + 1.17 2.05 & 0.86 & 0.88 & 0.88 & 0.85 & 0.69 & 0.16 & 0.00 + 2.05 3.60 & 0.87 & 0.87 & 0.85 & 0.77 & 0.43 & 0.04 & 0.00 + 3.60 6.32 & 0.81 & 0.78 & 0.67 & 0.46 & 0.14 & 0.00 & 0.00 + 6.32 11.09 & 0.70 & 0.56 & 0.35 & 0.19 & 0.02 & 0.00 & 0.00 + 11.09 19.47 & 0.42 & 0.23 & 0.07 & 0.02 & 0.00 & 0.00 & 0.00 + 19.47 34.18 & 0.12 & 0.03 & 0.02 & 0.00 & 0.00 & 0.00 & 0.00 + 34.18 60.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 + lccccccc ptabularcr ptabularcr ptabularcr 0.12 0.21 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.87 & 0.13 + 0.21 0.37 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.84 & 0.10 + 0.37 0.66 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.75 & 0.06 + 0.66 1.17 & 1.00 & 0.99 & 0.99 & 0.98 & 0.96 & 0.60 & 0.03 + 1.17 2.06 & 0.99 & 0.98 & 0.97 & 0.95 & 0.88 & 0.40 & 0.01 + 2.06 3.63 & 0.98 & 0.98 & 0.96 & 0.91 & 0.71 & 0.16 & 0.00 + 3.63 6.41 & 0.96 & 0.93 & 0.85 & 0.71 & 0.33 & 0.03 & 0.00 + 6.41 11.32 & 0.90 & 0.79 & 0.51 & 0.32 & 0.04 & 0.00 & 0.00 + 11.32 19.99 & 0.72 & 0.48 & 0.17 & 0.03 & 0.00 & 0.00 & 0.00 + 19.99 35.29 & 0.27 & 0.09 & 0.02 & 0.00 & 0.00 & 0.00 & 0.00 + 35.29 62.31 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 &
0.00 + 62.31110.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 + ptabularcr ptabularcr ptabularcr 0.07 0.12 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.88 & 0.13 + 0.12 0.22 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.86 & 0.12 + 0.22 0.38 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.81 & 0.08 + 0.38 0.66 & 1.00 & 1.00 & 1.00 & 0.99 & 0.98 & 0.70 & 0.05 + 0.66 1.17 & 1.00 & 0.99 & 0.98 & 0.97 & 0.93 & 0.53 & 0.02 + 1.17 2.05 & 0.99 & 0.98 & 0.97 & 0.93 & 0.84 & 0.31 & 0.00 + 2.05 3.60 & 0.98 & 0.97 & 0.94 & 0.89 & 0.61 & 0.09 & 0.00 + 3.60 6.32 & 0.96 & 0.92 & 0.82 & 0.64 & 0.24 & 0.02 & 0.00 + 6.32 11.09 & 0.89 & 0.80 & 0.50 & 0.32 & 0.03 & 0.00 & 0.00 + 11.09 19.47 & 0.76 & 0.45 & 0.16 & 0.03 & 0.00 & 0.00 & 0.00 + 19.47 34.18 & 0.35 & 0.12 & 0.03 & 0.00 & 0.00 & 0.00 & 0.00 + 34.18 60.00 & 0.06 & 0.03 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 + lccccccc ptabularcr ptabularcr ptabularcr 0.12 0.21 & 0.99 & 1.00 & 1.00 & 1.00 & 1.00 & 0.87 & 0.13 + 0.21 0.37 & 0.99 & 1.00 & 1.00 & 1.00 & 1.00 & 0.84 & 0.10 + 0.37 0.66 & 0.99 & 1.00 & 1.00 & 0.99 & 0.99 & 0.74 & 0.05 + 0.66 1.17 & 0.99 & 0.99 & 0.98 & 0.97 & 0.95 & 0.59 & 0.02 + 1.17 2.06 & 0.98 & 0.97 & 0.96 & 0.94 & 0.87 & 0.38 & 0.01 + 2.06 3.63 & 0.97 & 0.97 & 0.95 & 0.91 & 0.70 & 0.15 & 0.00 + 3.63 6.41 & 0.95 & 0.91 & 0.84 & 0.69 & 0.31 & 0.02 & 0.00 + 6.41 11.32 & 0.87 & 0.78 & 0.50 & 0.29 & 0.04 & 0.00 & 0.00 + 11.32 19.99 & 0.68 & 0.45 & 0.14 & 0.03 & 0.00 & 0.00 & 0.00 + 19.99 35.29 & 0.27 & 0.07 & 0.02 & 0.00 & 0.00 & 0.00 & 0.00 + 35.29 62.31 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 &
0.00 + 62.31110.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 + ptabularcr ptabularcr ptabularcr 0.07 0.12 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.85 & 0.12 + 0.12 0.22 & 1.00 & 1.00 & 1.00 & 1.00 & 1.00 & 0.84 & 0.11 + 0.22 0.38 & 1.00 & 1.00 & 1.00 & 1.00 & 0.99 & 0.78 & 0.07 + 0.38 0.66 & 1.00 & 1.00 & 1.00 & 0.99 & 0.98 & 0.68 & 0.04 + 0.66 1.17 & 1.00 & 0.99 & 0.98 & 0.97 & 0.93 & 0.50 & 0.02 + 1.17 2.05 & 0.99 & 0.98 & 0.97 & 0.93 & 0.83 & 0.29 & 0.00 + 2.05 3.60 & 0.98 & 0.97 & 0.94 & 0.89 & 0.59 & 0.08 & 0.00 + 3.60 6.32 & 0.96 & 0.92 & 0.82 & 0.63 & 0.23 & 0.01 & 0.00 + 6.32 11.09 & 0.89 & 0.80 & 0.49 & 0.32 & 0.03 & 0.00 &
0.00 + 11.09 19.47 & 0.76 & 0.43 & 0.16 & 0.03 & 0.00 & 0.00 & 0.00 + 19.47 34.18 & 0.35 & 0.10 & 0.03 & 0.00 & 0.00 & 0.00 & 0.00 + 34.18 60.00 & 0.06 & 0.03 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 + | we characterize the completeness of point source lists from _ spitzer space telescope _
surveys in the four infrared array camera ( irac ) bandpasses , emphasizing the galactic legacy infrared mid - plane survey extraordinaire ( glimpse ) programs ( glimpse i , ii , 3d , 360 ; deep glimpse ) and their resulting point source catalogs and archives .
the analysis separately addresses effects of incompleteness resulting from high diffuse background emission and incompleteness resulting from point source confusion ( i.e. , crowding ) . an artificial star addition and
extraction analysis demonstrates that completeness is strongly dependent on local background brightness and structure , with high - surface - brightness regions suffering up to five magnitudes of reduced sensitivity to point sources .
this effect is most pronounced at the irac 5.8 and 8.0 @xmath0 m bands where uv - excited pah emission produces bright , complex structures ( photodissociation regions ; pdrs ) . with regard to diffuse background effects
, we provide the completeness as a function of stellar magnitude and diffuse background level in graphical and tabular formats .
these data are suitable for estimating completeness in the low - source - density limit in any of the four irac bands in glimpse catalogs and archives and some other @xmath1 irac programs that employ similar observational strategies and are processed by the glimpse pipeline . by performing the same analysis on smoothed images we show that the point source incompleteness is primarily a consequence of _ structure _ in the diffuse background emission rather than photon noise . with regard to source confusion in the high - source - density regions of the galactic plane ,
we provide figures illustrating the 90% completeness levels as a function of point source density at each band .
we caution that completeness of the glimpse 360/deep glimpse catalogs is suppressed relative to the corresponding archives as a consequence of rejecting stars that lie in the psf wings of saturated sources .
this effect is minor in regions of low saturated star density , such as toward the outer galaxy ; this effect is significant along sightlines having a high density of saturated sources , especially for deep glimpse and other programs observing closer to the galactic center using 12 s or longer exposure times . |
in recent years , many cosmological observations have provided the strong evidence that the universe is flat and its energy density is dominated by the dark energy component whose negative pressure causes the cosmic expansion to accelerate @xcite@xmath0@xcite . in order to clarify the origin of the dark energy ,
one has tried to understand the connection of the dark energy with particle physics . in a dynamical model proposed by fardon , nelson and weiner ( mavans ) ,
relic neutrinos could form a negative pressure fluid and cause the cosmic acceleration @xcite . in this model , an unknown scalar field which is called `` acceleron '' is introduced and neutrinos are assumed to interact through a new scalar force .
the acceleron sits at the instantaneous minimum of its potential , and the cosmic expansion only modulates this minimum through changes in the neutrino density .
therefore , the neutrino mass is given by the acceleron , in other words , it depends on its number density and changes with the evolution of the universe .
the equation of state parameter @xmath1 and the dark energy density also evolve with the neutrino mass .
those evolutions depend on a model of the scalar potential and the relation between the acceleron and the neutrino mass strongly .
typical examples of the potential have been discussed in ref .
@xcite .
the variable neutrino mass was considered at first in ref .
@xcite , and was discussed for neutrino clouds @xcite . ref .
@xcite considered coupling a sterile neutrino to a slowly rolling scalar field which was responsible for the dark energy .
@xcite considered coupling of the dark energy scalar , such as the quintessence to neutrinos and discuss its impact on the neutrino mass limits from baryogenesis . in the context of the mavans scenario , there have been a lot of works @xcite@xmath0@xcite .
the origin of the scalar potential for the acceleron was not discussed in many literatures , however , that is clear in the supersymmetric mavans scenario @xcite . in this work , we present a model including the supersymmetry breaking effect mediated by the gravity .
then we show evolutions of the neutrino mass and the equation of state parameter in the model .
the paper is organized as follows : in section ii , we summarize the supersymmetric mavans scenario and present a model . sec .
iii is devoted to a discussion of the supersymmetry breaking effect mediated by the gravity in the dark sector . in sec .
iv , the summary is given .
in this section , we discuss the supersymmetric mass varying neutrinos scenario and present a model . the basic assumption of the mavans with supersymmetry is to introduce a chiral superfield @xmath2 in the dark sector , which is assumed to be a singlet under the gauge group of the standard model .
it is difficult to build a viable mavans model without fine - tunings in some parameters when one assumes one chiral superfield in the dark sector , which couples to only the left - handed lepton doublet superfield @xcite .
therefore , we assume that the superfield @xmath2 couples to both the left - handed lepton doublet superfield @xmath3 and the right - handed neutrino superfield @xmath4 . in this framework , we suppose the superpotential @xmath5 where @xmath6 are coupling constant of @xmath7 and @xmath8 , @xmath9 , @xmath10 and @xmath11 are mass parameters .
the scalar and spinor component of @xmath2 are @xmath12 , and the scalar component is assumed to be the acceleron which cause the present cosmic acceleration .
the spinor component is a sterile neutrino .
the helium-4 abundancy gives the most accurate determination of a cosmological number of neutrinos and does not exclude a fourth thermalized neutrino at @xmath13 @xcite .
the third term of the right - hand side in eq .
( [ w ] ) is derived from the yukawa coupling such as @xmath14 with @xmath15 , where @xmath16 is the higgs doublet . in the mavans scenario ,
the dark energy is assumed to be the sum of the neutrino energy density and the scalar potential for the acceleron : @xmath17 since only the acceleron potential contributes to the dark energy , we assume the vanishing vacuum expectation values of sleptons , and thus the effective scalar potential is given as @xmath18 we can write down a lagrangian density from eq .
( [ w ] ) , @xmath19 it is noticed that the lepton number conservation in the dark sector is violated because this lagrangian includes both @xmath20 and @xmath21 . after integrating out the right - handed neutrino , the effective neutrino mass matrix is given by @xmath22 in the basis of @xmath23 , where we assume @xmath24 .
the first term of the @xmath25 element of this matrix corresponds to the usual term given by the seesaw mechanism @xcite in the absence of the acceleron .
the second term is derived from the coupling between the acceleron and the right - handed neutrino but the magnitude of this term is negligible small because of the suppression of @xmath26 .
therefore , we can rewrite the neutrino mass matrix as @xmath27 where @xmath28 .
it is remarked that only the mass of a sterile neutrino is variable in the case of the vanishing mixing ( @xmath29 ) between the left - handed and a sterile neutrino on cosmological time scale . the finite mixing ( @xmath30 )
makes the mass of the left - handed neutrino variable .
we will consider these two cases of @xmath29 and @xmath30 later . in the mavans scenario ,
there are two constraints on the scalar potential .
the first one comes from observations of the universe , which is that the magnitude of the present dark energy density is about @xmath31 , @xmath32 being the critical density .
thus , the first constraint turns to @xmath33 where `` @xmath34 '' represents a value at the present epoch .
the second one is the stationary condition . in this scenario ,
the neutrino mass is assumed to be a dynamical field which is a function of the acceleron .
therefore , the dark energy density should be stationary with respect to the variation of the neutrino mass : @xmath35 if @xmath36 , this condition is equivalent to the usual stationary condition stabilized by an ordinary scalar field .
( [ stationary ] ) is rewritten by using the cosmic temperature @xmath37 : @xmath38 where @xmath39 , @xmath40 and @xmath41 we can get the time evolution of the neutrino mass from eq . ( [ stationary1 ] ) . since the stationary condition should be always satisfied in the evolution of the universe , this one at the present epoch is the second constraint on the scalar potential : @xmath42 in addition to two constraints for the potential , we also have two relations between the acceleron and the neutrino mass at the present epoch : @xmath43 ^ 2 + 4m_d^2 } } { 2},\\ m_\psi ^0&=&\frac{c+m_a+\lambda _ 1\phi ^0}{2}\nonumber\\ & & -\frac{\sqrt{[c-(m_a+\lambda _ 1\phi ^0)]^2 + 4m_d^2}}{2}. \end{aligned}\ ] ] next , we will consider the dynamics of the acceleron field . in order that the acceleron does not vary significantly on distance of inter - neutrino spacing , the acceleron mass at the present epoch must be less than @xmath44 @xcite . here and below
, we fix the present acceleron mass as @xmath45 once we adjust parameters which satisfy five equations ( [ v ] ) and ( [ stationary2])@xmath0([amass ] ) , we can have evolutions of the neutrino mass by using the eq .
( [ stationary1 ] ) .
the dark energy is characterized by the evolution of the equation of state parameter @xmath1 .
the equation of state in this scenario is derived from the energy conservation equation in the robertson - walker background and the stationary condition eq .
( [ stationary1 ] ) : @xmath46\rho _ \nu}{3\rho _ { \mbox{{\scriptsize de } } } } , \end{aligned}\ ] ] where @xmath47 it seems that @xmath1 in this scenario depend on the neutrino mass and the cosmic temperature .
this means that @xmath1 varies with the evolution of the universe unlike the cosmological constant . in the last of this section
, it is important to discuss the hydrodynamic stability of the dark energy from mavans .
the speed of sound squared in the neutrino - acceleron fluid is given by @xmath48 where @xmath49 is the pressure of the dark energy .
recently , it was argued that when neutrinos are non - relativistic , this speed of sound squared becomes negative in this scenario @xcite .
the emergence of an imaginary speed of sound shows that the mavans scenario with non - relativistic neutrinos is unstable , and thus the fluid in this scenario can not acts as the dark energy .
however , finite temperature effects provide a positive contribution to the speed of sound squared and avoid this instability @xcite .
then , a model should satisfy the following condition , @xmath50 where @xmath51 is the redshift parameter , @xmath52 , and @xmath53 actually , some models satisfy this condition @xcite .
in order to consider the effect of supersymmetry breaking in the dark sector , we assume a superfield @xmath54 , which breaks supersymmetry , in the hidden sector , and the chiral superfield @xmath2 in the dark sector is assumed to interact with the hidden sector only through the gravity .
this framework is shown graphically in fig .
[ fig:0 ] .
once supersymmetry is broken at tev scale , its effect is transmitted to the dark sector through the following operators : @xmath55 where @xmath56 is the planck mass .
then , the scale of the soft terms @xmath57-@xmath58 is expected .
such a framework was discussed in the `` acceleressence '' scenario @xcite .
now , we consider only one superfield which breaks supersymmetry for simplicity .
if one extend the hidden sector , one can consider a different mediation mechanism between the standard model and the hidden sector from one between the dark and the hidden sector .
we will return to this point later . in this framework , taking supersymmetry breaking effect into account ,
the scalar potential is given by @xmath59 where @xmath60 and @xmath61 are supersymmetry breaking parameters , and @xmath62 is a constant determined by the condition that the cosmological constant is vanishing at the true minimum of the acceleron potential .
this scalar potential is the same one presented in @xcite .
we consider two types of the neutrino mass matrix in this scalar potential .
they are the cases of the vanishing and the finite mixing between the left - handed and a sterile neutrino . when the mixing between the left - handed and a sterile neutrino is vanishing , @xmath29 in the neutrino mass matrix ( [ mm ] )
. then we have the mass of the left - handed and a sterile neutrino as @xmath63 in this case , we find that only the mass of a sterile neutrino is variable on cosmological time scale due to the second term in eq .
( [ sterilem ] ) .
let us adjust parameters which satisfy eqs .
( [ v ] ) and ( [ stationary2])@xmath0([amass ] ) .
( [ v ] ) , the scalar potential eq .
( [ v1 ] ) is used .
putting typical values for four parameters by hand as follows : @xmath64 we have @xmath65 there is a tuning between @xmath8 and @xmath61 in order to satisfy the constraint on the present accerelon mass of eq .
( [ amass ] ) . ) ] now , we can calculate evolutions of the mass of a sterile neutrino and the equation of state parameter @xmath1 by using these values .
the numerical results are shown in figs .
[ fig:2 ] , [ fig:3 ] and [ fig:4 ] . especially , the behavior of the mass of a neutrino near the present epoch is shown in fig .
[ fig:3 ] .
we find that the mass of a sterile neutrino have varied slowly in this epoch .
this means that the first term of the left hand side in eq .
( [ constraint ] ) , which is a negative contribution to the speed of sound squared , is tiny .
we can also check the positive speed of sound squared in a numerical calculation .
therefore , the neutrino - acceleron fluid is hydrodynamically stable and acts as the dark energy . ) ] ( @xmath66 ) ] next , we consider the case of the finite mixing between the left - handed and a sterile neutrino ( @xmath30 ) . in this case ,
the left - handed and a sterile neutrino mass are given by @xmath67 ^ 2 + 4m_d^2 } } { 2},\\ m_\psi&=&\frac{c+m_a+\lambda _ 1\phi } { 2}\nonumber\\ & & -\frac{\sqrt{[c-(m_a+\lambda _
1\phi ) ] ^2 + 4m_d^2}}{2}. \end{aligned}\ ] ] we can expect that both the mass of the left - handed and a sterile neutrino have varied on cosmological time scale due to the term of the acceleron dependence .
taking typical values for four parameters as @xmath68 we have @xmath69 where we required that the mixing between the active and a sterile neutrino is tiny . in our model , the small present value of the acceleron is needed to satisfy the constraints on the scalar potential in eqs .
( [ v ] ) and ( [ stationary2 ] ) . ) ] ) ] ( @xmath66 ) ] when we add following terms of the supersymmetry breaking effects to the scalar potential , @xmath70 a small gravitino mass ( @xmath71 ) is favored .
such a small gravitino mass has been given in the gauge mediation model of ref.@xcite .
therefore , a mediation mechanism between the standard and the hidden sector which leads to a small gravitino mass is suitable for our framework .
values of parameters in ( [ value1 ] ) are almost same as the case of the vanishing mixing ( [ value ] ) . however , the mass of the left - handed neutrino is variable unlike the vanishing mixing case .
the time evolution of the left - handed neutrino mass is shown in fig .
. the mixing does not affect the evolution of a sterile neutrino mass and the equation of state parameter , which are shown in figs .
[ fig:6 ] , [ fig:7 ] .
since the variation in the mass of the left - handed neutrino is not vanishing but extremely small , the model can also avoid the instability of speed of sound . the value of the sum of the left - handed and a sterile neutrino mass is within the limit provided by analyses from wmap three - years date @xcite . finally , we comment the smallness of the evolution of the neutrino mass at the present epoch . in our model ,
the mass of the left - handed and a sterile neutrino include the constant part . a variable part is a function of the acceleron . in the present epoch ,
the constant part dominates the neutrino mass because the present value of the acceleron should be small .
this smallness of the value of the acceleron is required from the cosmological observation and the stationary condition in eqs .
( [ v ] ) and ( [ stationary2 ] ) .
we presented a supersymmetric mavans model including the effects of the supersymmetry breaking mediated by the gravity .
evolutions of the neutrino mass and the equation of state parameter have been calculated in the model .
our model has a chiral superfield in the dark sector , whose scalar component causes the present cosmic acceleration , and the right - handed neutrino superfield . in our framework
, supersymmetry is broken in the hidden sector at tev scale and the effect is assumed to be transmitted to the dark sector only through the gravity .
then , the scale of soft parameters are @xmath72-@xmath73 is expected .
we considered two types of model .
one is the case of the vanishing mixing between the left - handed and a sterile neutrino .
another one is the finite mixing case . in the case of the vanishing mixing ,
only the mass of a sterile neutrino had varied on cosmological time scale . in the epoch of @xmath74 ,
the sterile neutrino mass had varied slowly .
this means that the speed of sound squared in the neutrino acceleron fluid is positive , and thus this fluid can act as the dark energy . in the finite mixing case ,
the mass of the left - handed neutrino had also varied .
however , the variation is extremely small and the effect of the mixing does not almost affect the evolution of the sterile neutrino mass and the equation of state .
therefore , this model can also avoid the instability .
t. yanagida , in proceedings of the _ `` workshop on the unified theory and the baryon number in the universe '' _ , tsukuba , japan , 1979 , edited by o. sawada and a. sugamoto , kek report no .
kek-79 - 18 , p. 95 ; prog .
* 64 * ( 1980 ) 1103 ; | we discuss the effect of the supersymmetry breaking on the mass varying neutrinos(mavans ) scenario . especially , the effect mediated by the gravitational interaction between the hidden sector and the dark energy sector is studied . a model including a chiral superfield in the dark sector and the right handed neutrino superfield is proposed .
evolutions of the neutrino mass and the equation of state parameter are presented in the model .
it is remarked that only the mass of a sterile neutrino is variable in the case of the vanishing mixing between the left - handed and a sterile neutrino on cosmological time scale .
the finite mixing makes the mass of the left - handed neutrino variable . |
signal reconstruction is usually formulated as a linear inverse problem @xmath1 where @xmath2 is the true signal , @xmath3 is an @xmath4 real matrix , @xmath5 is the observed data and @xmath6 represents the random noise . over the last decade
, sparse reconstruction has received great attentions @xcite . in general ,
the task of sparse reconstruction is to find a sparse @xmath7 from .
obviously , the most natural measure of sparsity is the @xmath0-norm , defined at @xmath2 as @xmath8 where @xmath9 stands for the cardinality of the set @xmath10 and @xmath11 denotes the support of @xmath7 , that is @xmath12 .
when the noise level of @xmath13 is known , we consider the following constrained @xmath0 regularization problem @xmath14 where @xmath15 denotes the @xmath16-norm , and @xmath17 is the noise level . specially , in the noiseless case , i.e. , when @xmath18 , problem reduces to seeking for the sparsest solutions from a linear system as follows @xmath19 finding a global minimizer of problem is known to be combinational and np - hard in general @xcite as it involves the @xmath0-norm . the number of papers dealing with the @xmath0-norm is large and several types of numerical algorithms have been adopted to approximately solve the problem . in this paper , we mainly focus on the penalty methods . before making a further discussion on the penalty methods , we briefly review other two types of approaches , greedy pursuit methods and relaxation of the @xmath0-norm .
greedy pursuit methods , such as orthogonal matching pursuit @xcite and cosamp @xcite , adaptively update the support of @xmath7 by exploiting the residual and then solve a least - square problem on the support in each iteration .
this type of algorithms are widely used in practice due to their relatively fast computational speed . however , only under very strict conditions , can they be shown theoretically to produce desired numerical solutions to problem .
relaxation methods first replace the @xmath0-norm by other continuous cost functions , including convex functions @xcite and nonconvex functions @xcite , such as the bridge functions @xmath20 with @xmath21 @xcite , capped-@xmath22 functions @xcite , minmax concave functions @xcite and so on .
then the majorization - minimization strategy is applied to the relaxation problem , which leads to solving a constrained reweighted @xmath22 or @xmath16 subproblems @xcite in each step .
the cost functions of the iterates are shown to be descent @xcite , while the convergence of the iterates themselves are generally unknown .
the penalty method is a classical and important approach for constrained optimization . for an overview on the penalty method
, we refer the readers to the book @xcite .
an extensively used penalty method for problem is to solve the following least square penalty problem @xmath23 where @xmath24 is a penalty parameter . as problem
still involve the @xmath0-norm , the greedy pursuit and relaxation methods mentioned before are applicable for approximately solving the problem .
however , it has relatively better structure than problem from the standpoint of algorithmic design .
more precisely , the proximal operator of @xmath0-norm has a closed form , while the least square penalty term in problem is differentiable with a lipschitz continuous gradient .
therefore , it can be directly and efficiently solved by proximal - gradient type algorithms @xcite or primal - dual active set methods @xcite .
one specific proximal - gradient type algorithm for problem is the iterative hard thresholding ( iht ) @xcite .
in general , it is shown in @xcite that for arbitrary initial point , the iterates generated by the algorithm converge to a local minimizer of problem . under certain conditions
, the iht converges with an asymptotic linear rate .
the structure of the least square penalty problem makes it easier to develop convergent numerical algorithms . the penalty parameter @xmath25 balances the magnitude of the residual and the sparsity of solutions to problem .
an important issue is that wether problem and share a common global minimizer for some penalty parameter @xmath25 .
it seems intuitive that there exists certain connections between these two problems .
however , to the best of our knowledge , there is very little theoretical results on penalty methods for problem in general .
recently , chen , lu and pong study the penalty methods for a class of constrained non - lipschitz optimization @xcite .
the non - lipschitz cost functions considered in @xcite does not contain the @xmath0-norm and the corresponding results can not be generalized to the constrained @xmath0 regularization problems .
nikolova in @xcite gives a detailed mathematical description of optimal solutions to problem and further in @xcite establishes their connections with optimal solutions to the following optimization problem @xmath26 where @xmath27 is a given positive integer .
problem also involves the @xmath0-norm and is usually called the sparsity constrained minimization problem .
however , optimal solutions of problem and their connections with problem have not been studied in these works .
this paper is devoted to describing global minimizers of a class of constrained @xmath0 regularization problems and the corresponding penalty methods . as in @xcite
, we use `` optimal solutions '' for global minimizers and `` optimal solution set '' for the set of all global minimizers in the remaining part of this paper .
the class of constrained @xmath0 regularization problems we considered has the following form @xmath28 where @xmath29 or @xmath30 . obviously , problem has two cases , the @xmath16 constrained problem and the @xmath22 constrained problem as follows @xmath31 the analysis of these two constrained problems and their penalty methods share many common points and thus we shall discuss them in a unified way .
moreover , we introduce a class of scalar penalty functions @xmath32 satisfying the following blanket assumption , where @xmath33 .
[ hypo:1 ] the function @xmath34 is continuous , nondecreasing with @xmath35 .
then the penalty problems for can be in general formulated as @xmath36 where @xmath24 is a penalty parameter . in the case of @xmath37 , when we set the penalty function @xmath38 for @xmath39 , the general penalty problem reduces to the specific problem
. we shall give a comprehensive study on connections between optimal solutions of problem and problem .
our main contributions are summarized below .
* we study in depth the existence , stability and cardinality of solutions to problems and .
in particular , we prove that the optimal solution set of problem is piecewise constant with respect to @xmath25 .
more precisely , we present a sequence of real number @xmath40 where @xmath41 . we claim that for all @xmath42 , problems with different @xmath43 share completely the same optimal solution set . *
we clarify the relationship between optimal solutions of problem and those of problem .
we find that , when @xmath44 problem probably * never * has a common optimal solution with problem for any @xmath24 .
this means that the penalty problem may fail to be an exact penalty formulation of problem .
a numerical example is presented to illustrate this fact in section [ sec : exp ] .
* we show that under a mild condition on the penalty function @xmath45 , problem has exactly the same optimal solution set as that of problem once @xmath46 for some @xmath47 . in particular , for @xmath37 we present a penalty function @xmath45 with which the penalty term in problem is differentiable with a lipschitz continuous gradient .
this penalty function enables us to design innovative numerical algorithms for problem .
the remaining part of this paper is organized as follows . in section [ sec : exist ] we show the existence of optimal solutions to problems and .
the stability for problems and are discussed in section [ sec : sta ] .
we analyze in section [ sec : relation ] the relationship between optimal solutions of problems and .
section [ sec : equal ] establishes a sufficient condition on @xmath45 , which ensures problems and have completely the same optimal solution set when the parameter @xmath25 is sufficiently large .
we further investigate the cardinality and strictness of optimal solutions to problems and in section [ sec : minimizer ] .
the theoretical results are illustrated by numerical experiments in section [ sec : exp ] .
we conclude this paper in section [ sec : con ] .
this section is devoted to the existence of optimal solutions to problems and .
the former issue is trivial , while the latter issue requires careful analysis . after a brief discussion on the existence of optimal solutions to problem
, we shall focus on a constructive proof of the existence of optimal solutions to problem .
since the objective function of problem is proper , lower semi - continuous and has finite values , it is clear that problem has optimal solutions as long as the feasible region is nonempty .
we present the result in the following theorem . for simplicity , we denote by @xmath48 the feasible region of problem , that is @xmath49 [ thm : existconst ] the optimal solution set of problem is nonempty if and only if @xmath50 , where @xmath48 is defined by . in the rest of this section , we dedicate to discussing the existence of optimal solutions to problem .
since the objective function of consists of two terms , we begin with defining several notations by alternating minimizing each term . to this end , for any positive integer @xmath51 , we define @xmath52 [ def : srhoomega ] given @xmath45 satisfying h[hypo:1 ] and @xmath53 , the integer @xmath54 , the sets @xmath55 , @xmath56 , @xmath57 are defined by the following iteration @xmath58 we first show that definition [ def : srhoomega ] is well defined .
it suffices to prove both the following two optimization problems @xmath59 and @xmath60 have optimal solutions for any @xmath61 and @xmath62 .
clearly , problem has an optimal solution since the objective function is piecewise constant and has finite values .
the next lemma tells us that the optimal solution set of problem is always nonempty for any @xmath63 .
we denote by @xmath64 the @xmath65-dimensional vector with @xmath66 as its components .
for an @xmath4 matrix @xmath67 and @xmath68 , let @xmath69 formed by the columns of @xmath67 with indexes in @xmath70 .
similarly , for any @xmath2 and @xmath68 , we denote by @xmath71 the vector formed by the components of @xmath7 with indexes in @xmath70 .
[ lema : l0constsolu ] given @xmath45 satisfying h[hypo:1 ] and @xmath53 , for any @xmath63 , the optimal solution set of problem is nonempty
. as function @xmath45 is nondecreasing in @xmath72 , it suffices to show that the optimization problem @xmath73 has an optimal solution .
we first prove that for any @xmath68 , the optimization problem @xmath74 has an optimal solution .
obviously , @xmath75 is an optimal solution to problem if and only if @xmath76 and @xmath77 in the case of @xmath37 , it is well - known that @xmath78 is an optimal solution to problem , where @xmath79 is the pseudo - inverse matrix of the matrix @xmath80 .
when @xmath29 , the optimal solution set of problem is nonempty since @xmath81 is a piecewise linear function and bounded below , see proposition 3.3.3 and 3.4.2 in @xcite .
therefore , problem has an optimal solution for any @xmath68 . with the help of problem
, problem can be equivalently written as @xmath82 clearly , the optimal solution set of problem is nonempty , therefore , problem has an optimal solution .
we then complete the proof . by lemma [ lema : l0constsolu ] , definition [ def : srhoomega ]
is well defined .
we present this result in the following proposition .
given @xmath45 satisfying h[hypo:1 ] and @xmath53 , the integer @xmath54 and @xmath83 for @xmath84 can be well defined by definition [ def : srhoomega ] .
we then provide some properties of @xmath54 and @xmath83 for @xmath85 defined by definition [ def : srhoomega ] in the following proposition .
[ prop : rhosomega ] given @xmath45 satisfying h[hypo:1 ] and @xmath53 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] .
then the following statements hold : * @xmath86 and @xmath87 , where @xmath88 denotes the rank of @xmath3 .
* @xmath89 .
* @xmath90 .
* @xmath91 , @xmath92 and @xmath93 , for @xmath94 , @xmath95 . * @xmath96 for @xmath97 .
items ( ii ) , ( iii ) , ( iv ) and ( v ) are clear by definition [ def : srhoomega ] .
we dedicate to proving item ( i ) following .
let @xmath98 . from the proof of lemma [ lema : l0constsolu ] ,
since @xmath45 is nondecreasing , @xmath100 for any @xmath101 . let @xmath102 , then @xmath103 .
let @xmath104 denote the @xmath105th column of @xmath3 , @xmath106 . since the rank of @xmath3 is @xmath88 , without loss of generality , we assume @xmath107 is the maximal linearly independent group of @xmath108
. then there exists @xmath109 such that @xmath110 .
we define @xmath111 as @xmath112 for @xmath113 and @xmath114 for @xmath115 .
it is obvious that @xmath116 and @xmath117 , implying @xmath118 .
therefore @xmath119 .
then by the definition of @xmath120 , we have @xmath121 . from the procedure of the iteration in definition [ def : srhoomega ] and @xmath122
, we obtain @xmath123 . with the help of definition [ def :
srhoomega ] , the euclid space @xmath124 can be partitioned into @xmath125 sets : @xmath126 , @xmath127 , @xmath128 .
therefore , in order to establish the existence of optimal solutions to problem , it suffices to prove the existence of optimal solutions to problem @xmath129 and @xmath130 for all @xmath131 .
we present these desired results in the following lemma .
[ lema : partitionsolution ] given @xmath45 satisfying h[hypo:1 ] and @xmath53 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] . then for any @xmath24 ,
the following statements hold : * the optimal solution sets of problems and for @xmath128 are not empty . *
the optimal value of problem is @xmath132 and the optimal solution set to problem is @xmath133 . *
the optimal value of problem is @xmath134 and the optimal solution set of problem is @xmath135 , for @xmath128 . we only need to prove items ( ii ) and ( iii ) since they imply item ( i ) .
we first prove item ( ii ) .
it is obvious that restricted to the set @xmath126 , the minimal value of the first term @xmath136 is @xmath120 and can be attained at any @xmath137 . by definition [ def : srhoomega ] , the minimal value of @xmath138 is @xmath139 and can be attained at any @xmath137 .
therefore , the optimal value of problem is @xmath132 .
let @xmath140 be the optimal solution set of problem
. clearly , @xmath141 .
we then try to prove @xmath142 .
it suffices to prove @xmath143 for any @xmath101 .
if not , there exists @xmath102 and @xmath144 , then @xmath145 .
then the objective function value at @xmath146 is no less than @xmath147 due to the definition of @xmath139 , contradicting the fact that @xmath146 is an optimal solution of problem .
then we get item ( ii ) .
item ( iii ) can be obtained similarly , we omit the details here . for convenient presentation ,
we define @xmath148 , for @xmath97 , at @xmath149 as @xmath150 now , we are ready to show the existence of optimal solutions to problem .
[ thm : exist ] given @xmath45 satisfying h[hypo:1 ] and @xmath53 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] .
let @xmath151 for @xmath152 be defined by .
then , the optimal solution set of problem is nonempty for any @xmath24 .
furthermore , the following statements hold for any fixed @xmath24 : * the optimal value of problem is @xmath153 .
* @xmath154 is the optimal solution set of problem , where @xmath155 .
we omit the proof here since it is a direct result of the fact that @xmath156 and lemma [ lema : partitionsolution ] . [ remark : l=0 ] according to definition [ def : srhoomega ] , if @xmath157 , then @xmath158 and @xmath159 .
then , by theorem [ thm : exist ] , if @xmath157 , the optimal value and the optimal solution set of problem are @xmath160 and @xmath161 respectively for any @xmath24 .
in this section , we study the stability for problems and , including behaviors of the optimal values and optimal solution sets with respect to changes in the corresponding parameters .
we prove that , the optimal value of problem and that of problem change piecewise constantly and piecewise linearly respectively as the corresponding parameters vary .
moreover , we prove that the optimal solution set of problem is piecewise constant with respect to the parameter @xmath25 .
we begin with introducing the notion of marginal functions , which plays an important role in optimization theory , see for example @xcite .
set @xmath162 where @xmath53 . by lemma [ lema : l0constsolu ] , we have @xmath163 is well defined .
let @xmath164 defined at @xmath165 as @xmath166 where @xmath48 is defined by .
clearly , for a fixed @xmath165 , @xmath167 is the optimal value of problem .
thus @xmath168 is well defined due to theorem [ thm : existconst ] .
the function @xmath168 is also called the marginal function of problem .
similarly , we can define the marginal function of problem .
given @xmath45 satisfying h[hypo:1 ] and @xmath53 , we define @xmath169 at @xmath24 as @xmath170 clearly , @xmath171 is well defined due to theorem [ thm : exist ] . in order to study the stability of the optimal solution sets of problems and
, we define @xmath172 at @xmath165 as the optimal solution set of the constrained problem , that is , @xmath173 where @xmath48 is defined by and @xmath174 collects all the subsets of @xmath124 . by theorem [ thm : existconst ] , @xmath175 is well defined .
we also define @xmath176 at @xmath24 as @xmath177 for a fixed @xmath24 , @xmath178 is the optimal solution set of problem , therefore , is also well defined due to theorem [ thm : exist ] .
then , our task in this section is establishing the properties of functions @xmath168 and @xmath171 as well as properties of mappings @xmath175 and @xmath179 .
this subsection is devoted to the stability for problem .
we explore the behaviors of the marginal function @xmath168 and optimal solution set @xmath175 with respect to @xmath180 .
we begin with a lemma which is crucial in our discussion .
[ lemma : lambdastrict ] given @xmath45 satisfying h[hypo:1 ] and @xmath53 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] . for any @xmath181 , @xmath97 , let @xmath182 .
then the following hold : * @xmath80 has full column rank . * if @xmath183 , then @xmath71 is the unique minimizer of @xmath184 , that is for any @xmath185 , @xmath186 .
we first prove item ( i ) .
we prove it by contradiction .
if not , there exists @xmath187 such that @xmath188 and @xmath189 such that @xmath190 .
therefore , @xmath191 .
this contradicts the definition of @xmath192
. then we get item ( i ) .
we next prove item ( ii ) .
since @xmath183 and @xmath45 is nondecreasing , it follows that @xmath193 .
thus , @xmath194 for any @xmath181 .
the problem @xmath184 is equivalent to the problem @xmath195 then from item ( i ) , the objective function of the convex optimization problem is strictly convex .
therefore , we obtain item ( ii ) .
we next show an important proposition which reveals the relationship between @xmath196 and the optimal solution set of problem . clearly , when @xmath197 , @xmath198 .
thus , we only discuss cases as @xmath199 .
[ prop : partialeauvalence ] given @xmath45 satisfying h[hypo:1 ] and @xmath53 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] .
let @xmath175 , @xmath48 and @xmath200 be defined by , and respectively .
suppose @xmath45 is strictly increasing .
then , the following statements hold : * for any @xmath201 , there must exist @xmath202 such that @xmath203 , and @xmath204 is a subset of @xmath205 , that is @xmath206 moreover , @xmath207 for any @xmath208 .
* particularly , if the @xmath51 in item ( i ) satisfies @xmath209 , then @xmath205 is equal to @xmath204 , that is @xmath210 * further , if @xmath37 and the @xmath51 in item ( i ) satisfies @xmath211 , then @xmath204 is a real subset of @xmath205 , that is @xmath212 we first prove item ( i ) .
since @xmath45 is strictly increasing , @xmath213 and @xmath214 , there must exist @xmath202 such that @xmath203 due to item ( iii ) of proposition [ prop : rhosomega ] .
we then prove @xmath215 .
let @xmath216 , we will show @xmath217 .
since @xmath218 , @xmath219 and @xmath45 is strictly increasing , we have @xmath220 . in order to prove @xmath217
, we need to show that for any @xmath221 there holds @xmath222 .
if not , there exists @xmath221 such that @xmath223 . by the definition of @xmath224 , we have @xmath225 , which contradicts the fact that @xmath226 .
therefore , @xmath216 implies @xmath217 .
moreover , since @xmath227 we have @xmath228 , where @xmath168 is the marginal function of problem , defined by . therefore , we have @xmath229 for any @xmath208 .
we then get item ( i ) .
next , we try to prove item ( ii ) . by item ( i ) of this proposition , we have @xmath207 for any @xmath208 . in order to prove , we only need to show @xmath230 for any @xmath208 due to item ( v ) of proposition [ prop : rhosomega ] .
we next show it by contradiction .
if there exists @xmath231 such that @xmath232 , then @xmath233 due to @xmath234 , the strictly increasing property of @xmath45 and @xmath209 .
since @xmath235 , by the definition of @xmath236 , we have @xmath237 , contradicting .
thus we get item ( ii ) of this proposition .
finally , we prove item ( iii ) .
we only need to show that there exists @xmath208 such that @xmath238 .
let @xmath216 , we have @xmath239 .
let @xmath182 .
since @xmath45 is continuous and @xmath240 , there exists @xmath241 such that as long as @xmath242 , @xmath243 .
let @xmath244 .
then @xmath245 . set @xmath246 .
set @xmath247 .
obviously , @xmath248 , @xmath249 and @xmath250 .
since @xmath45 is strictly increasing , we have @xmath221 . by item ( i ) of this proposition , @xmath208 due to @xmath207 and @xmath221 .
we then show @xmath238 .
it amounts to showing @xmath251 by item ( v ) of proposition [ prop : rhosomega ] . by item ( ii ) of lemma [ lemma : lambdastrict ] , we have @xmath252 since @xmath253 and @xmath7 share the same support @xmath70 .
then by the strictly increasing property of @xmath45 we get @xmath238 .
then item ( iii ) of this proposition follows .
[ remark : same ] one can easily check that definition [ def : srhoomega ] produces the same @xmath54 and @xmath254 for @xmath84 for different choices of @xmath45 once they are all strictly increasing .
therefore , it is reasonable that @xmath205 has connections with @xmath196 for @xmath84 , as shown in proposition [ prop : partialeauvalence ] .
now , we are ready to establish the main results of this subsection .
[ thm : staconst ] given @xmath255 defined at @xmath256 as @xmath257 and @xmath53 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] .
let @xmath168 and @xmath175 be defined by and respectively .
then the following statements hold : * @xmath258 * @xmath259 in particular , if @xmath37 , we have @xmath260 for @xmath261 , @xmath262 . *
@xmath168 is a piecewise constant function and has at most @xmath263 values .
further , @xmath168 is lower semicontinuous , right - continuous and nonincreasing . *
@xmath264 for any @xmath265 satisfying @xmath266 , @xmath262 . items ( i ) and ( ii ) follow from proposition [ prop : partialeauvalence ] .
then item ( iii ) of this theorem follows immediately . by item ( i ) and @xmath267 for any @xmath265 satisfying @xmath266
, we obtain item ( iv ) .
this subsection is for the stability of problem .
we shall study the behaviors of the marginal function @xmath171 and optimal solution set @xmath268 with respect to @xmath25 .
according to theorem [ thm : exist ] , @xmath269 with @xmath270 defined by .
obviously , each @xmath270 for @xmath97 is a line with slop @xmath271 and intercept @xmath192 .
therefore , by items ( ii ) and ( iii ) of proposition [ prop : rhosomega ] , it is easy to deduce that @xmath171 is continuous , piecewise linear and nondecreasing .
next we propose an iteration procedure to find the minimal value of @xmath270 for @xmath97 , i.e. , the marginal function @xmath272 , for @xmath273 .
[ def : lambda_i ] given @xmath45 satisfying h[hypo:1 ] and @xmath53 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] .
then the integer @xmath274 , @xmath275 are defined by the following iteration @xmath276 we present an example to illustrate the iterative procedure in definition [ def : lambda_i ] . [ exam:3.6 ] in this example , we set @xmath277 .
by proposition , we set @xmath278 and @xmath279
. then @xmath270 for @xmath280 can be obtained by equation [ def : fi ] .
we depict the plots of @xmath270 for @xmath280 in figure [ fig : ex1 ] .
we observe from figure [ fig : ex1 ] that , when @xmath281 , @xmath282 .
further , the first and largest critical parameter @xmath283 can be obtained by the largest intersection point of @xmath284 and @xmath270 for all @xmath285 . here
, we have that @xmath286 , which collects all the indexes that @xmath287 except @xmath288 . in order to search for the second critical parameter value @xmath289
, we first find that on the second interval @xmath290 , @xmath291 , where @xmath292 .
thus , @xmath289 can be obtained by the largest intersection point of @xmath293 and @xmath294 for @xmath295 . then @xmath296 .
thus , in this case , @xmath297 , @xmath298 . on interval @xmath299 , we have @xmath300 . in this example , @xmath301 for any @xmath24 . [
fig : ex1 ] the following proposition provides some basic properties of @xmath274 , @xmath302 and @xmath303 for @xmath304 by definition [ def : lambda_i ] .
[ prop : lambda_i ] given @xmath45 satisfying h[hypo:1 ] and @xmath53 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] .
let @xmath274 , @xmath302 and @xmath303 for @xmath304 be defined by definition [ def : lambda_i ] . then the following statements hold : * @xmath305 , in particular , if @xmath306 then @xmath307 .
* @xmath308 .
* @xmath309 . *
@xmath310 and @xmath311 , for all @xmath94 , @xmath312 .
the proof is outlined in appendix 8.1 . the main results of this subsection are presented in the following theorem . [ thm : stability ] given @xmath45 satisfying h[hypo:1 ] and @xmath53 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] .
let @xmath274 , @xmath302 and @xmath303 for @xmath304 be defined by definition [ def : lambda_i ] .
let @xmath270 for @xmath97 , @xmath171 and @xmath268 be defined by , and respectively . by remark [ remark : l=0 ] , if @xmath157 , then @xmath313 and @xmath314 . if @xmath306 , then the following statements hold : * @xmath315,\\ f_{t_i } , & \mathrm{~~if~~}\lambda\in ( \lambda_i , \lambda_{i-1 } ] , i\in\mathbb{n}_{k-1},\\ f_0(\lambda ) , & \mathrm{~~if~~}\lambda\in(\lambda_0 , + \infty ) .
\end{cases}\ ] ] * @xmath316 * @xmath171 is continuous , piecewise linear and nondecreasing , particularly , @xmath171 is strictly increasing in @xmath317 . to improve readability , the proof of theorem [ thm : stability ] and two auxiliary lemma and proposition are given in appendix 8.2 .
a direct consequence of theorem [ thm : stability ] is stated below .
[ crol : stability ] under the assumptions of theorem [ thm : stability ] , the following statements hold : * if @xmath318 or @xmath319 , @xmath320 , then @xmath321 and @xmath322 hold for any @xmath323 and any @xmath324 . * if @xmath325 , then @xmath222 and @xmath326 hold for any @xmath323 and any @xmath324 . item ( i ) can be obtained by item ( ii ) of theorem [ thm : stability ] as well as item ( v ) of proposition [ prop : rhosomega ] .
similarly , item ( ii ) follows from item ( ii ) of theorem [ thm : stability ] , items ( ii ) , ( iii ) , ( v ) of proposition [ prop : rhosomega ] as well as item ( ii ) of proposition [ prop : lambda_i ] . from theorem [ thm : stability ] , the optimal value of problem changes piecewise linearly while the optimal solution set of problem changes piecewise constantly as the parameter @xmath25 varies .
in addition , by corollary [ crol : stability ] , the optimal values of both the first and second terms of are piecewise constant with respect to changes in the parameter @xmath25 . we observe from theorem [ thm : staconst ] and [ thm : stability ] that , when @xmath197 , problem share the same optimal solution set @xmath161 as problem if @xmath25 is chosen to be small enough .
thus , in the remaining part of this paper , we only focus on the nontrivial case when @xmath327 , where @xmath200 is defined by .
in this section we explore the relationship between the constrained problem with @xmath328 and a corresponding special penalty problem . in the case of @xmath37
we shall study the relationship between problem and the least square penalty problem . in the case of @xmath29
, we shall discuss the connections between problem and the following @xmath22-penalty problem @xmath329 in fact , both problems and can be cast into the general penalty problem with @xmath45 defined at @xmath256 , for @xmath53 , as @xmath330 we exhibit the main results of this section in the following theorem .
[ thm : leasequv ] for @xmath45 defined by and @xmath328 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] .
let @xmath274 , @xmath302 and @xmath303 for @xmath304 be defined by definition [ def : lambda_i ] .
let @xmath331 and @xmath332 .
let @xmath200 , @xmath268 and @xmath175 be defined by , and respectively . for @xmath333 ,
let @xmath334 be the integer such that @xmath335 . * if @xmath336 and suppose @xmath337 , then @xmath338 further , if @xmath339 , we have @xmath340 for @xmath341 and @xmath342 for @xmath343 . * if @xmath344 , and suppose @xmath345 , then @xmath346 * if @xmath347 , then @xmath348 for any @xmath24 .
we first prove item ( i ) . from item ( ii ) of theorem
[ thm : stability ] , we have @xmath349 for @xmath341 . by item ( i ) of proposition [ prop : partialeauvalence ] , we get that @xmath350 as @xmath341 .
we also obtain that if @xmath339 , @xmath351 for @xmath341 from item ( ii ) of proposition [ prop : partialeauvalence ] . when @xmath352 or @xmath353 , by item ( ii ) of theorem [ thm : stability ] and @xmath354 , we have @xmath355 and @xmath356 for any @xmath357 .
since @xmath358 and @xmath227 for all @xmath359 , we get @xmath360 for @xmath352 or @xmath353 . clearly , if @xmath339 , then @xmath361 for @xmath352 or @xmath353 . when @xmath362 $ ] , @xmath356 for @xmath363 and thus @xmath364 .
we then prove item ( ii ) .
by item ( i ) of proposition [ prop : partialeauvalence ] , @xmath227 for any @xmath217 . from item ( ii ) of theorem [ thm : stability ] , we have @xmath365 due to @xmath345
. then we get item ( ii ) by the fact that @xmath356 for any @xmath366 and for any @xmath367 as @xmath368 .
we finally prove item ( iii ) .
similarly , we have @xmath227 for any @xmath217 .
however , by theorem [ thm : stability ] @xmath356 for any @xmath363 and @xmath24 since @xmath369 .
thus we get item ( iii ) .
[ remark : p=2 ] in the case of @xmath37 , by proposition [ prop : partialeauvalence ] , if @xmath211 and @xmath336 , then we have @xmath370 when @xmath341 .
if @xmath371 , then we have @xmath336 . in such case , item ( i ) of theorem [ thm : leasequv ] holds clearly . according to item ( iii ) of theorem [ thm : leasequv ] , in general
it is possible that problem has * no * common optimal solutions with problem for any @xmath24 .
this means that in general problem may not be an exact penalty method for problem .
however , in the noiseless case , we show in the next corollary that problem has the same optimal solution set as problem if @xmath25 is large enough .
[ coro : least ] let @xmath45 be defined by and @xmath328 . assume the feasible region of problem is nonempty .
then , there exists @xmath47 such that problems and with @xmath46 have the same optimal solution set .
since the feasible region of problem is nonempty , @xmath372 by definition [ def : srhoomega ] . clearly ,
problem is exactly problem with @xmath18 .
then , @xmath373 in this case .
let @xmath274 and @xmath374 , @xmath375 for @xmath376 be defined by definition [ def : lambda_i ] . from item ( ii ) of proposition
[ prop : lambda_i ] , @xmath288 and thus @xmath377 .
according to item ( i ) of theorem [ thm : leasequv ] , problems and have the same optimal solution set when @xmath378 .
set @xmath379 , we then obtain the desired results .
in this section , we first establish exact penalty conditions under which problem have the same optimal solution set as problem , provided that @xmath46 for some @xmath47 . based on the conditions developed , we propose several exact penalty formulations for problem . according to the definition of @xmath139 and @xmath133 in definition [ def : srhoomega ]
, the penalty term @xmath380 attains its minimal value at any @xmath137 .
we also know that @xmath381 attains its minimal value at any @xmath363 when @xmath281 by item ( ii ) of theorem [ thm : stability ] , where @xmath283 is defined by definition [ def : lambda_i ] .
since @xmath45 is nondecreasing , we have @xmath382 for any @xmath220
. then we obtain the following theorem .
[ thm : equivalence ] let @xmath45 satisfy h[hypo:1 ] and @xmath53 .
suppose the feasible region of problem is nonempty , that is @xmath50 , where @xmath48 is defined by .
if @xmath383 then there exists @xmath47 such that problems and with @xmath46 have the same optimal solution set . since holds , by the definition of @xmath139 and @xmath133 in definition [ def : srhoomega ] , @xmath384 .
thus @xmath133 is the optimal solution set of problem .
let @xmath379 , where @xmath283 is defined in definition [ def : lambda_i ] . then , from item ( ii ) of theorem [ thm : stability ] , @xmath133 is also the optimal solution set of problem with @xmath46 .
then we have that problems and with @xmath46 have the same optimal solution set . in general
, it is not easy to verify condition .
the next corollary states a simple and useful condition which also ensures the exact penalty property once @xmath25 is large enough .
[ crol : equv ] let @xmath45 satisfy h[hypo:1 ] and @xmath53 .
assume that the feasible region of problem is nonempty .
if @xmath45 satisfies @xmath385 then there exists @xmath47 such that problems and with @xmath46 have the same optimal solution set .
the desired results follows from the fact that condition implies condition .
next we present several examples of @xmath45 satisfying condition . for @xmath386 ,
we denote by @xmath387 the positive part of @xmath388 , that is , @xmath389 given @xmath328 .
let scalar function @xmath390 be defined at @xmath256 as @xmath391 .
then @xmath392 satisfies h1 and condition . with
the choice of @xmath393 , the penalty term in problem becomes @xmath394 the penalty term with @xmath37 is well exploited in penalty method for quadratically constrained @xmath395 minimization with @xmath21 in @xcite .
one can easily check that the penalty term is convex but non - differentiable for @xmath328 . given @xmath328 .
let @xmath396 be defined at @xmath256 as @xmath397 then @xmath398 satisfies h1 and condition .
with the choice of @xmath399 , the penalty term in problem becomes @xmath400 by choosing @xmath399 , for @xmath37 problem reads as @xmath401 according to corollary [ crol : equv ] , the following proposition concerning the exact penalization of problem holds .
[ prop : env ] suppose that the feasible region of problem is nonempty , then the penalty problem shares the same optimal solution set with the constrained problem , provided that @xmath46 for some @xmath47 .
in particular , when @xmath18 , problem reduces to the least square penalty problem . from proposition
[ prop : env ] , we deduce that has the same optimal solution set as the linear constrained problem when @xmath25 is sufficiently large .
this result coincides with that obtained in corollary [ coro : least ] .
in addition to the advantage of exact penalization , we prove in the next proposition that the penalty term in is differentiable with a lipschitz continuous gradient . for simplicity
, we define @xmath402 at @xmath2 as @xmath403 we denote by @xmath404 the largest singular value of @xmath405 . we also need to recall the notion of firmly nonexpansive .
operator @xmath406 is called firmly nonexpansive ( resp . ,
nonexpansive ) if for all @xmath407 @xmath408 obviously , a firmly nonexpansive operator is nonexpansive .
now , we are ready to show @xmath409 is differentiable with a lipshitz continuous gradient . [ prop : diff ] let @xmath402 be defined by .
then function @xmath409 is differentiable with a lipschitz continuous gradient . in particular , for @xmath2 * @xmath410 * @xmath411 is lipschitz continuous with the lipschitz constant @xmath412 .
we define @xmath413 , at @xmath414 as @xmath415 obviously , @xmath416 for any @xmath2 .
one can check that for any @xmath414 , @xmath417 thus @xmath418 and item ( i ) follows .
we next prove item ( ii ) .
we first show @xmath419 is nonexpansive .
let @xmath420 be the projection operator onto the set @xmath421 .
we have @xmath422 . since the set @xmath421 is nonempty closed convex , operator @xmath423 is firmly nonexpansive , by proposition 12.27 in @xcite .
thus , @xmath419 is nonexpansive . for any @xmath424
, we have that @xmath425 the above inequality leads to item ( ii ) .
as mentioned in the introduction , the differentiability of the penalty term has important effect on the algorithmic design for problem .
according to proposition [ prop : diff ] , by choosing @xmath399 , the penalty term is differentiable with a lipschitz continuous gradient in problem for @xmath37 . as a result
, we are able to develop efficient numerical algorithms with theoretical convergence guarantee for solving this problem , e.g. , proximal - gradient type algorithms .
in this section , we investigate the cardinality of optimal solution sets of problems and .
we also discuss the strictness of their optimal solutions as a byproduct . as it is shown in section [ sec : sta ] , optimal solution sets of these two problems are closely related to @xmath204 , for @xmath426 , defined by definition [ def : srhoomega ] .
therefore , we shall first consider the cardinality of @xmath204 .
the cases of @xmath29 and @xmath37 are discussed in separate subsections , since the analysis and results are different for these two cases . for the case of @xmath37 , we have the following proposition concerning the cardinality of @xmath204 .
[ prop : cardip2 ] for @xmath45 satisfying h[hypo:1 ] and @xmath37 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] . for @xmath427 , if @xmath428 , then we have @xmath429 by the definition of @xmath204 , we have @xmath227 for any @xmath216 .
let @xmath430 .
we will first prove that for any @xmath431 , @xmath432 , where @xmath433 . for @xmath431 ,
if there doest not exist @xmath216 such that @xmath434 , then we have @xmath435 , that is @xmath436 . otherwise , if there exists @xmath216 such that @xmath434 , we then prove @xmath437 . by the definition of @xmath204 we have @xmath438 for any @xmath439 . by item ( ii ) of lemma [ lemma : lambdastrict ]
, we have that @xmath440 for all @xmath441 .
then , we get @xmath442 for all @xmath441 due to @xmath443 .
thus , @xmath444 , that is @xmath445 .
therefore , we get @xmath432 for any @xmath431 .
further , since @xmath446 and @xmath447 , we get of this proposition .
the next corollary is obtained immediately .
[ coro : cardip2 ] for @xmath45 satisfying h[hypo:1 ] and @xmath37 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] .
if @xmath45 is strictly increasing , then @xmath448 for any @xmath427 .
we first present an important lemma . to this end , we define , for any @xmath449 and @xmath450 , @xmath451 . for convenience ,
we adopt an assumption for a vector in @xmath452 .
[ hypo:2 ] for a vector @xmath453 , there holds @xmath454 for all @xmath455 .
[ lema : l1 ] let @xmath67 an @xmath456 real matrix and @xmath5 .
let @xmath457 satisfy @xmath458 . then * if @xmath459 and @xmath67 satisfies h[hypo:2 ] , then @xmath460 , where @xmath461 ; * if @xmath462 and @xmath463 , then for any @xmath450 , there exists @xmath464 such that @xmath465 . for simplicity , we set @xmath466 .
we first prove item ( i ) .
let @xmath467 .
obviously @xmath468 . then for any @xmath469 , we set @xmath470 .
it is clear that @xmath471 due to h[hypo:2 ] .
then we have @xmath472
. therefore , @xmath460 .
we get item ( i ) .
we next prove item ( ii ) . since @xmath463 , we set @xmath473 , where @xmath474 is the component - wise signum function ( @xmath66 is assigned to @xmath66 ) . set @xmath475 .
we have @xmath476 .
then there exists @xmath241 such that for any @xmath477 , there holds @xmath478 for any @xmath479 .
let @xmath480 and @xmath481 . clearly , @xmath482 .
therefore , we have @xmath483 .
obviously , @xmath484 is an euclid linear space with dimension no less then @xmath485 .
thus , for any @xmath450 , there exists @xmath486 such that @xmath487 .
set @xmath488 .
then , we have @xmath489 , @xmath478 for @xmath479 and @xmath490 .
thus , we have @xmath491 . then , we get this lemma . now , we are ready to present a proposition on the cardinality of @xmath204 for @xmath29 . [ prop : cardip1 ] for @xmath45 satisfying h[hypo:1 ] and @xmath29 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] . for @xmath427 ,
the following statements hold : * for @xmath492 , @xmath493 . * for @xmath494 , if @xmath495 , all the columns of @xmath3 satisfy h[hypo:2 ] and @xmath496 , then @xmath497 . * for @xmath494 with @xmath498 or @xmath499 ,
if there exists @xmath216 such that @xmath500 , then for any @xmath450 there exists @xmath501 such that @xmath439 , therefore , @xmath502 . item ( i ) follows from the fact that @xmath503 .
we next prove item ( ii ) . by the definition of @xmath504 and @xmath495
, we have @xmath505 and @xmath506 for any @xmath507 . since @xmath508
, then we assume @xmath509 satisfies @xmath510 .
let @xmath511 for @xmath106 .
then @xmath512 . by item ( i ) of lemma [ lema : l1 ] , we have @xmath513 for @xmath106
. then we get item ( ii ) of this proposition .
we finally prove item ( iii ) .
let @xmath182 and @xmath514 .
since @xmath216 , we have @xmath227 and @xmath515 . set @xmath516 . for any @xmath450 , set @xmath517 .
then by item ( ii ) of lemma [ lema : l1 ] , there exists @xmath518 such that @xmath519 and @xmath520 .
then we have @xmath521 and @xmath522 by the definition of @xmath204 .
thus @xmath523 . item ( iii ) follows .
this subsection is devoted to the cardinality of the optimal solution set of problem when @xmath37 .
we begin with recalling the notion of strict minimizers . for a function @xmath524 and a set @xmath525 ,
@xmath526 is a strict minimizer of the problem @xmath527 if there exists @xmath450 such that for any @xmath528 there holds @xmath529 .
it is clear that as @xmath530 , @xmath531 and @xmath64 is the unique minimizer of problem .
we next only consider the case when @xmath532 .
[ thm : strictp2 ] for @xmath533 defined at @xmath39 as @xmath257 and @xmath37 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] .
let @xmath534 , where @xmath200 is defined by .
let @xmath535 be defined by .
then the following statements are equivalent : * there exits @xmath202 such that @xmath536 .
* @xmath537 for any @xmath217 . *
any @xmath359 is a strict minimizer of problem .
* @xmath538 is finite .
we first prove item ( i ) implies item ( iv ) . by item ( ii ) of theorem [ thm : staconst ] , we have @xmath539 since @xmath536 .
then we have @xmath538 is finite due to corollary [ coro : cardip2 ] .
thus , we get item ( iv ) .
it is obvious that item ( iv ) implies item ( iii ) .
we next prove item ( iii ) implies item ( ii ) by contradiction .
suppose there exists @xmath540 such that @xmath541 .
we will show @xmath146 is not a strict minimizer of problem .
let @xmath542 .
then there exists @xmath241 such that for any @xmath543 there holds @xmath544 .
let @xmath545 .
set @xmath546 .
then for any @xmath547 we have @xmath548 for @xmath479 and @xmath549 .
thus , @xmath550 for any @xmath551 . clearly , @xmath552 .
then , for any @xmath450 , there exists @xmath553 , such that @xmath554 , contradicting the fact that @xmath146 is a strict minimizer of problem .
therefore , we obtain that item ( iii ) implies item ( ii ) .
we finally show that item ( ii ) implies item ( i ) .
we also prove it by contradiction .
suppose item ( i ) doest not hold .
then by item ( i ) of proposition [ prop : partialeauvalence ] , there exists @xmath555 , where @xmath556 and @xmath557 due to item ( iii ) of proposition [ prop : partialeauvalence ] . by item ( ii ) , @xmath558 for any @xmath559 , contradicting the fact that @xmath560 due to the definition of @xmath561 .
thus we get item ( ii ) implies item ( i )
. then we complete the proof .
the next corollary follows from theorem [ thm : strictp2 ] . for @xmath533 defined at @xmath39 as @xmath257 and @xmath37 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] . for @xmath562 , @xmath334 , the following hold : * there exits @xmath217 such that @xmath7 is not a strict minimizer of problem .
* @xmath563 . in this subsection , we discuss cardinality of optimal solution set of problem when @xmath37 .
we state the main results in the following theorem .
[ thm : penacarp2 ] for @xmath45 satisfying h[hypo:1 ] and @xmath37 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] .
let @xmath274 , @xmath302 and @xmath303 for @xmath304 be defined by definition [ def : lambda_i ] .
let @xmath564 , @xmath565 for @xmath97 and @xmath268 defined by .
then for @xmath566 , the following statements hold : * if @xmath567 , then for any @xmath568 , @xmath569 is finite and any @xmath363 is a strict minimizer of problem . * if @xmath570 and @xmath571 for all @xmath572 , then @xmath573 is finite and any @xmath574 is a strict minimizer of problem .
we first prove item ( i ) .
since @xmath567 , @xmath575 is finite . by ( ii ) of theorem [ thm : stability ] , we have @xmath576 for any @xmath577
. then we get item ( i ) .
we then prove item ( ii ) .
since @xmath571 for all @xmath578 , @xmath579 is finite for all @xmath578 . then , by item ( ii ) of theorem [ thm : stability ] , @xmath573 is finite
. then we get item ( ii ) of this theorem .
a direct and useful consequence of theorem [ thm : penacarp2 ] is presented below .
[ coro : penacarp2 ] let @xmath45 satisfy h1 and @xmath37 .
if @xmath45 is strictly increasing , then @xmath580 is finite and every @xmath363 is a strict minimizer of problem for any @xmath24 . with the help of the above corollary
, we characterize the cardinality of optimal solution set of problem in the next proposition .
our result coincides with that obtained in @xcite .
the cardinality of the optimal solution set of problem is finite for any @xmath24 and every optimal solution is a strict minimizer .
problem is a special case of problem with @xmath581 for @xmath39 . since in this case
@xmath45 is strictly increasing , we obtain the conclusions by corollary [ coro : penacarp2 ] .
this subsection provides results on cardinality of optimal solution sets of problems and when @xmath29 .
[ thm : probcarp1 ] for @xmath45 satisfying h[hypo:1 ] and @xmath29 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] .
let @xmath274 , @xmath302 and @xmath303 for @xmath304 be defined by definition [ def : lambda_i ] .
let @xmath564 .
let @xmath268 and @xmath535 be defined by and respectively .
then the following statements hold : * @xmath582 for @xmath583 and @xmath584 . * for @xmath585 with @xmath586 or @xmath587 , if there exists @xmath588 such that @xmath589 , then we have @xmath590 for any @xmath591 and @xmath146 is not a strict minimizer of problem .
* suppose @xmath257 for @xmath39 , then @xmath592 for all @xmath562 , @xmath334 ; further , for @xmath593 and @xmath51 satisfying @xmath585 with @xmath594 or @xmath587 , if there exists @xmath595 such that @xmath596 , then we have @xmath597 , and @xmath146 is not a strict minimizer of problem . by item ( ii ) of theorem [ thm : stability ] and
the definition of @xmath598 , item ( i ) follows immediately . by theorem [ thm : stability ] , @xmath599 for @xmath591
. then we have item ( ii ) due to item ( iii ) of proposition [ prop : cardip1 ] . for @xmath562 , @xmath334 , from item ( ii ) of theorem [ thm : staconst ] , @xmath600 . then any @xmath601 satisfies @xmath602
let @xmath603 .
similar to the proofs of theorem [ thm : strictp2 ] , for any @xmath450 , there exits @xmath604 such that @xmath605 and @xmath606 .
thus , @xmath607 . therefore , @xmath597 for @xmath562 . for @xmath593 , we have @xmath608 .
then by item ( iii ) of proposition [ prop : cardip1 ] we obtain item ( iii ) of this theorem .
in this section , we provide numerical illustrations for some of our main theoretical findings , including : * the marginal function of problem is piecewise constant , while that of problem is piecewise linear ; * it is possible that the least square penalty problem has no common optimal solutions with the constrained problem for any @xmath24 ; * there exists @xmath47 such that problem is an exact penalty formulation of the constrained problem for @xmath46 .
all the @xmath0 optimization problems involved are solved by an exhaustive combinational search . for better readability ,
all numerical digits are round to four decimal places .
we present in detail a concrete example for sparse signal recovery with @xmath609 as follows @xmath610,\\ \label{eq : x } x^*=\left [ \begin{array}{ccccc } 0 & 1 & 1 & 0 & 0\\ \end{array } \right]^\top,\\ \label{eq : b } b = ax+\eta= \left[\begin{array}{ccccc } 14.43 & 7.21 & 4.49 & 13.02 \end{array}\right]^\top,\end{aligned}\ ] ] where @xmath611 denotes random noise following a nearly normal distribution .
the coefficients of @xmath3 , @xmath146 , and @xmath612 in , , and are exact .
the noise level of @xmath13 is @xmath613 .
as @xmath146 is sparse , we naturally expect recovering @xmath146 by solving problem with @xmath3 , @xmath612 in , and @xmath614 . to this end
, we shall first verify that @xmath146 is an optimal solution to problem .
let @xmath257 , @xmath615 and @xmath37 . then according to definition [ def : srhoomega ]
, we have that @xmath616 , @xmath617 and @xmath618 by theorem [ thm : staconst ] , the marginal function of problem has the form of @xmath619 specially , we find that @xmath620 when @xmath621 .
this together with @xmath622 implies that @xmath146 is an optimal solution to problem with @xmath614 .
next we consider the least square penalty problem with @xmath3 , @xmath612 in and . by remark [
remark : same ] , in this case the marginal function of problem is @xmath623 where @xmath277 , @xmath192 , @xmath271 , @xmath624 are given in and . then , @xmath274 , @xmath303 for @xmath376 , and @xmath625 for @xmath626 are produced by the iterative procedure in definition [ def : lambda_i ] .
we obtain that @xmath627 , @xmath628 @xmath629 @xmath630 the plots of @xmath631 for @xmath280 and their intersection points are depicted in figure [ fig : ex2 ] . by theorem [ thm : stability ] , @xmath272 in reads as @xmath632 we note that @xmath633 for any @xmath24 . by theorem [ thm : stability ]
, we deduce that @xmath634 for any @xmath363 and any @xmath24 .
therefore , given @xmath3 , @xmath612 in , , problem has no common optimal solution as problem with @xmath614 for any @xmath24 .
in other words , problem fails to recover or approximate the true signal @xmath146 . in the next numerical test ,
we discuss penalty problem with @xmath3 , @xmath612 in , and @xmath614 . as shown in section [ sec : equal ] , problem has the same optimal solution set with problem , provided that @xmath46 for some @xmath47 .
we shall find the exact numerical value of @xmath635 .
set @xmath399 as in . from the iteration in definition [ def : srhoomega ]
, we have that @xmath636 , @xmath637 @xmath638 according to the proof of theorem [ thm : equivalence ] , @xmath635 actually equals to @xmath639 produced by the iteration procedure in definition [ def : lambda_i ] . given @xmath640 , @xmath641 in , , @xmath642
, it holds that @xmath643 therefore , we conclude that for @xmath3 , @xmath612 in , and @xmath614 , penalty problem share the same optimal solution set as constrained problem when the penalty parameter @xmath644 .
the quadratically constrained @xmath0 regularization problem arises in many applications where sparse solutions are desired .
the least square penalty method has been widely used for solving this problem .
there is little results regarding properties of optimal solutions to problems and , although these two problems have been well studied mainly from a numerical perspective . in this paper
, we aim at investigating optimal solutions of a more general constrained @xmath0 regularization problem and the corresponding penalty problem .
we first discuss the optimal solutions of these two problems in detail , including existence , stability with respect to parameter , cardinality and strictness .
in particular , we show that the optimal solution set of the penalty problem is piecewise constant with respect to the penalty parameter .
then we clarify the relationship between optimal solutions of the two problems .
it is proven that , when the noise level @xmath44 there does not always exist a penalty parameter , such that problem and the constrained @xmath0 problem share a common optimal solution . under mild conditions on the penalty function , we prove that problem has the same optimal solution set with problem if the penalty parameter is large enough . with the help of the conditions , we further propose exact penalty problems for the constrained problem .
we expect our theoretical findings can offer motivations for the design of innovative numerical schemes for solving constrained @xmath0 regularization problems .
[ lemma : lam ] given @xmath45 satisfying h[hypo:1 ] and @xmath53 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] .
let @xmath274 , @xmath302 and @xmath303 for @xmath304 be defined by definition [ def : lambda_i ] . then @xmath645 conversely , for any @xmath652 , @xmath653 for @xmath650 . by the definition of @xmath375 , @xmath654 for @xmath650 and there exists @xmath655 such that @xmath656 .
thus , @xmath657 and @xmath658 for @xmath650 . by the definition of @xmath625
, one has @xmath646 .
we then prove item ( iii ) . from the definition of @xmath664 , @xmath665 for @xmath659 .
then by lemma [ lemma : lam ] , it follows that @xmath666 for any @xmath648 due to @xmath667
. then we have @xmath668 for @xmath648 .
according to the definition of @xmath669 , @xmath670 for @xmath659 . item ( iii ) follows immediately .
[ lema : lambdai ] given @xmath45 satisfying h[hypo:1 ] and @xmath53 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] .
let @xmath274 , @xmath302 and @xmath303 for @xmath304 be defined by definition [ def : lambda_i ] . for @xmath659 ,
next we prove item ( iii ) .
since @xmath681 , @xmath682 and @xmath683 , it holds that @xmath684 this means that holds for @xmath685 .
we assume for @xmath686 , holds when @xmath687 .
then we try to show holds as @xmath688 , that is @xmath689 one finds that @xmath690 then follows .
therefore , we obtain item ( iii ) immediately . next , we prove item ( iv ) . item ( ii ) implies that @xmath676 for all @xmath691 .
thus , we only need to prove @xmath678 for @xmath692 , which is equivalent to @xmath693 for @xmath692 . by item ( ii ) , for @xmath694 , it holds that @xmath695 for @xmath696 . from the fact that
@xmath697 and @xmath698 , we obtain @xmath699 holds for @xmath696 . by items ( i ) and
( iii ) , @xmath700 for @xmath694 .
this together with and item ( i ) implies that holds for @xmath679 .
[ prop : fi ] given @xmath45 satisfying h[hypo:1 ] and @xmath53 , let @xmath54 and @xmath83 for @xmath84 be defined by definition [ def : srhoomega ] .
let @xmath274 , @xmath302 and @xmath303 for @xmath304 be defined by definition [ def : lambda_i ] .
let @xmath270 for @xmath97 .
then the following statements hold : * for @xmath701 , @xmath702 for all @xmath703 .
* @xmath704 for @xmath701 . * @xmath705 for any @xmath378 and any @xmath706 . * @xmath707 for any @xmath708 and any @xmath709 . * for @xmath710 ,
there holds @xmath711 for any @xmath712 and any @xmath713 .
we first prove items ( i ) and ( ii ) . from the definition of @xmath714 and @xmath715 , for any @xmath703 @xmath716 . item ( i )
is obtained immediately by the definition of @xmath294 for @xmath717 in . item ( ii )
follows from @xmath718 .
we next prove item ( iii ) .
thanks to items ( ii ) and ( iii ) of proposition [ prop : rhosomega ] , item ( iii ) amounts to @xmath719 for all @xmath720 .
it is clear that holds by the definition of @xmath721 in definition [ def : lambda_i ] and @xmath288 . finally , we prove item ( v ) . by item ( iv ) of lemma [ lema : lambdai ] ,
@xmath725 for @xmath726 and @xmath727 .
thus , @xmath711 for any @xmath728 and any @xmath727 . by the definition of @xmath714
, we have @xmath729 for @xmath730 and @xmath731 .
then , it follows that @xmath711 for any @xmath732 and any @xmath733 . by item ( ii ) of theorem [ thm : exist ] and item ( i ) of proposition [ prop : fi ] , @xmath736 .
using this fact , it suffices to show @xmath737 for any @xmath738 . from items ( i ) and ( iv ) of lemma [ lema : lambdai ] , @xmath739 for @xmath740 . by the definition of @xmath625 and @xmath375
, we have @xmath741 for @xmath742 , which implies @xmath737 for @xmath742 .
hence @xmath737 for any @xmath738
. , _ convergence of descent methods for semi - algebraic and tame problems : proximal algorithms , forward - backward splitting , and regularized gauss - seidel methods _ , mathematical programming , 137 ( 2013 ) , pp .
91129 . height 2pt depth -1.6pt width 23pt , _ relationship between the optimal solutions of least squares regularized with @xmath0-norm and constrained by k - sparsity _ , applied and computational harmonic analysis , ( 2016 ) . | the constrained @xmath0 regularization plays an important role in sparse reconstruction .
a widely used approach for solving this problem is the penalty method , of which the least square penalty problem is a special case .
however , the connections between global minimizers of the constrained @xmath0 problem and its penalty problem have never been studied in a systematic way .
this work provides a comprehensive investigation on optimal solutions of these two problems and their connections .
we give detailed descriptions of optimal solutions of the two problems , including existence , stability with respect to the parameter , cardinality and strictness .
in particular , we find that the optimal solution set of the penalty problem is piecewise constant with respect to the penalty parameter
. then we analyze in - depth the relationship between optimal solutions of the two problems .
it is shown that , in the noisy case the least square penalty problem probably has no common optimal solutions with the constrained @xmath0 problem for any penalty parameter . under a mild condition on the penalty function
, we establish that the penalty problem has the same optimal solution set as the constrained @xmath0 problem when the penalty parameter is sufficiently large .
based on the conditions , we further propose exact penalty problems for the constrained @xmath0 problem .
finally , we present a numerical example to illustrate our main theoretical results . optimal solutions , constrained @xmath0 regularization , penalty methods , stability |
the problem of the lifetime of a metastable state has been addressed in a variety of areas , including first - order phase transitions , josephson junctions , field theory and chemical kinetics @xcite .
recent experimental and theoretical results show that long - live metastable states are observed in different areas of physics @xcite .
experimental and theoretical investigations have shown that the average escape time from metastable states in fluctuating potentials presents a nonmonotonic behavior as a function of the noise intensity with the presence of a maximum @xcite .
this is the noise enhanced stability ( nes ) phenomenon : the stability of metastable states can be enhanced and the average life time of the metastable state increases nonmonotonically with the noise intensity .
this resonance - like behavior contradicts the monotonic behavior of the kramers theory @xcite .
the occurrence of the enhancement of stability of metastable states by the noise has been observed in different physical and biological systems @xcite .
very recently nes effect was observed in an ecological system @xcite , an oscillator chemical system ( the belousov - zhabotinsky reaction ) @xcite and in magnetic systems @xcite .
interestingly in ref .
@xcite the stabilization of a metastable state due to noise is experimentally detected and a decreasing behavior of the maximum lyapunov exponent as a function of the noise intensity is observed .
a generalization of the lyapunov exponent for stochastic systems has been recently defined in ref .
@xcite to complement the analysis of the transient dynamics of metastable states .
this new measure of stability is the mean growth rate coefficient ( mgrc ) @xmath1 and it is evaluated by a similar procedure used for the calculation of the lyapunov exponent in stochastic systems @xcite . by linearizing the langevin equation of motion
( see next eq . [ eq ] ) , we consider the evolution of the separation @xmath2 between two neighboring trajectories of the brownian particle starting at @xmath3 and reaching @xmath4 @xmath5 and define @xmath6 as an instantaneous growth rate .
we note that , in eq .
( [ delta ] ) , @xmath7 is calculated onto the noisy trajectory @xmath8 $ ] @xcite . the growth rate coefficient @xmath9 ( for the @xmath10 noise realization ) , is then defined as the long - time average of the instantaneous @xmath11 coefficient over @xmath12 @xcite @xmath13 in the limit @xmath14 , eq .
( [ lyap ] ) coincides formally with the definition of the maximum lyapunov exponent , and therefore , the @xmath9 coefficient has the meaning of a finite - time lyapunov exponent .
this quantity is useful to characterize a transient dynamics in nonequilibrium dynamical systems @xcite .
the mean growth rate coefficient @xmath1 is then defined as the ensemble average of the growth rate coefficient @xmath9 @xmath15 over the noise realizations .
the mean growth rate coefficient has a nonmonotonic behavior as a function of the noise intensity for brownian particles starting from unstable initial positions @xcite .
this nonmonotonicity with a minimum indicates that @xmath1 can be used as a new suitable measure or signature of the nes effect . the inclusion of realistic noise sources , with a finite correlation time , impacts both the stationary and the dynamic features of nonlinear systems . for metastable thermal equilibrium systems
it has been demonstrated that colored thermal noise can substantially modify the crossing barrier process @xcite .
a rich and enormous literature on escape processes driven by colored noise was produced in the @xmath16 s @xcite .
more recently many papers investigated the role of the correlated noise on different physical systems @xcite , which indicates a renewed interest in the realistic noise source effects .
in this work we present a study of the average decay time of an overdamped brownian particle subject to a cubic potential with a metastable state .
we focus on the role of different unstable initial conditions and of colored noise in the average escape time .
the effect of the correlation time @xmath0 on the transient dynamics of the escape process is related to the characteristic time scale of the system , that is the relaxation time inside the metastable state @xmath17 . for @xmath18 , the dynamical regime of the brownian
particle is close to the white noise dynamics . for @xmath19
, we obtain : ( i ) a big shift of the increase of the average escape times towards higher noise intensities ; ( ii ) an enhancement of the value of the average escape time maximum with a broadening of the nes region in the plane ( @xmath20 ) , which becomes very large for high values of @xmath0 ; ( iii ) the shift of the peculiar initial position @xmath21 ( towards lower values ) , found in our previous studies @xcite , which separates the set of the initial unstable states producing divergency , for @xmath22 tending to zero , from those which give only a nonmonotonic behavior of the average escape time ; ( iv ) the entire qualitative behaviors ( i - iii ) can be applied to the standard deviation of the escape time ; ( v ) the shift of the minimum values in the curves of the mean growth rate coefficient @xmath1 ; ( vi ) trend to the disappearance of the minimum in the curves of @xmath1 , with a decreasing monotonic behavior for increasing @xmath0 ; ( vii ) trend to the disappearance of the divergent dynamical regime in @xmath23 , with increasing @xmath0 .
the paper is organized as follows . in the next section
we introduce the model . in the third section
we show the results and in the final section we draw the conclusions .
the starting point of our study is the langevin equation @xmath24 where @xmath25 is the ornstein - uhlenbeck process @xmath26 and @xmath27 is the wiener process with the usual statistical properties : @xmath28 and @xmath29 . the system of eqs .
( [ eq ] ) and ( [ eqou ] ) represents a two - dimensional markovian process , which is equivalent to a non - markovian langevin equation driven with additive gaussian correlated noise , with @xmath25 obeying the following statistical properties @xmath30 and @xmath31 , for @xmath32 and @xmath33 . here
@xmath34 is the correlation time of the process .
the integration of eq .
( [ eqou ] ) yields in the limit @xmath35 the white noise term @xmath36 and the stationary correlation function of the ornstein - uhlenbeck process gives in the limit @xmath35 the correlation function of the white noise : @xmath37 .
the potential @xmath38 used in eq .
( [ eq ] ) is @xmath39 , with @xmath40 , @xmath41 .
the potential profile has a local stable state at @xmath42 and an unstable state at @xmath43 ( see fig .
[ 1-fpot ] ) .
the relaxation time for the metastable state at @xmath42 is @xmath44_{x=0 } = 2 a$ ] , which is the characteristic time scale of our system . for our potential profile
we have @xmath45 .
as a function of noise intensity @xmath22 in the case of correlated noise with @xmath46 , for the four initial positions investigated ( see fig . [ 1-fpot ] ) .
inset : the related standard deviation as a function of the noise intensity @xmath22 .
the dotted straight line at @xmath47 separates the simulation data representing the brownian particles escaped within the maximum simulation time @xmath48 for @xmath49 , from those representing the particles partially trapped within the well for a time greater or equal to @xmath48 for @xmath50 .
panel b : mean growth rate coefficient @xmath1 as a function of the noise intensity @xmath22 , with the same initial positions of fig .
[ 1-fpot].,width=321 ] reaching the threshold position @xmath51 into the potential well , within the @xmath48 , as a function of noise intensity @xmath22 .
this threshold position @xmath52 corresponds to the flex point of the potential , where the instantaneous growth rate @xmath6 is equal to zero .
the correlation time of the noise is @xmath53 , with the same initial conditions of fig .
[ 1-fpot].,width=321 ] the calculations of the average escape time as a function of the colored noise intensity have been performed by averaging over * @xmath54 * realizations the numerical solution of the stochastic differential equation ( [ eq ] ) .
the absorbing boundary for the escape process is put on @xmath55 , and the maximum simulation time is @xmath56 .
for all the initial unstable states ( see fig .
[ 1-fpot ] ) and all the correlation times considered we find an enhancement of the mean first passage time ( mfpt ) @xmath23 with respect to the deterministic time . reaching the threshold position @xmath51 into the potential well , within the @xmath48 , as a function of noise intensity @xmath22 .
this threshold position @xmath52 corresponds to the flex point of the potential , where the instantaneous growth rate @xmath6 is equal to zero .
the correlation time of the noise is @xmath57 , with the same initial conditions of fig .
[ 1-fpot].,width=321 ] in fig .
[ 2-nes ] the calculations performed with low colored noise ( @xmath58 ) for the mean first passage time @xmath23 and the mean growth rate coefficient @xmath1 are shown .
we see as signatures of the nes effect a maximum in the curve of @xmath23 and a minimum in that of @xmath1 . in the inset of fig .
[ 2-nes]a the standard deviation of the first passage time as a function of noise intensity is reported .
we note that the behaviors of @xmath23 and @xmath1 in this low colored noise regime ( @xmath46 ) coincides with those obtained in the white noise case @xcite .
moreover by comparing the theoretical predictions of @xmath23 ( see eq .
( 3 ) of ref .
@xcite ) with direct numerical simulations of the langevin equation , a very good agreement is obtained ( see fig . 3 of ref .
@xcite ) . in fig .
[ 3-n_i ] the semi - log plots of the fraction of particles @xmath59 reaching the threshold position @xmath60 into the potential well , within the @xmath48 , as a function of noise intensity @xmath22 , with the same initial conditions of fig .
[ 1-fpot ] , are shown .
this threshold position @xmath52 corresponds to the concavity change of the potential and is considered for this reason as a reference indicator for the effective entrance of the particle into the well .
it is possible to observe that for very low noise intensity none particle enters into the well within the @xmath48 considered , and the estimation of the stability measures take their deterministic values .
we note that the behavior of the mean growth rate coefficient as a function of the noise intensity is strongly affected by the characteristic potential shape of a metastable state .
the curves shown in fig . [ 3-n_i ] clarify the behavior of @xmath1 in the limit of @xmath61 .
in fact the position @xmath51 is the flex point of the potential , where the instantaneous growth rate @xmath6 is equal to zero .
we see that for low noise intensities the fraction @xmath59 goes to zero , producing an increasing behavior of the mgrc ( see fig . [ 2-nes]b ) .
the behaviors of the mfpts as a function of the noise intensity @xmath22 with other values of @xmath0 are shown in fig .
[ 4-tau_c ] .
we clearly observe two dynamical regimes depending on the value of @xmath0 with respect to the relaxation time of the system ( @xmath62 ) : ( a ) weak colored noise ( @xmath63 ) and ( b ) strong colored noise ( @xmath64 ) . by observing fig .
[ 4-tau_c]a ( @xmath65 ) we can see that the qualitative behavior of mfpt shown in fig .
[ 2-nes]a is recovered . in the weak color noise regime
we can still observe the divergent behavior of mfpts for @xmath66 and a non monotonic behavior for @xmath67 , with @xmath68 . by increasing the value of the correlation time ( @xmath69 )
we observe a large displacement of the maximum of mfpt towards higher values of noise intensity and a shift of the peculiar initial position @xmath21 towards lower values . for @xmath70 , @xmath71 , and for @xmath72 , @xmath73 ( see figs . [ 4-tau_c]b and [ 4-tau_c]c ) , where @xmath74 is the peculiar initial position of the brownian particle in the presence of colored noise .
we note that @xmath68 is a fixed value for white noise case @xcite , while the position @xmath74 is a variable quantity for colored noise and it is depending on the value of the correlation time of the noise .
moreover , we observe a broadening of the nes region , which becomes very large for high values of the correlation time @xmath0 .
the nes region is the area where enhanced stability of a metastable state is observed . in other words
it is the area under each curve of @xmath23 _ vs _ @xmath22 ( see figs . [ 2-nes]a and [ 4-tau_c ] ) , where the values of @xmath23 are greater than the deterministic dynamical time related to the particular initial position investigated ( see also fig . 1 in _ mantegna and spagnolo _ , 1998 , ref
. @xcite ) .
the asymmetry of the potential profile with respect to the @xmath75 coordinate makes more effective the correlation of the noise for brownian particles moving from left to right .
this means that , at very low noise intensities of the colored noise , the particles inside the potential well will escape more easily with respect to the white noise case .
therefore , the trapping effect , which is responsible for the divergent behavior for any initial unstable state within the range @xmath76 will happen in a restricted range of initial positions , that is @xmath77 with @xmath78 .
specifically this peculiar position @xmath74 is shifted towards decreasing values of the @xmath75 coordinate for increasing correlation time @xmath0 of the noise source . in fig .
[ 2-nes]a and all panels of fig .
[ 4-tau_c ] the dotted straight line at @xmath47 separates the simulation data representing the brownian particles escaped within the maximum simulation time @xmath48 for @xmath49 , from those representing the particles partially trapped within the well for a time greater or equal to @xmath48 for @xmath50 .
this means that the simulation data obtained for @xmath50 underestimate the real data in the divergent dynamical regime .
in fact if we prolong the maximum simulation time @xmath48 we obtain more approximate values for @xmath23 and @xmath79 and the divergent behavior will be visible at lower noise intensities . as a consequence @xmath80 will be shifted towards lower values . for high values of the noise intensity
all the plots show a monotonic decrease behavior as a function of noise intensity collapsing in a unique curve .
moreover the slope of this limit curve becomes flatter by increasing the correlation time .
this means that the nes effect involves more and more orders of magnitude of the noise intensity .
the effect of the colored noise is therefore to delay the escape process or in other words to enhance more and more the stability of the metastable state for increasing values of the noise intensity . in fig .
[ 5-sigma ] the standard deviation @xmath79 of the first passage time distribution for @xmath57 is shown .
we see a huge increase of the @xmath79 for low values of noise intensity , demonstrating a strong enlargement of the distribution when the particle feels a noise intensity comparable with the height of potential barrier . similarly to the mfpts , _ color _ induces a shift in the divergent behavior of @xmath79 .
the relative measure of the width with respect to the mean value is shown in the inset of fig .
[ 5-sigma ] , where the ratio @xmath81 is plotted .
this ratio reveals a nonmonotonic behavior with a minimum , demonstrating the existence of a noise intensity for which the width of the first passage time distribution is the minimum related to its mean . in other words this value corresponds to a maximum of precision in the measure of @xmath23 .
this optimal noise intensity is shifted toward high noise values by increasing @xmath0 . the behavior of the mean growth rate coefficient @xmath1 as a function of the noise intensity @xmath22 for different values of the noise correlation time is shown in fig .
[ 6-lya_c ] . in the weak color noise regime
we observe a nonmonotonic behavior with a minimum for all the initial positions investigated with a shift in the position of the minimum towards higher noise intensities . in the strong color regime
the minimum , which represents a trapping phenomenon for a finite time , is visible for the divergent behavior of mfpts for @xmath82 and it is shifted towards higher noise intensities by increasing the correlation time . for initial positions @xmath83 ,
the minimum tends to disappear , but at the same time the @xmath1 parameter decreases monotonically with increasing noise intensity , showing a trapping phenomenon at higher noise intensities .
this trend to the disappearance of the minimum in the curves of @xmath1 , corresponds to the restricted range of the initial positions for which we observe a divergent behavior of @xmath23 , that is to the trend of disappearance of this divergent behavior .
we note that the behaviour of @xmath1 as a function of the noise intensity @xmath22 obtained in our analysis is in qualitative agreement with that obtained by the experimental investigation of the stabilization of a metastable state in an oscillatory chemical system ( the belousov - zhabotinsky reaction ) @xcite .
specifically the decreasing behavior of the maximum lyapunov exponent of fig . 2 of ref .
@xcite is in qualitative good agreement with the behavior of the curves in fig . [
6-lya_c]b and fig .
[ 6-lya_c]c .
this could be ascribed to the correlation time always present in noise sources used in any experimental set - up . in fig .
[ 7-n_i_c ] we report , for all the initial positions investigated and for @xmath57 , the semi - log plot of the fraction of particles @xmath59 entering into the potential well up to the position @xmath84 , within the @xmath48 , as a function of noise intensity @xmath22 . at very low noise intensities and for increasing values of the correlation time @xmath0 ,
the particles have difficulty to enter into the potential well , within the @xmath48 considered , shifting the entrance statistics towards higher values of the noise intensity .
in this work we analyzed the effect of the colored noise , generated by an ornstein - uhlenbeck process , on the enhancement of the mean first passage time in a cubic potential with a metastable state and on the minimum of the mean growth rate coefficient as a function of the noise intensity . we analyze different initial unstable states .
we obtain nes effect for all the initial positions investigated and an enhancement of the nes region for increasing values of correlation times .
the results obtained for a particle moving in a cubic potential are quite general , because we always obtain nes effect when a particle is initially located just to the right of a local potential maximum and next to a metastable state , in the escape region . in experiments real noise sources
are correlated with a finite correlation time . as a consequence
the nes effect can be observed at higher noise intensities with respect to the idealized white noise case .
the enhancement and the shift of the nes region , towards higher values of the noise intensity , allows to reveal experimentally the nes effect only by using a suitable correlation time @xmath0 in the noise source .
i. biazzo , f. caltagirone , g. parisi , and f. zamponi , phys .
lett . * 102 * , 195701 ( 2009 ) ; r. giacchetti and e. sorace , phys .
* 101 * , 190401 ( 2008 ) ; l. m. krauss and j. dent , phys .
. lett . * 100 * , 171301 ( 2008 ) .
n. v. agudov , b. spagnolo , phys .
e * 64 * , 035102(r ) ( 2001 ) ; n. v. agudov , a. a. dubkov , b. spagnolo , physica a * 325 * , 144 ( 2003 ) ; b. spagnolo , a. a. dubkov , n. v. agudov , acta physica polonica b , vol . *
35 * ( 4 ) , 1419 ( 2004 ) .
l. schimansky - geier and h. herzel , j. stat .
phys . * 70 * , 141 ( 1993 ) ; g. paladin , m. serva , and a. vulpiani , phys .
lett . * 74 * , 66 ( 1995 ) ; v. loreto , g. paladin , and a. vulpiani , phys .
e * 53 * , 2087 ( 1996 ) . | we present a study of the escape time from a metastable state of an overdamped brownian particle , in the presence of colored noise generated by ornstein - uhlenbeck process .
we analyze the role of the correlation time on the enhancement of the mean first passage time through a potential barrier and on the behavior of the mean growth rate coefficient as a function of the noise intensity .
we observe the noise enhanced stability effect for all the initial unstable states used , and for all values of the correlation time @xmath0 investigated .
we can distinguish two dynamical regimes characterized by weak and strong correlated noise respectively , depending on the value of @xmath0 with respect to the relaxation time of the system . |
the discovery of the first extrasolar planet in 1995 @xcite marked the beginning of a new era in planetary science , characterized by a rapidly expanding set of known extrasolar planets .
more than 200 exoplanets have been discovered so far @xcite . among these , giant gas planets in small orbits are in the majority since the primary tool for detection , radio velocity measurements , is most sensitive to finding heavy planets that rapidly orbit their parent star @xcite . from radius measurements of transient extrasolar planets , it is also known that most of these giant gas planets are like jupiter in consisting primarily of hydrogen and helium . modeling
the interior of such planets requires an accurate equation of state for hydrogen - helium mixtures at high pressure and temperature conditions similar to those in planetary interiors @xcite .
thus , the characterization of such system by first principle calculations will help us to answer questions concerning the inner structure of planets , their origin and evolution @xcite . in this article
, we focus on studying the transition from molecular to atomic hydrogen . in particular , we investigate the effect of different helium concentrations on this transition . in what follows ,
it will be shown how the structure of the fluid and its equation of state ( eos ) change with varying the hydrogen - helium mixing ratio .
, and the temperature is @xmath0 k. the density corresponds to a ) @xmath1 ( @xmath2 g/@xmath3 ) at pressure of @xmath4 gpa , and b ) @xmath5 ( @xmath6 g/@xmath3 ) at pressure of @xmath7 gpa .
these conditions correspond to a molecular phase and to a metallic regime of fluid hydrogen , respectively .
, title="fig : " ] a ) , and the temperature is @xmath0 k. the density corresponds to a ) @xmath1 ( @xmath2 g/@xmath3 ) at pressure of @xmath4 gpa , and b ) @xmath5 ( @xmath6 g/@xmath3 ) at pressure of @xmath7 gpa .
these conditions correspond to a molecular phase and to a metallic regime of fluid hydrogen , respectively .
, title="fig : " ] b ) we use density functional molecular dynamics ( dft - md ) to investigate the questions connected with the physics of giant gas planets . under conditions along jupiter s isentrope , the electrons can be considered to be in their ground state .
they are either chemically bound in molecules or atoms , or form a degenerate interacting electron gas .
density functional theory can thus be used to describe the electrons throughout .
the ions form a strongly coupled fluid which makes many - body simulations an obvious choice for their description . for the interaction of the classical ions and electrons ( quantum particles ) the born - oppenheimer ( bo ) approximation
can be invoked since the motions of electrons and ions follow different timescales .
the results presented here are obtained using the cpmd code @xcite .
we use simulations cells with @xmath8 electrons and the corresponding number of ions , and periodic boundary conditions .
the forces acting on the ions are provided by dft calculations within generalized gradient approximation ( gga ) for the electrons in the coulomb field of the ions .
we use troullier martin norm - conserving pseudopotentials for the electron - ion interactions @xcite . to check for finite - size effects , calculations with supercells ranging from 54 to 250 atoms
are carried out , but give no significant change in the results ( @xmath9 in pressure for supercells with 128 or more particles ) .
we have also performed simulations with different number of k points ( 1 to 512 ) to sample the brillouin zone , and it was confirmed that a single k point already gives converged results for the supercells and conditions used here .
recently , it was shown that in high pressure water effects arising from the treatment of the electrons at their true temperature instead of in the ground state can be important for the dynamics of the system @xcite .
therefore , special attention was given to the effects arising from a finite electron temperature .
we have performed additional dft - md simulations with the vasp code using the finite temperature mermin functional to model the influence of thermal occupation of electronic states @xcite . for jupiter
s mixing ratio of @xmath10 almost no difference could be detected at relevant temperatures in the metallic regime ( less than half a percent deviation ) .
figure [ fig : nice ] displays snapshots of the simulation cell during runs for two different situations . in fig .
[ fig : nice]a ) the molecular fluid at low density ( @xmath1 ) is characterized by larger distances between the molecules and by electronic densities concentrated around the nuclei .
such a molecular fluid can be found in the outer mantle regions of jupiter . the physical system reminiscent to the inner mantle of jupiter , shown in fig . [ fig : nice]b ) , can be characterized as a metallic fluid .
the density is much higher , molecules are dissociated as a result of the increased pressure .
the electrons are delocalized as a result of the pauli exclusion principle and form an electron cloud that extends through the simulation cell .
k across the molecular - atomic transition for various densities and mixing ratios .
the three rows of graphs show the hydrogen - hydrogen , hydrogen - helium , and helium - helium pair correlation functions . here ,
@xmath11 was multiplied by the concentration of hydrogen atoms so that the area under the peak at @xmath12 corresponds to the fraction of molecules in the fluid . in each graph , simulations for different mixing ratios have been combined : pure hydrogen ( @xmath13 ) , hydrogen and helium for jupiter s mixing ratio ( @xmath14 ) , a fluid with as many helium atoms as hydrogen molecules ( @xmath15 ) , and pure helium ( @xmath16 ) .
the columns show results from different electronic densities ( given in terms of the wigner seitz parameter @xmath17 ) , for which different mixing ratios have been compared .
, scaledwidth=100.0% ] for given temperature and density conditions , which of the two described phases can be found , namely , molecular or atomic / metallic , depends strongly on the helium concentration .
moreover , the helium fraction plays a crucial role in how the mixture transforms from the molecular to the atomic phase with increasing temperature or pressure .
changes in the structure of the fluid , i.e. , the transition from a molecular to a metallic fluid may be discussed in terms of the pair correlation function ( pcf ) @xmath18 pair correlation functions for different types of nuclei are shown in fig .
[ fig : grarray ] .
we discuss the changes in the pair correlations with respect to density at a constant temperature of @xmath19 k. let us first comment on the hydrogen - hydrogen pair correlations in pure hydrogen .
a pronounced peak at about @xmath20 demonstrates the presence of molecules in the fluid .
these pair correlations are normalized so that the area under the molecular peak gives the number of molecules found in the fluid .
the dissociation degree can thus be extracted either by integrating over the pcf or by invoking a distance criterion for two hydrogen atoms .
table [ valuetable ] reports degrees of dissociation that where derived by analyzing the md trajectories using the distance criterion and a minimum molecular lifetime of @xmath21 vibrational periods of the isolated hydrogen molecule .
the lifetime criteria leads to a decaying molecular fraction in the metallic regime , which would be overestimated based on the distance criterion alone .
a high peak followed by a minimum that approaches zero is a strong indication that the molecules are stable , for instance this is the case at @xmath1 for different values of @xmath22 .
indeed , this can be confirmed by directly analyzing the number and lifetime of the molecules @xcite . on the other hand , at high density ( @xmath23 , @xmath13 ) ,
no isolated first peak can be identified and the shape of the pcf is typical for a correlated atomic liquid . the first row of fig .
[ fig : grarray ] contains two additional graphs showing the pair correlations at densities and mixing ratios intermediate between the two extrema just discussed . in these cases ,
the molecular peak can still be identified but its height is reduced and it is widened so that the first minimum has a value greater than zero .
a shape like this is caused by a mixture of atoms and molecules ; where the definition of molecules here is somewhat ambiguous as they may have finite lifetimes .
let us now discuss the influence of helium on the pcf .
the general effect can be seen by comparing hydrogen - hydrogen pair correlation functions at identical electronic densities and different helium fractions .
more helium content in the mixture generally leads to a lower hydrogen dissociation .
the presence of helium atoms thus stabilizes the hydrogen molecules and the effect of pressure to dissociate hydrogen molecules is partially suppressed .
a more detailed analysis revealed that in mixtures of hydrogen and helium the bond length of the hydrogen molecule is shortened by about @xmath24 @xcite .
the influence of helium allows hydrogen molecules to exist in a parameter regime where otherwise pure hydrogen would be dissociated .
the microscopic reason for this effect is the double charge of the helium nuclei binding two electrons .
these electrons are bound stronger than the electrons in a hydrogen molecule and are more localized , which has profound effects on the high pressure properties of the fluid @xcite . in the mixture , atoms and molecules can be packed more closely before electronic wavefunctions begin to overlap .
when molecules are compressed and wave functions begin to overlap , their localization increases because of the exclusion principle , which means larger kinetic energy .
but the kinetic and potential energies scale differently with density , so eventually , it is energetically more favorable for the electrons to delocalize ( lower kinetic energy , higher potential energy ) .
we also examine helium - helium ( he - he ) and hydrogen - helium ( h - he ) pair correlation functions extracted from the dft - md simulations .
both , he - he and h - he pair correlation functions show a finite probability for finding a neighboring atom only for distances larger than @xmath25 .
this value reflects the bigger radius of a helium atom compared to a hydrogen atom .
hydrogen - helium correlations are similar to the continuous part of the hydrogen - hydrogen pcfs .
naturally , there is no molecular peak and significant contributions occur at distances larger than the first minimum of the hydrogen - hydrogen pcfs .
thereafter , similar weak short - range order expressed in oscillations around unity can be observed .
there are only minor differences between hydrogen - helium correlation functions of different mixing ratios . .
values for pressure ( 1st line ) and fraction of atoms bound in molecules ( 2nd line ) at various densities and mixing ratios at @xmath26 k. these values belong to the pair correlation functions in fig .
[ fig : grarray ] . [ cols="^,^,^,^,^",options="header " , ]
the correlations in the helium subsystem can be described as short - range hard sphere type interactions .
interestingly , the helium - helium pair correlations show a more pronounced first peak in the @xmath14 mixture than in the @xmath15 mixture and they again feature a slightly higher peak than the ones for pure helium .
there is a bigger total number of hydrogen atoms present for a lower mixing ratio @xmath22 at constant electronic density .
two hydrogen atoms or one hydrogen molecule have a bigger volume than one helium atom . the effective density for the helium atoms
is thus higher in a mixture with less helium .
this causes a general requirement to increased coordination among the atoms which in turn is reflected in the higher peak in the pair correlation .
the effect of hydrogen on helium can be understood as the opposite of the effect of helium on hydrogen .
the observed structure in helium - helium and hydrogen - helium pair correlations illustrates the importance of the interaction between the subsystems and is an indication that the linear mixing rule is only a first approximation for describing mixtures @xcite . , and helium for two different electronic density parameters @xmath17 : a ) @xmath27 corresponds to @xmath28 g/@xmath3 in hydrogen , and b ) @xmath23 corresponds to @xmath29 g/@xmath3 in hydrogen .
, title="fig : " ] a ) , and helium for two different electronic density parameters @xmath17 : a ) @xmath27 corresponds to @xmath28 g/@xmath3 in hydrogen , and b ) @xmath23 corresponds to @xmath29 g/@xmath3 in hydrogen .
, title="fig : " ] b ) naturally , changes in the structure of the fluid have an effect on the equation of state ( eos ) as presented in fig . [
fig : diffx ] .
the pressure isochores show some distinct features depending on the helium concentration .
since helium atoms are smaller than hydrogen molecules , one finds that at low temperature the pressure decreases substantially when hydrogen molecules are replaced by helium atoms at constant volume , which leaves the total electronic density unchanged . in pure hydrogen ,
the isochores exhibit a region with a negative slope @xmath30 , which is due to the delocalization of the electrons during the dissociation of hydrogen molecules .
a higher helium fraction reduces the magnitude of the drop in pressure ( e.g. for @xmath14 ) or even completely removes it ( @xmath15 ) . as a result
, one finds that the @xmath31 curves for @xmath13 , @xmath14 , and @xmath15 approximately converge in the dissociated regime .
there , the pressure depends only little on the helium concentration because the ( ideal ) ionic pressure increase and the pressure decrease due to electronic delocalization depend inversely on the mixing ratio . only for large helium concentration
this cancellation effect does not occur anymore as the comparison with the @xmath16 curve shows .
[ fig : diffx ] alone demonstrates temperature dissociation . by comparing the left and the right part of fig .
[ fig : diffx ] , it is possible to observe pressure dissociation , too .
the region with negative slopes indicating molecular dissociation is shifted to smaller temperatures for higher densities .
this means that , by increasing the density at constant temperature , one can transfer from a molecular fluid to an atomic fluid as well .
the structure of the fluid and the equation of state for hydrogen - helium mixtures were analyzed under conditions found in giant gas planets .
we have considered the structural changes in the fluid during dissociation as function of density and helium concentration .
dft - md proves to be an useful technique to study such questions and to obtain accurate equation of state data .
such data are needed in order to improve existing models for giant gas planets and answer longstanding question concerning their development and inner structure . in particular , we find that helium plays a crucial role in affecting hydrogen dissociation both as a function of temperature and pressure .
hydrogen molecules are stable at higher temperatures and pressures in mixtures of hydrogen and helium than in pure hydrogen .
the transition from the molecular to the atomic phase proceeds over a wider range of temperature and pressure .
we observe regions with negative slope of the pressure isochores in pure hydrogen as well as in hydrogen - helium mixtures , but only up to a critical mixing ratio .
this material is based upon work supported by nasa under the grant nng05gh29
g , by the carnegie institution of canada , and by the nsf under the grant 0507321 . i.t . and s.a.b .
acknowledge support by the nserc of canada . | equilibrium properties of hydrogen - helium mixtures under thermodynamic conditions found in the interior of giant gas planets are studied by means of density functional theory molecular dynamics simulations .
special emphasis is placed on the molecular - to - atomic transition in the fluid phase of hydrogen in the presence of helium .
helium has a substantial influence on the stability of hydrogen molecules .
the molecular bond is strengthened and its length is shortened as a result of the increased localization of the electron charge around the helium atoms , which leads to more stable hydrogen molecules compared to pure hydrogen for the same thermodynamic conditions .
the _ ab initio _ treatment of the mixture enables us to investigate the structure of the liquid and to discuss hydrogen - hydrogen , helium - helium , and hydrogen - helium correlations on the basis of pair correlation functions . |
in many applications such as complex system design or hyperparameter calibration for learning systems , the goal is to optimize some output value of a non - explicit function with as few evaluations as possible .
indeed , in such contexts , one has access to the function values only through numerical evaluations by simulation or cross - validation with significant computational cost .
moreover , the operational constraints generally impose a sequential exploration of the solution space with small samples .
the generic problem of sequentially optimizing the output of an unknown and potentially _ non - convex _ function is often referred to as _ global optimization _ ( @xcite ) , black - box optimization ( @xcite ) or derivative - free optimization ( @xcite ) .
there are several algorithms based on various heuristics which have been introduced in order to address complicated optimization problems with limited regularity assumptions , such as genetic algorithms , model - based algorithms , branch - and - bound methods ... see @xcite for a recent overview .
this paper follows the line of the approaches recently considered in the machine learning literature ( @xcite ) .
these approaches extend the seminal work on lipschitz optimization of @xcite and they led to significant relaxations of the conditions required for convergence , _
e.g. _ only the existence of a local _ smoothness _ around the optimum is required ( @xcite ) .
more precisely , in the work of @xcite and @xcite , specific conditions have been identified to derive a finite - time analysis of the algorithms .
however , these guarantees do not hold when the unknown function is not assumed to be locally smooth around ( one of ) its optimum . in the present work ,
we propose to explore concepts from ranking theory based on overlaying estimated level sets ( @xcite ) in order to develop global optimization algorithms that do not rely on the smoothness of the function .
the idea behind this approach is simple : even if the unknown function presents arbitrary large variations , most of the information required to identify its optimum may be contained in its induced ranking rule , _
i.e. _ how the level sets of the function are included one in another . to exploit this idea ,
we introduce a novel optimization scheme where the complexity of the function is characterized by the underlying pairwise ranking which it defines .
our contribution is twofold : first , we introduce two novel global optimization algorithms that learn the ranking rule induced by the unknown function with a sequential scheme , and second , we provide mathematical results in terms of statistical consistency and convergence to the optimum .
moreover , the algorithms proposed lead to efficient implementation and display good performance on the classical benchmarks for global optimization as shown at the end of the paper . the rest of the paper is organized as follows . in section [ sec : setup ]
we introduce the framework and give the main definitions . in section [ sec : rankopt ] we introduce and analyze the rankopt algorithm which requires a prior information on the ranking structure underlying the unknown function . in section [ sec : adarank ] , an adaptive version of the algorithm is presented .
companion results which establish the equivalence between learning algorithms and optimization procedures are discussed in section [ sec : equivalence ] as they support implementation choices . finally , the adaptive version of the algorithm is compared to other global optimization algorithms in section [ sec : implementation ] .
all proofs are postponed to the appendix section .
* setup . *
we consider the problem of sequentially maximizing an unknown real - valued function @xmath3 where @xmath1 is a compact and convex set . the objective is to identify some point @xmath4 with a minimal amount of function evaluations .
the setup we consider is the following : at each iteration @xmath5 , an algorithm selects an evaluation point @xmath6 which depends on the previous evaluations @xmath7 and receives the evaluation of the unknown function @xmath8 at this point .
after @xmath9 iterations , the algorithm returns the argument of the highest value observed so far : @xmath10 the analysis provided in the paper considers that the number @xmath9 of evaluation points is not fixed and it is assumed that function evaluations are noiseless .
* notations . * for any @xmath11 , we define the standard @xmath12-norm as @xmath13 , we denote by @xmath14 the corresponding inner product and we denote by @xmath15 the @xmath12-ball centered in @xmath16 of radius @xmath17 . for any bounded set @xmath1 , we define its inner - radius as @xmath18 , its diameter as @xmath19 and we denote by @xmath20 its volume where @xmath21 stands for the lebesgue measure .
we denote by @xmath22 the set of continuous functions defined on @xmath23 taking values in @xmath24 and we denote by @xmath25 the set of ( multivariate ) polynomial functions of degree @xmath26 defined on @xmath23 .
finally , we denote by @xmath27 the uniform distribution over a bounded measurable domain @xmath28 and we denote by @xmath29 the indicator function taking values in @xmath30 . in this section
, we introduce the ranking structure as a complexity characterization for a general real - valued function to be optimized .
first , we observe that real - valued functions induce an order relation over the input space @xmath23 , and the underlying ordering induces a ranking rule which records pairwise comparisons between evaluation points .
( induced ranking rule ) the ranking rule @xmath31 induced by a function @xmath3 is defined by : @xmath32 for all @xmath33 .
the key argument of the paper is that the optimization of any weakly regular real - valued function only depends on the nested structure of its level sets .
hence there is an equivalence class of real - valued functions that share the same induced ranking rule as shown by the following proposition .
( ranking rule equivalence ) [ prop : rankingequivalence ] let @xmath34 be any continuous function .
then , a function @xmath35 shares the same induced ranking rule with @xmath36 , i.e. , @xmath37 , @xmath38 if and only if there exists a strictly increasing ( not necessary continuous ) function @xmath39 such that @xmath40 . [ pics : sameranking ] proposition [ prop : rankingequivalence ] states that even if the unknown function @xmath0 admits non - continuous or large variations , up to a transformation @xmath42 , there might exist a simpler function @xmath43 that shares the same induced ranking rule .
figure [ pics : sameranking ] gives an example of three functions that share the same ranking while they display highly different regularity properties . as a second example
, we may consider the problem of maximizing the function @xmath44 if @xmath45 and @xmath46 otherwise over @xmath47 $ ] . in this case , the function @xmath0 is not _
smooth _ around its unique global maximizer @xmath48 but shares the same induced ranking rule with @xmath49 over @xmath23 .
+ we can now introduce a complexity characterization of real - valued functions of a set @xmath23 through the complexity class of its induced ranking rule .
we call this class a ranking structure .
( continuous ranking structure and continuous ranking rules ) let @xmath0 be a real - valued function .
we say that @xmath0 has a continuous ranking rule if @xmath50 where @xmath51 denotes the set of continuous ranking rules ( i.e. ranking rules induced by continuous functions ) . in the continuation of this definition , we further introduce two examples of more stringent ranking structures .
( polynomial ranking rules ) the set of polynomial ranking rules of degree @xmath52 is defined as @xmath53 we point out that even a polynomial function of degree @xmath54 may admit a lower degree polynomial ranking rule .
for example , consider the polynomial function @xmath55 .
since @xmath56 where @xmath57 is a strictly increasing function , the ranking rule induced by @xmath0 is a polynomial ranking rule of degree @xmath58 . the second class of ranking structures we introduce is a class of non - parametric ranking rules .
( convex ranking rules ) the set of convex ranking rules of degree @xmath59 is defined as @xmath60 it is easy to see that the ranking rule of a function @xmath0 is a convex ranking rule of degree @xmath54 if and only all the level sets of the function @xmath0 are unions of at most @xmath54 convex sets .
we now state two conditions that will be used in the theoretical analysis : the first condition is about the identifiability of the maximum of the function and the second is about the regularity of @xmath0 around its maximum .
( identifiability ) [ cond : id ] the maximum of a function @xmath35 is said to be identifiable if for any @xmath61 arbitrary small , @xmath62 condition [ cond : id ] prevents the function from having a jump on its maximum .
it will be useful to state asymptotic results of the type @xmath63 when @xmath64 .
( regularity of the level sets ) [ cond : levelset ] a function @xmath35 has @xmath65-regular level sets for some @xmath66 , @xmath67 if : 1 .
the global optimizer @xmath68 is unique 2 . for any @xmath69 @xmath70 ,
the iso - level set @xmath71 satisfies @xmath72 condition [ cond : levelset ] guarantees that the points associated with high evaluations are close to the unique optimizer with respect to the euclidean distance .
however , note that for any iso - level set @xmath73 with finite distance to the optimum , the condition is satisfied with @xmath74 and @xmath75 .
hence , this condition concerns the local behavior of the level sets when @xmath76 . as an example , the iso - level sets of three simple functions satisfying the condition with different values of @xmath77 are shown in figure [ pics : lvlset ] .
in this section , we consider the problem of optimizing an unknown function @xmath0 given the prior knowledge that its ranking @xmath79 belongs to a given ranking structure @xmath80 .
[ fig : rankopt ] * 1 .
initialization : * let @xmath81 + evaluate @xmath82 , @xmath83 + @xmath84 , @xmath85 + * 2 .
iterations : * repeat while @xmath86 : + let @xmath87 + if there exists @xmath88 such that @xmath89 + evaluate @xmath8 , @xmath90 + @xmath91 + @xmath92 + * 3
. output : * return @xmath93 the input of the rankopt algorithm ( displayed in figure [ fig : rankopt ] ) are a number @xmath9 of iterations , the unknown function @xmath0 , a compact and convex set @xmath1 and a ranking structure @xmath80 . at each iteration @xmath94 , a point @xmath95
is uniformly sampled over @xmath23 and the algorithm decides , whether or not , to evaluate the function at this point .
the decision rule involves the active subset of @xmath96 which contains the ranking rules that are consistent with the ranking rule induced by @xmath0 over the points sampled so far .
we thus set @xmath97 where @xmath98 is the empirical ranking loss : @xmath99 as a direct consequence of the definition of the active subset , if there does not exist any ranking rule @xmath88 satisfying @xmath100 , it also implies that @xmath101 which means that @xmath102 .
thus , the algorithm never evaluates the function at a point that will not return certainly an evaluation at least equal to the highest evaluation @xmath103 observed so far .
( approximation of the level sets ) it is noteworthy that the sampling area implicitly defined by the decision rule @xmath104 can be written in terms of unions of level sets , _
@xmath105 thus , since @xmath79 might be any ranking rule in @xmath106 , the sampling area @xmath107 is also the smallest subset that contains certainly the _ true _ level set of the highest evaluation observed so far @xmath103 .
( connection with active learning ) the proposed algorithm might be seen as an extension to ranking of the baseline _ active learning _
algorithm introduced in @xcite and further analyzed in @xcite .
however , in active learning we aim at estimating a binary classifier @xmath108 where the goal in global optimization is to estimate the winner of a tournament deriving from the ranking rule @xmath109 .
( adaptation to noisy evaluations ) [ rem : noisy ] following the steps of @xcite and @xcite , the algorithm can be extended to noisy evaluations by considering a relaxed version of the active subset @xmath110 where @xmath111 comes out of some standard generalization bound on @xmath112 .
we now state some convergence properties of the rankopt algorithm .
the results are stated in a probabilistic framework .
the source of randomness comes from the algorithm itself which generates uniform random variables and not from the evaluations which are assumed noiseless .
the next result will be important in order to formulate the consistency property of the algorithm .
[ th : fasterprs ] fix any @xmath113 , let @xmath1 be any compact and convex set , let @xmath80 be any ranking structure and let @xmath3 be any function such that @xmath114 .
then , if @xmath93 denotes the random output of the rankopt@xmath115 algorithm , we have that @xmath116 , @xmath117 where @xmath118 are @xmath9 independent random variables , uniformly distributed over @xmath23 .
combining the previous proposition with the identifiability condition gives the following asymptotic result .
( consistency ) [ coro : consistencyrank ] consider the same notations and assumptions as in proposition [ th : fasterprs ] .
then , if the function @xmath0 has an identifiable maximum ( condition [ cond : id ] ) , we have that @xmath119 we now provide our first finite - sample bound on the distance @xmath120 between the exact solution and its approximation .
( upper bound ) [ coro : upperbound ] consider the same assumptions as in proposition [ th : fasterprs ] .
then , if function @xmath0 has @xmath65-regular level sets ( condition [ cond : levelset ] ) , for any @xmath113 and any @xmath121 , with probability at least @xmath122 , @xmath123 where @xmath124 .
more surprisingly , a lower bound can also be derived by connecting the rankopt algorithm with a theoretical algorithm which uses the prior knowledge of the level sets of the unknown function pure adaptive search ( @xcite ) .
the next proposition shows the link between the two algorithms .
[ th : slowerpas ] fix any @xmath113 and let @xmath125 be a sequence of @xmath9 random variables distributed as the markov process defined by : @xmath126 where @xmath127 denotes the level set of @xmath128 .
then , under the same assumptions as in proposition [ th : fasterprs ] , we have that @xmath116 , @xmath129 equipped with proposition [ th : slowerpas ] , we are ready to establish our second finite - time bound .
( lower bound ) [ coro : lowerbound ] consider the same assumptions as in proposition [ th : fasterprs ] .
then , if the function @xmath0 has @xmath65-regular level sets ( condition [ cond : levelset ] ) , for any @xmath121 and any @xmath113 , with probability at least @xmath122 , @xmath130 where @xmath131 .
( on the performance criterion ) observe that the level set assumption used in theorem [ coro : upperbound ] and theorem [ coro : lowerbound ] is invariant to any strictly increasing composition @xmath42 ( i.e. , if @xmath0 has @xmath65-regular level sets , so has @xmath132 ) .
it implies that the bounds derived on the distance @xmath133 hold independently of the smoothness of the unkown function .
we now consider the problem of optimizing a function @xmath0 when no information is available on its induced ranking rule @xmath79 . the adarankopt algorithm ( shown in figure [ fig : adarank ] ) is an extension of the rankopt algorithm which involves model selection following the principle of structural risk minimization @xcite .
we consider a parameter @xmath134 and a nested sequence of ranking structures @xmath135 satisfying @xmath136 [ fig : adarank ] let @xmath81 + evaluate @xmath82 , @xmath137 + @xmath138 , @xmath85 + repeat while @xmath86 + let @xmath139 + if @xmath140 + let @xmath141 + if @xmath142 + let @xmath143 + evaluate @xmath8 , @xmath90 + @xmath92 + @xmath144 + @xmath145 + return @xmath93 the algorithm is initialized by evaluating the function at a point @xmath146 uniformly distributed over @xmath23 and by considering the smallest ranking structure @xmath147 of the sequence . at each iteration @xmath86 , a bernoulli random variable @xmath148 of parameter @xmath149 is sampled .
if @xmath140 , the algorithm _ explores _ the space by evaluating the function at a point uniformly sampled over @xmath23 .
if @xmath142 , the algorithm _ exploits _ the previous evaluations by making an iteration of the rankopt algorithm with the smallest ranking structure @xmath150 of the sequence that probably contains the true ranking @xmath79 . once a new evaluation @xmath8 has been made , the index @xmath151 is updated .
the parameter @xmath149 thus drives the trade - off between the exploitation phase and the exploration phase which prevents the algorithm from getting stuck in a local maximum .
we start by casting the consistency result for the algorithm .
( consistency ) [ prop : cons_adarank ] fix any @xmath134 and let @xmath152 be any nested sequence of ranking structures ( [ eq : nested ] ) .
then , for any function @xmath35 which has an identifiable maximum ( condition [ cond : id ] ) , if @xmath93 denotes the random output of the adarankopt @xmath153 algorithm , we have that @xmath119 informally , proposition [ prop : cons_adarank ] states that even if the algorithm is poorly tuned ( _ e.g. , _ an inappropriate sequence of ranking structures or a bad choice of @xmath149 ) , it remains consistent over the class of functions that have an identifiable maximum .
+ to derive a finite - time analysis of the algorithm , we start to investigate the number of iterations required to identify a ranking structure that contains the _ true _ ranking rule .
( stopping time ) let @xmath154 be the index of the smallest ranking structure that contains the true ranking rule and let @xmath155 be the sequence of random variables driving the model selection , defined in the algorithm .
we define the stopping time which corresponds to the number of iterations required to identify the index @xmath156 as @xmath157 in order to bound @xmath158 , we need to control the complexity of the sequence of ranking structures .
let us denote by @xmath159 the true ranking loss where @xmath160 is a couple of independent random variables uniformly distributed over @xmath23 and define the rademacher average of a ranking structure @xmath96 as @xmath161 where @xmath162 are @xmath9 independent copies of @xmath163 and @xmath164 are @xmath165 independent rademacher random variables ( _ i.e. , _ random symmetric sign variables ) , independent of @xmath166 .
( stopping time upper bound ) [ prop : model ] assume that the index @xmath167 is finite , assume that @xmath168 and assume that there exists a constant @xmath169 such that @xmath170 , the rademacher complexity of @xmath171 satisfies @xmath172 .
then , for any @xmath121 , with probability at least @xmath122 , @xmath173 in the situation described above , a ranking structure which contains the true ranking rule can be idenified in a finite number of iterations .
it is then possible to recover an upper bound similar to the one of theorem [ coro : upperbound ] where the structure structure is assumed to be known .
( upper bound ) [ coro : upper_ada ] consider the same assumptions as in proposition [ prop : model ] and assume further that the function @xmath0 has @xmath65-regular level sets ( condition [ cond : levelset ] ) .
then , if @xmath93 denotes the random output of adarankopt@xmath174 , for any @xmath121 and any @xmath175 , we have that @xmath176 with probability at least @xmath177 where @xmath178 is the same constant as in theorem [ coro : upperbound ] .
( on the complexity assumption ) as pointed out in @xcite ( remark 2 therein ) , standard vc - type arguments can be used in order to bound @xmath179 .
if the set of functions @xmath180 is a vc major class with finite vc dimension @xmath181 , then @xmath182 for a universal constant @xmath183 . in particular , this covers the case of polynomial ranking rules .
( related work ) to the best of our knowledge , this is the first analysis of an optimization algorithm which learn the ranking rule induced by the unknown function .
however , for a different approach to global optimization , we refer to the work of @xcite where the function is assumed to be locally smooth around ( one of ) its optimum . in the latter work , finite - time bounds on the difference @xmath184
are obtained when the smoothness is known ( doo ) and the smoothness is unknown ( soo ) .
we discuss here some technical aspects involved in the practical implementation of the adarankopt algorithm .
fix any nested sequence of ranking structures @xmath135 and suppose that we have collected a sample @xmath185 of @xmath9 evaluations we address the questions of ( i ) sampling @xmath186 uniformly over the non - trivial subset @xmath187 and ( ii ) updating the index @xmath188 once @xmath189 has been evaluated .
we start to show that both these steps can be done using a single generic procedure that tests , for a given @xmath190 , if there exists a ranking rule @xmath191 that perfectly ranks a sample of @xmath192 evaluations , _
i.e. , _ @xmath193
\(i ) sampling @xmath163 until @xmath194 allows to sample uniformly over @xmath195 . by definition of the non - trivial subset ,
we know that @xmath194 if and only if there exists a ranking rule @xmath196 such that @xmath197 or @xmath198 . rewriting the previous statement in terms of minimal error
gives that @xmath194 if and only if : * either @xmath199 where the empirical ranking loss is taken over the sample @xmath200 ( case @xmath201 ) , * or @xmath199 where @xmath202 is taken over the sample @xmath203 where @xmath183 is any strictly positive constant ( case @xmath204 ) .
\(ii ) assume now that @xmath189 has been evaluated .
since @xmath205 forms a nested sequence of ranking structures , it implies that the sequence @xmath206 is stricly increasing so we have that @xmath207 where the empirical loss is taken over the sample @xmath208 .
hence , @xmath188 can be updated by sequentially testing if @xmath209 for @xmath210 + as stated above , for any nested sequence of ranking structures , both the steps ( i ) and ( ii ) can be done using a generic procedure that tests if ( [ eq : test ] ) holds true .
we now provide some equivalences that can be used in order to design such a procedure for the two classes of ranking structures introduced in section [ sec : setup ] . for simplicity , we will assume in the sequel that all the evaluations of the sample are distinct , _
@xmath211 where @xmath212 denote the indexes of the corresponding reordering .
consider the sequence of polynomial ranking rules @xmath213 and let @xmath214 be the function that maps @xmath215 into the corresponding polynomial feature space of degree @xmath54 where @xmath216 , _
we provide here some equivalences that can be exploited in order to implement the algorithm with the polynomial rankings . as a starting point , we start to show that ( [ eq : test ] ) holds true if and only if the set of points @xmath218 satisfies some property . [ prop : binary ] fix any degree @xmath219 and assume that the @xmath192 evaluations of the sample are distinct ( [ eq : sample_ranked ] ) .
then , there exists a polynomial ranking rule of degree @xmath54 that perfectly ranks the sample if and only if there exists an axis @xmath220 satisfying : @xmath221 since only the _ existence _ of a separating axis is required , one can then use the following lemma , presented in the usual binary classification framework and illustrated in figure [ pics : separability ] . ( separability ) [ lem : zero ] let @xmath222 be any set of binary classification where @xmath223 , let ch@xmath224 be the convex hull of @xmath225 and let us denote by @xmath226 the zero vector .
then , there exists a separating axis @xmath227 satisfying : @xmath228 if and only if @xmath229 thus , applying lemma [ lem : zero ] on the sample @xmath230 and using the vertices representation of the associated convex hull leads us to the next equivalence .
[ coro : lp ] under the same assumptions as in proposition [ prop : binary ] , there exists a polynomial ranking rule of degree @xmath54 that perfectly ranks the sample if and only if the polyhedral set defined by @xmath231 is empty where @xmath232 is the @xmath233-matrix with its @xmath234-th column equal to @xmath235 and @xmath236 stands for the inequality @xmath237 component - wise @xmath238i.e .
, @xmath239 ) .
notice that , in practice , the problem of testing the emptiness of a polyhedral set admits a tractable solution , as discussed in remark [ rem : algo ] .
[ pics : separability ] consider the non - parametric sequence of convex ranking rules @xmath241 .
we provide here two equivalences related to the practical implementation of the algorithm with the convex rankings .
first , following the steps of @xcite allows to formulate the next equivalence .
( overlaying classifiers ) [ prop : binary_cvx ] fix any degree @xmath219 , let @xmath242 and assume that all the evaluations are distinct .
then , ( [ eq : test ] ) holds true if and only if there a exists a sequence of classifiers @xmath243 of the form @xmath244 satisfying : 1 .
@xmath245 , @xmath246 , 2 .
@xmath247 .
the problem of overlaying classifiers admits a tractable solution when @xmath242 . in the specific case where @xmath248 and @xmath249 , testing the existence of nested convex classifiers is equivalent to testing the emptiness of a cascade of polyhedral sets , as shown in the next proposition .
[ prop : lp_cvx ] fix any @xmath250 , let @xmath249 and assume that all the evaluations of the sample are distinct
. then , ( [ eq : test ] ) holds true if and only if for each @xmath251 the polyhedral set defined by : @xmath252 is empty where @xmath253 is the @xmath254-matrix where its @xmath234-th column is equal to @xmath255 .
( algorithmic aspects ) [ rem : algo ] testing the emptiness of a polyhedral set can be seen as the problem of finding a feasible point of a linear program . in particular
, this problem can be solved using the simplex algorithm .
for further details , we refer to chapter 11.4 in @xcite where practical examples as well as algorithmic solutions are discussed .
the purpose of this section is to show that the main algorithm of the paper is empirically competitive with existing state - of - the - art global optimization algorithms .
to do so , we compare the _ general _ performances of the adarankopt algotihm with four different algorithms , developed from various approaches of global optimization : * bayesopt ( @xcite ) is a bayesian optimization algorithm .
it uses a distribution over functions to build a surrogate model of the unkown function .
the parameters controlling the distribution are estimated during the optimization process . *
cma - es ( @xcite ) is an evolutionary algorithm . at each iteration ,
the new evaluation points are sampled according to a multivariate normal distribution .
the covariance matrix of the distribution is estimated over the previous evaluations . *
crs ( @xcite ) is a controlled random search with local mutations .
it starts with a random population and randomly evolves these points by an heuristic rule .
the algorithm was taken from the nlopt library ( @xcite ) . *
direct ( @xcite ) is a lipschitz optimization algorithm where the lipschitz constant is unknown .
it uses a deterministic splitting technique of the search space .
the algorithm was also taken from the nlopt library . for a fair comparison ,
the tuning parameters were all set to default and the adarankopt algorithm was used with the polynomial ranking rules and a parameter @xmath149 fixed to @xmath256 . + [ fig : comparison ] in the first series of experiments ( displayed in figure [ fig : comparison ] ) , four standard optimization problems were considered and the the horizon @xmath9 was set to @xmath258 evaluations . for each problem ,
the task consists in maximizing : * the mccormick function @xmath259 over the domain @xmath260 \times [ -3 , 4]$ ] which presents two local maxima . * a perturbed version of the styblinski - tang function @xmath261 which has four local maxima over @xmath262 ^ 2 $ ] . * the levi n.13 function @xmath263 over the domain @xmath264 ^ 2 $ ] .
the levi function has a strong global structure but presents about 100 local maxima . * the himmelblau function @xmath265 over @xmath266 ^ 2 $ ] .
this function has four global maxima .
[ fig : comparison2 ] in the second series of experiments ( displayed in figure [ fig : comparison2 ] ) , the horizon @xmath9 was set to @xmath267 evaluations and the algorithms were compared on four specific problems .
the objectives consist in maximizing : * the one - dimensional function @xmath268 over @xmath269 $ ] where @xmath270 .
this function has 17 local maxima and presents a large discontinuity around its unique optimum @xmath271 . for this problem ,
the convex rankings were temporally used . * the hlder table function @xmath272 over the domain @xmath273 ^ 2 $ ] .
this function has a very weak global structure and presents about 36 local maxima . *
the griewank function @xmath274 over the domain @xmath275 ^ 4 $ ] .
the griewank function has many widespread and regulary distributed local minima . * the function @xmath276 over @xmath277^{10}$ ] .
a special feature of this function is that it is maximized over the whole hyperplane @xmath278 . in most cases ,
we remark that the adarankopt converges fast and avoids falling in local maxima , see for instance * ( a ) * , * ( b ) * , * ( d ) * and * ( e)*. moreover , experiment * ( h ) * also suggests that the algorithm is robust against the dimensionality of the input space when the ranking structure of the unkown function is regular enough .
these performances can be explained by the ability of the algorithm to learn the ranking structure of the unkown function , which allows to adapt to a wide variety of functions .
however , experiments * ( c ) * and * ( g ) * highlight some limits of the algorithm . indeed , when the unkown function presents a strong global structure but has a large amount of local mimima , the algorithm succefully manage to learn the _ global _ ranking structure but fails at locally optimizing the function .
an approach to solve this problem would be to extend the algorithm to noisy evaluations , as stated in remark [ rem : noisy ] .
we have provided a global optimization strategy based on a sequential estimation of the ranking of the unknown function .
we introduced two algorithms : rankopt which requires a prior knowledge of the ranking rule of the unknown function and its adaptive version adarankopt which performs model selection .
a theoretical analysis is provided and the adaptive algorithm is shown to be empirically competitive with the state - of - the - art methods on synthetic optimization problems .
* proof of proposition [ prop : rankingequivalence ] * @xmath279 assume that there exists a strictly increasing function @xmath39 such that @xmath40 and let @xmath280 be the standard sign function defined by @xmath281 . since @xmath42 is a strictly increasing function , @xmath282 , @xmath283 + @xmath284 assume that @xmath285 , @xmath38 . if @xmath282 , @xmath286 , then @xmath287 and @xmath288 are constant over @xmath23 and we have that @xmath289 where @xmath290 is a strictly increasing function .
assume that @xmath0 is not constant over @xmath23 and introduce the function @xmath291 $ ] defined by : @xmath292 we start to show that there exists a strictly increasing function @xmath39 such that @xmath293 . to properly define the function @xmath42 ,
we show by contradiction that @xmath294 the function @xmath0 is constant over the iso - level set @xmath295 . fix any @xmath296 , pick any @xmath297 and assume , without loss of generality , that @xmath298 .
the equality of the rankings implies that ( i ) @xmath299 , @xmath300 and ( ii ) @xmath301 . putting ( i ) and ( ii ) altogether and using the continuity of @xmath36 leads to the contradiction @xmath302 then , denoting by @xmath303 the unique value of the function @xmath0 over @xmath304
, we can now introduce the function @xmath305 defined by @xmath306 since @xmath307 , @xmath308 , the continuity of @xmath36 implies that @xmath309 , @xmath310 .
therefore , @xmath293 where @xmath39 is any strictly increasing extension of the function @xmath311 over @xmath24 . reproducing the same steps
, one can show that there exists a strictly increasing function @xmath312 such that @xmath313 .
hence @xmath314 where @xmath315 is a strictly increasing function .
@xmath316 ( adarankopt process ) [ prop : adarankopt ]
let @xmath1 be any compact and convex set , let @xmath522 be any sequence of nested ranking structures , fix any @xmath134 and let @xmath35 be any function such that @xmath50 .
then , the adarankopt@xmath317 @xmath0 , @xmath23 , @xmath149 , @xmath523 algorithm evaluates the function on a sequence of random variables @xmath524 defined by : @xmath525 where @xmath526 , @xmath148 is a bernoulli random variable of parameter @xmath149 independent of @xmath527 and @xmath323 and @xmath528 are as defined as in the algorithm .
[ prop : rankconcentration ] ( from @xcite ) let @xmath524 be a sequence of @xmath9 independent copies of @xmath163 , let @xmath96 be any ranking structure , let @xmath35 be any real - valued function and let @xmath529 be the empricial ranking loss taken over @xmath530 . then , denoting by @xmath531 the rademacher average of @xmath96 and by @xmath532 the true ranking loss defined in section [ sec : adarank ] , for any @xmath121 , with probability at least @xmath122 , @xmath533 * proof of proposition [ prop : cons_adarank ] * fix any @xmath534 and let @xmath535 be the corresponding level set .
we show by induction that @xmath170 , @xmath536 since @xmath81 and @xmath537 , we have that @xmath538 and the result holds when @xmath325 .
assume that the statement holds for a given @xmath113 and let @xmath442 be the sequence of @xmath192 random variables defined in proposition [ prop : adarankopt ] .
conditioning on @xmath539 gives that @xmath540 .
\end{aligned}\ ] ] since @xmath541 ( see proposition[prop : adarankopt ] ) , we have that @xmath542 therefore , ( [ eq : consada ] ) is proved by plugging the induction assumption into the previous equation .
we finally get that @xmath543 using the fact that @xmath134 and that @xmath544 ( condition [ cond : id ] ) .
@xmath316 + * proof of proposition [ prop : model ] * fix any @xmath121 , let @xmath545 where @xmath546 is the integer part of the upper bound and let @xmath547 be the sequence of random variables defined in proposition [ prop : adarankopt ] . then , denoting by @xmath548 the empirical ranking loss taken over the first @xmath549 samples @xmath550 and since @xmath135 forms a nested sequence , we have that @xmath551 .
hence @xmath552 we start to lower bound the empirical risk by only keeping the first @xmath553 ( i.i.d . )
explorative samples : @xmath554 where @xmath555 .
conditioning on @xmath556 , we know that @xmath557 is a sequence of @xmath556 independent random variables , uniformly distributed over @xmath23 ( see proposition[prop : adarankopt ] ) .
therefore , on the event @xmath558 , the right hand term of inequality ( [ eq : model1 ] ) has the same distribution as @xmath559 where @xmath560 is a sequence of @xmath553 independent copies of @xmath163 , also independent of @xmath561 .
hence @xmath562 where the empirical ranking loss is taken over @xmath560 .
for any @xmath563 we have that @xmath564 and so , by proposition [ prop : concentration ] , we have with probability at least @xmath565 that @xmath566 hence , noticing that the right hand term of the previous inequality is stricly positive , due to the definition of @xmath553 , gives that @xmath567 .
finally , by hoeffding s inequality we have that @xmath568 and we deduce that @xmath569 . @xmath316 + * proof of theorem [ coro : upper_ada ] * fix any @xmath121 , let @xmath570 be the integer part of the upper bound of proposition [ prop : model ] ( with probability @xmath571 ) and let @xmath572 be the upper bound of the theorem [ coro : upperbound ] ( with probability @xmath571 ) .
now , fix any @xmath573 and let @xmath574 be the sequence of random variables defined in proposition [ prop : adarankopt ] .
then , denoting @xmath575 , applying the bayes rules gives that @xmath576 due to the definition of @xmath577 , we have that @xmath578 by proposition [ prop : model ] .
moreover , on the event @xmath579 , the true ranking structure @xmath580 is identified for any @xmath581 .
therefore , the distance @xmath582 can be bounded using the last @xmath583 samples with a similar technique as the one used in the proof of theorem [ coro : upperbound ] ( where the ranking structure is assumed to be known ) and we get that @xmath584 . finally , noticing that @xmath585 ends the proof .
@xmath316 pick any @xmath593 . by definition of the @xmath590-ball
, there exists @xmath594 such that @xmath595 and @xmath596 ( resp .
@xmath597 where @xmath598 and @xmath599 ) .
the convexity of @xmath23 implies that @xmath600 , @xmath601 hence @xmath602 and we deduce that @xmath603 is a convex set . [ lem : equi ] let @xmath80 be any continuous ranking structure and let @xmath604 be any sample satisfying @xmath605 .
then , if @xmath202 denotes the empirical ranking loss taken over the sample , we have the following equivalence : @xmath606 @xmath612
let @xmath613 be any ranking rule satisfying @xmath614 . since @xmath615 is a continuous ranking , there exists a function @xmath34 such that @xmath616 , @xmath37 .
pick any @xmath617 and assume , without loss of generality , that @xmath618 . using the function @xmath36 , we have that @xmath619 hence , @xmath620 , @xmath621 and so @xmath622 .
* proof of proposition [ prop : binary ] * @xmath623 assume that there exists @xmath624 satisfying @xmath622 . since @xmath397 is a polynomial ranking , there exists a polynomial function @xmath625 where @xmath626 such that @xmath285 , @xmath627 . since @xmath622 , applying lemma [ lem : equi ] gives that @xmath628 , @xmath629 hence , there exists @xmath630 s.t .
@xmath631 , @xmath632 .
+ @xmath633 assume that there exists @xmath220 s.t .
@xmath634 , @xmath635 .
pick any constant @xmath636 , define the polynomial function @xmath637 and let @xmath638 be its induced polynomial ranking rule . then , @xmath635 , we have that @xmath639 and applying lemma [ lem : equi ] gives that @xmath640 . @xmath316
+ * proof of lemma [ lem : zero ] * note that @xmath641 .
therefore , one can assume without loss of generality that @xmath642 , @xmath643 by replacing @xmath644 by @xmath645 .
+ @xmath623 assume that there exists @xmath227 such that @xmath635 , @xmath646 and assume by contradiction that @xmath647 . since @xmath647 , we know by proposition [ def : cvx ] that there exists @xmath648 such that @xmath649 , @xmath650 and @xmath651 , @xmath652 and it gives the contradiction @xmath653 @xmath654 assume that @xmath655
. since @xmath9 and @xmath656 are finite , @xmath587 is a closed , compact and convex set and @xmath657 exists and the condition @xmath658 implies that @xmath659 . now , let @xmath660 be the ( unique ) point of the convex hull satisfying @xmath661 .
we show by contradiction that @xmath662 , @xmath663 .
assume that there exists @xmath664 s.t .
the convexity of the convex hull implies that the whole line @xmath666 also belongs to the convex hull .
however , since @xmath667 and @xmath668 , the line @xmath669 is not tangent to the ball @xmath670 and intersects it .
hence , there exists @xmath671 s.t .
since @xmath673 belongs to the convex hull , it leads us to the contradiction @xmath674 we deduce that @xmath662 , @xmath675 .
finally , since @xmath676 , there exists @xmath677 such that @xmath678 , @xmath679 . @xmath316
+ * proof of corollary [ coro : lp ] * combining proposition [ prop : binary ] and lemma [ lem : zero ] gives the next equivalences : @xmath680 by proposition [ def : cvx ] , we know that @xmath681 if and only if there exists @xmath682 such that @xmath683 , @xmath650 and @xmath684 , @xmath685 .
putting those constraints into matricial form gives the result .
@xmath316 + * proof of proposition [ prop : binary_cvx ] * @xmath623 assume that there exists @xmath686 such that @xmath687 and let @xmath688 be the sequence of classifiers defined by @xmath689 . by lemma [ lem : equi ] , we know that @xmath635 , @xmath690 and so @xmath691 . on the other hand , by definition of the convex rankings of degree @xmath54 , we know that all the classifiers are of the form @xmath692 .
+ @xmath654assume that there exists a sequence of classifiers @xmath693 satisfying : ( i ) @xmath694 , ( ii ) @xmath695 and ( iii ) @xmath696 , @xmath697 .
define the non - continuous function @xmath698 and observe that @xmath699 , @xmath700 and so @xmath687 .
now , let @xmath701 be an approximation of the function @xmath702 where @xmath703\\ 1 - \frac{x - u}{\epsilon } & \ \ \
\text{if}\ \
\ x \in ] u , u + \epsilon ] \\ 0 & \ \ \
\text{otherwise}. \end{cases}\ ] ] first , by continuity of @xmath704 we know that @xmath705 is continuous .
second , note that @xmath706 and @xmath707 , we have that @xmath708 and so @xmath709 . finally , using the same decomposition , it is easy to see that @xmath710 , the level sets of @xmath705 are unions of at most @xmath54 segements ( convex sets ) . hence , there exists @xmath711 satisfying @xmath622 .
@xmath316 now , fix any @xmath719 and observe that the set @xmath720 is a convex set by convexity of the ranking rule @xmath397 .
however , since @xmath721 is the smallest convex set that contains @xmath722 ( see proposition [ def : cvx ] ) , we necessarily have that @xmath723 .
therefore , putting the previous statements altogether gives that @xmath724 , @xmath725 by proposition [ def : cvx ] , it implies that @xmath724 , there does not exist any @xmath726 such that @xmath727 , @xmath728 and @xmath684 .
putting those constraints into matricial form gives the result .
+ @xmath654 assume that all the polyhedrons are empty .
reproducing the same ( inverse ) steps as in the first equivalence gives that @xmath729 , @xmath730 and so @xmath731 now , define the non - continuous function @xmath732 , observe that @xmath733 and let @xmath734 be an approximation of the function @xmath0 , where @xmath735 , @xmath736 if @xmath737 and @xmath738 otherwise .
first , note that for any convex set @xmath23 , the function @xmath739 is continuous and so is the function @xmath705 .
second , note that @xmath740 and @xmath741 , we have that @xmath708 and so @xmath742 . finally , since the @xmath590-ball @xmath603 of any convex set @xmath23 is also a convex set ( see lemma [ lem : epsiball ] ) , @xmath307 and @xmath743 , the level set @xmath744 is a convex set .
hence , there exists @xmath745 satisfying @xmath746 . @xmath316
( rankopt process ) [ prop : chain ] let @xmath1 be any compact and convex set , let @xmath80 be any ranking structure and let @xmath3 be any function such that @xmath114 . then , the rankopt@xmath317 @xmath318 @xmath319 @xmath320 algorithm evaluates the function @xmath0 on a random sequence @xmath162 defined by : @xmath321 where @xmath322 denotes the sampling area and @xmath323 and @xmath106 are defined as in the algorithm .
the result is proved by induction . since @xmath324 ,
the result trivially holds when @xmath325 .
assume that the satement holds for a given @xmath113 .
by definition of the algorithm and using the induction assumption , we know that the rankopt@xmath326 algorithm evaluates the function on a sequence of random variables @xmath327 where @xmath328 rankopt@xmath329 , @xmath330 and @xmath331 is a sequence of independant random variables uniformly distributed over @xmath23 and independant of @xmath162 . therefore , for any borelian @xmath332 , @xmath333 and the result is proved by induction .
noticing that @xmath107 is a subset of @xmath23 gives the first inclusion . to state the second inclusion ,
pick any @xmath337 and observe that @xmath338 . since @xmath79 always perfectly ranks the sample ( _ i.e. , _
@xmath339 ) , there exists @xmath340 such that @xmath341 and we deduce that @xmath342 . [ def : pas ] ( pure adaptive search process , from @xcite ) .
let @xmath1 be any compact and convex set and let @xmath35 be any function such that @xmath50 .
we say that the sequence @xmath343 is distributed as a pure adaptive search process pas@xmath344 if it has the same distribution as the markov process defined by : @xmath345 where @xmath346 is the level set of the highest value observed so far .
[ lem : fasteruniform ] let @xmath347 be a sequence of @xmath9 random variables distributed as a pas@xmath344 process .
then , for any @xmath348 $ ] , we have that @xmath349 where @xmath350 are @xmath9 independent copies of @xmath351)$ ] and @xmath352 denotes the sampling area of the pas@xmath344 process at time @xmath9 .
note that if @xmath353 the result trivially holds for any @xmath354 and any @xmath113 .
therefore , we assume without loss of generality that @xmath355 and we set some additional notations : @xmath356 $ ] define @xmath357 and let @xmath358 be the corresponding level set . observing that @xmath356 $ ] , @xmath359 , we are ready to prove the lemma by induction .
+ fix any @xmath348 $ ] and let @xmath360)$ ] be a random variable independent of @xmath361 .
since @xmath362 and @xmath363 , we have that @xmath364 and the result holds when @xmath325 .
assume that the statement holds for a given @xmath113 , fix any @xmath348 $ ] and let @xmath365 be a sequence of @xmath192 random variables distributed as a pas@xmath326 process . since @xmath366 ( definition [ def : pas ] ) , conditioning on @xmath367 gives that @xmath368.\ ] ] since the level sets form a nested sequence , we have the following equivalences : ( i ) @xmath369 and ( ii ) @xmath370 .
hence @xmath371 and so @xmath372.\ ] ] now , let @xmath373)$ ] be a random variable independent of @xmath374 . since @xmath375 is uniformly distributed over @xmath376 $ ] and independent of @xmath377 , we have that @xmath378 therefore , @xmath379
= \p \left ( u_{n+1 } \cdot \frac{\mu(\x^{\star}_n)}{\mu(\x ) } \leq \frac{\mu(\x_{y_u})}{\mu(\x ) } \right).\ ] ] finally , using the fact that @xmath359 and plugging the induction assumption into the previous equation gives ( by independence ) that : @xmath380 where @xmath381 are @xmath192 independent copies of @xmath351)$ ] . taking the logarithm on both sides
gives that @xmath385 @xmath386 @xmath387 . since @xmath388)$ ] , we have that @xmath389 and so @xmath390 gamma@xmath391 .
the result follows from the application of concentration results of sub - gamma random variables ( see chapter 2.4 in @xcite ) .
[ lem : inclusion ] let @xmath1 be any compact and convex set and let @xmath392 be any function that has ( @xmath393)-regular level sets ( condition [ cond : levelset ] ) .
then , for any @xmath394 where @xmath271 denotes the unique optimizer of the function @xmath0 over @xmath23 , we have that @xmath395 where @xmath396 denotes the intersection of @xmath23 with the @xmath12-sphere centered in @xmath271 of radius @xmath397 . to prove the second inclusion , we first show that @xmath398 $ ] , there exists @xmath399 which satisfies @xmath400 . fix any @xmath401 $ ] ,
pick any @xmath402 and introduce the function @xmath403 \rightarrow \r$ ] defined by @xmath404 which returns the value of @xmath0 over the segment @xmath405 $ ] . note that the function is continuous and well - defined due to the convexity of @xmath23 and the continuity of @xmath0 .
since @xmath406 , @xmath407 and @xmath408 $ ] , applying the intermediate value theorem gives us that there exists @xmath409 $ ] such that @xmath410 .
hence , there exists @xmath411 such that @xmath412 .
assume now that there exists @xmath413 such that @xmath414 .
as a direct consequence , it implies that @xmath415 .
however , we also have that @xmath416 and so putting altogether the previous inequalities with the level set assumption leads us to the next contradiction : @xmath417 the contradiction holds for any @xmath418 $ ] .
we deduce that @xmath419 which proves the second inclusion .
+ we use a similar technique to prove the first inclusion .
assume that there exists @xmath420 s.t .
@xmath421 . by introducing the function @xmath422
, one can show that there exists @xmath423 s.t .
hence , @xmath425 . on the other hand ,
it leads to a similar contradiction and we deduce that @xmath427 .
introduce the similarity transformation @xmath431 defined by : @xmath432 and let @xmath433 be the image of @xmath23 by the similarity transformation . by definition
we have that @xmath434 .
hence , the convexity of @xmath23 implies that @xmath435 and we have that @xmath436 .
now , since @xmath437 is a similarity transformation ( and conserves the ratios of the volumes before / after transformation ) we have that @xmath438 noticing that @xmath439 , @xmath440 where @xmath441 stands for the standard gamma function gives the result .
* proof of proposition [ th : fasterprs ] * the statement is proved by induction . since @xmath81 ,
the result trivially holds when @xmath325 .
assume that the statement holds for a given @xmath113 and let @xmath442 be a sequence of @xmath192 random variables distributed as a rankopt@xmath443 process . since the result trivially holds @xmath444 , fix any @xmath445 $ ] and let @xmath446 be the corresponding level set .
applying proposition [ prop : chain ] gives that @xmath447 we start to bound the second term . since @xmath448 ( see proposition [ prop : chain ] ) ,
conditioning on @xmath449 gives that @xmath450 .
\end{aligned}\ ] ] lemma [ lem : rep ] gives the following inclusion of events : @xmath451 .
hence @xmath452 plugging ( [ eq : pop12 ] ) into ( [ eq : prop11 ] ) and noticing that @xmath453 leads to the next inequality : @xmath454 now , let @xmath455 be a sequence of @xmath192 independent copies of @xmath163 . reproducing the same steps as previously and using the fact that @xmath456 gives that @xmath457 finally , plugging the induction assumption into ( [ eq : prop13 ] ) and comparing the result with ( [ eq : prop14 ] ) yields to the desired result . @xmath316 + * proof of corollary [ coro : consistencyrank ] * pick any @xmath458 and let @xmath459 be the corresponding level set .
proposition [ th : fasterprs ] gives that @xmath170 , @xmath460 where @xmath461 are @xmath9 independent copies of @xmath462 .
therefore , we have that @xmath463 since @xmath464 by condition [ cond : id ] , @xmath465 @xmath316 * proof of theorem [ coro : upperbound ] * since @xmath466 is a continuous ranking , there exists @xmath34 such that @xmath37 , @xmath467 ( see proposition [ prop : rankingequivalence ] ) .
since all the arguments only use function comparisons , one can assume without loss generality that @xmath468 .
we also set some additional notations : let @xmath469 be the upper bound of the theorem , let @xmath470 and let @xmath471 @xmath472 @xmath473 @xmath474 @xmath475 . since the result trivially holds when @xmath469 @xmath237 @xmath476 , assume that @xmath477 which also implies by the level set assumption that @xmath478 . using the second inclusion of lemma [ lem : inclusion ] and proposition [ th : fasterprs ]
gives that @xmath479 where @xmath480 are @xmath9 independent copies of @xmath163 and @xmath481 is defined in lemma [ lem : inclusion ] . since the random variables @xmath482 are independent and uniformly distributed over @xmath23 , applying the first inclusion of lemma [ lem : inclusion ] gives that @xmath483 lemma [ lem : zab ] gives that @xmath484 .
therefore @xmath485 finally , using the fact that @xmath486 , @xmath487 yields to the desired result .
@xmath316 + * proof of proposition [ th : slowerpas ] * the statement is proved by induction . since @xmath146 and @xmath488 are both uniformly distributed over @xmath23 , the result holds when @xmath325 . assume that the statement holds for a given @xmath113 and let @xmath442 be a sequence of @xmath192 random variables distributed as the rankopt@xmath489 process . since the result trivially holds @xmath490 , fix any @xmath491 $ ] and let @xmath492 be the corresponding level set .
then , denoting by @xmath195 the sampling area of @xmath186 and by @xmath493 the level set of the highest value observed so far , reproducing the same steps as in the proof of proposition [ th : fasterprs ] gives that @xmath494 . \end{aligned}\ ] ] lemma [ lem : rep ] gives the following inclusion @xmath495 .
hence @xmath496 .
\end{aligned}\ ] ] for any random variable @xmath497 $ ] , we have that @xmath498 and so @xmath499 now , let @xmath500 be a sequence of @xmath192 random variables distributed as a pas@xmath501 process . reproducing the same steps as previously gives that @xmath502 = \frac{\mu(\x_y ) } { \mu(\x ) } + \int_{\frac{\mu(\x_y ) } { \mu(\x ) } } ^1 \p\left ( \mu(\x_{f(x^{\star}_{n } ) } ) < \frac{\mu(\x_y)}{t } \right)~dt.\ ] ] finally , by the induction assumption @xmath503 , we have that @xmath504 where @xmath505 and the proof is complete by plugging ( [ eq : slowerpas3 ] ) into ( [ eq : slowerpas1 ] ) and by comparing the result with ( [ eq : slowerpas2 ] ) .
@xmath316 + * proof of theorem [ coro : lowerbound ] * as mentioned in the proof of theorem [ coro : upperbound ] , since @xmath466 , there exists @xmath506 such that @xmath507 and one can assume without loss of generality that @xmath468 .
we also set some additional notations : let @xmath508 be the lower bound of the theorem , let @xmath509 and let @xmath510 . using the first inclusion of lemma [ lem : inclusion ] and proposition [ th : slowerpas ]
gives that @xmath511 where @xmath343 is a sequence of @xmath9 random variables distributed as the pas@xmath329 process . applying the second inclusion of lemma [ lem : inclusion ] and denoting by @xmath512
the level set of the highest value observed so far gives that @xmath513 since @xmath514 , there exists @xmath428 such that @xmath515 which implies that @xmath516
. therefore @xmath517 by definition of @xmath518 , we have that @xmath519 and so , applying lemma [ lem : fasteruniform ] gives that @xmath520 where @xmath350 are @xmath9 independent copies of @xmath351)$ ] .
we finally get that @xmath521 by lemma [ prop : concentration ] . | in this paper , we consider the problem of maximizing an _ unknown _ function @xmath0 over a compact and convex set @xmath1 using as few observations @xmath2 as possible .
we observe that the optimization of the function @xmath0 essentially relies on learning the induced bipartite ranking rule of @xmath0 .
based on this idea , we relate global optimization to bipartite ranking which allows to address problems with high dimensional input space , as well as cases of functions with weak regularity properties .
the paper introduces novel meta - algorithms for global optimization which rely on the choice of any bipartite ranking method .
theoretical properties are provided as well as convergence guarantees and equivalences between various optimization methods are obtained as a by - product .
eventually , numerical evidence is given to show that the main algorithm of the paper which adapts empirically to the underlying ranking structure essentially outperforms existing state - of - the - art global optimization algorithms in typical benchmarks .
global optimization , ranking , statistical analysis , convergence rate bounds |
the bright radio source pks 1830211 ( subrahmanyan et al . 1990 ; hereafter s90 , jauncey et al .
1991 ) has attracted much attention as the most detailed example of a lensed radio ring . among the classically - lensed qsos ,
its short time delay of 44 days ( van ommen et al .
1995 ) and clean lens geometry make it a good candidate for measuring h@xmath2 .
the lens , a gas rich galaxy at z=0.89 , was discovered in the millimeter via molecular absorption ( wiklind & combes 1996 ) , which is seen towards only one of the two flat spectrum hot spots ( wiklind & combes 1996 , frye et al .
the presence of a nearby m - star as well as heavy extinction along the line of sight ( b=-5.7 degrees ) had until now hampered the identification of the lens and the source . in this paper
we describe how the mcs deconvolution algorithm ( magain , courbin , & sohy 1998 ) was used to detect the counterparts of this bright radio ring in deep keck optical and infrared images .
near ir @xmath3 ( @xmath4=1.25 micron ) and @xmath5 ( @xmath4=2.15 micron ) images were obtained on the nights of 1997 april 14 and 15 with the irac2b camera on the eso / mpi 2.2 m telescope , which uses a nicmos3 256@xmath6256 hgcdte array . the good seeing ( 06 - 07 ) and the fact that a good sampling is desirable for deconvolution , led us to choose the smallest pixel size available on this instrument , i.e. , 0151 , resulting in a total field of view of 38 on a side .
the data were taken and reduced exactly in the same way as in courbin , lidman & magain ( 1998 ) .
several standard stars were observed during the night .
the standard deviation in the zero points was less than 0.02 magnitudes for both @xmath3 and @xmath5 .
the ir magnitudes reported in this paper are on the @xmath7 system of bessell & brett ( 1988 ) .
near - ir keck i data were obtained on the night of 1994 april 5 with nirc ( matthews & soifer , 1994 ) .
five 80 second @xmath8 images were combined using the sigma clipping algorithm available in iraf .
the pixel size is 0157 , similar to that used with irac2b .
the images were obtained under marginally non - photometric conditions , with a zero point uncertain by about 0.1 magnitude . due to the crowding of the field , and the low number of frames available , sky subtraction was imperfect but did not influence dramatically the quality of the data analysis .
six dithered @xmath9-band images were obtained during the full moon night of 1997 june 15 using the keck ii telescope and lris ( oke et al .
the ccd detector was a tektronix 2048@xmath62048 with a pixel size of 0215 .
the individual exposure times were restricted to 3 minutes in order to avoid saturation of the brightest stars in this extremely crowded field .
the background was very high .
the images were bias subtracted and flat fielded in the standard way .
all of them have a seeing close to 08 .
no standard star was observed , but a flux calibration could be performed relative to an @xmath9-band image taken at la silla with the 0.9 m telescope on the night of 1997 april 15 .
the mcs deconvolution code ( magain et al . 1997 ) was applied to all the images . due to the low signal - to - noise ( s / n ) and the numerous bad pixels in single ir images , these were medianed in groups of 9 dithered and sky subtracted frames .
the resulting images have better s / n and cosmetics .
two infrared - bright nearby stars , although outside the field of view , produced scattered light in the field , forcing us to reject a fair fraction ( @xmath10 percent ) of our observations in the @xmath5 band .
one of the culprits was the known source iras 18306 - 2106 .
two stacks were obtained in @xmath3 ( total exposure time of 1920 sec ) and four in @xmath5 ( total exposure time of 1440 sec ) .
only one such stack was obtained for the ir keck images since we had only 5 dithered frames to combine ( total exposure time of 400 sec ) .
the deconvolution process is the same as in courbin et al .
( 1998 ) . we chose a pixel scale in the deconvolved images that is a factor of two smaller than the pixel scale in the original data , to insure that the sampling theorem is satisfied in the deconvolved images .
after deconvolution , the resolution is fixed to 2 pixels full - width - half - maximum ( fwhm ) for all data .
the corresponding final pixel scales and resolutions on the plane of the sky are given in fig .
we constructed the infrared point - spread - function ( psf ) from the relatively isolated , bright star labelled h in djorgovski et al .
( 1992 ; hereafter d92 ) . in the optical images ,
the stars mentioned in d92 are all saturated .
consequently , the psf was constructed from 4 fainter stars about 30 away from the radio ring . crowding in the optical field
made the quality of the psf rather poor , especially in the outer wings of the hexagonal keck psf , introducing systematic residuals into the deconvolved image and residual maps ( see section 3.2 ) .
in each band , all the frames available were _ deconvolved simultaneously _ ,
i.e. 6 in @xmath9 , 2 in @xmath3 , 4 in @xmath11 and 1 in @xmath8 . in other words ,
the output of the procedure is a unique deconvolved image which is simultaneously compatible with all the images included in the data set .
the deconvolved image is given as a sum of point sources with _ known gaussian shape _ and a diffuse background smoothed on the length scale of the final resolution chosen by the user .
the photometry and the astrometry of the point sources are also obtained as byproducts of the deconvolution and are provided in tables 1 and 2 .
many deconvolution algorithms generate the so - called `` deconvolution artefacts '' as well as noise enhancement .
even if the mcs code does not produce artefacts , it still has to accomplish the difficult task of deconvolving the image from imperfect and noisy data .
an objective criterion has therefore been established to check the quality of the results .
a natural product of the mcs code is the rm ( residual map ) , which is the difference between the deconvolved and real images divided by the standard deviation of each pixel . whenever the simultaneous deconvolution capabality of the method is used , as in the present case , the program returns one rm for each frame in the data set .
a perfect deconvolved image should match the data at best in the sense of the @xmath12 and leave flat rm with a mean value of 1 ( one standard deviation ) all over the field .
rms guide the user in his / her choice of the smoothing to be applied to the image in order to avoid local under- or over - fitting of the data .
still , on the basis of the rms , the user can constrain the number of point sources to be involved in the deconvolution .
for example , missing a bright point source ( i.e. , with high s / n ) will result in a `` hot spot '' well above the critical value of 1 in the rms .
this is no longer true for very faint sources . in this case
, no discrimination can be made between an extended object and a point source , so the faint point sources are modelled as part of the deconvolved background .
we therefore always choose the minimum number of point sources needed in order to produce acceptable rms .
thus , different images of the same field can have different aspects , depending on whether individual objects were modeled as point sources or not .
objects near the frame edges are not well - fitted , especially in crowded fields such as these ( e.g. , objects labelled 1 to 4 in fig .
1 ) . object 1 leaves particularly significant structures in the rms of the @xmath9-band data , either when modeled as a point source or as part of the background in the deconvolved image .
we therefore conclude that it is extended or that it is a very strong blend of point sources . in order to run the mcs algorithm
, the user has to provide initial positions and intensities for all points sources .
the astrometric 1@xmath13 error bars were estimated as the dispersion of the results of several deconvolutions using different initial conditions .
the typical astrometric accuracy is 005 for the brightest point sources , while it can be as much as 02 for the faintest ones . since the peak intensity of a point source
is allowed to be different in each data frame , the photometric error bar is simply the dispersion of the peak intensities found after simultaneous deconvolution of a whole data set . in order to quantify the additional errors introduced by a poor knowledge of the psf in the keck @xmath9-band , the deconvolution program was run using 5 different psfs computed from different stars in the field .
in fig . 1 and plate 1 we present both the raw and deconvolved images in @xmath9 , @xmath5 and @xmath8 , with a resolution of the order of 015 - 020 .
a red point source is clearly detected at the position expected for the ne radio source of pks 1830211 .
another red object is observed close to the position of the sw radio source of the lensed system , but the extended nature of the source and the poor quality of the psf do not allow to sort out its detailed morphology . the photometry and astrometry of the field are presented in tables 1 and 2 and in fig .
2 , along with our estimates of the 1@xmath13 error bars . the red star in d92 dominates the total flux in the @xmath9-band , while in the infrared the component near the ne radio source takes over . with our high s
/ n we can show that its shape is compatible with a point source , and its color , @xmath14 , is much redder than any `` normal '' star ( e.g. , koornneef , 1983 ) .
the @xmath9 , @xmath5 , and @xmath8 positions are all within the 1@xmath13 radio error bars . in fig .
2 all objects are aligned with respect to the m - star so that we can compare the different positions measured for the ne and sw components at optical and ir wavelengths .
molecular absorption towards the two lensed images was spatially resolved at millimeter wavelengths , and the separation between the lensed images was found to be 098 ( frye et al .
1997 ) . in our optical and ir images
the sw component is @xmath15 and @xmath16 respectively , away from the ne component .
a plausible explanation for the apparent positional shift between the optical , ir and radio positions is that the sw component is a blend of two objects : the lensing galaxy and the heavily reddened sw component seen in the radio images .
this has been very recently confirmed by hst nicmos3 observations by the castles group ( kochanek et al .
1998 ) which show the lensing galaxy in the center of the `` hole '' of the radio ring , as predicted by lensclean modeling of the system by kochanek & narayan ( 1992 ) .
note also the very good general agreement between the hst images and our deconvolutions , although the latter were carried out without prior knowledge of the hst data .
the flux ratio between the two lensed images of the source is 0.973 @xmath17 0.016 at radio wavelengths ( s90 ) , and about 18.2 in @xmath8 ( [email protected] magnitudes ) .
the combination of a reddened sw radio source plus blue lens can explain the large flux ratios .
both ne and sw components are reddened , making pks 1830211 another good example of a dusty lens , along with mg 0414 + 0534 ( e.g. , annis & lupino 1993 ) and mg1131 + 0456 ( larkin et al . 1994 ) , the mean galactic extinction being far below the values obtained for pks 1830211 . at the lens redshift ,
the observed @xmath8-band corresponds to a central wavelength of 1.17 @xmath18 in the restframe . for the galactic reddening curve ,
the extinction in this band is @xmath19 ( sandage & mathis , 1981 ) .
thus , the differential extinction is about 8.5 magnitudes in the @xmath20-band and implies @xmath21 , in good agreement with the value of 2.4 independently derived from the hst observations ( kochanek et al .
1998 ) .
if the total extinction is comparable to the differential extinction between the two images , then the source is attenuated in the observed @xmath8-band by about 3 magnitudes . on the other hand , considering simple sis or point - mass models , the magnification by the lens may be similar , or somewhat less , maybe 2 magnitudes .
so , a reasonable value for the unobscured un - lensed @xmath8 magnitude of the source may be around 14 .
a typical quasar has @xmath22 .
so the unobscured @xmath20 magnitude of the source may be about 17 , which is bright , but not unreasonable for @xmath23 .
the main result of our study is the detection of the optical and near - ir counterpart to the ne radio source of pks 1830 - 211 and the possible detection of the sw component and lensing galaxy .
however , our sw component candidate might be the lensing galaxy alone or , given the crowding in the field , a red galactic object almost coincident with the position of the sw radio source .
the hypothesis of a demagnified third image of the source ( s90 ) between the 2 main lensed images is unlikely as in such a case extinction of the lens would have made it visible in the ir .
furthermore , the ir centroid of the sw component would have been shifted towards `` object e '' rather than towards the radio position of the sw component . the higher contrast between the ne component and nearby m - star in the infrared make near - ir spectroscopy necessary for finding the source redshift . deep , high resolution near - ir imaging is needed to reveal the exact nature of the faint sw component . however , even at the highest resolution attainable , 02- 03 in the ir with hst ( in particular in @xmath8 where the sw component of pks 1830 - 211 is best visible ) , deconvolution will be essential to discriminate between the sw component candidate , the lensing galaxy and additional faint galactic stars .
we thank jack welch , hy spinrad , jens hjorth , andreas jaunsen , chris kochanek and the anonymous referee for useful discussions and comments about the first version of the manuscript .
we also thank alfred rosenberg who provided the eso @xmath9-band frame used for the photometric calibration .
the expert help of the staff at eso and wmko during the observing runs was very much appreciated .
fc is supported by contracts arc 94/99 - 178 `` action de recherche concerte de la communaut franaise '' and ple dattraction interuniversitaire p4/05 ( sstc , belgium ) .
sgd and map acknowledge support from the nsf pyi award ast-9157412
. additional support for map was provided by nasa through grant number hf-01099.01 - 97a from stsci , which is operated by aura under nasa contract nas5 - 26555 . | new high quality keck and eso images of pks 1830211 are presented . applying a powerful new deconvolution algorithm to these optical and infrared data , both images of the flat spectrum core of the radio source
have been identified .
an extended source is also detected in the optical images , consistent with the expected location of the lensing galaxy .
the source counterparts are very red at @xmath0 , suggesting strong galactic absorption with additional absorption by the lensing galaxy at @xmath1 , and consistent with the detection of high redshift molecules in the lens . |
as soon as the class was discovered , the obvious similarities between the black hole x - ray transients ( bhxrts ) and their white dwarf analogues , dwarf novae ( dn ) , guided investigations into the mechanisms responsible for the dramatic outbursts exhibited by the former .
the outbursts in dn have been successfully explained as the result of temperature - dependent viscosity in the accretion disk : the disk instability model ( dim ) ( cannizzo 1993 ) .
the longer recurrence timescales for bhxrts and the shapes and durations of their outburst lightcurves , however , provide a challenge to the dim ( lasota 1996 ) .
the dim makes definite quantitative predictions for the temperature distribution , and hence the expected broad band spectrum , throughout the outburst cycle ( _ e.g. _ cannizzo , chen , & livio 1995 ) .
accretion disk emission is likely to dominate in the uv , so one of the primary motivations for spectroscopic observations of bhxrts with hst is , therefore , to observe the broad band spectral evolution , and hence address the question of the driving mechanism for the transient outbursts .
this paper reviews the uv - optical spectra of bhxrts obtained with hst , and describes the consequent deductions about the outburst mechanisms .
the first bhxrt to be observed with hst was a0620 - 00 , 16 years after the 1975 outburst .
mcclintock , horne , & remillard ( 1995 ) interpreted the @xmath0 hst spectrum in conjunction with a quiescent rosat observation . after subtracting the contribution of the k5 v mass donor star
, they found an optical - uv accretion spectrum which could be modeled as a 9000 k blackbody , with an area of only @xmath1 of the disk area . the low uv flux emitted by this accreting black hole was a surprise .
by analogy with quiescent dn a mass transfer rate into the outer disk of @xmath2 was inferred . meanwhile , the rosat soft x - ray flux implied a mass transfer rate through the inner disk of only @xmath3 .
qualitatively , therefore , these findings were in agreement with the dim , suggesting the accumulation of material in the quiescent outer disk .
the extremely low @xmath4 seemed improbable , however , and the authors pointed out that isolated black holes might well accrete more than this from the ism ! a new explanation was advanced by narayan , mcclintock , & yi ( 1996 ) , who postulated that the standard disk model is only applicable to the outer flow , and that within @xmath5 the flow is advective : _
i.e. _ the viscously - generated thermal energy is carried with the flow rather than being promptly radiated away . for black hole accretors , this advected energy can be carried through the event horizon . with this hypothesis , therefore , the extremely low quiescent accretion fluxes do not necessarily demand the extremely low mass transfer rates inferred from the standard accretion disk model .
this object was the first to be monitored in the uv - optical through the decline from outburst , though hst observations occurred only at one epoch , four months after the maximum .
the spectral evolution was analyzed by cheng et al .
the data appeared consistent with steady - state optically thick accretion disks and the deduced mass transfer rate fell monotically during the decline .
the dim predicts , however , that the declining mass transfer rate is accompanied by a cooling wave propagating through the disk as successive hot , high viscosity , annuli make the transition to the cool , low viscosity , state .
the consequent changing temperature distribution should have produced an observable cooling wave signature at the long wavelength end of the spectrum .
the cooling wave was not observed , however , suggesting problems with the straightforward application of the dim to bhxrts .
gro j1655 - 40 was discovered in 1994 july ; since then it has undergone repeated outbursts to a similar level and is apparently an atypical bhxrt .
superluminal radio jets were associated with the 1994 outburst ( hjellming , these proceedings ) . following the onset of x - ray activity in april 1996 ,
hst spectra were obtained on five separate visits from 1996 may 14 to july 22 .
a full description of these observations and the associated multiwavelength campaign is given in hynes et al .
( 1997 ) .
gro j1655 - 40 is a highly reddened source , so an accurate correction for interstellar extinction is a prerequisite to any analysis of the spectrum .
the @xmath6 feature gives a sensitive measure of the extinction : e(b - v)=@xmath7 , a value consistent with direct estimates of the visual extinction and with measurements of interstellar absorption lines ( hynes et al .
1997 ) .
figure 1 is the 1996 may 14 dereddened uv - optical spectrum .
though the uv portion of the spectrum is consistent with the @xmath8 power - law predicted by the steady - state blackbody disk model , the optical ( @xmath9 ) spectrum rises to longer wavelengths in contrast to the predictions of the model . ignoring the @xmath10 data
, a @xmath8 model can be fit to the uv data , leading us to deduce the mass transfer rate is @xmath11m@xmath12yr@xmath13m@xmath12yr@xmath14 , where the dominant source of uncertainty arises from interstellar extinction . taking a compact object mass of 7m@xmath15 and assuming an accretion efficiency of @xmath16
, the eddington rate is @xmath17m@xmath18yr@xmath14 , so near the peak of the outburst this interpretation of the uv spectrum implies @xmath19 .
[ v1 ] we need to invoke something other than a pure steady - state optically thick accretion disk in order to explain the optical light .
the shape of the spectrum is qualitatively suggestive of an irradiated disk ; irradiation can alter the temperature profile of the outer disk producing a rise in flux towards longer wavelengths as illustrated in figure 2 .
the multiwavelength lightcurves for the outburst ( hynes et al .
1997 , and hynes et al .
these proceedings ) do not , however , appear to support a simple irradiation model : the optical and uv flux declines while the x - ray flux rises ! [ mwplot ] in order to characterize the optical component of the spectra we fit blackbody spectra to the @xmath9 data for each visit ( figure 3 ) .
while the fluxes fell by about a factor of three between our first and last visit , the color temperature remained almost constant , dropping from 9800 k to 8700 k ; the emitting area dropped from @xmath20 to @xmath21 .
the system parameters for gro j1655 - 40 are well constrained ( orosz and bailyn 1997 , hereafter ob97 , hjellming and rupen 1995 ) and the total available emitting area of the disk and secondary star is @xmath22 , so it is possible to explain the optical emission at the peak of the outburst as thermal emission , but only if both the secondary star and the majority of the disk area have essentially the same isothermal temperature distribution . attributing the @xmath8 uv component to a steady - state disk is not necessarily ruled out , as this requires only the hot inner regions of the disk . in addition to the large isothermal area
, there is a suggestion that our optical spectra are more strongly peaked than a single temperature blackbody , so we considered non - thermal mechanisms .
self - absorbed synchrotron emission from a cloud of relativistic electrons produced good fits to the optical component ; the deduced electron energies are @xmath23 , the magnetic field is @xmath24 , and the size of the cloud is @xmath25 .
figure 3 shows both the blackbody and the synchrotron models .
while the synchrotron models fit better , they have an extra free parameter . attributing substantial optical flux to this compact nonthermal source
relieves the requirement for a large isothermal emitter in the system .
it is interesting to note that intrinsic vri band linear polarization ( @xmath26 ) was detected in july 1996 ( scaltriti et al . 1997 ) , consistent with the hypothesis of optical synchrotron emission . on the other hand
, the dim may _ require _ a large isothermal outer disk for a long - period system like gro j1655 - 40 .
since the disk in such a system is large , the temperature for a steady state disk falls below the minimum temperature for the hot , high viscosity , state long before the outer disk is reached . even for an eddington mass transfer rate in gro j1655 - 40
this limit is reached for a radius less than a quarter of the roche lobe radius .
this means that there is no global steady - state solution for @xmath27 .
it is not clear what will happen to the temperature distribution in such a case , but it is possible that in outburst much of the outer disk could be maintained just in the hot state . hence one might expect the outer disk to appear as an approximately constant temperature , shrinking area emitter as the decline proceeds .
this hypothesis need to be tested with self - consistent numerical modeling , but until this is done our data can not rule out a thermal interpretation .
the c iv @xmath28 , si iv @xmath29 , and si iii @xmath30 uv resonance lines shown in figure 4 all show likely p - cygni profiles .
the peak to trough separation in all three cases is @xmath31 , slightly larger than seen in outbursting dwarf novae .
line profiles produced by biconical accretion disk winds were calculated by shlosman and vitello ( 1993 ) who found ` classical ' p cygni profiles only for inclinations around 6070@xmath32 , in striking agreement with the inclination determined for gro j1655 - 40 ( ob97 ) .
we conclude that there was likely a biconical accretion disk wind at the peak of the uv outburst , when @xmath33 .
[ pcyg ] observations of this target were obtained in early quiescence approximately two years after the first observed outburst , and seven months after the last reported reflare ( callanan 1995 ) .
a total of 4.6 hours exposure time was spent obtaining the optical - uv spectrum with five overlapping spectrograph set - ups .
the target was extremely faint ( @xmath34 ) , so extreme care was taken with the data reduction .
the spectrum is shown in figure 5(a ) .
we note that even if the entire @xmath35 flux is attributed to the m2 v mass donor ( fillipenko , matheson , and ho 1995 ) , the flux at @xmath36 must be at least @xmath37 due to the accretion flow , and shortwards of @xmath38 the accretion light overwhelmingly dominates .
shrader et al .
( 1994 ) obtained multiwavelength data in outburst , and noted that the optical - uv energy distribution showed a pronounced break in slope at @xmath39 .
e(b - v)=0.4 was deduced from the outburst iue data .
the quiescent hst data , dereddened using this value for the interstellar extinction , is shown in figure 5(b ) . the spectral shape in fig .
5(b ) is unusual .
there is a very pronounced change in slope at @xmath40 and the shape bears a strong resemblance to that of gro j1655 - 40 in outburst ( figure 1 . ) by analogy with the synchrotron fits to the gro j1655 - 40 spectra , we speculate that the strongly rising optical accretion flux may be partly due to optical synchrotron as predicted by advection models ( _ e.g. _ narayan , barret , & mcclintock 1997 ) .
the observations described herein have provided some intriguing challenges to the theoretical models for transient outbursts . as we collect detailed multi - epoch data on more systems
it is becoming clear that the bhxrts exhibit complex and diverse behavior , and it seems we need to consider a variety of mechanisms in order to properly understand the wealth of observational phenomena .
support for this work was provided by nasa through grant numbers go-6017.01 - 94a and go-4377.02 - 92a from the space telescope science institute , which is operated by aura under nasa contract nas5 - 26555 .
we are grateful to our collaborators , coauthors on hynes et al .
1997 , for their contributions to the work described herein .
callanan , p.j .
garcia , m.r .
, mcclintock , j.e . ,
zhao , p. , remillard , r.a . , bailyn , c.d .
orosz , j.a .
harmon , b. a. , paciesas , w.s .
1995 , , 441 , 786 .
cannizzo j. k. , 1993 , in accretion disks in compact stellar systems , wheeler j. c. , world scientific publishing , singapore , p. 318
cannizzo j. k. , chen w. , livio m. , 1995 , apj , 454 , 880 cheng , f .- h . , horne , k. , panagia , n. , shrader , c.r . ,
gilmozzi , r. , paresce , f. , lund , n. 1992 , , 397 , 664 .
filippenko , a.v . , matheson , t. , ho , l.c .
1995 , , 455 , 614 .
hjellming r. m. , rupen m. p. , 1995
, nat , 375 , 464 hynes , r.i . , haswell , c.a .
, shrader , c.r . ,
chen , w. , horne , k. , harlaftis , e.t .
, obrien , k. , hellier , c. , and fender , r.p .
1997 , , submitted .
lasota , j.p .
1996 , in compact stars in binaries , j. van paradijs et al .
mcclintock , j. , horne , k. , and remillard , r. , 1995 , , 442 , 358 .
narayan , r. , barret , d. , mcclintock , j.e .
1997 , apj , 482 , 448 narayan , r. , mcclintock , j. , and yi , i. 1996 , , 457 , 821 .
orosz j. a. , bailyn c. d. , 1997 , apj , 477 , 876 ( ob97 ) scaltriti f. , bodo g. , ghisellini g. , gliozzi m. , and trussoni e. 1997 , , 325 , 29 .
shlosman i. , vitello p. , 1993
, , 409 , 372 shrader , c.r . ,
wagner , r.m .
, hjellming , r.m .
, han , x.h . , starrfield , s.g .
1994 , , 434 , 698 . | hubble space telescope observations of black hole x - ray transients are discussed in the context of the disk instability outburst model .
we focus on the multiwavelength campaign following gro j1655 - 40 through the summer 1996 outburst . |
of agent and the cooperation between agents in multi - agent system ( mas ) , a framework , named distributed constraint optimization problem ( dcop ) in terms of constraints that are known and enforced by distinct agents comes into being with it .
in last decade , the research effort of dcop has been dedicated on the following three directions : 1 ) the development of dcop algorithms which are able to better balance the computational complexity and the accuracy of solution , such as large neighborhood search method @xcite ; markov chain monte carlo sampling method @xcite @xcite and distributed junction tree based method @xcite 2 ) the extension of classical dcop model in order to make it more flexible and effective for practical application , such as expected regret dcop model @xcite , multi - variable agent decomposition model @xcite and dynamic dcop model @xcite 3 ) the application of dcop in modeling environmental systems , such as sensor networks @xcite , disaster evacuation @xcite , traffic control @xcite and resource allocation @xcite . in this paper ,
we take more attention to the application of dcop .
more precisely , we leverage dcop to solve user association problem in the downlink of multi - tier heterogeneous networks with the aim to assign mobile users to different base stations in different tiers while satisfying the qos constraint on the rate required by each user .
is generally regarded as a resource allocation problem @xcite in which the resource is defined by the resource blocks ( rbs ) . in this case , the more rbs allocated to a user , the larger rate achieved by the user .
the methods to solve the user association problem are divided into centralized controlled and distributed controlled . with regard to the centralized way , a central entity
is set up to collect information , and then used to decide which particular bs is to serve which user according to the collected information .
a classical representation of centralized method is max - sinr @xcite .
distributed controlled methods attract considerable attention in last decade since they do not require a central entity and allow bss and users to make autonomous user association decisions by themselves through the interaction between bss and users . among all available methods ,
the methods based on lagrange dual decomposation ( ldd ) @xcite and game theory @xcite have better performance .
hamidreza and vijay @xcite put forward a unified distributed algorithm for cell association followed by rbs distribution in a @xmath2-tier heterogeneous network .
with aid of ldd algorithm , the users and bss make their respective decisions based on local information and a global qos , expressed in terms of minimum achievable long - term rate , is achieved .
however , the constraint relaxation and the backtrack in almost each iteration are needed to avoid overload at the bss .
in addition , as we will show later , the number of out - of - service users will increase since a user always selects a best - rate thereby taking up a large number of rbs and leaving less for others .
nguyen and bao @xcite proposed a game theory based method in which the users are modeled as players who participate in the game of acquiring resources .
the best solution is the one which can satisfy nash equilibrium ( ne ) . loosely speaking ,
such solution is only a local optima .
in addition , it is difficult to guarantee the quality of the solution .
there is no research of modeling user association problem as mas .
however , some similar works have been done focusing on solving the resource management problem in the field of wireless networks or cognitive radio network @xcite by dcop framework .
these methods can not be directly applied to user association problem mainly due to the scale of the models for these practical applications is relatively small .
for instance , monteiro @xcite formalized the channel allocation in a wireless network as a dcop with no more than 10 agents considered in the simulation parts .
however , the amount of users and resource included in a hetnet is always hundreds and thousands . in this case ,
a good modeling process along with a suitable dcop algorithm is necessary . according to the in - depth analysis above
, it motivates us to explore a good way to solve user association problem by dcop .
the main contributions of this paper are as follows : * an ecav ( each connection as variable ) model is proposed for modeling user association problem using dcop framework .
in addition , we introduce a parameter @xmath0 with which we can control the scale ( the number of variables and constriants ) of the ecav model . * a dcop algorithm based on markov chain ( mc ) is proposed which is able to balance the time consumption and the quality of the solution . *
the experiments are conducted which show that the results obtained by the proposed algorithm have superior accuracy compared with the max - sinr algorithm .
moreover , it has better robustness than the ldd based algorithm when the number of users increases but the available resource at base stations are limited .
the rest of this paper is organized as follows . in section
[ preliminary ] , the definition of dcop and the system model of user association problem along with its mixed integer programming formulation are briefly introduced . in section [ formulation_with_dcop ]
, we illustrate the ecav-@xmath0 model .
after that , a mc based algorithm is designed in section [ markov_chain_algorithm ] .
we explore the performance of the dcop framework by comparing with the max - sinr and ldd methods in section [ experimental_evaluation ] .
finally , section [ conclusion ] draws the conclusion .
this section expounds the dcop framework and system model of user association problem along with its mixed integer programming formulation . the definitions of dcop have a little difference in different literatures @xcite . in this paper
, we formalize the dcop as a four tuples model @xmath3 where @xmath4 consists of a set of agents , @xmath5 is the set of variables in which each variable @xmath6 only belongs to an agent @xmath7 .
each variable has a finite and discrete domain where each value represents a possible state of the variable .
all the domains of different variables consist of a domain set @xmath8 , where @xmath9 is the domain of @xmath10 .
a constraint @xmath11 is defined as a mapping from the assignments of @xmath12 variables to a positive real value : @xmath13 the purpose of a dcop is to find a set of assignments of all the variables , denoted as @xmath14 , which maximize the utility , namely the sum of all constraint rewards : @xmath15 consider a @xmath2-tier hetnet where all the bss in the same tier have the same configurations .
for example , a two - tier network including a macro bs ( @xmath16 ) and a femto bs , ( @xmath17 ) , is shown in fig.[ecav_instance ] .
the set of all bss is denoted as @xmath18 where @xmath19 is the total number of bss .
all the bss in the @xmath2th tier transmit with the same power @xmath20 .
the total number of users is denoted by @xmath21 and the set of all users is @xmath22 . with ofdma technology in lte - advanced networks , the resource , time - frequency , is divided into blocks where each block is defined as a resource block ( rb ) including a certain time duration and certain bandwidth @xcite . in this paper ,
the resource configured at each bs is in the format of rb so that its available rbs are decided by the bandwidth and the scheduling interval duration allocated to that bs .
we assume the bss in the hetnet share the total bandwidth such that both intra- and inter - tier interference exist when the bss allocate rbs to the users instantaneously . assuming the channel state information is available at the bss , the @xmath23 experienced by user @xmath24 , served by @xmath25 in the @xmath2th tier is given by @xmath26 in ( [ sinr ] ) , @xmath27 is the channel power gain between @xmath24 and @xmath25
, @xmath28 represents all the bss in @xmath29 except @xmath25 , @xmath30 is the bandwidth and @xmath31 is noise power spectral density .
the channel power gain includes the effect of both path loss and fading .
path loss is assumed to be static and its effect is captured in the average value of the channel power gain , while the fading is assumed to follow the exponential distribution . from the above ,
the efficiency of user @xmath24 powered by bs @xmath25 , denoted as @xmath32 , is calculated as @xmath33 given the bandwidth @xmath30 , time duration @xmath34 and the scheduling interval @xmath35 configured at each rb , we attain the unit rate at @xmath24 upon one rb as follows @xmath36 on the basis of formula ( [ rate_rb ] ) , the rate received at @xmath24 with @xmath37 rbs provided by @xmath25 in the @xmath2th tier is @xmath38 associated with each user is a quality - of - service ( qos ) constraint .
this is expressed as the minimum total rate the user should receive . denoting the rate requiremnt of the @xmath39th user by @xmath40 , the minimum number of rbs required to satisfy @xmath40 is calculated by : @xmath41
in which @xmath42 is a ceiling function .
the formulations of user association problem by mixed linear programming are similar in a series of papers ( see the survey literature @xcite ) . in this paper
, we present a more commonly used formulation as follows @xmath43 the first constraint ensures the rate qos requirement from users . constraint ( [ formula_rb_constraint ] ) indicates that the amount of rbs consumed at the same bs is no more than the total rbs @xmath44 configurated at the bs . constraint ( [ formula_link_constraint ] )
guarantees one user associated with a unique bs .
constraint ( [ formula_nrb_constraint ] ) guarantees the number of rbs a bs allocates to a user falls within the range from zero and @xmath44 .
the last constraint ( [ formula_vector_constraint ] ) guarantees the connection between a user and a bs has two states denoted by a binary variable .
the objective function ( [ formula_objective ] ) refers to the sum of rate rather than a function acted on the rate such as @xmath45 ( e.g. @xmath46 ) in some references .
generally , two phases are needed to gain the solution including : 1 ) transforming original problem into a satisfied one through relaxing constraint ( [ formula_nrb_constraint ] ) by @xmath47 ; 2 ) the left rbs in each bs will be allocated to users in order to maximize the objective function .
in this section , we expound and illustrate the ecav model along with its modified version ecav-@xmath0 . before giving the formulation based on dcop
, we firstly introduce the definition of candidate bs : [ candidate_bs ] we declare @xmath48 is a candidate bs of @xmath49 if the rate at @xmath24 is above the threshold @xmath50 with @xmath51 rbs provided by @xmath25 .
simultaneously , @xmath51 should be less than the total number of rbs ( @xmath44 ) configurated at @xmath25 . after confirming the set of candidate bss of @xmath49 , denoted by , @xmath52
, @xmath24 sends messages to its candidate bss so that each @xmath53 gets knowledge of its possible connected users .
we define each possible connection between @xmath24 and its candidate bs @xmath25 as a variable , denoted by @xmath54 . in this case , all the variables are divided into @xmath19 groups according to the potential connection between users and different bss .
the domain of each variable @xmath54 , denoted by @xmath55 , where @xmath56 if no rb is allocated to @xmath24 , otherwise , @xmath57 .
we define each group as an agent .
thus , an @xmath58-ary constraint exists among @xmath58 variables ( intra - constraint ) to guarantee that there is no overload at @xmath25 .
note that a user may have more than one candidate bs , there are constraints ( inter - constraints ) connecting the variables affiliated to different agents on account of the assumption that a unique connection exists between a user and a bs .
generally speaking , the utility ( objective ) function in the dcop model is the sum of constraint rewards which reflects the degree of constraint violations .
we define the reward @xmath59 of inter- and intra - constraints in the ecav model as follows . for @xmath60 r(c)= [ ecav_constraint_1_b ] - , & @xmath61 + [ ecav_constraint_1_a ] 0 , &
otherwise for @xmath62 r(c)= [ ecav_constraint_2_a ] - , & @xmath63 + [ ecav_constraint_2_b ] , & otherwise in constraint ( [ ecav_constraint_1_b ] ) , @xmath64 is the subset of variables connected by constraint @xmath65 .
@xmath66 represents the assignment of @xmath54 .
a reward ( we use @xmath67 in this paper ) is assigned to the constraints if there at least two variables are non - zero at the same time ( unique connection between a user and a bs ) .
otherwise , the reward is equal to zero . in constraint ( [ ecav_constraint_2_a ] )
, the reward is @xmath67 once there is a overload at the bs . otherwise , the reward is the sum of the rates achieved at users .
it is easy to find that a variable in the ecav model with non - zero assignment covers constraint ( [ formula_rate_constraint ] ) and ( [ formula_nrb_constraint ] ) in the mixed integer programming formulation .
moreover , intra and inter - constraints respectively cover constraint ( [ formula_rb_constraint ] ) and constraint([formula_link_constraint ] ) .
therefore , the global optimal solution @xmath14 obtained from the ecav model is consistent with the one obtained from the mixed integer programming formulation , denoted as @xmath68 .
is consistent with @xmath68 when the total rate calculated by objective function ( [ formula_optimization_function ] ) and ( [ formula_objective ] ) is equal .
this is because there may be no more than one optimal solution . ] to better understand the modeling process , we recall the instance in fig.[ecav_instance ] where the candidate bss of @xmath69 and @xmath70 are the same , denoted as @xmath71 , while the candidate bss of @xmath72 and @xmath73 are respectively @xmath74 and @xmath75 .
we assume the total rbs configurated at @xmath16 and @xmath17 is 8 and 10 . for simplicity
, we assume the rate of each user served by one rb provided by @xmath16 is 0.8 bit / s . and 1 bit / s of each user is served by @xmath17 . then , the ecav model is shown in fig.[ecav_instance : b ] .
there are two agents named @xmath76 and @xmath77 .
the variables in @xmath76 are @xmath78 and @xmath79 where @xmath54 refers to a connection between user @xmath24 and @xmath25 .
similarly , the variables in @xmath77 are @xmath80 and @xmath81 . assuming the threshold rate is 3 bit / s
, we can calculate that at least @xmath82 rbs needed for the users served by @xmath16 , thus the domain of each variable in @xmath76 is @xmath83 . also , the domain of each variable in @xmath77 is @xmath84 .
the black lines in each agent are two 3-nry intra - constraints , thus @xmath85 .
the red lines connecting two agents are two intra - constraints , thus @xmath86 .
we use @xmath87 and @xmath88 to illustrate how the reward of constraint works in different conditions . considering @xmath87 ,
the reward is @xmath67 when all the variables associated with @xmath87 have the same assignment 4 .
thus the total number of rbs consumed by three users is 12 which is more than 8 rbs configurated at @xmath16 .
otherwise , the reward is 0.8 @xmath89 4 @xmath89 3 = 9.6 ( bit / s ) calculated according to ( [ rate ] ) . considering @xmath88 ,
the reward is @xmath67 when the assignment of @xmath90 is 3 and the assignment of @xmath91 is 4 because it means @xmath70 will connect with more than one bss ( @xmath16 and @xmath17 ) , which violates the assumption of unique connection .
otherwise , the reward is 0 ( [ ecav_constraint_1_a ] ) ) .
if there is no constraint violated , the final utility calculated by the objective function is the total rate in the whole hetnet ( constraint ( [ ecav_constraint_2_b ] ) ) .
the scale of an ecav model , referring to the number of agents and constraints , is related to the number of users , bss and the candidate bss hold at each user .
however , some candidate bss of the user can be ignored because these bss are able to satisfy the requirement of the user but with massive rbs consumed .
it can be illustrated by the number of rbs a bs allocate to a user is inversely proportional to the geographical distance between them . in this way , we introduce a parameter @xmath0 with which we limit the number of candidate bss of each user is no more than @xmath0 .
the following algorithms present the selection of top @xmath0 candidate bss ( denoted by @xmath92 ) and the modeling process of ecav-@xmath0 .
[ alg1allcsstart ] @xmath93 [ alg1allcsend ] bubblesort ( @xmath52 ) [ alg1sortstart ] @xmath94 @xmath95 [ alg1sortbysinrstart ] exchane @xmath96 and @xmath97 [ alg1sortbysinrend ] algorithm [ alg1 ] is the pseudo code for determining @xmath92 .
it is executed by each user distributely .
more precisely , a user estimates its total candidate bss @xmath98 by the procedure from line [ alg1allcsstart ] to [ alg1allcsend ] .
based on [ unit_efficiency ] to [ unit_rb ] , the candidate bss of a user is ordered according to the unit number of rbs consumed at such user served by different bss ( from line [ alg1sortbysinrstart ] to [ alg1sortbysinrend ] ) .
the time consumption of algorithm [ alg1 ] mainly consists of two parts .
one is the dermination of @xmath98 with time complexity @xmath99 .
the other is the ordering operation with time complexity @xmath100 . as a result ,
the total time expended of algorithm [ alg1 ] is @xmath101 . with @xmath92 ,
we present the pseudo code in relation to the building of ecav-@xmath0 model . @xmath102 @xmath103 @xmath104 [ ecav_agent ] [ ecav_user_start ] @xmath105 @xmath106 @xmath107 @xmath108 @xmath109 [ ecav_user_end ] @xmath110 [ ecav_intra_c_start ] @xmath111 [ ecav_intra_c_end ] as for algorithm [ alg2 ] , it firstly sets up the agents distributely ( line [ ecav_agent ] ) .
it takes @xmath112 .
after that , each user determines variables , domains as well as inter - constraints from line [ ecav_user_start ] to [ ecav_user_end ] .
this is also carried out in parallel with @xmath112 .
finally , the intra - constraints are constructed by each agent with @xmath112 ( line [ ecav_intra_c_start ] to [ ecav_intra_c_end ] ) .
the total time complexity is @xmath113 .
dcop , to some degree , is a combinatorial optimization problem in which the variables select a set of values to maximize the objective function without or with the minimum constraint violation .
we use @xmath114 to denote the set of all possible combination of assignments of variables .
also , we call each element @xmath115 as a * candidate solution*. considering an ecav-@xmath0 model in which the four tuples are as follows : * @xmath116 * @xmath117 a connection between @xmath24 and @xmath118 * @xmath119 * @xmath120 we are able to rewrite the model in the following way : [ transformation_1 ] @xmath121 after that , a convex log - sum - exp approximation of ( [ transformation_1_objective_function ] ) can be made by : @xmath122 where @xmath123 is a positive constant .
we then estimate the gap between log - sum - exp approximation and ( [ transformation_1_objective_function ] ) by the following proposition in @xcite : [ gap ] given a positive constant @xmath123 and @xmath58 nonnegative values @xmath124 , we have + @xmath125 in addition , the objective function ( [ transformation_1_objective_function ] ) has the same optimal value with the following transformation : @xmath126 in which @xmath127 is the reward with a candidate solution @xmath128 .
for simplicity , we use @xmath129 . hence , on the basis of formulations ( [ log_sum_exp_approximation ] ) and ( [ gap_equation ] ) , the estimation of ( [ transformation_1_objective_function ] ) can be employed by evaluating @xmath130 in the following way : @xmath131 assuming @xmath132 and @xmath133 are the primal and dual optimal points with zero duality gap . by solving the karush - kuhn - tucker ( kkt ) conditions
@xcite , we can obtain the following equations : [ kkt ] @xmath134 then we can get the solution of @xmath135 as follows : @xmath136 on the basis of above transformation , the objective is to construct a mc with the state space being @xmath114 and the stationary distribution being the optimal solution @xmath135 when mc converges . in this way ,
the assignments of variables will be time - shared according to @xmath135 and the system will stay in a better or best solution with most of the time .
another important thing is to design the nonnegative transition rate @xmath137 between two states @xmath128 and @xmath138 . according to @xcite , a series of methods
are provided which not only guarantee the resulting mc is irreducible , but also satisfy the balance equation : @xmath139 . in this paper
, we use the following method : @xmath140^{-1}\label{transit_rate}\ ] ] the advantage of ( [ transit_rate ] ) is that the transition rate is independent of the performance of @xmath138 . a distributed algorithm , named wait - and - hp in @xcite , is used to get the solution after we transform dcop into a mc .
however , as the existence of inter- and intra- constriants in , a checking through the way of message passing is made in order to avoid constraint violation .
in this section , we test the performance of the mc based algorithm with different assginments of @xmath0 in the ecav model .
a simulated environment including a three - tiers hetnet created within a @xmath141 square is considered . in the system , there is one macro bs , 5 pico bss and 10 femto bss with their transmission powers respectively 46 , 35 , and 20 dbm .
the macro bs is fixed at the center of the square , and the other bss are randomly distributed .
the path loss between the macro ( pico ) bss and the users is defined as @xmath142 , while the pass loss between femto bss and users is @xmath143 .
the parameter @xmath144 represents the euclidean distance between the bss and the users in meters .
the noise power refers to the thermal noise at room temperature with a bandwidth of 180khz and equals to -111.45 dbm .
one second scheduling interval is considered . without special illustration ,
200 rbs are configured at macro bs , as well as 100 and 50 rbs are configured at each pico and femto bs . in addition , all the results are the mean of 10 instances .
we firstly discuss the impact of different assignments of @xmath0 on the performance of ecav model from the point of view of the runtime and the quality of solution .
more precisely , we generate different number of users ranging from 20 to 100 with the step interval of 10 . the time consumed by the mc based algorithm
is displayed in fig.[runtime ] .
it is clear to see that more time is needed when the number of users increases .
also , the growth of runtime is depended on the value of @xmath0 .
specially , there is an explosive growth of runtime when we set @xmath0 from four to five .
as previously stated , this is caused by more candidate bss considered by each user .
however , the quality of the solutions with different values of @xmath0 is not obviously improved according the results in table [ table_model_variables ] . for instance , the average rate achieved at each user is only improved no more than 0.1 bit / s when the number of users are 100 with the values of @xmath0 are 3 and 5 .
it is difficult to make a theoretical analysis of the realationship between @xmath0 and the quality of the solution .
we leave this research in future works . from above analysis , we set @xmath1 in the following experiments in order to balance the runtime and performance of the solution .
in addition , we test the performance of the mc based algorithm comparing with its counterparts max - sinr and ldd based algorithms .
.the average rate ( bit / s ) achieved at each user [ cols="^,^,^,^,^,^",options="header " , ] ( @xmath145 ) with different number of users in the hetnet [ runtime ] ] ] ] in fig.[distributed_200 ] , we check the connection state between 200 users and bss in different tiers .
a phenomenon we can observe from the figure is that there are more or less some users out of service even we use different kinds of algorithms .
it is not only caused by the limited resource configured at each bss , but also related to the positions of such kinds of users .
they are located at the edge of the square and hardly served by any bs in the system .
further , more users are served by macro bs in max - sinr algorithm because a larger sinr always eixsts between the users and macro bs . as a result , the total non - served users in mmax - sinr algorithms are more than the other two if there is no scheme for allocating the left resource . on the other hand , the number of non - served users in mc are less than ldd when @xmath1 since the user @xmath24 will select a bs @xmath25 with the maximal @xmath146 in each iteration of the ldd algorithm . in other words ,
the users prefer to connect with a bs which can offer better qos even when more resources are consumed .
therefore , some bss have to spend more rbs which leads to the resource at these bss being more easily used up . ] ] in fig.[non served ] , we produce a statistic of the number of non - served users when we change the total number of users configured in the hetnet .
the average number of non - served users for each algorithm along with the standard deviation is presented in the figure . compared with fig.[distributed_200 ] , a more clear results imply that more than 60 ( at worst , around 70 ) non - served users in the max - sinr algorithm . the ldd based algorithm comes the second with approximate 20 users .
the best resutls are obtained by the mc algorithm with no more than 20 users even the total users in the hetnet is 240 . in fig.[cdf_200 ] , we compare the cumulative distribution function ( cdf ) of the rate . the rate of the users seldomly drops below the threshold ( 3 bit / s ) when we use the distributed algorithms ( ldd and mc based algorithms ) , while max - sinr algorithm is unable to satisfy the rate qos constraints .
moreover , the rate cdfs of the mc based algorithm never lie above the corresponding cdfs obtained by implementing the max - sinr algorithm ( the gap is between @xmath147 ) .
likewise , at worst @xmath148 gap eixts between the mc based algorithm and ldd when we set @xmath1 . at last ,
another intesest observation is made by configurating different number of rbs at macro bs ( fig.[number of rb in mac ] ) .
when we change the number of rbs from 150 to 250 at macro bs , it is clear to see that the total rate obtained by ldd is not sensitive to the variation of the resource hold by macro bs .
this result is also related to the solving process in which two phases are needed when employing a ldd based algorithm . as we have discuss in the introduction section , a solution which can satisfy the basic qos requirement
will be accepted by the ldd based algorithm .
it finally affects the allocation of left resource at marco bs . as a result
, the algorithm easily falls into the local optima .
this problem , to some degree , can be overcome by the ecav model since there is only one phase in the model . with the ecav model
, a constraint satisfied problem is transformed into a constraint optimizaiton problem . and
the advantage of dcop is successfully applied into solving user assocation problem .
an important breakthrough in this paper is that we take the dcop into the application of hetnet .
more preisely , we propose an ecav model along with a parameter @xmath0 to reduce the number of nodes and constraints in the model .
in addition , a markov basesd algorithm is applied to balance the quality of solution and the time consumed . from experimental results
, we can draw a conclusion that the quality of the solution obtained by the ecav-3 model solved with the mc based algorithm is better than the centralized algorithm , max - sinr and distributed one ldd , especially when the number of users increases but they are limited to the available rbs . in future work
, we will extend our research to the following two aspects : in some algorithms , like k - opt @xcite and adopt @xcite for dcop , there are already a theoretial analysis on the completeness of solution .
however , it is still a chanllenge job in most research of dcop algorithm , like the mc based algorithm proposed in this paper .
thus , we will explore the quality of the solution assoicated with different values of @xmath0 . in practice ,
the bss in small cells ( like pico / femto bss ) have properties of plug - and - play .
they are generally deployed in a home or small business where the environment is dynamic . in this way
, we should design a dcop model which is fit for the variations in the environment such as the mobility of users and different states ( active or sleep ) of bss . to this end
, a stochastic dcop model can be considered like the one in .
10 f. fioretto , f. campeotto , a. dovier , e. pontelli , and w. yeoh , large neighborhood search with quality guarantees for distributed constraint optimization problems , in _ proceedings of the 2015 international conference on autonomous agents and multiagent systems_.1em plus 0.5em minus 0.4eminternational foundation for autonomous agents and multiagent systems , 2015 , pp .
18351836 .
f. fioretto , w. yeoh , and e. pontelli , a dynamic programming - based mcmc framework for solving dcops with gpus , in _ international conference on principles and practice of constraint programming_.1em plus 0.5em minus 0.4emspringer , 2016 , pp .
813831 .
t. le , f. fioretto , w. yeoh , t. c. son , and e. pontelli , er - dcops : a framework for distributed constraint optimization with uncertainty in constraint utilities , in _ proceedings of the 2016 international conference on autonomous agents & multiagent systems_.1em plus 0.5em minus 0.4eminternational foundation for autonomous agents and multiagent systems , 2016 , pp . 606614 .
w. yeoh , p. varakantham , x. sun , and s. koenig , incremental dcop search algorithms for solving dynamic dcops , in _ the 10th international conference on autonomous agents and multiagent systems - volume 3_.1em plus 0.5em minus 0.4eminternational foundation for autonomous agents and multiagent systems , 2011 , pp .
10691070 .
g. mao and b. d. anderson , graph theoretic models and tools for the analysis of dynamic wireless multihop networks , in _ 2009 ieee wireless communications and networking conference_.1em plus 0.5em minus 0.4emieee , 2009 , pp .
a. a. kannan , b. fidan , and g. mao , robust distributed sensor network localization based on analysis of flip ambiguities , in _ ieee globecom 2008 - 2008 ieee global telecommunications conference_.1em plus 0.5em minus 0.4emieee , 2008 , pp .
k. kinoshita , k. iizuka , and y. iizuka , effective disaster evacuation by solving the distributed constraint optimization problem , in _ advanced applied informatics ( iiaiaai ) , 2013 iiai international conference on_.1em plus 0.5em minus 0.4emieee , 2013 , pp .
399400 .
f. amigoni , a. castelletti , and m. giuliani , modeling the management of water resources systems using multi - objective dcops , in _ proceedings of the 2015 international conference on autonomous agents and multiagent systems_.1em plus 0.5em minus 0.4eminternational foundation for autonomous agents and multiagent systems , 2015 , pp .
821829 .
p. rust , g. picard , and f. ramparany , using message - passing dcop algorithms to solve energy - efficient smart environment configuration problems , in _ international joint conference on artificial intelligence _ ,
n. guan , y. zhou , l. tian , g. sun , and j. shi , qos guaranteed resource block allocation algorithm for lte systems , in _ proc .
conf . wireless and mob .
, netw . and commun .
( wimob ) _ , shanghai , china , oct .
2011 , pp . 307312 .
d. liu , l. wang , y. chen , m. elkashlan , k .- k .
wong , r. schober , and l. hanzo , user association in 5 g networks : a survey and an outlook ,
_ ieee communications surveys & tutorials _ , vol . 18 , no . 2 , pp .
10181044 , 2016 . q. ye , b. rong , y. chen , c. caramanis , and j. g. andrews , towards an optimal user association in heterogeneous cellular networks , in _ global communications conference ( globecom ) , 2012 ieee_.1em plus 0.5em minus 0.4emieee , 2012 , pp .
41434147 .
t. l. monteiro , m. e. pellenz , m. c. penna , f. enembreck , r. d. souza , and g. pujolle , channel allocation algorithms for wlans using distributed optimization , _ aeu - international journal of electronics and communications _ , vol .
66 , no . 6 , pp .
480490 , 2012 .
t. l. monteiro , g. pujolle , m. e. pellenz , m. c. penna , and r. d. souza , a multi - agent approach to optimal channel assignment in wlans , in _ 2012 ieee wireless communications and networking conference ( wcnc)_.1em plus 0.5em minus 0.4emieee , 2012 , pp .
26372642 .
m. vinyals , j. a. rodriguez - aguilar , and j. cerquides , constructing a unifying theory of dynamic programming dcop algorithms via the generalized distributive law ,
_ autonomous agents and multi - agent systems _ , vol . 22 , no . 3 , pp .
439464 , 2011 . | multi - agent systems ( mas ) is able to characterize the behavior of individual agent and the interaction between agents .
thus , it motivates us to leverage the distributed constraint optimization problem ( dcop ) , a framework of modeling mas , to solve the user association problem in heterogeneous networks ( hetnets ) .
two issues we have to consider when we take dcop into the application of hetnet including : ( i ) how to set up an effective model by dcop taking account of the negtive impact of the increment of users on the modeling process ( ii ) which kind of algorithms is more suitable to balance the time consumption and the quality of soltuion . aiming to overcome these issues ,
we firstly come up with an ecav-@xmath0 ( each connection as variable ) model in which a parameter @xmath0 with an adequate assignment ( @xmath1 in this paper ) is able to control the scale of the model .
after that , a markov chain ( mc ) based algorithm is proposed on the basis of log - sum - exp function .
experimental results show that the solution obtained by dcop framework is better than the one obtained by the max - sinr algorithm . comparing with the lagrange dual decomposition based method ( ldd ) ,
the solution performance has been improved since there is no need to transform original problem into a satisfied one .
in addition , it is also apparent that the dcop based method has better robustness than ldd when the number of users increases but the available resource at base stations are limited . |
considerable interest has been recently devoted in finding exact solutions to schrdinger equations involving known potentials when the mass is position - dependent ( pdm ) . among them
, one may mention the morse and coulomb potentials @xcite .
moreover , it has been recently shown @xcite that to lowest order of perturbation theory , there exists a whole class of hermitian position - dependent - mass hamiltonians that are associated with pseudo - hermitian hamiltonians .
a great deal of interest has been paid to the interplay between these pseudo - hermitian pt - symmetric hamiltonians and their equivalent hermitian representations @xcite . in particular , mostafazadeh @xcite has considered the transition to the classical limit by showing that the relevant classical hamiltonian for the pt - symmetric cubic anharmonic oscillator plus a harmonic term , produces a behavior similar to a point particle with position - dependent - mass interacting with a quartic harmonic oscillator . indeed , many physical settings exist in which the effective mass can in principle depend on position .
for example , wang et al .
@xcite have recently shown that the schrdinger equation for a thin charged shell moving under the influence of its own gravitational field may be viewed as a position - dependent - mass problem .
displacement operators have already been introduced for systems with position - dependent - mass , for null or constant potentials from which generalized forms of the momentum operator have been obtained @xcite . in this contribution
, we demonstrate the possibility of transforming via similarity transformations , a position dependent mass hamiltonian into a hamiltonian with constant ( unity ) mass . by doing so
, these hamiltonians can then be solved ( if integrable ) using well - known techniques from quantum mechanics .
if on the other hand the potentials are not solvable , perturbative methods may be applied for their solution . in order to achieve this objective , we use aspects associated with some non - classical states of the harmonic oscillator , namely , squeezed states @xcite . for squeezed states ,
the uncertainty may be `` squeezed '' in one of the quadratures , while in the other canonical conjugate variable the uncertainty increases .
in what follows , we will first show how the constant mass may be eliminated from the kinetic energy in a hamiltonian . in this regard , consider the hamiltonian @xmath0 where the mass particle is @xmath1 and @xmath2 .
this hamiltonian is in turn transformed using the squeeze unitary operator @xcite @xmath3.\ ] ] to find how the operator @xmath4 transforms the position and the momentum operators , the hadamard lemma @xcite is used ; i.e. , that @xmath5 + \frac{1}{2!}\left [ \hat{a},\left [ \hat{a},\hat{b}\right ] \right]+ \frac{1}{3!}\left [ \hat{a},\left [ \hat{a},\left [ \hat{a},\hat{b}\right]\right ] \right]+ ... $ ] , from which we obtain that @xmath6 as a result , the transformed hamiltonian takes the form @xmath7 and thus the mass has been effectively eliminated from the kinetic energy term . based on this latter possibility
, one could ask if the mass can also be eliminated from the kinetic energy via a proper transformation , even if it is position dependent .
there is always some uncertainty as to the actual form of the kinetic energy term in a hamiltonian , when the mass is position dependent .
this is because @xmath8 no longer commutes with the momentum .
there are consequently several ways to write the kinetic part of the hamiltonian that must be kept hermitian ; for instance @xmath9 on the other hand , by choosing @xmath10 , we arrive to the ordering proposed by bendaniel and duke @xcite , @xmath11 while with the choice @xmath12 , @xmath13 , we get @xmath14.\ ] ] although there is no apparent reason in selecting any particular ordering for the kinetic position - dependent - mass hamiltonian , here we will choose to work with the bendaniel and duke proposal .
physical arguments supporting this choice were put forward by lvy - leblond @xcite .
we now consider the complete quantum hamiltonian of a particle with position - dependent mass @xmath15 we then use the transformation @xmath16 with @xmath17 \right\rbrace,\ ] ] where @xmath18 is a well behaved function that will depend on position .
using the hadamard lemma @xcite , one can show that the momentum operator transforms according to @xmath19,\ ] ] where @xmath20 for which @xmath21 on the other hand , for the position operator , we obtain @xmath22 where @xmath23 with @xmath24 from equation ( [ 11 ] ) , we note that @xmath25 from the above equations , we can then write @xmath26 where the transformed potential @xmath27 is given by @xmath28,\ ] ] and where @xmath29 up to this point , we have succeeded in eliminating the position dependency of the mass .
note that both hamiltonians , @xmath30 and @xmath31 have the same sets of eigenvalues since they are related by a similarity transformation .
therefore , by finding the eigenvalues of @xmath31 we can directly obtain the eigenvalues corresponding to the position dependent mass hamiltonian @xmath30 .
let us consider a mass that decays with the position in an exponential - like fashion ; i.e. , let @xmath32 figure 1 , depicts this mass dependence on position when @xmath33 , and for three different values of the parameter @xmath34 . ) for @xmath33 , and @xmath35 . ]
this particular dependence of the mass on position suggests the auxiliary function @xmath36 , in which case the similarity transformation takes the form @xmath37 .\ ] ] from here one finds that @xmath38 and @xmath39 that is consistent with @xmath40 .
+ with this particular choice for a position dependent mass ( [ 20 ] ) , we also choose the following potential @xmath41 with real arbitrary coefficients . if @xmath42 , @xmath43 , and @xmath44 , the transformed potential function @xmath27 is given by the morse potential @xmath45 ^ 2,\ ] ] where @xmath46 and @xmath47 . + in figure 2 , we plot the original potential ( [ 24 ] ) as a function of the position for the same values as in figure 1 , when @xmath48 . ) for @xmath33 , @xmath35 , and @xmath48 . ]
the solution of the transformed equation , can now be obtained given that the mass is constant and the potential involved is of the morse type that is known to admit analytical solutions .
by means of a squeeze - like unitary transformation , we have related a position dependent mass hamiltonian to a hamiltonian with constant mass . by doing so
, we can use standard methods for analyzing quantum mechanical problems of this type .
importantly , the eigenvalues of the transformed hamiltonian are the same as those associated with the original position dependent problem .
meanwhile the eigenfunctions of these two hamiltonians are related by a similarity transformation - given by the squeeze - like operator ( [ 10 ] ) .
a. de souza dutra and c. a. s. almeida , phys .
lett . a 275 , 25 ( 2000 ) .
b. bagchi , a. banerjee , c. quesne , czech .
56 , 893 ( 2006 ) .
b. roy and p. roy , j. phys .
a 35 , 3961 ( 2002 ) .
a. d. alhaidari , phys .
a 66 , 042116 ( 2002 ) .
l. dekar , l. chetouani , and t. f. hammann , j. math .
39 , 2551 ( 1998 ) .
v. milanovic and z. ikonic , j. phys . a 32 , 7001 ( 1999 ) b. bagchi , p. gorain , c. quesne , and r. roychoudhury , mod .
lett . a 19 , 2765 ( 2004 ) . c. quesne and v. m. tkachuk , j. phys .
a 37 , 4267 ( 2004 ) .
j. yu , s .- h .
dong , and g .- h .
sun , phys .
lett . a 322 , 290 ( 2004 ) .
g. chen and z. chen , phys .
lett . a 331 , 312 ( 2004 ) .
a. r. plastino , a. puente , m. casas , f. garcias , and a. plastino , rev .
46 , 78 ( 2000 ) . c. quesne , k. a. penson , v. m. tkachuk revisiting maths - type q - deformed coherent states in : proceedings of the 8th international conference on squeezed states and uncertainty relations .
puebla , mexico - june 9 - 13 , 2003 .
h. moya - cessa , r. juregui , s. hacyan and o. castaos .
( rinton press , nj ) 336 - 341 ( 2003 ) .
b. gnl , b. gnl , d. tutcu , and o.zer , mod .
lett . a 17 , 2057 ( 2002 ) . r. ko , m. koca , and e. krck , j. phys . a 35 , l527 ( 2002 ) . b. bagchi , a. banerjee , c. quesne , and v. m. tkachuk , j. phys . a 38 , 2929 ( 2005 ) .
dong and m. lozada - cassou , phys .
lett . a 337 , 313 ( 2005 ) .
l. jiang , l .- z .
yi , and c .- s .
jia , phys .
lett . a 345 , 279 ( 2005 ) .
j. f. carinena , m. f. ranada , and m. santander , ann .
322 , 434 ( 2007 ) c. weisbuch and b. vinter , quantum semiconductor heterostructures ( academic , new york , 1997 ) .
a. mostafazadeh , j. phys .
a 38 , 6557 ( 2005 ) .
a. mostafazadeh and a. batal , j. phys . a 37 11645 , ( 2004 ) . h. f. jones , j. phys .
a 38 , 1741 ( 2005 ) .
p. k. ghosh , j. phys .
a 38 , 7313 ( 2005 ) .
b. bagchi , c. quesne and r. roychoudhury , j. phys .
a 38 , l647 ( 2005 ) .
wang , e. greenwood and d. stojkovic , phys .
d 80,124027 ( 2009 ) .
costa filho , m.p .
almeida , g.a .
farias and j.s .
andrade jr . , phys .
a 84 , 050102(r ) ( 2011 ) .
mazharimousavi , phys .
rev . a 85 , 034102 ( 2012 ) .
r. loudon and p.l .
knight , j. of mod .
optics 34 , 709 ( 1987 ) .
h. moya - cessa and a. vidiella - barranco , j. of mod .
optics 39 , 2481 ( 1992 ) .
h. m. moya - cessa and f. soto - eguibar , introduction to quantum optics , rinton press ( 2011 ) .
d.j . bendaniel and c.b .
duke , phys .
152 , 683 ( 1966 ) .
lvy - leblond , phys .
a 52 , 1845 ( 1995 ) . | we provide a squeeze - like transformation that allows one to remove a position dependent mass from the hamiltonian . methods to solve the schrdinger equation
may then be applied to find the respective eigenvalues and eigenfunctions .
as an example , we consider a position - dependent - mass that leads to the integrable morse potential and therefore to well - known solutions . |
historically , the thick disc of the milky way was first identified through star counts toward the galactic poles as a vertically extended disc component with scale height @xmath5 1 kpc @xcite , in addition to the thin disc with scale height @xmath6 that dominates the disc stars in the immediate solar neighbourhood .
therefore , at the outset , the tentative definition of the thick - disc stars was those disc stars with large vertical orbital excursions and large vertical velocities . later
, spectroscopic studies ( e.g. , @xcite ) on kinematically - selected disc stars suggested that stars with large vertical motions ( which are likely to belong to the thick disc ) tend to show lower [ fe / h ] and higher [ @xmath7/fe ] than those with small vertical motions ( thin - disc stars ) .
these chemical properties suggest that the thick - disc stars are older than thin - disc stars ( lower [ fe / h ] ) and that the star formation timescale of the thick disc was shorter than that of the thin disc ( higher [ @xmath7/fe ] ) .
recently , klaus fuhrmann @xcite investigated a kinematically - unbiased volume - complete sample of solar - type disc and halo stars located within 25 pc of the sun .
the distribution of his sample stars in the [ fe / h]-[mg / fe ] space reveals two chemically distinct populations of disc stars ( see figure 15 of @xcite ) .
based on the kinematical properties of these populations , he identified the lower-[fe / h ] and higher-[mg / fe ] population to be the thick disc , and the other population to be the thin disc .
this clear separation between the thin and thick discs is also confirmed in the nearby ( heliocentric distance @xmath8 100 pc ) kinematically - unbiased sample of @xcite for which [ fe / h ] and [ @xmath7/fe ] are available through high - resolution spectroscopy .
these recent observations suggest that the thick disc is better defined by chemical compositions of stars , especially by [ fe / h ] and [ @xmath7/fe ] @xcite . in the past decades
, our understanding of the structure of the thick disc towards the galactic poles has been greatly improved @xcite .
the next step forward is to unravel its more global structure , such as the radial dependence of its vertical structure .
so far , many attempts have been made to fulfil this goal , and they are broadly categorised into two classes of studies . the first class of studies are based on nearby samples of disc stars .
one such example is @xcite , who fitted the distribution of local disc stars with his distribution function model .
he discussed some global structure of the thick disc by looking into his best - fit models .
although this kind of studies can predict almost everything if the best - fit models are reasonably correct , one critical problem with these studies is the validity of the functional forms of the assumed distribution functions . the second class of studies are based on in - situ samples of ( relatively ) distant disc stars .
this class can be further categorised into three sub - classes : those studies using ( i ) high - resolution spectroscopic samples ; ( ii ) medium - resolution spectroscopic samples ; or ( iii ) photometric samples .
the advantage of sub - class ( i ) studies is that we can define the thick disc purely by means of the stellar chemistry @xcite .
however , the number of stars that are currently available is less than a few hundred , and this small sample size makes it hard to obtain some statistical properties of distant disc stars
. also , since the errors in distance and proper motion are large , kinematical analyses are difficult for these stars . in the sub - class ( ii ) studies , much larger sample of stars are available than in the sub - class ( i ) studies .
a recent example of this sub - class is @xcite , who studied the density distribution of chemically defined disc populations by fitting the sdss / segue data with analytic models of the density profile .
however , since the chemical abundances of these samples are less accurate than high - resolution samples , some blending of the thin- and thick - disc stars is inevitable in this kind of studies . most of the sub - class ( iii ) studies are based on the star count method ( e.g. , @xcite ) .
photometric samples are advantageous in grasping the global structure of the stellar disc because the available sample size is the largest among these subclasses , and because the sample stars cover a wide range of heliocentric distance . however , since the photometric data lack chemical information for each sample star , it is not possible to separate the sample into stellar components .
this inseparability means that one needs to adequately model all the stellar components that contribute to the sample , which is harder than adequately modelling the thick disc only . by taking into account these problems , in this paper
we explore the possibility of constraining the global structure of the thick disc based on a small number of nearby chemically - selected stars , but not relying on analytic modelling of the thick disc . specifically , we reconstruct the distribution function of the thick disc within a certain portion of the action space which is accessible from local observations .
this paper is organised in the following manner .
first , we describe our sample stars in section [ section : sample ] . in section [ section : theory ] , we introduce the concepts of the observable action space @xmath9 and the observable distribution function @xmath10 . there , we discuss the possibility of inferring some information on the velocity dispersion of the thick disc by using @xmath10 . in section [ section : method ] , we present our method to reconstruct @xmath10 from a local sample of thick - disc stars . in section
[ section : analysis ] , we present our main results .
the implications from our study are discussed in section [ section : discussion ] , and section [ section : conclusion ] sums up our analyses .
in this paper , we assume that the milky way potential is axisymmetric and adopt the usual galactocentric cylindrical coordinate system @xmath11 to describe the three - dimensional positions of stars in the milky way .
the three - dimensional velocities along the respective coordinates are denoted by @xmath12 , where @xmath13 is positive outwards , @xmath14 is positive in the direction of galactic rotation , and @xmath15 is positive in the direction of the north galactic pole .
we assume that the local standard of rest ( lsr ) is in a circular orbit with a rotation speed of @xmath16 @xcite and that the galactocentric distance of the sun is @xmath17 .
we also adopt the peculiar motion of the sun with respect to the lsr determined by @xcite . in this paper , we assume that the thick disc of the milky way is a single stellar population and define the thick - disc stars to be those stars with @xmath18 } \leq -0.2 $ ] and @xmath19 } \geq 0.2 $ ] , by taking into account the solar neighbour observations of disc stars ( e.g. , @xcite ) .
we constructed a kinematically - unbiased sample of 127 thick - disc stars , for which accurate position , velocity , [ fe / h ] , and [ mg / fe ] are available .
84 stars in our sample were taken from @xcite , and the other 43 stars were taken from @xcite .
most of our sample stars are distributed within 100 pc from the sun .
@xcite published a kinematically - unbiased sample of 1111 stars , for which accurate [ fe / h ] and [ mg / h ] estimated from high - resolution spectroscopic observations as well as the three dimensional cartesian velocity based on hipparcos observations are available .
we derived the distances and the line - of - sight velocities for 1103 stars in their sample that were also included in the updated hipparcos catalog of @xcite .
in the course of this inversion , we took into account the solar peculiar motion adopted in @xcite and we used the three - dimensional cartesian velocity taken from @xcite and the position on the sky and the proper motion taken from @xcite .
we then calculated the three - dimensional velocity @xmath12 in our coordinate system , by using our assumption on @xmath20 . from these 1103 stars , we selected those stars that satisfy both the thick - disc chemical criteria ( @xmath18 } \leq -0.2 $ ] and @xmath19 } \geq 0.2 $ ] ) and the kinematical criteria of @xmath21 and @xmath22 .
the latter velocity cuts were applied to minimise possible contamination from halo stars . finally , we discarded those stars that had been flagged as non - single stars , and we obtained 84 stars .
@xcite published a catalog of 743 stars compiled from 11 catalogs of nearby stars and listed the chemical abundances of these stars . among these 11 catalogs ,
those catalogs taken from @xcite , @xcite , @xcite , @xcite , and @xcite are kinematically unbiased , so we first selected those stars that are included in either of these five catalogs .
then , we excluded those stars that happened to be included in @xcite .
the position and velocity of these stars were then calculated either by using the data from genova - copenhagen survey @xcite or by using the updated hipparcos catalog @xcite and simbad database .
finally , as in the previous subsection , we applied the chemical and velocity criteria and obtained 43 thick - disc stars .
as described above , we adopted a pair of velocity cuts to exclude halo stars . rigorously speaking
, these cuts potentially could have introduced some kinematical bias on our analyses .
however , we have confirmed that adopting a looser velocity constraint of @xmath23 ( without cuts on @xmath13 or @xmath15 ) has only negligible effects on our final results .
the distribution of our sample stars in the [ fe / h]-[mg / fe ] plane is shown in figure [ fig : feh_mgfe ] .
the uncertainties in the chemical abundances are small enough ( typical uncertainties are @xmath24 } = $ ] 0.03 dex and @xmath25 } = $ ] 0.05 dex for the @xcite sample ) compared with the scatter in this diagram .
therefore , our sample does not suffer from severe contamination from low-[mg / fe ] stars .
figure [ fig : velocity_distribution ] shows the velocity distribution of our sample in the @xmath26 , @xmath27 , @xmath28 and @xmath29 planes .
our sample stars show a mean velocity of @xmath30 , and thus lag from the lsr by nearly @xmath31 .
the three - dimensional velocity dispersion is @xmath32 . in figure
[ fig : separability ] , we show the value of @xmath3 as functions of @xmath14 and @xmath33 . to make these plots , we first sort the sample in @xmath14 or @xmath33 , evaluate the dispersion @xmath3 for a binned sample of @xmath34 stars , and move through the sample in steps of 4 stars ( so that any pair of adjacent bins share 36 stars ) .
the derived value of @xmath3 for each bin is plotted against the median value of @xmath14 or @xmath33 of the bin .
although the bin size of 40 stars is rather large compared with the total sample size of 127 stars , the result of this figure suggests that there is no strong trend in @xmath3 as a function of @xmath33 or @xmath14 .
this property suggests that the vertical kinematics ( @xmath3 ) is not strongly correlated with the in - plane motion ( @xmath33 or @xmath14 ) .
+ + + + +
most of the thick - disc stars in the solar neighbourhood are on non - circular orbits .
more specifically , the majority of nearby thick - disc stars have guiding radii smaller than the galactocentric distance of the sun ( @xmath35 ) , due to the asymmetric drift @xcite . therefore ,
if the thick disc is a static stellar system , kinematics of nearby thick - disc stars carry some information on the thick disc outside the solar neighbourhood , especially on the inner part of the thick disc .
of course , nearby thick - disc stars do not carry all the information of the thick disc , since these stars are inherently biased in that they represent a portion of the thick - disc stars whose orbits penetrate through the solar neighbourhood .
in other words , a nearby sample only covers a certain fraction of action space @xmath9 where the corresponding orbits bring stars to the solar neighbourhood @xcite .
hence , the best we can do with nearby samples is to reconstruct the distribution function @xmath36 within this _ observable action space _ @xmath9 , i.e. to reconstruct the _ observable distribution function _ defined by @xmath37
we here note that @xmath38 is different from the as - observed action distribution . the fraction of its orbital period that a given star spends in the solar neighbourhood depends on its orbit , so that the probability of finding the star in the solar neighbourhood depends on its action . in section [
section : method ] , we take into account this probability to reconstruct the observable distribution function @xmath38 . in order to have a physical insight into @xmath9 , let us consider the velocity space that corresponds to @xmath9 by assuming a potential model of the milky way .
figure [ fig : observable ] shows the allowed range of @xmath4 for disc stars located at @xmath39 kpc with zero vertical velocity to reach the solar neighbourhood . since only a portion of the disc stars at @xmath40 kpc can reach the solar neighbourhood , what we can learn from @xmath10 about the thick disc at @xmath40 kpc is the velocity distribution of stars within the ` observable ' region in figure [ fig : observable ] . with the information on @xmath10 , we can estimate the _ observable vertical velocity dispersion _ defined by @xmath41 at any given radius @xmath42 on the disc plane .
this dispersion corresponds to the vertical velocity dispersion of those stars located at @xmath42 whose velocities are within the observable region . thus @xmath43 is conceptually different from the ( ordinary ) vertical velocity dispersion given by @xmath44 which is the vertical velocity dispersion of all the stars located at @xmath42 .
-space for disc stars at @xmath45 kpc with @xmath46 to reach the solar neighbourhood , calculated for the potential model in [ section : analysis ] .
the boundary curve between the observable region and unobservable region corresponds to @xmath47 ( @xmath48 is the apocentric distance of the orbit ) , and thus the circular orbit at @xmath40 kpc ( red dot ) lies within the unobservable region . when @xmath49 , the boundary curve slightly shifts downwards in this diagram .
however , this difference is not significant for a typical value of @xmath50 of the thick - disc stars . in this plot
, we assume a realistic potential model of the milky way described in section [ section : analysis ] . ]
here we describe how to reconstruct @xmath38 of the thick disc from a nearby sample of thick - disc stars , based on the idea of @xcite .
we note that @xcite investigated a rigorous treatment of an as - observed distribution function in which survey selection function is taken into account ( see appendix ) .
we do not follow their formulation here since we do not know the exact survey selection function for our sample stars ( but we do know that our sample stars are not kinematically selected ) .
although we do not take into account the survey selection function , our approach has an advantage that we do not have to assume the functional form of the distribution function . in the following ,
we regard disc stars as test particles moving in the milky way potential .
also , we assume that we have a kinematically - unbiased sample of @xmath51 thick - disc stars which happen to be passing near the sun . under an axisymmetric potential @xmath52 , each star has three actions @xmath53 @xcite .
let us denote the actions of the @xmath54th star as @xmath55 .
then , the time - averaged distribution function of a single orbit characterised by @xmath55 is expressed as @xmath56 and the corresponding spatial density is given by @xmath57 the value of @xmath58 can be interpreted as the probability density of finding a star with action @xmath55 at position @xmath59 , since @xmath60 @xcite . by following @xcite , we approximate the observable distribution function @xmath10 to be the weighted sum of single orbit distribution functions given by @xmath61 here , @xmath62 is a selection index that determines whether the @xmath54th star should be included ( @xmath63 ) or not ( @xmath64 ) , which shall be introduced to stabilise our analyses .
the factor @xmath65 is the orbit weighting factor to be determined .
although @xcite evaluated @xmath66 through bayesian statistics , here we numerically evaluate @xmath67 and simply set @xmath68 .
this simple choice of @xmath66 corresponds to a limiting case where all of our sample stars have identical position @xcite .
of finding a given star in the solar neighbourhood is equal to the fraction of its orbital period to stay in the solar neighbourhood .
thus , if such a star is observed in the solar neighbourhood , it represents @xmath69 stars in the same orbit ( including those with different orbital phases ) .
] since most of our sample stars are located within 100 pc away from the sun , our simple choice of @xmath66 is well justified .
once @xmath10 is reconstructed , we can estimate the observable vertical velocity dispersion @xmath43 at a given point on the disc plane , by using equation ( [ eq : def_sigz_fobs ] ) .
it can be shown that @xmath70 if the assumed potential is correct and a large enough number of local sample stars are available .
thus , we can perform a sanity check of the adopted potential by seeing if these dispersions reasonably agree with each other .
we adopt a realistic potential model of the milky way , consisting of halo , bulge , and disc components . in our model ,
the halo potential is given by the navarro - frenk - white potential @xcite of the form @xmath71 where @xmath72 is the galactocentric distance , @xmath73 , and @xmath74 .
the bulge potential is given by @xmath75 @xcite .
we adopt @xmath76 and @xmath77 .
we have confirmed that adopting a slightly flattened bulge potential hardly affects our results . for the disc potential
, we assume the miyamoto - nagai potential @xcite given by @xmath78 where @xmath79 , @xmath80 , and @xmath81 .
the values of parameters are chosen by using the publicly available code of `` @xcite , while fixing @xmath17 , @xmath82 , and @xmath77 .
after setting up the potential models , we derived the weighting factor for the @xmath54th star , @xmath66 , in the following manner .
we first integrated the orbit of the @xmath54th star forward in time for a long enough period .
we then calculated the fraction @xmath83 of the integration time which the star spent in a volume defined by @xmath84 , @xmath85 , and @xmath86 . here , we denote the observed position of the @xmath54th star on the meridional plane as @xmath87 and we adopted @xmath88 . by taking into account the relation @xmath89 , we derived @xmath90 and set @xmath91 .
we set @xmath64 for those stars that satisfy @xmath92 , where @xmath93 and @xmath94 are the mean and the standard deviation of @xmath95 , respectively .
this final procedure was included to prevent our results being affected by small number of stars with large @xmath66 .
figure [ fig : sigz_data ] shows the reconstructed profile of @xmath43 .
here we use an approximation of @xmath96 we adopt @xmath97 and calculate @xmath98 at @xmath99 in steps of @xmath100 .
the radial range of @xmath42 is set so that more than 42 stars ( i.e. , more than @xmath101 stars ) have non - zero contribution to the integrals in equation ( [ eq : approximate_sigz_realistic ] ) , i , e . , have positive values of @xmath102 . in figure
[ fig : sigz_data ] , the red dots show the reconstructed values of @xmath103 , and the shaded region represents the uncertainty which is derived from a bootstrap resampling of the contributing stars .
the velocity dispersion declines nearly exponentially as a function of @xmath42 , and the best fit exponential profile of @xmath43 is shown with a black straight line .
we find that the best - fit scale length defined by @xmath104 is @xmath105 .
the random error is the uncertainty in the fit and the systematic error is estimated by performing mock catalog analyses .
we interpolate the values of @xmath43 at @xmath106 kpc and @xmath107 kpc to obtain @xmath108 .
this value is in agreement with the local value of @xmath109 in our sample , indicating that the adopted potential is reasonable at least in the solar neighbourhood .
we vary the parameters @xmath110 within the range of @xmath111 and @xmath112 and find that the nearly exponential profile of @xmath43 is a generic feature independent of the assumed parameters .
@xmath113 mildly depends on these parameters , but it only increases from 7.5 kpc to 9.4 kpc when we vary @xmath114 from 2 kpc to 4 kpc ; and from 8.1 kpc to 9.1 kpc when we vary @xmath115 from 0.15 kpc to 0.5 kpc .
+ as discussed in section [ subsection : observable_dispersion ] , @xmath116 represents the vertical velocity dispersion of stars located at @xmath117 whose velocity is within the observable region in the velocity space ( figure [ fig : observable ] ) . in order to see the variation of the vertical kinematics within the observable region , we create the following groups of stars from our sample : ( 1 ) those stars with @xmath118 ( @xmath119 stars ) ; and ( 2 ) those stars with @xmath120 ( @xmath121 stars )
here , we define @xmath122 and @xmath48 as the minimum and maximum value of @xmath42 for each star s orbit .
these samples happen to be non - overlapping , and the observed vertical velocity dispersion for groups 1 and 2 are respectively @xmath123 and @xmath124 .
all of these stars spend some time at @xmath125 , so in figure [ fig : group12](a ) , we show the distribution of @xmath33 and @xmath14 when they pass through this location .
the stars in groups 1 and 2 are distributed in distinct areas in this velocity space , reflecting the different orbital eccentricities . in figure
[ fig : group12](b ) , we show the vertical velocity dispersion @xmath116 of these groups , calculated by using @xmath126 derived in section [ section : analysis ] .
we see that the radial profiles of @xmath116 for these groups are quite similar to each other at @xmath127 .
this result suggests that the vertical kinematics of the stars in the observable region is more or less homogeneous at least at @xmath127 . ) and group 2 ( @xmath120 ) in the @xmath128 plane when they pass through @xmath129 kpc .
( b ) the reconstructed profile of @xmath43 for these groups . as a reference
, we show the exponential profile shown in figure [ fig : sigz_data ] ( black solid line ) .
, title="fig : " ] + +
as seen in figures [ fig : velocity_distribution](a ) , [ fig : velocity_distribution](b ) , and [ fig : separability ] , the vertical velocity dispersion of the thick disc does not have a strong dependency on the in - plane velocities ( @xmath13 and @xmath14 ) in the solar neighbourhood . in order to have some physical insight into this finding ,
let us suppose , as a working hypothesis , that this apparent homogeneity of vertical velocity distribution across the @xmath4-space is the case not only near the sun but also outside the solar vicinity
. then it is implied that the distribution of the vertical action @xmath130 is not strongly dependent on the radial action ( @xmath131 ) or azimuthal action ( @xmath132 ) .
it follows that there is a typical value of vertical action @xmath133 that characterises the vertical motions of the thick - disc stars . on the other hand ,
the maximum height @xmath134 above the disc plane that a given thick - disc star can reach at galactocentric radius @xmath42 is determined mainly by its vertical action @xmath130 . in realistic potential models of the milky way
, @xmath135 is an increasing function of @xmath42 , since the gravitational pull towards the disc plane is weaker at larger @xmath42 .
thus , if the distribution of @xmath130 of the thick - disc stars is nearly independent of @xmath136 , then the thick disc needs to flare and the scale height of the thick disc roughly follows the radial behaviour of @xmath137 . then , under what circumstances does the distribution of @xmath130 become ( approximately ) independent of @xmath136 ? here
we consider such an example .
let us suppose that the thick disc was formed through radial migration @xcite .
some numerical simulations suggest that the radial migration conserves the stellar vertical action @xmath130 as the star changes its in - plane motion @xcite .
thus , radial migration redistributes the stars in @xmath136-space while keeping @xmath130 unchanged .
this redistribution in the action space is expected to erase the initial dependency of @xmath130-distribution on @xmath136 @xcite , unless the initial dependency is too strong to erase . if we take into account that the thick - disc stars are generally very old ( @xmath138 gyr ; @xcite ) , it may be likely that the radial migration ends up with a homogeneous distribution of @xmath130 as a function of @xmath131 or @xmath132 .
it is of interest to note that some numerical simulations predict that the thick disc flares if it was formed though radial migration @xcite .
if @xmath3 is really independent of @xmath4 anywhere on the disc plane , we have @xmath139 . in this case
we can estimate the thickness @xmath140 of the thick disc at galactocentric radius @xmath42 by using our reconstructed profile of @xmath43 . as an example
, we define the thickness @xmath141 by @xmath142 where @xmath52 is the galactic potential . by plugging @xmath43 into @xmath3 in equation ( [ eq : def_h ] )
, we obtain @xmath140 shown in figure [ fig : h_data ] .
although the @xmath140 profile mildly depends on the adopted potential , we have confirmed that @xmath141 increases as a function of @xmath42 independent of the potential parameters as long as they are within the range explored in section [ subsection : result_potential_dependece ] .
+ as we have discussed in the previous subsection , the radial dependence of the thickness of the thick disc is related to the dependency of the vertical kinematics on the in - plane motions . by taking into account some observational evidence of the thick - disc component in external galaxies with a constant scale height @xcite ,
let us consider what we would need to keep the scale height @xmath141 constant .
here we concentrate our discussion on the situation at a certain radius @xmath143 in the inner disc ( @xmath144 ) . in order to satisfy @xmath145 , we require @xmath146 so that the vertical velocity distribution is inhomogeneous in the @xmath4 space . since @xmath43 represents the vertical velocity distribution of observable stars ( those stars
whose velocity is inside the observable region ) and @xmath3 represents the vertical velocity distribution of all the stars , this requirement implies the following situation : in the inner disc , those thick - disc stars with large @xmath132 ( large guiding centre radii ) need to have smaller vertical action than those thick - disc stars with small @xmath132 and @xmath131 . with small @xmath132 to be unobservable , it has to have small @xmath33 ( see figure [ fig : observable ] ) and thus small @xmath131 .
] this situation may be attained if there are enough low - eccentricity , shell - like orbits with large vertical motions in the inner disc .
we have investigated an orbit - based method to reconstruct the observable distribution function @xmath10 , which describes the distribution of action within the portion of action space covered by a local disc sample . by applying our method to 127 chemically - selected thick - disc stars in the solar neighbourhood
, we found that the vertical velocity dispersion @xmath43 that corresponds to @xmath10 decreases nearly exponentially as a function of @xmath42 with a scale length of @xmath147 kpc .
we also found that the vertical velocity dispersion @xmath3 of the local thick - disc stars shows only weak dependency on the in - plane motions @xmath4 .
this apparent homogeneity of @xmath3 in @xmath4-space may be naturally explained if the distribution of the vertical action @xmath130 is only weakly dependent on the radial action @xmath131 and the azimuthal angular momentum @xmath132 .
one possible mechanism to produce such action distribution is radial migration , where the thick - disc stars are re - distributed in the @xmath136-space while keeping their @xmath130 unchanged @xcite .
kh is supported by japan society for the promotion of science ( jsps ) through a postdoctoral fellowship for research abroad .
kh thanks yuzuru yoshii , jason sanders and shigeki inoue for useful feedback on the original manuscript , and thanks jo bovy for making `` publicly available . also , the authors thank the referees for their fruitful comments . 99 adibekyan , v. z. , sousa , s. g. , santos , n. c. , et al .
2012 , a&a , 545 , a32 allende prieto , c. , barklem , p. s. , lambert , d. l. , & cunha , k. 2004 , a&a , 420 , 183 bensby , t. , feltzing , s. , & lundstrm , i. 2003 , a&a , 410 , 527 bensby , t. , feltzing , s. , & lundstrm , i. 2004 , a&a , 415 , 155 bensby , t. , alves - brito , a. , oey , m. s. , yong , d. , & melndez , j. 2011 , apj , 735 , l46 binney , j. 2010 , mnras , 401 , 2318 binney , j. 2012 , mnras , 426 , 1328 binney , j. , & mcmillan , p. 2011 , mnras , 413 , 1889 binney , j. , & tremaine , s. 2008 , galactic dynamics : second edition , by james binney and scott tremaine .
isbn 978 - 0 - 691 - 13026 - 2 ( hb ) . published by princeton university press , princeton , nj usa , 2008 .
bovy , j. , rix , h .- w . ,
liu , c. , et al .
2012 , apj , 753 , 148 bovy , j. , & rix , h .- w .
2013 , apj , 779 , 115 bovy , j. 2015 , apjs , 216 , 29 brook , c. b. , kawata , d. , gibson , b. k. , & freeman , k. c. 2004 , apj , 612 , 894 brook , c. b. , gibson , b. k. , martel , h. , & kawata , d. 2005 , apj , 630 , 298 chen , y. q. , nissen , p. e. , zhao , g. , zhang , h. w. , & benoni , t. 2000 , a&as , 141 , 491 chiba , m. , & beers , t. c. 2001 , apj , 549 , 325 comern , s. , elmegreen , b. g. , knapen , j. h. , et al .
2011 , apj , 741 , 28 edvardsson , b. , andersen , j. , gustafsson , b. , et al .
1993 , a&a , 275 , 101 fuhrmann , k. 1998 , a&a , 338 , 161 fuhrmann , k. 2004 , astronomische nachrichten , 325 , 3 fuhrmann , k. 2008 , mnras , 384 , 173 fuhrmann , k. 2011 , mnras , 414 , 2893 gilmore , g. , & reid , n. 1983 , mnras , 202 , 1025 gilmore , g. , wyse , r. f. g. , & kuijken , k. 1989 , ara&a , 27 , 555 haywood , m. , di matteo , p. , lehnert , m. d. , katz , d. , & gmez , a. 2013 , a&a , 560 , a109 hernquist , l. 1990 , apj , 356 , 359 holmberg , j. , nordstrm , b. , & andersen , j. 2009 , a&a , 501 , 941 jensen , e. b. , & thuan , t. x. 1982 , apjs , 50 , 421 kuijken , k. , & gilmore , g. 1989 , mnras , 239 , 571 loebman , s. r. , rokar , r. , debattista , v. p. , et al .
2011 , apj , 737 , 8 masseron , t. , & gilmore , g. 2015 , arxiv:1503.00537 mcmillan , p. j. , & binney , j. 2012 , mnras , 419 , 2251 may , a. , & binney , j. 1986 , mnras , 221 , 857 minchev , i. , famaey , b. , quillen , a. c. , et al .
2012 , a&a , 548 , a127 mishenina , t. v. , soubiran , c. , kovtyukh , v. v. , & korotin , s. a. 2004 , a&a , 418 , 551 miyamoto , m. , & nagai , r. 1975 , pasj , 27 , 533 navarro , j. f. , frenk , c. s. , & white , s. d. m. 1996 , apj , 462 , 563 reddy , b. e. , tomkin , j. , lambert , d. l. , & allende prieto , c. 2003 , mnras , 340 , 304 reddy , b. e. , lambert , d. l. , & allende prieto , c. 2006 , mnras , 367 , 1329 reid , m. j. , menten , k. m. , brunthaler , a. , et al .
2014 , apj , 783 , 130 rix , h .- w . , & bovy , j. 2013 , a&a rev . , 21 , 61 rokar , r. , debattista , v. p. , stinson , g. s. , et al .
2008 , apj , 675 , l65 rokar , r. , debattista , v. p. , & loebman , s. r. 2013 , mnras , 433 , 976 robin , a. c. , reyl , c. , fliri , j. , et al .
2014 , a&a , 569 , aa13 sanders , j. l. , & binney , j. 2015 , mnras , 449 ,
3479 sellwood , j. a. , & binney , j. j. 2002 , mnras , 336 , 785 schnrich , r. , & binney , j. 2009 , mnras , 396 , 203 schnrich , r. , & binney , j. 2009 , mnras , 399 , 1145 schnrich , r. , binney , j. , & dehnen , w. 2010 , mnras , 403 , 1829 solway , m. , sellwood , j. a. , & schnrich , r. 2012 , mnras , 422 , 1363 sommer - larsen , j. , & zhen , c. 1990 , mnras , 242 , 10 soubiran , c. , & girard , p. 2005
, a&a , 438 , 139 van der kruit , p. c. , & searle , l. 1981 , a&a , 95 , 105 van leeuwen , f. 2007 , a&a , 474 , 653 yoshii , y. 1982 , pasj , 34 , 365 yoshii , y. 2013 , planets , stars and stellar systems .
volume 5 : galactic structure and stellar populations , 393
in our analyses , we followed the approach of @xcite to relate the observed kinematic data to the underlying distribution function .
we note here that our approach owes much to the fact that our sample stars are located at the immediate solar neighbourhood ( @xmath148 ) .
let us consider a sample of stars taken from an imaginary magnitude - limited survey in which all the stars with apparent magnitude brighter than a certain value are observed . in this case
, those sample stars at large distances are high - luminosity stars , while the sample covers a wider luminosity range for nearby stars .
thus , such a sample puts more weight on nearby stars .
if the kinematical properties are highly inhomogeneous within the surveyed region , this sample is kinematically biased . in general , if the surveyed region is large , the survey selection function has to be taken into account in order to estimate the kinematic properties of the system ( mcmillan & binney 2012 ) .
this means that our methodology is not applicable to a sample that covers a large volume unless the completeness is taken into account , since our method implicitly requires that the sample stars are fair representatives of the stellar population .
however , since our sample is distributed in a small region ( @xmath149 pc ) , it is safe to assume that the kinematical properties of the thick - disc stars are homogeneous inside this region ( the thick disc has a scale height of @xmath150 kpc and a scale length of @xmath151-@xmath152 kpc ) .
therefore , the effect from the ( poorly - known ) survey selection function is relatively small in our sample . | we explored a method to reconstruct the distribution function of the galactic thick disc within the action space where nearby thick - disc stars are distributed . by applying this method to 127 chemically - selected thick - disc stars in the solar neighbourhood
, we found that the vertical velocity dispersion that corresponds to the reconstructed distribution function declines approximately as @xmath0 at @xmath1 , with @xmath2 kpc .
also , we found that the vertical velocity dispersion @xmath3 of our local thick - disc stars shows only weak dependency on radial and azimuthal velocities @xmath4 .
we discuss possible implications of these results on the global structure of the milky way thick disc .
[ firstpage ] galaxy : disc galaxy : evolution galaxy : formation galaxy : kinematics and dynamics galaxy : structure solar neighbourhood |
i am very grateful to pauchy hwang and the organizers for their invitation and warm hospitality .
i also thank pauchy hwang and bingkan xue for the collaborated results in this talk .
this work is partially supported by national natural science foundation of china ( nos . 10721063 , 10575003 , 10528510 ) , by the key grant project of chinese ministry of education ( no .
305001 ) , and by the research fund for the doctoral program of higher education ( china ) .
j. r. horandel , j. phys .
g * 29 * , 2439 ( 2003 ) . c. a. ayre _ et al .
_ , j. phys .
a * 5 * , l102 ( 1972 ) .
t. hebbeker and c. timmermans , astropart .
phys . * 18 * , 107 ( 2002 ) .
s. ostapchenko , nucl .
suppl . * 151 * , 147 ( 2006 ) .
see , e.g. , w. d. apel _ et al .
_ [ kascade collaboration ] , arxiv : astro - ph/0510810 .
p. hansen , t. k. gaisser , t. stanev and s. j. sciutto , phys .
d * 71 * , 083012 ( 2005 ) ; a. cillis and s. j. sciutto , arxiv : astro - ph/9908002 . | the muon charge ratio of ultrahigh energy cosmic rays may provide information to detect the composition of the primary cosmic rays .
we propose to extract the charge information of high energy muons in very inclined extensive air showers by analyzing their relative lateral positions in the shower transverse plane .
the most high energy particles can be observed by human being are from cosmic rays .
the study of them belongs to frontiers of human knowledge in combination of cosmology , astrophysics , and particle physics , and can provide better understanding of the universe from most small to most big , i.e. , connecting quarks to the cosmos .
the universe is not empty , but full of background relic particles from the big bang .
it has long been anticipated that the highest energy cosmic rays would be protons from outside the galaxy , and there is an upper limit of the highest energy in the observed proton spectrum , commonly referred to as the gzk cutoff @xcite , as the protons traveling from intergalactic distances should experience energy losses owing to pion productions by the photons in the cosmic background radiation .
although there have been attentions for the cosmic ray events above the gzk cutoff , it is natural to expect that these ultrahigh energy cosmic rays come from sources within the gzk zone @xcite , i.e. , not far from us in more than tens of mpc .
recently there are also reports on the observation of the gzk cut - off by new experiments @xcite .
however , questions about the composition of such ultrahigh energy cosmic ray particles , e.g. , whether they are protons , neutrons , or anti - nucleons @xcite , are still open to investigations . muons in the air showers are mainly from decays of pions and kaons produced in the interactions of the primary cosmic rays with the atmosphere .
the very high energy secondary pion and kaon cosmic rays can be considered as from the current fragmentation of partons in deep inelastic scattering of the primary cosmic rays with the nucleon targets of the atmosphere in a first approximation @xcite .
we also consider only the favored fragmentation processes , i.e. , the @xmath0 , which is composed of valence @xmath1 and @xmath2 quarks , is from the fragmentation of @xmath1 and @xmath2 quarks in the nucleon beam , and the @xmath3 , which is composed of valence @xmath4 and @xmath5 quarks , is from the fragmentation of @xmath4 and @xmath5 quarks @xcite .
similarly , the @xmath6 , which is composed of valence @xmath1 and @xmath7 , is from the fragmentation of @xmath1 and @xmath7 quarks , and the @xmath8 , which is composed of valence @xmath4 and @xmath9 , is from the fragmentation of @xmath4 and @xmath9 quarks .
the @xmath10 is from the decay of a @xmath0 or a @xmath6 and the @xmath11 is from the decay of a @xmath3 or a @xmath8 .
we can roughly estimate the muon charge ratio by @xmath12+\kappa \left[u(x)+\bar{s}(x)\right]\right\}}{\int_0 ^ 1 { \mathrm d } x \left\{\left[d(x)+\bar{u}(x)\right]+\kappa \left[\bar{u}(x)+s(x)\right]\right\ } } , \label{qme}\ ] ] where @xmath13 is the quark distribution with flavor @xmath14 for the incident hadron beam and @xmath15 is a factor reflecting the relative muon flux and fragmentation behavior of @xmath16 .
secondary collisions do not influence the above estimation , since the current parton beams still keep their flavor content and act as the current partons after the strong interactions with the partons in the atmosphere targets . adopting a simple model estimation of the parton flavor content in the nucleon without any parameter @xcite , we find that @xmath17 for proton and @xmath18 for neutron .
this simple evaluation is in agreement with the empirical expectation of @xmath19 for proton and @xmath20 for neutron @xcite as well as that in an extensive monte carlo calculation @xcite , thus it provides a clear picture to understand the dominant features for the muon charge ratio by the primary hadronic cosmic rays . for the @xmath21 ratio for antiproton , it is equivalent to the @xmath22 ratio for proton by using eq .
( [ qme ] ) , thus we find @xmath23 for antiproton , which is close to that for neutron .
the @xmath21 ratio for antineutron is also equivalent to the @xmath22 ratio for neutron , and it is @xmath24 , which is close to that for proton .
it is hard to distinguish between the primary neutrons and antiprotons ( or protons and antineutrons ) by the @xmath21 ratio of the air shower , unless very high precision measurement is performed and also our knowledge of the muon charge ratio for each nucleon species is well established .
the study of cosmic rays with primary energies above @xmath25gev are typically based on the measurements of extensive air showers ( eas ) that they initiate in the atmosphere .
the ground detector array records the secondary particles produced in shower cascades , including photons , electrons ( positrons ) , muons , and some hadrons .
then their arrival times and density profiles are used to infer the primary energy and composition of the incident cosmic ray particle , usually through comparison with simulated results .
photons , electrons and positrons are the most numerous secondary particles in an eas event .
however , for very inclined showers , these electromagnetic components would travel a long slant distance and are almost completely absorbed before they reach the ground . on the other hand ,
muons are decay products of charged mesons in shower hadronic cascades .
most high energy muons survive their propagation through the slant atmospheric depth , during which they lose typically a few tens of gev s energy .
these high energy muons carry important information about the nature of the primary cosmic ray hadron , which will be extracted from their energy spectrum and lateral distribution .
as discussed by hwang and i @xcite , the ratio of positive versus negative muons @xmath26 is a significant quantity which can help to discern the primary composition , and at high energies this charge ratio also reflects important features of hadronic meson production in cosmic ray collisions . in order to obtain such muon charge information
, we would need a way to distinguish between positive and negative high energy muons .
unfortunately , existing muon detectors available at shower arrays , usually scintillators and water erenkov detectors , are not commonly equipped with magnetized steel to differentiate the muon charges .
even if they were , the limited region of the magnetic field prevents definite determination of high energy muons track curvature .
this invites us to think of the geomagnetic field as a huge natural detector for muon charge information . apparently , after being produced high in the atmosphere , a positively charged muon would bend east on its way down while a negatively charged muon would bend west , introducing an asymmetry into the density profile of the shower front . if their separation is large enough as compared with other circularly symmetric `` background '' deviations , it will be possible to distinguish the positive muons from the negative ones . to see such an effect , xue and i @xcite analyzed the possibility of obtaining the charge information of high energy muons in very inclined extensive air showers .
we have demonstrated that positive and negative high energy muons in sufficiently inclined air showers can be distinguished from each other through their opposite geomagnetic deviations in the transverse plane .
we developed a revised heitler model to calculate this distinct double - lobed distribution , and studied the condition for the two lobes of either positive or negative muons to be separable with confidence . from our criterion of resolvability
, we concluded that a zenith angle @xmath27 will be most suitable for our approach .
there are already some results from full air shower simulations that take into account the geomagnetic effect on muon propagation @xcite .
they illustrated remarkable double - lobed muon lateral density profile in very inclined air showers , which is in agreement with our expectation qualitatively .
however , no present study has fully considered the high energy part of muon content , which can be used to compare with our results .
thus we would like to propose future simulations of very inclined extensive air showers that focus on the behavior of high energy muons .
they also have to keep track of the muon charges and the relation to their lateral positions . for more detailed analysis and discussion
, please refer to ref.@xcite . in summary
, we propose to extract the charge information of high energy muons in very inclined extensive air showers by analyzing their relative lateral positions in the shower transverse plane .
this muon charge information is helpful to detect the composition of cosmic rays , e.g. , the neutron or antiproton content of the ultrahigh energy cosmic rays . |
unambiguous proof for disks in massive star formation is still missing .
millimeter continuum observations suggest flattened structures without providing velocity information ( e.g. , @xcite ) , and molecular line studies suggest rotational motions but are often confused outflows and ambient gas ( e.g. , @xcite and beuther et al .
, this volume ) .
maser studies show disk signatures in some cases but are mostly not unambiguous as well ( e.g. , @xcite ) .
the best evidence yet for genuine disk emission comes from ch@xmath1cn observations in iras20126 + 4104 @xcite . in this case
, the velocity gradient defining the presence of the disk is aligned perpendicular to the bipolar outflow , consistent with the common disk / jet paradigm . to further investigate possible disk emission and its association with molecular jets
, we used the submillimeter array ( sma ) to observe the jet tracer sio(54 ) and the hot - core tracer hcooch@xmath1(2019 ) in a massive star - forming region .
the source iras18089 - 1732 is a young high - mass protostellar object ( hmpo ) which has been studied in detail over recent years .
the source is part of a sample of 69 hmpos selected mainly via infrared color - color criteria and the absence of strong cm emission @xcite .
iras18089 - 1732 is approximately at a distance of 3.6kpc and its bolometric luminosity is about @xmath5l@xmath2 @xcite .
millimeter continuum observations reveal a massive core @xmath6m@xmath2 with h@xmath7o and ch@xmath1oh maser emission , and a weak 1mjy source is detected at 3.6 cm @xcite . as part of a single - dish co outflow study , wing emission indicative of molecular outflows was detected but the co map was too confused to define a bipolar outflow @xcite . during these observations , @xcite also observed sio(21 ) at 3 mm , and bipolar structure was detected in the north - south direction .
furthermore , @xcite reported the detection of the hot - core - tracing molecules ch@xmath1cn and ch@xmath1oh .
this letter focuses on the jet / disk observations and the ( sub-)mm continuum data .
a description of the line forest observed simultaneously is presented in an accompanying paper ( beuther et al .
, this volume ) .
iras18089 - 1732 was observed with the sma between may and july 2003 in two different configurations with 3 to 5 antennas in the array .
the phase reference center of the observations was r.a.[j2000 ] 18:11:51.4 and dec.[j2000 ] @xmath8:31:28.5 .
the frequency was centered on the sio(54 ) line at 217.105ghz , the hcooch@xmath1(2019 ) line at 216.967ghz could be observed simultaneously in the same band .
the hcooch@xmath1 line consists of 8 distinct components but is dominated by 4 of them which are separated by 2.5mhz ( corresponding to 3.5 km s@xmath9 ) .
the correlator bandwidth at that time was 1ghz with a frequency resolution of 0.825mhz .
we smoothed the sio(54 ) data to a spectral resolution of 3kms@xmath9 and the hcooch@xmath1(2019 ) data to 2kms@xmath9 to increase the signal - to - noise ratio .
the continuum was constructed via averaging the line - free channels in the upper side - band .
the beam size at 217ghz was @xmath10 and at 354 ghz @xmath11 .
system temperatures in the 850@xmath0 m band were between 300 - 900k and in the 1 mm band around 200k .
the continuum rms at 217ghz was @xmath12mjy and at 354ghz 40mjy .
the flux calibration was estimated to be accurate to @xmath13 . for more details on the sma ,
the observations and data reduction , see the accompanying papers by ho , moran & lo and beuther et al .
( this volume ) .
figure [ continuum ] compares the ( sub-)mm continuum observations and shows additional cm continuum and h@xmath7o and ch@xmath1oh maser data @xcite . even in the highest - spatial - resolution data at 850@xmath0 m ,
the dust emission remains singly peaked , i.e. , it does not split up into multiple sources as observed in other massive star - forming regions , e.g. , iras19410 + 2336 @xcite .
nevertheless , in our 1 mm data we resolve elongated emission in the south and north - west , which demonstrates that iras18089 - 1732 has a compact mm core with extended halo emission ( fig .
[ continuum ] ) .
the halo emission is not seen in the 850@xmath0 m observations because of the reduced sensitivity and uv - coverage .
while the weak 3.6 cm peak and the h@xmath7o maser position coincide exactly with the ( sub-)mm continuum peak , the ch@xmath1oh maser position is about @xmath14 to the south .
the latter could indicate that there might be a second source at the position of the ch@xmath1oh maser which we can not distinguish .
table [ para ] shows the derived peak and integrated fluxes ( @xmath15 and @xmath16 ) at 1 mm and 850@xmath0 m . comparing the sma 1 mm data with single - dish observations of the region @xcite
, we find that about @xmath17 of the flux is filtered out in the interferometric data .
it is difficult to derive a spectral index from the continuum images because the different uv - coverages filter out different amounts of flux .
however , we can measure fluxes @xmath18 in the uv - plane .
ideally , one would select the same regions in the uv - plane , but as this would reduce the amount of available data even more , it is reasonable to compare the values for matching baseline ranges ( in units of @xmath19 ) .
we selected one range of short baselines ( @xmath20 , corresponding to spatial scales between @xmath21 and @xmath22 ) and one range of longer baselines ( @xmath23 , corresponding to spatial scales between @xmath24 and @xmath25 ) where there were sufficient data in both frequency bands : the flux values are shown in table [ para ] .
the 3.6 cm flux is only 0.9mjy @xcite , and assuming free - free emission its contribution to the sub-(mm ) observations is negligible . assuming a power - law relation @xmath26 in the rayleigh - jeans limit with the dust opacity index @xmath27 , we find @xmath28 for short baselines corresponding to large spatial scales and @xmath29 for large baselines corresponding to small spatial scales .
these values are lower than the canonical value of 2 @xcite and decrease to small spatial scales .
the exact values should be taken with caution due to the large calibration uncertainty ( estimated to be within @xmath13 ) , but the trend of a low @xmath27 decreasing with decreasing spatial scales appears real . without mapping the selected sets of uv - data in the image - plane , we can not determine whether the smaller spatial scales correspond to the central core or to another unresolved structure .
however , the 850@xmath0 m continuum image , which is based on larger baselines , is more compact than the 1.3 mm emission ( fig .
[ continuum ] ) indicating that the largest baselines do trace the emission from the core center
. a low value of @xmath27 has also recently been found in another high - mass star - forming region @xcite , and @xcite report that @xmath27 decreases in l1489 from 1.5 - 2 in the envelope down to @xmath30 at the inner peak .
this trend may be due to grain growth within the central core / disk or high optical depth .
assuming optically thin dust emission at mm wavelength , we calculate the mass and peak column density using the 1.3 mm data following the procedure outlined for the single - dish dust continuum data by @xcite .
we use a grain emissivity index @xmath31 and a temperature typical for hot - core - like sources of 100k .
the central core mass then is @xmath32m@xmath2 and the peak column density @xmath33 @xmath34 , corresponding to a visual extinction @xmath35 } = n_{\rm{h_2}}/(10^{21}\rm{cm}^{-2 } ) \sim 800 $ ] . as discussed by @xcite ,
the errors are dominated by systematics , e.g. , exact knowledge of @xmath27 or the temperature .
we estimate the masses and column densities to be correct within a factor of 5 .
figure [ channel ] presents a channel map of sio(54 ) , and we find blue and red emission ( systemic velocity 34.9kms@xmath9 , [ hcooch3 ] ) north of the continuum core .
the sio emission by itself allows an interpretation of a bipolar outflow centered @xmath36 north of the core .
however , we do not detect any sub - mm continuum source there ( @xmath37 m @xmath38 mass sensitivity @xmath39m@xmath2 ) , and a massive outflow without a sub - mm continuum source at the driving center is very unlikely .
therefore , we favor the interpretation of an outflow emanating from the core toward the north with a position - angle ( p.a . ) of @xmath40 .
the fact that we see blue and red sio emission toward the northern lobe indicates that the outflow axis is near the plane of the sky . for the red lobe
, we find an increase in velocity with distance from the center of the core resembling the hubble - law within molecular outflows ( e.g. , @xcite ) .
previous unpublished single - dish sio(21 ) observations with @xmath41 resolution also show a bipolar outflow in the north - south direction .
the main difference is that the southern part of the sio(21 ) outflow is not observed at the higher frequency and higher spatial resolution with the sma .
we can attribute this difference to several possible reasons .
first , interferometers filter out the large scale emission , and without additional short spacing information , only the more compact jet - like emission is detected with the sma .
second , the excitation of the 54 transition is higher with respect to the 21 transition .
while the sio(21 ) line traces the lower temperature gas , the sio(54 ) line is sensitive to the more excited gas being associated with the more collimated component of the outflow .
furthermore , an asymmetry in the small - scale distribution of the dense gas can also contribute to the differences of the large - scale sio(21 ) and small - scale sio(54 ) observations .
the high - density - gas tracing molecular hcooch@xmath1(2019 ) emission is confined to the central core traced by the ( sub-)mm continuum ( fig . [ channel ] ) .
a gaussian fit to the central spectrum results in a systemic velocity of @xmath42kms@xmath9 , and we observe a velocity gradient across the core .
it is possible to fit the peak position of each spectral channel shown in figure [ channel ] to a higher accuracy than the nominal spatial resolution , down to 0.5hpbw/(s / n ) , with s / n the signal - to - noise ratio @xcite .
we performed these fits in the uv - plane to avoid artifacts due to inverting and cleaning the data , the resulting positions are shown in figure [ pos_velo ] .
the derived velocity gradient follows a p.a .
of @xmath43 from blue to red velocities .
the separation of the most blue and most red positions is @xmath44 , corresponding to @xmath45au .
figure [ pos_velo ] also presents a position - velocity diagram for the fitted channels at the p.a .
of @xmath46 confirming the space - velocity shift .
this velocity gradient is neither in the direction of the sio outflow nor directly perpendicular to it .
a velocity gradient solely due to a rotating disk would be perpendicular to the outflow .
the sio outflow is near the plane of the sky complicating an accurate determination of its p.a . , and the outflow might also be affected by precession ( e.g. , iras20126 + 4104 , @xcite ) . regarding the hcooch@xmath1 emission , it is unlikely that this high - density tracer depicts a different outflow which would remain undetected by the sio .
therefore , we regard neither the partially uncertain outflow p.a . nor a second outflow sufficient to explain the difference in p.a . between the sio and the hcooch@xmath1 emission . however , as shown in the accompanying paper ( beuther et al . ,
this volume ) , most molecular line data are influenced by the molecular outflow , and it is likely that hcooch@xmath1 is also affected by the outflow .
furthermore , @xcite have shown that infall can influence the dense - gas velocity signature as well .
assuming that the inner core contains a rotating disk , which has an orientation perpendicular to the outflow , and that the hcooch@xmath1 velocity signature is influenced by the disk , the outflow and infall , it is plausible that the observed axis of the hcooch@xmath1 velocity gradient is offset from the expected disk orientation .
we do not observe the inner region of the disk where the outflow / jet is accelerated , but we may have detected the outer parts of the accretion - disk .
assuming equilibrium between centrifugal and gravitational forces at the outer radius of the disk , we calculate the dynamical mass enclosed within the central @xmath47 to be @xmath48m@xmath2 ( with @xmath49 the unknown inclination angle between the disk plane and the line of sight , and an hcooch@xmath1 velocity shift . taking into account the 4 dominant hcooch@xmath1 line components spanning 3.5kms@xmath9 ( [ obs ] ) ,
the hcooch@xmath1 velocity shift is @xmath50 . ] of 4.25kms@xmath9 ) , of the order of the core mass derived from the dust emission ( table [ para ] ) .
thus , in contrast to t tauri systems where the disk mass is negligible compared with the protostellar mass , in iras18089 - 1732 a considerable fraction of the enclosed mass seems to be part of a large accretion disk and/or rotating envelope .
this result fits the picture that iras18089 - 1732 is in a very young evolutionary state , with the central ( proto)star still accreting material from the surrounding core / disk .
the combined sio(54 ) and hcooch@xmath1(2019 ) observations toward iras18089 - 1732 support a massive star formation scenario where high - mass stars form in a similar fashion as their low - mass counterparts , i.e. , via disk accretion accompanied by collimated jets / outflows .
the hcooch@xmath1 observations still barely resolve the disk / envelope structure and the interpretation is not entirely unambiguous , but the data indicate rotation which might stem at least partly from an accretion disk . higher resolution observations in different disk tracers are needed to investigate the disk / envelope conditions in more detail . furthermore , the continuum data indicate a lower @xmath27 than the standard value of 2 , and we observe a decreasing @xmath27 with decreasing spatial scales . as nearly all the gas observed in dust emission takes part in the assumed dynamical rotation observed in hcooch@xmath1 this low @xmath27 might be due to grain growth or high opacity within a disk - like structure ( e.g. , @xcite ) . | sma observations of the massive star - forming region iras18089 - 1732 in the 1 mm and 850@xmath0 m band reveal outflow and disk signatures in different molecular lines .
the sio(54 ) data show a collimated outflow in the northern direction .
in contrast , the hcooch@xmath1(2019 ) line , which traces high - density gas , is confined to the very center of the region and shows a velocity gradient across the core .
the hcooch@xmath1 velocity gradient is not exactly perpendicular to the outflow axis but between an assumed disk plane and the outflow axis .
we interpret these hcooch@xmath1 features as originating from a rotating disk that is influenced by the outflow and infall .
based on the ( sub-)mm continuum emission , the mass of the central core is estimated to be around 38m@xmath2 . the dynamical mass derived from the hcooch@xmath1 data is 22m@xmath2 , of about the same order as the core mass .
thus , the mass of the protostar / disk / envelope system is dominated by its disk and envelope .
the two frequency continuum data of the core indicate a low dust opacity index @xmath3 in the outer part , decreasing to @xmath4 on shorter spatial scales . |
these notes contain some details about talks that were presented in the international conference `` quantum topology '' organized by laboratory of quantum topology of chelyabinsk state university in july 2014 .
they are based on the notes of five lectures presented in the icpam - ictp research school of mekns in may 2012 .
i thank the organizers of these two great events .
i also thank catherine gille and kvin corbineau for useful comments on these notes .
these notes have been written in an introductory way , in order to be understandable by graduate students .
in particular , sections [ seclk ] , [ sectheta ] and [ secmorepar ] provide an elementary self - contained presentation of the @xmath3invariant .
the notes also contain original statements ( theorems [ thmmain],[thmunivone ] , [ thmkeyuniv ] and [ thmmainstraight ] ) together with sketches of proofs .
complete proofs of these statements , which generalize known statements , will be included in a monograph @xcite .
the modern powerful invariants of links and @xmath0manifolds that will be defined in section [ secconstcsi ] can be thought of as generalizations of the linking number . in this section
, we warm up with several ways of defining this classical basic invariant .
this allows us to introduce conventions and methods that will be useful througout the article .
let @xmath5 denote the unit circle of @xmath6 .
@xmath7 consider two @xmath8 embeddings @xmath9 ( .5,.5 ) rectangle ( 4,3.5 ) ; ( .8,1.5 ) .. controls ( .8,1 ) and ( 1.7,1 ) .. ( 1.7,2 ) ; ( 4,2 ) .. controls ( 4,3 ) and ( 2.5,2.7 ) .. ( 2.5,2.5 ) ( 3.5,2 ) .. controls ( 3.5,1.8 ) and ( 2.5,1.7 ) .. ( 2.5,1.5 ) ; ( 1.2,3.3 ) .. controls ( 1,3.3 ) and ( .7,3.2 ) .. ( .7,3 ) .. controls ( .7,2.8 ) and ( 1.5,.7 ) .. ( 2.2,.7 ) .. controls ( 2.4,.7 ) and ( 3,.9 ) .. ( 3,1.1 )
( 3,2.9 ) .. controls ( 3,3.1 ) and ( 2.4,3.3 ) .. ( 2.2,3.3 ) ; ( 2.5,2.5 ) .. controls ( 2.5,2.3 ) and ( 3.5,2.2 ) .. ( 3.5,2 ) ( 2.5,1.5 ) .. controls ( 2.5,1 ) and ( 4,1 ) .. ( 4,2 ) ; ( 2.2,3.3 ) .. controls ( 1.5,3.3 ) and ( .8,2 ) .. ( .8,1.5 ) ; ( 1.7,2 ) .. controls ( 1.7,2.7 ) and ( 1.5,3.3 ) .. ( 1.2,3.3 ) ; ( .8,1.6)(.8,1.3 ) ; ( 2.5,1.6)(2.5,1.3 ) ; ( .6,1.5 ) node@xmath10 ( 3.7,2 ) node@xmath11 ; and the associated _ gauss map _
@xmath12 ( 0,-.5 ) rectangle ( 4,1 ) ; ( 0,0 ) .. controls ( 0,-.5 ) and ( 1.3,-.9 ) .. ( 2,-.9 ) .. controls ( 2.7,-.9 ) and ( 4,-.5 ) .. ( 4,0 ) .. controls ( 4,.5 ) and ( 2.7,.9 ) .. ( 2,.9 ) .. controls ( 1.3,.9 ) and ( 0,.5 ) .. ( 0,0 ) ; plot[smooth ] coordinates(1.4,.1 ) ( 1.6,0 ) ( 2,-.1 ) ( 2.4,0 ) ( 2.6,.1 ) ; ( 1.6,0 ) to ( 2.4,0 ) ; ( 2.8,-.15 ) ( 3.1,-.15 ) node[right]@xmath13 ; ( 2.8,-.15 ) ( 2.8,.15 ) node[above]@xmath1 ; @xmath14 ( -1,-.5 ) rectangle ( 1,1 ) ; ( 0,0 ) circle ( .9 ) ; ( -.9,0 ) to ( .9,0 ) ; ( -.9,0 ) to ( .9,0 ) ; ( -.15,-.7 ) ( .15,-.7 ) node[right]@xmath13 ; ( -.15,-.7 ) ( -.15,-.4 ) node[left]@xmath1 ; denote the standard area form of @xmath15 by @xmath16 so that @xmath17 is the homogeneous volume form of @xmath15 such that @xmath18 . in 1833 , gauss defined the _ linking number _ of the disjoint _ knots _ @xmath19 and @xmath20 , simply denoted by @xmath10 and @xmath11 , as an integral @xcite . with modern notation
, his definition reads @xmath21 it can be rephrased as _
@xmath22 is the degree of the gauss map @xmath23_. a _
topological @xmath24dimensional manifold @xmath25 without boundary _ is a hausdorff topological space that is a countable union of open subsets @xmath26 labeled in a set @xmath27 ( @xmath28 ) , where every @xmath26 is identified with an open subset @xmath29 of @xmath30 by a homeomorphism @xmath31 , called a _
_ manifolds are considered up to homeomorphism so that homeomorphic manifolds are considered identical . for @xmath32 , the topological manifold @xmath25 _
has a @xmath33structure _ or _ is a @xmath33manifold _ , if , for each pair @xmath34 , the map @xmath35 defined on @xmath36 is a @xmath33diffeomorphism to its image . the notion of @xmath37maps , @xmath38 , from such a manifold to another one
can be naturally deduced from the known case where the manifolds are open subsets of some @xmath30 , thanks to the local identifications provided by the charts .
@xmath33manifolds are considered up to @xmath33diffeomorphisms .
an _ orientation _ of a real vector space @xmath39 of positive dimension is a basis of @xmath39 up to a change of basis with positive determinant .
when @xmath40 , an orientation of @xmath39 is an element of @xmath41 .
for @xmath42 , an orientation of @xmath30 identifies @xmath43 with @xmath44 .
( in these notes , we freely use basic algebraic topology , see @xcite for example . ) a homeomorphism @xmath45 from an open subset @xmath46 of @xmath30 to another such @xmath39 is _ orientation - preserving _ at a point @xmath47 , if @xmath48 is orientation - preserving .
if @xmath45 is a diffeomorphism , @xmath45 is orientation - preserving at @xmath47 if and only if the determinant of the jacobian @xmath49 is positive .
if @xmath30 is oriented and if the transition maps @xmath35 are orientation - preserving ( at every point ) for @xmath34 , the manifold @xmath25 is _
oriented_. for @xmath50 , @xmath13 , @xmath1 or @xmath0 , any topological @xmath24-manifold may be equipped with a unique smooth structure ( up to diffeomorphism ) ( see theorem [ thmstructhree ] , below ) . unless otherwise mentioned ,
our manifolds are _ smooth _
( i. e. @xmath8 ) , oriented and compact , and considered up oriented diffeomorphisms .
products are oriented by the order of the factors . more generally , unless otherwise mentioned , the order of appearance of coordinates or parameters orients manifolds .
a point @xmath51 is _ a regular value _ of a smooth map @xmath52 between two smooth manifolds @xmath25 and @xmath53 , if for any @xmath54 the tangent map @xmath55 at @xmath47 is surjective .
according to the morse - sard theorem @xcite , the set of regular values of such a map is dense . if @xmath25 is compact , it is furthermore open
. when @xmath25 is oriented and compact , and when the dimension of @xmath25 coincides with the dimension of @xmath53 , the _ differential degree _ of @xmath56 at a regular value @xmath51 of @xmath53 is the ( finite ) sum running over the @xmath54 of the signs of the determinants of @xmath55 . in our case where @xmath25 has no boundary , this differential degree is locally constant on the set of regular values , and it is the _ degree _ of @xmath56 , if @xmath53 is connected .
see ( * ? ? ?
* chapter 5 ) . finally , recall a homological definition of the degree .
let @xmath57 $ ] denote the class of an oriented _ closed _
( i.e. compact , connected , without boundary ) @xmath24manifold in @xmath58 .
@xmath59 $ ] . if @xmath25 and @xmath53 are two closed oriented @xmath24manifolds and if @xmath60 is a ( continuous ) map , then @xmath61)=\mbox{deg}(f)[n]$ ] . in particular , for the gauss map @xmath23 of subsection [ subdefgauss ] , @xmath62)=lk(j , k)[s^2].\ ] ] since the differential degree of the gauss map @xmath23 is locally constant , @xmath63 for any @xmath1-form @xmath64 on @xmath15 such that @xmath65 .
let us compute @xmath22 as the differential degree of @xmath23 at the vector @xmath66 that points towards us .
the set @xmath67 is made of the pairs of points @xmath68 where the projections of @xmath69 and @xmath70 coincide , and @xmath69 is under @xmath70 .
they correspond to the _ crossings _ ( -.6,.1 ) rectangle ( .6,.55 ) ; ( .2,0 ) ( -.2,.4 ) node[left]@xmath10 ; ( -.2,0 ) ( .2,.4 ) node[right]@xmath11 ; ( -.2,0 ) ( .2,.4 ) ; and ( -.6,.1 ) rectangle ( .6,.55 ) ; ( -.2,0 ) ( .2,.4 ) node[right]@xmath10 ; ( .2,0 ) ( -.2,.4 ) node[left]@xmath11 ; ( .2,0 ) ( -.2,.4 ) ; of the diagram . in a diagram ,
a crossing is _ positive _ if we turn counterclockwise from the arrow at the end of the upper strand to the arrow of the end of the lower strand like ( -.5,.1 ) rectangle ( .5,.55 ) ; ( .2,0 ) ( -.2,.4 ) ; ( -.2,0 ) ( .2,.4 ) ; ( -.2,0 ) ( .2,.4 ) ; ( .1,.45 ) .. controls ( 0,.5 ) .. ( -.1,.45 ) ; .
otherwise , it is _ negative _ like ( -.5,.1 ) rectangle ( .5,.55 ) ; ( -.2,0 ) ( .2,.4 ) ; ( .2,0 ) ( -.2,.4 ) ; ( .2,0 )
( -.2,.4 ) ; . for the positive crossing ( -.6,.1 ) rectangle ( .6,.55 ) ; ( .2,0 ) ( -.2,.4 ) node[left]@xmath10 ; ( -.2,0 ) ( .2,.4 ) node[right]@xmath11 ; ( -.2,0 ) ( .2,.4 ) ; , moving @xmath69 along @xmath10 following the orientation of @xmath10 , moves @xmath71 towards the south - east direction , while moving @xmath70 along @xmath11 following the orientation of @xmath11 , moves @xmath71 towards the north - east direction , so that the local orientation induced by the image of @xmath23 around @xmath72 is ( -.1,.1 ) rectangle ( 1.5,.55 ) ; ( 0,.2 )
( .2,0 ) node[right]@xmath73 ; ( 0,.2 ) ( .2,.4 ) node[right]@xmath74 ; , which is ( -.1,.1 ) rectangle ( .5,.55 ) ; ( 0,.2 ) ( .2,0 ) node[right]@xmath13 ; ( 0,.2 ) ( .2,.4 ) node[right]@xmath1 ; .
therefore , the contribution of a positive crossing to the degree is @xmath13 .
similarly , the contribution of a negative crossing is @xmath75 .
we have just proved the following formula @xmath76 ( .2,0 ) -- ( -.2,.4 ) node[left]{\tiny $ j$ } ; \draw [ ->,draw = white , double = gray , very thick ] ( -.2,0 ) -- ( .2,.4 ) node[right]{\tiny $ k$ } ; \draw [ ->,draw = gray ] ( -.2,0 ) -- ( .2,.4 ) ; \end{tikzpicture}}- \sharp { \begin{tikzpicture } \useasboundingbox ( -.6,.1 ) rectangle ( .6,.55 ) ; \draw [ - > ] ( -.2,0 ) -- ( .2,.4 ) node[right]{\tiny $ j$ } ; \draw [ ->,draw
= white , double = gray , very thick ] ( .2,0 ) -- ( -.2,.4 ) node[left]{\tiny $ k$ } ; \draw [ ->,draw = gray ] ( .2,0 ) -- ( -.2,.4 ) ; \end{tikzpicture}}\ ] ] where @xmath77 stands for the cardinality here @xmath78 ( .2,0 ) -- ( -.2,.4 ) node[left]{\tiny $ j$ } ; \draw [ ->,draw = white , double = gray , very thick ] ( -.2,0 ) -- ( .2,.4 ) node[right]{\tiny $ k$ } ; \draw [ ->,draw = gray ] ( -.2,0 ) -- ( .2,.4 ) ; \end{tikzpicture}}$ ] is the number of occurences of @xmath79 ( .2,0 ) -- ( -.2,.4 ) node[left]{\tiny $ j$ } ; \draw [ ->,draw = white , double = gray , very thick ] ( -.2,0 ) -- ( .2,.4 ) node[right]{\tiny $ k$ } ; \draw [ ->,draw = gray ] ( -.2,0 ) -- ( .2,.4 ) ; \end{tikzpicture}}$ ] in the diagram so that @xmath80 ( .2,0 ) -- ( -.2,.4 ) node[left]{\tiny $ j$ } ; \draw [ ->,draw = white , double = gray , very thick ] ( -.2,0 ) -- ( .2,.4 ) node[right]{\tiny $ k$ } ; \draw [ ->,draw = gray ] ( -.2,0 ) -- ( .2,.4 ) ; \end{tikzpicture}}- \sharp { \begin{tikzpicture } \useasboundingbox ( -.6,.1 ) rectangle ( .6,.55 ) ; \draw [ - > ] ( -.2,0 ) -- ( .2,.4 ) node[right]{\tiny $ j$ } ; \draw [ ->,draw = white , double = gray , very thick ] ( .2,0 ) -- ( -.2,.4 ) node[left]{\tiny $ k$ } ; \draw [ ->,draw = gray ] ( .2,0 ) -- ( -.2,.4 ) ; \end{tikzpicture}}.\ ] ] similarly , @xmath81 ( .2,0 ) -- ( -.2,.4 ) node[left]{\tiny $ k$ } ; \draw [ ->,draw = white , double = black , very thick ] ( -.2,0 ) -- ( .2,.4 ) node[right]{\tiny $ j$ } ; \draw [ - > ] ( -.2,0 ) -- ( .2,.4 ) ; \end{tikzpicture}}- \sharp { \begin{tikzpicture } \useasboundingbox ( -.6,.1 ) rectangle ( .6,.55 ) ; \draw [ ->,draw = gray ] ( -.2,0 ) -- ( .2,.4 ) node[right]{\tiny $ k$ } ; \draw [ ->,draw = white , double = black , very thick ] ( .2,0 ) -- ( -.2,.4 ) node[left]{\tiny $ j$ } ; \draw [ - > ] ( .2,0 ) -- ( -.2,.4 ) ; \end{tikzpicture}}$ ] so that @xmath82 ( .2,0 ) -- ( -.2,.4 ) node[left]{\tiny $ k$ } ; \draw [ ->,draw = white , double = black , very thick ] ( -.2,0 ) -- ( .2,.4 ) node[right]{\tiny $ j$ } ; \draw [ - > ] ( -.2,0 ) -- ( .2,.4 ) ; \end{tikzpicture}}- \sharp { \begin{tikzpicture } \useasboundingbox ( -.6,.1 ) rectangle ( .6,.55 ) ; \draw [ ->,draw = gray ] ( -.2,0 ) -- ( .2,.4 ) node[right]{\tiny $ k$ } ; \draw [ ->,draw = white , double = black , very thick ] ( .2,0 ) -- ( -.2,.4 ) node[left]{\tiny $ j$ } ; \draw [ - > ] ( .2,0 ) -- ( -.2,.4 ) ; \end{tikzpicture}}= \frac12\left(\sharp { \begin{tikzpicture } \useasboundingbox ( -.6,.1 ) rectangle ( .6,.55 ) ; \draw [ - > ] ( .2,0 ) -- ( -.2,.4 ) node[left]{\tiny $ j$ } ; \draw [ ->,draw = white , double = gray , very thick ] ( -.2,0 ) -- ( .2,.4 ) node[right]{\tiny $ k$ } ; \draw [ ->,draw = gray ] ( -.2,0 ) -- ( .2,.4 ) ; \end{tikzpicture}}+\sharp { \begin{tikzpicture } \useasboundingbox ( -.6,.1 ) rectangle ( .6,.55 ) ; \draw [ ->,draw = gray ] ( .2,0 ) -- ( -.2,.4 ) node[left]{\tiny $ k$ } ; \draw [ ->,draw = white , double = black , very thick ] ( -.2,0 ) -- ( .2,.4 ) node[right]{\tiny $ j$ } ; \draw [ - > ] ( -.2,0 ) -- ( .2,.4 ) ; \end{tikzpicture}}- \sharp { \begin{tikzpicture } \useasboundingbox ( -.6,.1 ) rectangle ( .6,.55 ) ; \draw [ - > ] ( -.2,0 ) -- ( .2,.4 ) node[right]{\tiny $ j$ } ; \draw [ ->,draw = white , double = gray , very thick ] ( .2,0 ) -- ( -.2,.4 ) node[left]{\tiny $ k$ } ; \draw [ ->,draw = gray ] ( .2,0 ) -- ( -.2,.4 ) ; \end{tikzpicture}}- \sharp { \begin{tikzpicture } \useasboundingbox ( -.6,.1 ) rectangle ( .6,.55 ) ; \draw [ ->,draw
= gray ] ( -.2,0 ) -- ( .2,.4 ) node[right]{\tiny $ k$ } ; \draw [ ->,draw = white , double = black , very thick ] ( .2,0 ) -- ( -.2,.4 ) node[left]{\tiny $ j$ } ; \draw [ - > ] ( .2,0 ) -- ( -.2,.4 ) ; \end{tikzpicture}}\right)\ ] ] and @xmath83 . in our first example , @xmath84 .
let us draw some further examples . for the _ positive hopf link _ ( -1,-.5 ) rectangle ( 1,.5 ) ; ( 0,0 )
( 0,.2 ) .. controls ( 0,.4 ) and ( -.15,.5 ) .. ( -.3,.5 ) .. controls(-.45,.5 ) and ( -.6,.2 ) .. ( -.6,0 ) node[left]@xmath10 ; ( -.6,0 ) .. controls ( -.6,-.2 ) and ( -.45,-.5 ) .. ( -.3,-.5 ) .. controls ( -.15,-.5 ) and ( 0,-.35 ) .. ( 0,-.2 ) ; ( -.1,.2 ) .. controls ( -.2,.2 ) and ( -.25,.15 ) .. ( -.25,.05 ) .. controls ( -.25,-.05 ) and ( -.2,-.1 ) .. ( 0,-.1 ) .. controls ( .2,-.1 ) and ( .25,-.05 ) .. ( .25,.05 ) .. controls ( .25,.15 ) and ( .2,.2 ) .. ( .1,.2 ) node[right]@xmath11 ; , @xmath85 . for the _ negative hopf link _ ( -1,-.5 ) rectangle ( 1,.5 ) ; ( 0,0 ) ( 0,.2 ) .. controls ( 0,.4 ) and ( -.15,.5 ) .. ( -.3,.5 ) .. controls(-.45,.5 ) and ( -.6,.2 ) .. ( -.6,0 ) ; ( -.6,0 ) .. controls ( -.6,-.2 ) and ( -.45,-.5 ) .. ( -.3,-.5 ) .. controls ( -.15,-.5 ) and ( 0,-.35 ) .. ( 0,-.2 ) ; ( -.1,.2 ) .. controls ( -.2,.2 ) and ( -.25,.15 ) .. ( -.25,.05 ) .. controls ( -.25,-.05 ) and ( -.2,-.1 ) .. ( 0,-.1 ) .. controls ( .2,-.1 ) and ( .25,-.05 ) .. ( .25,.05 ) .. controls ( .25,.15 ) and ( .2,.2 ) .. ( .1,.2 ) ; , @xmath86 . for the _ whitehead link _ ( -.5,-1.1 ) rectangle ( 1.5,1 ) ;
( .4,0 ) ( .4,.2 ) .. controls ( .4,.45 ) and ( .6,.6 ) .. ( .8,.6 ) .. controls ( 1,.6 ) and ( 1.4,.2 ) .. ( 1.4,-.1 ) ; ( 0,0 ) ( 0,.2 ) .. controls ( 0,.6 ) and ( .5,.9 ) .. ( .8,.9 ) .. controls ( 1.1,.9 ) and ( 1.7,.5 ) .. ( 1.7,-.1 ) ; ( .3,.2 ) ( .1,.2 ) ( -.1,.2 ) .. controls ( -.2,.2 ) and ( -.35,.15 ) .. ( -.35,.05 ) .. controls ( -.35,-.05 ) and ( 0,-.1 ) .. ( .2,-.1 ) .. controls ( .4,-.1 ) and ( .75,-.05 ) .. ( .75,.05 ) .. controls ( .75,.15 ) and ( .6,.2 ) .. ( .5,.2 ) ; ( .2,-.3 ) .. controls ( -.1,-.3 ) and ( .2,-1.1 ) .. ( .8,-1.1 ) .. controls ( 1.1,-1.1 ) and ( 1.7,-.7 ) .. ( 1.7,-.1 ) ; ( 0,-.2 ) .. controls ( 0,-.3 ) and ( .1,-.5 ) .. ( .2,-.5 ) .. controls ( .3,-.5 ) and ( .4,-.3 ) .. ( .4,-.2 ) ; ( .2,-.3 ) .. controls ( .3,-.3 ) and ( .5,-.8 ) .. ( .8,-.8 ) .. controls ( 1.1,-.8 ) and ( 1.4,-.4 ) .. ( 1.4,-.1 ) ; , @xmath87 .
a _ topological @xmath24dimensional manifold @xmath25 with possible boundary _ is a hausdorff topological space that is a union of open subsets @xmath26 with subscripts in a set @xmath27 , ( @xmath28 ) , where every @xmath26 is identified with an open subset @xmath29 of @xmath88-\infty,0]^k \times { \mathbb{r}}^{n - k}$ ] by a chart @xmath31 .
the _ boundary _ of @xmath88-\infty,0]^k \times { \mathbb{r}}^{n - k}$ ] is made of the points @xmath89 of @xmath88-\infty,0]^k \times { \mathbb{r}}^{n - k}$ ] such that there exists @xmath90 such that @xmath91 .
the _ boundary _ of @xmath25 is made of the points that are mapped to the boundary of @xmath88-\infty,0]^k \times { \mathbb{r}}^{n - k}$ ] . for @xmath92 ,
the topological manifold @xmath25 _ is a @xmath33manifold with ridges ( or with corners ) ( resp . with boundary ) _ , if , for each pair @xmath34 , the map @xmath35 defined on @xmath36 is a @xmath33diffeomorphism to its image ( resp . and
if furthermore @xmath93 , for any @xmath94 )
. then the _ ridges _ of @xmath25 are made of the points that are mapped to points @xmath89 of @xmath88-\infty,0]^k \times { \mathbb{r}}^{n - k}$ ] so that there are at least two @xmath90 such that @xmath91 .
the tangent bundle to an oriented submanifold @xmath95 in a manifold @xmath25 at a point @xmath47 is denoted by @xmath96 .
the _ normal bundle _
@xmath97 to @xmath95 in @xmath25 at @xmath47 is denoted by @xmath98 .
it is oriented so that ( a lift of an oriented basis of ) @xmath98 followed by ( an oriented basis of ) @xmath96 induce the orientation of @xmath99 .
the boundary @xmath100 of an oriented manifold @xmath25 is oriented by the _
outward normal first _ convention .
if @xmath101 is not in a ridge , the outward normal to @xmath25 at @xmath47 followed by an oriented basis of @xmath102 induce the orientation of @xmath25 .
for example , the standard orientation of the disk in the plane induces the standard orientation of the circle , counterclockwise , as the following picture shows .
( -.8,-.8 ) rectangle ( .8,.8 ) ; ( 0,0 ) circle ( .7 ) ; ( -.15,-.3 )
( .15,-.3 ) node[right]@xmath13 ; ( -.15,-.3 ) ( -.15,0 ) node[left]@xmath1 ; ( .7,0 ) ( 1,0 ) node[right]@xmath13 ; ( .7,0 ) ( .7,.3 ) node[left]@xmath1 ; ( -.7,0 ) arc ( -180:180:.7 ) ; as another example , the sphere @xmath15 is oriented as the boundary of the ball @xmath103 , which has the standard orientation induced by ( thumb , index finger ( 2 ) , middle finger ( 3 ) ) of the right hand .
( -1,-.7 ) rectangle ( 1,1 ) ; ( 0,0 ) circle ( .9 ) ; ( -.9,0 ) to ( .9,0 ) ; ( -.9,0 ) to ( .9,0 ) ; ( -.15,-.7 ) ( .15,-.7 ) node[right]@xmath1 ; ( -.15,-.7 ) ( -.15,-.4 ) node[left]@xmath0 ; two submanifolds @xmath95 and @xmath104 in a manifold @xmath25 are transverse if at each intersection point @xmath47 , @xmath105 .
the transverse intersection of two submanifolds @xmath95 and @xmath104 in a manifold @xmath25 is oriented so that the normal bundle to @xmath106 is @xmath107 , fiberwise
. if the two manifolds are of complementary dimensions , then the sign of an intersection point is @xmath108 if the orientation of its normal bundle coincides with the orientation of the ambient space , that is if @xmath109 ( as oriented vector spaces ) , this is equivalent to @xmath110 ( as oriented vector spaces again , exercise ) .
otherwise , the sign is @xmath111 .
if @xmath95 and @xmath104 are compact and if @xmath95 and @xmath104 are of complementary dimensions in @xmath25 , their _ algebraic intersection _ is the sum of the signs of the intersection points , it is denoted by @xmath112 .
when @xmath25 is an oriented manifold , @xmath113 denotes the same manifold , equipped with the opposite orientation . in a manifold @xmath25 , a @xmath114-dimensional _ chain ( resp .
rational chain ) _ is a finite combination with coefficients in @xmath115 ( resp . in @xmath116 ) of smooth @xmath114-dimensional oriented submanifolds @xmath117 of @xmath25 with boundary and ridges , up to the identification of @xmath118 with @xmath119 . again , unless otherwise mentioned , manifold are oriented .
the boundary @xmath120 of chains is a linear map that maps a smooth submanifold to its oriented boundary .
the canonical orientation of a point is the sign @xmath108 so that @xmath121= \{1\}-\{0\}$ ] . [ lemsignint ] let @xmath95 and @xmath104 be two transverse submanifolds of a @xmath122dimensional manifold @xmath25 , of respective dimensions @xmath123 and @xmath124 , with disjoint boundaries . then @xmath125 proof : note that @xmath126 . at a point @xmath127
, @xmath128 is oriented by @xmath129 , where @xmath130 is the outward normal to @xmath95 .
if @xmath131 , then @xmath130 is also an outward normal to @xmath106 , and @xmath132 is cooriented by @xmath133 while @xmath134 is cooriented by @xmath135 . at a point @xmath136
, @xmath132 is cooriented by @xmath133 like @xmath137 .
let @xmath10 and @xmath11 be two rationally null - homologous disjoint cycles of respective dimensions @xmath138 and @xmath114 in a @xmath122-manifold @xmath25 , where @xmath139 .
there exists a rational @xmath140chain @xmath141 bounded by @xmath10 transverse to @xmath11 , and a rational @xmath142chain @xmath143 bounded by @xmath11 transverse to @xmath10 and for any two such rational chains @xmath141 and @xmath143 , @xmath144 . in particular , @xmath145 is a topological invariant of @xmath146 , which is denoted by @xmath147 and called the _ linking number _ of @xmath10 and @xmath11 .
@xmath148 proof : since @xmath11 is rationally null - homologous , @xmath11 bounds a rational @xmath142chain @xmath143 .
without loss , @xmath143 is assumed to be transverse to @xmath141 so that @xmath149 is a rational @xmath13chain ( which is a rational combination of circles and intervals ) .
( as explained in ( * ? ? ?
* chapter 3 ) , generically , manifolds are transverse . ) according to lemma [ lemsignint ] , @xmath150 furthermore , the sum of the coefficients of the points in the left - hand side must be zero , since this sum vanishes for the boundary of an interval .
this shows that @xmath151 , and therefore that this rational number is independent of the chosen @xmath141 and @xmath143 . since @xmath152 , @xmath153 in particular , the _ linking number _ of two rationally null - homologous disjoint links @xmath10 and @xmath11 in a @xmath0-manifold @xmath25 is the algebraic intersection of a rational chain bounded by one of the knots and the other one . for @xmath154 or @xmath116 , a _ @xmath155-sphere or ( integer or rational ) homology @xmath0-sphere _
a _ @xmath155-ball _ ) is a smooth , compact , oriented @xmath0-manifold , without ridges , with the same @xmath155-homology as the sphere @xmath156 ( resp . as a point ) .
in such a manifold , any knot is rationally null - homologous so that the linking number of two disjoint knots always makes sense .
a _ meridian _ of a knot @xmath11 is the ( oriented ) boundary of a disk that intersects @xmath11 once with a positive sign .
since a chain @xmath141 bounded by a knot @xmath10 disjoint from @xmath11 in a @xmath0manifold @xmath25 provides a rational cobordism between @xmath10 and a combination of meridians of @xmath11 , @xmath157=lk(j , k)[m_k]$ ] in @xmath158 where @xmath159 is a meridian of @xmath11 .
( -1,-.3 ) rectangle ( 1,1.3 ) ; ( .5,0 ) .. controls ( .8,0 ) and ( .9,.3 ) .. ( .9,.6);(1.05,.8 ) .. controls ( 1.05,.85 ) and ( 1,.9 ) .. ( .9,.9 ) .. controls ( .8,.9 ) and ( .75,.85 ) .. ( .75,.8 ) ; ( .9,.6 ) .. controls ( .9,.9 ) and ( .7,1.2 ) .. ( .5,1.2 ) .. controls ( .3,1.2 ) and ( -.2,1.15 ) .. ( -.2,1);(.75,.8 ) .. controls ( .75,.75 ) and ( .8,.7 ) .. ( .9,.7 ) .. controls ( 1,.7 ) and ( 1.05,.7 ) .. ( 1.05,.8 ) ; ( .75,.8 ) .. controls ( .75,.75 ) and ( .8,.7 ) .. ( .9,.7 ) .. controls ( 1,.7 ) and ( 1.05,.7 ) .. ( 1.05,.8 ) node[right]@xmath159 ; ( .4,.5 ) .. controls ( .4,.3 ) and ( .2,0 ) .. ( 0,0 ) node[below]@xmath11 ; ( .2,1 ) .. controls ( .2,1.15 ) and ( -.3,1.2 ) .. ( -.5,1.2 ) .. controls ( -.8,1.2 ) and ( -.9,.9 ) .. ( -.9,.6 ) .. controls ( -.9,.3 ) and ( -.8,0 ) .. ( -.5,0 ) .. controls ( -.3,0 ) and ( -.2,.4 ) .. ( 0,.4 ) .. controls ( .2,.4 ) and ( .3,0 ) .. ( .5,0 ) ; ( 0,0 ) .. controls ( -.2,0 ) and ( -.4,.3 ) .. ( -.4,.5 ) .. controls ( -.4,.7 ) and ( .2,.85 ) .. ( .2,1);(-.2,1 ) .. controls ( -.2,.85 ) and ( .4,.7 ) .. ( .4,.5 ) ; [ lemdeflkhom ] when @xmath11 is a knot in a @xmath116-sphere or a @xmath116-ball @xmath25 , @xmath160 $ ] , so that the equation @xmath157=lk(j , k)[m_k]$ ] in @xmath158 provides an alternative definition for the linking number .
proof : exercise .
the reader is also invited to check that @xmath161 as an exercise though it will be proved in the next subsection , see proposition [ propdeflkeq ] .
[ lempstt ] the map @xmath162 is a homotopy equivalence . in particular
@xmath163 is an isomorphism for all @xmath94 , @xmath164 is a homology @xmath15 , and @xmath165=\left(h_{2}(p_{s^2})\right)^{-1}[s^2]$ ] is a canonical generator of @xmath166.\ ] ] proof : @xmath164 is homeomorphic to @xmath1670,\infty [ \times
s^2 $ ] via the map @xmath168 as in subsection [ subdefgauss ] , consider a two - component link @xmath169 .
this embedding induces an embedding @xmath170 the map @xmath23 of subsection [ subdefgauss ] reads @xmath171 , and since @xmath172=\mbox{deg}(p_{jk})[s^2]=lk_g(j , k)[s^2]$ ] in @xmath173 $ ] , @xmath174=h_2(j \times k)[s^1 \times s^1]=lk_g(j , k)[s]\ ] ] in @xmath175.$ ] we will see that this definition of @xmath176 generalizes to links in rational homology @xmath0spheres and then prove that our generalized definition coincides with the general definition of linking numbers in this case . for a @xmath0-manifold @xmath25 , the normal bundle to the diagonal of @xmath177 in @xmath177
is identified with the tangent bundle to @xmath25 , fiberwise , by the map @xmath178 a _ parallelization _ @xmath179 of an oriented @xmath0-manifold @xmath25 is a bundle isomorphism @xmath180 that restricts to @xmath181 as an orientation - preserving linear isomorphism from @xmath181 to @xmath99 , for any @xmath182 .
it has long been known that any oriented @xmath0-manifold is parallelizable ( i.e. admits a parallelization ) .
( it is proved in subsection [ subproofpar ] . ) therefore , a tubular neighborhood of the diagonal in @xmath177 is diffeomorphic to @xmath183 . [ lemhomstwo ]
let @xmath25 be a rational homology @xmath0sphere , let @xmath184 be a point of @xmath25 .
let @xmath185 .
then @xmath186 has the same rational homology as @xmath15 .
let @xmath104 be a ball in @xmath187 and let @xmath47 be a point inside @xmath104 , then the class @xmath165 $ ] of @xmath188 is a canonical generator of @xmath189 $ ] .
proof : in this proof , the homology coefficients are in @xmath116 . since @xmath187 has the homology of a point , the knneth formula implies that @xmath190 has the homology of a point .
now , by excision , @xmath191 @xmath192 using the long exact sequence of the pair @xmath193 , we get that @xmath194 .
define the _ gauss linking number _ of two disjoint links @xmath10 and @xmath11 in @xmath187 so that @xmath195=lk_g(j , k)[s]\ ] ] in @xmath196 .
note that the two definitions of @xmath176 coincide when @xmath197 .
[ propdeflkeq ] @xmath198 proof : first note that both definitions make sense when @xmath10 and @xmath11 are disjoint links : @xmath199=lk_g(j , k)[s]$ ] and @xmath147 is the algebraic intersection of @xmath11 and a rational chain @xmath141 bounded by @xmath10 .
if @xmath11 is a knot , then the chain @xmath141 of @xmath187 provides a rational cobordism @xmath117 between @xmath10 and a combination of meridians of @xmath11 in @xmath200 , and a rational cobordism @xmath201 in @xmath186 , which allow us to see that @xmath202 and @xmath203 linearly depend on @xmath157 \in h_1(\check{m}\setminus k)$ ] .
thus we are left with the proof that @xmath204 .
since @xmath205 linearly depends on @xmath206 \in h_1(\check{m}\setminus m_k)$ ] , we are left with the proof @xmath207 when @xmath11 is a meridian of @xmath159 .
now , there is no loss in assuming that our link is a hopf link in @xmath4 so that the equality follows from the equality for the positive hopf link in @xmath4 . for a @xmath1component link @xmath146 in @xmath4 ,
the definition of @xmath147 can be rewritten as @xmath208 for any regular value @xmath66 of @xmath23 , and for any @xmath1-form @xmath64 of @xmath15 such that @xmath65 .
thus , @xmath147 is the evaluation of a @xmath1-form @xmath209 of @xmath210 at the @xmath1-cycle @xmath211 $ ] , or it is the intersection of the @xmath1-cycle @xmath211 $ ] with a @xmath2-manifold @xmath212 , which will later be seen as the interior of a prototypical propagator .
we will adapt these definitions to rational homology @xmath0spheres in subsection [ subprop ] . the definition of the linking number that we will generalize in order to produce more powerful invariants is contained in lemma [ lemlkprop ] .
propagators will be the key ingredient to define powerful invariants from graph configurations in section [ secconstcsi ] .
they are defined in subsection [ subprop ] below after needed preliminaries .
they allow us to define the @xmath3invariant as an invariant of parallelized homology @xmath0balls in subsection [ subthetapar ] .
the @xmath3invariant is next turned to an invariant of rational homology @xmath0spheres in subsection [ subthetarat ] .
let @xmath95 be a submanifold of a smooth manifold @xmath104 , and let @xmath213 denote its unit normal bundle .
the fiber @xmath214 of @xmath213 is oriented as the boundary of a unit ball of @xmath215 . here , _ blowing up _
such a submanifold @xmath95 of codimension @xmath216 of @xmath104 means replacing @xmath95 by @xmath213 . for small open subspaces
@xmath217 of @xmath95 , @xmath2180,\infty [ \times s^{c-1 } ) \times u_a\right)$ ] is replaced by @xmath219 , so that the blown - up manifold @xmath220 is homeomorphic to the complement in @xmath104 of an open tubular neighborhood ( thought of as infinitely small ) of @xmath95 . in particular ,
@xmath221 is homotopy equivalent to @xmath222 .
furthermore , the blow up is canonical , so that the created boundary is @xmath223 and there is a canonical smooth projection from @xmath220 to @xmath104 such that the preimage of @xmath224 is @xmath225 . if @xmath95 and @xmath104 are compact , then @xmath220 is compact , it is a smooth compactification of @xmath222 . in the following figure , we see the result of blowing up @xmath226 in @xmath227 , and the closures in @xmath228 of @xmath229 , @xmath230 and the diagonal of @xmath227 , successively .
( -2,-1 ) rectangle ( 14,1 ) ; ( -1,-1 ) rectangle ( 1,1 ) ; ( -1,0 ) ( 1,0 ) node[right]@xmath231 ; ( 0,-1 ) ( 0,1 ) node[right]@xmath232 ; ( -1,-1 )
( 1,1 ) node[right]diag ; ( 3,0 ) nodeblow up @xmath226 ; ( 2,-.2 )
( 4,-.2 ) ; ( -1,-1 ) rectangle ( 1,1 ) ; ( -1,0 ) ( 1,0 ) ; ( 0,-1 ) ( 0,1 ) ; ( -1,-1 ) ( 1,1 ) ; ( 0,0 ) circle ( .3 ) ; ( .5,-.5 ) ( .3,-.3 ) ; ( .2,-.6 ) node[right]unit normal bundle to @xmath226 ; ( 3.4,0 ) nodeblow up the blown - up lines ; ( 2,-.2 ) ( 4,-.2 ) ; ( -1,-1 ) rectangle ( 1,1 ) ; ( -.93,-1 ) ( 1,.93 ) ( -1,-.93 ) ( .93,1 ) ; ( -.12,-1 ) rectangle ( .12,1 ) ( -1,-.12 ) rectangle ( 1,.12 ) ; ( 0,0 ) circle ( .3 ) ; ( -1,-1 ) ( -.93,-1 ) ( 1,.93 ) ( 1,1 ) ( .93,1 ) ( -1,-.93 ) ( -1,-1 ) ; ( -.1,-1.1 ) rectangle ( .1,1.1 ) ( -1.1,-.1 ) rectangle ( 1.1,.1 ) ; see @xmath156 as @xmath233 or as two copies of @xmath4 identified along @xmath234 by the ( exceptionally orientation - reversing ) diffeomorphism @xmath235 .
then @xmath236 where the unit normal bundle @xmath237 to @xmath184 in @xmath156 is canonically diffeomorphic to @xmath15 via @xmath238 , where @xmath239 is the limit of a sequence of points of @xmath4 approaching @xmath184 along a line directed by @xmath240 .
@xmath241 fix a rational homology @xmath0sphere @xmath25 , a point @xmath184 of @xmath25 , and @xmath242 .
identify a neighborhood of @xmath184 in @xmath25 with the complement @xmath243 of the closed ball @xmath244 of radius @xmath13 in @xmath4 .
let @xmath245 be the complement of the closed ball @xmath246 of radius @xmath1 in @xmath4 , which is a smaller neighborhood of @xmath184 in @xmath25 via the understood identification .
then @xmath247 is a compact rational homology ball diffeomorphic to @xmath248 .
define the _ configuration space _
@xmath249 as the compact @xmath250manifold with boundary and ridges obtained from @xmath177 by blowing up @xmath251 , the closures in @xmath252 of @xmath253 , @xmath254 and the diagonal of @xmath190 , successively
. then @xmath255 contains the unit normal bundle @xmath256 to the diagonal of @xmath190 .
this bundle is canonically isomorphic to the unit tangent bundle @xmath257 to @xmath187 ( again via the map @xmath258 \mapsto [ y - x]\right)$ ] ) .
[ lemstrucctwo ] let @xmath259 .
the open manifold @xmath260 is @xmath261 and the inclusion @xmath262 is a homotopy equivalence .
in particular , @xmath249 is a compactification of @xmath261 homotopy equivalent to @xmath261 .
the manifold @xmath249 is a smooth compact @xmath250-dimensional manifold with boundary and ridges .
there is a canonical smooth projection @xmath263 .
@xmath264 proof : let @xmath265 be the complement of the open ball of radius one of @xmath4 in @xmath156 . blowing up @xmath251 in @xmath266
transforms a neighborhood of @xmath251 into the product @xmath267 .
explicitly , there is a map @xmath2680,1 [ , ( x \neq 0,y \neq 0 ) \in s^5 \subset ( { \mathbb{r}}^3)^2 ) & \mapsto & ( \frac1{\lambda\norm{x}^2}x,\frac1{\lambda\norm{y}^2}y ) \end{array}\ ] ] that is a diffeomorphism onto its image , which is a neighborhood of the preimage of @xmath251 under the blow - down map @xmath269 .
this neighborhood intersects @xmath270 , @xmath271 , and @xmath272 as @xmath2730,1 [ \times 0 \times s^2)$ ] , @xmath2730,1 [ \times s^2\times 0)$ ] and @xmath2730,1 [ \times ( s^5 \cap \mbox{diag}(({\mathbb{r}}^3)^2)))$ ] , respectively .
in particular , the closures of @xmath270 , @xmath271 , and @xmath272 in @xmath252 intersect the boundary @xmath274 of @xmath252 as three disjoint spheres in @xmath275 , and they read @xmath276 , @xmath277 and @xmath278 . thus , the next steps will be three blow - ups along these three disjoint smooth manifolds .
these blow - ups will preserve the product structure @xmath279 .
therefore , @xmath249 is a smooth compact @xmath250-dimensional manifold with boundary , with three _ ridges _
@xmath280 in @xmath281 .
a neighborhood of these ridges in @xmath249 is diffeomorphic to @xmath282 .
[ lemextproj ] the map @xmath283 of lemma [ lempstt ] smoothly extends to @xmath284 , and its extension @xmath283 satisfies : @xmath285 where @xmath286 and @xmath287 denote the projections on the first and second factor with respect to the above expressions .
proof : near the diagonal of @xmath4 , we have a chart of @xmath284 @xmath288 that maps @xmath2890,\infty [ , y \in s^2)$ ] to @xmath290 . here , @xmath283 extends as the projection onto the @xmath15 factor .
+ consider the orientation - reversing embedding @xmath291 @xmath292 note that this chart induces the already mentioned identification @xmath293 of the ill - oriented unit normal bundle @xmath294 to @xmath295 in @xmath156 with @xmath15 .
when @xmath296 , @xmath297 then @xmath283 can be smoothly extended on @xmath298 ( where @xmath299 ) by @xmath300 similarly , set @xmath301 now , with the map @xmath302 of the proof of lemma [ lemstrucctwo ] , when @xmath47 and @xmath51 are not equal to zero and when they are distinct , @xmath303 when @xmath304 .
this map naturally extends to @xmath252 outside the boundaries of @xmath276 , @xmath277 and @xmath305 by keeping the same formula when @xmath306 .
let us check that @xmath283 smoothly extends over the boundary of the diagonal of @xmath248 .
there is a chart of @xmath249 near the preimage of this boundary in @xmath249 @xmath307 that maps @xmath3080,\infty [ , \mu \in ] 0,\infty [ , x \in s^2 , y \in s^2)$ ] to @xmath309 where @xmath283 reads @xmath310 and therefore smoothly extends when @xmath299 .
we similarly check that @xmath283 smoothly extends over the boundaries of @xmath311 and @xmath312 .
let @xmath313 denote the standard parallelization of @xmath4 .
say that a parallelization @xmath314 of @xmath187 that coincides with @xmath313 on @xmath243 is _ asymptotically standard .
_ according to subsection [ subproofpar ] , such a parallelization exists .
such a parallelization identifies @xmath257 with @xmath315 .
[ propprojbord ] for any asymptotically standard parallelization @xmath179 of @xmath187 , there exists a smooth map @xmath316 such that @xmath317 where @xmath286 and @xmath287 denote the projections on the first and second factor with respect to the above expressions .
proof : this is a consequence of lemma [ lemextproj ] .
since @xmath249 is homotopy equivalent to @xmath318 , according to lemma [ lemhomstwo ] , @xmath319 $ ] where the canonical generator @xmath165 $ ] is the homology class of a fiber of @xmath320 . for a @xmath1-component link @xmath146 of @xmath187 , the homology class @xmath211 $ ] of @xmath321 in @xmath322 reads @xmath323 $ ] , according to subsection [ subgengaussdef ] and to proposition [ propdeflkeq ] .
define an _
asymptotic rational homology @xmath4 _ as a pair @xmath324 where @xmath187 is @xmath0-manifold that reads as the union over @xmath881,2 ] \times s^2 $ ] of a rational homology ball @xmath325 and the complement @xmath243 of the unit ball of @xmath4 , and @xmath179 is an asymptotically standard parallelization of @xmath187 .
since such a pair @xmath324 canonically defines the rational homology @xmath0sphere @xmath326 , `` let @xmath324 be an asymptotic rational homology @xmath4 '' is a shortcut for `` let @xmath25 be a rational homology @xmath0sphere equipped with an asymptotically standard parallelization @xmath179 of @xmath187 '' .
let @xmath324 be an asymptotic rational homology @xmath4 .
a _ propagating chain _ of @xmath327 is a @xmath2chain @xmath328 of @xmath249 such that @xmath329 for some @xmath330 .
a _ propagating form _ of @xmath327 is a closed @xmath1-form @xmath331 on @xmath249 whose restriction to @xmath255 reads @xmath332 for some @xmath1-form @xmath64 of @xmath15 such that @xmath65 .
propagating chains and propagating forms are simply called _ propagators _ when their nature is clear from the context . recall the map @xmath333 of lemma [ lemextproj ] . for any @xmath330
, @xmath334 is a propagating chain of @xmath335 , and for any @xmath1-form @xmath64 of @xmath15 such that @xmath65 , @xmath209 is a propagating form of @xmath335 .
propagating chains exist because the @xmath0-cycle @xmath336 of @xmath255 bounds in @xmath249 since @xmath337 .
dually , propagating forms exist because the restriction induces a surjective map @xmath338 since @xmath339 .
explicit constructions of propagating chains associated to morse functions or heegaard diagrams can be found in @xcite .
[ lemlkprop ] let @xmath324 be an asymptotic rational homology @xmath4 .
let @xmath117 be a two - cycle of @xmath249 .
for any propagating chain @xmath328 of @xmath327 transverse to @xmath117 and for any propagating form @xmath331 of @xmath327 , @xmath340=\int_{c}\omega_p[s]=\langle c , { { \cal p}}\rangle_{c_2(m)}[s]$ ] in @xmath319 $ ] .
in particular , for any two - component link @xmath146 of @xmath187 .
@xmath341 proof : fix a propagating chain @xmath328 , the algebraic intersection @xmath342 only depends on the homology class @xmath340 $ ] of @xmath117 in @xmath249 .
similarly , since @xmath331 is closed , @xmath343 only depends on @xmath340 $ ] .
( indeed , if @xmath117 and @xmath344 cobound a chain @xmath345 , @xmath346 and @xmath347 cobound @xmath348 , and @xmath349 according to the stokes theorem . )
furthermore , the dependance on @xmath340 $ ] is linear .
therefore it suffices to check the lemma for a cycle that represents the canonical generator @xmath165 $ ] of @xmath322 .
any fiber of @xmath257 is such a cycle .
note that the intersection of transverse ( oriented ) submanifolds is an associative operation , so that @xmath350 is well defined .
furthermore , for a connected manifold @xmath53 , the class of a @xmath351-cycle in @xmath352={\mathbb{q}}$ ] is a well - defined number , so that the _ algebraic intersection _ of several transverse submanifolds whose codimension sum is the dimension of the ambient manifold is well defined as the homology class of their ( oriented ) intersection .
this extends to rational chains , multilinearly .
thus , for three such transverse submanifolds @xmath95 , @xmath104 , @xmath117 in a manifold @xmath345 , their algebraic intersection @xmath353 is the sum over the intersection points @xmath354 of the associated signs , where the sign of @xmath354 is positive if and only if the orientation of @xmath345 is induced by the orientation of @xmath355 .
let @xmath324 be an asymptotic rational homology @xmath4 .
let @xmath356 , @xmath357 and @xmath358 be three pairwise transverse propagators of @xmath327 with respective boundaries @xmath336 , @xmath359 and @xmath360 for three distinct points @xmath354 , @xmath361 and @xmath216 of @xmath15 , then @xmath362 does not depend on the chosen propagators @xmath356 , @xmath357 and @xmath358 .
it is a topological invariant of @xmath363 .
for any three propagating chains @xmath364 , @xmath365 and @xmath366 of @xmath327 , @xmath367 proof : since @xmath368 , if the propagator @xmath356 is changed to a propagator @xmath369 with the same boundary , @xmath370 bounds a @xmath371-dimensional chain @xmath372 transverse to @xmath373 . the @xmath13-dimensional chain @xmath374 does not meet @xmath255 since @xmath373 does not meet @xmath255 .
therefore , up to a well - determined sign , the boundary of @xmath374 is @xmath375 .
this shows that @xmath376 is independent of @xmath356 when @xmath354 is fixed .
similarly , it is independent of @xmath357 and @xmath358 when @xmath361 and @xmath216 are fixed .
thus , @xmath376 is a rational function on the connected set of triples @xmath377 of distinct point of @xmath15 .
it is easy to see that this function is continuous .
thus , it is constant .
let us similarly prove that @xmath378 is independent of the propagating forms @xmath364 , @xmath365 and @xmath366 .
assume that the form @xmath364 , which restricts to @xmath255 as @xmath379 , is changed to @xmath380 , which restricts to @xmath255 as @xmath381 .
[ lemetactwo ] there exists a one - form @xmath382 on @xmath15 such that @xmath383 .
for any such @xmath382 , there exists a one - form @xmath384 on @xmath249 such that @xmath385 , and the restriction of @xmath384 to @xmath255 is @xmath386 .
proof of the lemma : since @xmath364 and @xmath380 are cohomologous , there exists a one - form @xmath384 on @xmath249 such that @xmath387 .
similarly , since @xmath388 , there exists a one - form @xmath382 on @xmath15 such that @xmath383 .
on @xmath255 , @xmath389 .
thanks to the exact sequence @xmath390 @xmath391 .
therefore , there exists a function @xmath392 from @xmath255 to @xmath44 such that @xmath393 on @xmath394 .
extend @xmath392 to a @xmath8 map on @xmath249 and change @xmath384 into @xmath395 .
then @xmath396 since any @xmath371-form on @xmath15 vanishes .
thus , @xmath378 is independent of the propagating forms @xmath364 , @xmath365 and @xmath366 .
now , we can choose the propagating forms @xmath364 , @xmath365 and @xmath366 , poincar dual to @xmath356 , @xmath357 and @xmath358 , and supported in very small neighborhoods of @xmath356 , @xmath357 and @xmath358 , respectively , so that the intersection of the three supports is a very small neighborhood of @xmath397 , where it can easily be seen that @xmath398 .
in particular , @xmath399 reads @xmath400 for any propagating chain @xmath64 of @xmath327 .
since such a propagating chain represents the linking number , @xmath399 can be thought of as the _ cube of the linking number with respect to @xmath179 . _
when @xmath179 varies continuously , @xmath399 varies continuously in @xmath116 so that @xmath399 is an invariant of the homotopy class of @xmath179 . in this subsection
, @xmath25 denotes a smooth , compact oriented @xmath0-manifold with possible boundary @xmath100 .
recall that such a @xmath0-manifold is parallelizable .
let @xmath401 denote the group of orientation - preserving linear isomorphisms of @xmath4 .
let @xmath402 denote the set of maps @xmath403 from @xmath25 to @xmath401 that send @xmath100 to the unit @xmath13 of @xmath401 .
let @xmath404 $ ] denote the group of homotopy classes of such maps , with the group structure induced by the multiplication of maps , using the multiplication in @xmath401 .
for a map @xmath405 in @xmath402 , set @xmath406 let @xmath407 be a parallelization of @xmath25 .
then any parallelization @xmath179 of @xmath25 that coincides with @xmath408 on @xmath100 reads @xmath409 for some @xmath410 .
thus , fixing @xmath408 identifies the set of homotopy classes of parallelizations of @xmath25 fixed on @xmath100 with the group @xmath404 $ ] .
since @xmath401 deformation retracts onto the group @xmath411 of orientation - preserving linear isometries of @xmath4 , @xmath404 $ ] is isomorphic to @xmath412 $ ] .
see @xmath156 as @xmath413 where @xmath103 is the standard ball of radius @xmath414 of @xmath4 seen as @xmath415 \times s^2)/(0 \sim \{0\ } \times s^2)$ ] .
let @xmath416 map @xmath417 , v \in s^2)$ ] to the rotation @xmath418 with axis directed by @xmath419 and with angle @xmath420 .
this map induces the double covering @xmath421 , which orients @xmath411 and which allows one to deduce the first three homotopy groups of @xmath411 from the ones of @xmath156 .
they are @xmath422 , @xmath423 and @xmath424 $ ] . for @xmath425 ,
@xmath426 is generated by the class of the loop that maps @xmath427 to the rotation @xmath418 .
note that a map @xmath405 from @xmath428 to @xmath429 has a degree @xmath430 , which may be defined as the differential degree at a regular value ( different from @xmath13 ) of @xmath405 .
it can also be defined homologically , by @xmath431 = \mbox{deg}(g)[so(3),1]$ ] .
the following theorem is proved in section [ secmorepar ] .
[ thmpone ] for any smooth compact connected oriented @xmath0-manifold @xmath25 , the group @xmath432\ ] ] is abelian , and the degree @xmath433 \longrightarrow { \mathbb{z}}\ ] ] is a group homomorphism , which induces an isomorphism @xmath433 \otimes_{{\mathbb{z}}}{\mathbb{q}}\longrightarrow { \mathbb{q}}.\ ] ] when @xmath434 , ( resp .
when @xmath435 ) , there exists a canonical map @xmath286 from the set of homotopy classes of parallelizations of @xmath25 ( resp .
that coincide with @xmath313 near @xmath15 ) such that for any map @xmath405 in @xmath436 , for any trivialization @xmath179 of @xmath437 @xmath438 the definition of the map @xmath286 is given in subsection [ subdefpont ] , it involves relative pontrjagin classes .
when @xmath434 , the map @xmath286 coincides with the map @xmath45 that is studied by kirby and melvin in @xcite under the name _ hirzebruch defect .
_ see also @xcite .
since @xmath412 $ ] is abelian , the set of parallelizations of @xmath25 that are fixed on @xmath100 is an affine space with translation group @xmath412 $ ] .
recall that @xmath416 maps @xmath417 , v \in s^2)$ ] to the rotation with axis directed by @xmath419 and with angle @xmath420 .
let @xmath25 be an oriented connected @xmath0-manifold with possible boundary .
for a ball @xmath103 embedded in @xmath25 , let @xmath439 be a ( continuous ) map that coincides with @xmath440 on @xmath103 and that maps the complement of @xmath103 to the unit of @xmath411 .
the homotopy class of @xmath441 is well - defined .
[ lemdegrho ] @xmath442 proof : exercise .
recall that an asymptotic rational homology @xmath4 is a pair @xmath324 where @xmath187 is @xmath0-manifold that reads as the union over @xmath881,2 ] \times s^2 $ ] of a rational homology ball @xmath325 and the complement @xmath243 of the unit ball of @xmath4 , and that is equipped with an asymptotically standard parallelization @xmath179 . in this subsection
, we prove the following proposition .
[ propthetap ] let @xmath324 be an asymptotic rational homology @xmath4 .
for any map @xmath405 in @xmath443 trivially extended to @xmath187 , @xmath444 theorem [ thmpone ] allows us to derive the following corollary from proposition [ propthetap ] .
[ corthetap ] @xmath445 is an invariant of @xmath116-spheres .
[ lemtheprimhom ] @xmath446 is independent of @xmath179 .
set @xmath447 .
then @xmath448 is a homomorphism from @xmath449 $ ] to @xmath116 .
proof : for @xmath450 , @xmath361 or @xmath216 , the propagator @xmath451 of @xmath327 can be assumed to be a product @xmath452 \times p_{\tau |ub_m}^{-1}(d)$ ] on a collar @xmath452 \times ub_m$ ] of @xmath453 in @xmath249 . since @xmath454 \times ub_m;{\mathbb{q}})=0 $ ] , @xmath455 \times p_{\tau |ub_m}^{-1}(d ) ) \setminus ( 0 \times p_{\tau |ub_m}^{-1}(d ) ) \right ) \cup ( 0 \times p_{\tau \circ \psi_{{\mathbb{r}}}(g ) |ub_m}^{-1}(d))$ ] bounds a chain @xmath456 .
the chains @xmath457 , @xmath458 and @xmath459 can be assumed to be transverse .
construct the propagator @xmath460 of @xmath461 from @xmath451 by replacing @xmath452 \times p_{\tau |ub_m}^{-1}(d)$ ] by @xmath456 on @xmath452 \times ub_m$ ] .
then @xmath462 \times ub_m}.\ ] ] using @xmath179 to identify @xmath453 with @xmath463 allows us to see that @xmath446 is independent of @xmath179 .
then it is easy to observe that @xmath448 is a homomorphism from @xmath464 $ ] to @xmath116 .
according to theorem [ thmpone ] and to lemma [ lemdegrho ] , it suffices to prove that @xmath465 in order to prove proposition [ propthetap ] .
it is easy to see that @xmath466 .
thus , we are left with the proof of the following lemma .
[ lemtheprimrho ] @xmath467 . again ,
see @xmath103 as @xmath415 \times s^2)/(0 \sim \{0\ } \times s^2)$ ] .
we first prove the following lemma : [ leml ] let @xmath354 be the north pole .
the point @xmath468 is regular for the map @xmath469 and its preimage ( cooriented by @xmath15 via @xmath470 ) is the knot @xmath471 , where @xmath472 is the equator that bounds the southern hemisphere .
proof : it is easy to see that @xmath473 .
( -1,-1.2 ) rectangle ( 1,1.2 ) ; ( 0,0 ) circle ( 1 ) ( -1.1,-.05 ) node[above]@xmath47 ( 0,1 ) node[above]@xmath354 ( 0,-.95 ) node[below]@xmath474 ( 0,-.2 ) node[below]@xmath475 ; ( -1,0 ) .. controls ( -.95,-.1 ) and ( -.3,-.25 ) .. ( 0,-.25 ) ; ( 0,-.25 ) .. controls ( .3,-.25 ) and ( .95,-.1 ) .. ( 1,0 ) ; ( -1,0 ) to ( 1,0 ) ; ( -1,0 ) ( -1,-.3 ) node[below]@xmath13 ; ( 0,-1 ) ( .6,-1 ) node[right]@xmath476 ; ( -1,0 ) ( -1.3,0 ) node[left]@xmath1 ; ( 0,1 ) circle ( 1.5pt ) ( 0,-1 ) circle ( 1.5pt ) ; let @xmath477 .
when @xmath478 moves along the great circle that contains @xmath354 and @xmath47 from @xmath47 towards @xmath468 in @xmath479 , @xmath480 moves from @xmath468 in the same direction , which will be the direction of the tangent vector @xmath476 of @xmath15 at @xmath468 , counterclockwise in our picture , where @xmath47 is on the left .
then in our picture , @xmath15 is oriented at @xmath468 by @xmath476 and by the tangent vector @xmath481 at @xmath468 towards us . in order to move @xmath482 in the @xmath481 direction ,
one increases @xmath420 so that @xmath475 is cooriented and oriented like in the figure .
proof of lemma [ lemtheprimrho ] : we use the notation of the proof of lemma [ lemtheprimhom ] and we construct an explicit @xmath457 in @xmath452 \times ub^3 \stackrel{{{\tau_s } } } { = } [ -1,0 ] \times b^3 \times s^2 $ ] . when @xmath483 , there is a unique geodesic arc @xmath484 $ ] with length @xmath485 from @xmath354 to @xmath486 . for @xmath487 $ ] ,
let @xmath488 $ ] be such that the length of @xmath489 $ ] is @xmath490 .
this defines @xmath491 on @xmath492 , @xmath493 .
let us show how the definition of @xmath491 smoothly extends on the manifold @xmath494 obtained from @xmath103 by blowing up @xmath475 . the map @xmath470 maps the normal bundle to @xmath475 to a disk of @xmath15 around @xmath468 , by an orientation - preserving diffeomorphism on every fiber ( near the origin ) .
in particular , @xmath470 induces a map @xmath495 from the unit normal bundle to @xmath475 to the unit normal bundle to @xmath468 in @xmath15 , which preserves the orientation of the fibers .
then for an element @xmath51 of the unit normal bundle to @xmath475 in @xmath25 , define @xmath496 as before on the half great circle @xmath497_{\tilde{\rho}_a(-y)}$ ] from @xmath354 to @xmath468 that is tangent to @xmath498 at @xmath468 ( so that @xmath498 is an outward normal to @xmath497_{\tilde{\rho}_a(-y)}$ ] at @xmath468 ) .
this extends the definition of @xmath491 , continuously .
the whole sphere is covered with degree @xmath75 by the image of @xmath499 \times u{{\mathfrak{v}}}_x(l_a))$ ] , where the fiber @xmath500 of the unit normal bundle to @xmath475 is oriented as the boundary of a disk in the fiber of the normal bundle .
let @xmath501 be the closure of @xmath502 , m\in ( b^3 \setminus l_a ) } \left(m,{x_t}(m)\right)\right)$ ] in @xmath503 .
@xmath504 , m\in { b\!\ell(b^3,l_a)}}\left(p_{b^3}(m),{x_t}(m)\right).\ ] ] then @xmath505}x_{t}(-\partial { b\!\ell(s^3,l_a)})\ ] ] where @xmath506 is oriented like @xmath507 so that the last summand reads @xmath508 because the sphere is covered with degree @xmath75 by the image of @xmath499 \times u{{\mathfrak{v}}}_x(l_a))$ ] .
let @xmath509 be a disk bounded by @xmath475 in @xmath103 .
set @xmath510 so that @xmath511 . now let @xmath512 be the endomorphism of @xmath503 over @xmath103 that maps a unit vector to the opposite one . set @xmath513\times b^3\times a & + \{-2/3\}\times g(a ) & + [ -2/3,0]\times \cup_{m\in b^3}(m,\rho_a(m))\\ \mbox{and } & g_{-a}&=[-1,-1/3]\times b^3 \times ( -a ) & + \{-1/3\}\times \iota(g(a ) ) & + [ -1/3,0]\times \cup_{m\in b^3}(m,\rho(m)(-a ) ) .
\end{array}\ ] ] then @xmath514 \times l_a \times ( -a ) + \{-2/3\ } \times d_a \times ( -a ) -\{-1/3\ } \times \cup_{m\in d_a}(m,\rho_a(m)).\ ] ] finally , @xmath515 is the algebraic intersection of @xmath516 with @xmath517 in @xmath249 .
this intersection coincides with the algebraic intersection of @xmath516 with any propagator of @xmath327 according to lemma [ lemlkprop ] .
therefore @xmath518\times s^2 \times b^3}=-\mbox{deg}_a(\rho_a \colon d_a \rightarrow s^2).\ ] ] the orientation of @xmath475 allows us to choose @xmath519 as the northern hemisphere , the image of this hemisphere under @xmath470 covers the sphere with degree @xmath13 so that @xmath467 .
this section contains the needed background from the theory of finite type invariants .
it allows us to introduce the target space generated by feynman - jacobi diagrams , for the general invariants presented in section [ secconstcsi ] , in a progressive way .
theories of finite type invariants are useful to characterize invariants . such a theory allowed greg kuperberg and dylan thurston to identify @xmath520 with the casson invariant @xmath521 for integer homology @xmath0-spheres , in @xcite .
the invariant @xmath521 was defined by casson in 1984 as an algebraic count of conjugacy classes of irreducible representations from @xmath522 to @xmath523 .
see @xcite .
the kuperberg
thurston result above was generalized to the case of rational homology @xmath0-spheres in ( * ? ? ?
* theorem 2.6 and corollary 6.14 ) .
thus , for any rational homology @xmath0sphere @xmath25 , @xmath524 where @xmath521 is the walker generalization of the casson invariant to rational homology @xmath0spheres , which is normalized like in @xcite for integer homology @xmath0spheres , and like @xmath525 for rational homology @xmath0spheres with respect to the walker normalisation @xmath526 of @xcite . for invariants of knots and links in @xmath4 , the base of the theory of finite type invariants
was mainly established by bar - natan in @xcite .
a more complete review of this theory has been written by chmutov , duzhin and mostovoy in @xcite .
for integer homology @xmath0spheres , the theory was started by ohtsuki in @xcite and further developed by goussarov , habiro , le and others .
see @xcite .
delphine moussard developed a theory of finite type invariants for rational homology @xmath0spheres in @xcite .
her suitable theory is based on the lagrangian - preserving surgeries defined below .
[ defhhh ] an _ integer ( resp .
rational ) homology handlebody _ of genus @xmath405 is a compact oriented @xmath0-manifold @xmath95 that has the same integral ( resp .
rational ) homology as the usual solid handlebody @xmath527 below .
( 0,-1 ) rectangle ( 11,1 ) ; ( 2,.9 ) .. controls ( 1.9,.9 ) and ( 1.75,.7 ) .. ( 1.75,.5 ) node[left]@xmath528 ; ( 1.75,.5 ) .. controls ( 1.75,.3 ) and ( 1.9,.1 ) .. ( 2,.1 ) ; ( 2,.9 ) .. controls ( 2.1,.9 ) and ( 2.25,.7 ) .. ( 2.25,.5 ) .. controls ( 2.25,.3 ) and ( 2.1,.1 ) .. ( 2,.1 ) ; plot[smooth ] coordinates(1.4,.1 ) ( 1.6,0 ) ( 2,-.1 ) ( 2.4,0 ) ( 2.6,.1 ) ; plot[smooth ] coordinates(1.6,0 ) ( 2,.1 ) ( 2.4,0 ) ; plot[smooth ] coordinates(6.5,.4 ) ( 6.3,.5 ) ( 5,.9 ) ( 3.5,.4 ) ( 2,.9 ) ( .6,.35 ) ( .6,-.35 ) ( 2,-.9 ) ( 3.5,-.4 ) ( 5,-.9 ) ( 6.3,-.5 ) ( 6.5,-.4 ) ; ( 6.5,-.4 ) ( 7.5,-.4 ) ( 6.5,.4 ) ( 7.5,.4 ) ; plot[smooth ] coordinates(7.5,.4 ) ( 7.7,.5 ) ( 9,.9 ) ( 10.4,.35 ) ( 10.4,-.35 ) ( 9,-.9 ) ( 7.7,-.5 ) ( 7.5,-.4 ) ; ( 2,.9 ) .. controls ( 1.9,.9 ) and ( 1.75,.7 ) .. ( 1.75,.5 ) node[left]@xmath529 ; ( 1.75,.5 ) .. controls ( 1.75,.3 ) and ( 1.9,.1 ) .. ( 2,.1 ) ; ( 2,.9 ) .. controls ( 2.1,.9 ) and ( 2.25,.7 ) .. ( 2.25,.5 ) .. controls ( 2.25,.3 ) and ( 2.1,.1 ) .. ( 2,.1 ) ; plot[smooth ] coordinates(1.4,.1 ) ( 1.6,0 ) ( 2,-.1 ) ( 2.4,0 ) ( 2.6,.1 ) ; plot[smooth ] coordinates(1.6,0 ) ( 2,.1 ) ( 2.4,0 ) ; ( 2,.9 ) .. controls ( 1.9,.9 ) and ( 1.75,.7 ) .. ( 1.75,.5 ) node[left]@xmath530 ; ( 1.75,.5 ) .. controls ( 1.75,.3 ) and ( 1.9,.1 ) .. ( 2,.1 ) ; ( 2,.9 ) .. controls ( 2.1,.9 ) and ( 2.25,.7 ) .. ( 2.25,.5 ) .. controls ( 2.25,.3 ) and ( 2.1,.1 ) .. ( 2,.1 ) ; plot[smooth ] coordinates(1.4,.1 ) ( 1.6,0 ) ( 2,-.1 ) ( 2.4,0 ) ( 2.6,.1 ) ; plot[smooth ] coordinates(1.6,0 ) ( 2,.1 ) ( 2.4,0 ) ; show that if @xmath95 is a rational homology handlebody of genus @xmath405 , then @xmath531 is a genus @xmath405 surface .
the _ lagrangian _ @xmath532 of a compact @xmath0-manifold @xmath95 is the kernel of the map induced by the inclusion from @xmath533 to @xmath533 . in the figure ,
the lagrangian of @xmath527 is freely generated by the classes of the curves @xmath534 .
[ deflagsur ] an _ integral ( resp .
rational ) lagrangian - preserving ( or lp ) surgery _
@xmath535 is the replacement of an integral ( resp .
rational ) homology handlebody @xmath95 embedded in the interior of a @xmath0-manifold @xmath25 by another such @xmath536 whose boundary is identified with @xmath531 by an orientation - preserving diffeomorphism that sends @xmath532 to @xmath537 .
the manifold @xmath538 obtained by such an lp - surgery reads @xmath539 ( this only defines the topological structure of @xmath538 , but we equip @xmath538 with its unique smooth structure . ) if @xmath535 is an integral ( resp .
rational ) lp - surgery , then the homology of @xmath538 with @xmath115-coefficients ( resp . with @xmath116-coefficients ) is canonically isomorphic to @xmath540 ( resp . to @xmath541 ) .
if @xmath25 is a @xmath116-sphere , if @xmath535 is a rational lp - surgery , and if @xmath146 is a two - component link of @xmath542 , then the linking number of @xmath10 and @xmath11 in @xmath25 and the linking number of @xmath10 and @xmath11 in @xmath538 coincide .
proof : exercise .
let @xmath543 or @xmath44 .
a @xmath155valued _ invariant _ of oriented @xmath0-manifolds is a function from the set of @xmath0-manifolds , considered up to orientation - preserving diffeomorphisms to @xmath155 .
let @xmath544 denote a disjoint union of @xmath24 circles , where each @xmath545 is a copy of @xmath5 . here ,
an _ @xmath24component link _ in a @xmath0-manifold @xmath25 is an equivalence class of smooth embeddings @xmath546 under the equivalence relation that identifies two embeddings @xmath547 and @xmath548 if and only if there is an orientation - preserving diffeomorphism @xmath45 of @xmath25 such that @xmath549 .
a _ knot _ is a one - component link .
a _ link invariant _
knot invariant _ ) is a function of links ( resp .
for example , @xmath3 is an invariant of @xmath116-spheres and the linking number is a rational invariant of two - component links in rational homology @xmath0spheres in order to study a function , it is usual to study its derivative , and the derivative of its derivative ... the derivative of a function is defined from its variations . for a function @xmath392 from @xmath550 to @xmath155 , one can define its first order derivatives @xmath551 by @xmath552 and check that all the first order derivatives of @xmath392 vanish if and only if @xmath392 is constant . inductively define an @xmath24-order derivative as a first order derivative of an @xmath553-order derivative for a positive integer @xmath24 .
then it can be checked that all the @xmath554-order derivatives of a function vanish if and only if @xmath392 is a polynomial of degree not greater than @xmath24 . in order to study topological invariants , we can similarly study their variations under _ simple operations . _ below
, @xmath555 denotes one of the following sets * @xmath556 , * the set @xmath557 of knots in @xmath4 , the set @xmath558 of @xmath24-component links in @xmath4 , * the set @xmath559 of @xmath115-spheres , the set @xmath560 of @xmath116-spheres .
and @xmath561 denotes a set of _ simple operations _ acting on some elements of @xmath555 . for @xmath562 ,
@xmath561 will be made of the operations @xmath563 for knots or links in @xmath4 , the _ simple operations _ will be _ crossing changes . _
a _ crossing change ball _ of a link @xmath547 is a ball @xmath104 of the ambient space , where @xmath564 is a disjoint union of two arcs @xmath565 and @xmath566 properly embedded in @xmath104 , and there exist two disjoint topological disks @xmath567 and @xmath568 embedded in @xmath104 , such that , for @xmath569 , @xmath570 and @xmath571 . after an isotopy , the projection of @xmath572 reads @xmath573 ( -.2,0 ) -- ( .2,.4 ) ; \draw [ ->,draw = white , double = black , very thick ] ( .2,0 ) -- ( -.2,.4 ) ; \draw [ - > ] ( .2,0 ) -- ( -.2,.4 ) ; \draw [ dash pattern = on 2pt off 2pt ] ( 0,.2 ) circle ( .283 ) ; \end{tikzpicture}}$ ] or @xmath574 ( .2,0 ) -- ( -.2,.4 ) ; \draw [ ->,draw
= white , double = black , very thick ] ( -.2,0 ) -- ( .2,.4 ) ; \draw [ - > ] ( -.2,0 ) -- ( .2,.4 ) ; \draw [ dash pattern = on 2pt off 2pt ] ( 0,.2 ) circle ( .283 ) ; \end{tikzpicture}}$ ] ( the corresponding pairs ( ball , arcs ) are isomorphic , but they are regarded in different ways ) , a _ crossing change _ is a change that does not change @xmath547 outside @xmath104 and that modifies @xmath547 inside @xmath104 by a local move @xmath575 ( -.2,0 ) -- ( .2,.4 ) ; \draw [ ->,draw = white , double = black , very thick ] ( .2,0 ) -- ( -.2,.4 ) ; \draw [ - > ] ( .2,0 ) -- ( -.2,.4 ) ; \draw [ dash pattern = on 2pt off 2pt ] ( 0,.2 ) circle ( .283 ) ; \end{tikzpicture}}\rightarrow { \begin{tikzpicture } \useasboundingbox ( -.5,.1 ) rectangle ( .5,.55 ) ; \draw [ - > ] ( .2,0 ) -- ( -.2,.4 ) ;
\draw [ ->,draw = white , double = black , very thick ] ( -.2,0 ) -- ( .2,.4 ) ; \draw [ - > ] ( -.2,0 ) -- ( .2,.4 ) ; \draw [ dash pattern = on 2pt off 2pt ] ( 0,.2 ) circle ( .283 ) ; \end{tikzpicture}})$ ] or @xmath576 ( .2,0 ) -- ( -.2,.4 ) ; \draw [ ->,draw
= white , double = black , very thick ] ( -.2,0 ) -- ( .2,.4 ) ; \draw [ - > ] ( -.2,0 ) -- ( .2,.4 ) ; \draw [ dash pattern = on 2pt off 2pt ] ( 0,.2 ) circle ( .283 ) ; \end{tikzpicture}}\rightarrow { \begin{tikzpicture } \useasboundingbox ( -.5,.1 ) rectangle ( .5,.55 ) ; \draw [ - > ] ( -.2,0 ) -- ( .2,.4 ) ; \draw [ ->,draw = white , double = black , very thick ] ( .2,0 ) -- ( -.2,.4 ) ; \draw [ - > ] ( .2,0 ) -- ( -.2,.4 ) ; \draw [ dash pattern = on 2pt off 2pt ] ( 0,.2 ) circle ( .283 ) ; \end{tikzpicture}})$ ] . for the move @xmath575 ( -.2,0 ) -- ( .2,.4 ) ; \draw [ ->,draw = white , double = black , very thick ] ( .2,0 ) -- ( -.2,.4 ) ; \draw [ - > ] ( .2,0 ) -- ( -.2,.4 ) ; \draw [ dash pattern = on 2pt off 2pt ] ( 0,.2 ) circle ( .283 ) ; \end{tikzpicture}}\rightarrow { \begin{tikzpicture } \useasboundingbox ( -.5,.1 ) rectangle ( .5,.55 ) ; \draw [ - > ] ( .2,0 ) -- ( -.2,.4 ) ; \draw [ ->,draw = white , double = black , very thick ] ( -.2,0 ) -- ( .2,.4 ) ; \draw [ - > ] ( -.2,0 ) -- ( .2,.4 ) ; \draw [ dash pattern = on 2pt off 2pt ] ( 0,.2 ) circle ( .283 ) ; \end{tikzpicture}})$ ] , the crossing change is _ positive _ , it is _ negative _ for the move @xmath576 ( .2,0 ) -- ( -.2,.4 ) ; \draw [ ->,draw
= white , double = black , very thick ] ( -.2,0 ) -- ( .2,.4 ) ; \draw [ - > ] ( -.2,0 ) -- ( .2,.4 ) ; \draw [ dash pattern = on 2pt off 2pt ] ( 0,.2 ) circle ( .283 ) ; \end{tikzpicture}}\rightarrow { \begin{tikzpicture } \useasboundingbox ( -.5,.1 ) rectangle ( .5,.55 ) ; \draw [ - > ] ( -.2,0 ) -- ( .2,.4 ) ; \draw [ ->,draw = white , double = black , very thick ] ( .2,0 ) -- ( -.2,.4 ) ; \draw [ - > ] ( .2,0 ) -- ( -.2,.4 ) ; \draw [ dash pattern = on 2pt off 2pt ] ( 0,.2 ) circle ( .283 ) ; \end{tikzpicture}})$ ] . for integer ( resp .
rational ) homology @xmath0spheres , the simple operations will be integral ( resp .
rational ) _ @xmath577-surgeries of genus @xmath0 .
_ say that crossing changes are _ disjoint _ if they sit inside disjoint @xmath0-balls .
say that @xmath577-surgeries @xmath535 and @xmath578 in a manifold @xmath25 are _ disjoint _ if @xmath95 and @xmath104 are disjoint in @xmath25 .
two operations on @xmath556 are always _ disjoint _ ( even if they look identical ) .
in particular , disjoint operations commute , ( their result does not depend on which one is performed first ) .
let @xmath579 .
consider the vector space @xmath580 freely generated by @xmath555 over @xmath155 .
for an element @xmath47 of @xmath555 and @xmath24 pairwise disjoint operations @xmath581 acting on @xmath47 , define @xmath582=\sum_{i \subseteq \underline{n } } ( -1)^{\sharp i}x((o_i)_{i \in i } ) \in { { \cal f}}_0(x)\ ] ] where @xmath583 denotes the element of @xmath555 obtained by performing the operations @xmath584 for @xmath28 on @xmath47 . then define @xmath585 as the @xmath155-subspace of @xmath580 generated by the @xmath586 $ ] , for all @xmath587 equipped with @xmath24 pairwise disjoint simple operations . since @xmath588= [ x;o_1,\dots , o_n ] - [ x(o_{n+1});o_1,\dots , o_n],\ ] ] @xmath589 , for all @xmath590 .
a @xmath155valued function @xmath392 on @xmath555 , uniquely extends as a @xmath155linear map of @xmath591 which is still denoted by @xmath392 . for an integer @xmath590 , the invariant ( or function ) @xmath392 is of _ degree @xmath592 _ if and only if @xmath593 . the _ degree _ of such an invariant is the smallest integer @xmath590 such that @xmath593 .
an invariant is of _ finite type _ if it is of degree @xmath24 for some @xmath590 .
this definition depends on the chosen set of operations @xmath561 .
we fixed our choices for our sets @xmath555 , but other choices could lead to different notions . see @xcite .
let @xmath594 be the space of invariants of degree at most @xmath24 .
of course , for all @xmath590 , @xmath595 .
@xmath596 is the space of polynomials of degree at most @xmath24 on @xmath556 .
( exercise ) . any @xmath24-component link in @xmath4
can be transformed to the trivial @xmath24-component link below by a finite number of disjoint crossing changes .
( -.5,-.6 ) rectangle ( 5.5,.6 ) ; ( -.5,0 ) arc ( -180:180:.5 ) ; ( -.5,0 ) node[left]@xmath597 ; ( 2,0 ) arc ( -180:180:.5 ) ; ( 2,0 ) node[left]@xmath598 ; ( 3.5,0 ) node ; ( 5,0 ) arc ( -180:180:.5 ) ; ( 5,0 ) node[left]@xmath599 ; proof : let @xmath547 be an @xmath24-component link in @xmath4 . since @xmath4 is simply connected , there is a homotopy that carries @xmath547 to the trivial link .
such a homotopy @xmath600\times \coprod_{i=1}^n s^1_i\rightarrow { \mathbb{r}}^3 $ ] can be chosen , so that @xmath601 is an embedding except for finitely many times @xmath602 , @xmath603 where @xmath604 is an immersion with one double point and no other multiple points , and the link @xmath601 changes exactly by a crossing change when @xmath605 crosses a @xmath602 .
( for an alternative elementary proof of this lemma , see ( * ? ? ? * subsection 7.1 ) ) . in particular , a degree @xmath351 invariant of @xmath24-component links of @xmath4 must be constant , since it is not allowed to vary under a crossing change .
check that @xmath606 , where @xmath607 is the constant map that maps any knot to @xmath13 .
check that the linking number is a degree @xmath13 invariant of @xmath1component links of @xmath4 .
+ 3 . check that @xmath608 , where @xmath607 is the constant map that maps any two - component link to @xmath13 .
let @xmath392 be a knot invariant of degree at most @xmath24 .
we want to evaluate @xmath609)$ ] where the @xmath584 are disjoint negative crossing changes ( -.5,.1 ) rectangle ( .5,.55 ) ; ( .2,0 ) ( -.2,.4 ) ; ( -.2,0 ) ( .2,.4 ) ; ( -.2,0 ) ( .2,.4 ) ; @xmath610 ( -.5,.1 ) rectangle ( .5,.55 ) ; ( -.2,0 ) ( .2,.4 ) ; ( .2,0 ) ( -.2,.4 ) ; ( .2,0 )
( -.2,.4 ) ; to be performed on a knot @xmath11 .
such a @xmath611 $ ] is usually represented as a _ singular knot with @xmath24 double points _ that is an immersion of a circle with @xmath24 transverse double points ( -1,-.5 ) rectangle ( 1,.7 ) ; ( 30:.7 ) to ( 150:.3 ) ; ( 150:.3 ) to ( -90:.7 ) ; ( -90:.7 ) to ( 30:.3 ) ; ( 30:.3 ) to ( 150:.7 ) ; ( 150:.7 ) to ( -90:.3 ) ; ( -90:.3 ) to ( 30:.7 ) ; ( -30:.39 ) circle ( 2.5pt ) ( 90:.39 ) circle ( 2.5pt ) ( -150:.39 ) circle ( 2.5pt ) ; , where each double point ( -.5,.1 ) rectangle ( .5,.55 ) ; ( .2,0 ) ( -.2,.4 ) ; ( -.2,0 ) ( .2,.4 ) ; ( 0,.2 ) circle ( 1.5pt ) ; can be desingularized in two ways , the positive one ( -.5,.1 ) rectangle ( .5,.55 ) ; ( .2,0 ) ( -.2,.4 ) ; ( -.2,0 ) ( .2,.4 ) ; ( -.2,0 ) ( .2,.4 ) ; and the negative one
( -.5,.1 ) rectangle ( .5,.55 ) ; ( -.2,0 ) ( .2,.4 ) ; ( .2,0 ) ( -.2,.4 ) ; ( .2,0 ) ( -.2,.4 ) ; , and @xmath11 is obtained from the singular knot by desingularizing all the crossings in the positive way , which is ( -1,-.5 ) rectangle ( 1,.7 ) ; ( 30:.7 ) to ( 150:.3 ) ; ( -90:.7 ) to ( 30:.3 ) ; ( 150:.7 ) to ( -90:.3 ) ; ( 150:.3 ) to ( -90:.7 ) ; ( 30:.3 ) to ( 150:.7 ) ; ( -90:.3 ) to ( 30:.7 ) ; ( -90:.3 ) to ( 30:.7 ) ; in our example .
note that the sign of the desingularization is defined from the orientation of the ambient space .
define the _ chord diagram _
@xmath612)$ ] associated to @xmath611 $ ] as follows .
draw the preimage of the associated singular knot with @xmath24 double points as an oriented dashed circle equipped with the @xmath613 preimages of the double points and join the pairs of preimages of a double point by a plain segment called a
_ chord_. @xmath614 \useasboundingbox ( -1,-.5 ) rectangle ( 1,.7 ) ; \draw [ out=120,in=60 ] ( 30:.7 ) to ( 150:.3 ) ; \draw [ out=240,in=180 ] ( 150:.3 ) to ( -90:.7 ) ; \draw [ out=0,in=-60 ] ( -90:.7 ) to ( 30:.3 ) ; \draw [ out=120,in=60 ] ( 30:.3 ) to ( 150:.7 ) ; \draw [ out=240,in=180 ] ( 150:.7 ) to ( -90:.3 ) ; \draw [ out=0,in=-60,- > ] ( -90:.3 ) to ( 30:.7 ) ; \fill ( -30:.39 ) circle ( 2.5pt ) ( 90:.39 ) circle ( 2.5pt ) ( -150:.39 ) circle ( 2.5pt ) ; \end{tikzpicture}})={\begin{tikzpicture } \useasboundingbox ( -.4,-.2 ) rectangle ( .4,.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( .3,0 ) arc ( 0:360:.3 ) ; \draw ( -150:.3 ) -- ( 30:.3 ) ; \draw ( -30:.3 ) -- ( 150:.3 ) ; \draw ( 90:.3 ) -- ( -90:.3 ) ; \fill ( -30:.3 ) circle ( 1.5pt ) ( 30:.3 ) circle ( 1.5pt ) ( -90:.3 ) circle ( 1.5pt ) ( 90:.3 ) circle ( 1.5pt ) ( -150:.3 ) circle ( 1.5pt ) ( 150:.3 ) circle ( 1.5pt ) ; \end{tikzpicture}}\ ] ] formally , a _ chord diagram _ with @xmath24 chords is a cyclic order of the @xmath613 ends of the @xmath24 chords , up to a permutation of the chords and up to exchanging the two ends of a chord .
[ leminvtyf ] when @xmath392 is a knot invariant of degree at most @xmath24 , @xmath609)$ ] only depends on @xmath612)$ ] . proof : since @xmath392 is of degree @xmath24 , @xmath609)$ ] is invariant under a crossing change outside the balls of the @xmath584 , that is outside the double points of the associated singular knot .
therefore , @xmath609)$ ] only depends on the cyclic order of the @xmath613 arcs involved in the @xmath584 on @xmath11 .
let @xmath615 be the @xmath155-vector space freely generated by the @xmath24 chord diagrams on @xmath5 .
@xmath616 ( .3,0 ) arc ( 0:360:.3 ) ; \end{tikzpicture}}$ ] , @xmath617 ( .3,0 ) arc ( 0:360:.3 ) ; \draw ( 0,-.3 ) -- ( 0,.3 ) ; \fill ( 0,-.3 ) circle ( 1.5pt ) ( 0,.3 ) circle ( 1.5pt ) ; \end{tikzpicture}}$ ] , @xmath618 ( .3,0 ) arc ( 0:360:.3 ) ; \draw ( -45:.3 ) .. controls ( -90:.1 ) .. ( -135:.3);\draw ( 45:.3 ) .. controls ( 90:.1 ) .. ( 135:.3 ) ; \fill ( -45:.3 ) circle ( 1.5pt ) ( 45:.3 ) circle ( 1.5pt ) ( -135:.3 ) circle ( 1.5pt ) ( 135:.3 ) circle ( 1.5pt ) ; \end{tikzpicture}}\oplus { \mathbb{k}}{\begin{tikzpicture } \useasboundingbox ( -.4,-.2 ) rectangle ( .4,.3 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( .3,0 ) arc ( 0:360:.3 ) ; \draw ( -45:.3 ) -- ( 135:.3 ) ; \draw ( -135:.3 ) -- ( 45:.3 ) ; \fill ( -45:.3 ) circle ( 1.5pt ) ( 45:.3 ) circle ( 1.5pt ) ( -135:.3 ) circle ( 1.5pt ) ( 135:.3 ) circle ( 1.5pt ) ; \end{tikzpicture}}$ ] , @xmath619 ( .3,0 ) arc ( 0:360:.3 ) ; \draw ( -30:.3 ) .. controls ( -60:.1 ) .. ( -90:.3 ) ; \draw ( 30:.3 ) .. controls ( 60:.1 ) .. ( 90:.3 ) ; \draw ( -150:.3 ) .. controls ( 180:.1 ) .. ( 150:.3 ) ; \fill ( -30:.3 ) circle ( 1.5pt ) ( 30:.3 ) circle ( 1.5pt ) ( -90:.3 ) circle ( 1.5pt ) ( 90:.3 ) circle ( 1.5pt ) ( -150:.3 ) circle ( 1.5pt ) ( 150:.3 ) circle ( 1.5pt ) ; \end{tikzpicture}}\oplus { \mathbb{k}}{\begin{tikzpicture } \useasboundingbox ( -.4,-.2 ) rectangle ( .4,.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( .3,0 ) arc ( 0:360:.3 ) ; \draw ( -30:.3 ) -- ( 150:.3 ) ; \draw ( 30:.3 ) .. controls ( 60:.1 ) .. ( 90:.3 ) ; \draw ( -150:.3 ) .. controls ( -120:.1 ) .. ( -90:.3 ) ; \fill ( -30:.3 ) circle ( 1.5pt ) ( 30:.3 ) circle ( 1.5pt ) ( -90:.3 ) circle ( 1.5pt ) ( 90:.3 ) circle ( 1.5pt ) ( -150:.3 ) circle ( 1.5pt ) ( 150:.3 ) circle ( 1.5pt ) ; \end{tikzpicture}}\oplus { \mathbb{k}}{\begin{tikzpicture } \useasboundingbox ( -.4,-.2 ) rectangle ( .4,.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( .3,0 ) arc ( 0:360:.3 ) ; \draw ( -150:.3 ) .. controls ( 180:.1 ) .. ( 150:.3 ) ; \draw ( 30:.3 ) -- ( -90:.3 ) ; \draw ( 90:.3 ) -- ( -30:.3 ) ; \fill ( -30:.3 ) circle ( 1.5pt ) ( 30:.3 ) circle ( 1.5pt ) ( -90:.3 ) circle ( 1.5pt ) ( 90:.3 ) circle ( 1.5pt ) ( -150:.3 ) circle ( 1.5pt ) ( 150:.3 ) circle ( 1.5pt ) ; \end{tikzpicture}}\oplus { \mathbb{k}}{\begin{tikzpicture } \useasboundingbox ( -.4,-.2 ) rectangle ( .4,.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( .3,0 ) arc ( 0:360:.3 ) ; \draw ( -150:.3 ) -- ( -30:.3 ) ; \draw ( 90:.3 ) -- ( -90:.3 ) ; \draw ( 150:.3 ) -- ( 30:.3 ) ; \fill ( -30:.3 ) circle ( 1.5pt ) ( 30:.3 ) circle ( 1.5pt ) ( -90:.3 ) circle ( 1.5pt ) ( 90:.3 ) circle ( 1.5pt ) ( -150:.3 ) circle ( 1.5pt ) ( 150:.3 ) circle ( 1.5pt ) ; \end{tikzpicture}}\oplus { \mathbb{k}}{\begin{tikzpicture } \useasboundingbox ( -.4,-.2 ) rectangle ( .4,.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( .3,0 ) arc ( 0:360:.3 ) ; \draw ( -150:.3 ) -- ( 30:.3 ) ; \draw ( -30:.3 ) -- ( 150:.3 ) ; \draw ( 90:.3 ) -- ( -90:.3 ) ; \fill ( -30:.3 ) circle ( 1.5pt ) ( 30:.3 ) circle ( 1.5pt ) ( -90:.3 ) circle ( 1.5pt ) ( 90:.3 ) circle ( 1.5pt ) ( -150:.3 ) circle ( 1.5pt ) ( 150:.3 ) circle ( 1.5pt ) ; \end{tikzpicture}}$ ] .
the map @xmath620 from @xmath615 to @xmath621 that maps @xmath622 to some @xmath611 $ ] whose diagram is @xmath622 is well - defined and surjective
. proof : use the arguments of the proof of lemma [ leminvtyf ] .
@xmath623 ( .3,0 ) arc ( 0:360:.3 ) ; \draw ( -150:.3 ) -- ( 30:.3 ) ; \draw ( -30:.3 ) -- ( 150:.3 ) ; \draw ( 90:.3 ) -- ( -90:.3 ) ; \fill ( -30:.3 ) circle ( 1.5pt ) ( 30:.3 ) circle ( 1.5pt ) ( -90:.3 ) circle ( 1.5pt ) ( 90:.3 ) circle ( 1.5pt ) ( -150:.3 ) circle ( 1.5pt ) ( 150:.3 ) circle ( 1.5pt ) ; \end{tikzpicture}})=[{\begin{tikzpicture}[scale=.5 ] \useasboundingbox ( -1,-.5 ) rectangle ( 1,.7 ) ; \draw [ out=120,in=60 ] ( 30:.7 ) to ( 150:.3 ) ; \draw [ out=240,in=180 ] ( 150:.3 ) to ( -90:.7 ) ; \draw [ out=0,in=-60 ] ( -90:.7 ) to ( 30:.3 ) ; \draw [ out=120,in=60 ] ( 30:.3 ) to ( 150:.7 ) ; \draw [ out=240,in=180 ] ( 150:.7 ) to ( -90:.3 ) ; \draw [ out=0,in=-60,- > ] ( -90:.3 ) to ( 30:.7 ) ; \fill ( -30:.39 ) circle ( 2.5pt ) ( 90:.39 ) circle ( 2.5pt ) ( -150:.39 ) circle ( 2.5pt ) ; \end{tikzpicture}}].\ ] ] the kernel of the composition of @xmath624 and the restriction below @xmath625 is @xmath626 .
thus , @xmath627 injects into @xmath628 and @xmath629 is finite dimensional for all @xmath24 .
furthermore , @xmath630 an _ isolated chord _ in a chord diagram is a chord between two points of @xmath5 that are consecutive on the circle . [ lem14 t ] let @xmath345 be a diagram on @xmath5 that contains an isolated chord .
then @xmath631 let @xmath632 , @xmath633 , @xmath634 , @xmath635 be four n - chord diagrams that are identical outside three portions of circles where they look like : @xmath636 ( 70:.5 ) arc ( 70:110:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 170:.5 ) arc ( 170:240:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( -60:.5 ) arc ( -60:10:.5 ) ; \draw ( 90:.5 ) --(-40:.5 ) ( -10:.5 ) -- ( 205:.5 ) ; \fill ( 90:.5 ) circle ( 1.5pt ) ( 205:.5 ) circle ( 1.5pt ) ( -40:.5 ) circle ( 1.5pt ) ( -10:.5 ) circle ( 1.5pt ) ; \end{tikzpicture}},\;\;\;\ ; d^2= { \begin{tikzpicture } \useasboundingbox ( -.8,-.4 ) rectangle ( .8,.6 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 70:.5 ) arc ( 70:110:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 170:.5 ) arc ( 170:240:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( -60:.5 ) arc ( -60:10:.5 ) ; \draw ( 90:.5 ) -- ( 220:.5 ) ( 190:.5 ) -- ( -25:.5 ) ; \fill ( 90:.5 ) circle ( 1.5pt ) ( 220:.5 ) circle ( 1.5pt ) ( -25:.5 ) circle ( 1.5pt ) ( 190:.5 ) circle ( 1.5pt ) ; \end{tikzpicture}},\;\;\;\ ; d^3= { \begin{tikzpicture } \useasboundingbox ( -.8,-.4 ) rectangle ( .8,.6 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 70:.5 ) arc ( 70:110:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 170:.5 ) arc ( 170:240:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( -60:.5 ) arc ( -60:10:.5 ) ; \draw ( 90:.5 ) -- ( -10:.5 ) ( -40:.5 ) -- ( 205:.5 ) ; \fill ( 90:.5 ) circle ( 1.5pt ) ( 205:.5 ) circle ( 1.5pt ) ( -40:.5 ) circle ( 1.5pt ) ( -10:.5 ) circle ( 1.5pt ) ; \end{tikzpicture}}\;\;\;\mbox{and}\;\;\ ; d^4={\begin{tikzpicture } \useasboundingbox ( -.8,-.4 ) rectangle ( .8,.6 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 70:.5 ) arc ( 70:110:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 170:.5 ) arc ( 170:240:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( -60:.5 ) arc ( -60:10:.5 ) ; \draw ( 90:.5 ) --(190:.5 ) ( 220:.5 ) -- ( -25:.5 ) ; \fill ( 90:.5 ) circle ( 1.5pt ) ( 220:.5 ) circle ( 1.5pt ) ( -25:.5 ) circle ( 1.5pt ) ( 190:.5 ) circle ( 1.5pt ) ; \end{tikzpicture}},\ ] ] then @xmath637 proof : for the first assertion , observe that @xmath638 ( -90:.3 ) arc ( -90:90:.3 ) ; \draw ( -45:.3 ) -- ( 45:.3 ) ; \fill ( -45:.3 ) circle ( 1.5pt ) ( 45:.3 ) circle ( 1.5pt ) ; \end{tikzpicture}})=[{\begin{tikzpicture } \useasboundingbox ( -.1,-.1 ) rectangle ( .5,.2 ) ; \draw [ out=-90,in=-90,- > ] ( .4,0 ) to ( 0,.25 ) ; \draw [ out=90,in=90,draw = white , double = black , very thick ] ( 0,-.25 ) to ( .4,0 ) ; \end{tikzpicture}}]-[{\begin{tikzpicture } \useasboundingbox ( -.1,-.1 ) rectangle ( .5,.2 ) ; \draw [ out=90,in=90 ] ( 0,-.2 ) to ( .4,0 ) ; \draw [ out=-90,in=-90,draw = white , double = black , very thick ] ( .4,0 ) to ( 0,.2 ) ; \draw [ out=-90,in=-90,- > ] ( .4,0 ) to ( 0,.2 ) ; \end{tikzpicture}}]$ ] .
for the second one , see ( * ? ? ? * lemma 2.21 ) , for example .
let @xmath639 denote the quotient of @xmath615 by the _ four term relation , _ which is the quotient of @xmath615 by the vector space generated by the @xmath640 for all the @xmath2-tuples @xmath641 as above .
call @xmath642 the relation that identifies a diagram with an isolated chord with @xmath351 so that @xmath643 is the quotient of @xmath639 by the vector space generated by diagrams with an isolated chord . according to lemma [ lem14 t ] above
, the map @xmath620 induces a map @xmath644 the fundamental theorem of _ vassiliev invariants _ ( which are finite type knot invariants ) can now be stated .
[ thmbn ] there exists a family of linear maps @xmath645 such that * @xmath646 , * @xmath647 induces the inverse of @xmath648 from @xmath621 to @xmath643 .
in particular @xmath649 and @xmath650 .
this theorem has been proved by kontsevich and bar - natan in @xcite using the _ kontsevich integral _
@xmath651 described in @xcite and in ( * ? ? ?
* chapter 8) , for @xmath652 .
it is also true when @xmath543 .
* note * the kontsevich integral has been generalized to a functor from the category of framed tangles to a category of jacobi diagrams by le and murakami in @xcite .
le and murakami showed how to derive the reshetikhin - turaev quantum invariants of framed links in @xmath4 defined in @xcite from their functor , in ( * ? ? ? * theorem 10 ) .
[ defunitrivgra ] a _ uni - trivalent graph _
@xmath622 is a @xmath250-tuple @xmath653 where @xmath654 , @xmath655 , @xmath656 and @xmath657 are finite sets , which are called the set of half - edges of @xmath622 , the set of edges of @xmath622 , the set of univalent vertices of @xmath622 and the set of trivalent vertices of @xmath622 , respectively , @xmath658 is a two - to - one map ( every element of @xmath655 has two preimages under @xmath659 ) and @xmath660 is a map such that every element of @xmath656 has one preimage under @xmath661 and every element of @xmath657 has three preimages under @xmath661 , up to isomorphism . in other words , @xmath622 is a set @xmath654 equipped with two partitions , a partition into pairs ( induced by @xmath659 ) , and a partition into singletons and triples ( induced by @xmath661 ) , up to the bijections that preserve the partitions .
these bijections are the _ automorphisms _ of @xmath622 .
[ defdia ] let @xmath117 be an oriented one - manifold .
a _ jacobi diagram @xmath622 with support @xmath117 , _ also called jacobi diagram on @xmath117 , is a finite uni - trivalent graph @xmath622 equipped with an isotopy class of injections @xmath662 of the set @xmath656 of univalent vertices of @xmath622 into the interior of @xmath117 .
vertex - orientation _ of a jacobi diagram @xmath622 is an _ orientation _ of every trivalent vertex of @xmath622 , which is a cyclic order on the set of the three half - edges which meet at this vertex . a jacobi diagram is _ oriented _ if it is equipped with a vertex - orientation .
such an oriented jacobi diagram @xmath622 is represented by a planar immersion of @xmath663 where the univalent vertices of @xmath656 are located at their images under @xmath662 , the one - manifold @xmath117 is represented by dashed lines , whereas the diagram @xmath622 is plain .
the vertices are represented by big points .
the local orientation of a vertex is represented by the counterclockwise order of the three half - edges that meet at it .
here is an example of a picture of a jacobi diagram @xmath622 on the disjoint union @xmath664 of two circles : ( -1.4,-.2 ) rectangle ( 1.4,.5 ) ; ( 0,-.3 ) arc ( -90:270:.3 ) ; ( -1,-.3 ) arc ( -90:270:.3 ) ; ( -1.3,0 ) ( 30:.3 ) ; ( 0,0 ) ( -140:.3 ) ( 0,0 ) ( -40:.3 ) ( 0,0 ) ( 0,.3 ) ; ( -40:.3 ) circle ( 1.5pt ) ( -140:.3 ) circle ( 1.5pt ) ( 90:.3 ) circle ( 1.5pt ) ( 30:.3 ) circle ( 1.5pt ) ( -1.3,0 ) circle ( 1.5pt ) ( 0,0 ) circle ( 1.5pt ) ( 1.2,0 ) circle ( 1.5pt ) ( .8,0 ) circle ( 1.5pt ) ; ( 1,0 ) circle ( .2 ) ( .8,0 ) ( 1.2,0 ) ; the _ degree _ of such a diagram is half the number of all the vertices of @xmath622 . of course , a chord diagram of @xmath615 is a degree @xmath24 jacobi diagram on @xmath5 without trivalent vertices .
let @xmath665 denote the @xmath155-vector space generated by the degree @xmath24 oriented jacobi diagrams on @xmath117 .
@xmath666 ( .3,0 ) arc ( 0:360:.3 ) ; \draw ( 0,-.3 ) -- ( 0,.3 ) ; \fill ( 0,-.3 ) circle ( 1.5pt ) ( 0,.3 ) circle ( 1.5pt ) ; \end{tikzpicture}}\oplus { \mathbb{k}}{\begin{tikzpicture } \useasboundingbox ( -.4,-.2 ) rectangle ( .4,.3 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( .3,0 ) arc ( 0:360:.3 ) ; \draw ( 0,-.3 ) -- ( 0,-.1 ) ( 0,.05 ) circle ( .15 ) ; \fill ( 0,-.3 ) circle ( 1.5pt ) ( 0,-.1 ) circle ( 1.5pt ) ; \end{tikzpicture}}\oplus { \mathbb{k}}{\begin{tikzpicture } \useasboundingbox ( -.4,-.2 ) rectangle ( .4,.3 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( .3,0 ) arc ( 0:360:.3 ) ; \draw ( -.2,0 ) -- ( .2,0 ) ( 0,0 ) circle ( .2 ) ; \fill ( -.2,0 ) circle ( 1.5pt ) ( .2,0 ) circle ( 1.5pt ) ; \end{tikzpicture}}\oplus { \mathbb{k}}{\begin{tikzpicture } \useasboundingbox ( -.4,-.2 ) rectangle ( .4,.3 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( .3,0 ) arc ( 0:360:.3 ) ; \draw ( .2,0 ) arc ( 0:180:.2 ) ( -.2,0 ) -- ( -.05,0 ) .. controls ( .05,0 ) and ( .2,-.3 ) .. ( .2,0 ) ( -.2,0 ) arc ( -180:-90:.2 ) ( 0,-.2 ) .. controls ( .15,-.2 ) and ( 0,.1 ) .. ( .2,0 ) ; \fill ( -.2,0 ) circle ( 1.5pt ) ( .2,0 ) circle ( 1.5pt ) ; \end{tikzpicture}}\oplus { \mathbb{k}}{\begin{tikzpicture } \useasboundingbox ( -.4,-.2 ) rectangle ( .4,.3 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( .35,0 ) arc ( 0:360:.35 ) ; \draw ( 0,-.07 ) -- ( 0,.07 ) ( 0,.19 ) circle ( .12 ) ( 0,-.19 ) circle ( .12 ) ; \fill ( 0,-.07 ) circle ( 1.5pt ) ( 0,.07 ) circle ( 1.5pt ) ; \end{tikzpicture}}\ ] ] let @xmath667 denote the quotient of @xmath665 by the following relations as , jacobi and stu : as : ( -.4,-.3 ) rectangle ( .4,.4 ) ; ( 0,0 ) ( 0,-.4 ) ( 0,0 ) ( 50:.4 ) ( 0,0 ) ( 130:.4 ) ; ( 0,0 ) circle ( 1.5pt ) ; + ( -.4,-.2 ) rectangle ( .4,.3 ) ; ( 0,0 ) ( 0,-.2 ) ; ( 0,0 ) .. controls ( 50:.3 ) .. ( 130:.5 ) ; ( 0,0 ) .. controls ( 130:.3 ) .. ( 50:.5 ) ; ( 0,0 ) circle ( 1.5pt ) ; @xmath668 + jacobi : ( -.4,-.3 ) rectangle ( .4,.4 ) ; ( -90:.4 ) ( 0,0 ) ( 30:.4 ) ( 0,0 ) ( 150:.4 ) ; ( -.3,-.25 ) ( 0,-.25 ) ; ( 0,-.25 ) circle ( 1.5pt ) ( 0,0 ) circle ( 1.5pt ) ; + ( -.4,-.3 ) rectangle ( .4,.4 ) ; ( -90:.4 ) ( 0,0 ) ( 30:.4 ) ( 0,0 ) ( 150:.4 ) ; ( -.3,-.25 ) to ( 30:.25 ) ; ( 30:.25 ) circle ( 1.5pt ) ( 0,0 ) circle ( 1.5pt ) ; + ( -.4,-.3 ) rectangle ( .4,.4 ) ; ( -90:.4 ) ( 0,0 ) ( 30:.4 ) ( 0,0 ) ( 150:.4 ) ; ( -.3,-.25 ) .. controls ( -.2,-.25 ) and ( .25,-.2 ) .. ( .25,0 ) .. controls ( .25,.2 ) and ( 120:.35 ) .. ( 150:.25 ) ; ( 150:.25 ) circle ( 1.5pt ) ( 0,0 ) circle ( 1.5pt ) ; @xmath668 + stu : ( -.1,-.25 ) rectangle ( .7,.4 ) ; ( 0,-.2 ) ( .6,-.2 ) ; ( .3,-.2 ) ( .3,0 ) ( .15,.2 ) ( .3,0 ) ( .45,.2 ) ; ( .3,-.2 ) circle ( 1.5pt ) ( .3,0 ) circle ( 1.5pt ) ; @xmath669 ( -.1,-.25 ) rectangle ( .7,.4 ) ; ( 0,-.2 ) ( .6,-.2 ) ; ( .15,-.2 ) ( .15,.2 ) ( .45,.2 ) ( .45,-.2 ) ; ( .15,-.2 ) circle ( 1.5pt ) ( .45,-.2 ) circle ( 1.5pt ) ; - ( -.1,-.25 ) rectangle ( .7,.4 ) ; ( 0,-.2 ) ( .6,-.2 ) ; ( .15,.2 ) ( .45,-.2 ) ; ( .15,-.2 ) ( .45,.2 ) ; ( .15,-.2 ) circle ( 1.5pt ) ( .45,-.2 ) circle ( 1.5pt ) ; as before , each of these relations relate oriented jacobi diagrams which are identical outside the pictures where they are like in the pictures .
lie algebras provide nontrivial linear maps , called _ weight systems _ from @xmath667 to @xmath155 , see @xcite and ( * ? ? ?
* section 6 ) . in the weight system constructions ,
the jacobi relation for the lie bracket ensures that the maps defined for oriented jacobi diagrams factor through the jacobi relation . in @xcite , pierre vogel proved that the maps associated to lie ( super)algebras are sufficient to detect nontrivial elements of @xmath667 until degree @xmath670 , and he exhibited a non trivial element of @xmath671 that can not be detected by such maps .
the jacobi relation was originally called ihx by bar - natan in @xcite because , up to as , it can be written as @xmath672 \useasboundingbox ( -.4,-.3 ) rectangle ( .4,.4 ) ; \draw ( -.3,-.3 ) -- ( .3,-.3 ) ( -.3,.3 ) -- ( .3,.3 ) ( 0,-.3 ) -- ( 0,.3 ) ; \fill ( 0,-.3 ) circle ( 2pt ) ( 0,.3 ) circle ( 2pt ) ; \end{tikzpicture}}={\begin{tikzpicture}[scale=.7 ] \useasboundingbox ( -.4,-.3 ) rectangle ( .4,.4 ) ; \draw ( -.3,-.3 ) -- ( -.3,.3 ) ( .3,-.3 ) -- ( .3,.3 ) ( -.3,0 ) -- ( .3,0 ) ; \fill ( -.3,0 ) circle ( 2pt ) ( .3,0 ) circle ( 2pt ) ; \end{tikzpicture}}- { \begin{tikzpicture}[scale=.7 ] \useasboundingbox ( -.4,-.3 ) rectangle ( .4,.4 ) ; \draw ( -.3,-.3 ) -- ( .3,.3 ) ( .3,-.3 ) -- ( -.3,.3 ) ( -.15,-.15 ) -- ( .15,-.15 ) ; \fill ( -.15,-.15 ) circle ( 2pt ) ( .15,-.15 ) circle ( 2pt ) ; \end{tikzpicture}}.$ ] set @xmath673 .
when @xmath674 , let @xmath675 denote the quotient of @xmath676 by the vector space generated by the diagrams that have at least one connected component without univalent vertices .
then @xmath677 is generated by the oriented jacobi diagrams whose ( plain ) connected components contain at least one univalent vertex .
the natural map from @xmath615 to @xmath678 induces an isomorphism from @xmath639 to @xmath678 .
sketch of proof : the natural map from @xmath615 to @xmath678 factors though @xmath679 since , according to @xmath680 , @xmath681 ( 70:.5 ) arc ( 70:110:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 170:.5 ) arc ( 170:240:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( -60:.5 ) arc ( -60:10:.5 ) ; \draw ( 90:.5 ) -- ( -10:.5 ) ( -40:.5 ) -- ( 205:.5 ) ; \fill ( 90:.5 ) circle ( 1.5pt ) ( 205:.5 ) circle ( 1.5pt ) ( -40:.5 ) circle ( 1.5pt ) ( -10:.5 ) circle ( 1.5pt ) ; \end{tikzpicture}}- { \begin{tikzpicture } \useasboundingbox ( -.8,-.4 ) rectangle ( .8,.6 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 70:.5 ) arc ( 70:110:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 170:.5 ) arc ( 170:240:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( -60:.5 ) arc ( -60:10:.5 ) ; \draw ( 90:.5 ) --(-40:.5 ) ( -10:.5 ) -- ( 205:.5 ) ; \fill ( 90:.5 ) circle ( 1.5pt ) ( 205:.5 ) circle ( 1.5pt ) ( -40:.5 ) circle ( 1.5pt ) ( -10:.5 ) circle ( 1.5pt ) ; \end{tikzpicture}}= { \begin{tikzpicture } \useasboundingbox ( -.8,-.4 ) rectangle ( .8,.6 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 70:.5 ) arc ( 70:110:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 170:.5 ) arc ( 170:240:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( -60:.5 ) arc ( -60:10:.5 ) ; \draw ( 0,0 ) -- ( -25:.5 ) ( 0,0 ) -- ( 90:.5 ) ( 0,0 ) -- ( 205:.5 ) ; \fill ( 90:.5 ) circle ( 1.5pt ) ( -25:.5 ) circle ( 1.5pt ) ( 205:.5 ) circle ( 1.5pt ) ( 0,0 ) circle ( 1.5pt ) ; \end{tikzpicture}}= { \begin{tikzpicture } \useasboundingbox ( -.8,-.4 ) rectangle ( .8,.6 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 70:.5 ) arc ( 70:110:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 170:.5 ) arc ( 170:240:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( -60:.5 ) arc ( -60:10:.5 ) ; \draw ( 90:.5 ) --(190:.5 ) ( 220:.5 ) -- ( -25:.5 ) ; \fill ( 90:.5 ) circle ( 1.5pt ) ( 220:.5 ) circle ( 1.5pt ) ( -25:.5 ) circle ( 1.5pt ) ( 190:.5 ) circle ( 1.5pt ) ; \end{tikzpicture}}-{\begin{tikzpicture } \useasboundingbox ( -.8,-.4 ) rectangle ( .8,.6 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 70:.5 ) arc ( 70:110:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 170:.5 ) arc ( 170:240:.5 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( -60:.5 ) arc ( -60:10:.5 ) ; \draw ( 90:.5 ) -- ( 220:.5 ) ( 190:.5 ) -- ( -25:.5 ) ; \fill ( 90:.5 ) circle ( 1.5pt ) ( 220:.5 ) circle ( 1.5pt ) ( -25:.5 ) circle ( 1.5pt ) ( 190:.5 ) circle ( 1.5pt ) ; \end{tikzpicture}}\ ] ] in @xmath682 .
since stu allows us to inductively write any oriented jacobi diagram whose connected components contain at least a univalent vertex as a combination of chord diagrams , the induced map from @xmath639 to @xmath678 is surjective . in order to prove injectivity ,
one constructs an inverse map
. see ( * ? ? ? * subsection 3.4 ) .
the le fundamental theorem on _ finite type invariants of @xmath115-spheres _ is the following one .
[ thmle ] there exists a family @xmath683 of linear maps such that * @xmath684 , * @xmath685 induces an isomorphism from @xmath686 to @xmath687 , * @xmath688 .
in particular @xmath689 and @xmath690 .
this theorem has been proved by le @xcite using the le - murakami - ohtsuki invariant @xmath691 of @xcite . as explained in @xcite ,
this lmo invariant contains the quantum witten - reshetikhin - invariants of rational homology @xmath0spheres defined in @xcite .
in @xcite , delphine moussard obtained a similar fundamental theorem for _ finite type invariants of @xmath116-spheres _ using the configuration space integral @xmath692 described in @xcite , @xcite and in theorem [ thmmain ] below .
as in the knot case , the hardest part of these theorems is the construction of an invariant @xmath693 that has the required properties .
we will define such an invariant by `` counting jacobi diagram configurations '' in subsection [ subconstcsi ] and explain why it satisfies the required so - called universality properties in subsection [ subuniv ] .
set @xmath694 and @xmath695 .
assume that a one - manifold @xmath117 is decomposed as a union of two one - manifolds @xmath696 whose interiors in @xmath117 do not intersect .
define the _ product associated to this decomposition _ : @xmath697 as the continuous bilinear map which maps @xmath698,[\gamma_2])$ ] to @xmath699 $ ] , if @xmath700 is a diagram with support @xmath701 and if @xmath702 is a diagram with support @xmath703 , where @xmath704 denotes their disjoint union .
in particular , the disjoint union of diagrams turns @xmath705 into a commutative algebra graded by the degree , and it turns @xmath706 into a @xmath705-module , for any @xmath13-dimensional manifold @xmath117 . an orientation - preserving diffeomorphism from a manifold @xmath117 to another one @xmath344 induces an isomorphism from @xmath677 to @xmath707 , for all @xmath24 . let @xmath708 $ ] be the compact oriented interval .
if @xmath709 , and if we identify @xmath27 with @xmath710 $ ] and with @xmath711 $ ] with respect to the orientation , then the above process turns @xmath712 into an algebra where the elements with non - zero degree zero part admit an inverse .
[ propdiagcom ] the algebra @xmath713)$ ] is commutative .
the projection from @xmath714 $ ] to @xmath715/(0\sim 1)$ ] induces an isomorphism from @xmath716)$ ] to @xmath678 for all @xmath24 , so that @xmath717 inherits a commutative algebra structure from this isomorphism .
the choice of a connected component @xmath718 of @xmath117 equips @xmath719 with an _
@xmath713)$]-module structure @xmath720 _ , induced by the inclusion from @xmath714 $ ] to a little part of @xmath718 outside the vertices , and the insertion of diagrams with support @xmath714 $ ] there .
in order to prove this proposition , we present a useful trick in diagram spaces . first adopt a convention .
so far , in a diagram picture , or in a chord diagram picture , the plain edge of a univalent vertex , has always been attached on the left - hand side of the oriented one - manifold .
now , if @xmath114 plain edges are attached on the other side on a diagram picture , then we agree that the corresponding represented element of @xmath721 is @xmath722 times the underlying diagram . with this convention , we have the new antisymmetry relation in @xmath721 : @xmath723 ( 0,-.2 ) -- ( .6,-.2 ) ; \draw ( .3,.2 ) -- ( .3,-.2 ) ; \fill ( .3,-.2 ) circle ( 1.5pt ) ; \end{tikzpicture}}+{\begin{tikzpicture } \useasboundingbox ( -.1,-.25 ) rectangle ( .7,.4 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 0,-.2 ) -- ( .6,-.2 ) ; \draw ( .3,.2 ) .. controls ( .1,0 ) and ( .05,-.35 ) .. ( .15,-.35 ) .. controls ( .25,-.35 ) .. ( .3,-.2 ) ; \fill ( .3,-.2 ) circle ( 1.5pt ) ; \end{tikzpicture}}=0\ ] ] and we can draw the stu relation like the jacobi relation : @xmath724 ( 0,-.2 ) -- ( 1,-.2 ) ; \draw ( .3,.3 ) -- ( .3,-.2 ) ( .7,.3 ) -- ( .7,-.2 ) ; \fill ( .3,-.2 ) circle ( 1.5pt ) ( .7,-.2 ) circle ( 1.5pt ) ; \end{tikzpicture}}+ { \begin{tikzpicture } \useasboundingbox ( -.1,-.25 ) rectangle ( 1.1,.4 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 0,-.2 ) -- ( 1,-.2 ) ; \draw ( .4,.3 ) .. controls ( .1,-.1 ) and ( .3,-.35 ) .. ( .4,-.35 ) .. controls ( .5,-.35 ) .. ( .7,-.2 ) ( .7,.3 ) -- ( .4,-.2 ) ; \fill ( .4,-.2 ) circle ( 1.5pt ) ( .7,-.2 ) circle ( 1.5pt ) ; \end{tikzpicture}}+ { \begin{tikzpicture } \useasboundingbox ( -.1,-.25 ) rectangle ( 1.1,.4 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 0,-.2 ) -- ( 1,-.2 ) ; \draw ( .2,.3 ) .. controls ( .2,-.2 ) and ( .7,-.1 ) .. ( .7,0 ) .. controls ( .7,.1 ) .. ( .5,.15 ) ( .5,.3 ) -- ( .5,-.2 ) ; \fill ( .5,-.2 ) circle ( 1.5pt ) ( .5,.15 ) circle ( 1.5pt ) ; \end{tikzpicture}}=0.\ ] ] [ lemcom ] let @xmath700 be a jacobi diagram with support @xmath117 .
assume that @xmath725 is immersed in the plane so that @xmath725 meets an open annulus @xmath95 embedded in the plane exactly along @xmath726 embedded arcs @xmath565 , @xmath566 , , @xmath727 and @xmath124 , and one vertex @xmath419 so that : 1 .
the @xmath728 may be dashed or plain , they run from a boundary component of @xmath95 to the other one , 2 .
@xmath124 is a plain arc which runs from the boundary of @xmath95 to @xmath729 , 3 .
the bounded component @xmath345 of the complement of @xmath95 does not contain a boundary point of @xmath117 .
let @xmath730 be the diagram obtained from @xmath700 by attaching the endpoint @xmath419 of @xmath124 to @xmath728 instead of @xmath565 on the same side , where the side of an arc is its side when going from the outside boundary component of @xmath95 to the inside one @xmath731 .
then @xmath732 in @xmath706 .
@xmath733 ( 0,0 ) circle ( .9 ) ; \draw [ fill = white , draw = white ] ( 0,0 ) circle ( .5 ) ( .6,-.2 ) rectangle ( .8,.2 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 0,-1 ) -- ( 0,1 ) ; \draw ( .7,0 ) node{\tiny $ a$ } ( 0,-.2 ) -- ( -.3,0 ) -- ( 0,0 ) ( -.3,0 ) -- ( 0,.2 ) ( -.5,-1 ) -- ( 0,-.7 ) ( .1,-.6 ) node{\tiny $ v$ } ( .2,.75 ) node{\tiny $ \alpha_2 $ } ( .2,-.8 ) node{\tiny $ \alpha_1 $ } ( .25,0 ) node{\tiny $ d$ } ( -.3,-.7 ) node{\tiny $ \beta$ } ( .8,-.8 ) node{$\gamma_1 $ } ; \fill ( 0,-.7 ) circle ( 1.5pt ) ( 0,-.2 ) circle ( 1.5pt ) ( 0,0 ) circle ( 1.5pt ) ( 0,.2 ) circle ( 1.5pt ) ( -.3,0 ) circle ( 1.5pt ) ; \end{tikzpicture}}+ { \begin{tikzpicture } \useasboundingbox ( -1.3,-.8 ) rectangle ( 1.3,1.2 ) ; \draw [ fill = gray!20,draw = white ] ( 0,0 ) circle ( .9 ) ; \draw [ fill = white , draw = white ] ( 0,0 ) circle ( .5 ) ( .6,-.2 ) rectangle ( .8,.2 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 0,-1 ) -- ( 0,1 ) ; \draw ( .7,0 ) node{\tiny $ a$ } ( 0,-.2 ) -- ( -.3,0 ) -- ( 0,0 ) ( -.3,0 ) -- ( 0,.2 ) ( -.5,-1 ) .. controls ( .1,-.7 ) and ( .55,-.4 ) .. ( .55,0 ) .. controls ( .55,.4 ) and ( .2,.7 ) .. ( 0,.7 ) ( -.1,.8 ) node{\tiny $ v$ } ( .2,.75 ) node{\tiny $ \alpha_2 $ } ( .25,0 ) node{\tiny $ d$ } ( .2,-.8 ) node{\tiny $ \alpha_1 $ } ( -.3,-.7 ) node{\tiny $ \beta$ } ( .8,-.8 ) node{$\gamma_2 $ } ; \fill ( 0,.7 ) circle ( 1.5pt ) ( 0,-.2 ) circle ( 1.5pt ) ( 0,0 ) circle ( 1.5pt ) ( 0,.2 ) circle ( 1.5pt ) ( -.3,0 ) circle ( 1.5pt ) ; \end{tikzpicture}}=0\ ] ] @xmath733 ( 0,0 ) circle ( .9 ) ; \draw [ fill = white , draw = white ] ( 0,0 ) circle ( .5 ) ( .6,-.2 ) rectangle ( .8,.2 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 0,-1 ) -- ( 0,1 ) ; \draw ( .7,0 ) node{\tiny $ a$ } ( -1,0 ) -- ( 0,0 ) ( -.5,-1 ) -- ( 0,-.7 ) ( .1,-.6 ) node{\tiny $ v$ } ( .25,0 ) node{\tiny $ d$ } ( .2,.75 ) node{\tiny $ \alpha_2 $ } ( .2,-.8 ) node{\tiny $ \alpha_1 $ } ( -.7,.15 ) node{\tiny $ \alpha_3 $ } ( -.3,-.7 ) node{\tiny $ \beta$ } ( .8,-.8 ) node{$\gamma_1 $ } ; \fill ( 0,-.7 ) circle ( 1.5pt ) ( 0,0 ) circle ( 1.5pt ) ; \end{tikzpicture}}+ { \begin{tikzpicture } \useasboundingbox ( -1.3,-.8 ) rectangle ( 1.3,1.2 ) ; \draw [ fill = gray!20,draw = white ] ( 0,0 ) circle ( .9 ) ; \draw [ fill = white , draw = white ] ( 0,0 ) circle ( .5 ) ( .6,-.2 ) rectangle ( .8,.2 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 0,-1 ) -- ( 0,1 ) ; \draw ( -.5,-1 ) .. controls ( .1,-.7 ) and ( .55,-.4 ) .. ( .55,0 ) .. controls ( .55,.4 ) and ( .2,.7 ) .. ( 0,.7 ) ( .7,0 ) node{\tiny $ a$ } ( -1,0 ) -- ( 0,0 ) ( -.1,.8 ) node{\tiny $ v$ } ( .25,0 ) node{\tiny $ d$ } ( .2,.75 ) node{\tiny $ \alpha_2 $ } ( .2,-.8 ) node{\tiny $ \alpha_1 $ } ( -.7,.15 ) node{\tiny $ \alpha_3 $ } ( -.3,-.7 ) node{\tiny $ \beta$ } ( .8,-.8 ) node{$\gamma_2 $ } ; \fill ( 0,.7 ) circle ( 1.5pt ) ( 0,0 ) circle ( 1.5pt ) ; \end{tikzpicture}}+ { \begin{tikzpicture } \useasboundingbox ( -1.3,-.8 ) rectangle ( 1.3,1.2 ) ; \draw [ fill = gray!20,draw = white ] ( 0,0 ) circle ( .9 ) ; \draw [ fill = white , draw = white ] ( 0,0 ) circle ( .5 ) ( .6,-.2 ) rectangle ( .8,.2 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 0,-1 ) -- ( 0,1 ) ; \draw ( -.5,-1 ) .. controls ( .1,-.7 ) and ( .55,-.4 ) .. ( .55,0 ) .. controls ( .55,.4 ) and ( .2,.7 ) .. ( 0,.7 ) .. controls ( -.25,.7 ) and ( -.7,.35 ) .. ( -.7,0 ) ( .7,0 ) node{\tiny $ a$ } ( -1,0 ) -- ( 0,0 ) ( -.8,.1 ) node{\tiny $ v$ } ( .25,0 ) node{\tiny $ d$ } ( .2,.75 ) node{\tiny $ \alpha_2 $ } ( .2,-.8 ) node{\tiny $ \alpha_1 $ } ( -.7,-.15 ) node{\tiny $ \alpha_3 $ } ( -.3,-.7 ) node{\tiny $ \beta$ } ( .8,-.8 ) node{$\gamma_3 $ } ; \fill ( -.7,0 ) circle ( 1.5pt ) ( 0,0 ) circle ( 1.5pt ) ; \end{tikzpicture}}=0\ ] ] proof : the second example shows that the stu relation is equivalent to this relation when the bounded component @xmath345 of @xmath734 intersects @xmath700 in the neighborhood of a univalent vertex on @xmath117 . similarly , the jacobi relation is easily seen as given by this relation when @xmath345 intersects @xmath700 in the neighborhood of a trivalent vertex . also note that as corresponds to the case when @xmath345 intersects @xmath700 along a dashed or plain arc .
now for the bar - natan ( * ? ? ?
* lemma 3.1 ) proof .
see also ( * ? ? ?
* lemma 3.3 ) .
assume without loss that @xmath419 is always attached on the right - hand - side of the @xmath123 s .
add to the sum the trivial ( by jacobi and stu ) contribution of the sum of the diagrams obtained from @xmath700 by attaching @xmath419 to each of the three ( dashed or plain ) half - edges of each vertex @xmath735 of @xmath725 in @xmath345 on the left - hand side when the half - edges are oriented towards @xmath735 .
now , group the terms of the obtained sum by edges of @xmath725 where @xmath419 is attached , and observe that the sum is zero edge by edge by as .
proof of proposition [ propdiagcom ] : to each choice of a connected component @xmath718 of @xmath117 , we associate an _ @xmath712-module structure @xmath720 on @xmath719 _ , which is given by the continuous bilinear map : @xmath736 such that : if @xmath737 is a diagram with support @xmath117 and if @xmath622 is a diagram with support @xmath27 , then @xmath738,[\gamma^{\prime}])$ ] is mapped to the class of the diagram obtained by inserting @xmath622 along @xmath718 outside the vertices of @xmath622 , according to the given orientation . for example , @xmath739 ( 0,-.8 ) -- ( 0,.8 ) ; \draw ( 0,-.2 ) -- ( -.3,0 ) -- ( 0,0 ) ( -.3,0 ) -- ( 0,.2 ) ; \fill ( 0,-.2 ) circle ( 1.5pt ) ( 0,0 ) circle ( 1.5pt ) ( 0,.2 ) circle ( 1.5pt ) ( -.3,0 ) circle ( 1.5pt ) ; \end{tikzpicture}}{\begin{tikzpicture } \useasboundingbox ( -1,-.8 ) rectangle ( 1,1 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( -.8,0 ) arc ( -180:180:.8 ) ; \draw ( -135:.8 ) -- ( 0,-.15 ) -- ( 0,.15 ) -- ( 135:.8 ) ( 0,.15 ) -- ( 45:.8 ) ( 0,-.15 ) -- ( -45:.8 ) ; \fill ( -135:.8 ) circle ( 1.5pt ) ( 135:.8 ) circle ( 1.5pt ) ( 45:.8 ) circle ( 1.5pt ) ( -45:.8 ) circle ( 1.5pt ) ( 0,-.15 ) circle ( 1.5pt ) ( 0,.15 ) circle ( 1.5pt ) ; \end{tikzpicture}}= { \begin{tikzpicture } \useasboundingbox ( -1,-.8 ) rectangle ( 1,1 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( -.8,0 ) arc ( -180:180:.8 ) ; \draw ( -135:.8 ) -- ( 0,-.15 ) -- ( 0,.15 ) -- ( 135:.8 ) ( 0,.15 ) -- ( 60:.8 ) ( 0,-.15 ) -- ( -60:.8 ) ; \draw ( -30:.8 ) -- ( .4,0 ) -- ( 30:.8 ) ( .4,0 ) -- ( .8,0 ) ; \fill ( -135:.8 ) circle ( 1.5pt ) ( 135:.8 ) circle ( 1.5pt ) ( 60:.8 ) circle ( 1.5pt ) ( -60:.8 ) circle ( 1.5pt ) ( 30:.8 ) circle ( 1.5pt ) ( 0:.8 ) circle ( 1.5pt ) ( -30:.8 ) circle ( 1.5pt ) ( .4,0 ) circle ( 1.5pt ) ( 0,-.15 ) circle ( 1.5pt ) ( 0,.15 ) circle ( 1.5pt ) ; \end{tikzpicture}}={\begin{tikzpicture } \useasboundingbox ( -1,-.8 ) rectangle ( 1,1 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( -.8,0 ) arc ( -180:180:.8 ) ; \draw ( -135:.8 ) -- ( 0,-.15 ) -- ( 0,.15 ) -- ( 150:.8 ) ( 0,.15 ) -- ( 30:.8 ) ( 0,-.15 ) -- ( -45:.8 ) ; \draw ( 60:.8 ) -- ( 0,.45 ) -- ( 120:.8 ) ( 0,.45 ) -- ( 0,.8 ) ; \fill ( -135:.8 ) circle ( 1.5pt ) ( 150:.8 ) circle ( 1.5pt ) ( 30:.8 ) circle ( 1.5pt ) ( -45:.8 ) circle ( 1.5pt ) ( 60:.8 ) circle ( 1.5pt ) ( 90:.8 ) circle ( 1.5pt ) ( 120:.8 ) circle ( 1.5pt ) ( 0,.45 ) circle ( 1.5pt ) ( 0,-.15 ) circle ( 1.5pt ) ( 0,.15 ) circle ( 1.5pt ) ; \end{tikzpicture}}\ ] ] as shown in the first example that illustrates lemma [ lemcom ] , the independence of the choice of the insertion locus is a consequence of lemma [ lemcom ] , where @xmath700 is the disjoint union @xmath740 and @xmath700 intersects @xmath345 along @xmath741 .
this also proves that @xmath712 is a commutative algebra .
since the morphism from @xmath712 to @xmath717 induced by the identification of the two endpoints of @xmath27 amounts to quotient out @xmath712 by the relation that identifies two diagrams that are obtained from one another by moving the nearest univalent vertex to an endpoint of i near the other endpoint , a similar application of lemma [ lemcom ] also proves that this morphism is an isomorphism from @xmath712 to @xmath717 .
( in this application , @xmath124 comes from the inside boundary of the annulus . )
in this section , we finally describe the promised invariants , which generalize both the linking number and @xmath3 .
these invariants count configurations of jacobi diagrams with support some link , in an asymptotic rational homology @xmath4 . in subsection
[ subdefconfspace ] , we introduce the relevant configuration spaces . in subsection
[ subconfint ] , we define integrals over these spaces from propagating forms .
the wanted invariants are obtained by combining these integrals in subsection [ subconstcsi ] .
these integrals will be expressed in terms of algebraic intersections , which involve propagating chains , in subsection [ subrat ] .
important universality properties of the constructed invariants are presented in subsection [ subuniv ] .
let @xmath324 be an asymptotic rational homology @xmath4 .
let @xmath117 be a disjoint union of @xmath114 circles @xmath545 , @xmath742 and let @xmath743 denote a @xmath8 embedding from @xmath117 to @xmath187 .
let @xmath622 be a jacobi diagram with support @xmath117 .
let @xmath744 denote the set of univalent vertices of @xmath622 , and let @xmath745 denote the set of trivalent vertices of @xmath622 .
a _ configuration _ of @xmath622 is an embedding @xmath746 whose restriction @xmath747 to @xmath46 may be written as @xmath748 for some injection @xmath749 in the given isotopy class @xmath750 $ ] of embeddings of @xmath46 into the interior of @xmath117 .
denote the set of these configurations by @xmath751 , @xmath752 ,
c_{|u}=l \circ j\right\}.\ ] ] in @xmath751 , the univalent vertices move along @xmath753 while the trivalent vertices move in the ambient space , and @xmath751 is naturally an open submanifold of @xmath754 . an _ orientation _ of a set of cardinality at least @xmath1 is a total order of its elements up to an even permutation
. cut each edge of @xmath622 into two half - edges .
when an edge is oriented , define its _ first _ half - edge and its _ second _ one , so that following the orientation of the edge , the first half - edge is met first . recall that @xmath654 denotes the set of half - edges of @xmath622 .
when @xmath622 is equipped with a vertex - orientation , orientations of the manifold @xmath751 are in canonical one - to - one correspondence with orientations of the set @xmath654 . proof : since @xmath751 is naturally an open submanifold of @xmath754 , it inherits @xmath755-valued charts from @xmath44-valued orientation - preserving charts of @xmath117 and @xmath4-valued orientation - preserving charts of @xmath187 . in order to define the orientation of @xmath755
, one must identify its factors and order them ( up to even permutation ) .
each of the factors may be labeled by an element of @xmath654 : the @xmath44-valued local coordinate of an element of @xmath117 corresponding to the image under @xmath138 of an element of @xmath46 sits in the factor labeled by the half - edge of @xmath46 ; the @xmath0 cyclically ordered ( by the orientation of @xmath187 ) @xmath44-valued local coordinates of the image under a configuration @xmath216 of an element of @xmath756 live in the factors labeled by the three half - edges that are cyclically ordered by the vertex - orientation of @xmath622 , so that the cyclic orders match .
the dimension of @xmath751 is @xmath757 where @xmath758 denotes the set of edges of @xmath622 .
since @xmath759 , @xmath760 a _ numbered _ degree @xmath24 jacobi diagram is a degree @xmath24 jacobi diagram @xmath622 whose edges are oriented , equipped with an injection @xmath761 . such an injection numbers the edges
. note that this injection is a bijection when @xmath656 is empty .
let @xmath762 denote the set of numbered degree @xmath24 jacobi diagrams with support @xmath117 without _ looped edges _ like ( -.2,-.1 ) rectangle ( .6,.2 ) ; ( .5,0 ) arc ( 0:360:.15 ) ; ( -.1,0 ) ( .2,0 ) ; ( .2,0 ) circle ( 1.5pt ) ; .
let @xmath622 be a numbered degree @xmath24 jacobi diagram .
the orientations of the edges of @xmath622 induce the following orientation of the set @xmath654 of half - edges of @xmath622 : order @xmath655 arbitrarily , and order the half - edges as ( first half - edge of the first edge , second half - edge of the first edge , , second half - edge of the last edge ) .
the induced orientation is called the _ edge - orientation _ of @xmath654 .
note that it does not depend on the order of @xmath655 .
thus , as soon as @xmath622 is equipped with a vertex - orientation @xmath763 , the edge - orientation of @xmath622 orients @xmath751 .
an edge @xmath764 oriented from a vertex @xmath476 to a vertex @xmath481 of @xmath622 induces the following canonical map
@xmath765 for any @xmath766 , let @xmath767 be a propagating form of @xmath327 . define @xmath768 where @xmath769 denotes the manifold @xmath751 equipped with the orientation induced by the vertex - orientation @xmath763 and by the edge - orientation of @xmath622 .
the convergence of this integral is a consequence of the following proposition , which will be proved in subsection [ subcompconf ] .
[ propcompext ] there exists a smooth compactification @xmath770 of @xmath751 where the maps @xmath771 smoothly extend . according to this proposition
, @xmath772 smoothly extends to @xmath770 , and @xmath773 for any three propagating forms @xmath774 , @xmath775 and @xmath776 of @xmath327 , @xmath777 ( -.5,0 ) arc ( -180:180:.15 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( .5,0 ) arc ( 0:360:.15 ) ; \draw ( -.2,0 ) -- ( .2,0 ) ; \fill ( -.2,0 ) circle ( 1.5pt ) ( .2,0 ) circle ( 1.5pt ) ( .85,0 ) node{$s^1_j$ } ( -.85,0 ) node{$s^1_i$ } ; \end{tikzpicture}},(\omega(i))_{i \in \underline{3}})=lk(k_i , k_j)\ ] ] and @xmath778 for any numbering of the ( plain ) diagrams ( exercise ) .
let us now study the case of @xmath779 ( 0,-.2 ) arc ( -90:270:.2 ) ; \draw [ - > ] ( -.2,0 ) -- ( .2,0 ) ; \draw ( .5,0 ) node{$s^1_j$ } ( 0,.1 ) node{\tiny $ k$ } ; \fill ( -.2,0 ) circle ( 1.5pt ) ( .2,0 ) circle ( 1.5pt ) ; \end{tikzpicture}},(\omega(i))_{i \in \underline{3}})$ ] , which depends on the chosen propagating forms , and on the diagram numbering .
a _ dilation _ is a homothety with positive ratio
. let @xmath780 denote the fiber space over @xmath781 made of the tangent vectors to the knot @xmath781 of @xmath187 that orient @xmath781 , up to dilation .
the fiber of @xmath780 is made of one point , so that the total space of this _ unit positive tangent bundle to @xmath781 _ is @xmath781 .
let @xmath782 denote the fiber space over @xmath781 made of the opposite tangent vectors to @xmath781 , up to dilation . for a knot @xmath781 in @xmath187 , @xmath783 ( 0,.2 ) arc ( -270:90:.2 ) ;
\draw ( -.2,0 ) -- ( .2,0 ) ( .5,0 ) node{$s^1_j$ } ; \fill ( -.2,0 ) circle ( 1.5pt ) ( .2,0 ) circle ( 1.5pt ) ; \end{tikzpicture}})=\{(k_j(z),k_j(z\exp(i\theta ) ) ) ; ( z,\theta ) \in
s^1 \times ] 0,2\pi[\ } .\ ] ] let @xmath784 ( 0,.2 ) arc ( -270:90:.2 ) ; \draw ( -.2,0 ) -- ( .2,0 ) ( .5,0 ) node{$s^1_j$ } ; \fill ( -.2,0 ) circle ( 1.5pt ) ( .2,0 ) circle ( 1.5pt ) ; \end{tikzpicture}})$ ] be the closure of @xmath785 ( 0,.2 ) arc ( -270:90:.2 ) ; \draw ( -.2,0 ) -- ( .2,0 ) ( .5,0 ) node{$s^1_j$ } ; \fill ( -.2,0 ) circle ( 1.5pt ) ( .2,0 ) circle ( 1.5pt ) ; \end{tikzpicture}})$ ] in @xmath249 .
this closure is diffeomorphic to @xmath786 $ ] where @xmath787 is identified with @xmath780 , @xmath788 is identified with @xmath782 and @xmath789 ( 0,.2 ) arc ( -270:90:.2 ) ; \draw ( -.2,0 ) -- ( .2,0 ) ( .5,0 ) node{$s^1_j$ } ; \fill ( -.2,0 ) circle ( 1.5pt ) ( .2,0 ) circle ( 1.5pt ) ; \end{tikzpicture } } ) = u^+k_j - u^-k_j$ ] .
[ lemvaritheta ] for any @xmath790 , let @xmath767 and @xmath791 be propagating forms of @xmath327 , where @xmath792 is a one - form on @xmath249 .
@xmath793 ( 0,-.2 ) arc ( -90:270:.2 ) ; \draw [ - > ] ( -.2,0 ) -- ( .2,0 ) ; \draw ( .5,0 ) node{$s^1_j$ } ( 0,.1 ) node{\tiny $ k$ } ; \fill ( -.2,0 ) circle ( 1.5pt ) ( .2,0 ) circle ( 1.5pt ) ; \end{tikzpicture}},(\omega^{\prime}(i))_{i \in \underline{3}})-i({\begin{tikzpicture } \useasboundingbox ( -.3,-.1 ) rectangle ( .8,.2 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 0,-.2 ) arc ( -90:270:.2 ) ; \draw [ - > ] ( -.2,0 ) -- ( .2,0 ) ; \draw ( .5,0 ) node{$s^1_j$ } ( 0,.1 ) node{\tiny $ k$ } ; \fill ( -.2,0 ) circle ( 1.5pt ) ( .2,0 ) circle ( 1.5pt ) ; \end{tikzpicture}},(\omega(i))_{i \in \underline{3}})=\int_{u^+k_j}\eta(k ) -\int_{u^-k_j}\eta(k).\ ] ] proof : apply the stokes theorem to @xmath794 .
find a knot @xmath781 of @xmath4 and a form @xmath795 of @xmath796 such that the right - hand side of lemma [ lemvaritheta ] does not vanish .
( use lemma [ lemetactwo ] , hints can be found in subsection [ subsecstraight ] . )
say that a propagating form @xmath64 of @xmath327 is _ homogeneous _ if its restriction to @xmath255 is @xmath797 for the homogeneous volume form @xmath17 of @xmath15 of total volume @xmath13 .
[ lemdefitheta ] for any @xmath790 , let @xmath767 be a homogeneous propagating form of @xmath327
. then @xmath779 ( 0,-.2 ) arc ( -90:270:.2 ) ; \draw [ - > ] ( -.2,0 ) -- ( .2,0 ) ; \draw ( .5,0 ) node{$s^1_j$ } ( 0,.1 ) node{\tiny $ k$ } ; \fill ( -.2,0 ) circle ( 1.5pt ) ( .2,0 ) circle ( 1.5pt ) ; \end{tikzpicture}},(\omega(i))_{i \in \underline{3}})$ ] does not depend on the choices of the @xmath767 , it is denoted by @xmath798 .
proof : apply lemma [ lemetactwo ] with @xmath799 , so that @xmath800 in lemma [ lemvaritheta ] .
let @xmath652 .
let @xmath801 $ ] denote the class in @xmath667 of a numbered jacobi diagram @xmath622 of @xmath762 equipped with a vertex - orientation @xmath763 , then @xmath802 \in { { \cal a}}^t_n(c)$ ] is independent of the orientation of @xmath763 , it will be simply denoted by @xmath803 $ ] .
[ thmmain ] let @xmath324 be an asymptotic rational homology @xmath4 .
let @xmath804 be an embedding .
for any @xmath805 , let @xmath767 be a homogeneous propagating form of @xmath327
. set @xmath806 \in { { \cal a}}_n^t(\coprod_{j=1}^ks^1_j).\ ] ] then @xmath807 is independent of the chosen @xmath767 , it only depends on the diffeomorphism class of @xmath808 , on @xmath809 and on the @xmath798 , for the components @xmath781 of @xmath547 .
more precisely , set @xmath810 there exist two constants @xmath811 and @xmath812 such that @xmath813 only depends on the diffeomorphism class of @xmath808 . here
@xmath814 acts on @xmath815 , on the copy @xmath816 of @xmath5 as indicated by the subscript @xmath138 .
@xmath817 furthermore , if @xmath197 , then the projection @xmath818 of @xmath819 on @xmath820 is a _
universal _ finite type invariant of links in @xmath4 , i.e. @xmath821 satisfies the properties stated for @xmath822 in theorem [ thmbn ] .
it is the configuration space invariant studied by altschler , freidel @xcite , dylan thurston @xcite , sylvain poirier @xcite and others .
if @xmath823 , then @xmath824 is the kontsevich configuration space invariant @xmath825 , which is a universal invariant for @xmath115-spheres according to a theorem of kuperberg and thurston @xcite , and which was completed to a universal finite type invariant for @xmath116-spheres by delphine moussard @xcite .
the proof of this theorem is sketched in section [ secproofinv ] . under its assumptions ,
let @xmath826 be a homogeneous propagating form of @xmath327 , let @xmath512 be the involution of @xmath249 that permutes two elements in @xmath827 , set @xmath828 , and set @xmath829 for any @xmath94 .
let @xmath830 be the set of automorphisms of @xmath622 , which is the set of permutations of the half - edges that map a pair of half - edges of an edge to another such and a triple of half - edges that contain a vertex to another such , and that map half - edges of univalent vertices on a component @xmath781 to half - edges of univalent vertices on @xmath781 so that the cyclic order among such vertices is preserved .
set @xmath831 then @xmath832=\sum_{\gamma \,\mbox{\tiny unnumbered , unoriented } } \frac1{\sharp \mbox{aut}(\gamma)}i(\gamma,(\omega)_{i \in \underline{3n}})[\gamma]\ ] ] where the sum of the right - hand side runs over the degree @xmath24 jacobi diagrams on @xmath117 without looped edges . indeed , for a numbered graph @xmath622 , there are @xmath833 ways of renumbering it , and @xmath834 of them will produce the same numbered graph .
[ thmunivone ] let @xmath835 , @xmath836 .
recall @xmath837 .
set @xmath838 .
let @xmath187 be an asymptotically standard @xmath116-homology @xmath4 .
let @xmath547 be a link in @xmath187 .
let @xmath839 be a collection of pairwise disjoint balls in @xmath187 such that every @xmath840 intersects @xmath547 as a ball of a crossing change that contains a positive crossing @xmath841 , and let @xmath842 be the link obtained by changing the positive crossings @xmath841 to negative crossings .
let @xmath843 be a collection of pairwise disjoint rational homology handlebodies in @xmath844 .
let @xmath845 be rational lp surgeries in @xmath187 .
set @xmath846 $ ] and @xmath847 if @xmath848 , then @xmath849 vanishes . sketch of proof : as in @xcite , one can use ( generalized ) propagators for the @xmath850 that coincide for different @xmath27 wherever it makes sense ( for example , for configurations that do not involve points in surgered pieces @xmath851 ) .
see also @xcite .
then contributions to the alternate sum of the integrals over parts that do not involve at least one point in an @xmath851 or in an @xmath852 , for all @xmath354 cancel .
assume that every crossing change is performed by moving only one strand .
again , contributions to the alternate sum of the integrals that do not involve at least one point on a moving strand cancel .
furthermore , if the moving strand of @xmath841 is moved very slightly , and if no other vertex is constrained to lie on the other strand in the ball of the crossing change , then the alternate sum is close to zero .
thus in order to produce a contribution to the alternate sum , a graph must have at least @xmath853 vertices .
see @xcite or ( * ? ? ?
* section 5.4 ) , and ( * ? ? ?
* section 3 ) for more details .
this implies that @xmath821 is of degree at most @xmath24 for links in @xmath4 , and that @xmath854 is of degree at most @xmath613 for @xmath115-spheres or @xmath116-spheres .
now , under the hypotheses of theorem [ thmunivone ] , assume that @xmath851 is the standard genus @xmath0 handlebody with three handles with meridians @xmath855 and longitudes @xmath856 such that @xmath857 .
see @xmath851 as a thickening of the trivalent graph below .
( -2.5,-2.5 ) rectangle ( 2.5,2.2 ) ; ( 30:.4 )
( 0,0 ) ( 0,-.4 ) ( 150:.4 )
( 0,0 ) ( 30:1.1 ) circle ( .7 ) ( 150:1.1 ) circle ( .7 ) ( -90:1.1 ) circle ( .7 ) ; ( 30:.6 ) arc ( -150:210:.5 ) ; ( 150:.6 ) arc ( -30:330:.5 ) ; ( -90:.6 ) arc ( -270:90:.5 ) ; ( 30:.8 ) node@xmath858 ( 150:.8 ) node@xmath859 ( -90:.8 ) node@xmath860 ( 30:2.2 ) node@xmath861 ( 149:2.25 ) node@xmath862 ( -90:2.2 ) node@xmath863 ; ( 30:1.6 ) .. controls ( 29:1.6 ) and ( 26:1.7 ) .. ( 26:1.8 ) .. controls ( 26:1.9 ) and ( 29:2 ) .. ( 30:2 ) ; ( 26:1.8 ) .. controls ( 26:1.9 ) and ( 29:2 ) .. ( 30:2 ) ; ( 30:2 ) .. controls ( 31:2 ) and ( 34:1.95 ) .. ( 34:1.9 ) ; ( 30:1.6 ) .. controls ( 31:1.6 ) and ( 34:1.65 ) .. ( 34:1.7 ) ; ( 150:1.6 ) .. controls ( 149:1.6 ) and ( 146:1.7 ) .. ( 146:1.8 ) .. controls ( 146:1.9 ) and ( 149:2 ) .. ( 150:2 ) ; ( 146:1.8 ) .. controls ( 146:1.9 ) and ( 149:2 ) .. ( 150:2 ) ; ( 150:2 ) .. controls ( 151:2 ) and ( 154:1.95 ) .. ( 154:1.9 ) ; ( 150:1.6 ) .. controls ( 151:1.6 ) and ( 154:1.65 ) .. ( 154:1.7 ) ; ( -90:1.6 ) .. controls ( -91:1.6 ) and ( -94:1.7 ) .. ( -94:1.8 ) .. controls ( -94:1.9 ) and ( -91:2 ) .. ( -90:2 ) ; ( -94:1.8 ) .. controls ( -94:1.9 ) and ( -91:2 ) .. ( -90:2 ) ; ( -90:2 ) .. controls ( -89:2 ) and ( -86:1.95 ) .. ( -86:1.9 ) ; ( -90:1.6 ) .. controls ( -89:1.6 ) and ( -86:1.65 ) .. ( -86:1.7 ) ; also assume that @xmath852 is an integer homology handlebody . in @xmath864
, there is a surface @xmath865 such that @xmath866 .
assume that @xmath867 .
( for example , choose @xmath852 such that @xmath868 , like in the case of the matveev borromean surgery of @xcite . )
assume that the @xmath856 bound surfaces @xmath869 in @xmath187 .
assume that the collection of surfaces @xmath870 reads @xmath871 so that + for any @xmath872 , for @xmath873 , if @xmath874 , the interior of @xmath875 intersects @xmath876 only in @xmath877 .
note that @xmath878 .
note that these assumptions are realised in the following case .
start with an embedding of a jacobi diagram @xmath622 whose univalent vertices belong to chords ( plain edges between two univalent vertices ) on @xmath879 in @xmath187 .
assume that the trivalent vertices of @xmath622 are labeled in @xmath880 , and assume that its chords are labeled in @xmath881 .
apply the following operations + replace edges ( -1.2,-.2 ) rectangle ( .3,.4 ) ; ( -1,.1 )
( -.9,0 )
( -1,-.1 ) ( -.9,0 )
( .2,.1 ) ( .1,0 ) ( .2,-.1 ) ; ( -.9,0 ) circle ( 1.5pt ) ( .1,0 ) circle ( 1.5pt ) ; without univalent vertices by ( -1.2,-.2 ) rectangle ( .3,.4 ) ; ( -.3,0 ) arc ( 0:90:.2 ) ; ( -.1,0 ) arc ( 0:90:.2 ) ; ( -.1,0 ) arc ( 0:-200:.2 ) ( -.5,.2 ) arc ( 90:270:.2 ) ( -.3,0 ) arc ( 0:-20:.2 ) ; ( -1,.1 )
( -.9,0 )
( -1,-.1 ) ( -.9,0 )
( -.7,0 ) ( -.1,0 )
( .2,.1 ) ( .1,0 ) ( .2,-.1 ) ; ( -.9,0 ) circle ( 1.5pt ) ( .1,0 ) circle ( 1.5pt ) ; , + replace a chord ( -.7,-.2 ) rectangle ( .2,.4 ) ; ( 0,-.3 ) ( 0,.3 ) ; ( -.5,.3 ) ( -.5,-.3 ) ; ( -.5,0 ) ( 0,0 ) ; ( -.5,0 ) circle ( 1.5pt ) ( 0,0 ) circle ( 1.5pt ) ; labeled by @xmath361 by a crossing change @xmath841 @xmath882 ( 0,-.3 ) -- ( 0,.3 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( -.4,.3 ) -- ( -.4,-.3 ) ; \end{tikzpicture}}\rightarrow { \begin{tikzpicture } \useasboundingbox ( -.7,-.2 ) rectangle ( .4,.4 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( 0,0 ) -- ( 0,.3 ) ; \draw ( 0,-.3 ) -- ( 0,-.2 ) ; \draw [ ->,dash pattern = on 2pt off 2pt ] ( .1,.1 ) .. controls ( .15,.1 ) and ( .25,.05 ) .. ( .25,0 ) .. controls ( .25,-.05 ) and ( .1,-.1 ) .. ( 0,-.1 ) .. controls ( -.1,-.1 ) and ( -.5,-.2 ) .. ( -.5,-.3 ) ; \draw ( 0,-.3 ) -- ( 0,-.2 ) ; \draw [ dash pattern = on 2pt off 2pt ] ( -.5,.3 ) .. controls ( -.5,.2 ) and ( -.2,.1 ) .. ( -.1,.1 ) ; \end{tikzpicture}}$ ] in a ball @xmath840 that is a neighborhood of the plain edge .
+ thicken the trivalent graph ( -.4,-.2 ) rectangle ( .4,.2 ) ; ( 150:.15 )
( 0,0 ) ( -90:.15 ) ( 0,0 )
( 30:.15 ) ( 30:.25 ) circle ( .1 ) ( 150:.25 ) circle ( .1 ) ( -90:.25 ) circle ( .1 ) ; ( 0,0 ) circle ( 1pt ) ; associated to the trivalent vertex labeled by @xmath354 , and call it @xmath851 . then the surfaces @xmath869 are the disks bounded by the small loops of ( -.4,-.2 ) rectangle ( .4,.2 ) ; ( 150:.15 ) ( 0,0 ) ( -90:.15 ) ( 0,0 ) ( 30:.15 ) ( 30:.25 ) circle ( .1 ) ( 150:.25 ) circle ( .1 ) ( -90:.25 ) circle ( .1 ) ; ( 0,0 ) circle ( 1pt ) ; .
conversely , under the assumptions before the example , define the following vertex - oriented jacobi diagram @xmath883)\ ] ] on @xmath879 , with * two univalent vertices joined by a chord for each crossing change ball @xmath840 at the corresponding places on @xmath879 ( in @xmath884 ) , * one trivalent vertex for each @xmath851 , where the three adjacent half - edges of the vertex correspond to the three @xmath869 , with the fixed cyclic order , such that any pair of half - edges corresponding to some @xmath885 and its friend @xmath886 forms an edge between two trivalent vertices .
[ thmkeyuniv ] under the assumptions above , let @xmath846 $ ] .
when @xmath887 , @xmath888 \;\mbox{mod 1t}\;\left(\mbox{or in}\ ; \frac{{{\cal a}}_n^t(\coprod_{j=1}^ks^1_j)}{(1t)}\right).\ ] ] sketch of proof : when @xmath889 , the proof of theorem [ thmunivone ] can be pushed further in order to prove the result like in @xcite or ( * ? ? ?
* section 5.4 ) . in general , when @xmath890 , it is a consequence of the main theorem in @xcite ( theorem 2.4 ) .
the general result can be obtained by mixing the arguments of ( * ? ? ?
* section 3 ) with the arguments of the link case .
this theorem is the key to proving the universality of @xmath891 among vassiliev invariants for links in @xmath4 and to proving the universality of @xmath892 among finite type invariants of @xmath115-spheres .
this universality implies that all finite type invariants factor through @xmath892 .
theorem [ thmkeyuniv ] with @xmath893 instead of @xmath892 is proved in @xcite , when @xmath890 , when the @xmath845 are matveev s borromean surgeries and when the @xmath869 are disks such that @xmath894 .
then the main theorem of @xcite implies theorem [ thmkeyuniv ] with @xmath893 instead of @xmath892 , when @xmath890 and when the @xmath851 and the @xmath852 are integral homology handlebodies .
in this section , we state theorem [ thmmainstraight ] .
this is another version of theorem [ thmmain ] , which leads to a definition of @xmath892 involving algebraic intersections rather than integrals in subsection [ subrat ] .
it is based on the concept of _ straight links _ introduced in subsection [ subsecstraight ] .
this section also contains sketches of proofs of theorems [ thmmain ] and [ thmmainstraight ] .
we begin with the introduction of appropriate compactifications of configuration spaces to justify the convergence of our integrals stated in proposition [ propcompext ] .
let @xmath895 be a finite set .
see the elements of @xmath896 as maps @xmath897 . for a non - empty @xmath898 ,
let @xmath899 be the set of maps that map @xmath27 to @xmath184 .
for @xmath898 such that @xmath900 , let @xmath901 be the set of maps that map @xmath27 to a single element of @xmath25 .
when @xmath27 is a finite set , and when @xmath39 is a vector space of positive dimension , @xmath902 denotes the space of injective maps from @xmath27 to @xmath39 up to translation and dilation . when @xmath900 , @xmath902 embeds in the compact space @xmath903 of non - constant maps from @xmath27 to @xmath39 up to translation and dilation .
the fiber of the unit normal bundle to @xmath901 in @xmath896 over a configuration @xmath478 is @xmath904 .
proof : exercise .
let @xmath905 denote the space of injective maps from @xmath895 to @xmath187 .
define a compactification @xmath906 of @xmath905 by generalizing the previous construction of @xmath907 as follows .
start with @xmath896 .
blow up @xmath908 , which is the point @xmath909 such that @xmath910 . +
then for @xmath911 , in this decreasing order , successively blow up the ( closures of the preimages under the composition of the previous blow - down maps of the ) @xmath912 such that @xmath913 ( choosing an arbitrary order among them ) and , next , the ( closures of the preimages under the composition of the previous blow - down maps of the ) @xmath914 such that @xmath915 ( again , choosing an arbitrary order among them ) . [ lemcompconf ]
the successive manifolds that are blown - up in the above process are smooth and transverse to the boundaries .
the manifold @xmath906 is a smooth compact @xmath916-manifold independent of the possible order choices in the process . for @xmath917 , @xmath918 ,
the map @xmath919 smoothly extends to @xmath920 .
sketch of proof : a configuration @xmath921 of @xmath896 induces the following partition @xmath922 of @xmath923 pick disjoint neighborhoods @xmath924 in @xmath25 of the points @xmath47 of @xmath925 that are furthermore in @xmath187 for @xmath47 in @xmath187 and that are identified with balls of @xmath4 by @xmath8-charts .
consider the neighborhood @xmath926 of @xmath921 in @xmath896 .
the first blow - ups that transformed this neighborhood are * the blow - up of @xmath927 if @xmath928 , which changed ( a smaller neighborhood of @xmath929 in ) @xmath930 to @xmath931 , * and the blow - ups of the @xmath932 , for the @xmath933 such that @xmath934 , which changed ( a smaller neighborhood of @xmath935 in ) @xmath936 to @xmath937 , where @xmath938 and @xmath939 fibers over @xmath940 , and the fiber over @xmath941 is @xmath942 .
when considering how the next blow - ups affect the preimage of a neighborhood of @xmath921 , we can restrict to our new factors .
first consider a factor @xmath937 .
picking @xmath943 and fixing a riemannian structure on @xmath944 identifies @xmath942 with the space of maps @xmath945 such that @xmath946 and @xmath947 .
then @xmath948 is identified with @xmath949 in @xmath950 ( where @xmath924 is identified with an open subset of @xmath4 ) , for @xmath304 .
now , @xmath937 must be blown - up along its intersections with the preimage closures of the @xmath901 such that @xmath900 , @xmath951 and @xmath27 is maximal .
these intersections respect the product structure by @xmath952 and the fibration over @xmath940 so that we only need to understand the blow - ups of the intersections of the @xmath901 with a fiber of @xmath939 .
these are nothing but configurations in a ball of @xmath4 , and we can iterate our process .
now consider the possible factor @xmath931 and blow up its intersections with the preimage closures of the @xmath914 for @xmath953 maximal and with the preimage closures of the @xmath901 with @xmath954 in an order compatible with the algorithm . here , @xmath955 is the unit sphere of @xmath956 .
a point @xmath957 is in the preimage closure of @xmath914 under the previous blow - up if @xmath958 .
in particular , the @xmath914 and the @xmath901 again read as products by @xmath959 , and we study what happens near a given @xmath122 of @xmath955 .
for such a @xmath122 , we proceed as before if @xmath960 .
otherwise the factor of @xmath961 must be treated differently , namely by blowing up @xmath962 in @xmath955 . then iterate .
this produces a compact manifold @xmath963 with boundary and ridges , which is finally independent of the order of the blow - ups ( when this order is compatible with the algorithm ) , since it is locally independent .
the interior of @xmath963 is @xmath964 . since the blow - ups separate all the pairs of points at some scale , @xmath771 naturally extends there .
the introduced local coordinates show that the extension is smooth .
see ( * ? ? ?
* section 3 ) for more details .
[ lemcompconfl ] the closure of @xmath751 in @xmath965 is a smooth compact submanifold of @xmath965 , which is denoted by @xmath966 .
proof : exercise .
proposition [ propcompext ] is a consequence of lemmas [ lemcompconf ] and [ lemcompconfl ] .
a one - chain @xmath216 of @xmath15 is _ algebraically trivial _ if for any two points @xmath47 and @xmath51 outside its support , the algebraic intersection of an arc from @xmath47 to @xmath51 transverse to @xmath216 with @xmath216 is zero , or equivalently if the integral of any one form of @xmath15 along @xmath216 is zero .
let @xmath324 be an asymptotic rational homology @xmath4 .
say that @xmath781 is _ straight _ with respect to @xmath179 if the curve @xmath967 of @xmath15 is algebraically trivial ( recall the notation from proposition [ propprojbord ] and subsection [ subconfint ] ) .
a link is _ straight _ with respect to @xmath179 if all its components are .
if @xmath781 is straight , then @xmath968 ( 0,.2 ) arc ( -270:90:.2 ) ; \draw ( -.2,0 ) -- ( .2,0 ) ( .5,0 ) node{$s^1_j$ } ; \fill ( -.2,0 ) circle ( 1.5pt ) ( .2,0 ) circle ( 1.5pt ) ; \end{tikzpicture}}))$ ] is algebraically trivial .
recall @xmath969 ( 0,.2 ) arc ( -270:90:.2 ) ; \draw ( -.2,0 ) -- ( .2,0 ) ( .5,0 ) node{$s^1_j$ } ; \fill ( -.2,0 ) circle ( 1.5pt ) ( .2,0 ) circle ( 1.5pt ) ; \end{tikzpicture}})$ ] , @xmath970 .
if @xmath971 is algebraically trivial , then for any propagating chain @xmath328 of @xmath327 transverse to @xmath718 and for any propagating form @xmath331 of @xmath327 , @xmath972 where @xmath798 is defined in lemma [ lemdefitheta ] . in particular , @xmath973 and @xmath974 when @xmath25 is an integer homology @xmath0sphere .
proof : exercise .
recall lemmas [ lemetactwo ] and [ lemvaritheta ] .
let @xmath187 be an asymptotically standard @xmath116-homology @xmath4 .
for any parallel @xmath975 of a knot @xmath11 in @xmath187 , there exists an asymptotically standard parallelization @xmath976 homotopic to @xmath179 , such that @xmath11 is straight with respect to @xmath976 , and @xmath977 or @xmath978 . for any embedding @xmath979 that is straight with respect to @xmath179 , @xmath980 is the linking number of @xmath11 and a parallel of @xmath11 .
sketch of proof : for any knot embedding @xmath11 , there is an asymptotically standard parallelization @xmath976 homotopic to @xmath179 such that @xmath981 is one point .
thus @xmath11 is straight with respect to @xmath982 .
then @xmath976 induces a parallelization of @xmath11 , and @xmath983 is the linking number of @xmath11 with the parallel induced by @xmath976 .
( exercise ) . in general , for two homotopic
asymptotically standard parallelizations @xmath179 and @xmath976 such that @xmath11 is straight with respect to @xmath179 and @xmath976 , @xmath984 is an even integer ( exercise ) so that @xmath980 is always the linking number of @xmath11 with a parallel of @xmath11 . in @xmath4 equipped with @xmath313 , any link is represented by an embedding @xmath547 that sits in a horizontal plane except when it crosses under , so that the non - horizontal arcs crossing under are in vertical planes .
then the non - horizontal arcs have an algebraically trivial contribution to @xmath967 , while the horizontal contribution can be changed by adding kinks @xmath985 ( .4,0 ) to ( 0,.25 ) ; \draw [ out=90,in=90,draw = white , double = black , very thick ] ( 0,-.25 ) to ( .4,0 ) ; \end{tikzpicture}}$ ] or @xmath986 ( 0,-.2 ) to ( .4,0 ) ; \draw [ out=-90,in=-90,draw = white , double = black , very thick ] ( .4,0 ) to ( 0,.2 ) ; \draw [ out=-90,in=-90 ] ( .4,0 ) to ( 0,.2 ) ; \end{tikzpicture}}$ ] so that @xmath547 is straight with respect to @xmath313 . in this case
@xmath987 is the _ writhe _ of @xmath781 , which is the number of positive self - crossings of @xmath781 minus the number of negative self - crossings of @xmath781 . in particular , up to isotopy of @xmath547 , @xmath987 can be assumed to be @xmath988 ( exercise ) .
similarly , for any number @xmath512 that is congruent mod @xmath989 to @xmath980 there exists an embedding @xmath990 isotopic to @xmath11 and straight such that @xmath991 ( exercise ) .
let us state another version of theorem [ thmmain ] using straight links instead of homogeneous propagating forms .
recall @xmath992 [ thmmainstraight ] let @xmath324 be an asymptotic rational homology @xmath4 .
let @xmath804 be a straight embedding with respect to @xmath179 .
for any @xmath805 , let @xmath767 be a propagating form of @xmath327 .
set @xmath993 \in { { \cal a}}_n^t(\coprod_{j=1}^ks^1_j).\ ] ] then @xmath994 is independent of the chosen @xmath767 .
in particular , with the notation of theorem [ thmmain ] , @xmath995 this version of theorem [ thmmain ] allows us to replace the configuration space integrals by algebraic intersections in configuration spaces , and thus to prove the rationality of @xmath892 for straight links as follows . for any @xmath766 ,
let @xmath996 be a propagating chain of @xmath327 .
say that a family @xmath997 is _ in general @xmath998 position _ with respect to @xmath547 if for any @xmath999 , the @xmath1000 are pairwise transverse chains in @xmath770 . in this case , define @xmath1001 as the algebraic intersection in @xmath1002 of the codimension @xmath1 rational chains @xmath1000 . if the @xmath767 are propagating forms of @xmath327 poincar dual to the @xmath996 and supported in sufficiently small neighborhoods of the @xmath996 , then @xmath1003 for any @xmath999 , and @xmath1004 is rational , in this case .
the constants @xmath1005 and @xmath1006 of theorem [ thmmain ] are called _ anomalies . _
the anomaly @xmath1007 is the opposite of the constant @xmath1008 defined in ( * ? ? ?
* section 1.6 ) , @xmath1009 for any integer @xmath24 , and @xmath1010 $ ] according to ( * ? ? ?
* proposition 2.45 ) .
the computation of @xmath1011 can also be deduced from corollary [ corthetap ] .
we define @xmath123 below .
let @xmath1012 .
let @xmath1013 denote the linear map @xmath1014 let @xmath622 be a numbered jacobi diagram on @xmath44 .
define @xmath1015 like in subsection [ subdefconfspace ] where the line @xmath1013 of @xmath4 replaces the link @xmath547 of @xmath187 .
let @xmath1016 be the quotient of @xmath1015 by the translations parallel to @xmath1013 and by the dilations .
then the map @xmath1017 associated to an edge @xmath764 of @xmath622 maps a configuration to the direction of the vector from its origin to its end in @xmath15 .
it factors through @xmath1016 , which has two dimensions less .
now , define @xmath1018 as the total space of the fibration over @xmath15 whose fiber over @xmath419 is @xmath1016 .
the configuration space @xmath1018 carries a natural smooth structure , it can be compactified as before , and it can be oriented as follows , when a vertex - orientation @xmath763 is given .
orient @xmath1015 as before , orient @xmath1016 so that @xmath1015 is locally homeomorphic to the oriented product ( translation vector @xmath1019 in @xmath1020 , ratio of homothety @xmath10210,\infty[$ ] ) @xmath1022 and orient @xmath1018 with the @xmath1023 convention .
( this can be summarized by saying that the @xmath15-coordinates replace @xmath1024 . )
[ propanom ] for @xmath766 , let @xmath1025 be a two - form of @xmath15 such that @xmath1026 .
define @xmath1027 let @xmath1028 denote the set of connected numbered diagrams on @xmath44 with at least one univalent vertex , without looped edges .
set @xmath1029 \;\ ; \in { \cal a}({\mathbb{r}}).\ ] ] then @xmath727 does not depend on the chosen @xmath1025 , @xmath1030 ( 0,-.3 ) -- ( 0,.4 ) ; \draw ( 0,.15 ) arc ( 90:270:.15 ) ; \fill ( 0,-.15 ) circle ( 1pt ) ( 0,.15 ) circle ( 1pt ) ; \end{tikzpicture}}\right]$ ] and @xmath1031 for all @xmath1032 .
the series @xmath1033 is called the _ bott and taubes anomaly_. proof : the independence of the choices of the @xmath1025 will be a consequence of lemma [ leminvone ] below .
let us prove that @xmath1031 for all @xmath1032 .
let @xmath622 be a numbered graph and let @xmath1034 be obtained from @xmath622 by reversing the orientations of the @xmath1035 edges of @xmath622 . consider the map @xmath1036 from @xmath1037 to @xmath1038 that composes a configuration by the multiplication by @xmath75 in @xmath4 .
it sends a configuration over @xmath1012 to a configuration over @xmath1039 , and it is therefore a fibered map over the orientation - reversing antipode of @xmath15 .
equip @xmath622 and @xmath1034 with the same vertex - orientation
. then our map @xmath1036 is orientation - preserving if and only if @xmath1040 is even .
furthermore for all the edges @xmath764 of @xmath1034 , @xmath1041 , then since @xmath1042 , @xmath1043 it is known that @xmath1044 and @xmath1045 @xcite . furthermore , according to @xcite , @xmath1046 is a combination of diagrams with two univalent vertices , and @xmath1047 is obtained from the kontsevich integral by inserting @xmath122 times the plain part of @xmath1048 on each degree @xmath122 connected component of a diagram .
the variation of @xmath1049 when some @xmath1050 is changed to @xmath1051 for a one - form @xmath384 on @xmath249 reads @xmath1052 according to the stokes theorem , it reads @xmath1053 where the integral along @xmath1054 is actually the integral along the codimension one faces of @xmath770 , which are considered as open .
such a codimension one face only involves one blow - up . for any non - empty subset @xmath104 of @xmath1055 , the codimension one face associated to the blow - up of @xmath1056 in @xmath1057
is denoted by @xmath1058 , it lies in the preimage of @xmath1059 , in @xmath770 .
the other codimension one faces are associated to the blow - ups of the @xmath1060 in @xmath1057 , for subsets @xmath104 of @xmath1055 of cardinality at least @xmath1 . the face of @xmath770 associated to @xmath1060 is denoted by @xmath1061 .
let @xmath1062 .
assume that @xmath1063 if @xmath1064 .
the image of @xmath1061 in @xmath1057 is in the set of maps @xmath478 of @xmath1060 that define an injection from @xmath1065 to @xmath187 , which factors through an injection isotopic to the restriction of @xmath662 on @xmath1066 .
this set of maps @xmath1067 is a submanifold of @xmath1068 .
thus , @xmath1061 is a bundle over @xmath1069 .
when @xmath104 has no univalent vertices , the fiber over a map @xmath478 is the space @xmath1070 of injective maps from @xmath104 to @xmath1071 up to translations and dilations .
when @xmath104 contains univalent vertices of a component @xmath781 , the fiber over @xmath478 is the submanifold @xmath1072 of @xmath1073 , made of the configurations that map the univalent vertices of @xmath104 to a line of @xmath1074 directed by @xmath780 at @xmath1075 , in an order prescribed by @xmath622 .
if @xmath104 does not contain all the univalent vertices of @xmath622 on @xmath816 , this order is unique .
otherwise , @xmath1061 has @xmath1076 connected components corresponding to the total orders that induce the cyclic order of @xmath1077 .
when @xmath104 is a subset of the set of vertices @xmath1055 of a numbered graph @xmath622 , @xmath1078 denotes the set of edges of @xmath622 between two elements of @xmath104 ( edges of @xmath622 are plain ) , and @xmath1079 is the subgraph of @xmath622 made of the vertices of @xmath104 and the edges of @xmath1078 .
[ leminvone ] let @xmath324 be an asymptotic rational homology @xmath4 .
let @xmath1080 . for @xmath766 ,
let @xmath767 be a closed @xmath1-form on @xmath714 \times c_2(m)$ ] whose restriction to @xmath1081 is denoted by @xmath1082 , for any @xmath487 $ ] .
assume that for @xmath487 $ ] , @xmath1082 restricts to @xmath1083 as @xmath1084 , for some two - form @xmath1085 of @xmath15 such that @xmath1086 .
set @xmath1087 \in { { \cal a}}_n^t(\coprod_{j=1}^ks^1_j).\ ] ] then @xmath1088 where @xmath1089\times f(\gamma , b)}\bigwedge_{e \in e(\gamma)}p_e^{\ast}(\omega(j_e(e)))[\gamma].\ ] ] under the assumptions of theorem [ thmmain ] ( where the @xmath767 are homogeneous ) or theorem [ thmmainstraight ] ( where @xmath547 is straight with respect to @xmath179 ) , when @xmath1090 is fixed , @xmath807 is independent of the chosen @xmath767 . in particular , when @xmath823 , @xmath1091 coincides with the kontsevich configuration space integral invariant described in @xcite .
furthermore , the @xmath727 of proposition [ propanom ] are also independent of the forms @xmath1025 .
sketch of proof : according to the stokes theorem , for any @xmath999 , @xmath1092\times f } \bigwedge_{e \in e(\gamma)}p_e^{\ast}(\omega(j_e(e)))\ ] ] where the sum runs over the codimension one faces @xmath1093 of @xmath770 .
below , we sketch the proof that the only contributing faces are the faces @xmath1061 such that @xmath1094 and @xmath1079 is a connected component of @xmath622 , or equivalently , that the other faces do not contribute .
like in ( * ? ?
* lemma 2.17 ) faces @xmath1058 do not contribute .
when the product of all the @xmath771 factors through a quotient of @xmath714\times f(\gamma , b)$ ] of smaller dimension , the face @xmath1061 does not contribute .
this allows us to get rid of * the faces @xmath1061 such that @xmath104 is not a pair of univalent vertices of @xmath622 , and @xmath1079 is not connected ( see ( * ? ? ?
* lemma 2.18 ) ) , * the faces @xmath1061 such that @xmath1095 where @xmath1079 has a univalent vertex that was trivalent in @xmath622 ( see ( * ? ? ?
* lemma 2.19 ) ) .
we also have faces that cancel each other , for graphs that are identical outside their @xmath1079 part . *
the faces @xmath1061 ( that are not already listed ) such that @xmath1079 has at least a bivalent vertex cancel ( mostly by pairs ) by the parallelogram identification ( see ( * ? ? ?
* lemma 2.20 ) ) .
* the faces @xmath1061 where @xmath1079 is an edge between two trivalent vertices cancel by triples , thanks to the jacobi ( or ihx ) relation ( see ( * ? ? ?
* lemma 2.21 ) ) . * similarly , two faces where @xmath104 is made of two ( necessarily consecutive in @xmath117 ) univalent vertices of @xmath622 cancel @xmath1096 faces @xmath1097 where @xmath1098 is an edge between a univalent vertex of @xmath622 and a trivalent vertex of @xmath622 , thanks to the stu relation .
thus , we are left with the faces @xmath1061 such that @xmath1079 is a ( plain ) connected component of @xmath622 , and we get the wanted formula for @xmath1099 .
in the anomaly case , the same analysis of faces leaves no contributing faces , so that the @xmath727 are independent of the forms @xmath1025 in proposition [ propanom ] . back to the behaviour of @xmath815 under the assumptions of theorem
[ thmmain ] or theorem [ thmmainstraight ] , assume that @xmath1090 is fixed and apply the formula of the lemma to compute the variation of @xmath807 when some propagating chain @xmath1100 of @xmath327 is changed to some other propagating chain @xmath1101 .
according to lemma [ lemetactwo ] , under our assumptions , @xmath384 can be chosen so that @xmath1102 on @xmath255 and @xmath1103 if @xmath1100 and @xmath1104 are homogeneous .
define @xmath1105 on @xmath714\times c_2(m)$ ] ( @xmath1106 $ ] ) , and extend the other @xmath1107 trivially .
then @xmath1099 vanishes if @xmath1100 and @xmath1104 are homogeneous , as all the involved @xmath1108 do , so that @xmath807 is independent from the chosen homogeneous propagating forms @xmath767 of @xmath1109 in theorem [ thmmain ] .
now , assume that @xmath547 is straight . when @xmath1110 , the integrand of @xmath1108 factors through the natural projection of @xmath714\times f(\gamma , b)$ ] onto @xmath1061 , and @xmath1111 , consequently .
assume @xmath1112 , then @xmath1113\times f(\gamma , b)}p_{e_i}^{\ast}(d(t \eta ) ) \wedge \bigwedge_{e \in e(\gamma)\setminus e_i}p_e^{\ast}(\omega(j_e(e)).\ ] ] the form @xmath1114 pulls back through @xmath714\times f(\gamma_b , b)$ ] , and through @xmath1115 when @xmath1116 , so that , for dimension reasons , @xmath1108 vanishes unless @xmath1117 . therefore , we assume @xmath1117 .
when @xmath104 contains no univalent vertices , @xmath1108 factors through @xmath1118 \times \cup_{m(b)\in \check{m}}\check{s}_b(t_{m(b)}m ) } p_{e_i}^{\ast}(d(t \eta ) ) \wedge \bigwedge_{e
\in e(\gamma_b)\setminus e_i}p_e^{\ast}(\omega(j_e(e))).\ ] ] here the parallelization @xmath179 identifies the bundle @xmath1119 with @xmath1120 , and the integrand factors through the projection of @xmath714\times \check{m } \times \check{s}_b({\mathbb{r}}^3)$ ] onto @xmath1121\times \check{s}_b({\mathbb{r}}^3)$ ] whose dimension is smaller ( by @xmath0 ) .
in particular , @xmath1111 in this case , the independence of the choice of the @xmath767 is proved when @xmath823 ( when the link is empty ) , and @xmath1091 coincides with the kontsevich configuration space integral invariant described in @xcite .
let us now study the sum of the @xmath1108 , where @xmath1122 is a fixed labeled graph and @xmath1079 is a fixed numbered connected diagram with at least one univalent vertex on @xmath816 .
this sum factors through @xmath1123 \times \cup_{m(b)\in k_j}\check{s}_b(t_{m(b)}m,\gamma)}p_{e_i}^{\ast}(d(t \eta ) ) \wedge \bigwedge_{e \in e(\gamma_b ) \setminus e_i}p_e^{\ast}(\omega(j_e(e)))$ ] . at a collapse ,
the univalent vertices of @xmath1079 are equipped with a linear order , which makes @xmath1079 a numbered graph @xmath1124 on @xmath44 .
the corresponding connected component of @xmath714 \times \cup_{m(b)\in k_j}\check{s}_b(t_{m(b)}m,\gamma)$ ] reads @xmath714 \times \cup_{x \in u^+k_j } \check{q}(p_{\tau}(x);\tilde{\gamma}_b)$ ] ( @xmath1125 was defined in subsection [ subanom ] ) .
this allows us to see the contribution of such a connected component as the integral of a one - form ( defined by partial integrations ) over @xmath967 .
such an integral is zero when @xmath781 is straight .
now , theorem [ thmmainstraight ] is a corollary of theorem [ thmmain ] ( which is not yet completely proved ) .
recall that @xmath667 splits according to the number of connected components without univalent vertices of the graphs .
then it is easy to observe that @xmath1126 where @xmath891 is obtained from @xmath892 by sending the graphs with components that have no univalent vertices to @xmath351 , and @xmath1127 . according to ( * ? ? ?
* theorem 1.9 ) , @xmath1128 is a topological invariant of @xmath25 .
here , we will now focus on @xmath1129 , and define it with a given homogeneous propagating form , @xmath1130 for all i , so that @xmath1129 is an invariant of the diffeomorphism class of @xmath1131 .
we study its variation under a continuous deformation of @xmath179 and we prove the following lemma .
[ lemvartauanom ] let @xmath1132}$ ] define a smooth homotopy of asymptotically standard parallelizations of @xmath187 .
@xmath1133 proof : set @xmath1134 , observe that @xmath854 ( which is valued in a finite - dimensional vector space ) is differentiable thanks to the expression of @xmath1135 in lemma [ leminvone ] ( any function @xmath1136\times c}\omega$ ] for a smooth compact manifold @xmath117 and a smooth form @xmath64 on @xmath714\times c$ ] is differentiable with respect to @xmath605 ) .
now , the forms associated to edges of @xmath1079 do not depend on the configuration of @xmath1137 .
they will be integrated along @xmath714 \times ( \cup_{m(b)\in k_j}\check{s}_b(t_{m(b)}m,\gamma_b))$ ] , while the other ones will be integrated along @xmath1138 at @xmath1139 $ ] .
therefore , the global variation @xmath1140 reads @xmath1141\sharp_j\right ) z(u ) du\ ] ] where @xmath1142 define @xmath1143,c \in \cup_{m(b)\in k_j}\check{s}_b(t_{m(b)}m,\gamma_b ) } \bigwedge_{e \in e(\gamma_b)}p_e^{\ast}(\omega_{s^2})(u , c)\ ] ] so that @xmath1144 is the integral of @xmath1145 along @xmath1146 and @xmath1147\sharp_j\right ) z(u)\right ) du.\ ] ] therefore , @xmath1148\sharp_j\right ) z(t)\ ] ] and we are left with the computation of @xmath1149 .
the restriction of @xmath1150 from @xmath714\times u^+k_j$ ] to @xmath15 induces a map @xmath1151 \times \cup_{m(b)\in k_j}\check{s}_b(t_{m(b)}m,\gamma_b ) \rightarrow \check{q}({\gamma}_b)\ ] ] for any @xmath1079 .
@xmath1152 integrating @xmath1153 $ ] along the fiber in @xmath1154 yields a two form on @xmath15 , which is homogeneous , because everything is .
thus this form reads @xmath1155 $ ] where @xmath1156 , and where @xmath1157=\alpha$ ]
. therefore @xmath1158\times u^+k_j}p_{\tau(.)}^{\ast}(\omega_{s^2}).\ ] ] since @xmath1159\times u^+k_j}p_{\tau(.)}^{\ast}(\omega_{s^2})=\frac12 \frac{\partial}{\partial t}i_{\theta}(k_j,\tau(t))$ ] , we conclude easily .
then the derivative of @xmath1160 vanishes so that this expression does not change when @xmath179 smoothly varies .
thanks to ( * ? ? ?
* theorem 1.9 ) , in order to conclude the ( sketch of ) proof of theorem [ thmmain ] , we are left with the proof that @xmath1161 does not depend on the homotopy class of @xmath179 . when @xmath179 changes in a ball that does not meet the link , the forms can be changed only in the neighborhoods of the unit tangent bundle to this ball . using lemma [ leminvone ] again
, the variation will be seen on faces @xmath1061 , where @xmath1079 has at least one univalent vertex , and where the forms associated to the edges of @xmath1079 do not depend on the parameter in @xmath714 $ ] so that their product vanishes . in particular
, @xmath1161 is invariant under the natural action of @xmath1162 on the homotopy classes of parallelizations .
we now examine the effect of the twist of the parallelization by a map @xmath1163 . without loss , assume that @xmath1164 for some @xmath419 of @xmath15 and that @xmath405 maps @xmath781 to rotations with axis @xmath419 .
we want to compute @xmath1165 .
identify @xmath453 with @xmath463 via @xmath179 .
there exists a form @xmath64 on @xmath714\times b_m \times s^2 $ ] that reads @xmath797 on @xmath1166\times b_m \times s^2 ) \setminus ( 1 \times b_m \times s^2)$ ] ) and that reads @xmath1167 on @xmath1168 .
extend this form to a form @xmath1169 on @xmath714\times c_2(m)$ ] , that restricts to @xmath1170 as @xmath797 , and to @xmath1171 as @xmath1167 , where @xmath1172 on @xmath463 so that @xmath1173 , there .
let @xmath1174 denote the set of diagrams of @xmath762 without components without univalent vertices .
define @xmath1175 \in { { \cal a}}_n(\coprod_{j=1}^ks^1_j).\ ] ] for @xmath1176 , set @xmath1177,c \in \cup_{m(b)\in k_j}\check{s}_b(t_{m(b)}m,\gamma_b ) } \bigwedge_{e \in e(\gamma_b)}p_e^{\ast}(\omega)[\gamma_b].\ ] ] set @xmath1178 and @xmath1179 .
thanks to lemma [ leminvone ] , like in the proof of lemma [ lemvartauanom ] , @xmath1180 is differentiable , and @xmath1181 . by induction on the degree ,
it is easy to see that this equation determines @xmath1180 as a function of the @xmath1182 and @xmath1183 whose degree @xmath351 part is @xmath13 , and that @xmath1184 .
extend @xmath1169 over @xmath1185\times c_2(m)$ ] so that its restriction to @xmath1186\times b_m \times s^2 $ ] is obtained by applying @xmath1187 to the @xmath1169 translated , and extend all the introduced maps , then @xmath1188 because everything is carried by @xmath1187 . in particular @xmath1189 .
now , @xmath1190 , since @xmath1191 is homotopic to the trivial map outside a ball ( see lemma [ lempreptrivun ] , 2 ) . by induction on the degree of diagrams , this shows @xmath1192 .
conclude by observing that under our assumptions , where @xmath1193 is the linking number of @xmath781 and its parallel induced by @xmath1194 , @xmath1195 this finishes the ( sketch of ) proof of theorem [ thmmain ] in general . 1 .
a vassiliev invariant is _ odd _ if it distinguishes some knot from the same knot with the opposite orientation .
are there odd vassiliev invariants ? 2 . more generally , do vassiliev invariants distinguish knots in @xmath156 ? 3 . according to a theorem of bar - natan and lawrence @xcite ,
the lmo invariant fails to distinguish rational homology @xmath0spheres with isomorphic @xmath1196 , so that , according to a moussard theorem @xcite , rational finite type invariants fail to distinguish @xmath116-spheres .
do finite type invariants distinguish @xmath115-spheres ?
4 . find relationships between @xmath892 or other finite type invariants and heegaard floer homologies .
see @xcite to get propagators associated to heegaard diagrams
. also see related work by shimizu and watanabe @xcite .
5 . compare @xmath892 with the lmo invariant @xmath1197 .
compute the anomalies @xmath123 and @xmath1007 .
find surgery formulae for @xmath892 .
kricker defined a lift @xmath1198 of the kontsevich integral @xmath1199 ( or the lmo invariant ) for null - homologous knots in @xmath116-spheres @xcite .
the kricker lift is valued in a space @xmath1200 that is mapped to @xmath678 by a map @xmath1201 , which allows one to recover @xmath1199 from @xmath1198 .
the space @xmath1200 is a space of trivalent diagrams whose edges are decorated by rational functions whose denominators divide the alexander polynomial .
compare the kricker lift @xmath1198 with the equivariant configuration space invariant @xmath1202 of @xcite valued in the same diagram space @xmath1200 .
see @xcite for alternative definitions and further properties of @xmath1202 . 9 .
is @xmath892 obtained from @xmath1202 in the same way as @xmath1199 is obtained from @xmath1198 ?
[ secmorepar ] in order to make the definition of @xmath3 complete , we give a detailed self - contained presentation of @xmath809 . in this section ,
@xmath25 is a smooth oriented connected @xmath0-manifold with possible boundary .
again , see @xmath156 as @xmath413 and see @xmath103 as @xmath415 \times s^2)/(0 \sim \{0\ } \times s^2)$ ] . recall that @xmath416 maps @xmath417 , v \in s^2)$ ] to the rotation @xmath418 with axis directed by @xmath419 and with angle @xmath420 .
also recall that the group structure of @xmath412 $ ] is induced by the multiplication of maps , using the multiplication of @xmath411 .
any @xmath1203 induces a map @xmath1204 since @xmath1205 @xmath1206 , and the image of @xmath1207 under the above isomorphisms is denoted by @xmath1208 .
( formally , this @xmath1208 denotes the image of the generator of @xmath1209 under @xmath1208 in @xmath1210 . ) [ lempreptrivun ] let @xmath25 be an oriented connected @xmath0-manifold with possible boundary . recall that @xmath439 is a map that coincides with @xmath440 on a ball @xmath103 embedded in @xmath25 and that maps the complement of @xmath103 to the unit of @xmath411 . 1 . any homotopy class of a map @xmath405 from @xmath428 to @xmath429 , such that @xmath1208 is trivial , belongs to the subgroup @xmath1211>$ ] of @xmath412 $ ] generated by @xmath1212 $ ] .
2 . for any @xmath1213 \in [ ( m,\partial m),(so(3),1)]$ ] , @xmath1213 ^ 2 \in < [ \rho_m(b^3)]>.$ ] 3 .
the group @xmath412 $ ] is abelian .
proof : let @xmath1203 .
assume that @xmath1208 is trivial .
choose a cell decomposition of @xmath25 with respect to its boundary , with only one three - cell , no zero - cell if @xmath1214 , one zero - cell if @xmath1215 , one - cells , and two - cells .
then after a homotopy relative to @xmath100 , we may assume that @xmath405 maps the one - skeleton of @xmath25 to @xmath13 .
next , since @xmath1216 , we may assume that @xmath405 maps the two - skeleton of @xmath25 to @xmath13 , and therefore that @xmath405 maps the exterior of some @xmath0-ball to @xmath13 .
now @xmath405 becomes a map from @xmath1217 to @xmath411 , and its homotopy class is @xmath1218 $ ] in @xmath424 $ ] .
therefore @xmath405 is homotopic to @xmath1219 .
this proves the first assertion .
since @xmath1220 is trivial , the second assertion follows . for the third assertion , first note that @xmath1212 $ ] belongs to the center of @xmath412 $ ] because it can be supported in a small ball disjoint from the support ( preimage of @xmath1221 ) of a representative of any other element .
therefore , according to the second assertion any square will be in the center .
furthermore , since any commutator induces the trivial map on @xmath522 , any commutator is in @xmath1211>$ ] .
in particular , if @xmath392 and @xmath405 are elements of @xmath412 $ ] , @xmath1222 where the first factor equals @xmath1223 . exchanging @xmath392 and @xmath405 yields @xmath1224
. then the commutator , which is a power of @xmath1212 $ ] , has a vanishing square , and thus a vanishing degree
. then it must be trivial . in this subsection , we prove the following standard theorem .
the spirit of our proof is the same as the kirby proof in @xcite .
but instead of assuming familiarity with the obstruction theory described by steenrod in ( * ? ? ?
* part iii ) , we use this proof as an introduction to this theory .
[ thmpar ] any oriented @xmath0-manifold is parallelizable .
the restriction of the tangent bundle @xmath437 to an oriented @xmath0-manifold @xmath25 to any closed ( non - necessarily orientable ) surface @xmath1225 immersed in @xmath25 is trivializable .
proof : let us first prove that this bundle is independent of the immersion .
it is the direct sum of the tangent bundle to the surface and of its normal one - dimensional bundle .
this normal bundle is trivial when @xmath1225 is orientable , and its unit bundle is the @xmath1-fold orientation cover of the surface , otherwise .
( the orientation cover of @xmath1225 is its @xmath1-fold orientable cover , which is trivial over annuli embedded in the surface ) .
then since any surface @xmath1225 can be immersed in @xmath4 , the restriction @xmath1226 is the pull - back of the trivial bundle of @xmath4 by such an immersion , and it is trivial . then using stiefel - whitney classes , the proof of theorem [ thmpar ] quickly goes as follows .
let @xmath25 be an orientable smooth @xmath0-manifold , equipped with a smooth triangulation .
( a theorem of whitehead proved in the munkres book @xcite ensures the existence of such a triangulation . ) by definition , the _ first stiefel - whitney class _
@xmath1227 seen as a map from @xmath522 to @xmath1228 maps the class of a loop @xmath216 embedded in @xmath25 to @xmath351 if @xmath1229 is orientable and to @xmath13 otherwise .
it is the obstruction to the existence of a trivialization of @xmath437 over the one - skeleton of @xmath25 .
since @xmath25 is orientable , the first stiefel - whitney class @xmath1230 vanishes and @xmath437 can be trivialized over the one - skeleton of @xmath25 .
the _ second stiefel - whitney class _ @xmath1231 seen as a map from @xmath1232 to @xmath1228 maps the class of a connected closed surface @xmath1225 to @xmath351 if @xmath1226 is trivializable and to @xmath13 otherwise .
the second stiefel - whitney class @xmath1233 is the obstruction to the existence of a trivialization of @xmath437 over the two - skeleton of @xmath25 , when @xmath1234 .
according to the above lemma , @xmath1235 , and @xmath437 can be trivialized over the two - skeleton of @xmath25 . then since @xmath1236 , any parallelization over the two - skeleton of @xmath25 can be extended as a parallelization of @xmath25 .
we detail the involved arguments below without mentioning stiefel - whitney classes , ( actually by almost defining @xmath1233 ) . the elementary proof below
can be thought of as an introduction to the obstruction theory used above .
elementary proof of theorem [ thmpar ] : let @xmath25 be an oriented @xmath0-manifold . choose a triangulation of @xmath25 .
for any cell @xmath216 of the triangulation , define an arbitrary trivialization @xmath1237 such that @xmath1238 induces the orientation of @xmath25 .
this defines a trivialization @xmath1239 of @xmath25 over the @xmath351-skeleton @xmath1240 of @xmath25 .
let @xmath1241 be the set of @xmath114cells of the triangulation .
every cell is equipped with an arbitrary orientation . for an edge @xmath1242 of the triangulation , on @xmath1243
, @xmath1244 reads @xmath1245 for a map @xmath1246 .
since @xmath401 is connected , @xmath1247 extends to @xmath764 , and @xmath1248 extends @xmath1244 to @xmath764 .
doing so for all the edges extends @xmath1244 to a trivialization @xmath1249 of the one - skeleton @xmath1250 of @xmath25 . for an oriented triangle @xmath605 of the triangulation , on @xmath1251
, @xmath1249 reads @xmath1252 for a map @xmath1253 .
let @xmath1254 be the homotopy class of @xmath1255 in @xmath1256 , @xmath1254 is independent of @xmath1257
. then @xmath1258 is a cochain . when @xmath1259 , @xmath1249 may be extended to a trivialization @xmath1260 over the two - skeleton of @xmath25 , as before . since @xmath1236 , @xmath1260
can next be extended over the three - skeleton of @xmath25 , that is over @xmath25 .
let us now study the obstruction cochain @xmath1261 whose vanishing guarantees the existence of a parallelization of @xmath25 .
if the map @xmath1247 associated to @xmath764 is changed to @xmath1262 for some @xmath1263 for every edge @xmath764 , define the associated trivialization @xmath1264 , and the cochain @xmath1265 that maps @xmath764 to the homotopy class of @xmath1266 . then @xmath1267 is the coboundary of @xmath1268 . let us show that @xmath1261 is a cocycle . consider a @xmath0-simplex @xmath756 , then @xmath1244 extends to @xmath756 . without loss of generality , assume that @xmath1269 coincides with this extension , that for any face @xmath605 of @xmath756 , @xmath1257 is the restriction of @xmath1269 to @xmath605 , and that the above @xmath1264 coincides with @xmath1269 on the edges of @xmath1270 . then @xmath1271 . since a coboundary also maps @xmath1270 to @xmath351 , @xmath1272 .
now , it suffices to prove that the cohomology class of @xmath1261 ( which is actually @xmath1233 ) vanishes in order to prove that there is an extension @xmath1264 of @xmath1244 on @xmath1250 that extends on @xmath25 .
since @xmath1273 , it suffices to prove that @xmath1261 maps any @xmath1dimensional @xmath1228-cycle @xmath117 to @xmath351 .
we represent the class of such a cycle @xmath117 by a non - necessarily orientable closed surface @xmath1225 as follows .
let @xmath1274 and @xmath1275 be small regular neighborhoods of @xmath1240 and @xmath1250 in @xmath25 , respectively , such that @xmath1276 is a disjoint union , running over the edges @xmath764 , of solid cylinders @xmath1277 identified with @xmath880,1 [ \times d^2 $ ] .
the core @xmath880,1 [ \times \{0\}$ ] of @xmath12780,1 [ \times d^2 $ ] is a connected part of the interior of the edge @xmath764 .
( @xmath1275 is thinner than @xmath1274 . )
construct @xmath1225 in the complement of @xmath1279 as the intersection of the support of @xmath117 with this complement .
then the closure of @xmath1225 meets the part @xmath714 \times s^1 $ ] of every @xmath1280 as an even number of parallel intervals from @xmath1281 to @xmath1282 .
complete @xmath1225 in @xmath1283 by connecting the intervals pairwise in @xmath1280 by disjoint bands .
after this operation , the boundary of the closure of @xmath1225 is a disjoint union of circles in the boundary of @xmath1274 , where @xmath1274 is a disjoint union of balls around the vertices .
glue disjoint disks of @xmath1274 along these circles to finish the construction of @xmath1225 .
extend @xmath1244 to @xmath1274 , assume that @xmath1249 coincides with this extension over @xmath1284 , and extend @xmath1249 to @xmath1275 .
then @xmath1226 is trivial , and we may choose a trivialization @xmath1285 of @xmath437 over @xmath1225 that coincides with our extension of @xmath1244 over @xmath1274 , over @xmath1286 .
we have a cell decomposition of @xmath1287 with only @xmath13-cells and @xmath1-cells , where the @xmath1-cells of @xmath1225 are in one - to - one canonical correspondence with the @xmath1-cells of @xmath117 , and one - cells bijectively correspond to bands connecting two - cells in the cylinders @xmath1277 .
these one - cells are equipped with the trivialization of @xmath437 induced by @xmath1249 .
then we can define @xmath1dimensional cochains @xmath1288 and @xmath1289 from @xmath1290 to @xmath1228 as before , with respect to this cellular decomposition of @xmath1225 , where @xmath1291 is again a coboundary and @xmath1292 so that @xmath1293 , and since @xmath1294 , @xmath1295 and we are done .
let @xmath1225 be a non - necessarily orientable closed surface embedded in the interior of @xmath25 , and let @xmath179 be a parallelization of @xmath25 .
we define a twist @xmath1296 below .
the surface @xmath1225 has a tubular neighborhood @xmath1297 , which is a @xmath1298$]bundle over @xmath1225 that admits ( orientation - preserving ) bundle charts with domains @xmath1298\times d$ ] for disks @xmath345 of @xmath1225 so that the changes of coordinates restrict to the fibers as @xmath1299
. then @xmath1300 is the continuous map that maps @xmath1301 to @xmath13 such that @xmath1302\times d)$ ] is the rotation with angle @xmath1303 and with axis @xmath1304 where @xmath1305 \times s)$ ] is the tangent vector to the fiber @xmath1298 \times s$ ] at @xmath1306 .
since this rotation coincides with the rotation with opposite axis and with opposite angle @xmath1307 , our map @xmath1308 is a well - defined continuous map . clearly , the homotopy class of @xmath1308 only depends on the homotopy class of @xmath179 and on the isotopy class of @xmath1225 . when @xmath1309 , when @xmath179 is the standard parallelization of @xmath4 , and when @xmath1310 denotes the sphere @xmath1311 inside @xmath103 , the homotopy class of @xmath1312 coincides with the homotopy class of @xmath440 . [ lemgstau ]
@xmath1313 is the mod @xmath1 intersection with @xmath1225 .
+ @xmath1314 \rightarrow h^1(m,\partial m;{\mathbb{z}}/2{\mathbb{z}})$ ] is onto .
proof : the first assertion is obvious , and the second one follows since @xmath1210 is the poincar dual of @xmath1232 and since any element of @xmath1232 is the class of a closed surface . [ lemdeg ] the degree is a group homomorphism @xmath433 \longrightarrow { \mathbb{z}}\ ] ] and @xmath1315 . proof : it is easy to see that @xmath1316 when @xmath392 or @xmath405 is a power of @xmath1212 $ ] .
let us prove that @xmath1317 for any @xmath392 . according to lemma [ lemgstau ] , there is an unoriented embedded surface @xmath1318 of the interior of @xmath117 such that @xmath1319 for some trivialization @xmath179 of @xmath437 .
then , according to lemma [ lempreptrivun ] , @xmath1320 is homotopic to some power of @xmath441 , and we are left with the proof that the degree of @xmath1191 is @xmath1321 for @xmath1322 .
this can easily be done by noticing that @xmath1191 is homotopic to @xmath1323 where @xmath1324 is the boundary of the tubular neighborhood of @xmath1318 . in general , @xmath1325 , and
the lemma is proved .
lemmas [ lempreptrivun ] and [ lemdeg ] imply the following lemma .
[ lemdegtwo ] the degree induces an isomorphism @xmath433 \otimes_{{\mathbb{z}}}{\mathbb{q}}\longrightarrow { \mathbb{q}}.\ ] ] any group homomorphism @xmath1326 \longrightarrow { \mathbb{q}}$ ] reads @xmath1327 .
let @xmath1328 or @xmath6 .
let @xmath590 .
the stabilization maps induced by the inclusions @xmath1329 will be denoted by @xmath94 .
elements of @xmath1330 are represented by matrices whose columns contain the coordinates of the images of the basis elements , with respect to the standard basis of @xmath1331 see @xmath156 as the unit sphere of @xmath1332 so that its elements are the pairs @xmath1333 of complex numbers such that @xmath1334 .
the group @xmath523 is identified with @xmath156 by the homeomorphism @xmath1335 \end{array}\ ] ] so that the first non trivial homotopy group of @xmath523 is @xmath1336.\ ] ] the long exact sequence associated to the fibration @xmath1337 shows that @xmath1338 is an isomorphism for @xmath1339 and @xmath1340 , and in particular , that @xmath1341 for @xmath1342 and @xmath1343\ ] ] where @xmath1344 is the following map @xmath1345 \end{array}.\ ] ] let @xmath1346 and @xmath1347 be two compact connected oriented @xmath0-manifolds whose boundaries have collars that are identified by a diffeomorphism .
let @xmath1348 and @xmath1349 be two parallelizations ( which respect the orientations ) that agree on the collar neighborhoods of @xmath1350 . then the _ relative pontrjagin number @xmath1351 _ is the pontrjagin obstruction to extending the trivialization of @xmath1352 induced by @xmath1353 and @xmath1354 across the interior of a signature @xmath351 cobordism @xmath372 from @xmath1346 to @xmath1347 .
details follow .
let @xmath25 be a compact connected oriented @xmath0-manifold .
a _ special complex trivialization _ of @xmath437 is a trivialization of @xmath1355 that is obtained from a trivialization @xmath407 that induces the orientation of @xmath25 by composing @xmath1356 by @xmath1357 for a map @xmath1358 .
the definition and properties of relative pontrjagin numbers , which are given with more details below , are valid for pairs of special complex trivializations .
the _ signature _ of a @xmath2-manifold is the signature of the intersection form on its @xmath1359 ( number of positive entries minus number of negative entries in a diagonalised version of this form ) . also recall that any closed oriented three - manifold bounds a compact oriented @xmath2-dimensional manifold whose signature may be arbitrarily changed by connected sums with copies of @xmath1360 or @xmath1361 .
cobordism from @xmath1346 to @xmath1347 _ is a compact oriented @xmath2-dimensional manifold @xmath372 with corners such that @xmath1362 \times \partial m_0 ) \cup_{\partial m_1 \sim 1 \times \partial m_0 } m_1,\ ] ] is identified with an open subspace of one of the products @xmath1363 or @xmath880,1 ] \times m_1 $ ] near @xmath1364 , as the following picture suggests .
( -3,-.7 ) rectangle ( 5.5,2.5 ) ; ( 0,0 )
( 4,0 ) ( 4,2 )
( 0,2 ) ( 2,1 ) node@xmath1365 ( 0,1 ) node[left]@xmath1366 ( 4,1 ) node[right]@xmath1367 ; ( 0,0 ) ( 0,2 ) ( 4,0 )
( 4,2 ) ; ( 1.2,0 ) ( .8,-.4 ) ( 1.2,2 ) ; ( .9,-.5 ) node[left]@xmath714 \times ( -\partial m_0)$ ] ; ( 4,.1 ) node@xmath1368 ( 3.5,.1 ) node@xmath1368 ( 3,.1 ) node@xmath1368 node[below](3,-.1 ) @xmath1369 ; let @xmath1370 be such a cobordism from @xmath1346 to @xmath1347 , with signature @xmath351 . consider the complex @xmath2-bundle @xmath1352 over @xmath372 .
let @xmath1371 be the tangent vector to @xmath714 \times \{\mbox{pt}\}$ ] over @xmath1364 ( under the identifications above ) , and let @xmath1372 denote the trivialization of @xmath1352 over @xmath1364 that is obtained by stabilizing either @xmath1353 or @xmath1354 into @xmath1373 or @xmath1374 .
then the obstruction to extending this trivialization to @xmath372 is the relative first _
pontrjagin class _
@xmath1375 \in h^4(w,\partial w;{\mathbb{z}}=\pi_3(su(4)))={\mathbb{z}}[w,\partial w]\ ] ] of the trivialization .
now , we specify our sign conventions for this pontrjagin class .
they are the same as in @xcite .
in particular , @xmath286 is the opposite of the second chern class @xmath1376 of the complexified tangent bundle .
see @xcite .
more precisely , equip @xmath1346 and @xmath1347 with riemannian metrics that coincide near @xmath1377 , and equip @xmath372 with a riemannian metric that coincides with the orthogonal product metric of one of the products @xmath714 \times m_0 $ ] or @xmath714 \times m_1 $ ] near @xmath1364 .
equip @xmath1352 with the associated hermitian structure .
the determinant bundle of @xmath1378 is trivial because @xmath372 is oriented , and @xmath1379 is also trivial .
our parallelization @xmath1372 over @xmath1364 is special with respect to the trivialization of @xmath1379 .
up to homotopy , assume that @xmath1372 is unitary with respect to the hermitian structure of @xmath1352 and the standard hermitian form of @xmath1380 .
since @xmath1381 when @xmath1382 , the trivialization @xmath1372 extends to a special unitary trivialization @xmath179 outside the interior of a @xmath2-ball @xmath1383 and defines @xmath1384 over the boundary @xmath1385 of this @xmath2-ball @xmath1383 . over this @xmath2-ball @xmath1383 ,
the bundle @xmath1352 admits a trivialization @xmath1386 then @xmath1387 , for a map @xmath1388 whose homotopy class reads @xmath1389=-p_1(w;\tau(\tau_0,\tau_1))[i^2(m^{{\mathbb{c}}}_r ) ] \in \pi_3(su(4)).\ ] ] define @xmath1390 .
[ proppontdef ] the first _ pontrjagin number _
@xmath1351 is well - defined by the above conditions .
proof : according to the nokivov additivity theorem , if a closed ( compact , without boundary ) @xmath2-manifold @xmath66 reads @xmath1391 where @xmath1392 and @xmath1393 are two @xmath2-manifolds with boundary , embedded in @xmath66 that intersect along a closed @xmath0manifold @xmath555 ( their common boundary , up to orientation ) then @xmath1394 according to a rohlin theorem ( see @xcite or @xcite ) , when @xmath66 is a compact oriented @xmath2manifold without boundary , @xmath1395 .
we only need to prove that @xmath1351 is independent of the signature @xmath351 cobordism @xmath372 .
let @xmath1396 be a @xmath2-manifold of signature @xmath351 bounded by @xmath1397 .
then @xmath1398 is a @xmath2-dimensional manifold without boundary whose signature is @xmath1399 by the novikov additivity theorem .
according to the rohlin theorem , the first pontrjagin class of @xmath1398 is also zero . on the other hand
, this first pontrjagin class is the sum of the relative first pontrjagin classes of @xmath372 and @xmath1396 with respect to @xmath1372 .
these two relative pontrjagin classes are opposite and therefore the relative first pontrjagin class of @xmath372 with respect to @xmath1372 does not depend on @xmath372 .
similarly , it is easy to prove the following proposition .
[ proppontdefsignon ] under the above assumptions except for the assumption on the signature of the cobordism @xmath372 , @xmath1400 in this subsection , we describe @xmath1401 and the natural maps from @xmath1162 to @xmath1401 and to @xmath1402 . the _ quaternion field @xmath1403 _ is the vector space @xmath1404 equipped with the multiplication that maps @xmath1405 to @xmath1406 , and with the conjugation that maps @xmath1407 to @xmath1408 .
the norm of @xmath1407 is the square root of @xmath1409 , it is multiplicative . setting @xmath1410 , @xmath1411 is an orthogonal basis of @xmath1403 with respect to the scalar product associated to the norm .
the unit sphere of @xmath1403 is the sphere @xmath156 , which is equipped with the corresponding group structure .
there are two group morphisms from @xmath156 to @xmath1412 induced by the multiplication in @xmath1403 .
@xmath1413 @xmath1414 together , they induce the group morphism @xmath1415 the kernel of this group morphism is @xmath1416 so that this morphism is a two - fold covering . in particular , @xmath1417 \oplus { \mathbb{z}}[\overline{m_r}].\ ] ] for @xmath1328 or @xmath6 and @xmath590 , the @xmath155 ( euclidean or hermitian ) oriented vector space with the direct orthonormal basis @xmath1418 is denoted by @xmath1419 .
there is also the following group morphism @xmath1420 whose kernel is @xmath1421 .
this morphism @xmath1422 is also a two - fold covering .
this definition of @xmath1422 coincides with the previous one , up to homotopy .
proof : it is clear that the two maps coincide up to homotopy , up to orientation since both classes generate @xmath1423 . we take care of the orientation using the outward normal first convention to orient boundaries , as usual . an element of @xmath156 reads @xmath1424 for a unique @xmath1425 $ ] and a unit quaternion @xmath419 with real part zero , which is unique when @xmath1426 .
in particular , this defines a diffeomorphism @xmath1427 from @xmath880,\pi[\times s^2 $ ] to @xmath1428 .
we compute the degree of @xmath1427 at @xmath1429 .
the space @xmath1403 is oriented as @xmath1430 , where @xmath1431 is oriented by the outward normal to @xmath15 , which coincides with the outward normal to @xmath156 in @xmath1432 , followed by the orientation of @xmath15 . in particular since @xmath1433 is an orientation - reversing diffeomorphism at @xmath1434 , the degree of @xmath1427 is @xmath13 and @xmath1427 preserves the orientation .
now @xmath1435 where @xmath1436 is a rotation with axis @xmath419 for any @xmath419 .
since @xmath1437 , the two maps @xmath1422 are homotopic .
one can check that they are actually conjugate .
define @xmath1438 [ lempitroissoquatrestr ] in @xmath1417 \oplus { \mathbb{z}}[\overline{m_r}],\ ] ] @xmath1439)=[m_{\ell } ] + [ \overline{m}_r]= [ m_{\ell } ] - [ { m}_r].\ ] ] proof : the @xmath1440-product in @xmath1401 coincides with the product induced by the group structure of @xmath1412 .
[ lem237 ] recall that @xmath1441 denotes the map from the unit sphere @xmath156 of @xmath1403 to @xmath1442 induced by the right - multiplication . denote the inclusions @xmath1443 by @xmath216 .
then in @xmath1402 , @xmath1444)=2[i^2(m^{{\mathbb{c}}}_r)].\ ] ] proof : let @xmath1445 denote the complexification of @xmath1446 . here , @xmath1447 . when @xmath1448 and @xmath1449 , @xmath1450 , and @xmath1451 .
let @xmath1452 , define @xmath1453 consider the quotient @xmath1454 . in this quotient ,
@xmath1455 , @xmath1456 , and since @xmath1451 , @xmath1457 and @xmath1458 .
therefore this quotient is isomorphic to @xmath1403 as a real vector space with its complex structure @xmath1459 .
then it is easy to see that @xmath1460 maps @xmath1461 to @xmath351 in this quotient .
thus @xmath1462 .
now , observe that @xmath1445 is the orthogonal sum of @xmath1463 and @xmath1464 . in particular , @xmath1461 is isomorphic to the quotient @xmath1465 , which is isomorphic to @xmath1466 and @xmath1460 acts on it by the right multiplication .
therefore , with respect to the orthonormal basis @xmath1467 , @xmath1460 reads @xmath1468.\ ] ] therefore , the homotopy class of @xmath1460 is the sum of the homotopy classes of @xmath1469 \;\ ; \mbox{and}\;\ ; ( z_1+z_2j ) \mapsto \left[\begin{array}{cc } 1 & 0 \\ 0 & m^{{\mathbb{c}}}_r \circ \iota(z_1,z_2 ) \end{array } \right]\ ] ] where @xmath1470 . since the first map is conjugate by a fixed element of @xmath1471 to @xmath1472 , it is homotopic to @xmath1472 , and since @xmath512 induces the identity on @xmath1473 , the second map is homotopic to @xmath1472 , too .
the following lemma finishes to determine the maps @xmath1474 and @xmath1475 [ lempitroissoquatre ] @xmath1476)=c_{\ast}([m_{\ell}])=-2[i^2(m^{{\mathbb{c}}}_r)].\ ] ] @xmath1477))=-4[i^2(m^{{\mathbb{c}}}_r)].\ ] ] proof : according to lemma [ lempitroissoquatrestr ] , @xmath1478)=[m_{\ell } ] + [ \overline{m}_r]= [ m_{\ell } ] - [ { m}_r].$ ] using the conjugacy of quaternions , @xmath1479 .
therefore @xmath1480 is conjugated to @xmath1481 via the conjugacy of quaternions , which lies in @xmath1482 .
since @xmath1483 is connected , the conjugacy by an element of @xmath1483 induces the identity on @xmath1402 .
thus , @xmath1484)=c_{\ast}([\overline{m}_{r}])=-c_{\ast}([{m}_{r}]),\ ] ] and @xmath1477))=-2 c_{\ast}([{m}_{r}]).\ ] ] we finish proving theorem [ thmpone ] by proving the following proposition .
[ proppontdeg ] let @xmath1346 and @xmath25 be two compact connected oriented @xmath0-manifolds whose boundaries have collars that are identified by a diffeomorphism .
let @xmath1485 and @xmath1486 be two special complex trivializations ( which respect the orientations ) that coincide on the collar neighborhoods of @xmath1487 .
let @xmath1488 $ ] denote the group of homotopy classes of maps from @xmath25 to @xmath1489 that map @xmath100 to @xmath13 .
for any @xmath1490 define @xmath1491 then @xmath1492 is independent of @xmath1353 and @xmath179 , @xmath1493 induces an isomorphism from the group @xmath1488 $ ] to @xmath115 , and , if @xmath405 is valued in @xmath411 , then @xmath1494 in order to prove this proposition , we first prove the following lemma . [ lempunind ] under the hypotheses of proposition [ proppontdeg ] , @xmath1495 is independent of @xmath1353 and @xmath179 .
proof : indeed , @xmath1495 can be defined as the obstruction to extending the following trivialization of the complexified tangent bundle to @xmath714 \times m$ ] restricted to the boundary .
this trivialization is @xmath1496 \oplus \tau$ ] on @xmath1497 \times \partial m)$ ] and @xmath1496 \oplus \tau \circ \psi(g)$ ] on @xmath1498 . but
this obstruction is the obstruction to extending the map @xmath1499 from @xmath1500 \times m)$ ] to @xmath1471 that maps @xmath1497 \times \partial m)$ ] to @xmath13 and that coincides with @xmath1501 on @xmath1498 , regarded as a map from @xmath1500 \times m)$ ] to @xmath1471 , over @xmath499 \times m)$ ] .
this obstruction , which lies in @xmath1402 since @xmath1502 , for @xmath1382 , is independent of @xmath1353 and @xmath179 .
proof of proposition [ proppontdeg ] : lemma [ lempunind ] guarantees that @xmath1493 defines two group homomorphisms to @xmath115 from @xmath1488 $ ] and from @xmath1503 $ ] .
since @xmath1504 is trivial for @xmath1382 and since @xmath1505 , the group of homotopy classes @xmath1506 $ ] is generated by the class of a map that maps the complement of a @xmath0-ball @xmath104 to @xmath13 and that factors through a map that generates @xmath1507 . by definition of the pontrjagin classes
, @xmath1493 sends such a generator to @xmath988 and it induces an isomorphism from @xmath1488 $ ] to @xmath115 . according to lemma [ lempreptrivun ] and to lemma [ lemdegtwo ] , the restriction of @xmath1493 to @xmath1508 $ ] must read @xmath1509 , and we are left with the proof of the following lemma .
[ lemvarpun ] @xmath1510 let @xmath1511 , we can extend @xmath1499 ( defined in the proof of lemma [ lempunind ] ) by the constant map with value 1 outside @xmath1512 \times b^3 \cong b^4 $ ] and , in @xmath1402 @xmath1513=-p_1(\tau,\tau \circ \psi(g))[i^2(m^{{\mathbb{c}}}_r)].\ ] ] since @xmath1514 is homotopic to @xmath1515 , lemma [ lempitroissoquatre ] allows us to conclude .
piecewise linear ( or pl ) @xmath24manifolds can be defined as the @xmath1516-manifolds of subsection [ subbackground ] by replacing @xmath1516 with piecewise linear ( or pl ) . when @xmath1517 , the above notion of pl - manifold coincides with the notions of smooth and topological manifold , according to the following theorem .
this is not true anymore when @xmath1518 .
see @xcite .
[ thmstructhree ] when @xmath1517 , the category of topological @xmath24manifolds is isomorphic to the category of pl @xmath24manifolds and to the category of @xmath33 @xmath24manifolds , for @xmath92 .
for example , according to this statement , which contains several theorems ( see @xcite ) , any topological @xmath0manifold has a unique @xmath8structure .
below @xmath1519 .
the equivalence between the @xmath1520-categories follows from work of whitney in 1936 @xcite . in 1934 , cairns @xcite provided a map from the @xmath1521category to the pl category , which shows the existence of a triangulation for @xmath1521manifolds , and he proved that this map is onto ( * ? ? ?
* theorem iii ) in 1940 .
moise @xcite proved the equivalence between the topological category and the pl category in 1952 .
this diagram was completed by munkres ( * ? ? ?
* theorem 6.3 ) and whitehead @xcite in 1960 by their independent proofs of the injectivity of the natural map from the @xmath1522category to the topological category .
2 chern - simons perturbation theory , _ proceedings of the xxth international conference on differential geometric methods in theoretical physics , vol . 1 , 2 ( new york , 1991 ) _ , world sci .
publ . , river edge , nj , 1992 , p. 345 .
( ) _ la recherche de la topologie perdue _ , progress in mathematics , vol .
62 , birkhuser boston inc . ,
boston , ma , 1986 , i. du ct de chez rohlin .
ii . le ct de casson .
[ i. rokhlin s way .
casson s way ] .
an invariant of integral homology @xmath0-spheres which is universal for all finite type invariants , solitons , geometry , and topology : on the crossroad , amer .
179 , amer .
soc . , providence , ri , 1997 , p. 75100 .
invariants of knots and 3-manifolds derived from the equivariant linking pairing , chern - simons gauge theory : 20 years after , ams / ip stud . adv .
50 , amer . math .
soc . , providence , ri , 2011 , p. 217242 .
_ topology from the differentiable viewpoint _ ,
princeton landmarks in mathematics , princeton university press , princeton , nj , 1997 , based on notes by david w. weaver , revised reprint of the 1965 original . | the finite type invariant concept for knots was introduced in the 90 s in order to classify knot invariants , with the work of vassiliev , goussarov and bar - natan , shortly after the birth of numerous quantum knot invariants .
this very useful concept was extended to @xmath0manifold invariants by ohtsuki .
these introductory lectures show how to define finite type invariants of links and @xmath0-manifolds by counting graph configurations in @xmath0-manifolds , following ideas of witten and kontsevich .
the linking number is the simplest finite type invariant for @xmath1component links .
it is defined in many equivalent ways in the first section . as an important example , we present it as the algebraic intersection of a torus and a @xmath2-chain called a _ propagator _ in a configuration space . in the second section ,
we introduce the simplest finite type @xmath0manifold invariant , which is the casson invariant ( or the @xmath3invariant ) of integer homology @xmath0spheres .
it is defined as the algebraic intersection of three propagators in the same two - point configuration space . in the third section
, we explain the general notion of finite type invariants and introduce relevant spaces of feynman jacobi diagrams . in sections 4 and 5
, we sketch an original construction based on configuration space integrals of universal finite type invariants for links in rational homology @xmath0spheres and we state open problems .
our construction generalizes the known constructions for links in @xmath4 and for rational homology @xmath0spheres , and it makes them more flexible . in section 6 , we present the needed properties of parallelizations of @xmath0manifolds and associated pontrjagin classes , in details .
* keywords : * knots , @xmath0-manifolds , finite type invariants , homology @xmath0spheres , linking number , theta invariant , casson - walker invariant , feynman jacobi diagrams , perturbative expansion of chern - simons theory , configuration space integrals , parallelizations of @xmath0manifolds , first pontrjagin class + * msc : * 57m27 57n10 55r80 57r20 |
one of the major results of the _ infrared astronomical satellite _ ( iras ) mission came from its low resolution spectrometer ( lrs ) with the detection of a new broad feature around 21 @xmath0 m @xcite in the spectra of stars that have gone through the asymptotic giant branch ( agb ) phase .
the existence of this feature was not recognized at once , because it is weak in the first object in which it was observed , iras 22272 + 5435 , which was listed in the lrs atlas as having a low - temperature continuum and silicate absorption @xcite . in the first paper about this feature ,
four sources are listed ( namely iras 07134 + 1005 , 23304 + 6147 , 22272 + 5435 , and 04296 + 3429 ) ; subsequent ground - based ( united kingdom infrared telescope ) , airborne ( kuiper airborne observatory ) , and space - based ( infrared space observatory ) observations confirmed those detections and also added new members to the group of 21-@xmath0 m emitters .
observations with the _ infrared space observatory _ ( iso ) led to conclude that the feature actually peaks around 20.1 @xmath0 m @xcite , although it is still today commonly referred to as the 21 @xmath0 m feature .
the feature has been detected almost exclusively in c - rich proto - planetary nebulae ( ppne ) , intermediate - mass stars ( 18 m@xmath1 ) transiting from the agb to the planetary nebula ( pn ) stage . during the agb phase ,
stars lose most of their initial mass , with mass loss rates ranging from 10@xmath2 up to 10@xmath3 m@xmath1 yr@xmath4 .
the agb stellar atmospheres are subject to periodic he - flashes , which imply that enriched material produced by nucleosynthesis in the bottom layers is carried to the surface . if this dredge - up of the star is efficient enough , the chemistry of the cooled material in the circumstellar envelope ( cse )
will be based on c rather than o , as normally expected .
depending on their @xmath5c@xmath6/[$]o@xmath6 $ ] ratio , we can then distinguish between o - rich and c - rich cses .
when the star has evolved past the tip of the agb , the intense mass loss that characterizes the agb has ceased , but the dust and gas in the cse is typically so thick that the star can not be detected at optical wavelengths .
the expansion and subsequent dilution of the envelope leads to the optical detection of a post - agb star , which typically has a double - peaked spectral energy distribution ( sed ) , as the result of the radiation from the central source and its cse .
the central star then evolves towards hotter temperatures , while its cse continues to expand and dilute .
if the central star does not evolve too slowly , a few thousand years after the agb phase it is hot enough to ionize its cse ( t@xmath7 @xmath82030@xmath9 k ) , which can lead to the destruction of molecules and even dust grains .
the point at which the ionization of the cse occurs is in fact considered as the beginning of the planetary nebula phase .
all of the objects showing the 21 @xmath0 m feature are c - rich and metal poor with enhanced abundances of s - processed elements @xcite , and typically with f
g spectral type . besides the unidentified feature
, their mid - ir spectra typically show a broad feature around 30 @xmath0 m that is usually attributed to mgs @xcite and has been found to consist of two features centered at 26 and 33 @xmath0 m , though such a distinction is not observed in all sources @xcite . as of this writing
, 16 ppne are known to exhibit the unidentified 21-@xmath0 m feature @xcite . besides these sources , the feature has been detected weakly in three pne @xcite and three agb stars @xcite . it has not been observed in iras 01005 + 7910 @xcite , a c - rich ppn with a hot central star ( b0 i ) that has already started to ionize its circumstellar shell @xcite .
other hot post - agb stars showing features of polycyclic aromatic hydrocarbons ( pah ) in their mid - ir spectra and therefore likely to be c - rich do not exhibit the feature either @xcite .
this leads to the conclusion that the carrier must form during the agb phase or very early afterwards , and is probably easily destroyed as the radiation field of the central star hardens .
since its first detection , many different molecules have been considered as possible carriers of the 21-@xmath0 m feature .
the connection with the c - rich nature of the envelopes makes it obvious to take into account c - bearing molecules , for instance tic , doped - sic , pah , hydrogenated amorphous carbon ( hac ) , and fullerenes . yet , non c - bearing molecules have been indicated as possible carriers as well , for example sis@xmath10 , feo , fe@xmath10o@xmath11 , and fe@xmath11o@xmath12 @xcite .
the main issues that arise with almost all of the proposed carriers are the presence of secondary features that are not detected in the astrophysical spectra , or the anomalous abundances necessary to account for the observed intensity of the feature . in this paper , we present observations obtained within a program aimed at characterizing the dust emission in objects transiting from the agb to the pn stage with a
g spectral types , as a follow - up to our project presented in @xcite , which focused on sources with hot central stars ( b spectral type ) .
while the whole sample of transition targets will be presented in a separate paper along with data from other spitzer programs ( hart et al , in preparation ) , here we will focus on three new sources that show the unidentified 21-@xmath0 m feature .
these are the only targets in our sample , where we find evidence for this feature .
the new observations presented in this paper were carried out with the infrared array camera ( irac ; * ? ? ? * ) and the infrared spectrograph ( irs ; * ? ? ?
* ) on - board the spitzer space telescope @xcite , within program 50116 ( pi : g. fazio ) in march and april 2009 .
the irs had four modules with different spectral resolutions .
the short - low ( sl ) and long - low ( ll ) modules covered the 5.214.5 and 14.038.0 @xmath0 m ranges respectively , with spectral resolution r@xmath860120 .
the short - high ( sh ) and long - high ( lh ) modules covered the 9.919.6 and 18.737.2 @xmath0 m ranges with r@xmath8600 .
our spectral data were typically acquired in low resolution mode .
targets with iras fluxes that would imply saturation with the irs at longer wavelengths were observed in low resolution only between 5 and 14 @xmath0 m ( sl modules ) , while the 1038 @xmath0 m range was observed in high spectral resolution ( sh and lh modules ) . in the latter case ,
a dedicated background observation was also performed for each target .
for the three sources presented here , low resolution observations were performed for iras 13245 - 5036 and 15482 - 5741 , while low and high resolution data were collected for iras 14429 - 4539 .
iras 13245 - 5036 was not centered on the sl slit , therefore only ll data are available for this target .
basic calibrated data were retrieved from the spitzer web archive .
the observations of iras 13245 - 5036 were processed with the 18.7.0 pipeline by the spitzer science center ( ssc ) , while version 18.18.0 was applied to 14429 - 4539 and 15482 - 5741 .
the data were reduced following the standard recipe in the on - line irs cookbook : the background subtraction was performed from the off - target observations for low resolution data ( i.e. , when one module is on the target , the other one is observing off - source ) and from the dedicated background observation for the high resolution data sets .
the spectra were cleaned using irsclean , extracted and averaged together in smart @xcite .
the high spectral resolution data were extracted over the full slit , while the optimal extraction algorithm @xcite was applied to the sl and ll observations .
the irac observations were performed in high dynamic range mode with a small dither pattern , as very extended emission was not expected . in this observing mode ,
two sets of bcds are created by default , corresponding to a long and a short exposure ; in our case the exposures were respectively 12s and 0.6s long .
all of the irac bcds retrieved from the spitzer archive were processed with the s18.18.0 pipeline .
the bcds were treated for artifact correction with the idl code available from the ssc web site and then multi - frame aperture photometry was performed with the mosaicker and point source extractor ( mopex ) using a 10-pixel aperture and 12 - 20 pixel annulus , which is the standard for the irac calibration . as the irac
absolute flux calibrators are known to about 23% , and irac repeatability is at about the 1.5% level , we assume in the following 5% as our photometric error . since our targets are quite bright for irac , the photometry extraction was performed on the short frames , to avoid any saturation effect , which was evident in iras 14429 - 4539 at 8.0 @xmath0 m even in the short frames .
the flux densities are listed in table [ tab : irac ] .
.photometry of our targets obtained with irac .
the absolute accuracy is about 5% . [
cols="<,^,^,^,^ " , ] the continuum obtained as a combination of gray bodies is typically unable to properly isolate the feature at 21 @xmath0 m , because of the substantial contributions from the features in the 2633 @xmath0 m range , which are often blended together .
emission in the 5.518 @xmath0 m range is present in all of the targets , as shown in the _ left - side _ close - ups in figures [ fig : old1 ] and [ fig : old2 ] . after converting the spectra into w m@xmath13 @xmath0m@xmath4 ,
we integrated the continuum - subtracted spectra to estimate the flux in the 5.518 @xmath0 m and 18.523.5 @xmath0 m ranges ( the exact edges of the ranges were adapted to the emission in each source , turning out in differences of about 0.3 @xmath0 m ) . to assess the emission in the 21-@xmath0 m feature
, a straight baseline was fitted to its red and blue edges and then subtracted before performing the integration , as shown in two examples in figure [ fig : baseline ] .
the emission in the 518 @xmath0 m range is mostly due to pahs but contributions from hacs are possible , while the 18.523.5 @xmath0 m range is obviously dominated by the 21-@xmath0 m feature . in figure
[ fig : intensity ] , we plot the flux estimated in the 5.518 @xmath0 m bands , indicated as the sum of the pah 7 and 11 @xmath0 m bands , versus that in the 21-@xmath0 m feature . as can be seen in the plot , a correlation between the fluxes in the two wavelength ranges is evident ; we can calculate a linear correlation coefficient of [email protected] .
an explanation for the correlation may be that the abundances of the hydrocarbons and of the carrier of the 21-@xmath0 m feature are linked together , possibly because of a common formation process and therefore the correlation would just reflect the different amounts of hydrocarbons and 21-@xmath0 m carriers produced in the cses .
the colors of the dots in figure [ fig : intensity ] are proportional to the temperatures of the central stars , as derived from their spectral classifications , and range from red ( coldest ) to blue ( hottest ) .
no clear correlation is observed between the intensity of the features and the stellar temperature . in principle , if the observed correlation were due only to a common excitation process for the molecules involved and then essentially to the hardening of the radiation field , we would expect to find a correlation with the temperature of the central star
. there may be several reasons for this lack of correlation .
for example , extinction effects can play a role .
also , one of the explanations for the existence of the 21-@xmath0 m feature implies radiation - induced decomposition of large hac grains into smaller pahs @xcite . in this scenario ,
the carrier of the 21-@xmath0 m feature would be a short - lived decomposition product .
the efficiency of the processing of large grains would depend on both t@xmath15 and the location of the dust layer .
a massive star will be hot enough to cause processing earlier than a low - mass star , with its dust still close to the central star .
a low - mass star will reach the same temperature at a later age , with its dust farther away and will therefore need an even larger temperature to attain comparable processing .
one of the main results of the iras mission was its discovery of the existence of a broad mid - ir feature around 20 @xmath0 m .
the feature has been detected almost exclusively in the circumstellar envelopes of c - rich proto - pne , with a few weak detections in their precursors ( agb stars ) and successors ( pne ) .
the nature of the feature is still unclear today , but it must be linked to the c - rich nature of the sources : its carriers are likely to be produced towards the very end of the agb and are easily destroyed by uv radiation . in this paper
we report about the detection of three new c - rich proto - pne that exhibit this unidentified feature through observations performed with the irs instrument on - board spitzer . besides the 21 @xmath0 m feature
, our targets also show mid - ir bands from pahs and a broad feature around 30 @xmath0 m commonly attributed to mgs .
as the dispersion of the dust formed during the agb creates a very extended shell around the central star , it can be expected that evidence for such large dust regions can be found in infrared images , but our irac observations do not point to any extended emission . by a combination of catalog data and our new observations , we reconstruct the seds of our targets and model them with the code dusty . with a r@xmath13 density distribution ,
a good match of the observational points is possible , if the model assumes a circumstellar environment where the mass loss process has not been constant with time .
the use of feo as carrier of the 21-@xmath0 m feature does not accurately reproduce the spectral shape of the feature .
we have retrieved the infrared spectra of the other 21-@xmath0 m sources available in the iso and spitzer archives .
a gray - body continuum has been subtracted from the spectra to enhance the features . after continuum subtraction
, we have calculated the flux in the features from 5 to 18 @xmath0 m , due to hydrocarbons , and that in the 21-@xmath0 m feature , finding a clear correlation between them .
we have not found any correlation with stellar temperatures .
the correlation found between the 518 and 21-@xmath0 m fluxes can point to a common origin of the carriers of the features .
we are grateful to pedro garca - lario for carefully reading this manuscript and providing us with his helpful comments .
+ this work is based on observations made with the spitzer space telescope , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa .
support for this work was provided by nasa through an award issued by jpl / caltech .
the guide star catalogue - ii is a joint project of the space telescope science institute and the osservatorio astronomico di torino .
space telescope science institute is operated by the association of universities for research in astronomy , for the national aeronautics and space administration under contract nas5 - 26555 .
the participation of the osservatorio astronomico di torino is supported by the italian council for research in astronomy .
additional support is provided by european southern observatory , space telescope european coordinating facility , the international gemini project and the european space agency astrophysics division .
the denis project has been partly funded by the science and the hcm plans of the european commission under grants ct920791 and ct940627 .
it is supported by insu , men and cnrs in france , by the state of baden - wrttemberg in germany , by dgicyt in spain , by cnr in italy , by ffwfbwf in austria , by fapesp in brazil , by otka grants f-4239 and f-013990 in hungary , and by the eso c&ee grant a-04 - 046 .
jean claude renault from iap was the project manager .
observations were carried out thanks to the contribution of numerous students and young scientists from all involved institutes , under the supervision of p. fouqu , survey astronomer resident in chile .
this publication makes use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by the national aeronautics and space administration and the national science foundation .
this research is based in part on observations with akari , a jaxa project with the participation of esa .
+ _ facilities : _ , . | among its great findings , the iras mission showed the existence of an unidentified mid - ir feature around 21 @xmath0 m . since its discovery
, this feature has been detected in all c - rich proto - pne of intermediate spectral type ( a g ) and - weakly - in a few pne and agb stars , but the nature of its carriers remains unknown . in this paper , we show the detection of this feature in the spectra of three new stars transiting from the agb to the pn stage obtained with the spitzer space telescope . following a recent suggestion ,
we try to model the seds of our targets with amorphous carbon and feo , which might be responsible for the unidentified feature .
the fit thus obtained is not completely satisfactory , since the shape of the feature is not well matched . in the attempt to relate the unidentified feature to other dust features
, we retrieved mid - ir spectra of all the 21-@xmath0 m sources currently known from iso and spitzer on - line archives and noticed a correlation between the flux emitted in the 21-@xmath0 m feature and that emitted at 7 and 11 @xmath0 m ( pah bands and hac broad emission ) .
such a correlation may point to a common nature of the carriers . |
the _ quark gluon plasma _ ( @xmath0 ) usually is defined as the phase of _ quantum chromodynamics _ ( @xmath2 )
in which the quarks and gluons degrees of freedom , that is normally confined within the hadrons , are mostly liberated .
the possible phases of the @xmath2 and the precise locations of critical boundaries or points are currently being actively studied .
in fact , revealing the @xmath2 phase transition structure is one of the central aims of the ongoing and the future theoretical and the experimental research in the field of the hot and/or the dense @xmath2 @xcite .
it is about thirty years since the study of the hot and the dense nuclear matter in the form of the @xmath0 has been started .
the experiments at the @xmath3 s _ super proton synchrotron _ ( @xmath4 ) first tried to create the @xmath0 in the 1980s and 1990s : the first hints of the formation of a new state of matter was obtained from the @xmath4 data in terms of the global observable , the event - by - event fluctuations , the direct photons and the di - leptons .
the current experiments at the @xmath5 national laboratory s _ relativistic heavy ion collider _ ( @xmath6 ) are still continuing these efforts . in the april 2005 ,
the formation of the quark matter was tentatively confirmed by the results obtained at the @xmath6 .
the consensus of the four @xmath6 research groups was in favor of the creation of the quark - gluon liquid at the very low viscosity @xcite .
since , in this new phase of matter , the quarks and the gluons are in the asymptotic freedom region , one expects that they interact weakly .
so , the _ perturbative _ methods can be used for such a system .
the asymptotic freedom suggests two procedures for the creation of the @xmath0 : @xmath7i@xmath8 the recipe for the @xmath0 at high temperature .
if one treats the quarks and the gluons as the massless noninteracting gas of molecules , such that the baryon density vanishes , the critical temperature which above that , the hadronic system dissolves into a system of quarks and gluons ( @xmath0 ) is @xmath9 @xcite .
however , the modern lattice @xmath2 calculation estimates the critical temperature , @xmath10 , to be about @xmath11 @xmath12 @xcite @xmath7ii@xmath8 the recipe for the @xmath0 at high baryon density . at zero temperature ,
the critical baryon density required the transition to take place in @xmath13 @xcite i.e. four times the empirical nuclear matter density . on these grounds , one should expect to find the @xmath0 in two places in the nature : firstly , in the early universe , about @xmath14 after the _ cosmic big bang _ or secondly , at the core of _ super - dense stars _ such as the neutron and the quark stars .
this new phase of matter can also be created in the initial stage of the _ little big bang _ by means of the relativistic nucleus - nucleus collisions in the heavy ion accelerators @xcite .
the critical temperature ( critical baryon density ) at the zero baryon density ( zero temperature ) has been obtained simply for the non - interacting , massless up and down quarks and gluons in the reference @xcite .
some primary works in the zero baryon density and in the framework of the bag model have been also presented in the reference @xcite ( and the reference therein ) . for the more complicated field
theoretical approaches see the references @xcite and the references therein .
so it would be interesting to perform a similar calculation to reference @xcite , but with the non zero interaction , for the region in which both @xmath15 and @xmath16 .
so , in this work we generalize above calculations for the region with the finite temperature and the baryon density by considering the weakly interacting quarks in the framework of the _ one gluon exchange _ scheme .
since it is intended to use the @xmath17 method in our calculation , it is assumed that @xmath18 .
the dependence of @xmath19 on the temperature and the baryon density is ignored @xcite .
we perform calculations beyond the zero hadronic pressure approximation of reference @xcite and the pressure of the excluded volume hadron @xcite model is also taken into account to find the corresponding @xmath0 transition phase diagram .
so , the paper is organized as follows . in the section
@xmath20 we review the derivation of the _ one gluon exchange _ interaction formulas @xcite based on the landau fermi liquid picture @xcite .
the section @xmath21 is devoted to the calculation of the thermodynamics properties of the @xmath0 .
the result and its comparison with the strange quark matter @xcite are given in the section @xmath22 .
finally a conclusion and summary are presented in the section @xmath23 .
the landau fermi liquid theory describes the relativistic systems such as : the nuclear matter under the extreme conditions , the quark matter , the quark gluon plasma , and other relativistic plasmas .
the basic framework of the landau theory of relativistic fermi liquids is given by @xmath24 and @xmath25 @xcite . in this framework
, one can evaluate the energy density of a weakly interacting quarks in @xmath0 by the following formulas : @xmath26 with @xmath27 where @xmath28 , is the degeneracy of quarks .
@xmath29 , @xmath30 and @xmath31 are the kinetic and the potential energies and the familiar fermi - dirac liquid distributions of our quasi - particles , respectively . in the equation ( [ tatsumi ] ) , @xmath32 is the landau - fermi interaction function , which is a criterion for the interaction between the two quarks ( the `` @xmath33 '' superscript refers to the unpolarized quark matter @xcite ) .
it is related to the two - particle forward scattering amplitude i.e. : @xmath34 @xmath35 is the commonly defined lorentz invariant matrix element .
now , using the _ feynman rules _ for the @xmath2 , @xmath36 can be calculated . since
, the direct term is proportional to the trace of the _ gelman matrices _ , the color symmetric matrix element is only given by exchange contribution : @xmath37 where @xmath38 ( our choice is @xmath39 ) . since our system is unpolarized , it is possible to sum over all the spin states and get the average of this quantity as : @xmath40 then for massless up and down quarks , @xmath41 so , the total energy of interacting quarks can be evaluated by using the equations ( 1 ) , ( 2 ) , ( 3 ) and ( 7 ) . in the next section , the various thermodynamic quantities of the interacting @xmath0 like the internal energy , the pressure , the entropy , etc are calculated .
we begin with calculation of the free ultra - relativistic massless quarks partition function density , which follows by its non - interacting gluon contribution s .
since we are interested in the heavy ion processes , it is assumed that the initial state to be the nuclear matter with the equal neutron and proton densities i.e. @xmath42 .
so , @xmath43 and the contributions of @xmath44 and @xmath45 quarks are equal . then write the @xmath46 partition function for each quarks is written as ( @xmath47 is the volume ) , @xmath48\frac{4\pi g_{q}}{(2\pi)^{3 } } k^{2 } dk.\ ] ] this integral can be evaluated analytically i.e. , @xmath49 where @xmath50 and @xmath51 ( in the @xmath52 factor unit ) . and the strange quark matter , respectively.,width=415 ] and the strange quark matter , respectively.,width=415 ] and the strange quark matter , respectively.,width=415 ] and the strange quark matter , respectively.,width=415 ] equation of state as a function of baryon density at two fixed temperatures .
the solid and dash curves are for the weakly interacting @xmath0 and the strange quark matter , respectively , but without the bag constant.,width=415 ] .,width=415 ] by taking into account the nuclear matter excluded volume pressure at four different bag pressures ( @xmath12).,width=415 ] now we are in the position to evaluate the various thermodynamic quantities , such as the energy density , the entropy and the free energy from the above partition function , especially we have , @xmath53_{\cal t } & & \nonumber\\ & & \hspace{-44 mm } = \frac{4\pi g_{q}}{3(2\pi)^{3 } } ( \pi^{2 } { \cal t}^{2 } \mu+\mu^{3}),\end{aligned}\ ] ] where @xmath54 is the quark degeneracy for two , up and down quarks , flavors .
all of the thermodynamic quantities are obtained as a function of chemical potential ( @xmath55 ) and temperature ( @xmath56 ) .
the temperature is a suitable experimental quantity but the chemical potential is not .
so , it is better to rewrite the thermodynamic quantities as a function of the temperature and the baryon density , instead of the chemical potential . since the baryon number of a quark is @xmath57 , we have , @xmath58 , and : @xmath59 then , the chemical potential can be found as a function of @xmath60 and @xmath61 , from the above equation i.e. , @xmath62 so our thermodynamic quantities become a function of baryon density and temperature . for the energy density of gluons we have , @xmath63 where again @xmath64 is the degeneracy of gluons .
our system is ultra - relativistic , so , there is a simple relation between the pressure and the energy density : @xmath65 the bag pressure and the vacuum energy contributions should be included , in addition to the quark and the gluon contribution energies , @xmath66 having the pressure , the entropy density of system is evaluated , @xmath67 to study the phases of the @xmath0 , one should concentrate on the @xmath0 and the hadronic pressures , @xmath68 and @xmath69 @xcite , respectively , i.e. for @xmath70 , the system is in the confinement phase and the quarks and the gluons are inside the bag , but for @xmath71 , the quarks and the gluons pressures can overcome to the bag and the hadronic pressures and the system is in the de - confinement phase , i.e. the @xmath0 phase is created .
so , the phase diagram can be extracted by solving the following equation : @xmath72 where in this work , both @xmath73 @xcite and @xmath74 @xcite cases are considered . for @xmath74 , the excluded volume effect for the nuclear matter equation of state
is used @xcite to calculate the hadronic pressure ( equation ( 29 ) of reference @xcite ) i.e. , @xmath75},\label{n2}\ ] ] where @xmath76 is the pressure of free ideal nucleonic ( fermion ) matter @xcite ( here , degeneracy is 4 , @xmath77 , @xmath78 and @xmath79 , e.g see the equation ( 9 ) of the reference @xcite ) .
now , the interaction energy between the quarks can be calculated by using the landau fermi liquid model and the results derived by us in the previous section .
the potential energy density of the interacting massless up or down quarks is found by using the equations ( 3 ) and ( 7 ) , @xmath80 where @xmath81 is the quark degeneracy of one flavor . by using the fermi - dirac distribution and the value of each quark flavor potential energy i.e. the equation ( 21 ) , an analytical formula for the total potential energy density of the two quark flavours
is written as follows : @xmath82 note that , the potential energy for massless quarks is always positive .
so , the interaction between quarks inside the bag is repulsive and it helps the interacting quarks and the gluons to penetrate from the bag more easily , rather than the noninteracting case , and further more , the _
one gluon exchange _ interaction , because of its repulsive properties , makes the conditions easier for the system to make the transition to the @xmath0 phase .
the internal energy density of the @xmath0 is evaluated by performing the summation over the interacting and the non - interacting parts of the energy density of quarks , the vacuum energy and the gluon energy density which were calculated before , and having that , the other thermodynamic quantities of the @xmath0 are found .
as it was pointed out before , it is assumed that the bag ( hadronic ) pressure is the one has been used in the reference @xcite i.e. @xmath83@xmath84 ( zero ) , in order to compare our phase diagram with this reference , and since it is intended to compare our results with the reference @xcite , the @xmath2 coupling constant os chosen to be @xmath85 . on the other hand , for the none zero hadronic pressure
, the baryonic chemical potential and the bag pressure are varied to reach to the critical temperature of the lattice @xmath2 predictions @xcite , i.e. @xmath86 @xmath12 .
we begin by presenting the calculated free energy per baryon for both the @xmath0 and the strange quark matter @xcite in the figure 1 ( 2 ) , as a function of baryon density ( temperature ) at the two different temperatures ( baryon densities ) . the free energy for the @xmath0 is larger than those of strange quark matter , since we know that the strange quark matter should be more stable than the @xmath0 and therefore the strange quark matter free energy should be smaller than that of the @xmath0 . as one should expect , the free energy increases ( decreases ) by increasing the baryon density ( temperature ) . while , the @xmath0 has less temperature dependent with respect to the strange quark matter , they have similar density dependent at fixed temperature .
similar comparisons are made for the entropies in the figures 3 and 4 .
the entropy per baryon for the @xmath0 is an increasing ( decreasing ) function of temperature ( baryon density ) and it is smaller than that of the strange quark matter @xcite .
again their dependence on the density is the same , but they behave especially differently at larger temperatures . the plots of the equation of states of both the @xmath0 ( without the effect of constant bag pressure ) and the strange quark matter as a function of baryon density at the two different temperatures are given in the figure 5 .
the @xmath0 equation of state is much harder than that of strange quark matter at the same baryon density and temperature .
the pressure of weakly interacting @xmath0 for the two different @xmath2 coupling constants , and the noninteracting @xmath0 ( without the effects of constant bag pressure ) at zero temperature as a function of baryon density is shown in the figure 6 .
the increase in the interaction strength makes the pressure to rise , and therefore at the smaller baryon densities , the pressure of quarks becomes equal to the bag pressure .
so , the interaction facilitates the quarks transition to the deconfined phase at lower density .
the @xmath2 coupling constant also plays the same role , i.e it will reduce the transition density . finally , the phase diagrams for both the interacting and the noninteracting @xmath0 are shown in the figure 7 for @xmath87 .
the _ one gluon exchange _ interaction which is repulsive , causes to get the @xmath0 at the smaller baryon densities and temperatures .
as it was pointed out before , the reason is very simple , the repulsive interaction between quarks helps them to escape from the bags .
so the formation of the @xmath0 happens much easier for the interacting quarks than the noninteracting one .
but , the critical temperature is about @xmath88 , which is much less than the lattice @xmath2 suggestion of @xmath89 . in the figure 8
, the hadronic pressure has been also taken into the account ( see the equations ( [ n1 ] ) and the reference [ n2 ] ) .
the slashed area is the forbidden region , i.e. for the bag pressure approximately larger than @xmath90 there is no critical temperature with the zero baryonic chemical potential ( density ) . with the bag pressure about @xmath91 (
note that the bag pressure estimated to as large as @xmath92 @xmath12 @xcite ) and the low baryonic chemical potential ( density ) it is possible to get results near to the lattice @xmath2 prediction , i.e. @xmath86 @xmath12 .
in conclusion , the _ one gluon exchange _ interaction was used to evaluate the strength of potential energy of the @xmath0 in the fermi liquid model . by calculating the @xmath0 partition function , the different thermodynamic properties of the @xmath0 as a function of baryon density and temperature for the both interacting and the non - interacting cases were discussed .
it was found that the @xmath0 internal and free energies are much larger than those of strange quark matter . on the other side ,
if we consider the massive quarks like the strange quarks in our the @xmath0 , the potential energy becomes a negative quantity , but for the massless quarks it is always a positive quantity , therefore the internal and the free energy densities of strange quark matter become smaller than the @xmath0 ones .
we have seen how the _ one gluon exchange _ interaction for the massless quarks affects the phase diagram of the @xmath0 and causes the system to reach to the deconfined phase at the smaller baryon densities and temperatures .
our results depend on the values of bag pressure and the @xmath2 coupling constant which the latter does not have a dramatic effect on our results .
the increase of the hadronic and the bag pressure can improve our results toward the lattice @xmath2 calculations . in
the future works , we could adjust our phase diagram to get the relation between the bag constant and the @xmath2 coupling constant . on the other hand it is possible to generalize our method for the non - constant @xmath2 coupling and the bag pressure .
finally , we can also add the interaction between the gluons to our present calculations . | the thermodynamic properties of the _ quark gluon plasma _ ( @xmath0 ) as well as its phase diagram are calculated as a function of baryon density ( chemical potential ) and temperature .
the @xmath0 is assumed to be composed of the light quarks only , i.e. the up and the down quarks , which interact weakly and the gluons which are treated as they are free .
the interaction between quarks is considered in the framework of the _ one gluon exchange _ model which is obtained from the fermi liquid picture .
the bag model is used , with fixed bag pressure ( @xmath1 ) for the _ nonperturbative _ part and the _ quantum chromodynamics _ ( @xmath2 ) coupling is assumed to be constant i.e. no dependence on the temperature or the baryon density .
the effect of weakly interacting quarks on the @xmath0 phase diagram are shown and discussed .
it is demonstrated that the _ one gluon exchange _
interaction for the massless quarks has considerable effect on the @xmath0 phase diagram and it causes the system to reach to the confined phase at the smaller baryon densities and temperatures . the pressure of excluded volume hadron gas model is also used to find the transition phase - diagram .
our results depend on the values of bag pressure and the @xmath2 coupling constant which the latter does not have a dramatic effect on our calculations . finally , we compare our results with the thermodynamic properties of strange quark matter and the lattice @xmath2 prediction for the @xmath0 transition critical temperature . |
the authors would like to thank professor m.e .
grypeos for an informative discussion in atomic clusters 99 i. bialynicki - birula and j. mycielski , commun . meth .
phys . * 44 * 129 ( 1975 ) s. r. gadre , phys . rev .
a * 30 * 620 ( 1984 ) s. r. gadre , s. b. sears , s. j. chacravorty and r. d. bendale , phys .
a bf 32 2602 ( 1985 ) s. r. gadre and r. d. bendale , phys .
a * 36 * 1932 ( 1987 ) m. ohya and p.petz , `` quantum entropy and its use '' ( springer- verlag , 1993 ) a. nagy and r. g. parr , int .
j. quantum chem . * 58 * 323 ( 1996 ) v. majernic and t. opatrny , j. phys .
a : math . gen . * 29 * 2187 ( 1996 ) c. p. panos and s. e. massen , int .
j. of mod .
e * 6 * 497 ( 1997 ) g. a. lalazissis , s. e. massen , c. p. panos and s. s. dimitrova , los alamos e - print archive nucl - th/9711058 , int .
j. of mod .
e ( 1998 ) in press i.i hirschman jr . , am . j. math . * 79 * 152 ( 1957 ) w. ekardt , phys . rev .
b * 29 * 1558 ( 1984 ) h. nishioka , k. hansen and b. r. mottelson , phys .
b * 42 * 9377 ( 1990 ) b.a . kotsos and m.e .
grypeos , physica b * 229 * 173 ( 1997 ) c.b .
dover and n. van giai , nucl .
phys . a * 190 * 373 ( 1972 ) | the position and momentum space information entropies of the electron distributions of atomic clusters are calculated using a woods saxon single particle potential .
the same entropies are also calculated for nuclear distributions according to the skyrme parametrization of the nuclear mean field .
it turns out that a similar functional form @xmath0 for the entropy as function of the number of particles @xmath1 holds approximately for atoms , nuclei and atomic clusters .
it is conjectured that this is a universal property of a many - fermion system in a mean field .
it is also seen that there is an analogy of our expression for @xmath2 to boltzmann s thermodynamic entropy @xmath3 .
pacs : 89.70.+c ; 36.40.+d ; 31.10.+z ; 21.60.-n information theoretical methods have played in recent years an important role in the study of quantum mechanical systems @xcite in two cases : first in the clarification of fundamental concepts of quantum mechanics and second in the synthesis of probability densities in position and momentum spaces . in the first case
an important step was the discovery of an entropic uncertainty relation ( eur ) by bialynicki
birula and mycielski @xcite which for a three dimensional system has the form : @xmath4 ( see also ref .
@xcite for the one dimensional case ) . in ( [ ( 1 ) ] )
@xmath5 is the shannon information entropy in position
space : @xmath6 @xmath7 is the corresponding entropy in momentum - space : @xmath8 and @xmath9 , @xmath10 are the position and momentum
space density distributions respectively , which are normalized to one . however , for a normalization to the number of particles @xmath1 , the following eur holds @xcite : @xmath11 inequality ( 1 ) , for the information entropy sum in conjugate spaces , is a joint measure of uncertainty of a quantum mechanical distribution , since a highly localized @xmath9 is associated with a diffuse @xmath10 , leading to low @xmath5 and high @xmath7 and vice versa .
expression ( [ ( 1 ) ] ) is an information theoretical uncertainty relation stronger than heisenberg s @xcite .
we also note that expression ( [ ( 1 ) ] ) does not depend on the unit of length in measuring @xmath9 and @xmath10 i.e. the sum @xmath12 is invariant to uniform scaling of coordinates .
gadre @xcite derived the following approximate expression for the information entropies of electron distributions in atoms : @xmath13 using thomas - fermi theory and gadre et al @xcite derived : @xmath14 with hartee - fock calculations . here , @xmath1 is the number of electrons .
panos and massen @xcite found the following expression for nuclear distributions , employing the simple harmonic oscillator ( ho ) model of the nucleus : @xmath15 where @xmath1 is the number of nucleons in nuclei .
relations of the same functional form hold for @xmath5 and @xmath7 separately but the important quantity is @xmath16 . there is a striking similarity of ( 5 ) , ( 6 ) and ( 7 ) with the eur ( 4 ) , indicating that the functional form @xmath17 is universal for a many - fermion system in a mean field .
however , the above relations were derived for a normalization of @xmath9 and @xmath10 to the number of particles @xmath1 . in the following
we find it more convenient to normalize to one .
there is a simple relationship between the two cases and we can easily transform one case to the other according to the relations : @xmath18=\frac{s_r[norm = n]}{n}+\ln n\ ] ] @xmath19=\frac{s_k[norm = n]}{n}+\ln n\ ] ] hence , we have for normalization to one , the following expressions : @xmath20 @xmath21 @xmath22 in the present letter we extend our calculations for two other cases : the distribution of the valence electrons in atomic clusters using a woods saxon single particle potential and the nuclear distribution in nuclei employing the skyrme parametrization of the nuclear mean field . in atomic ( metallic ) clusters
the effective radial electronic potential was derived by ekardt @xcite in his spherical jellium background
model study of the self consistent charge density and the self consistent effective one particle potential , using the local density approximations .
ekardt s potentials for neutral sodium clusters were parametrized in ref .
@xcite by a woods saxon potential of the form : @xmath23 } \label{(11)}\ ] ] with @xmath24 , @xmath25 , @xmath26 and @xmath27 . for a detailed study regarding the parametrization of ekardt
s potentials see ref .
@xcite .
we solved numerically the schrdinger equation for atomic clusters with @xmath28 , @xmath29 , @xmath30 , @xmath31 , @xmath32 , @xmath33 , @xmath34 , @xmath35 valence electrons in the potential ( [ ( 11 ) ] ) and found the wave functions of the single particle states in configuration space and by fourier transform the corresponding ones in momentum space .
using the above wave functions , we calculated the electron density @xmath9 in position space and @xmath10 in momentum space , which were inserted into equations ( [ ( 2 ) ] ) , ( [ ( 3 ) ] ) and gave us the values of the information entropies @xmath5 and @xmath7 .
then we fitted the form @xmath0 to these values and obtained the expressions : @xmath36 next the nuclear densities @xmath9 and @xmath10 for several nuclei were obtained with hartee fock calculations using the skyrme parametrization of the nuclear mean field .
there are various parametrizations of the skyrme interaction , but they affect slightly the information entropies @xcite .
thus we used the skiii interaction @xcite .
finally , we fitted the form @xmath37 to the values obtained with skiii interaction and we found the expressions : @xmath38 the fit is in reasonably good agreement with its h.o .
counterpart ( comparison of relation ( [ ( 17 ) ] ) to ( [ ( 10 ) ] ) ) , though the individual entropies @xmath5 and @xmath7 do not match with the respective ho ones @xcite that well .
it seems that there is a delicate balance between the coordinate and momentum spaces , so that the interesting quantity is the sum @xmath39 ( the net information content of the system ) and not the individual entropies @xmath5 and @xmath7 . in figure 1
we plot our fitted form @xmath40 for atoms ( with hartee - fock , relation ( [ ( 9 ) ] ) , upper curve ) , atomic neutral na clusters ( relation ( [ ( 14 ) ] ) , middle curve ) and nuclei ( with skyrme , relation ( [ ( 17 ) ] ) , lower curve ) .
these lines correspond to our fitted expressions , while the corresponding values of our numerical calculations are denoted by solid circles ( clusters ) and open circles ( nuclei with skiii interaction ) .
concluding , in the present letter we derive an interesting characteristic of information entropies @xmath5 and @xmath7 for various systems i.e. atoms ( thomas fermi theory , hartee fock ) , nuclei ( harmonic oscillator model , skyrme force ) and atomic clusters(woods
saxon potential ) .
for all of these systems the entropies can be represented well by a function , which incorporates @xmath41 linearly i.e. @xmath42 where @xmath1 is the number of electrons in atoms or nucleons in nuclei or electrons in atomic clusters .
we may conjecture that this is a universal property of a many - fermion system in a mean field . as stated in ref .
@xcite , the information entropies seem to be a hidden treasure , as yet remains mostly unexplored .
a final comment seems appropriate : there is an analogy of our expression @xmath43 to boltzmann s thermodynamic entropy @xmath3 ( @xmath44 is boltzmann s constant ) .
this can be seen as follows : our relation can be written in the form : @xmath45 where @xmath46 .
the above entropy is the information entropy @xmath47 , which is related to the physical entropy @xmath48 through jaynes relation : @xmath49 . in our case
, we have : @xmath50 here @xmath51 and @xmath52 depend on the system under consideration ( atom , cluster or nucleus ) , but this dependence is not strong .
one can take that @xmath53 ( actually equals @xmath54 for atoms , 0.907 for clusters and
0.860 for nuclei ) and @xmath55 ( because @xmath56 for atoms , @xmath57 for nuclei and @xmath58 for clusters ) .
thus we obtain the following rough estimate for the entropy of a many fermion system : @xmath59 the number @xmath60 corresponds to the number of equiprobable microstates @xmath61 of boltzmann s relation .
hence , starting from shannon s information entropies of quantum mechanical distributions , we arrived at relation ( [ ( 18 ) ] ) , which can be considered as a quantum - mechanical analogue to boltzmann s entropy @xmath3 . |
magnetic moment bearing , rare earth containing , quasicrystals , being an example of well ordered solids with sharp diffraction peaks but with conventional requirement of translational symmetry lifted , present a rare example of an `` ideal '' spin glass , in which the spin glass state probably arises from the multiplicity of the r - r distances in the quasicrystalline lattice , @xcite as opposed to a substitutional disorder in crystalline metallic spin glasses .
@xcite successful growth of large , single grain , r@xmath1mg@xmath2zn@xmath3 ( r = rare earth ) icosahedral quasicrystals @xcite allowed for detailed studies of the physical properties of the spin glass state in these materials , leading , in particular , to a clear delineation of the experimental differences between heisenberg and non - heisenberg spin glasses .
@xcite it was shown that the freezing temperature , @xmath4 , is lower for the gd - based , heisenberg spin glasses , than for r = tb - er , non heisenberg spin glasses for the samples with the same values of the de gennes factor [ @xmath5 , or the weiss temperature . as a consequence , for e.g. ( tb@xmath6gd@xmath7)@xmath1mg@xmath2zn@xmath3 and ( dy@xmath6gd@xmath7)@xmath1mg@xmath2zn@xmath3 pseudo - ternary solid solution
the maximum in @xmath4 was observed for @xmath8 when a crossover from from heisenberg to non - heisenberg behavior occurs .
@xcite based on the large data sets in refs .
@xcite it was suggested that two factors give rise to the spin glass state in the r@xmath1mg@xmath2zn@xmath3 quasicrystals : distribution of r - r distances and distribution of easy axis ( or easy plane ) in the non - heisenberg members of the family . since (
i ) a heisenberg quasicrystal gd@xmath1mg@xmath2zn@xmath3 has a spin glass low temperature state , @xcite and ( ii ) an attempt to design spin glass just by mixing rare earths with different anisotropies in a crystalline structure has failed so far , @xcite the distribution of the r - r distances is apparently a _ necessary _ and _ sufficient _ condition for the formation of a spin glass , and distribution of magnetic anisotropies is neither . the frequency dependence of the freezing temperature ( the position of the peak in ac susceptibility ) has been observed and discussed in a number of spin glass systems . @xcite the fractional relative change in the freezing temperature per decade of frequency , @xmath9 $ ] ( @xmath10 is frequency ) was noticed to vary by more than an order of magnitude for different spin glasses .
@xcite it was suggested @xcite that the experimental data could be well described by the empirical vogel - fulcher law , @xmath11\ ] ] where @xmath12 is the boltzmann constant , and @xmath13 , @xmath14 , and @xmath15 are the fitting parameters . from the very beginning , though , it was understood that the usual set of @xmath16 experimental data would not be suitable for obtaining all three parameters , and the value of @xmath17 was either obtained from other measurements or estimated by some independent procedure ( e.g. remnant magnetization measurements @xcite were used to determine @xmath17 in @xcite ) .
r@xmath1mg@xmath2zn@xmath3 quasicrystals , with magnetic moment bearing rare earths give us an opportunity to address trends in the frequency dependence of the freezing temperature in the family of spin glasses with heisenberg and non - heisenberg members , a task that is rather difficult to undertake in dilute , substitutional , crystalline spin glasses . to the best of out knowledge frequency dependent measurements in this family were performed and analyzed so far only for tb@xmath1mg@xmath2zn@xmath3 .
large single grain r@xmath1mg@xmath2zn@xmath3 quasicrystals were grown from ternary or pseudo - ternary melt as described in detail in refs . @xcite .
the actual samples used in this work were taken from the batches extensively studied in the past .
@xcite the samples with r = tb , dy , gd , gd@xmath18y@xmath19 , gd@xmath20y@xmath21 , and gd@xmath22tb@xmath22 , were chosen for this work thus covering heisenberg , diluted ( disordered ) heisenberg , and non - heisenberg spin glasses .
low temperature , low field ( 25 oe ) zero field cooled - warming and field cooled - warming dc susceptibility measurements were done in a quantum design mpms-7 squid magnetometer so as to ensure that sample chosen indeed have a low temperature spin glass state and are not rhombohedral approximants with a long range magnetic order . @xcite .
the low temperature ac susceptibility was measured in 3 - 5 oe ac field at frequencies in the range of 10 hz to 10 khz , in zero dc field using the acms option of quantum design ppms-14 instrument .
a criterion @xmath23 ( @xmath24 is the real part of the ac susceptibility ) was used to infer a value for @xmath4 .
examples of low temperature ac susceptibility measurements are shown in fig .
the maximum in @xmath24 clearly shifts to higher temperatures at higher frequencies for the non - heisenberg spin glasses tb@xmath1mg@xmath2zn@xmath3 and dy@xmath1mg@xmath2zn@xmath3 and is almost unchanged for the heisenberg spin glass gd@xmath1mg@xmath2zn@xmath3 . fig .
[ f2 ] summarizes all of such measurements performed in this work : whereas with three orders of magnitude frequency change , the @xmath4 of non - heisenberg spin glasses ( r = tb , dy ) increases by about 20% , the @xmath4 of the heisenberg spin glass ( r = gd ) increases by mere @xmath25% .
if substitutional disorder is added to the heisenberg spin glass ( r = gd@xmath18y@xmath19 , gd@xmath20y@xmath21 ) the change is slightly larger , but still very much below those for the non - heisenberg systems . for a mixture of heisenberg and non - heisenberg magnetic moments , r = gd@xmath22tb@xmath22 ,
the relative change in the @xmath4 is in between the extreme cases but actually quite close to the r = tb data . the fractional relative change in the freezing temperature per decade of frequency , @xmath9 $ ] results are summarized in the table [ t1 ] .
the values are within the broad range reported for different spin glasses .
@xcite [ t1 ] the volgel - fulcher law can be used to fit the @xmath4 - frequency data .
the fits are rather insensitive to the value of @xmath17 . here , for all samples , we have used @xmath26 hz obtained for tb@xmath1mg@xmath2zn@xmath3 @xcite , although the values of @xmath17 as high as @xmath27 hz give a comparable quality fits .
fits for r = tb , dy , and gd are shown in fig .
[ f3 ] as an example .
the parameters of the fits for the ternary and pseudo - ternary quasicrystals in this work are listed in the table [ t1 ] . for the non - heisenberg
spin glasses , @xmath28 , as it was reported for a number of spin glasses .
@xcite for the heisenberg and disordered heisenberg spin glasses , @xmath29 , and the values are closer to each other .
the significance of this difference is not clear at this point .
it might just point out to some limitations of the vogel - fulcher fits for a small span of the freezing temperatures and a fixed value of @xmath17 throughout the family .
the systematic study of the frequency dependence of the spin glass freezing temperatures in the ternary and pseudo - ternary r@xmath1mg@xmath2zn@xmath3 quasicrystals revealed a distinct difference between non - heisenberg and heisenberg members of the family , the latter showing significantly weaker response to the measurement frequency change ( at least in the studied 10 hz to 10 khz range ) .
it appears that the distribution of magnetic anisotropies ( easy axis or plane ) in the non - heisenberg members changes the ( low frequency ) magnetic dynamics of these spin glasses .
similar trend was suggested ( based on a very limited experimental data set ) in crystalline rb@xmath30 spin glasses .
@xcite it would be of interest to see if such distinct behavior is observed in other families of magnetic rare earth containing spin glasses , and if so , what theoretical models can be used or developed to account for such behavior .
help of i.r .
fisher , a.f .
panchula and k.o .
cheon in samples synthesis and decadal room temperature annealing sudies is greatly appreciated .
work at the ames laboratory was supported by the department of energy , basic energy sciences , division of materials sciences and engineering under contract no .
de - ac02 - 07ch11358 .
s.l.b . acknowledges partial support from the state of iowa through iowa state university . | we present ac susceptibility measurements with the frequency spanning three orders of magnitude on single grain , icosahedral r - mg - zn ( r = rare earth ) quasicrystals .
the freezing temperature in gd - based , heisenberg spin glasses in this family increases by @xmath0 with a frequency increase from 10 hz to 10 khz , whereas the freezing temperature in the non - heisenberg members of the family is significantly more responsive to the frequency change ( by 16 - 22 % ) , suggesting that an additional magnetic anisotropy distribution in the non - heisenberg spin glasses causes changes in the low frequency magnetic dynamics . spin glass ; quasicrystals ; freezing temperature ; frequency dependence |
a variety of related models of the nucleon emff @xcite were fitted to the complete set of data available before september 2001 .
one group of models included variants of the basic gk model of @xmath0 , @xmath1 , and @xmath2 vector meson pole terms with hadronic form factors and a term with pqcd behavior which dominates at high @xmath9 @xcite .
four varieties of hadronic form factor parameterization ( of which two are used in @xcite ) were compared .
in addition to the gk type models we considered a group of models ( generically designated dr - gk ) that use the analytic approximation of @xcite to the dispersion integral approximation for the @xmath0 meson contribution ( similar to that of @xcite ) , modified by the four hadronic form factor choices used with the gk model , and the addition of the well established @xmath3 ( 1450 ) pole . every model had an electric and a magnetic coupling parameter for each of the three pole terms , four cut - off " masses for the hadronic form - factors and the qcd scale mass scale , @xmath11 for the logarithmic momentum transfer behavior in pqcd .
in addition the effect of a normalization parameter was sometimes considered for the dispersion relation behavior of the @xmath0 meson in the dr - gk models . when the set of parameters in each of the eight models was fitted to the full set of data available before publication , for @xmath6 , @xmath12 , @xmath7 , @xmath13 and the lower @xmath9 values of @xmath14 , three gk and all four dr - gk models attained reasonable @xmath15 ( when the inconsistency of some low @xmath9 @xmath7 and @xmath13 data was taken into account ) , but the extended dr - gk models had significantly lower @xmath15 .
furthermore @xmath11 was reasonable for three of the dr - gk models but for only the one of the gk models that had an unreasonably large anomalous magnetic coupling @xmath16 .
it was concluded that the three dr - gk models were the best nucleon emff to use in prediction of nuclear electromagnetic properties .
all thee were found to be moderately consistent in their predictions up to @xmath9 of 8 gev@xmath10 . however the part of the above data set from a recent @xmath4 ratio data @xcite for @xmath17 gev@xmath18 gev@xmath19 , swamped statistically by all the other data , was systematically lower than the fitted models ( fig.5 of @xcite ) contributing disproportionately to @xmath15 .
this ratio is determined by an asymmetry measurement in the scattering of polarized electrons on protons .
multiplied by the well determined values of @xmath12 one obtains values for @xmath6 which are not subject to the uncertainty inherent in the rosenbluth separation measurements in which @xmath6 is obtained by subtracting the much larger contribution of @xmath12 from the unpolarized cross section .
as expected the @xmath6 derived from the measured @xmath4 are consistently below those of the older rosenbluth separation values .
it is plausible to expect that the old @xmath6 data is responsible for restricting the best fit of the models to be substantially above the experimental @xmath4 values . with this in mind
the particularly high data of @xcite was omitted from the fit to the model type dr - gk@xmath20(1 ) of @xcite and the flexibility of a @xmath0 meson dispersion integral normalization parameter n was included . in this article the original version
is designated as gkex(01 ) and when fitted to the smaller data set as gkex(01- ) . as seen in tables [ t1 ] and [ t2 ] and figs .
[ elfig1 ] and [ elfig2 ] , there is only a small change in the fit to @xmath6 and @xmath4 , although the parameters of the fit change substantially .
@xmath21 @xmath22 after the publication of @xcite new data @xcite extended the measurements of @xmath4 up to @xmath23 gev@xmath10 , exacerbating the discrepancy with the predictions of the best models in @xcite .
very recently @xmath24 has been obtained directly @xcite by the scattering of polarized electrons on deuterium and detecting the polarized recoil neutron at @xmath9 = 0.45 , 1.15 and 1.47 gev@xmath10 .
the preliminary results are consistent with the galster @xcite parameterization from lower @xmath9 data @xmath25 which , in parallel to the situation for @xmath4 , implies much lower values of @xmath7 in their @xmath9 range when coupled with @xmath13 values ( either the precision data of @xcite or the model fits ) . in this paper , in addition to the above comparison of gkex(01 ) and gkex(01- ) , we fit the model of type dr - gk@xmath20(1 ) , with the added isoscalar vector meson @xmath8(1419 ) pole , to the following data sets , chosen to determine the effect of the old @xmath7 and @xmath6 data in direct conflict with the values of @xmath5 and @xmath4 from modern polarization measurements : 1 . the fit gkex(02l ) from the full data set of @xcite with the addition of @xcite and @xcite , the omission of @xcite ( as above for gkex(01- ) ) and the @xmath7 values for @xmath26 gev@xmath10 of @xcite , @xcite , and @xcite .
the fit of gkex(02s ) to the same data set as above except for the omission of the @xmath6 values for @xmath27 gev@xmath10 of @xcite .
it will be seen that the omission of the conflicting @xmath7 data , gkex(02l ) , has a much bigger influence than the omission of @xcite , gkex(01- ) , enabling a much better fit to @xmath4 in addition to a very good fit to @xmath5 , compared to gkex(01 ) . with the removal of the conflicting @xmath6 data , gkex(02s ) ,
the fit to all the remaining data , including @xmath4 , is very satisfactory . in section [ s:2 ]
we will specify the models and parameters used in this article , and the data sets used in section [ s:3 ] . in section [ s:4 ] we present the results of the four gkex fits in comparison with each other .
we extrapolate beyond the present experimental range of momentum transfer where necessary for predicting available deuteron emff data .
the model gkex(02s ) fits the modern and consistent older data well and meets the requirements of dispersion relations and of qcd at low and high momentum transfer .
conclusions are presented in section [ s:5 ] .
in fitting the nucleon emff data including the new @xmath5 and @xmath4 results we have chosen to use the extended gk model dr - gk@xmath20(1 ) of ref .
@xcite with the addition of a pole term for the well established isoscalar vector meson @xmath8(1419 ) , whose mass is lower than that of the already included isovector vector meson @xmath3(1450 ) .
we denote this model as gkex . the choice of the particular parameterization dr - gk@xmath20(1 )
was made because of its low @xmath15 value and the fact that its predicted values of @xmath4 were a little closer to the data than those of the other extended models .
in addition dr - gk@xmath20(1 ) has the following good physical properties : 1 .
it uses the qcd cut - off @xmath28 for the helicity flip meson - nucleon form factors , rather than the meson cut - off @xmath29 used by dr - gk(3 ) and dr - gk@xmath20(3 ) .
the evolution of the logarithmic dependence on @xmath9 is controlled by the quark - nucleon cut - off @xmath30 , along with @xmath11 .
dr - gk(1 ) and dr - gk(3 ) use @xmath28 instead of @xmath30 .
3 . fitted to the data
set of @xcite it finds @xmath11 = .1163 , close to the expected value .
the form factors are not very sensitive to this parameter which is fixed at .15 for the fits to the new data sets . so that the reader need not make constant reference to @xcite we repeat the relevant formulas here together with the new @xmath8(1419 ) terms .
the emff of a nucleon are defined by the matrix elements of the electromagnetic current @xmath31 @xmath32 where @xmath33 is the neutron , @xmath34 , or proton , @xmath35 , and @xmath36 is the square of the invariant momentum transfer . @xmath37 and @xmath38 are respectively the dirac and pauli form factors , normalized at @xmath39 as @xmath40 the sachs form factors , most directly obtained from experiment , are then @xmath41 expressed in terms of the isoscalar and isovector electromagnetic currents @xmath42 the gkex model has the following form for the four isotopic emff : @xmath43 where the pole terms are those of the @xmath0 , @xmath3 , @xmath1 , @xmath8 , and @xmath2 mesons , and the final term of each equation is determined by the asymptotic properties of pqcd .
the @xmath44 , @xmath45 , @xmath1 , or @xmath2 are the meson - nucleon form factors , while the @xmath46 are effectively quark - nucleon form factors . for gkex
the above hadronic form factors are parameterized in the following way : @xmath47 \noalign{\mbox{where $ \alpha=\rho,\omega$ and $ \lambda_{1,d}$ is $ \lambda_{1}$ for $ f^\alpha_i$ , $ \lambda_d$ for $ f^d_i$,}\medskip }
f^\phi_1(q^2 ) & = & f^\alpha_1\bigl(\frac{q^2}{\lambda^2_1 + q^2 } \bigr)^{1.5}\ , \quad f^\phi_1(0 ) = 0\nonumber\\
f^\phi_2(q^2 ) & = & f^\alpha_2\bigl(\frac{\lambda^2_1}{\mu^2_\phi}\frac{q^2+\mu^2_\phi}{\lambda^2_1 + q^2 } \bigr)^{1.5 } \label{e7 } \\
\mbox{with } \tilde q^2 & = & q^2\frac{\ln\bigl[(\lambda^2_d + q^2)/\lambda^2_{\mathrm{qcd}}\bigr]}{\ln(\lambda^2_d/\lambda^2_{\mathrm{qcd}})}\ . \nonumber\end{aligned}\ ] ] this parameterization , together with eq .
( [ e6 ] ) , guarantees that the normalization conditions of eq .
( [ e2 ] ) are met and that asymptotically @xmath48^{-2}\nonumber\\ f^i_2 & \sim f^i_1/q^2 \label{e8}\\ i & = is , iv \nonumber\end{aligned}\ ] ] as required by pqcd .
the form factor @xmath49 vanishes at @xmath50 , and it and @xmath51 decrease more rapidly at large @xmath9 than the other meson form factors .
this conforms to the zweig rule imposed by the @xmath52 structure of the @xmath2 meson @xcite .
this model has at most 14 free parameters : 1 . 8 couplings to the pole terms , the 4 @xmath53 and the 4 @xmath54 for the @xmath3 , @xmath1 , @xmath8 and @xmath2 mesons .
2 . 4 cut - off masses in the hadronic form factors , @xmath29 , @xmath28 , @xmath30 , and @xmath55 .
3 . the mass determing the size of the logarithmic @xmath9 behavior , @xmath11 .
the normalization factor
n for the dispersion relation contribution of the @xmath0 meson . however at most 12 of these parameters are freely varied in any of the fits made in the next section to the chosen data sets .
as previously stated , gkex(01 ) is the same as dr - gk@xmath20(1 ) of @xcite .
this model had the best fit to the full data set available at the publication of @xcite with @xmath56 = @xmath57 = 0 and with n=1 . for gkex(01- )
the 4 data points of @xcite were omitted from that data set . in this case @xmath56 and @xmath57 were still supressed but n was freely varied .
in the fits gkex(02l ) and gkex(02s ) @xmath56 and @xmath57 were freely varied , but these fits fixed n=1 again ( implying negligible error in the dispersion relation evaluation ) and @xmath11 was fixed at the physical value of 0.15 gev@xmath58 . the important difference from the data set of gkex(01- )
is the addition of the higher q @xmath4 data points of @xcite and the @xmath5 data points of @xcite and the omission of the @xmath7 values for @xmath26 gev@xmath10 of @xcite , @xcite and @xcite . in the shorter data
set of gkex(02s ) the @xmath6 values for @xmath27 gev@xmath10 of @xcite are also omitted .
the free parameters were optimized using a mathematica program that incorporates the levenberg - marquardt method . [ t1 ]
.model parameters .
common to all models are @xmath59 , @xmath60 , @xmath61 gev , @xmath62 gev , @xmath63 gev , @xmath64 gev and @xmath65 gev . [ cols="^,^,^,^,^",options="header " , ]
table [ t1 ] presents the parameters which minimize @xmath15 for the above 4 cases . for all 4 parameter sets the hadronic form factor cut - off masses , @xmath29 , @xmath28 , @xmath30 , and @xmath55 .
are reasonable .
the relatively large value of @xmath28 , which controls the spin - flip suppression in qcd , is consistent with the slow approach to asymptopia observed in polarized hadron scattering .
for the two cases in which @xmath66 is a fitted parameter , as well as the two for which it is fixed , it is consistent with high energy experiment .
the addition of the @xmath8(1.419 ) meson in gkex(02l ) and gkex(02s ) has moved @xmath67 closer to the expected small negative value than all earlier fits , but there is still the implication of some effect from a higher mass isoscalar meson .
the adequacy of the fits is an indication that the form factors with more poles would be similar to those already obtained . in table [ t2 ]
the values of @xmath15 are listed for the 4 cases and the contribution from each of the six form factor classes of measurement are detailed .
also shown are the values of @xmath15 when any data points omitted from the fit are re - inserted .
@xmath68 @xmath69 we note , as can also be seen in figs .
[ elfig3 ] and [ elfig4 ] that the quality of fit to the magnetic form factors , @xmath12 and @xmath13 changes negligibly as we refit to the datas sets that differ in the electric form factors and the electric to magnetic form factor ratios . as discussed in @xcite , the large excess of @xmath15 over the number of data points for @xmath13 is due to obvious inconsistencies in the data set for @xmath13 at @xmath70 .
the displacement of adjacent data points well beyond their error bars in this range is evident in the figures and contributes about 90 to the @xmath15 of @xmath13 .
the interesting changes are , of course , in the fits to @xmath6 , @xmath7 , @xmath4 and @xmath5 . as noted in the introduction ,
removing the 4 very high values of @xmath6 data @xcite does surprisingly little to allow a better fit to the @xmath4 data already in gkex(01 ) .
several of the parameters , all three @xmath54 , @xmath28 and @xmath30 , have large changes ( see table [ t1 ] ) , but this results in a small shift between the predictions of gkex(01 and gkex(01- ) as is evident in table [ t2 ] and figs . [ elfig1 ] and [ elfig2 ] .
the figures show a slightly better fit to @xmath4 correlated with a very slightly worse fit to @xmath6 .
the former is reflected in table [ t2 ] by the decrease in the @xmath15 contribution of the 17 @xmath4 points to which those cases were optimized from 29.0 t0 22.6 .
when the 4 higher @xmath9 of @xcite are added the @xmath15 contribution is much larger than the number of points ( 21 ) .
the drop in the @xmath15 contribution to @xmath6 from 67.2 to 48.2 is entirely due to the omission of the 4 data points of @xcite , but the @xmath15 for the full set of 48 points is a little larger because of the competition with @xmath4 .
the implication is that there is a constraint on the fit to @xmath4 from data independent of @xmath6 , arising from the model correlations between all the nucleon emff .
this is shown to be the case below . substituting the new @xmath5 values for the conflicting @xmath7 data of @xcite , @xcite and @xcite
causes a large difference between the gkex(02l ) and gkex(01- ) fits to @xmath6 , @xmath7 , @xmath4 and @xmath5 , as seen in table [ t2 ] and fig .
[ elfig5 ] - fig .
[ elfig8 ] . in particular for gkex(02l )
the @xmath15 contribution for all 21 @xmath4 data points is 23.1 .
[ elfig6 ] shows the strong improvement in the fit to @xmath4 .
the figure also shows that the goodness of the @xmath15 value is somewhat misleading because that fit is systematically high for the 3 highest @xmath9 data points . on the other hand the @xmath15 contribution for all 44 @xmath6 data points increases from 48.2 in gkex(01- ) to 75.3 in gkex(02l ) because of the compromise of better fitting @xmath4 .
the @xmath15 contribution for the 3 @xmath5 points now included is only 0.6 .
@xmath7 now contributes 24.1 for the remaining 15 data points ( which still include highly scattered low @xmath9 data as discussed in @xcite ) instead of 64.8 for the 23 data points in gkex(01- ) . for the gkex(02s )
case the @xmath6 data of @xcite , which is clearly inconsistent with the new @xmath4 data @xcite and @xcite , is also omitted .
the results are very good if the modern data is chosen when in conflict with the older rosenbluth separation results .
the @xmath15 contribution from the remaining 36 @xmath6 points is only 30.6 and for all 21 @xmath4 points only11.8 . for the remaining types of form factor measurements
there is a negligible change of @xmath15 between the gkex(02l ) and gkex(02s ) cases .
@xmath71 figs .
[ elfig5 ] , [ elfig6 ] , [ elfig7 ] , and [ elfig8 ] show the successive improvements in @xmath6 , @xmath4 , @xmath7 , and @xmath5 as the optimization data sets are varied from gkex(01 ) to gkex(02l ) and to gkex(02s ) . to demonsrate the correlation between the electric form factors and the ratio of electric to magnetic form factors we have , in figs .
[ elfig5 ] and [ elfig7 ] , entered ( as circles ) the electric form factor values obtained by multiplying the experimental @xmath4 and @xmath5 values by the case gkex(02s ) model values of the magnetic form factors normalized by the magnetic moments .
the correlation with the model prediction for the electric form factors is excellent .
@xmath72 @xmath73 @xmath74 the figures show the model extrapolations of @xmath4 , @xmath13 , @xmath7 and @xmath5 up to @xmath9 of 8 gev@xmath10 for the guidance of future experiments and because of their relevance to deuteron and other nuclear electromagnetic scattering predictions .
the extrapolation is sensitive to the weight given to the polarized vs. the rosenbluth separation data in the fits .
the resolution of this dichotomy will , in the context of the physical model employed here , greatly restrict the nucleon emff over a large range of momentum transfer .
the polarization measurements of @xmath4 and @xmath5 may soon be extended to larger @xmath9 , so it is of interest to examine the predictions of the good fit gkex(02s ) as @xmath9 increases . as seen in fig .
[ elfig5 ] the model curve is , as is the data , approximately linear in the range @xmath75 gev@xmath76 gev@xmath10 , but the model curve s slope is gradually decreasing in manitude . a linear fit to the data would change sign near @xmath77 gev@xmath10 where the model predicts 0.19 .
the model crosses zero near @xmath78 gev@xmath10 with a very small slope .
[ elfig8 ] shows the galster curve , @xmath79 of eq .
( [ e1 ] ) , to compare with the model and the data .
the model fits the data but deviates from the galster curve after that .
the model increases faster , reaching a maximum of 0.426 at @xmath80 gev@xmath10 where the galster value is only 0.309 , while @xmath79 increases monotonically to an asymptotic value of 0.342 .
a measurement of the present quality at @xmath80 gev@xmath10 could distinguish between the two .
the gkex model , consistent with vector meson dominance and perturbative qcd in the appropriate momentum transfer regions , represents well a consistent set of neutron and proton emff .
this set includes polarization measurements , which are directly related to the ratios of electric to magnetic form factors , and differential cross section measurements of the magnetic form factors .
the values of the electric form factors from the rosenbluth separation of the differential cross section is , in our final selection gkex(02s ) , only used for the lower range of @xmath9 where the magnetic contributions are less dominant . because of the physical properties of the model and the good quality of the fit we expect that the model predictions are sufficiently accurate to be used for predictions of the electromagnetic properties of nuclei .
the model values may also be useful in planning future experiments .
the above conclusions are only valid to the extent that adequate physics is included in the gkex models .
only the @xmath0 meson exchange includes the width of the vector mesons ( from dispersion relation results ) .
there will be corrections from the widths of the other exchanged vector mesons .
however the next most important , the @xmath1 and @xmath2 , have very narrow widths .
the higher masses of the @xmath3 ( 1450 ) and the @xmath8 ( 1420 ) reduces the importance of their substantial widths because of their distance from the physical region and their partial replacement by the pqcd term . in assuming vector dominance we have neglected the multi - meson exchange continuum contributions . in particular the two - pion continuum may have an influence at very low @xmath81 gev@xmath10 .
indeed , as remarked in@xcite and can be seen in fig .
5b , the @xmath6 data of @xcite has a more negative slope for @xmath82 gev@xmath10 than the higher @xmath9 data and the model fit .
the addition of a 2-pion exhange term to the model may enable a change of slope between the two regions , but would have little effect on the model fit for @xmath83 gev@xmath10 .
one may also want to consider some higher mass vector mesons .
this would have some importance in the fits of @xcite and@xcite , but are much less important in these gk type models because of the transition to pqcd behavior .
earle l. lomon , phys . rev .
* c64 , * 035204 ( 2001 ) .
gari and w. krmpelmann , phys
* b274 , * 159 ( 1992 ) ; erratum , phys . lett .
* b282 , * 483 ( 1992 ) .
p. mergell , ulf - g .
meissner and d. drechsel , nucl . phys .
* a596 , * 367 ( 1996 ) .
g. h@xmath84hler et al . , nucl . phys .
* b114 , * 505 ( 1976 ) .
jones et al . , phys .
* 84 , * 1398 ( 2000 ) .
walker et al . , phys . rev .
* d49 , * 5671 ( 1994 ) .
l. andivahis et al .
, phys . rev .
* d50 , * 5491 ( 1994 ) .
f. borkowski et al . , nucl .
* b93 , * 461 ( 1975 ) .
hanson et al . , phys . rev .
* d8 , * 753 ( 1973 ) .
murphy et al . , phys . rev .
* c9 , * 2125 ( 1974 ) .
berger et al . , phys .
* 35b , * 87 ( 1971 ) .
w. bartel et al . , nucl .
* b58 , * 429 ( 1973 ) .
o. gayou et al .
* 88 , * 092301 ( 2002 ) .
richard madey , private communication ; bull .
aps*46 , * db 10 , p 34 , oct . 2001 .
s. galster et al . , nucl . phys .
* b32 , * 221 ( 1971 )
. h. anklin et al .
, phys . lett .
* b428 , * 248 ( 1998 ) .
a. lung et al .
* 70 , * 718 ( 1993 ) .
w. bartel et al . , phys . lett .
* b39 , * 407 ( 1972 ) .
a.f . sill et al . , phys . rev .
* d48 , * 29 ( 1993 ) .
bosted et al .
* c42 , * 38 ( 1990 ) .
h. anklin et al . , phys . lett .
* b336 , * 313 ( 1994 ) .
e.e.w.bruins et al . ,
* 75 , * 21 ( 1995 ) .
h. gao et al . ,
phys . rev .
* c50 , * r546 ( 1994 ) .
p. markowitz et al . , phys . rev .
* c48 , * r5 ( 1993 ) .
esaulov et al . ,
* 45 , * 410 ( 1987 ) [ sov .
* 45 , * 258 ( 1987 ) ] .
w. bartel et al .
, phys . lett .
* 30b , * 285 ( 1969 ) .
w. xu et al . ,
rev . lett .
* 85 , * 2900 ( 2000 ) .
j. golak et al . , phys . rev .
* c63 , * 1034006 ( 2001 ) . c. herberg et al .
j. * a5 , * 131 ( 1999 ) .
m. ostrick et al .
, phys . rev .
* 83 , * 276 ( 1999 ) .
i. passchier et al . , phys .
* 82 , * 4988 ( 1999 ) .
d. rohe et al . ,
* 83 , * 4257 ( 1999 ) .
t. eden et al . , phys . rev .
* c50 , * r1749 ( 1994 ) .
m. meyerhoff et al . , phys .
* b327 , * 201 ( 1994 ) .
jones - woodward et al . , phys . rev .
* c44 , * r571 ( 1999 ) .
h. zhu et al .
* 87 , * 081801 ( 2001 ) .
t. walcher ( private communication ) .
s. kopecky et al . , phys .
* 74 , * 2427 ( 1995 ) . | the gari - krmpelmann ( gk ) models of nucleon electromagnetic form factors , in which the @xmath0 , @xmath1 , and @xmath2 vector meson pole contributions evolve at high momentum transfer to conform to the predictions of perturbative qcd ( pqcd ) , was recently extended to include the width of the @xmath0 meson by substituting the result of dispersion relations for the pole and the addition of the @xmath3 ( 1450 ) isovector vector meson pole .
this extended model was shown to produce a good overall fit to all the available nucleon electromagnetic form factor ( emff ) data .
since then new polarization data shows that the electric to magnetic ratios @xmath4 and @xmath5 obtained are not consistent with the older @xmath6 and @xmath7 data in their range of momentum transfer . the model is further extended to include the @xmath8 ( 1419 ) isoscalar vector meson pole .
it is found that while this gkex can not simultaneously fit the new @xmath4 and the old @xmath7 data , it can fit the new @xmath4 and @xmath5 well simultaneously .
an excellent fit to all the remaining data is obtained when the inconsistent @xmath6 and @xmath7 is omitted .
the model predictions are extended beyond the data , if needed , to momentum transfer squared , @xmath9 , of 8 gev@xmath10 . |
for dimensions @xmath2 the weyl tensor vanishes identically and for @xmath3 it has very special properties .
it is of interest to determine which of the properties of four - dimensional ( 4d ) spacetimes can be straightforwardly generalized to higher dimensions and which need to be modified or do not hold at all .
recently a classification of algebraic tensor types in lorentzian manifolds of arbitrary dimension was developed @xcite .
for the weyl tensor in 4d this classification reproduces the petrov classification and for the ricci tensor in 4d the segre classification . in 4d
it follows from the bianchi identities . in the first case
@xmath4 acts on three indices and stands for @xmath5 . in the other case
@xmath4 acts on four indices and is given by ( [ zavorka ] ) .
] r_ab\{cd;e}=0 [ bia ] that in algebraically special vacuum spacetimes the multiple principal null direction of the weyl tensor is geodesic and shearfree . in this paper
a higher dimensional frame formalism is developed in order to study implications of the bianchi identities , which are given in [ appbianchi ] , for vacuum spacetimes of algebraic types n and iii in higher dimensions .
although in most applications it is necessary to perform calculations in a given spacetime dimension @xmath0 , in this paper we present results without specifying the dimension and hence these results are valid in any dimension . in 4d , for algebraically special vacuum spacetimes some of the tetrad components of the bianchi identities in the newman - penrose formalism @xcite lead to simple algebraic equations ( i.e. equations with no derivatives ) . in higher dimensions
these algebraic equations are much more complex and the number of independent equations , as well as the number of unknowns , depend on the dimension of the spacetime .
we show that in vacuum type iii and n spacetimes of arbitrary dimension the multiple principal null direction ( pnd ) is geodesic . for type n spacetimes ( sec .
[ typen ] ) the symmetric expansion matrix @xmath6 has just one non - vanishing doubly degenerate eigenvalue .
consequently , the principal geodesic null congruence expands isotropically in two dimensions and does not expand in @xmath7 dimensions .
thus , shear does not vanish for @xmath8 .
the antisymmetric twist matrix @xmath9 has only one independent component .
all other components may be set to zero by appropriately choosing the frame . for type iii ,
we prove similar results in the generic case ( sec .
[ piii ] ) and in the non - twisting case ( [ appiiinon ] ) .
the complete proof , including all possible degenerate cases , is presented for five - dimensional spacetimes ( [ app5d ] ) . in section [ goldberg ]
, we show that for a vacuum spacetime the properties of @xmath6 and @xmath9 matrices mentioned above imply that the spacetime is algebraically special .
we also show that an arbitrary vacuum spacetime admitting a non - expanding and non - twisting geodesic null congruence ( i.e. , a higher dimensional generalization of the vacuum kundt class ) is algebraically special .
these statements can not be regarded as a generalization of the goldberg
sachs theorem for higher dimensions ( see sec .
[ goldberg ] for details ) since in [ typed ] it is shown that the myers perry metric , which is of the algebraic type d , has the expansion matrix @xmath6 with one doubly degenerate eigenvalue and also another non - vanishing eigenvalue . in section [ secdisc ] ,
we conclude with discussion of potential applications .
the newman penrose formalism @xcite in 4d is based on using a null tetrad @xmath10 and @xmath11 , where @xmath12 and @xmath13 are real null vectors and @xmath14 and @xmath15 are complex null vectors instead of an orthonormal basis , thus taking advantage of the null cone structure of spacetimes . for @xmath0-dimensional calculations it seems to be more practical to choose a pair of null vectors @xmath16 and an orthonormal set of real spacelike vectors @xmath17 .
we thus need two types of indices : indices @xmath18 with values @xmath19 and indices @xmath20 going from 2 to @xmath21 .
we will observe einstein s summation convention for both of these types of indices .
however , for indices @xmath22 , there is no difference between covariant and contravariant components an thus we will not distinguish between subscripts and superscripts .
the frame ^(0)= , ^(1)= , ^(i ) [ tetrad ] thus satisfies ^a _
a= n^a n_a = ^a m^(i)_a = n^a m^(i)_a= 0 , ^a n_a = 1 , m^(i)am^(j)_a=_ij , and the metric has the form g_a b = 2_(an_b ) + _ ij m^(i)_a m^(j)_b .
[ metric ] if one would like to completely generalize the newman
penrose formalism for higher dimensions it would be necessary to denote ricci rotation - coefficients and all independent components of the riemann , weyl and ricci tensors and then rewrite all frame components of the bianchi identities ( [ bia ] ) and the ricci identities v_a;bc = v_a;cb+r^s _ abc v_s , where @xmath23 is an arbitrary vector , as well as expressions for commutators @xcite of covariant derivatives in directions of the frame vectors d ^a _ a , n^a _ a , _
i m^(i)a _ a .
however , in this paper we are only interested in studying consequences of the bianchi identities , which are given in detail in [ appbianchi ] , and thus we do not introduce the ricci identities except for equation ( [ riccilmm ] ) given in section [ goldberg ] . in order to construct a basis in the space of 4-rank tensors with symmetries r_abcd= ( r_[ab ] [ cd ] + r_[cd ] [ ab ] ) , [ rsym1 ] we introduce the operation \ { } w_\{a x_b y_c z_d } ( w_[a x_b ] y_[c z_d]+ w_[c x_d ] y_[a z_b ] ) .
[ zavorka ] when decomposing the riemann tensor in terms of the frame vectors we also have to take into account that r_a\{bcd}=0 .
[ rsym2 ] now let us decompose the riemann tensor in its frame components and sort them by their boost weights ( see @xcite ) : r_abcd = ^2 + + ^1 + l + 4 r_0101 n^_\{a ^ _
b n^ _ c ^ _ d } + 4 r_01ij n^_\{a ^ _
c m^(j ) _ d } + + 8 r_0i1j n^_\{a m^(i ) _ b
^_c m^(j ) _ d } + r_ijkl m^(i)_\{a m^(j ) _ b m^(k ) _ c m^(l ) _ d } } ^0 [ eq : rscalars ] + + ^-1 + + ^-2 .
the riemann frame components in this relation are subject to constraints following from ( [ rsym1 ] ) and ( [ rsym2 ] ) , & & r_0[i|0|j ] = 0 , + & & r_0i(jk)=r_0\{ijk } = 0 , + & & r_ijkl = r_\{ijkl},r_i\{jkl}=0 , r_01ij= 2 r_0[i|1|j ] , [ eq : rcomps ] + & & r_1i(jk)=r_1\{ijk } = 0 , + & & r_1[i|1|j ] = 0 .
let us check that we have an appropriate number of independent frame components .
it is well known that an @xmath0-dimensional riemann tensor has @xmath24 independent components . counting independent frame components of various boost weights ,
we obtain 2 ^ 2,-2 + 2 ^ 1,-1 + ^0 , where @xmath25 .
this is in agreement with ( [ eq : rdim ] ) .
similarly , it is possible to decompose the weyl tensor but due to its tracelessness we have the following additional conditions : [ eq : rfree ] & & c_0i0i = c_1i1i = 0 , + & & c_010i = c_0jij , c_101i = c_1jij , + & & 2c_0i1j = c_01ij- c_ikjk , c_0101 = -c_ijij .
it is well known that an @xmath0-dimensional weyl tensor has @xmath26 independent components . by counting the independent scalars of various boost weights , we obtain ^2,-2 + ^1,-1 + ^0 , which is in agreement with ( [ eq : wdim ] ) .
the primary algebraic classification of the weyl tensor in higher dimensions @xcite is based on whether all weyl frame components of a boost weight higher then a specified number can be transformed away by an appropriate choice of the null direction @xmath12 . if it is indeed possible , we call the corresponding direction a weyl aligned null direction ( wand ) of an appropriate order ( 0 , 1 , 2 , 3 )
. if the set of wands of a given order is discrete , we call the corresponding directions principal null directions ( pnd ) .
type iii and n spacetimes admitting a wand of order 2 and 3 , respectively , were introduced in @xcite .
we will use the notation based on that given in @xcite which is suitable for these algebraic classes but becomes rather cumbersome in more general cases .
let us in accordance with @xcite state that a spacetime is of the algebraic type iii if there exists a frame ( [ tetrad ] ) in which the weyl tensor has the form c_abcd = 8 _ i ^_\{a n^ _ b ^ _ c m^(i ) _ d } + 8 _ ijk m^(i)_\{a m^(j ) _ b ^ _ c m^(k ) _ d } + 8 _ ij ^_\{a m^(i ) _
b ^ _ c m^(j ) _ d } [ weyliiin ] , with @xmath27 . the case with @xmath28 ( and consequently from ( [ eqpsi_i ] ) also @xmath29 ) is of the algebraic type n. note , however , that in this paper the operation @xmath30 differs by a factor @xmath31 from that given in @xcite .
the components of the weyl tensor @xmath32 and @xmath33 are given by _
i = c_101i , _ ijk= c_1kij , _ ij = c_1i1j .
note that @xmath33 is symmetric and traceless .
@xmath34 is antisymmetric in the first two indices and from ( [ eq : rcomps ] ) and ( [ eq : rfree ] ) it also follows that _ i=2 _ ijj , [ eqpsi_i ] + _ \{ijk}=0 . [ eqpsicykl ] let us denote components of covariant derivatives of the frame vectors @xmath35 by @xmath36 and @xmath37 , respectively , _
a;b = l_cd m^(c)_a m^(d)_b , n_a;b = n_cd m^(c)_a m^(d)_b , m^(i)_a;b=_cd m^(c)_a m^(d)_b .
since the norm of all frame vectors is constant , it follows that l_0a = n_1a=_ia=0 . also from the fact that all scalar products of the frame vectors are constant , we get n_0a+l_1a=0 , _ 0a + l_ia = 0 , _ 1a+n_ia=0 , _
ja+_ia=0 . [ const - scalar - prod ] we thus arrive at _ a ; b & = & l_11 _ a _ b + l_10 _ a n_b + l_1i _ a m^(i ) _ b + l_i1 m^(i)_a _ b + l_i0 m^(i ) _ a n_b + l_ij m^(i ) _ a m^(j ) _ b , [ dl ] + n_a ; b & = & - l_11 n_a _ b - l_10 n_a n_b - l_1i n_a m^(i ) _ b + n_i1 m^(i ) _ a _ b + n_i0 m^(i ) _ a n_b + n_ij m^(i ) _ a m^(j ) _ b , [ dn ] + m^(i)_a ; b & = & - n_i1 _ a _ b -n_i0 _ a n_b -l_i1 n_a _ b -l_i0 n_a n_b -n_ij _ a m^(j ) _ b + & & + _
j1 m^(j ) _ a _ b -l_ij n_a
m^(j ) _ b + _
j0 m^(j ) _ a n_b + _ kl m^(k ) _ a m^(l ) _ b . [ dm ] in sections [ typen ] and [ piii ] , we will show that the multiple pnd in type iii and n vacuum spacetimes is geodesic .
let us thus study properties of null geodesic congruences in higher dimensions .
analogous but in some cases not fully equivalent definitions are given in the appendix of the paper @xcite .
the congruence corresponding to @xmath12 is geodesic if _ a ; b ^b _ a l_i0 = 0 .
it is always possible to rescale @xmath12 ( and consequently also @xmath13 ) in such a way that @xmath38 and thus also @xmath39 .
from now on we will use this parametrization .
then the covariant derivative of the vector @xmath12 is _ a ; b = l_11 _ a _ b + l_1i _ a m^(i)_b + l_i1 m^(i)_a _ b + l_ij m^(i)_a m^(j)_b [ dlgeod ] .
let us decompose @xmath40 into its symmetric and antisymmetric parts , @xmath41 and @xmath42 , l_ij = s_ij + a_ij , s_ij = s_ji , a_ij= -a_ji , where s_ij=_(a ; b ) m^(i)a m^(j)b , a_ij=_[a ; b ] m^(i)a m^(j)b .
we define the expansion @xmath43 and the shear matrix @xmath44 as follows : & & ^a _ ; a = [ ] , + _ ij & & ( _ ( a ; b ) - _ kl m^(k)_a m^(l)_b ) m^(i)a m^(j)b = s_ij-_ij .
[ shearmatrix ] we will also denote @xmath45 . let us , for simplicity , call @xmath9 the _ twist matrix _ and @xmath6 the _ expansion matrix _ , though @xmath6 contains information about both , expansion and shear
. for simplicity we also introduce quantities , a_ija_ij .
for type n spacetimes the weyl tensor ( [ weyliiin ] ) has the form c_abcd = 8 _ ij ^_\{a m^(i ) _ b ^ _ c m^(j ) _ d } , [ weyln ] where @xmath46 is symmetric and traceless .
bianchi equations , given in [ appbianchi ] , now reduce to d _
ij&=&-2 _ k(i _ j)0-
_ ik l_kj -2 _ ij l_10 , [ bian1 ] + _ [ k _ j]i & = & - _ is _ [ jk ] + _
i[j_k]s + _ i[j l_k]1 + 2 l_1[j
_ k]i , [ bian2 ] + 0&= & l_k[i _ j]k,[bian3 ] + 0&= & _ i[jl_k]0 , [ bian4 ] + 0&= & l_k[j _ m]i + l_i[m _ j]k , [ bian5 ] + 0&= & _ i\{ka_jm}. [ eqbiaa ] let us first show that from equation ( [ bian4 ] ) it follows that the multiple pnd @xmath12 in type n vacuum spacetimes is geodesic . for simplicity we will denote @xmath47 as @xmath48 .
the contraction of @xmath49 with @xmath50 in ( [ bian4 ] ) leads to _
ij l_i = 0 .
now the contraction of ( [ bian4 ] ) with @xmath51 gives _ ik _ ik l_j = 0 .
note that in type n spacetimes @xmath52 and thus the previous equation implies @xmath53 and @xmath12 is indeed geodesic . by substituting @xmath54 into ( [ bian5 ] ) and using ( [ eqbiaa ] )
, we obtain s_k[j _ m]i + s_i[m
_ j]k = 0 .
[ eqbis ] let us now study in detail consequences of equations ( [ bian3 ] ) , ( [ eqbiaa ] ) and ( [ eqbis ] ) .
it is possible to find a general solution of these equations by taking into account their various contractions ( hereafter we assume @xmath55 and @xmath56 ; the non - twisting case can be obtained as a special case with @xmath57 ) .
contracting @xmath49 with @xmath58 in ( [ eqbiaa ] ) gives _ ik a_im = _ i m a_ik + = 0 [ anticoma ] and contracting @xmath50 with @xmath58 in ( [ eqbis ] ) leads to _
ij s_jm + _ jm s_ij = 2 _ i m + = 2. [ anticoms ] previous two equations imply + = 2 . by contracting @xmath49 with @xmath59 in ( [ anticoms ] )
we get _ ij s_ij = 0 _ ij l_ij = 0 .
[ conps ] from ( [ bian3 ] ) and ( [ anticoma ] ) it follows that _ ki s_kj = _ kj s_ki - = 0 .
[ coms ] let us denote a trace @xmath60 $ ] as @xmath61 .
now we are in a position to formulate the following lemma .
[ lem ] in vacuum type n spacetimes with @xmath62 following relations are satisfied : ( a ) & & = , + ( b ) & & = , + ( c ) & & = = , + ( d ) & & = - , + ( e ) & & 2 = p , + ( f ) & & ^t =
^t = ( ^2 + ) . *
( a ) * this equation is a direct consequence of ( [ anticoms ] ) and ( [ coms ] ) . + * ( b ) * by multiplying ( [ eqbis ] ) by @xmath63 and using lemma 1(a ) we get _ kj s_im s_ip+ _ mp s_kj=_km s_ij s_ip+ _ jp s_km . [ eqbias ] an appropriate linear combination of this equation with ( [ eqbis ] ) leads to _ kj ( s_im s_ip - s_mp ) = _ km ( s_ij s_ip - s_pj ) , which , denoting @xmath64 takes the form _
kj x_mp = _ km x_jp .
[ eqpsix ] note that the matrix is symmetric .
now contracting @xmath50 with @xmath58 gives _
jm x_jp = 0 .[prodpsix ] multiplying ( [ eqpsix ] ) with @xmath65 and using ( [ prodpsix ] ) leads to @xmath66 which , contracting @xmath61 with @xmath67 , gives @xmath68 and consequently @xmath69 .
+ * ( c ) * multiplying ( [ eqbiaa ] ) with @xmath70 and using lemma 1(a ) leads to _
ij a_km s_kp + _ ip a_mj + _ i m a_jk s_kp = 0 .
[ eqlemc ] now contracting @xmath49 with @xmath59 and using ( [ anticoma ] ) gives _
ij y_ip = 0 , [ prodpsiy ] where @xmath71 now substituting @xmath72 into ( [ eqlemc ] ) and using ( [ eqbiaa ] ) leads to _ ij y_mp
= _ i m y_jp .
[ eqpsiy ] multiplying ( [ eqpsiy ] ) with @xmath73 using ( [ prodpsiy ] ) gives @xmath74 and consequently @xmath75 .
+ * ( d ) * let us define another symmetric matrix by @xmath76 note that @xmath77=-2 \a .
$ ] multiplying ( [ anticoma ] ) by @xmath78 and again applying ( [ anticoma ] ) leads to _
ik b_ir = _ ir b_ik .
[ eqprodpsib ] multiplying ( [ eqbiaa ] ) by @xmath78 leads to _ ij b_kr - _ ik b_jr + _ i m a_jk a_mr=0 , [ eqpsib ] which after contracting @xmath50 with @xmath67 gives _ ik b_jk = - _ ij .[eqpsibrov ] multiplying ( [ eqpsib ] ) with @xmath79 and using ( [ eqprodpsib ] ) , ( [ eqpsibrov ] ) and ( [ anticoma ] ) leads to @xmath80 and similarly multiplying ( [ eqpsib ] ) by @xmath81 results in @xmath82 .
let us now define a symmetric matrix by q_ij b_ij + s_ij .
using previous formulae it turns out that @xmath83 and thus @xmath84 .
+ * ( e ) * this follows from multiplying ( [ eqbis ] ) by @xmath85 .
+ * ( f ) * this follows directly from lemma 1(b)(d ) .
let us now use lemma 1 to prove following lemmas for vacuum type n spacetimes with @xmath86 : the matrix @xmath6 has at most two eigenvalues @xmath87 and @xmath88 .
let us denote the eigenvector of @xmath6 by @xmath89 .
we thus have @xmath90 . by
multiplying lemma 1(b ) by @xmath89 we obtain s_ij s_jk _ k = s_ik _ k ^2 = and thus @xmath87 or @xmath91 .
the following statements are equivalent : + ( a ) vector @xmath89 is an eigenvector of @xmath6 for the eigenvalue @xmath87 .
+ ( b ) vector @xmath89 is an eigenvector of @xmath9 for the eigenvalue @xmath87 .
+ ( c ) vector @xmath89 is an eigenvector of @xmath46 for the eigenvalue @xmath87 .
\(a ) @xmath92 ( b ) : suppose that @xmath93 .
then by multiplying lemma 1(c ) in the form @xmath94 by @xmath95 we obtain @xmath96 .
+ ( b ) @xmath92 ( a ) : suppose that @xmath97 .
then by multiplying lemma 1(d ) in the form @xmath98 by @xmath99 we obtain @xmath100 .
+ ( a ) @xmath92 ( c ) : suppose that @xmath101 .
then by multiplying lemma 1(a ) in the form @xmath102 by @xmath99 we obtain @xmath103 .
+ ( c ) @xmath92 ( a ) : suppose that @xmath104 .
then by multiplying lemma 1(e ) in the form @xmath105 by @xmath99 we obtain @xmath100 .
every eigenvector of @xmath46 is also an eigenvector of @xmath6 .
the case @xmath87 is solved in lemma 3 .
let us now suppose that @xmath106 , @xmath107 . then by multiplying lemma 1(a ) in the form @xmath108 by @xmath95 we obtain @xmath109 .
the only possible eigenvalues of @xmath46 are @xmath87 and @xmath110 let us now suppose that @xmath106 , @xmath107 .
than thanks to lemma 4 @xmath111 . by multiplying lemma 1(e ) in the form @xmath105 by @xmath99
we obtain @xmath112 . for every symmetric matrix there exists an orthonormal basis of its eigenvectors @xmath113 in which w_ij = _ a _ a ^a_i ^a_j .
let us denote eigenvectors of @xmath46 corresponding to @xmath87 , @xmath114 and @xmath115 by @xmath116 , @xmath117 and @xmath118 , respectively . here
indices @xmath119 , @xmath120 and @xmath121 distinguish between vectors @xmath122 , @xmath123 and @xmath124 .
now we have _
ij = ( v^a_i v^a_j - w^_i w^_j ) . [ decompsumpsi ]
note that for indices @xmath125 , @xmath120 and @xmath121 we also observe einstein s summation convention . thanks to lemmas 3 and 4 s_ij =
( v^a_i v^a_j + w^_i w^_j ) .
[ decompsums ] let us now multiply ( [ eqbis ] ) by @xmath126 .
this leads to ^ab ( s_il + _ il ) = v^a_l v^b_i which for @xmath127 gives @xmath128 and thus just one value of index @xmath129 is possible and consequently @xmath130 . let us thus denote @xmath131 by @xmath132 .
similarly , we can find that @xmath133 and that there is thus just one vector @xmath134 .
equations ( [ decompsumpsi ] ) and ( [ decompsums ] ) now take form _
ij = ( v_i v_j - w_i w_j ) , s_ij =
( v_i v_j + w_i w_j ) .
[ decomppsis ] let us now introduce a new vector by v_i = a_ij v_j .
thanks to lemma 1(c ) and ( d ) we have s_ij v_i = s_ij a_ik v_k = v_j , v_i v_i = 0 , v_i v_i = and thus is just a multiple of @xmath135 @xmath136 and we have freedom to set @xmath137 then we have a_ij v_j = a w_i , a_ij w_j = -a v_i . now by multiplying equation ( [ eqbiaa ] ) by @xmath138 we get a_kl = a ( w_k v_l - v_k w_l ) [ decompa ] .
note that @xmath139 and @xmath9 given by ( [ decomppsis ] ) and ( [ decompa ] ) satisfy equations ( [ eqbiaa ] ) , ( [ eqbis ] ) and thus represent the general solution of these equations for @xmath140
for type iii vacuum spacetimes the weyl tensor is given by ( [ weyliiin ] ) , where @xmath34 satisfies ( [ eqpsi_i ] ) and ( [ eqpsicykl ] ) . the algebraic equations _
ijk l_k = l_[i _ j ] , [ eqgeoda ] + _
kl[j l_i ] = _ ij[k l_l ] , [ eqgeodb ] + l_i[j _ k ] + 2 l_s[j| _ si|k ] = 0 , [ bia33 ] + 2 a_\{ij| _ kl|m } + l_l\{i _ jm}k - l_k\{i _ jm}l=0 , [ bia35 ] where we denote @xmath47 as @xmath48 , follow from the bianchi equations given in [ appbianchi ] .
let us show that from ( [ eqgeoda ] ) and ( [ eqgeodb ] ) it follows that the multiple pnd @xmath12 in vacuum type iii spacetimes is geodesic . by contracting @xmath141 with @xmath58 in ( [ eqgeodb ] ) and using ( [ eqgeoda ] ) , ( [ eqpsi_i ] ) , and ( [ eqpsicykl ] )
we obtain 2 _ ijk l_j = _ k l_i .
[ eqgeodc ] one can see that for the class with @xmath142 it implies ( as we can see by contracting this formula with @xmath48 ) that @xmath143 .
similarly , for the case @xmath144 these equations also imply that @xmath143 ( contract ( [ eqgeodb ] ) with @xmath48 and employ ( [ eqgeoda ] ) and ( [ eqgeodc ] ) ) .
thus , the multiple pnd @xmath12 in vacuum type iii spacetimes is geodesic .
now the task is to solve ( [ bia33 ] ) and ( [ bia35 ] ) .
this turns out to be quite complicated .
we will be able to find a general solution in an arbitrary dimension in the non - twisting case with @xmath145 = 0 ( see [ appiiinon ] ) .
we also present a solution for the twisting case , but we can prove that it is a general solution only if we assume that @xmath34 is in a ` general form ' . in [ app5d ] , we show that this solution is indeed general in all degenerate cases in five dimensions .
let us start with extracting some information from ( [ bia33 ] ) and ( [ bia35 ] ) by various contractions . by contracting ( [ bia35 ] ) and using ( [ bia33 ] )
we get l _ ijk + 2 l_[i|s _
sk|j ] - 2 s_sk _ ijs + l_[i|k _ |j ] = 0 , [ bia335 ] where @xmath146 is the trace of @xmath147 . after adding ( [ bia33 ] ) ( where we replace indices @xmath148 by @xmath149 , respectively ) to ( [ bia335 ] ) we obtain l _ ijk + 4 s_[i|s _
sk|j ] - 2 s_sk _ ijs + 2 s_[i|k _
|j ] = 0 , [ bs1 ] which does not contain @xmath9 the antisymmetric part of @xmath147 , and similarly l _ ijk - 4 _
sk[i a_j]s - 2 _ [ i a_j]k - 2 _ ijs s_sk = 0 .
[ eqpsia ] contraction of @xmath49 with @xmath58 in ( [ bia33 ] ) leads to l _ k - 2 l_ik _
i + 2 l_si _ sik = 0 [ eqlpsi1 ] and contraction of @xmath50 with @xmath58 in ( [ bs1 ] ) ( using ( [ eqpsi_i ] ) and ( [ eqpsicykl ] ) ) to l_ij _ j = 2 l_js _ ijs [ eqlpsij ] and ( [ eqlpsi1 ] ) , using ( [ eqpsicykl ] ) , with ( [ eqlpsij ] ) imply l_i - l_ij_j-2l_ji_j+2l_kj_ijk=0.[eqlpsi2 ] by substituting @xmath150 into ( [ eqlpsi1 ] ) and ( [ eqlpsi2 ] ) we get & & l_i-2s_ij_j-2a_ji_j-4a_jk_ijk=0,[preas1 ] + & & l_i-3s_ij_j - a_ji_j+2(s_jk - a_jk)_ijk=0[preas2 ] and their linear combinations give & & l_i+4a_ij_j-4(s_jk+2a_jk)_ijk=0,[as2 ] + & & l_i-4s_ij_j+4s_jk_ijk=0.[as1 ] by multiplying ( [ bia33 ] ) by @xmath151 we arrive to l_ij l_ik _ j = _ k , [ eqell ] where @xmath152 . inspired by the type n case , we again choose an orthonormal basis of eigenvectors of @xmath6
. we will denote vectors corresponding to non - zero eigenvalues of @xmath6 as @xmath153 and vectors corresponding to the eigenvalue 0 as @xmath154 .
we thus have s_ij = _ a=1^ _ a v_i^a v_j^a . [ decompiiisa ] unfortunately , there are three indices @xmath129 in this formula and we thus will not use einstein s summation convention for indices @xmath155 ( though we will still use it for indices @xmath156 ) .
let us now decompose other quantities appearing in ( [ bia33 ] ) and ( [ bia35 ] ) _ i = _ a a_a v_i^a + b _
u_i^ , [ alb ] + a_ij = _ ab _ ab ( v_i^a v_j^b - v_j^a v_i^b ) + _ a _ a ( v_i^a u_j^ - v_j^a u_i^ ) + _ ( u_i^ u_j^ - u_j^ u_i^ ) , + _ ijk = _ abc _ abc ( v_i^a v_j^b - v_j^a v_i^b ) v_k^c + _ ab _ ab ( v_i^a v_j^b - v_j^a v_i^b ) u_k^ + + _ ac _ a c ( v_i^a u_j^ - v_j^a u_i^ ) v_k^c + _ a _ a ( v_i^a u_j^ - v_j^a u_i^ ) u_k^ + + _ c _ c ( u_i^ u_j^ - u_j^ u_i^ ) v_k^c + _ ( u_i^ u_j^ - u_j^ u_i^ ) u_k^ , where @xmath157 and @xmath158 are antisymmetric in first two indices .
now we need to rewrite some of the equations from this section in terms of these new quantities ( more complicated equations are given in [ appdecomp ] ) . from ( [ eqpsi_i ] ) and ( [ as1 ] )
, it follows that a_a & = & 4 _ b _ abb+ 2 _ a , [ tra ] + b _ & = & -2 _ b _ bb + 4 _ , [ trb ] and ( l-4_a)a_a & = & -8 _ b _
b_abb , [ tra2 ] + l b _
& = & 4 _ b_b _ bb , [ trb2 ] respectively . equation ( [ bs1 ] ) implies l _ = 0 [ cbs11 ] , + ( l-2_c ) _ c = 0 [ cbs12 ] , + l _ b + 2 _ b _ b = 0 [ cbs13 ] , + ( l-2_c ) _ a c + 4 _ a _ ac + _ a b _ _ ac = 0 [ cbs14 ] , + 2 _ a _ a b - 2 _ b _ ba + 2l _ ab = 0 [ cbs15 ] , + ( l - 2 _ c ) _ abc + 2 _ a _ acb + 2_b _ cba + _ c _ c[aa_b ] = 0,[cbs16 ] and ( [ eqpsicykl ] ) leads to _ a [ ] + _ a = 0 , [ bia01 ] + _ [ a| | c ] + _ ca = 0 , [ bia02 ] + _ \ { abc } = 0 , [ bia03 ] + _ \ { abc } = 0 .[bia04 ] in the twisting case , the equations ( [ bia33 ] ) and ( [ bia35 ] ) , without specifying dimension , are quite complex , and there are too many distinct cases to be solved .
we thus present the solution provided @xmath159 and provided that for every pair @xmath129 , @xmath160 ( @xmath161 ) there exists @xmath162 for which @xmath163 .
it turns out that the only eigenvalues of @xmath6 that are compatible with ( [ bia33 ] ) and ( [ bia35 ] ) correspond to @xmath164 , @xmath165 , @xmath166 .
we also checked other cases ( but not all of them ) and they also lead to the same conclusion . for completeness , we treat all other possible cases in five dimensions in [ app5d ] . a linear combination of ( [ cbs15 ] ) and ( [ bia02 ] ) , [ ( [ cbs15 ] ) - @xmath167 ( [ bia02 ] ) ] , gives ( 2 _ a + l ) _ a b = ( 2 _ b + l ) _
b a. [ eqo1 ] for @xmath168 [ ( [ cbs14])-@xmath169 ( [ bia02 ] ) ] leads to ( l + 2 _ a -2 _ b ) _ a b= 2 _ a _ b a _ a = b , [ eqo2 ] and [ @xmath170 ( [ eqo1])@xmath171([eqo2 ] ) ] gives ( _ a^2 + _ b^2 - _ a _ b - ) _ ab = 0 _ a = b .
[ eqlambdao ] if @xmath172 , we obtain from ( [ eqlambdao ] ) ( _ a^2 + _ b^2 - _ a _ b - ) = 0 _ a = b .
[ eqlambda ] if for each pair @xmath129 , @xmath173 there exists @xmath174 such that @xmath175 then ( [ eqlambda ] ) is valid for all @xmath129 , @xmath173 , @xmath176 . in the case
@xmath164 , we have @xmath177 and equation ( [ eqlambdao ] ) leads to @xmath178 and consequently @xmath179 .
in the case @xmath180 , by subtracting ( _ a^2 + _ c^2 - _ a _ c - ) = 0 _ a = c [ eqlambdac ] from ( [ eqlambda ] ) we obtain ( _ b - _ c ) ( _ b + _ c - _ a ) = 0 _ a = b , a = c , b = c and thus _ b - _ c=0 [ equaleigenvals ] or
_ b + _ c - _ a _ a = b , a = c , b = c=0 .
however , in the second case one can show that simply by summing this equation with interchanged indices @xmath181 which can not happen as @xmath182 by definition and thus ( [ equaleigenvals ] ) is satisfied and by substituting @xmath183 in ( [ eqlambda ] ) we obtain @xmath184 note however that @xmath185 and thus the case @xmath186 is excluded and @xmath164 .
thus we can conclude with for vacuum type iii spacetimes with @xmath187 , providing that for every pair @xmath129 , @xmath160 ( @xmath161 ) there exists @xmath162 for which @xmath163 , the expansion matrix @xmath6 has just one non - vanishing eigenvalue @xmath188 which has multiplicity 2 .
let us now study the case with two non - vanishing eigenvalues of @xmath6 , @xmath189 , in general , i.e. without any assumptions about @xmath190 . from ( [ cbs11 ] ) and ( [ cbs13 ] ) it follows that _
= 0 , _ a = - _ a .
[ antsymp ] by assuming that the index @xmath192 in ( [ cbs14 ] ) we obtain _ ac=0 . since @xmath193 is antisymmetric in the first two indices , @xmath194 for all combinations of indices . by substituting @xmath195 in ( [ cbs14 ] )
we obtain b_=0 and thus @xmath196 is an eigenvector of @xmath6 corresponding to @xmath197 .
equation ( [ cbs14 ] ) now implies _ ac = _ c a and ( [ trb ] ) implies _ a _ a a = 0 .
[ sumo ] substituting @xmath198 into ( [ cbs16 ] ) leads to _ abc=0 _ a = c , b = c , ( [ tra ] ) leads to a_a = 4 _ b _ abb a_1= 4 _ 122 , a_2=4 _ 211 [ summ ] and ( [ bia01 ] ) gives _ a + _ a = 0 .
[ prrel ] substituting the index @xmath199 into ( [ bia3-bc ] ) , summing over the index @xmath129 and using ( [ bia3-h ] ) , ( [ summ ] ) , ( [ sumo ] ) and ( [ antsymp ] ) we arrive at @xmath200 .
consequently , the vector @xmath201 _
j = _ b 2 _ ab a_b v_i^a , which is orthogonal to @xmath196 , is also an eigenvector of @xmath6 .
we thus have two orthogonal eigenvectors of @xmath6 , @xmath196 and @xmath202 .
let us denote @xmath203 by @xmath204 and @xmath205 by @xmath206 .
now we choose an orthogonal basis with these two vectors which corresponds to v^1_i , v^2_i , . in this basis we will denote components corresponding to @xmath207 and @xmath208 by indices @xmath209 and @xmath210 , respectively .
@xmath6 now takes the form s_ij = ( + ) .
[ vysls ] this equation implies @xmath211 . now using @xmath212 it follows from the previous equations and ( [ eqell ] ) that a_ik _
i = _ k and thus ^2 = _ i _ i = _ i a_ik _ k = _ k _ k = ^2 .
@xmath145 as an antisymmetric matrix takes the general form a_ij = & & _ pf ( _ i _ j - _ j _
( _ i u_j^ - _ j u_i^ ) + & + & _ f ( _ i u_j^ - _ j u_i^ ) + _ ( u_i^ u_j^ - u_j^ u_i^ ) but from @xmath213 and @xmath214 we get _
p=0 , _ pf = - , _ f=0 and thus a_ij = ( _ i _ j - _ j _
i ) + _ ( u_i^ u_j^ - u_j^ u_i^ ) .
thanks to ( [ antsymp ] ) ( [ prrel ] ) we can rewrite @xmath34 in the form _ ijk= ( _ i _ j - _ j _
i ) _ k + ( _ i u_j^-_j u_i^ ) _ k - ( _ i u_j^-_j u_i^ ) _ k + + ( _ i u_j^-_j u_i^ ) _ k + ( _ i u_j^-_j u_i^ ) _ k + ( _ i u_j^ - _ j u_i^ ) u_k^ + + ( _ i u_j^ - _ j u_i^ ) u_k^ - ( u_i^ u_j^- u_j^ u_i^ ) _ k - ( u_i^ u_j^- u_j^ u_i^ ) _ k . [ psiform ] by substituting ( [ psiform ] ) into ( [ bia5-d ] ) we obtain @xmath215 , @xmath216 and since @xmath217 we also get @xmath218
. then ( [ bia5-f ] ) or ( [ bia5-eg ] ) gives @xmath219 which , since @xmath220 does not vanish , implies @xmath221 .
in this case we thus have a_ij = ( _ i _ j - _ j _
i ) , [ vysla ] and _ ijk&= & ( _ i _ j - _ j _
i ) _ k + & + & ( _ i u_j^-_j u_i^ ) _ k - ( _ i u_j^-_j u_i^ ) _ k + & + & ( _ i u_j^-_j u_i^ ) _ k + ( _ i u_j^-_j u_i^ ) _ k .
[ vyslb ] we do not need to examine the rest of equations in [ appdecomp ] since @xmath222 and @xmath34 given by ( [ vysls ] ) , ( [ vysla ] ) and ( [ vyslb ] ) , respectively , already satisfy both equations ( [ bia33 ] ) and ( [ bia35 ] ) and thus represent their solution . note that from ( [ vysls ] ) , ( [ vysla ] ) and ( [ vyslb ] ) we obtain for type iii spacetimes relations equal or analogous to equations given in lemma [ lem ] : ( a ) & & _ ijss_sk=_ijk , + ( b ) & & s_iks_jk=s_ij , + ( c ) & & a_iks_kj = s_ika_kj=a_ij , + ( d ) & & a_ika_kj= - s_ij , + ( e ) & & l_ikl_jk=l_kil_kj = ( ^2 + ) s_ij .
in sections [ typen ] and [ piii ] , it is shown that for type n and iii vacuum spacetimes the expansion and twist matrices @xmath6 and @xmath9 have very specific properties given by ( [ decomppsis ] ) , ( [ decompa ] ) and ( [ vysls ] ) , ( [ vysla ] ) for type n and iii , respectively . note , however , that while in higher dimensions as well as in 4d the multiple pnd is geodesic , it is not shearfree for @xmath8 .
the question thus arises whether there exist some properties of matrices @xmath6 and @xmath9 that are satisfied ( together with the condition that the spacetime possesses a geodesic null congruence ) if and only if the vacuum spacetime is algebraically special .
the answer is unclear at present .
moreover , the conditions for @xmath6 and @xmath9 that hold for types n and iii are not satisfied for type d spacetimes ( see [ typed ] ) .
let us here , as a first step towards such possible generalization , show that from the ricci identities it follows that [ lemgs ] \a ) suppose that an otherwise arbitrary vacuum spacetime admits a non - expanding and non - twisting geodesic null congruence ( i.e. @xmath223 ) .
then , the spacetime in question is algebraically special .
+ b ) suppose that an otherwise arbitrary vacuum spacetime admits an expanding and twisting geodesic null congruence , and that its @xmath6 and @xmath9 matrices , in appropriately chosen frame , have the form ( [ decomppsis ] ) , ( [ decompa ] ) .
then , the spacetime in question is algebraically special .
the contraction of the ricci identities @xmath224 with @xmath225 assuming that @xmath226 is geodesic with an affine parametrization leads to dl_ij+l_isl_sj+l_is_j0+l_sj_i0+r_0i0j=0 .[riccilmm ] if @xmath227 then @xmath228 for all @xmath49 , @xmath58 and the spacetime is thus algebraically special .
this proves the part ( a ) of lemma [ lemgs ] .
if we switch @xmath49 and @xmath58 in ( [ riccilmm ] ) and add and subtract the two equations we obtain 2ds_ij+l_isl_sj+l_jsl_si+2s_is_j0 + 2s_sj_i0 + 2r_0i0j=0 , [ riccis ] + 2da_ij+l_isl_sj - l_jsl_si+2a_is_j0 + 2a_sj_i0=0 .[riccia ] we further assume that we have chosen vectors @xmath17 in such a way that @xmath229 is diagonal with the only non - zero entries being @xmath230 and that @xmath9 has only two non - vanishing components @xmath231 .
then ( [ riccis ] ) takes the form 2ds_ij+ ( ^2-)s_ij+2s_is_j0 + 2s_sj_i0 + 2r_0i0j=0 .
[ ricciss ] if @xmath232 then @xmath233 . if @xmath234 then ( [ ricciss ] ) implies @xmath235 which using ( [ const - scalar - prod ] ) leads to @xmath236 . if @xmath237 then again from ( [ ricciss ] ) we get @xmath238 which together with @xmath239 gives @xmath240 . for @xmath241 and @xmath242 equations ( [ ricciss ] ) and ( [ riccia ] )
turn to be 2s_is_j0 + 2r_0i0j=0 s_22_j0+r_020j=0 , s_33_j0+r_030j=0 , + 2a_is_j0=0 a_23_j0=a_32_j0=0 _ j0=_j0=0 that leads to @xmath243 and hence @xmath228 for all @xmath49 , @xmath58 and the spacetime is thus algebraically special .
this proves the part ( b ) of lemma [ lemgs ] .
note also that by contracting @xmath49 with @xmath58 in ( [ ricciss ] ) we obtain for higher dimensions the same relation which is also valid for 4d d=-^2 .
in this paper , we present a higher dimensional frame formalism .
the complete set of frame components of the bianchi identities , which are in this context usually called the bianchi equations , is given in [ appbianchi ] . for algebraically general
spacetimes these equations are quite complicated .
however , for algebraically special cases they are much simpler ( e.g. , see section 3 for the type n case ) . in 4d
it is possible to use the bianchi and ricci equations to construct many algebraically special solutions of einstein s field equations .
the hope is that it is possible to do a similar thing in higher dimensions , at least for the simplest algebraically special spacetimes .
the vast majority of today s known higher dimensional exact solutions are simple generalizations of 4d solutions . the present approach may lead to new , genuinely higher dimensional exact solutions .
type n and d ( see @xcite for the definition of type d in higher dimensions ) solutions may be of particular physical interest . in particular , we study the consequences of the algebraic bianchi equations for type n and iii spacetimes .
it turns out that the principal null direction @xmath12 in these spacetimes is geodesic . for
vacuum type n spacetimes with non - vanishing expansion or twist we also prove that the corresponding components of the weyl tensor and expansion and twist matrices , in an appropriately chosen frame , can be expressed as _
ij = ( v_i v_j - w_i w_j ) , s_ij =
( v_i v_j + w_i w_j ) , a_kl = a ( w_k v_l - v_k w_l ) .
note that we do not obtain any constraints for @xmath33 in non - twisting and non - expanding type n spacetimes .
we also establish similar results for vacuum type iii spacetimes .
the weyl tensor is said to be reducible if we can decompose it into two parts @xmath244 where indices @xmath245 have values from 0 to @xmath21 , whereas indices @xmath246 have values from 0 to @xmath247 , and indices @xmath248 from n to @xmath21 @xcite . in this sense
the weyl tensor in vacuum type n spacetimes with non - vanishing expansion or twist is reducible , with a nontrivial four - dimensional part and a vanishing @xmath249-dimensional part . in 4d the well - known goldberg
sachs theorem @xcite states that ` a vacuum metric is algebraically special if and only if it contains a shearfree geodesic null congruence ' .
this theorem ( and also its non - vacuum generalizations see @xcite ) is very useful for constructing algebraically special exact solutions . at present
it is unclear if and to what extent this theorem may be generalized for higher dimensions ( see sec .
[ goldberg ] for details ) .
recently , all 4d spacetimes with vanishing curvature invariants ( i.e. , vanishing invariants constructed from the riemann tensor and its covariant derivatives of an arbitrary order vsi spacetimes ) were determined in @xcite . in @xcite , the generalization to higher dimensions was discussed .
all these spacetimes are of type iii , n or o. the results presented in this paper will enable us to explicitly express curvature invariants involving derivatives of the riemann tensor in higher dimensions .
for example ,
for type n vacuum spacetimes we obtain , thanks to lemma 1 , i = c^a b c d ; pq c_a m c n ; pq c^t m u n ; rs c_t b u d ; rs = 3 ^ 2 2 ^ 10 p^2 ( s^2+a^2)^4 . consequently , we will be able to prove that higher dimensional vsi spacetimes are expansion and twist free , thereby proving the assertion made in @xcite ( see @xcite ) .
we are grateful to the referee for constructive criticism .
ap and vp would like to thank dalhousie university for its hospitality while part of this work was carried out .
rm and ac were partially supported by a research grant from nserc .
vp was supported by grant gacr-202/03/p017 and ap by grant kjb1019403 .
ap and vp also thank to nlf for support .
equation ( [ bia33 ] ) gives _
+ _ f _ f[|_f|]+4_=0 , [ bia3-h ] + 2 _ a_c + _ c b _ -2 _ c _ c + _ f 4_cf _ f + 2 _ f _ f _ fc + + 8 _ _
c + 4 _ c _ = 0 , [ bia3-fg ] + _ a [ b _ ] + 4 _ f _ f [ | _ fa| ] - 4 _ = 0 , [ bia3-e ] + _ b _ b c - _ c _ c b + _ [ c | a_|b ] + 4 _ f _ f[b| _ f |c ] + 4 _ [ b| _ |c]=0 , [ bia3-d ] + _ a _ ab b _ - _ a a_b + 2 _ ab b _ + 4 _ b _ ba + 8 _ f _ fb _ fa + + 2 _ b _ a - 4 _ f _ f _ fab + 4 _ _ a b = 0 , [ bia3-bc ] + _ a _ a[b a_c ] + 2 _ a[ba_c ] + 2 _ b _ bac - 2 _ c _ cab + + 8_f _ f [ b| _ fa|c]+2 _ [ b| _ a|c]=0.[bia3-a ] equation ( [ bia35 ] ) leads to 2_\ { | _ | } + _ \ { _ } -_\ { _ } = 0 , [ bia5-n ] + 2 _ _ a + _ + _ + 2 _ a [ _ | ] + _ a [ _ | ] = 0 , [ bia5-m ] + 2 _ \{| _ c | } + 2 _ \ { _ } c - _ c \ { _ } = 0 , [ bia5-l ] + - 4 _ [ a| _ |b ] + _ [ b| _ |a ] - _ [ b| _ |a ] + 4 _ [ | _ ab| ] + 4 _ ab_= 0 , [ bia5-k ] + 2 _ a [ | _ c | ] + _ c [ | _ a | ] + 2 _ _ c a - 2 _ c - _ a _
c + -(_a_ac+2_ca ) _ = 0,[bia5-ij ] + 4 _ \{| _ cd| } + _ d \ { _ } c - _ c \ { _ } d = 0 , [ bia5-h ] + 4 _ _ cde + 4 _ e [ | _ cd| ] + _ d[| _ e | ] c - _ c[| _ e | ] d + + 2 _ e _ e[d _ |c]+ 4 _ [ d|e _ |c ] = 0 , [ bia5-eg ] + 2 _ _ bed + _ d _ be + 2 _ [ b| _ d |e ] + _ [ b| _ e |d ] + - 2 _ be _ d + 2 _ d[b _ e ] + _ d _ d[b _ e ] = 0 , [ bia5-f ] + _ [ a| _ b]e + _ [ b| _ a]e + _ e [ _ ab| ] + 2 _ \{ab| _ |e } = 0 , [ bia5-d ] + 4 _ [ e| _ cd|b ] + 2 _ [ d| _ be|c ] + 4 _ be _ cd + 2 _ d[b _ e]c + 2 _ c[e _ b]d + + _ b _ b[d| _ e|c ] + _ e _ e[c| _ b|d ] = 0,[bia5-c ] + _ \ { a |_be } d-2_\ { ab|_d|e}+2_d\ { a_be}+_d_d\ { a _ be } = 0 , [ bia5-b ] + 4_\ { ab| _ cd|e } + 2_d\ { a _ be } c-2_c\ { a _ be } d + + _ d_d\ { a _ be } c -_c_c\ { a _ be } d=0.[bia5-a ]
d r_101i + r_010i-_i r_0101= 2r_010[i l_1]1- r_101i l_10 + - r_01is l_s1- 2r_101s l_si + 2r_[1|i1s l_s|0 ] - r_0101 n_i0 - 2r_010s n_si + + r_01is n_s0 + 2r_0s[0|i n_s|1 ] - r_010s _ i1 - r_101s _ i0 , - r_01ij+2_[j|r_101|i ] = 2r_101[j| l_1|i ] + 2r_101[i| l_|j]1 + 2r_1[i|1s l_s|j]+ r_1sij l_s1 + + 2r_0101 n_[ji]+ 2r_010[j n_i]1 + 2r_01[i|s n_s|j ] + 2r_0s1[j| n_s|i ] - r_0sij n_s1 + + 2r_01[i|s _ |j]1 + 2r_101s _ [ ji ] , -d r_01ij + 2_[i|r_010|j]= 2r_0101 l_[ij]+2r_010[j| l_1|i ] + 2r_01[i|s l_s|j ] + + 2r_0[i|1s l_s|j ] + 2r_101[i l_j]0 + r_1sij l_s0 + 2r_010[j n_i]0 + 2r_0[j|0s n_s|i ] + - r_0sij n_s0 + 2r_01[i|s _ |j]0 + 2r_010s _ [ ij ] , d r_1i1j - r_0j1i-_j r_101i= 2r_101i l_[1j]- 2r_1i1j l_10 + 2r_1i[j|s l_s|1 ] + + 2r_010[1| n_i|j]- 2r_101(i| n_|j)0 + 2r_0[j|is n_s|1]+ r_0s1i n_sj - 2r_1(ij)s n_s0 + + 2 r_[0|j1s _ i|1]+ 2 r_[0|s1i _ j|1 ] + r_101s _ ij , d r_0i1j - r_0i0j+_j r_010i= 2r_010[j| l_i|1]+ 2r_010i l_(1j ) - 2r_0i0j l_11 + + r_101j l_i0 + 2r_0i[j|s l_s|1]+2r_0[j|is l_s|1]- r_1jis l_s0 + r_010i n_j0 + 2r_0i[0|s n_s|j ] + - r_010s _ ij + 2 r_0s[0|j _ i|1]+2 r_0i[0|s
_ j|1 ] , -d r_1kij+r_0kij-_k r_01ij= - 2r_0[1|ij l_|k]1 + 2r_0k1[i l_j]1 + 2r_101[i l_j]k + - 2r_1[i|1k l_|j]0 + r_1kij l_10 + 2 r_[1|sij l_s|k ] - 2r_010[i n_j]k- r_01ij n_k0 + + 2r_0[i|0k n_|j]1 + 2r_0[j|1k n_|i]0 + 2r_[k|sij n_s|0 ] + + 2r_01[i|s _ |j]k + 2r_[0|kjs _ i|1 ] + 2r_[1|kis _ j|0 ] + 2r_[1|sij _ k|0 ] , r_0ijk+2_[k| r_0i1|j]= 2r_0i1[j| l_|k]1 + 2r_0[i|jk l_|1]1 + 2r_101[j| l_i|k ] + 2r_1[k|is l_s|j ] + - r_isjk l_s1 + 2r_010i n_[jk ] + 2r_0i0[j n_k]1 + 2r_0i[k|s n_s|j ] + + 2r_0i1s _ [ kj ] + 2r_0s1[k| _ i|j ] + 2r_0i[k|s _ |j]1 - r_0sjk _ i1 , d r_0ijk+2_[k|r_0i0|j]=- 2r_010i l_[jk]+2r_010[k| l_i|j ] + 4r_0i0[j| l_1|k]+ 2r_0i1[j l_k]0 + + 2r_0[i|jk l_|1]0 + 2r_0i[k|s l_s|j ] + 2r_0[k|is l_s|j ] - r_isjk l_s0 + 2r_0i0[j| n_|k]0 + + 2r_0i0s _ [ kj ] + 2 r_0[k|0s _ i|j ] + 2 r_0i[k|s _ |j]0 - r_0sjk _ i0 , dr_1ijk+2_[k|r_0|j]1i=2r_101i l_[jk ] + 2r_1i1[j l_k]0- r_1ijk l_10 + 2r_1i[k|s l_s|j ] + + r_010j n_ik + 2r_01[j|k n_i|0 ] - 2r_0[k|1i n_|j]0 + 2r_0[k|is n_s|j]- r_isjk n_s0 + + 2r_0[k|1s _ i|j ] - 2r_0s1i _ [ jk]+ 2r_1i[k|s _ |j]0- r_1sjk _ i0 , r_1ijk+2_[k|r_1i1|j]= 2r_1i1[j l_k]1 + 4r_1i1[k| l_1|j]- r_1ijk l_11 + + r_01jk n_i1 + 2r_0[j|1i n_|k]1 + 2r_101[k| n_i|j ] + 2r_101i n_[kj ] + 2r_1i[k|s n_s|j]+ 2r_1[k|is n_s|j ] + - r_isjk n_s1 - 2r_1i1s _ [ jk ] + 2r_1[k|1s _ i|j ] + 2r_1i[k|s _ |j]1 - r_1sjk _ i1 , -_\{i|r_01|jk}= r_101\{i l_jk } - r_101\{i l_kj } + r_1s\{ij| l_s|k } + r_010\{i n_kj } + - r_010\{i n_jk}- r_0sij n_sk+ 2r_0s[i|k n_s|j ] + r_01\{i|s _ |jk } - r_01\{i|s _ |kj } , -dr_ijkm+2_[k|r_0|m]ij= 2r_01ij l_[km ] + 2r_0k[1|j l_|i]m + 2r_0[m|1i l_j|k ] + + 2r_0m[i|j l_|1]k + 2r_1[j|km l_|i]0 + 2r_1[m|ij l_|k]0 + 2r_ij[k|s l_s|m ] + + 2r_0i0[m| n_j|k ] + 2r_0j0[k| n_i|m ] + 2r_0[j|km n_|i]0 + 2r_0[m|ij n_|k]0 + 2r_0[k|is _ j|m ] + + 2r_0[m|js _ i|k ] + 2r_0sij _ [ km ] + 2r_[i|skm _ |j]0 + 2r_ij[k|s _ |m]0 , -r_ijkm+2_[k|r_1|m]ij= 2r_1i[1|m l_j|k ] + 2r_1[j|1k l_|i]m + 2r_1j[k|m l_i|1 ] + + 2r_1kij l_[1 m ] + 2r_1mij l_[k1 ] - 2r_01ij n_[km ] + 2r_0[i|1 m n_|j]k + 2r_0[j|1k
n_|i]m + + 2r_0[j|km n_|i]1 + 2r_0[m|ij n_|k]1 + 2r_ij[k|s n_s|m ] + 2r_1k[i|s _ |j]m + + 2r_1m[j|s
_ |i]k + 2r_1sij _ [ km ] + 2r_ij[k|s _ |m]1 + 2r_[i|skm _ |j]1 , _ \{j|r_1i|mk}= r_1i1\{j l_mk } - r_1i1\{j l_km } + r_1i\{jk| l_1|m } + r_01\{jm| n_i|k } - r_0\{j|1i n_|km } + + r_0\{j|1i n_|mk } + r_is\{jk| n_s|m } - r_1i\{j|s _ |mk } + r_1i\{j|s _ |km } + r_1s\{jk| _ i|m } , _ \ { j|r_0i|mk}= r_01\{jk| l_i|m } - r_0i1\{j l_km } + r_0i1\{j l_mk } + r_0i\{jm| l_1|k}+ r_is\{jk| l_s|m } + + r_0i0\{j n_mk } - r_0i0\{j n_km } + r_0i\{j|s _ |km } - r_0i\{j|s _ |mk } + r_0s\{jk| _ i|m } , _ \ { k|r_ij|nm}= r_1j\{km| l_i|n}- r_1i\{km| l_j|n } - r_1\{k|ij l_|mn } + r_1\{k|ij l_|nm } + + r_0j\{km| n_i|n } - r_0i\{km| n_j|n } + r_0\{k|ij n_|nm } - r_0\{k|ij n_|mn } + + r_ij\{k|s _ |mn } - r_ij\{k|s _
|nm}+ r_is\{km| _
j|n}- r_js\{km| _
in this case , @xmath9=0 and thus @xmath250 is symmetric ( we assume @xmath251 since in the non - twisting and non - expanding case both sides of equations ( [ bia33 ] ) and ( [ bia35 ] ) vanish ) . from ( [ eqlpsi1 ] )
we obtain s_ij _ j = _
i. [ eqnt1 ] note that providing @xmath252 from ( [ eqnt1 ] ) and from ( [ eqell ] ) we obtain @xmath253 and thus the case @xmath254 leads to @xmath255 .
equation ( [ eqpsia ] ) leads to l _ ijk = 2 _ ijs s_sk , which using the decomposition ( [ alb ] ) implies l_ab=0 , l_a=0 , l_=0 , [ aonpsl ] + ( l-2_c ) _
abc=0,(l-2_c ) _
ac=0,(l-2_c ) _ c=0 . first study the case with @xmath254 . from equations ( [ aonpsl ] )
we get @xmath256 , @xmath257 , @xmath258 and thus from ( [ cbs13 ] ) and ( [ bia02 ] ) @xmath259 , @xmath260 and further from ( [ bia5-ij ] ) @xmath261 which imply @xmath262 ,
i.e. , a type n or o spacetime and thus there are no non - trivial solutions of ( [ bia33 ] ) and ( [ bia35 ] ) of type iii with @xmath254 .
now we may proceed to the case @xmath263 .
then ( [ aonpsl ] ) and ( [ bia01 ] ) imply _
ac=0 , _ = 0 , _
c=0.[aonps ] from ( [ eqnt1 ] ) we get _ a a_a = a_a , b_=0 , [ nontwist_b ] which leads to a_a=0 _ a=.[aal ] 1 . if all @xmath264 then all @xmath265 ( [ aal ] ) and thus @xmath164 , @xmath189
. the corresponding solution is given by ( [ res_nontwist ] ) .
if at least one @xmath266 then corresponding @xmath265 and from ( [ cbs16 ] ) with @xmath267 and ( [ bia3-a ] ) with @xmath268 and after interchanging @xmath129 with @xmath173 we get a_a&=&_abb_b = a , + a_a&=&4_abb_b = a .[aam ] comparing the right hand sides we obtain @xmath269 and thus @xmath270 for all @xmath271 .
however , the remaining @xmath265 and in this case again @xmath164 , @xmath189 .
if all @xmath272 then ( [ aam ] ) implies _
abb=0[nontwist_mabb ] for all @xmath129 , @xmath173 .
we treat separately the case with all @xmath256 and the case with some @xmath273 .
+ ( a ) + if all @xmath256 than there exists at least one non - vanishing @xmath274 in order to have a type iii weyl tensor . + for @xmath275 , from equation ( [ bia5-c ] ) for @xmath276 , @xmath277 , @xmath278 , @xmath279 we get @xmath280 which implies that all @xmath281 . equation ( [ cbs14 ] ) with @xmath282 , i.e. , @xmath283 , implies _
aa=0 _
.[nontwist_oaa ] from equation ( [ bia5-c ] ) ( for @xmath276 , @xmath284 , @xmath277 ) _
c_dd=-_d_cc_d = c.[nontwist_ooaa ] if there exists @xmath285 then from ( [ nontwist_oaa ] ) @xmath286 and thus from ( [ nontwist_ooaa ] ) all @xmath287 and @xmath262 , the spacetime is of the type n or o. the other possibility that all @xmath265 is not compatible with @xmath288 and so @xmath289 .
+ for @xmath164 , ( [ trb ] ) with ( [ nontwist_ooaa ] ) imply _
2=_1= _ 22=-_11[nontwist_o1122 ] or @xmath290 . in the second case ,
equation ( [ bia02 ] ) implies _
21=_12[nontwist_o12 ] and from ( [ cbs15 ] ) ( or ( [ bia3-d ] ) ) we get @xmath291 . + for @xmath130 , from ( [ nontwist_oaa ] ) we immediately get @xmath292 , type n or o spacetimes . + to summarize , the case @xmath256 ( @xmath293 @xmath129 , @xmath173 , @xmath160 ) always leads to @xmath164 , @xmath189 or to type n or o spacetimes .
+ ( b ) + let us now assume , for @xmath294 , that there exists @xmath295
. then ( [ bia5-a ] ) for @xmath278 , @xmath296 , @xmath161 , @xmath297 , @xmath277 and @xmath279 implies @xmath298 and thus all @xmath299 which is in contradiction with our assumptions and consequently @xmath300 .
+ equation ( [ cbs16 ] ) with @xmath301 and @xmath302 and ( [ bia3-a ] ) gives ( l-2_c+2_a ) _ cab & = & ( l-2_c+2_b ) _
cba_c = a , c = b[mcabmcba ] , + _ a_cab&=&_b_cba .
these two equations leads to ( l-2_c)_cab=(l-2_c)_cba and so either _
c= _ a_cab=_b_cba[nontwist_ll2_mabc ] or _
cab=_cba _ a_cab=_b_cba [ nontwist_lnl2_mabc ] i.e. , _ cab=_cba=0 _ cab=_cba= 0 _ a=_b .
+ let us assume that @xmath303 .
let @xmath304 .
then the case ( a ) @xmath305 ( [ nontwist_lnl2_mabc ] ) implies @xmath306 ( [ bia03 ] ) which is the case @xmath256 studied above ( a ) .
the condition ( b ) @xmath307 ( [ nontwist_lnl2_mabc ] ) also implies @xmath306 ( [ bia03 ] ) and for @xmath308 we get @xmath309 ( [ nontwist_lnl2_mabc ] ) and thus also @xmath310 and this is again the case @xmath256 .
however , if @xmath166 then @xmath311 ( [ nontwist_ll2_mabc ] ) and since @xmath306 we again obtain the case @xmath256 .
+ if @xmath165 then @xmath166 immediately implies @xmath312 and @xmath308 corresponds to the case analyzed in the previous paragraph .
+ @xmath313 again leads to all @xmath256 ( [ nontwist_mabb ] ) . to summarize ,
the general solution of ( [ bia33 ] ) and ( [ bia35 ] ) in the non - twisting case may be written , substituting ( [ aonps ] ) , ( [ nontwist_b ] ) , ( [ aam ] ) , ( [ nontwist_o1122 ] ) and ( [ nontwist_o12 ] ) into ( [ alb ] ) , in the form _ i&=&4_122v^1_i + 4_211v^2_i , + s_ij&=&(v^1_i v^1_j + v^2_i v^2_j),[res_nontwist ] + _ ijk&= & 2_122 ( v^1_i v^2_j - v^1_j v^2_i ) v^2_k + 2_211 ( v^2_i v^1_j - v^2_j v^1_i ) v^1_k + & + & _
11(v^1_i u_j^-v^1_j u_i^ ) v^1_k - _ 11 ( v^2_i u_j^-v^2_j u_i^ ) v^2_k + & + & _ 12 ( v^1_i u_j^-v^1_j u_i^ ) v^2_k + _ 12 ( v^2_i u_j^-v^2_j u_i^ ) v^1_k . in section [ piii ] , we studied possible non - zero eigenvalues @xmath314 , @xmath315 , of @xmath6 which are compatible with equations ( [ bia33 ] ) and ( [ bia35 ] ) , providing that for every pair @xmath129 , @xmath160 ( @xmath161 ) there exists @xmath162 for which @xmath163 .
it turned out that then it follows from ( [ bia33 ] ) and ( [ bia35 ] ) that @xmath164 and @xmath189 . here
we study all other possible cases in 5d and show that the only case with non - vanishing eigenvalues of @xmath6 is @xmath164 and @xmath189 .
corresponding form of @xmath34 , @xmath196 , @xmath6 and @xmath9 is given in section [ piii ] .
in 5d , @xmath316 can have values 0 , 1 , 2 or 3 .
we will treat these cases separately ( the cases @xmath317 and @xmath130 can be easily solved in an arbitrary dimension ) . 1 .
the case @xmath317 ( globally ) corresponds to @xmath255 and one of the ricci equations gives @xmath318 , which implies ( take trace of both sides ) that @xmath319 .
thus in an arbitrary dimension , as in 4d , there are no non - expanding , twisting spacetimes .
2 . if @xmath130 , i.e. , @xmath320 , then @xmath256 , @xmath260 and @xmath321 . from ( [ cbs11 ] ) and ( [ cbs12 ] ) ,
it follows @xmath261 and @xmath258 .
equations ( [ bia01 ] ) and ( [ cbs13 ] ) give @xmath259 . from ( [ cbs14 ] ) and ( [ trb ] )
, it follows @xmath322 and @xmath323 .
thus , the case @xmath130 leads to a type n or o spacetime .
3 . in the case
@xmath164 , @xmath324 and thus , thanks to antisymmetry , @xmath258 and @xmath261 . then from ( [ bia01 ] ) @xmath325 and from ( [ cbs13 ] )
@xmath326 or @xmath327 .
+ let us proceed further assuming that at least one non - vanishing @xmath328 exists : + from ( [ cbs14 ] ) and ( [ bia02 ] ) , it follows that @xmath329 is equivalent to @xmath330 .
similarly , from ( [ cbs14 ] ) and ( [ trb ] ) , @xmath331 is equivalent to @xmath332 . + the case @xmath333 was analyzed in section [ piii ] and leads to @xmath189 .
+ if @xmath334 and @xmath335 then ( [ bia02 ] ) gives @xmath260 and from ( [ cbs14 ] ) and ( [ trb ] ) , b_= _
11=_22 = -2(_11+_22 ) , follows @xmath189 .
next we have to analyze possible cases with @xmath257 @xmath336 .
+ let us now assume that all @xmath328 s are zero and at least one @xmath337 is non - vanishing . from ( [ cbs16 ] ) for @xmath276 and @xmath338 and from ( [ tra2 ] )
we get & & _ 2 a_1=2(3_1-_2)_122 , ( -3_1+_2 ) a_1=-8_2_122,[a1 ] + & & _ 1 a_2=2(3_2-_1)_211 , ( -3_2+_1 ) a_2=-8_1_211.[a2 ] if @xmath339 ( and thus also @xmath340 ) then the ratio of equations ( [ a1 ] ) gives @xmath341 which has solutions _ 2=_1 _ 2=-3_1.[a1sol ] similarly , if @xmath342 , @xmath343 then the ratio of equations ( [ a2 ] ) gives @xmath344 with solutions @xmath345 or @xmath346 .
thus for @xmath339 and @xmath342 , the only solution is @xmath291 .
+ if only one @xmath264 , without loss of generality @xmath347 and @xmath348 , then @xmath349 and from ( [ tra ] ) @xmath350 and from ( [ bia5-f ] ) it follows @xmath351 . for the second solution in ( [ a1sol ] ) , @xmath352 , from ( [ a1 ] ) and ( [ tra ] ) we obtain @xmath353 and @xmath354 , respectively , and thus this case does not occur .
+ if @xmath355 then ( [ a1 ] ) , ( [ a2 ] ) imply that all @xmath337 s are zero and from ( [ tra ] ) also all @xmath356 s vanish , which leads to a type n or o spacetime .
4 . for @xmath303 ,
the only components of @xmath34 , @xmath196 and @xmath357 are @xmath358 , @xmath359 and @xmath360 , respectively .
+ from ( [ cbs16 ] ) with @xmath361 we obtain the equation a_a=_abb_b= a that together with ( [ tra ] ) gives a_1&=&_122=_133=4(_122+_133 ) , [ 3m_a_1 ] + a_2&=&_211=_233=4(_211+_233 ) , [ 3m_a_2 ] + a_3&=&_311=_322=4(_311+_322 ) .
[ 3m_a_3 ] + \(a ) if all @xmath266 then the only solution is @xmath189 and @xmath312 , which corresponds to @xmath164 discussed above .
+ \(b ) let us now study the case with one @xmath359 vanishing ( we can assume @xmath362 without loss of generality ) . then from ( [ 3m_a_3 ] )
either @xmath363 and @xmath364 , which implies @xmath164 ,
@xmath312 , @xmath189 , or @xmath365 and then from ( [ 3m_a_1 ] ) and ( [ 3m_a_2 ] ) we get 3_1 ^ 2-_2 ^ 2-_3 ^ 2 + 2(-_1_2 -_1_3+_2_3)&=&0 , [ eqfroma1 ] + -_1 ^ 2 + 3_2 ^ 2-_3 ^ 2 + 2(-_1_2 + _ 1_3-_2_3)&=&0 .
their difference gives 4(_1-_2)(_1+_2-_3)=0 and thus either @xmath366 and from ( [ eqfroma1 ] ) @xmath312 or @xmath367 and from ( [ eqfroma1 ] ) @xmath368 .
this case thus again leads to @xmath164 .
+ \(c ) if only one @xmath359 is non - vanishing ( we again choose @xmath369 and @xmath370 without loss of generality ) then ( [ 3m_a_2 ] ) , ( [ 3m_a_3 ] ) imply either @xmath371 and @xmath372 , which is not possible , or @xmath371 , which gives @xmath189 and @xmath312 , or @xmath372 , which gives @xmath373 and @xmath374 , or @xmath375 .
all these cases , except the last one are inconsistent with our assumption @xmath303 .
we thus need to check the last case @xmath376 , which has two possible branches corresponding to @xmath310 and @xmath377 .
+ \(a ) @xmath310 : from ( [ cbs16 ] ) ( in the form ( [ mcabmcba ] ) ) ( -_1 + 3_2+_3)_123&=&(-_1+_2 + 3_3)_132,[mabc1 ] + ( 3_1-_2+_3)_213&=&(_1-_2 + 3_3)_231,[mabc2 ] + ( 3_1+_2-_3)_312&=&(_1 + 3_2 - 3_3)_321,[mabc3 ] we get @xmath312 and @xmath378 for @xmath379 , or @xmath380 .
then from ( [ bia3-a ] ) we get @xmath321 and @xmath381 which together with ( [ 3m_a_1 ] ) gives @xmath382 , @xmath383 .
+ \(b ) @xmath377 : from ( [ mabc1])([mabc3 ] ) and ( [ bia03 ] ) we get _ 1 ^ 2+_2 ^ 2+_3 ^ 2 - 2(_1_2+_1_3+_2_3)=0[mabc ] with the solution @xmath384 . by substituting this result into ( [ eqfroma1 ] )
we get @xmath385 and thus either @xmath383 and @xmath382 or @xmath189 and @xmath386 ( @xmath387 does not satisfy ( [ eqfroma1 ] ) ) .
+ \(d ) if all @xmath272 then from ( [ 3m_a_1])([3m_a_3 ] ) either one of @xmath388 and the other two are equal to @xmath197 or all @xmath389 .
then if + \a ) @xmath310 from ( [ mabc1])([mabc3 ] ) and ( [ bia03 ] ) we get either one of @xmath388 and other two are again equal to @xmath197 or all @xmath256 , which corresponds to a type n or o spacetime .
+ \b ) if @xmath377 then from ( [ mabc1])([mabc3 ] ) and ( [ bia03 ] ) we get ( [ mabc ] ) .
further from ( [ bia3-a ] ) we get all @xmath321 and then & & _ 2_123=_3_132 , + & & _ 1_123=-_3_231 , + & & _ 1_132=_2_231 , which yields & & 0=(_2-_3)(_1-_2-_3 ) , + & & 0=-_1 ^ 2+_1_2 + 3_1_3 + 2_2_3 - 2_3 ^ 2,[bia3al2 ] + & & 0=-_1 ^ 2 + 3_1_2+_1_3 + 2_2_3 - 2_2 ^ 2 , with the solution either @xmath390 ( then ( [ bia3al2 ] ) implies @xmath391 or @xmath392 is zero ) or @xmath393 and ( [ bia3al2 ] ) implies @xmath383 or @xmath394 ; however , from ( [ bia5-a ] ) we get @xmath395 which is contradiction .
the five - dimensional rotating black - hole metric in boyer - lindquist coordinates has the form @xcite ( we use the notation of @xcite ) ds^2=&&x^2 + ^2 d ^2 -dt^2 + ( x+a^2 ) ^2 ^2 + ( x+b^2 ) ^2 ^2 + & & + ( dt + a ^2 + b ^2 ) ^2 , where ^2=x+a^2 ^2 + b^2 ^2 , = ( x+a^2)(x+b^2)-r_0 ^ 2 x .
after normalizing two null vectors @xmath396 and @xmath397 given in ( 4.23 ) in @xcite , one can appropriately choose the rest of the frame vectors @xmath398 , @xmath399 and @xmath400 .
it turns out that all components of the weyl tensor with the boost weights 2 , 1 , -1 , -2 vanish and hence the spacetime is of the algebraic type d @xcite .
one can also explicitly calculate the matrix @xmath6 . while the form of @xmath6 depends on the choice of @xmath398 , @xmath399 and @xmath400 , the characteristic polynomial of @xmath6 , @xmath401 ,
for @xmath401 we obtain p _ ( ) = ( - ) ( - ) ^2 and the diagonal form of @xmath6 is _ diag= ( ccc & 0&0 + 0&&0 + 0&0 & ) from ( [ shearmatrix ] ) it follows that shear is = ( 1- ) , which is in accordance with equation ( 4.25 ) in @xcite . note that for @xmath402 all eigenvalues of @xmath6 are equal and shear is thus zero .
whereas for @xmath403 the matrix @xmath6 has two equal and one distinct eigenvalue and shear does not vanish .
note also that similarly as for the types iii and n , this algebraically special vacuum solution has non - vanishing shear ( see also footnote 4 in paper @xcite ) .
moreover , in this spacetime @xmath6 does not have properties proved in sections [ typen ] and [ piii ] for vacuum type n and type iii spacetimes .
99 r. milson , a. coley , v. pravda and a. pravdov , _ alignment and algebraically special tensors in lorentzian geometry _ , gr - qc/0401010 .
e. t. newman and r. penrose , _ an approach to gravitational radiation by a method of spin coefficients _ , j. math . phys . * 3 * , 566 ( 1962 ) .
a. coley , r. milson , v. pravda and a. pravdov , _ classification of the weyl tensor in higher dimensions _ , class .
quantum grav .
* 21 * , l35 ( 2004 ) .
j. n. goldberg and r. k. sachs , _ a theorem on petrov types _ , acta .
polon . , suppl , * 22 * , 13 ( 1962 ) .
h. stephani , d. kramer , m. maccallum , c. hoenselaers and e. herlt , exact solutions to einstein s field equations , second edition , cambridge university press , cambridge ( 2003 ) .
v. pravda , a. pravdov , a. coley and r. milson , _ all spacetimes with vanishing curvature invariants _ , class .
quantum grav .
* 19 * , 6213 ( 2002 ) .
a. coley , r. milson , n. pelavas , v. pravda , a. pravdov and r. zalaletdinov , _ generalizations of pp - wave spacetimes in higher dimensions _ ,
d * 67 * , 104020 ( 2003 ) .
a. coley , r. milson , v. pravda and a. pravdov , _ higher dimensional vsi spacetimes _ ( in preparation ) .
r. c. myers and m. j. perry , _ black holes in higher dimensional space - times _ , ann .
phys . * 172 * , 304 ( 1986 ) .
v. frolov and d. stojkovi , _ particle and light motion in space - time of a five - dimensional rotating black hole _ ,
d * 68 * , 064011 ( 2003 ) ; v. frolov and i. novikov , _ black hole physics : basic concepts and new developments _ , kluwer , dodrecht ( 1998 ) . | a higher dimensional frame formalism is developed in order to study implications of the bianchi identities for the weyl tensor in vacuum spacetimes of the algebraic types iii and n in arbitrary dimension @xmath0 .
it follows that the principal null congruence is geodesic and expands isotropically in two dimensions and does not expand in @xmath1 spacelike dimensions or does not expand at all .
it is shown that the existence of such principal geodesic null congruence in vacuum ( together with an additional condition on twist ) implies an algebraically special spacetime .
we also use the myers - perry metric as an explicit example of a vacuum type d spacetime to show that principal geodesic null congruences in vacuum type d spacetimes do not share this property . |
the accumulated experience in helioseismology of inverting mode frequency data provides a good starting point for asteroseismic inversion . in some rather superficial ways
the circumstances of the two may seem very different : helioseismologists can use many more mode frequencies than will ever be possible for a more distant star , so that the resolution that can be achieved in helioseismic inversion is beyond the grasp of asteroseismology ( an exception may be situations such as where modes are trapped in a narrow range of depth within the star ) .
another difference is that global parameters of the sun such as its mass , radius and age are much better known than they are for other stars ; hence the structure of the sun is constrained _ a priori _ much more accurately than it is for a distant star , even if the input physics had been known precisely , which of course is not the case for the sun or for other stars . in more fundamental ways , though , helioseismic and asteroseismic inversion are much more similar than they are different .
helioseismology has not always been blessed with such a wealth of data , and helioseismologists in the early days of their subject learned the value of model calibration and asymptotic description ( in particular of low - degree modes ) for making inferences about the sun .
they also learned some of the dangers and limitations of drawing inferences from real data .
the modal properties in many asteroseismic targets will be similar to the low - degree solar modes .
moreover , the principles of inversion are the same in both fields : obtaining localized information about the unseen stellar interior ; assessing resolution ; taking account of the effects of data errors ; assessing what information the data really contain about the object of study ; judiciously adding additional constraints or assumptions in making inferences from the data .
this paper touches on a few of these points with regard to inverting asteroseismic data , including the usefulness of optimally localized averaging ( ola ) kernels and model calibration using large and small frequency separations for solar - type oscillation spectra .
much more detail on those two topics can be found in the papers by basu et al .
( 2001 ) and monteiro et al .
( 2001 ) , both in these proceedings .
the inversion problem predicates that one is able to perform first the forward problem . in the present context ,
that means the computation of the oscillation data ( the mode frequencies ) from the assumed structure of the star . solving
the forward problem always involves approximations or simplifying assumptions , because we can not model the full complexities of a real star .
the inverse procedure , inferring the structure ( or dynamics ) of the star from the observables , is ill - posed because , given one solution , there will almost invariably formally be an infinite number of solutions that fit the data equally well .
the art of inverse theory is in no small part concerned with how to select from that infinity of possibilities .
actually , in various asteroseismic applications to date , e.g. to @xmath0 boo and certain @xmath1 scuti stars , the problem is rather that one has so far been unable to find even a single solution that fits the data ( see christensen - dalsgaard , bedding & kjeldsen 1995 , pamyatnykh et al .
this may indicate that the some of the approximations made in the forward problem are inappropriate ; it could also indicate that the errors in the data have been assessed incorrectly . in the present work ,
unless otherwise stated , we assume both that the forward problem can be solved correctly and that the statistical properties of the data errors are correctly known .
a relatively straightforward approach to the inverse problem is model calibration . at its conceptually simplest ,
this entails computing the observables for a set of models , possibly a sequence in which one or more physical parameters vary through a range of values , and choosing from among that set the one that best fit the data . `
best ' here is often taken to mean that model which minimizes the chi - squared value of a least - squares fit to the data .
given today s fast computers , searches through large sets of models are feasible , either blindly or by using some search algorithm such as genetic algorithm or monte carlo . approaches similar to this have been undertaken for white dwarf pulsators ( metcalfe , nather & winget 2000 ; metcalfe 2001 ) @xmath1 scuti stars ( pamyatnykh et al .
1998 ) , and pulsating sdb stars ( charpinet 2001 ) .
depending on the way in which the method is applied , the resulting model may be constrained to be a member of the discrete set of calibration models , or could lie `` between '' them if interpolation in the set of models is permitted .
model calibration is powerful and dangerous .
it is powerful because it allows one to incorporate prejudice into the search for a solution and , suitably formulated , it can always find a best fit .
it is dangerous for the same reaons : perhaps unwittingly on the part of the practitioner , it builds prejudice into the space of solutions that is considered . also , even a satisfactory fit to the data does not mean that the solution model is necessarily like the real star : one can make a one - parameter model calibration to a single datum , but it is unclear what aspects of the resulting stellar model the datum is actually able to constrain .
examples of prejudice that model calibration may incorporate , for good or ill , are the choice of physics used to construct the models , and possible assumptions about the smoothness of the structure of the star .
an insidious problem is that , if the approximations in the forward modelling introduce errors into the computed model observables , this can introduce a systematic error into the result of the model calibration .
the sun and solar - type stars provide a good example . here
the near - surface structure and the mode physics in that region are poorly modelled at present , and the simple approximations made introduce a systematic shift in the computed frequencies : low - frequency p modes have their upper turning point relatively deep in the star and are almost unaffected by the treatment of the surface layers , whereas modes of higher frequency have turning points closer to the surface and the error in their frequencies grows progressively bigger . in these circumstances ,
it is preferable not to calibrate to the frequency data themselves but rather to data combinations chosen to be relatively insensitive to the known deficiency in the forward modelling . in the near - surface layers , the eigenfunctions of solar p modes of low or intermediate degree @xmath2 are essentially independent of @xmath2 ,
so the error in the frequencies introduced by incorrectly modelling this region is just a function of frequency , scaled by the inverse of the mode inertia .
this suggests that one should calibrate frequencies of solar - type stars to data combinations chosen to be insensitive to such an error .
asymptotic analysis suggests other data combinations which can be used for model calibration and which are more discriminating than the raw frequency data .
monteiro et al . ( 2001 ) consider in some detail the use of the so - called large and small separations : @xmath3 respectively . as monteiro
demonstrate , the results of such a calibration , even for such global characteristics as the mass and age of the star , depend on the other physics assumed .
model calibration is just one approach to the inverse problem of inferring the stellar structure , which we may indicate schematically as @xmath4 , from the frequency data @xmath5 .
how else may the nonlinear dependence @xmath6 be inverted ?
a way which then allows application of a variety of techniques is to linearize about a reference model @xmath7 : inversion takes as input @xmath8 , the difference between the observed data and the corresponding values that model @xmath7 predicts , and produces as output @xmath9 , the estimated difference in structure between the star and @xmath7 . since the structure of @xmath7 is known , the structure of the star can then be reconstructed . of course , that is a naive hope , because of the inherent nonuniqueness discussed earlier , quite apart from issues of data errors .
but one may at least produce a refinement on the initial model and indeed this new model can then be used as reference for a subsequent inversion as the next step of an iterative approach .
depending on the technique adopted , it is helpful and may be essential to have a reasonable starting guess in the form of the reference model @xmath7 .
model calibration using the large and small separations is a reasonable way to find such a model for solar - type stars .
model calibration produces a solution @xmath7 that is in the span ( suitably defined ) of the calibration set of models , illustrated schematically as a surface in fig .
[ fig : modelspace ] .
inversion of the kind just described may then be used to proceed from @xmath7 to a new model @xmath10 which may be ` close ' to @xmath7 but outside the span of the original calibration models .
one hopes that the final model depends only weakly on the initial model , so that if by some other calibration ( assuming different physics when calibrating the large and small separations , for example ) one produces some other model @xmath11 , then the inversion step takes one close once more to the same final model @xmath10 .
one can imagine making all structural quantities in both the reference model and the target star dimensionless by taking out appropriate factors of the gravitational constant @xmath12 and the stars masses @xmath13 and radii @xmath14 .
then the frequency differences @xmath15 between the same mode of order @xmath16 and degree @xmath2 in the two stars can be related to the differences in dimensionless structural quantities by an equation such as @xmath17 where here the choice has been made to express the structural differences in terms of @xmath18 , the ratio of pressure @xmath19 to density @xmath20 , and @xmath21 , the helium abundance by mass : for discussion of this choice , see basu et al .
( 2001 ) . here
, @xmath22 is the difference in dimensionless @xmath18 between the two stars , the differences being evaluated at fixed fractional radius @xmath23 in the star .
the first term on the right - hand side is a constant and just reflects the @xmath24 homologous dependence of the frequencies .
the final term is some function of frequency , divided by mode inertia @xmath25 , which absorbs uncertainties from the near - surface layers .
for some details of the construction of the kernel functions @xmath26 and @xmath26 , which are known functions derivable from the reference model , see the appendix . in an inversion
, the left - hand side will be known ; all terms on the right - hand side are to be inferred , including the difference in @xmath27 between the reference model and target star .
to motivate the rest of the paper we consider first the structural differences between a few stellar models .
these will already indicate that rather small uncertainties in global parameters of stars , e.g. mass or age , can lead to large uncertainties in their structure .
of course the positive side of that is that the stellar structure is sensitive to those parameters and so , if the observable mode frequencies are in turn sensitive to those aspects of the structure then we may have some hope of using the observations to constrain e.g. the mass and age of the star rather precisely .
-0.2 cm panel ( a ) of figure [ fig : prhodiffs ] shows the relative differences in pressure @xmath19 and density @xmath20 , at fixed fractional radius , between two zams stars , of masses @xmath28 and @xmath29 .
we note that even for two stars with rather similar masses the differences are large , of order unity .
panel ( b ) shows the corresponding differences after the homology scaling has been taken out : this scaling is assumed taken out in the formulation presented in eq .
( [ eq : diffeq ] ) .
the effect is essentially to shift the two curves by a constant : although the differences are smaller , they are still large .
these changes arise from nonhomologous differences in the surface layers which change the entropy of the convection zone . indeed , writing @xmath30 in the stars convective envelopes ( in this context only , @xmath16 denotes polytropic index ) , and noting that sound speed there is essentially determined by surface gravity , one finds that in that region @xmath31 possibly of more direct relevance for inversion of eq .
( [ eq : diffeq ] ) is the difference in adiabatic sound speed @xmath32 ( @xmath33 , where @xmath34 is the first adiabatic exponent ) .
[ fig : fig3 ] shows , the relative differences in @xmath35 are smaller than the differences in pressure and density , but still quite substantial for stars that differ in mass by as little 10 per cent .
homology scaling has little effect on the differences in this case .
for a reasonable range of masses , these differences scale linearly with the mass difference , so for example the corresponding scaled differences between @xmath36 and @xmath28 zams stars are half those illustrated here , to a very good approximation .
a number of different linearized inversion methods have been developed in helioseismology , geophysics and diverse other areas of inversion applications .
two flexible approaches are optimally localized averages ( ola ) and regularized least squares ( see e.g. christensen - dalsgaard et al .
1990 for a description ) .
the ola method explicitly constructs a linear combination of kernels that is localized at some location in the star and is small elsewhere : the corresponding linear combination of relative frequency differences is then a measure of the localized average of the structural differences .
the application of sola in the asteroseismology of solar - type stars is discussed and illustrated by basu et al .
we emphasize here that ola reveals the true extent to which the seismic data alone can resolve the aspect of the stellar interior under study .
methods which on specific classes of problems appear to have superior performance to ola in resolving aspects of the stellar interior are introducing nonseismic information or assumptions in addition to the frequency data .
the form of least - squares method used most extensively in helioseismology is regularized least - squares : the idea is to represent the solution with a set of basis functions more finely than can be resolved by the data and with ideally no bias about the form of the solution built into the basis ; but then to minimize the sum of chi - squared fit to the data and a penalty term which is large if the solution has undesirable characteristics .
the most used penalty function is the integral over the star of the squared second derivative of the function under study with respect to radius . a different way of regularizing the least - squares solution , without introducing a penalty term , is to choose a drastically smaller set of basis functions .
this alternative , which we do not claim is intrinsically superior , may give apparently better results from few data if the basis functions are chosen with appropriate intuition or good fortune .
the reason is that one can introduce a huge amount of prejudice into the solution by forcing it to have a form determined by the basis functions . such a basis could , for example , force the buoyancy frequency to be zero in a convective core and permit a single discontinuity at the core boundary but not elsewhere : such assumptions may be reasonable , but it should be realized that they are additional to the seismic data .
if the star actually had a second discontinuity , or a more gradual variation at the core boundary , such an inversion would not generally reveal those features .
a helioseismic example is in finding the location of the base of the convective envelope , where remarkable precision ( @xmath37 taking into account uncertainties in abundance profiles , @xmath38 if the only uncertainty comes from data noise ) has been claimed ( basu & antia 1997 ) : this is credible only insofar as the base of the convection zone has precisely the form assumed in the inversion , because the true resolution at that location is much coarser than that .
the true resolution ( e.g. vertically ) is essentially limited by the reciprocal of the largest vertical wavenumber of the eigenfunctions corresponding to the available data ( thompson 1993 ) .
as a simple illustration of the apparent ability of least - squares inversion with a limited basis to infer structure even where the mode set has little resolving power , we take a basis of five functions chosen for algebraic convenience .
they are approximately ( but only approximately ) able to represent the difference in @xmath18 between zams models of solar - like stars .
the basis represents a function that is quadratic in @xmath39 for @xmath40 with zero derivative at @xmath41 , piecewise constant between radii @xmath42 , @xmath43 and @xmath44 , quadratic between @xmath44 and @xmath45 and zero for @xmath46 .
the values of @xmath44 and @xmath45 , which are initially set to @xmath47 and @xmath48 respectively , are adjusted by hand to find a minimum of the chi - squared fit . in the first of two illustrative applications
, we consider the inversion of frequency data ( 65 modes : @xmath49 , @xmath50 ; @xmath51 , @xmath52 , @xmath53 , @xmath54 , @xmath55 , @xmath56 ) from a @xmath36 zams star . in a real application
we would first calibrate the star by computing large and small separations from the data and using the approach of monteiro et al .
( 2001 ) to arrive at a reference model ; in fact we simply took a reference model which was a @xmath28 zams star . before discussing the results of the inversion ,
we first consider the data , i.e. the relative frequency differences , shown in fig .
[ fig : fig4 ] .
( for clarity , no noise has been added to the data , though it was added for the inversions . )
the dominant trends are that the values are negative , around @xmath57 , because of the difference in @xmath27 between the two stars ( in fact , @xmath58 for these two stellar models ) ; and there is a roughly linear trend with frequency , in this frequency range , coming from near - surface differences .
these two contributions can be estimated and removed by fitting an expression of the form ( [ eq : diffeq ] ) : for illustration , we show in fig . [ fig : fig5 ] what would remain : this is the signal from the interior , which contains the information that the inversion will use to infer conditions inside the star .
the impression now is that the data contain a signal which has some oscillatory component ( from the rather abrupt change in the structural differences at the base of the convective envelope ) but is otherwise only weakly a function of frequency and increases with decreasing @xmath2 , indicating that the more deeply penetrating ( i.e. lower-@xmath2 modes ) sense @xmath59 increasing in the deep interior as one gets closer to the centre of the star .
these features are indeed revealed by the inversion ( fig .
[ fig : fig6 ] ) , which compares the exact @xmath18-differences with the least - squares solutions for noise - free data and for data with gaussian noise with zero mean and uniform standard deviations @xmath60hz ,
@xmath61hz , and @xmath62hz .
the noise realization in the two panels is different , but within each panel the noise differs from case to case only by a multiplicative scaling .
it can be seen that for low noise levels this very small basis enables the differences to be recovered rather well , including the structure beneath the convection zone and the variation of @xmath22 in the core . for larger noise levels the artificiality of the basis functions
becomes more apparent ; also the solution for the higher noise levels is rather different for the two noise realizations , which gives some indication of the uncertainty even in this highly constrained solution .
a second example is an application to the same mode set but with data from model @xmath63 of monteiro et al .
( 2001 ) and gaussian data errors with @xmath64hz .
this star is slightly more massive and more evolved than the sun .
again , we omitted the calibration step and inverted relative to a reference model of the present - age sun .
the results are shown in fig .
[ fig : fig7 ] .
again the qualitative behaviour of the differences is recovered reasonably , including the downturn in the core .
the discrepancies are perhaps partly attributable to the fact that the basis functions are not so well suited for representing this case as the zams case .
it is a prejudice of some stellar astrophysicists ( it was indeed expressed a few times in cordoba during the workshop ) that helioseismic experience with the sun provides a poor example when it comes to asteroseismic inversion .
but the similarities between the two applications are much more significant than their differences .
an extremely important aspect of inversion in any context is to assess what the data really tell you and what information is being introduced by other assumptions or constraints .
ola kernels ( basu et al . 2001 ) indicate what resolution can really be achieved without additional assumptions .
but other approaches can advance our knowledge by allowing introduction of reasonable prejudices : e.g. , looking for signatures of sharp features , or introducing specific basis functions .
the best apparent results are likely to be achieved if those functions are physically motivated , because the solution will accord with our physical intuition ( prejudice )
. this may of course be dangerous .
our example of a highly constrained least - squares inversion illustrates that qualitatively reasonable results can be obtained throughout a star by introducing assumptions about the form of the solution , even in regions where in fact the mode set provides little or no localized information ( cf .
basu et al .
one may have other grounds on which to believe that such a solution is plausible , but on the basis of the data alone it should be viewed sceptically .
model calibration is a useful tool in its own right and for obtaining possible starting models for linearized asteroseismic inversions .
carefully used , model calibration allows one to build in some prejudice ; and the combination of calibration and inversion extends the space of solutions that one explores . calibrating on large and small separations can be effective ( monteiro et al . 2001 ) , always assuming that one has not neglected some important aspect of the physics : in that regard , the effects of rotation and magnetic fields need to be borne in mind ( see dziembowski & goupil 1998 ) , in solar - type stars as in many other pulsating stars .
finally we note that mode identification may be a problem , even in solar - type stars .
the asymptotic pattern of high - order p - mode frequencies may allow @xmath2 to be determined , but there may be some uncertainty in @xmath16 .
such an uncertainty can be allowed for in the inversion by adding to the right - hand side an extra term @xmath65 where @xmath12 is some function of frequency .
we have verified that , with our zams example , even restricting @xmath12 to be a constant function removes quite satisfactorily the effect of a misidentification of @xmath16 .
more generally , @xmath12 could judiciously be chosen to reflect the variation of the large separation with frequency .
this work was supported by the danish national research foundation through the establishment of the theoretical astrophysics center , and by the uk particle physics and astronomy research council .
basu , s. , antia , h. m. , 1997 , mnras , 287 , 189 basu , s. , christensen - dalsgaard , j. , thompson , m. j. , 2001 , these proceedings charpinet , s. , 2001 , in proc .
iau colloq .
185 , radial and nonradial pulsations as probes of stellar physics , eds aerts , c. , bedding , t. r. & christensen - dalsgaard , j. , in press christensen - dalsgaard , j. , bedding , t. r. , kjeldsen , h. , 1995 , apj , 443 , l29 christensen - dalsgaard , j. , schou , j. , thompson , m. j. , 1990 , mnras , 242 , 353 dziembowski , w. a. , goupil , m .- j . , 1998 , in proc .
`` workshop on science with a small space telescope '' , eds kjeldsen h. & bedding t.r . , aarhus university , p. 69
gough , d. o. , 1993 , in astrophysical fluid dynamics , les houches session xlvii , eds zahn , j .- p . & zinn - justin , j. , elsevier , amsterdam , p. 399
gough , d. o. , thompson , m. j. , 1991 , in solar interior and atmosphere , eds cox , a. n. , livingston , w. c. & matthews , m. , space science series , university of arizona press , p. 519 kosovichev , a. g. , 1999 , journal of computational and applied mathematics , 109 , 1 metcalfe , t. s. , 2001 , computational asteroseismology , ph.d .
thesis , university of texas at austin , usa metcalfe , t. s. , nather , r. e. , winget , d. e. , 2000 , apj , 545 , 974 monteiro , m. j. p. f. g. , christensen - dalsgaard , j. , thompson , m. j. , 2001 , these proceedings pamyatnykh , a. a. , dziembowski , w. a. , handler , g. , pikall , h. , 1998 , a&a , 333 , 141 thompson , m. j. , 1993 , in proc .
gong 1992 : seismic investigation of the sun and stars , ed .
brown , t. m. , asp conference series , san francisco , vol .
the derivation of kernels relating the linearized differences in structure to the differences in frequency ( cf .
eq . [ eq : diffeq ] ) has been discussed by , e.g. , gough & thompson ( 1991 ) , gough ( 1993 ) , and kosovichev ( 1999 ) .
as all those authors show , the kernels for either of the pairs of variables @xmath66 or @xmath67 are quite straightforward to derive from the equations of linear adiabatic oscillations together with the linearized equation of hydrostatic support .
obtaining kernels for various other pairs , including the pair @xmath68 used in this paper , can be accomplished by first obtaining kernels for one of the other two pairs and then using the following piece of manipulation .
it is sufficiently ubiquitous ( occurring often when one wishes to transform from kernels for a pair including @xmath20 to some other variable pair ) that we write it rather generally .
let @xmath69 be a solution of @xmath70 where prime denotes differentiation with respect to @xmath71 ( or @xmath23 if everything including @xmath71 is expressed in dimensionless variables ) , for a given function @xmath72 ( so @xmath73 is a functional of @xmath74 ) , with boundary conditions @xmath75 and @xmath76 .
then , provided @xmath77 and @xmath78 , where @xmath79 is the mass interior to radius @xmath71 , and letting @xmath80 denote integration from @xmath81 to @xmath82 ( @xmath41 to @xmath83 ) , @xmath84 note that @xmath77 holds in general ; and in our present application , we scale all structural quantities by @xmath12 , @xmath13 and @xmath14 to make them dimensionless , so @xmath78 is forced to be true .
these two conditions also mean that @xmath85 is zero , so any multiple of @xmath86 may be added to a kernel multiplying @xmath87 .
note that such an additional contribution to @xmath72 makes no change to the right - hand side of eq .
( [ eq : a1 ] ) and hence the contribution to @xmath73 from such an addition is zero .
obtaining kernels for @xmath21 additionally requires an assumption about the equation of state through @xmath34 , since the oscillations do not know directly about the chemical abundances . in the following we write @xmath88
then for convenience we record the following transformations : @xmath89 with @xmath90 ; @xmath91 with @xmath92 ; and @xmath93 with @xmath94 . | some issues of inverting asteroseismic frequency data are discussed , including the use of model calibration and linearized inversion .
an illustrative inversion of artificial data for solar - type stars , using least - squares fitting of a small set of basis functions , is presented .
a few details of kernel construction are also given . |
the unremitting interest in investigations of atmospheric acoustic - gravity waves ( agw ) over more than four decades dating back to hines pioneering work ( hines , 1960 , 1967 ) is dictated by the important role played by these waves in the dynamics of the earth s atmosphere .
these research efforts have been addressed in a large number of publications , including a series of through reviews ( hocke and schlegel , 1996 ; oliver et al . , 1997 ) .
agw typically show up in the ionosphere in the form of traveling ionospheric disturbances ( tids ) and are detected by various radio techniques .
tids are classified as large- and medium - scale disturbances differing by their horizontal phase velocity which is larger ( in the large - scale case ) or smaller ( for the medium scale ) velocity of sound in the lower thermosphere ( on the order of 300 m / s ) , with periods within 0.53.0 h and 1040 min , respectively .
medium - scale tids ( ms tids ) are observed predominantly during the daytime hours and are associated with agw which are generated in the lower atmosphere .
large - scale tids are predominant in the night - time hours and are closely associated with geomagnetic and auroral activity .
it is known that the sources of medium - scale agw can include natural processes of a different origin : magnetic storms , auroral phenomena , weather factors , tropospheric turbulence and jet flows , the solar terminator , strong earthquakes , volcanic eruptions , as well as anthropogenic influences ( rocket launchings , explosions , nuclear tests ) . as a consequence
the observed picture of the electron density disturbance is essentially a net interference wave field of the agw of a different origin .
identifying of the agw of a definite type from this field is a highly involved and generally an almost intractable problem .
the most reliable measurements of the main parameters of medium - scale agw ( parameters of the wave vector of the agw , spectral and dispersion characteristics , etc . ) can therefore be made only for a very rare , unusual type of ms tids , i.e. quasi - periodic ( monochromatic ) oscillations which are sometimes recorded as corresponding variations of the frequency doppler shift @xmath0 of the ionosphere - reflected hf radio signal ( davies and jones , 1971 ; waldock and jones , 1987 ; jacobson and carlos , 1991 ; yakovets et al . , 1999 ) .
experiments of this kind were instrumental in investigating the spatial - temporal characteristics of ms tids in the form of a wave process , because such recordings are easy to identify visually with monochromatic individual agw .
unfortunately , this was possible to accomplish for a very limited body of experimental data .
thus , jacobson and carlos ( 1991 ) managed to identify only a few monochromatic tids from their data obtained for more than 100 hours of observation .
yakovets et al .
( 1999 ) also recorded only a few realizations of monochromatic tids for two observing periods from the winter months of 1989 and 1990 .
yakovets et al .
( 1999 ) are likely to be the first to use the term `` wave packets '' to designate the quasi - monochromatic variations of @xmath0 , and they made an attempt to explain their origin on the basis of studying the phase structure of the oscillations .
the authors of the cited reference observed two types of @xmath0-variations : quasi - stochastic tids , and monochromatic tids in the form of wave packets .
they arrived at the conclusion that quasi - stochastic tids are characterized by a random phase behavior , a short length of coherence , and by a large vertical phase velocity .
wave packets show quasi - monochromatic oscillations of @xmath0 , a larger length of coherence , and a smaller vertical phase velocity . following yakovets et al .
( 1999 ) , we chose to utilize the term `` wave packets '' by expanding it to the term `` traveling wave packets '' ( twps ) . the investigation made in this paper
has brought out clearly that this designation describes most adequately the phenomenon involved .
some authors associate the variations of the frequency doppler shift @xmath0 with ms tids that are generated during the passage of atmospheric fronts , tornadoes , and hurricanes ( baker and davies , 1969 ; bertin et al .
, 1975 ; 1978 ; hung et al . , 1978 ; kersly and rees , 1982 ; stobie et al . , 1983
; huang et al . , 1985 ) .
it is only in some cases that these experiments observed quasi - monochromatic variations of @xmath0 with periods of about 10 min ( huang et al . , 1985 ) .
thus , in spite of the many years of experimental and theoretical studied , so far there is no clear understanding not only of the physical origin of the quasi - monochromatic ms tids but even of their morphology as well ( the occurrence frequency as a function of geographical location , time , level of geomagnetic and meteorological activity , etc . ) . to address these issues requires obtaining statistically significant sets of experimental data with good spatial resolution in order to study not only the morphological but also dynamic characteristics of quasi - monochromatic ms tids ( the direction of their displacement , their propagation velocity , and the location of the possible disturbance source ) .
another important requirement implies the continuous , global character of observations , because such phenomena are temporally highly rare and spatially random .
such a possibility is , for the first time , afforded by the use of the international ground - based network of two - frequency receivers of the navigation gps system which at the beginning of 2002 consisted of no less than 1000 sites , with its data posted on the internet , which opens up a new era of a global , continuous and fully computerized monitoring of ionospheric disturbances of a different class .
analysis and identification of twps became possible through the use of the technology ( developed at the istp ) for global detection and determination of parameters of ionospheric disturbances of a different class .
the objective of this paper is to study the morphology and spatial - temporal properties of twps using the data from the global network of gps receivers .
section 2 provides general information about the experiment and gives a brief description of the method of twps detecting .
section 3 presents our new evidence characterizing the twps morphology .
section 4 is devoted to a detailed analysis of the spatial - temporal properties of twps by considering an example of the most pronounced manifestation of twps on october 18 , 2001 as observed in california , usa . the discussion of our results compared with the findings reported by other authors is given in section 5 .
this paper presents , for the first time , the morphology of twps for 105 days of 19982001 , with a different level of geomagnetic activity and with the number of stations of the global gps network ranging from 10 to 300 . a total number of he tec series used in the analysis , corresponding to the observation along a single receiver - satellite line - of - sight ( los ) , with a duration of each series of about 2.3 hours , exceeded 700000 . for a diversity of reasons
, slightly differing sets of gps stations were selected for different events to be analyzed ; however , the geometry of experiment for all events was virtually identical .
the stations coordinates are not given here for reasons of space .
this information may be found at the electronic address http://lox.ucsd.edu/cgi-bin/allcoords.cgi ? .
the global gps covers rather densely north america and europe , and to a much lesser extent asia .
gps stations are more sparsely distributed on the pacific and atlantic oceans .
however , such coverage of the surface of the globe makes it possible , already today , to tackle the problem of global detection of disturbances with hither to unprecedented spatial accumulation .
thus , in the western hemisphere the corresponding number of stations can reach no less than 500 already today , and the number of los to the satellite can be no less than 20003000 .
the area of california , usa , is particular convenient for our investigations because of the large number of gps stations ( no less than 300 ) located over a relatively small area , which makes it possible to obtain a great variety of gps arrays of a different configuration for a reliable determination of the dynamic twps parameters using the method of gps interferometry of tids ( afraimovich et al . , 1998 ; 2000 ) .
the comparison of twps characteristics with geomagnetic field variations was based on using the data from the intermagnet network ( intermagnet , http://www.intermagnet.org/ ) .
the standard gps technology provides a means for wave disturbances detecion based on phase measurements of tec @xmath1 ( hofmann - wellenhof et al . , 1992 ) :
@xmath2\ ] ] where @xmath3 and @xmath4 are additional paths of the radio signal caused by the phase delay in the ionosphere , ( m ) ; @xmath5 and @xmath6 represent the number of phase rotations at the frequencies @xmath7 and @xmath8 ; @xmath9 and @xmath10 stand for the corresponding wavelengths , ( m ) ; @xmath11 is the unknown initial phase ambiguity , ( m ) ; and @xmath12 are errors in determining the phase path , ( m ) .
phase measurements in the gps can be made with a high degree of accuracy corresponding to the error of tec determination of at least @xmath13 m@xmath14 when averaged on a 30-second time interval , with some uncertainty of the initial value of tec , however ( hofmann - wellenhof et al . , 1992 ) .
this makes possible detecting ionization irregularities and wave processes in the ionosphere over a wide range of amplitudes ( up to @xmath15 of the diurnal tec variation ) and periods ( from 24 hours to 5 min ) .
the unit of tec , which is equal to @xmath16 m@xmath14 ( @xmath17 ) and is commonly accepted in the literature , will be used in the following . in some instances a convenient way for comparing tec response characteristics from the gps data with those obtained by analyzing the frequency doppler shift in the hf range ( davies and jones , 1971 ; waldock and jones , 1987 ; jacobson and carlos , 1991 ; yakovets et al . , 1999 )
is to estimate the frequency doppler shift @xmath0 from tec series obtained by formula ( 1 ) . to an approximation sufficient for the purpose of our investigation ,
a corresponding relationship was obtained by ( davies , 1969 ) : @xmath18 where @xmath19 stands for the time derivative of tec .
primary data include series of `` oblique '' values of tec @xmath20 , as well as the corresponding series of elevations @xmath21 and azimuths @xmath22 along los to the satellite calculated using our developed convtec program which converts the gps system standard rinex - files on the internet ( gurtner , 1993 ) . for twps characteristics to be determined continuous series of @xmath20 series of a duration of no less than 2.3 hours
are chosen . to normalize the response amplitude we converted the `` oblique '' tec to an equivalent `` vertical '' value ( klobuchar , 1986 ) : @xmath23,\ ] ] where @xmath24 is the earth s radius , and
@xmath25 km is the height of the @xmath26-layer maximum .
the most reliable results from the determination of twps parameters correspond to high values of elevations @xmath21 of the beam to the satellite because sphericity effects become reasonably small .
all results in this study were obtained for elevations @xmath21 larger than 30@xmath27 . to exclude the variations of the regular ionosphere , as well as trends introduced by the motion of the satellite , we employ the procedure of removing the linear trend by preliminarily smoothing the initial series with a convenient time window .
the technology for global detection of twps that was developed at the istp sb ras makes it possible to select from a large amount of experimental material in the automatic mode the tec disturbances which can be assigned to a class of twps .
the selection of tec series which could be ascribed to a class of twps was carried out by two criteria ( fig .
first of all , tec variations were selected , for which the value of the standard deviation exceeded a given threshold @xmath28 ( in the present case @xmath28 = 0.1 @xmath17 ) . in addition , for each filtered series , we verified the fulfillment of the `` quasi - monochromaticity '' condition of tec oscillations , for which the ratio @xmath29 of a total spectral signal power in the selected frequency band @xmath30 in the neighborhood of a maximum value of the power @xmath31 , to a total spectral signal power outside the frequency band @xmath30 under consideration exceeded a given threshold @xmath32 ( in the present case @xmath32 = 2 ) .
1 illustrates the selection process of the twps .
1a gives an example of weakly disturbed variations of the `` vertical '' tec @xmath33 as recorded on july 15 , 2001 at station darw ( @xmath34e ; @xmath35s ; satellite number prn05 ) .
1b presents the @xmath36-variations that were filtered from the initial @xmath33-series .
thin horizontal lines show the specified threshold @xmath28 .
the standard deviation of the @xmath36-variations is 0.019 @xmath17 , that is , does not reach the specified threshold @xmath28 = 0.1 @xmath17 .
1c illustrates the @xmath37 spectrum of the series @xmath36 from fig .
thin vertical lines show the boundaries of the frequency range @xmath30 . for this
spectrum @xmath29 = 0.66 is smaller than the specified @xmath32 = 2 and , hence , the series @xmath36 does not satisfy the condition of `` quasi - monochromaticity '' .
1d , 1e and 1f plots the same dependencies as in fig .
1a , 1b and 1c but for station tow2 ( @xmath38e ; @xmath39s ; satellite number prn09 ) .
it is evident from fig .
1d that at the background of the slow tec variations there are clearly identifiable ( unusual for background tec disturbances ) oscillations in the form of a wave packet of a duration of about 1 hour and with a typical period @xmath40 in the range from 10 to 18 min .
the oscillation amplitude of the detected wave packet exceeds one order of magnitude ( as a minimum ) the intensity of background tec fluctuations of this range of periods ( afraimovich et al . , 2001a ) .
the relative amplitude of such a response @xmath41 is considerable , 4 @xmath42 . as the background value of @xmath1 we used the absolute `` vertical '' tec value of @xmath20 for the site located at @xmath43s ; @xmath38e , obtained from ionex tec maps ( mannucci , 1998 ) . since the main contribution to the modulation of the tec is made by the region near the ionospheric @xmath44-region peak , the relative amplitude of the local electron density disturbance @xmath45 in this region can be several times larger than @xmath41 .
it is worthwhile to note that the two examples described above refer to the same time interval and to the stations spaced by a distance of no more than 1900 km from one another .
this suggests a local character of the phenomenon and is in agreement with the overall sample statistic characterizing its spatial correlation ( see below ) . the standard deviation of the series @xmath36 , shown in fig .
1e , is 0.114 @xmath17 , which is larger than the specified threshold @xmath28=0.1 @xmath17 , and this series satisfies the condition for the standard deviation .
1f presents the spectrum @xmath37 of the series @xmath36 , shown in fig .
1e . for this spectrum @xmath29 = 3.71 , which is larger than the specified @xmath32 = 2 , that is , in this case the series @xmath36 satisfies the condition of `` quasi - monochromaticity '' . panel * e * shows the maximum value of the amplitude @xmath46 of the packet and the time @xmath47 corresponding to this amplitude .
when the filtered di(t)-series satisfied the conditions described above , such an event was recognized as twp . furthermore , for each such event , a special file stored information about the name , geographical latitude @xmath48 and longitude @xmath49 of the gps station ; the gps satellite prn number ; the time @xmath47 corresponding to the maximum value of the amplitude @xmath46 of the twp ; the amplitude @xmath46 ; the twp oscillation period @xmath40 ; the @xmath29 ratio ; and about the value of the elevation @xmath50 and the azimuth @xmath51 of the los to the satellite calculated for the time @xmath47 . the sample statistic , presented below , was obtained by processing such files for our selected value of @xmath28=0.1 @xmath17 .
the method outlined above was used to obtain a series of twps totaling about 1300 cases , or about 0.2@xmath42 of the total number of the radio paths considered ( receiver - satellite los ) . as has been pointed out above
, the radio paths that were considered totaled over 700000 .
an analysis of the resulting statistic revealed a number of dependencies of the twps parameters on different factors .
first of all , we consider the seasonal dependence of the density and amplitude of the twps ( fig .
2 ) . fig .
2a plots the dependence of the number of days of observation @xmath52 on the time of the year .
it is evident that statistically , the autumn season is represented best .
2b shows the seasonal dependence of the number of twps @xmath53 .
2c plots the number of twps @xmath54 per day as a function of time of the year .
this dependence has maxima in winter and in autumn .
the relative twps density @xmath55 , obtained as the ratio of the number of twps @xmath53 to the number of receiver - satellite los , is presented in fig .
as is apparent from the figure , twps are observed in no more than 0.1 - 0.4@xmath42 of the total number of radio paths , and much more frequently in winter ( over 0.4@xmath42 ) and autumn ( up to 0.3@xmath42 ) than in spring and summer ( less than 0.1@xmath42 ) .
diamonds in fig .
2d show the mean values of @xmath56 of the maximum twps amplitudes @xmath46 for each season , and vertical lines show their standard deviations .
thick horizontal line shows the threshold in amplitude @xmath28=0.1 @xmath17 .
the most probable value of @xmath56 with a small scatter varies around the value 0.3 @xmath17 , irrespective of the season .
3d presents the normalized occurrence probability distribution of twps with the specified maximum amplitude of the packet @xmath46 .
the vertical dashed line shows the threshold in amplitude @xmath28=0.1 @xmath17 .
it was found that the most probable value of the amplitude @xmath57 , also shown in fig .
3d , is about 0.3 @xmath17 , and the half - width of the distribution is 0.2 @xmath17 . as was shown by afraimovich et al .
( 2001a ) , the mean values of the tec variation amplitude with the period of 20 min for the magnetically quiet and magnetically disturbed days do not exceed 0.01 @xmath17 and 0.07 @xmath17 , respectively .
thus the most probable value of the amplitude @xmath57 of twps exceeds the mean value of the tec variation amplitude by a factor of 46 as a minimum .
this estimate is consistent with the variation amplitude of the frequency doppler shift @xmath0 reported by yakovets et al .
( 1999 ) .
3b presents the diurnal distribution @xmath58 of the times @xmath47 corresponding to the maximum value of the amplitude @xmath46 of the wave packet of twps .
it is evident that most of the twps ( about 87@xmath42 ) are observed during the daytime , from about 7:00 to 16:00 of local time , lt .
3a plots the dependence @xmath59 of the number of twps on the values of the geomagnetic activity index @xmath60 taken by the modulus .
there is a general tendency of the number of twps to increase with the decreasing level of geomagnetic activity .
most ( about 92@xmath42 ) of the twps occur when the values of the @xmath60-index are smaller than 100 nt .
the availability of a large number of gps stations in some regions of the globe ( in california , usa and west europe , for example ) makes it possible to determine not only the temporal but also spatial characteristics of twps . to estimate the radius of spatial correlation of events of this type we calculated the number of cases where
the twps within a single 2.3-hour time interval were observed at any two gps stations separated by @xmath61 .
3c presents the histogram of values of @xmath62 as a function of distance @xmath61 .
it was found that the localization of the twps in space is strongly pronounced . in 82@xmath42 of cases
the distance @xmath61 does not exceed 500 km .
using the method of gps interferometry of tids ( afraimovich et al . , 1998 ; 2000 ) , we carried out a detailed analysis of the spatial - temporal properties of twps by considering an example of the most pronounced manifestation of twps on october 18 , 2001 over california , usa .
numerous traveling ionospheric disturbances of the type of twps were recorded on that day between 15:00 and 18:00 ut using signals from several satellites , at many gps stations located in california .
the area of california within @xmath63e ; @xmath64n is convenient for our investigations because of a large number of gps stations located there , which makes it possible to obtain a great variety of gps arrays of a different configuration for determining the tid parameters and provides a means of verifying the reliability of calculated data .
it is also important that for the above - mentioned time interval 15:18 ut and for the chosen longitude range the local time varied from 08 to 11 lt , which reduces the level of background tec fluctuations characteristic for the night - time . fig .
4 illustrates the geometry of the experiment of october 18 , 2001 .
heavy dots show the gps stations , and small dots show the position of subionospheric points for gps receiver - satellite los .
since each receiver site observes simultaneously several ( no less than four ) gps satellites , the number of radio paths far exceeds the number of stations , which enhances the capabilities of analysis . fig .
4a presents the entire set of gps stations used in the experiment .
fig . 4b and
4c show the stations and the subionospheric points where the tec variations revealed twps with an amplitude exceeding the specified threshold @xmath28=0.05 @xmath17 ( fig .
4b ) and @xmath28=0.1 @xmath17 ( fig .
4c ) . with the sole exception ,
twps were recorded along the paths over and , predominantly in the north - eastward direction . as is evident from fig .
4 , an increase of the recording threshold by a factor of two reduced the number of recorded events by a factor of two .
stations shown in fig .
4c were used as the elements of the gps arrays in calculating the twps parameters . the geomagnetic situation on october 18 , 2001
may be characterized as a weakly disturbed one , which must lead to some increase of the level of background tec fluctuations yet can not cause large - scale changes in electron density characteristic for a geomagnetically disturbed ionosphere .
geomagnetic field @xmath60-variations for october 18 , 2001 are plotted in fig .
5a . in the analysis of the geomagnetic situation we used also the data from magnetic observatory victoria ( @xmath65n ; @xmath66e ) where for the time interval of our interest a weak geomagnetic disturbance was recorded , which implied a decrease of the horizontal component @xmath67 of the magnetic field by 60 nt ( fig .
5b ) . there was a concurrent , small decrease of the fluctuation amplitude of the @xmath68-component in the range of periods 2 - 20 min ( fig .
the range of variation of the geomagnetic @xmath60-index for the selected time interval was also relatively small ( no more than 20 nt ) , yet the period 15:30 - 18:00 ut showed a clearly pronounced decrease of the @xmath60-index coinciding with the period of the decrease of the @xmath69-component of the magnetic field ( fig .
5d presents the distribution @xmath70 of the number of twps detected for that day by all stations of the global gps system analyzed here , with the standard deviation in excess of @xmath28 = 0.1 @xmath17 . fig .
5e illustrates the dynamic amplitude spectrum @xmath71 of tec variations in the range of periods 560 min obtained by using the method of spatial averaging of the spectra for the entire california region ( afraimovich et al . , 2001a ) .
overall , the tec variations correlate with geomagnetic field variations .
between 15:00 and 19:00 ut , the enhancement of the oscillations of the @xmath69-component was accompanied by an expansion of the spectrum and by an increase of the tec fluctuation amplitude . the highest intensity is shown by the tec oscillations with periods of 1217 min between 15:30 and 17:00 ut . the largest number of twps was also recorded during the same period of time ( fig .
5d ) . to check that twps were observed on that day somewhere else on the globe and not only between 15 and 19 ut , we processed the data with different values of the threshold @xmath28 for the entire global gps network .
6 presents the sample statistic of the pulsations identified for october 18 , 2001 as a function of ut and local time lt , calculated for the longitude of @xmath72e corresponding to the middle of the california region : fig .
6a - from twps with the standard deviation higher than @xmath28 = 0.1 @xmath17 obtained from all stations of the global gps network used in the study ( copy of fig .
5d ) ; fig . 6b and fig .
6c - same as in fig .
6a but for @xmath28 = 0.05 @xmath17 and @xmath28 = 0.01 @xmath17 ; fig .
6d - standard deviation higher than @xmath28 = 0.01 @xmath17 for the data from the california region only .
an analysis of the fig .
6 data leads us to conclude that the twps on that day were observed mainly in california only and only over the time interval 1517 ut .
the methods of determining the form and dynamic characteristics of tids that are used in this study are based on those reported in ( mercier , 1986 ; afraimovich , 1995 ; 1997 ; afraimovich et al .
, 1998 ; 1999 ) .
we determine the velocity and direction of motion of the phase interference pattern ( phase front ) in terms of some model of this pattern , an adequate choice of which is of critical importance . in the simplest form ,
space - time variations in phase of the transionospheric radio signal that are proportional to tec variations @xmath73 in the ionosphere , at each given time @xmath74 can be represented in terms of the phase interference pattern that moves without a change in its shape ( the non dispersive disturbances ) : @xmath75 where @xmath76 and @xmath77 are the velocities of intersection of the phase front of the axes x ( directed to the east ) and y ( directed to the north ) , respectively .
a special case of ( 4 ) is the most often used model for a solitary , plane travelling wave of tec disturbance : @xmath78 where @xmath79 are space - time variations of tec ; @xmath80 $ ] the amplitude ;
@xmath81 , @xmath82 , @xmath83 are the @xmath84- and @xmath85- projections of the wave vector @xmath86 , and the angular frequency of the disturbance , respectivly ; @xmath87 is the initial disturbance phase ; @xmath88 is the time when the disturbance has a maximum amplitude ; @xmath89 is the half - thickness of the wave packet. it should be noticed , however , that in real situations neither of these ideal models ( 4 ) , ( 5 ) are realized in a pure form .
this is because that the agw that cause tids propagate in the atmosphere in the form of a dispersing wave packet with a finite value of the width of the angular spectrum .
but in the first approximation on short time interval of averaging compared to time period of filtered variations of tec , the phase interference pattern moves without a substantial change in its shape .
mercier ( 1986 ) suggested a statistical method to analyze the phase interference pattern .
primary data comprise time dependencies of the spatial phase derivatives @xmath90 and @xmath91 along the directions @xmath85 and @xmath84 .
method mercier ( 1986 ) involves determining a series of instantaneous values of the direction @xmath22 @xmath92 and constructing subsequently , on a chosen time interval , the distribution function of azimuth @xmath93 .
the central value of @xmath94 is used by mercier as an estimate of the azimuth of prevailing propagation of tids ( modulo 180@xmath95 ) .
the other method is based on analyzing the phase interference pattern anisotropy in the antenna array plane by determining the contrast @xmath96 of the interference pattern ( mercier , 1986 ) . in this case
the ratio @xmath97 is calculated as follows : @xmath98 where @xmath99 and @xmath100 are series of the transformed values of @xmath91 and @xmath90 obtained by rotating the original coordinate system ( @xmath84 , @xmath85 ) by the angle @xmath101 : @xmath102 and @xmath103 and @xmath104 are r. m. s. of the corresponding series .
mercier ( 1986 ) showed that it is possible to find such a value of the rotation angle @xmath105 , at which the ratio @xmath97 will be a maximum and equal to the value of contrast @xmath96 .
this parameter characterizes the degree of anisotropy of the phase interference pattern .
the angle @xmath105 in this case indicates the direction of elongation , and the angle @xmath106 indicates the direction of the wave vector @xmath86 coincident ( modulo 180@xmath95 ) with the propagation direction of the phase front .
the method of mercier ( 1986 ) essentially makes it possible to determine only the anisotropy and the direction of irregularity elongation of the phase interference pattern ( modulo 180@xmath95 ) . a statistical ,
angle - of - arrival and doppler method ( sadm ) is proposed by afraimovich ( 1995 , 1997 ) for determining characteristics of the dynamics of the phase interference pattern in the horizontal plane by measuring variations of phase derivatives with respect not only to spatial coordinates @xmath91 and @xmath90 - proportional to the angles of arrival variations , but additionally to time @xmath107 - proportional to the frequency doppler shift variations .
this permits to ascertain the unambiguous orientation of @xmath22 of the wave - vector @xmath86 in the range 0360@xmath95 and determine the horizontal velocity modulus @xmath108 at each specific in - stant of time .
afraimovich et al.(1998 , 1999 ) described updating of the sadm algorithm for gps - arrays ( sadm - gps ) based on a simple model for the displacement of the phase interference pattern that travels without a change in the shape and on using current information about angular coordinates of the gps satellites : the elevation @xmath50 and the azimuth @xmath51 .
of course , such an approximation is acceptable only for large values of the los elevation @xmath109 .
the method sadm - gps permits to determine the horizontal velocity @xmath108 and the azimuth @xmath22 of tid displacement at each specific in - stant of time ( the wave - vector orientation @xmath86 ) in a fixed coordinate system ( @xmath84 , @xmath85 ) : @xmath110 where @xmath111 and @xmath112 are the propagation velocities of the phase front along the axes @xmath84 and @xmath85 in a frame of reference related to the gps - array ; @xmath113 and @xmath114 are the @xmath84 and @xmath85 projections of the velocity @xmath115 of the subionospheric point ( for taking into account the motion of the satellite ) .
the coordinates of the subionospheric point @xmath116 and @xmath117 at @xmath118 in the chosen topocentric coordinate system vary as : @xmath119 and the @xmath84- and @xmath85- components of the displacement velocity @xmath115 : @xmath120 let us take a brief look at the sequence of data handling procedures . out of a large number of gps stations ,
three points ( a , b , c ) are chosen in such a way that the distances between them do not exceed about one - half the expected wavelength @xmath121 of the disturbance .
the point b is taken to be the center of a topocentric frame of reference .
such a configuration of the gps receivers represents a gps - array ( or a gps - interferometer ) with a minimum of the necessary number of elements . in regions with a dense network of gps - points ,
we can obtain a broad range of gps - arrays of a different configuration , which furnishing a means of testing the data obtained for reliability ; in this paper we have taken advantage of this possibility .
the input data include series of the vertical tec @xmath122 , @xmath123 , @xmath124 , as well as corresponding series of values of the elevation @xmath50 and the azimuth @xmath51 of the los .
series of the values of the elevation @xmath50 and azimuth @xmath51 of the los are used to determine the location of the subionospheric point , as well as to calculate the elevation @xmath125 of the wave vector @xmath126 of the disturbance from the known azimuth @xmath94 ( see formula ( 12 ) ) .
since the distance between gps - array elements ( from several tens to a few hundred of kilometers ) is much smaller than that to the gps satellite ( over 20000 km ) , the array geometry at the height near the main maximum of the @xmath26-layer is identical to that on the ground .
linear transformations of the differences of the values of the filtered tec @xmath127 and @xmath128 at the receiving points a , b and c are used to calculate the components of the tec gradient @xmath129 and @xmath130 ( afraimovich et al . , 1998 ) .
the time derivative of tec @xmath19 is determined by differentiating @xmath131 at the point b. the resulting series are used to calculate instantaneous values of the horizontal velocity @xmath108 and the azimuth @xmath22 of tid propagation .
next , the series @xmath108 and @xmath22 are put to a statistical treatment .
this involves constructing distributions of the horizontal velocity @xmath132 and direction @xmath93 which are analyzed to test the hypothesis of the existence of the preferred propagation direction .
if such a direction does exist , then the corresponding distributions are used to calculate the mean value of the horizontal velocity @xmath133 , as well as the mean value of the azimuth @xmath134 of tid propagation .
as a first approximation , the transionospheric sounding method is responsive only to tids with the wave vector @xmath126 perpendicular to the direction @xmath135 of the los . a corresponding condition for elevation @xmath125 and azimuth @xmath94 has the form @xmath136
hence the phase velocity modulus @xmath137 can be defined as @xmath138 the aspect dependence ( 12 ) of the tec disturbance amplitude is of significant importance for investigating wave disturbances .
the condition ( 12 ) imposes a constraint on the number of los for which a reliable detection of tids at the background of noise is possible .
the aspect effect causes the disturbance maximum to be displaced along the time axis , which can introduce errors in determining the tid displacement if the velocity is calculated from time delays . furthermore , as will be shown below , the aspect effect will give rise to structures of the type of wave packets `` observed '' in tec variations , which do not exist in the reality .
afraimovich et al .
( 1992 ) showed that for the gaussian ionization distribution the tec disturbance amplitude @xmath52 is determined by the aspect angle @xmath139 between the vectors @xmath126 and @xmath135 , as well as by the ratio of the wavelength of the disturbance @xmath121 to the half - thickness of the ionization maximum @xmath140 : @xmath141 \label{twp - eq14}\ ] ] in this paper the influence of aspect effects on the character of tec behavior and on the accuracy of the calculated parameters of twps was investigated and the reliability of the determination of twps characteristics was verified by modeling the wave disturbances of electron density for the observing conditions of october 18 , 2001 .
thus , on the basis of using the transformations described in this section , for each of the gps arrays chosen for the analysis we obtained the average ( for the time interval of about 1 - 2 hours ) values of the following twps parameters : @xmath142 , the amplitude of the tec disturbance ; @xmath134 and @xmath143 the azimuth and elevation of the wave vector @xmath126 ; @xmath133 and @xmath144 the horizontal component and the phase velocity modulus , the contrast @xmath96 , and the azimuth of a normal to the wave front @xmath145 from the method reported by mercier ( 1986 ) . fig .
7a plots the typical time dependencies of tec @xmath33 for gps satellite prn14 for three gps stations : bran , chms , and dups in california .
the three gps sites constitute a typical gps array , the data from which were processed by the technique described in the preceding section . for the same stations fig .
7b presents the tec variations @xmath36 that were filtered from the initial series @xmath33 using the band - pass filter with the boundaries from 5 to 20 min .
the filtered series for the period 15:0016:00 ut show the presence of significant tec pulsations of the type of single wave packet with a duration of about one hour and with the amplitude @xmath146 = 0.5 @xmath17 .
not only does the range of the filtered oscillations @xmath36 far exceed the error of phase measurements ( @xmath147 @xmath17 ) , but it also exceeds nearly an order of magnitude the level of background tec variations .
tec variations from the three spatially separated gps stations show a high degree of similarity and have a small time shift .
this suggests that we are dealing here with the same traveling disturbance .
results of a calculation ( using the sadm - gps algorithm ) of the mean values of the azimuth @xmath134 and the horizontal velocity @xmath133 of the disturbance for each 30-s time interval are presented in fig . 7c and 7d . as is evident from the figure ,
this wave packet was traveling predominantly in the south - eastward direction with the mean velocity of about 180 m / s .
the scatter of the counts is caused by the incomplete correspondence of the actual picture of an ideal twp model in the form of a monochromatic packet ( 5 ) and , in particular , by the presence of background non - correlated tec fluctuations ( afraimovich et al . ,
1998 ) . a processing of the data from the other gps arrays in the same region ( fig .
4c ) by use of the sadm - gps method provided distributions of the main twp parameters recorded on october 18 , 2001 in california .
the various combinations of gps arrays for the time interval 1516 ut totaled 231 .
statistical data show an agreement of the mean values of the calculated parameters to within their standard deviations , which indicates a good stability of the data obtained , irrespective of the particular configuration of a gps array .
distributions of the mean values of the twp parameters calculated for each of the 231 gps arrays are presented in panels ( * a * * c * ) of fig .
according to our data , the value of the horizontal propagation velocity @xmath148 of twp ( fig .
8a ) varies from 60 to 270 m / s , on the average , with the most probable value 190 m / s .
the twp wavelength @xmath121 with the mean oscillation period of about 1000 s is on the order of 150200 km .
an analysis of the distribution of the azimuths @xmath93 ( fig .
8b ) shows a clearly pronounced south - eastward direction of twp displacement @xmath149 . the average ( for 231 gps arrays )
value of the contrast @xmath96 = 7 suggests a strong anisotropy of twp .
8b shows also the distribution of the azimuth of a normal to the wave front @xmath150 deduced by the method of mercier ( 1986 ) .
this distribution virtually coincides with @xmath93 , suggesting that the twp travel across their elongation .
thus the typical size of the entire wave packet along the propagation direction is about 300500 km , and along the wave front it is as long as 1000 m. the arrow in fig .
4c schematically shows the wave vector @xmath126 of the twp .
the values of @xmath94 and @xmath148 , presented in fig .
4c , correspond to the most probable values of the propagation azimuth and the modulus of the horizontal velocity of the twp .
the elevation of the twp wave vector , determined from the aspect condition ( 12 ) , has mostly a small positive value ( fig .
8c ) : i.e. the oscillation phase propagates upward .
since the mean value of the elevation is only @xmath151 , it can be assumed that tec pulsations represent an almost horizontal wave .
accordingly , estimations of the modulus of the phase velocity @xmath137 ( fig .
8a ) give values close to the value of its horizontal projection 50270 m / s , with the largest probable value 180 m / s .
in this paper the influence of aspect effects on the character of tec behavior and on the accuracy of the calculated of tec pulsations was investigated and the reliability of the determination of the twp characteristics was verified by modeling the wave disturbances of electron density for the observing conditions of october 18 , 2001 .
our developed model of tec measurements with the gps interferometer makes it possible to calculate as realistic a spatial and temporal distribution of the local electron density @xmath152 in the ionosphere as possible and then , using the coordinates of the receiver sites and of the satellites , to integrate @xmath152 along the receiver - satellite los with a given step over time ( afraimovich et al . , 1998 ) . as a result
we obtain time series of tec similar to input experimental data which can be processed by the same methods as used to process experimental data .
the ionization model takes into account the height distribution of @xmath152 , the seasonal and diurnal density variations determined by the zenith angle of the sun , as well as irregular disturbances of @xmath152 of a smaller amplitude and smaller spatial scales in the form of a discrete superposition of plane traveling waves . in this paper , with the purpose of elucidating the origin of twp , three types of disturbances were modeled : \a ) the disturbance in the form of a single plane wave with amplitude @xmath153=3% in the ionization maximum , with the period @xmath154=15 min and the wavelength @xmath5=172 km .
the elevation of the wave @xmath155=10@xmath95 , and the azimuth @xmath156=146@xmath95 ; \b ) the disturbance in the form of a superposition of two plane waves with periods @xmath154=15 min and @xmath157=12 min , and with the wavelengths @xmath5=172 km and @xmath6=138 km .
the elevations , azimuths and azimuths of the waves were specified identical : @xmath155=@xmath158=10@xmath95 , @xmath156=@xmath159=146@xmath95 , @xmath153=@xmath160=3% of the value of @xmath152 in the ionization maximum ; \c ) the disturbance in the form of a single wave packet , with the semi - thickness @xmath161=20 min and a maximum amplitude at the time @xmath47=15.5 ut .
the oscillations inside the packet had the period @xmath154=15 min and the wavelength @xmath5=172 km .
disturbance parameters were taken to be close to those obtained from experimental data using the technique from ( afraimovich et al . , 1998 ) .
tec pulsations were modeled with the purpose of ( 1 ) verifying the reliability of the calculated twp characteristics , and ( 2 ) elucidating the origin of twp .
a detailed description of the model used is given in ( afraimovich et al .
, 1998 ) .
( panels * e**h * at the right ) presents the results of calculations in terms of the model of the twp in the form of a single wave packet for the bran , chsm , dupr array on october 18 , 2001 .
parameters of the wave packet were taken to be close to experimental data ( fig .
distributions of pulsation parameters obtained in a similar modeling of twp for the other california gps arrays are presented in fig . 8 ( panels * d**f * at the right ) . fig .
7 illustrates a good similarity of the experimental and model tec variations @xmath33 , @xmath36 and the dependencies @xmath108 and @xmath22 .
noteworthy is the weaker inclination of the model tec @xmath33 ( fig .
7e ) when compared with the experimental one ( fig .
7a ) . this is because for the sake of simplicity and for illustrative purposes , diurnal variations in ionization in the model are proportional to the cosine of the zenith angle of the sun . in actual conditions the dependence is more complicated , which gives a faster temporal growth of the tec .
because the trend is removed in the subsequent discussion and only the relative tec variations @xmath36 are considered , the above - mentioned difference in the behavior of the model and experimental tec will not affect results obtained .
a comparison of the tec disturbance parameters specified in the model with the corresponding values obtained following a processing by formulas of the sadm - gps method shows a relatively good agreement of these values .
the azimuth of the wave vector of twp was taken to be @xmath162 , and the mean value of the azimuth calculated by the sadm - gps method is @xmath163 ( fig . 7 g , and fig .
8e ) ; the horizontal velocity was specified equal to 180 m / s , the calculated mean value of the velocity was 207.5 m / s ( fig . 7h , and fig .
8d ) ; the elevation in the model was @xmath164 , and the most probable value of the calculated elevation was @xmath165 ( fig .
the azimuth of twp propagation determined by the sadm - gps method was also close to the azimuth values calculated by analyzing the contrast of the phase interference pattern ( mercier , 1986 ) fig .
8b , e. all this demonstrates the validity of the sadm - gps technique and confirms the reliability of the twp parameters obtained using this technique .
let us now consider the possible mechanisms that are responsible for the formation of structures of the type of pulsations in observed tec variations .
the recorded tec pulsation represents a single wave packet with a duration of about 60 min , the oscillation period inside the packet of 1215 min , and with the amplitude exceeding the level of background fluctuations by a factor of 25 ( fig .
structures of such a type can be produced in tec variations in different ways . if a monochromatic wave with a period of about 15 min ( the period of observed tec variations ) propagates in the ionosphere , then tec oscillations of the type of pulsations could be produced through the aspect effect .
9a , d illustrates such a situation .
panels * a * and * d * present the filtered tec series obtained by modeling a single plane wave under different conditions of its observation . at the bran site ( fig .
9a ) the elevations of the prn14 satellite are close to 90@xmath27 , and the wave vector of the disturbance is perpendicular to the receiver - satellite los throughout the observing period .
this is indicated by the character of the theoretical dependence @xmath166 calculated by formula ( 14 ) for this los : @xmath166 is close to 1 over the entire observing interval ( fig .
waves disturbances of the tec close to the specified monochromatic wave are therefore observed along the bran
prn14 los . at the bru1 site for prn05 ,
the detection conditions for disturbances are significantly worse . in this case for the first half an hour the wave propagates virtually along the los ( @xmath167 , fig .
9f ) , and at the end of the observing interval he bru1prn15 los is perpendicular to the propagation direction of the wave but has low elevations ( @xmath168 , fig .
9f ) . owing to this
, tec variations show an increase in the oscillation amplitude from 0 to 0.7 @xmath17 there arises a feature resembling a wave packet that was absent in the initial specified model of the wave .
however , as is evident from the figure , in this case the increase in the tec variation amplitude has an almost linear character while the experiment ( fig .
7b ) shows a nonlinear amplitude modulation .
an investigation of the character of the aspect dependence for all gps arrays that were used in this study , showed that in 97% of cases ( including those shown in fig .
7b ) the situation is realized , which is depicted in fig .
9c , i.e. @xmath166 is close to 1 throughout this observing interval . considering all that has been said above , one can draw the conclusion that the recorded tec pulsations are indeed caused by the propagation of an actual wave packet of a local density disturbance in the ionosphere rather than resulting from the recording conditions .
another possibility of the production of a twp is the combination of two or several monochromatic waves with close periods propagating in the ionosphere ( yakovets et al . , 1999 ) .
we modeled the propagation of two quasi - horizontal waves with periods @xmath154=15 min and @xmath157=12 min by identical amplitudes , velocities and directions of propagation .
the resulting tec variations along two los : bran prn14 , and bru1prn05 , with the trend removed , are presented in fig . 9b and 9e , respectively . as would be expected ,
there arise amplitude - modulated tec oscillations with the modulation period of about @xmath169=60 min .
the period is close to the length of the wave packet recorded experimentally ; however , a simple superposition of two waves gives not one but a whole chain of wave packets . in the case where the aspect effect affects little the character of tec variations ( fig .
9c ) these packets have , in addition , different amplitudes as well ( fig .
a phase change of the combined waves affects little the picture presented here by altering the form and position of the amplitude minimum only slightly . the aspect dependence in fig .
9f , however , introduces an additional modulation into the recorded signal - there occurs a significant enhancement of one of the chain s packets ( fig .
9e ) , and the picture approaches what is observed experimentally in fig .
nevertheless , it can not be believed that the observed single twps are the result of the aspect effect .
firstly , in the case of the aspect modulation , because of the slow(nearly linear ) change of the amplitude , not one but at least two wave packets are observed .
secondly , as has been pointed out above , the character of the aspect dependence in most cases of twp recordings was such that it affected little the amplitude of disturbances . to obtain the closets tec oscillations to those observed experimentally we had to introduce an artificial modulation , i.e.
the amplitude of the initial monochromatic wave of the disturbance @xmath152 was specified not constant but it had a time dependence in the form of a gaussian function .
it is such a model of tec variations that is shown in fig .
sequences of wave packets similar to the ones shown in fig . 6 at the left , were recorded in frequency doppler variations with the multifrequency doppler ionosonde at almaty ( yakovets et al . , 1999 ) .
the recordings of the frequency doppler shift presented clearly show chains of wave packets ( on the average , 23 wave packets with the record length of about 8 hours ) .
the duration of the wave packets averaged 90 min , and the oscillation period in a packet averaged 16 min .
the authors of the cited reference observed two types of fd - variations : quasi - stochastic tids , and monochromatic tids in the form of wave packets .
they concluded that quasi - stochastic tids are characterized by a random behavior of the phase , a small length of coherence , and by a large vertical phase velocity .
wave packets show quasi - monochromatic oscillations , a larger length of coherence , and a smaller vertical phase velocity .
our estimate of the radius of spatial correlation ( on the order of several hundred kilometers ) is in reasonably good agreement with the data reported by yakovets et al .
( 1995 ) .
yakovets et al .
( 1999 ) argued that the observed wave packets are the superposition of the direct and ground - reflected wave whose source lies in the troposphere .
the analysis made in this paper did not confirm the validity of such an explanation for the october 18 , 2001 twp . by comparing our detected tec pulsations with the data from ( yakovets et al .
, 1999 ; hines , 1960 ; waldock and jones , 1984 ; francis , 1974 ) as well as with the findings of our modeling , it can be assumed that in the atmosphere there exists an additional amplitude modulation mechanism for wave processes which make sit possible to obtain tec oscillations in the form of a single wave packet .
francis ( 1974 ) , by considering the auroral electrojet to be the source of tids , showed that upon propagating through the atmosphere into the @xmath44-region , the ground - reflected waves acquire the properties of a wave packet .
on the basis of the dispersion relation hines ( 1960 ) and waldock and jones ( 1984 ) showed that tids that are associated with a tropospheric jet flow occur in the @xmath44-region in the form of a wave packet with quasi - monochromatic oscillations , the period of which is a function of inclination angle of the wave vector during the propagation of the wave from the source to the place of observation in the @xmath44-region .
our data on the twp displacement velocity and direction correspond to those of mid - latitude medium - scale traveling ionospheric disturbances ( ms tids ) obtained previously in the analysis of phase characteristics of hf radio signals ( kalikhman , 1980 ; afraimovich , 1982 ; waldock and jones , 1984 ; 1986 ; 1987 ; jacobson and carlos , 1991 ) , as well as signals from first - generation navigation satellites ( spoelstra , 1992 ) , geostationary satellites ( afraimovich et al . , 1997b ; jacobson et al . , 1995 ) and
discrete space radio sources ( mercier , 1986 ; 1996 ; velthoven et al . ,
main results of this study may be summarized as follows : 1 .
the most of the twps are observed during the daytime in no more than 0.1 - 0.4@xmath42 of the total number of radio paths , most commonly in winter and autumn in a weakly disturbed or quiet geomagnetic situation .
the distance between any two gps stations where the twps within a single 2.3-hour time interval were observed does not exceed 500 km .
2 . twps in the time range represent quasi - periodic oscillations of tec of a length on the order of 1 hour with the oscillation period in the range 1020 min and the amplitude exceeding the amplitude of `` background '' tec fluctuations by one order of magnitude , as a minimum and is 0.3 @xmath17 .
3 . the dynamical parameters of the twp observed on october 18 , 2001 over california
, usa were determined .
the twp traveled with the elevation @xmath125 = 10@xmath95 and the azimuth @xmath94 = 146@xmath95 .
its mean velocity @xmath144 = 180 m / c corresponds to the velocity of medium - scale agw .
it is probably that twp origins are medium - scale agw .
we are indebted to dr a.s .
potapov for participating in discussions .
we thank o.s .
lesyuta for help in organizing the experiment .
we are also grateful to v.g .
mikhalkovsky for his assistance in preparing the english version of the manuscript .
this work was done with support under rfbr grant of leading scientific schools of the russian federation no . 00 - 15 - 98509 and intas grant no .
99 - 1186 as well as russian foundation for basic research grants no .
01 - 05 - 65374 , 00 - 05 - 72026 .
afraimovich , e. l. , a. i. terekhov , m. yu . udodov , and s. v. fridman , refraction distortions of transionospheric radio signals caused by changes in a regular ionosphere and by travelling ionospheric disturbances , _ j. atmos . and terr .
phys . , 54 _ , 10131020 , 1992 .
afraimovich .
e. l. dynamics and anisotropy of traveling ionospheric disturbances as deduced from transionospheric sounding data . - statistical angle - of - arrival and doppler method ( sadm ) .
preprint istp , 1995 , n5 - 95 , 54p .
afraimovich e. l. statistical angle - of - arrival and doppler method ( sadm ) for determining characteristics of the dynamics of the transionospheric radio signal interference pattern .
_ acta geod .
geoph . hung .
_ , 1997 , v.32 , n3 - 4 , 461468
. afraimovich e. l. , boitman o. n. , zhovty e. i. , kalikhman a. d. and pirog t. g. dynamics and anisotropy of medium - scale traveling ionospheric disturbances as deduced from transionospheric sounding data .
_ acta geod .
geoph . hung .
_ , 1997 , v.32 , n3 - 4 , 301308 .
afraimovich e. l. , kosogorov e. a. , lesyuta o. s. , yakovets a. f. , ushakov i. i. geomagnetic control of the spectrum of traveling ionospheric disturbances based on data from a global gps network .
geophys . , 19 , n 7 _ , 723731 , 2001 . bertin f. , testud j. , kersley l. and rees p.r . , the meteorological jet stream as a source of medium scale gravity waves in the thermosphere : an experimental study .
_ j atmos .
phys . 40 _ , 11611183 , n 1011 , 1978 .
huang yinn - nien , cheng kang , and sen - wen chen . on the detection of acoustic - gravity waves generated by typhon by use of real time hf doppler frequency shift sounding system , _ radio sc . , 20 , n4 _ , 897906 , 1985 .
jacobson , a. r. , r. c. carlos , r. s. massey , and g. wu , observations of traveling ionospheric disturbances with a satellite - beacon radio interferometer : seasonal and local time behavior , _ j. geophys .
res . , 100 _ , 16531665 , 1995 .
velthoven , p.f.j .
, c. mercier , and h. kelder , simultaneous observations of travelling ionospheric disturbances by twodimensional radio interferometry and the differential doppler technique applied to satellite signals , _
, 52 _ , 305312 , 1990 . | we identified a new class of mid - latitude medium - scale traveling ionospheric disturbances ( ms tids ) , viz . traveling wave packets ( twps ) of total electron content ( tec ) disturbances .
for the first time , the morphology of twps is presented for 105 days from the time interval 1998 - 2001 with a different level of geomagnetic activity , with the number of stations of the global gps network ranging from 10 to 300 .
the radio paths used in the analysis total about 700000 .
these data were obtained using the globdet technology for global detection and monitoring of ionospheric disturbances of natural and technogenic origin from measurements of tec variations acquired by a global network of receivers of the navigation gps system .
the globdet technology was developed at the istp sb ras . using the technique of gps interferometry of tids we carried out a detailed analysis of the spatial - temporal properties of twps by considering an example of the most conspicuous manifestation of twps on october 18 , 2001 over california , usa .
it was found that twps are observed no more than in 0.1 - 0.4% of all radio paths , most commonly during the daytime in winter and autumn at low geomagnetic activity .
twps in the time range represent quasi - periodic oscillations of tec of a length on the order of 1 hour with the oscillation period in the range 10 - 20 min and the amplitude exceeding the amplitude of `` background '' tec fluctuations by one order of magnitude , as a minimum .
the radius of spatial correlation of twps does not exceed 500600 km ( 35 wavelengths ) .
the velocity and direction of twps correspond to those of mid - latitude medium - scale traveling ionospheric disturbances ( ms tids ) obtained previously from analyzing the phase characteristics of hf radio signals as well as signals from geostationary satellites and discrete cosmic radio sources . |
merging double neutron stars ( dnss ) that eject highly neutron rich material are prime candidates for the production sites of heavy @xmath0(apid)-process elements @xcite .
the overall amount of heavy @xmath0-process material in the milky way is consistent with the expectations of mass ejection in numerical merger simulations ( e.g. * ? ? ?
* ; * ? ? ?
* ) with their expected rates as estimated from galactic dnss ( see e.g. * ? ? ?
* ) or from the rate @xcite of short gamma - ray bursts ( sgrbs ) .
discoveries of @xmath0-process driven macronova ( kilonova ) candidates associated with sgrbs @xcite provided further observational evidence of the dns merger scenario ( e.g. * ? ? ?
following these developments , several recent works @xcite have shown that under reasonable assumptions dns mergers can account for the history of r - process enrichment in the galaxy .
however , recently , @xcite have used the observations of @xmath0-process elements in dwarf satellite galaxies to question the dns merger scenario for @xmath0-process production .
@xcite , and independently , @xcite reported the discovery of an @xmath0-process enriched ultra - faint dwarf ( ufd ) galaxy reticulum ii , the total stellar luminosity of reticulum ii is @xmath5 and the line of sight velocity dispersion is @xmath6 km / s @xcite .
@xcite suggested that the kick given to the dns during the second collapse would eject the binary from such a small galaxy .
a second problem that arises is that ufds are composed of very old stellar population @xcite , suggesting that the chemical abundances have been frozen since @xmath7 gyr ago .
this requires that the @xmath0-process formation should take place relatively soon after the formation of the first stars .
this raises the question whether mergers could take place sufficiently rapidly so that their @xmath0-process material would be able to enrich the old stellar population .
@xcite suggested , therefore , that a different mechanism must have produced the observed @xmath0-process material in these galaxies .
a significant population of rapid mergers " ( @xmath8gyr ) is natural , and in fact is expected from observations of dns systems in our galaxy .
two of the ten observed dns systems in our galaxy ( that do nt reside in globular clusters and for which the masses are well constrained ) , the double pulsar , j0737 - 3039 , and the original binary pulsar , b1913 - 16 , will merge in less than a few hundred myr .
given the spin down time of the pulsars in these systems , we can constrain the life time of these systems since the formation of the dns to less than 140myr and 460myr respectively . indeed
the existence rapid mergers has previously been suggested using population synthesis models @xcite .
furthermore , the observed small proper motion of j0737 - 3039 , @xmath9 @xcite , implies that some rapid mergers " move slowly enough to remain confined even within ufds . and @xmath10 .
this , in turn , constrains the separation before the collapse , @xmath11 , to @xmath12<a_0<a_1(1+e_1)\approx9.5\times10^{10}$]cm . ]
* hereafter bp16 ) used the observed orbital parameters of the galactic dns population to constrain the distributions of mass ejection and kick velocities associated with the formation of the second neutron star in the binary .
while the smallness of the sample and unknown selection effects do nt allow an accurate estimate of these distributions a clear picture emerges , there are two distinct types of neutron star formation .
the majority of the systems , about two thirds , involve a minimal mass ejection ( @xmath13 ) and low kick velocities ( @xmath14 kms@xmath15 ) .
the double pulsar system , psr j0737 - 3039 , is a prime candidate of this kind of collapse with @xmath16 and @xmath17 @xcite .
such a population of collapses with low mass ejection and kicks has been suggested on both observational @xcite and theoretical @xcite grounds .
subsequent addition of a low mass ejection channel of neutron star formation ( via electron capture sne ) to population synthesis models @xcite improved the fit of the models to the observed dns population .
a large fraction of dnss , born via the same mechanism , remain bound to their dwarf hosts . on a related topic ,
@xcite have argued that many dnss are expected to remain confined and merge within globular clusters , which have comparable escape velocities to ufds . to explore these ideas we begin , in [ sec : kicks ] with a simulation of the typical velocities of dns systems using the implied distributions of mass ejection and kicks from bp16 .
we then address the delay times between formation and merger in [ rapid ] .
we show that a significant fraction of dns systems will remain confined in ufds and merge rapidly , demonstrating the viability of dns mergers as sources of @xmath0-process material in ufds . in [ sec : impgalaxy ] we consider the implications of these findings to the related problem of observation of heavy @xmath0-process material in some very low metallicity stars in the galaxy .
we summarize our results and the case for dns mergers as the source of heavy @xmath0-process nucleosynthesis in
[ sec : summary ] .
consider the second collapse leading to the formation of the younger neutron star in a neutron star binary . assuming an initially circular orbit , the change in the centre of mass velocity , @xmath18 , due to the second collapse is given by : @xmath19 where @xmath20 and @xmath21 are the masses of the pulsar and the ( collapsing ) companion today , @xmath22 is the companion mass right before the collapse
, @xmath23 is the initial keplerian velocity and @xmath24 is the kick velocity imparted on the companion by the ejected mass . naturally , if both the ejected mass , @xmath25 and the kick , @xmath26 are sufficiently small , the change in the cm velocity is also small .
an inevitable mass loss , of order @xmath27 @xcite , arises due to emission of neutrinos . plugging these values into eq .
[ eq : vcm ] with an initial separation @xmath28 cm , final neutron star masses of @xmath29 and no asymmetric kick , we find : @xmath30 .
this is an approximate minimal value for the change in the cm velocity due to the collapse .
a kick with a similar magnitude in approximately the opposite direction ( or increasing @xmath31 , although see [ sec : introduction ] ) reduces somewhat this value .
bp16 considered the orbital parameters and masses of the galactic dns population and constrained the two distributions of mass ejection and kick velocities directly from observations ( with no a - priori assumptions regarding evolutionary models and/or types of collapse involved ) .
they ve shown that there is strong evidence for two distinct types of stellar collapses : the majority of systems have small eccentricities ( we refer to these as small @xmath32 systems " ) implying a small mass ejection ( @xmath33 ) and a low kick velocity ( @xmath34kms@xmath15 ) . only a minority of the systems ( large @xmath32 systems
" ) have been formed via the standard sn scenario involving a larger mass ejection of @xmath35 and kick velocities of up to @xmath36kms@xmath15 . to calculate the fraction of systems with a given @xmath18 we performed a monte carlo simulation using the best fit distributions of collapse parameters found in bp16 .
as mentioned earlier , given the rather small number of observed dns systems in the galaxy and possible observational biases ( see bp16 ) , it is impossible to determine the exact shapes of these distributions .
the following calculations should thus be treated as an order of magnitude estimates .
note , however , that bp16 considered several functional shapes for these distributions and the results depend only weakly on the exact functional shape chosen .
we assume that @xmath37 of the systems are small @xmath32 systems " with a log - normal mass ejection distribution ( above the minimum neutrino mass loss , which we take here to be @xmath38 ) peaking at @xmath39 and a log - normal kick distribution peaking at @xmath40kms@xmath15 while @xmath41 of the systems are large @xmath32 systems " with a log - normal mass ejection distribution peaking at @xmath42 and a log - normal kick distribution peaking at @xmath43kms@xmath15 . in all of these distributions
the standard deviations are half the magnitude of the corresponding peaks .
the direction of the kick velocities were chosen randomly .
in addition we take @xmath44 , and initially circular orbits with constant separations .
[ fig : vcmdist ] depicts the fraction of dns systems with @xmath45 as a function of @xmath46 for different initial separations .
the initial keplerian velocity decreases as the initial separation increases , resulting in a larger fraction of systems with small cm velocities for larger separations .
the shape of each curve is composed of two sharp rises , the first corresponding to the small @xmath32 systems " with weak kicks and the second to the large @xmath32 systems " with strong kicks . for initial separations in the range @xmath47 cm ,
a significant fraction of dns systems ( between @xmath48 ) have a cm velocity @xmath49kms@xmath15 ( smaller than a typical escape velocity of ufds ) .
this suggests that if these systems originated in such galaxies , they would , most likely , remain bound .
the time until merger of a dns depends strongly on the initial separation and eccentricity .
generally an explosion occurring in one of the members of a binary , will cause the binary to become less bound .
even if no asymmetric kick is imparted due to the explosion , the ejection of mass causes the stars to be less gravitationally bound .
this can also be seen from the equation for energy conservation with no kick : @xmath50^{-1}=\left[1-\frac{\delta m}{m_c+m_p}\right]^{-1}>1\ .\ ] ] nevertheless , if the kick is oriented such that the relative velocity of the stars decreases , the system can become more bound after the explosion . since for a circular orbit , the kinetic energy is exactly minus that of the total orbital energy , the most energy that can be extracted is by reducing the kinetic to zero , thus doubling the ( negative ) energy and leaving the binary with half of the initial semi - major axis and a very eccentric ( @xmath51 ) orbit . for this to occur , the kick has to be oriented appropriately and its magnitude should be comparable to the keplerian velocity . a dns system with a kick of this type would merge on a much shorter time - scale than implied by the original separation .
we carried out a monte carlo simulation as described in
[ sec : kicks ] . for each realization
, we calculate the conditions after the second collapse and use the separation and eccentricity to find the time until merger due to gw emission . the fraction of dns systems merging by a time @xmath52 for both the small @xmath32 " and
large @xmath32 " groups is shown in fig .
[ fig : probtmerger ] . for @xmath53 cm ( @xmath54 cm ) ,
corresponding to a merger time of @xmath55gyr ( @xmath56gyr ) , @xmath57 of large e " systems ( corresponding to @xmath58 of the entire dns population ) merge in less than @xmath59myr ( @xmath60gyr ) ( see * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
this population has important effects on the typical delay between sne and dns mergers leading to a faster rise in the abundance of heavy r - process elements as a compared to the iron abundance @xcite .
so far we have addressed the questions of merger time and cm velocities separately .
however , systems with confining kicks are preferentially those that undergo larger changes in the cm velocity .
the decrease in the gravitational orbital energy is compensated by an increase in the kinetic energy of the cm .
thus , although important for larger galaxies , confining kicks are not relevant for ufds .
for the latter , it is important to consider the combined probability of having a rapid merger with a small cm velocity .
[ fig : probtandv ] describes the fraction of dns systems merging before 1gyr with @xmath61 as a function of @xmath46 .
we consider an initially circular orbit with a separation @xmath53 cm ( @xmath54 cm ) . for @xmath53
cm , @xmath62 of the systems merge by 1gyr and have @xmath63kms@xmath15 ( comparable to the lowest escape velocities of ufds ) .
however , this fraction decreases significantly , down to @xmath64 , for @xmath54 cm .
as expected the distribution is dominated by the small @xmath32 " group of systems which merge on a time - scale that is dictated by their initial separation .
the existence of rapid mergers as well as evidence for the typical initial separations used here is supported by observations of known dns systems in our galaxy ( see [ sec : impgalaxy ] for details ) .
another line of evidence for the existence of rapid mergers arises from observations of sgrb hosts .
since sgrbs are most likely associated with dns mergers @xcite , one may expect that in the absence of rapid mergers , sgrbs would be predominantly observed in the more massive elliptical galaxies .
in fact , a significant fraction of sgrbs are observed in star forming galaxies @xcite .
the prevalence of sgrbs in star forming galaxies is naturally expected if a significant population of dns systems merge rapidly .
our main aim in this paper is to explore the possibility of a dns mergers taking place in ufds . still , this work and in particular the parts corresponding to rapid mergers has also implications to nucleosynthesis in the milky way .
abundances of @xmath0-process elements have been observed in halo and disk stars covering a metallicity range @xmath65}\approx-3.1 - 0.5 $ ] @xcite .
this requires a significant fraction of @xmath0-process nucleosynthesis to take place within a few hundred myr or less .
this has been used as an argument against dns mergers as the origin of heavy @xmath0-process material in the galaxy .
the argument is roughly the following : dns mergers will take place a few gyr after star formation at which stage all stars would already be contaminated " by a significant amount of iron .
furthermore , dns mergers require sne and those will produce iron before the @xmath0-process material is produced .
most galactic dns systems were born in the past when the galactic sfr was larger ( e.g. * ? ? ? * ) . however
, all the rapid mergers " that formed at that time have already merged .
hence the total fraction of rapid mergers is larger than what is currently observed in the galactic binary pulsar population .
this result is consistent with the global fast decay in the rate of sgrbs in the universe @xcite .
it is also consistent with the observation that the current rate of deposition of radioactive @xmath66pu on earth is much lower than what it was 4.6 gyr ago when the solar system was formed @xcite .
we have shown here that dns mergers can take place rapidly enough and enrich at least some parts of our galaxy on very short time scales .
specifically , @xmath57 of the large @xmath32 systems with @xmath53 cm have merger times less than @xmath59 myr .
they play important roles to account for a fast rise in the abundances of heavy @xmath0-process elements of metal poor stars in the galaxy .
the large fluctuations in the eu / fe ratio in metal poor stars , supports a rare enrichment process like dns merger .
regarding the problem of contamination " of the ism by iron from sne that took place prior to the formation of the dns system , we note that in a significant fraction of the cases , the dns systems are produced with very little mass ejection .
in other cases , the proper motion of the dns relative to their birth rest frame suggests that even in a few hundred myr they could move sufficiently within the galaxy to avoid the ism that has been contaminated by their progenitor " sne .
such an effect should be taken into account in chemical evolution studies .
an open question related to this involves the not yet understood topic of possible large scale turbulent mixing in the galaxy .
we have examined here dns mergers as sources of heavy @xmath0-process material in dwarf galaxies .
we have shown that both arguments raised by @xcite , that dns systems may be ejected from their galaxies due to strong kicks received at formation , and that dns would not be able to merge rapidly enough before the star formation in these galaxies has stopped , are naturally overcome .
first , due to a significant population of dns receiving weak kicks at birth ( bp16 ) , a large fraction of dns systems would have cm velocities @xmath49kms@xmath15 ( comparable to the lowest escape velocities from ufds ) .
second , given limits on the separation of the double pulsar system before its second collapse , a significant fraction of systems are expected to both remain confined in ufd galaxies and merge within less than a gyr . moving from ufds to the galaxy , where small velocities are not required to remain confined ,
the fraction of rapid mergers becomes even larger due to the contribution of confining kicks " .
limits on the time between the second collapse until eventual merger , of two out of the ten observed dns systems in the galaxy , imply that many dns systems are expected to merge within less than a few hundred myr .
this is also supported by observations of a rapid decay in the sgrb rate following the peak of star formation in the universe .
this population of rapid mergers is composed mainly of systems that were formed with small amounts of mass ejection @xmath67 , ( bp16 ) .
moreover , the mergers take place far away from the place of the progenitor sne .
this implies that dns mergers can naturally account for the observations of significant amounts of @xmath0-process material in some low metallicities halo stars in the galaxy .
this invalidates the argument used against dns mergers as the origin of @xmath0-process nucleosynthesis in our galaxy .
in fact , dns mergers are more easily compatible with the large fluctuations in the eu / fe ratio in metal poor stars which are more easily accounted for by rare events . in a companion paper @xcite
we provide further quantitative evidence that @xmath0-process material is produced in rare events in both our own galaxy and ufds .
we thank nicholas stone and todd thompson for useful discussions and comments .
this work was supported in part by an israel space agency ( sela ) grant , the templeton foundation and the i - core center for excellence
origins " of the isf . | merging neutron star binaries are prime candidate sources for heavy @xmath0-process nucleosynthesis .
the amount of heavy @xmath0-process material is consistent with the mass ejection and rates of mergers , and abundances of relic radioactive materials suggest that heavy @xmath0-process material is produced in rare events .
observations of possible macronovae provide further support for this model .
still , some concerns remain .
one is the observation of heavy @xmath0-process elements in ultra faint dwarf ( ufd ) galaxies .
the escape velocities from ufds are so small that the natal kicks , taking place at neutron stars birth , might eject such binaries from ufds .
furthermore the old stellar populations of ufds requires that @xmath0-process nucleosynthesis must have taken place very early on , while it may take several gyr for compact binaries to merge .
this last problem arises also within the milky way where heavy @xmath0-process materials has been observed in some low metallicity stars .
we show here that @xmath1 of neutron star binaries form with a sufficiently small proper motion to remain bound even in a ufd . furthermore , approximately @xmath2 of dnss with an initial separation of @xmath3
cm merge within 300myrs and @xmath4 merge in less than 100myrs .
this population of rapid mergers " explains the appearance of heavy @xmath0-process material in both ufds and in the early milky way . |
a floer homology is the morse homology of an action functional @xmath0 defined on the infinite dimensional space @xmath1 of smooth maps from a manifold @xmath2 to a target manifold @xmath3 .
the floer chain complex is generated by critical points of @xmath0 which are smooth maps from @xmath2 to @xmath3 solving a pde ( respectively an ode if @xmath2 is 1-dimensional ) .
the floer boundary operator is defined by counting gradient flow lines of @xmath0 .
these are smooth maps from @xmath4 to @xmath3 again solving a pde . in the special case where @xmath5 is the circle and @xmath0 is the classical hamiltonian action functional the critical points of @xmath0 are the periodic orbits of the hamiltonian system .
this was used by floer in his proof of the arnold conjecture @xcite .
floer flow lines satisfy a bogomolnyi type selfduality equation and hence the solutions of various problems in physics can be interpreted as floer flow lines . in this paper
we prove existence of solutions of a generalized abelian higgs equation discovered by schroers @xcite .
this equation is a special case of the symplectic vortex equations introduced by cieliebak , gaio , mundet , and salamon @xcite .
the method of our proof is the following .
we consider an action functional @xmath0 which is invariant under the action of a finite group @xmath6 which acts freely on the critical set of @xmath0 .
this gives rise to two floer homologies where the chain complex of the first one is generated by critical points of @xmath0 and the chain complex of the second one is generated by @xmath6-orbits of critical points of @xmath0 .
we show that the first floer homology is trivial and the second one is non - trivial which implies the existence of floer flow lines . + + this paper is organized as follows . in section [ ffl ]
we show how floer flow lines can be interpreted as selfduality equations and how they are related to problems in physics . in section [ statement ]
we introduce our floer action functional and state our main theorem . in section [ proof ]
we first explain how one can find floer flow lines if one is able to compute two floer homologies .
we then introduce our two floer homologies and compute them .
we finally consider the classical vortex equations as an example .
+ + * acknowledgements : * i would like to thank dietmar salamon for useful discussions .
this paper was written during my stay at fim of eth zrich .
i wish to thank fim for its kind hospitality .
assume that @xmath7 is a riemannian manifold and @xmath8 is a morse function on @xmath3 . for two
critical points @xmath9 let @xmath10 be the space of paths from @xmath11 to @xmath12 .
the energy functional @xmath13 is defined by @xmath14 the energy functional satisfies the following bogomolnyi equation @xcite , @xmath15 it follows from ( [ bogomolnyi ] ) that downwards and upwards gradient flow lines , i.e. solutions of the following first order ode @xmath16 are the absolute minimizers of the energy functional .
note that the euler - lagrange equations for @xmath17 are given by the following second order ode @xmath18 where @xmath19 denotes the hessian of @xmath20 .
of course solutions of ( [ dugrad ] ) are also solutions of ( [ euler - l ] ) .
this can be directly checked by differentiating ( [ dugrad ] ) .
equations describing the absolute minima of a functional which are first order while the euler - lagrange equations of the functional are second order are called * ( anti)-selfduality equations*. selfduality equations are of great importance in physics .
we show how some of the selfduality equations physicists are interested in can be interpreted as gradient flow lines of an action functional on an infinite dimensional space .
the basic example are the classical vortex equations on the cylinder , see @xcite as a basic reference . to describe them recall that the standard @xmath21-action on the complex plane @xmath22 given by @xmath23 is hamiltonian with respect to the standard symplectic structure @xmath24 on @xmath22 .
a moment map for this action is given by @xmath25 the vortex equations on the cylinder for a triple @xmath26 are @xmath27 if one thinks of @xmath28 as a connection on the trivial bundle over the cylinder then the second equation in ( [ ve ] ) reads @xmath29 where @xmath30 denotes the hodge operator on the cylinder @xmath31 and @xmath32 is the curvature of the connection @xmath33 .
the gauge group @xmath34 acts on the solutions of ( [ ve ] ) via @xmath35 in particular , every solution of ( [ ve ] ) is gauge equivalent to a solution of ( [ ve ] ) with @xmath36 , i.e. a pair @xmath37 satisfying @xmath38 the equations ( [ vet ] ) are called vortex equations in temporal gauge .
the gauge group @xmath39 acts on solutions of ( [ vet ] ) by @xmath40 and the map @xmath41 from solutions of ( [ vet ] ) to solutions of ( [ ve ] ) induces a bijection @xmath42 the equation ( [ vet ] ) can be interpreted as the gradient flow equation of an action functional @xmath0 defined on an infinite dimensional space @xmath43 . set @xmath44 and define @xmath45 as @xmath46 where @xmath47 is the liouville one - form satisfying @xmath48 .
then the gradient flow lines of @xmath0 with respect to the @xmath49-metric @xmath50 on @xmath43 , i.e.@xmath51 are precisely the equations ( [ vet ] ) . +
another example are the self - dual chern - simons vortices discovered by hong - kim - pac and jackiw - weinberg , see @xcite .
they read @xmath52 in temporal gauge they are given by @xmath53 the equations ( [ csvt ] ) are again gradient flow lines of @xmath0 , but with respect to a different metric , namely the following warped product metric on @xmath54 given for @xmath55 and @xmath56 by the formula @xmath57
assume that for @xmath58 the torus @xmath59 acts on the complex vector space @xmath60 via the action @xmath61 for some @xmath62-matrix @xmath33 with integer entries .
we endow the lie algebra of the torus @xmath63 with its standard inner product .
the action of the torus on @xmath60 is hamiltonian with respect to the standard symplectic structure @xmath64 .
let @xmath65 be an element of the lie algebra @xmath66 . denoting by @xmath67 the transposed matrix of @xmath33 a moment map @xmath68
is given by @xmath69 i.e. @xmath70 for the vector field @xmath71 on @xmath60 given by the infinitesimal action @xmath72 we will assume throughout this paper the following hypothesis , ( h ) : : the moment map @xmath73 is proper , @xmath74 is not empty , and @xmath75 acts freely on it .
it follows from ( h ) that the marsden - weinstein quotient @xmath76 is a compact symplectic manifold of dimension @xmath77 where the symplectic structure is induced from the standard symplectic structure on @xmath60 .
let @xmath78 for @xmath79 $ ] be a smooth family of @xmath75-invariant hamiltonians .
we denote by @xmath80 the hamiltonian vector field of @xmath81 .
furthermore , let @xmath82 for @xmath79 $ ] be a smooth family of @xmath75-invariant @xmath83-compatible almost complex structures , i.e. @xmath84 is a riemannian metric on @xmath60 . in this paper
we are proving existence of solutions @xmath85,\mathbb{c}^n\times \mathfrak{t}^k)$ ] of the following pde @xmath86 schroers @xcite discovered these equations for the special case where @xmath82 is the standard complex structure given by multiplication with @xmath87 and the hamiltonian @xmath88 vanishes identically , see also @xcite .
note that for the standard circle action on 1-dimensional complex space they reduce to the classical vortex equations .
they are examples of the symplectic vortex equations in temporal gauge on the cylinder .
the symplectic vortex equations can be defined more generally for hamiltonian group actions on a symplectic manifold .
they were introduced by cieliebak , gaio , mundet , and salamon in @xcite .
abbreviate @xmath89 the isotropy subgroup of the lagrangian submanifold @xmath90 . since by our standing assumption ( h ) there is a point @xmath91 which is fixed only by the identity in @xmath75 it follows that @xmath92 the gauge group @xmath93,t^k ) : h(0 ) , h(1 ) \in t^k_{\mathbb{r}^n}\}\ ] ] acts on the solutions of ( [ sve ] ) by @xmath94 we define the energy of a solution of ( [ sve ] ) by @xmath95 we finally introduce the space of admissible families of hamiltonians @xmath96 and the space of admissible families of almost complex structures @xmath97 .
the space @xmath96 consists of smooth families of @xmath75 invariant hamiltonians @xmath81 for which there exists a compact set @xmath98 such that the support of @xmath81 for every @xmath79 $ ] is contained in @xmath99 .
the space @xmath97 consists of smooth families of @xmath75-invariant @xmath83-compatible almost complex structures @xmath82 such that there exists a compact set @xmath100 such that @xmath82 equals outside of @xmath99 the standard complex structure , i.e. the multiplication by @xmath87 . +
our main result is the following theorem .
+ + * theorem a : * _ for generic pairs @xmath101 there exists at least two not gauge equivalent solutions of ( [ sve ] ) whose energy is positive and finite . _
let @xmath43 be a space on which a finite group @xmath6 acts and let @xmath102 be a @xmath6-invariant action functional all whose critical points are nondegenerate .
assume that the action of @xmath6 on @xmath103 is free .
we then can define two floer chain complexes @xmath104 and @xmath105 where the first one is generated by critical points of @xmath0 and the second one is generated by @xmath6-orbits of critical points of @xmath0 .
obviously , @xmath106 suppose that we are able to define floer homologies @xmath107 and @xmath108 out of the chain complexes above .
we are now in position to formulate our principle .
+ + * principle : * _ assume that @xmath109 then there exists a floer flow line of @xmath0 of positive , finite energy . _ + + to illustrate our principle we consider an example were @xmath43 is a finite dimensional manifold .
let @xmath43 be the two - sphere @xmath110 and let @xmath111 act on @xmath110 as the antipodal involution .
then the quotient @xmath112 is @xmath113 .
we consider the morse homology with @xmath114-coefficients . in this case
the assumption of our principle is satisfied , i.e.latexmath:[\[\mathrm{rk}\big(hm(s^2;\mathbb{z}_2)\big ) \neq understand what is going on we consider the standard morse function on @xmath113 which has one maximum , one critical point of index one , and one minimum , two morse flow lines from the maximum to the critical point of index one as well as two morse flow lines from the critical point of index one to the minimum . since we are considering morse homology with @xmath114-coefficients
the boundary operator on the morse chain complex on @xmath113 is zero .
however , on @xmath110 the two flow lines split off and the boundary operator of the morse complex on @xmath110 is not trivial anymore .
we first show how solution of ( [ sve ] ) arise as flow lines of an action functional .
we introduce the following path space @xmath116,\mathbb{c}^n \times \mathfrak{t}^k ) \colon v(0 ) , v(1 ) \in \mathbb{r}^n\}.\ ] ] for a smooth family of @xmath75-invariant hamiltonians @xmath81 we define the action functional @xmath117 by @xmath118 where @xmath119 is the liouville 1-form satisfying @xmath120 .
recall the gauge group @xmath93,t^k ) : h(0 ) , h(1 ) \in t^k_{\mathbb{r}^n}\}.\ ] ] it acts on @xmath121 by @xmath94 note that the differential of @xmath0 is invariant under the action of @xmath122 on @xmath121 . for a smooth family of @xmath75-invariant @xmath83-compatible almost complex structures @xmath82 define the metric @xmath123 on @xmath121 for @xmath124 and @xmath125 by @xmath126
then the gradient flow lines of @xmath127 with respect to @xmath123 are precisely the solutions of ( [ sve ] ) .
recall that @xmath96 denotes the space of families of admissible hamiltonians , i.e. torus invariant hamiltonians with compact support , and @xmath97 denotes the space of families of admissible almost complex structures , i.e families of torus invariant @xmath83-compatible almost complex structures which agree outside of a compact set with the standard complex structure on @xmath60 . for generic @xmath128
it was shown in @xcite that the floer homologies @xmath129 and @xmath130 are well defined .
theorem a follows from the following theorem .
* proof : * that @xmath133 is not zero follows from the results in @xcite .
there it was shown that the homology @xmath133 is given by the morse homology with coefficients in @xmath114 of the induced lagrangian in the marsden - weinstein quotient @xmath134 tensored with a novikov ring which is itself an infinite dimensional vector space over the field @xmath114 . since @xmath74 is not empty it follows that @xmath135 is not empty and hence @xmath133 is not zero . to show that @xmath136 is zero we make use of the fact that there exists a hamiltonian isotopy @xmath137 of @xmath60 generated by a smooth family of hamiltonians @xmath81 with support in a compact set @xmath138 such that @xmath139 to define the corresponding floer homotopy care has to be taken since the hamiltonians @xmath81 are not @xmath75-invariant anymore .
we therefore define the floer homotopy not on @xmath121 but on a coulomb section @xmath140 . here
is how this works .
there is a natural splitting @xmath141 where the contractible infinite dimensional lie group @xmath142 is given by elements @xmath143 satisfying @xmath144 .
the gauge group @xmath142 acts freely on @xmath121 .
moreover , for each path @xmath145,\mathfrak{t}^k)$ ] there is a unique element @xmath146 which puts @xmath147 into coulomb gauge @xmath148 i.e. @xmath149 is constant .
hence we may think of @xmath150,\mathbb{c}^n)\colon v(0),v(1 ) \in \mathbb{r}^n\}\times \mathfrak{t}^k\ ] ] as a global section in the principal @xmath142-bundle @xmath121 , i.e. the following diagram is commutative + + the @xmath49-metric @xmath50 on @xmath121 is @xmath142-invariant , hence it induces a quotient metric @xmath156 $ ] on @xmath157 and we denote by @xmath158\ ] ] its pullback on @xmath151 .
@xmath142-equivalence classes of gradient flow lines of @xmath127 with respect to @xmath50 are in natural one - to - one correspondence with gradient flow lines of the restriction of @xmath127 to @xmath151 with respect to the metric @xmath159 . to compute the metric
we introduce the following notation . for @xmath160 denote by @xmath161 the linear map defined by @xmath162 and by @xmath163 its dual with respect to the inner products on @xmath66 and on @xmath60
. then for @xmath164 where @xmath165 the metric @xmath159 is given by @xmath166 where @xmath167 is determined by @xmath168 by abuse of notation we will denote by @xmath127 also the restriction of @xmath127 to the coulomb section @xmath151 . in the following
we will drop the assumption that @xmath81 is @xmath75-invariant . to compute the gradient of @xmath127 with respect
to @xmath159 it is useful to abbreviate for @xmath169,\mathbb{c}^n)$ ] @xmath170 note that @xmath171 and @xmath172 vanish identically if @xmath88 is @xmath75-invariant . using this notation we get for the gradient @xmath173 where @xmath174 is given by @xmath175 hence gradient flow lines
satisfy the following equation @xmath176 up to the factor @xmath177 which measures how far is @xmath88 from being @xmath75-invariant these equations are examples of symplectic vortex equations .
however , since @xmath88 has compact support there exists a constant @xmath178 not depending on @xmath179 such that @xmath180 so the compactness result in @xcite can be adapted to our situation to prove that the floer homotopy is well defined . in our situation
the proof simplifies considerably , though .
this is firstly due to the fact that our gauge group is abelian so that we do not need the full strength of uhlenbeck s compactness theorem .
secondly thanks to the now established abstract perturbation techniques we do not have to bother about transversality any more and can work with the standard complex structure on @xmath60 .
since there are no derivatives of the standard complex structure the bubbling analysis can be avoided by directly using the elliptic estimate for the @xmath181-operator .
this proves the theorem .
@xmath182 we consider the special case of the standard @xmath21-action on @xmath22 .
we set the hamiltonian equal to zero and consider the standard complex structure on @xmath22 .
then the gradient flow lines are solutions of the classical vortex equations on the strip with lagrangian boundary condition
. the connected components of the critical set of @xmath0 are parametrised by @xmath183 .
it was shown by jaffe and taubes @xcite that between critical components whose difference in the @xmath184-factor is one there is exactly one solution up to gauge equivalence .
hence if one considers the floer complex modulo @xmath114 , the isotropy group of the lagrangian @xmath185 , then the floer boundary operator vanishes and the homology is nontrivial . on the other
if one does not mod out by @xmath114 , then the same phenomenon as in the case of the standard morse function on @xmath113 happens .
the flow lines split off and the floer homology is zero .
99 e.bogomolnyi , _ the stability of classical solutions _ , sov.j.nucl.phys . * 24 * ( 1976 ) , 449454 .
k.cieliebak , r.gaio , i.mundet , d.salamon , _ the symplectic vortex equations and invariants of hamiltonian group actions _ , j.symplectic geom . * 1 * ( 2002 ) , no.3 , 543 - 645 .
k.cieliebak , r.gaio , d.salamon , _
@xmath186-holomorphic curves , moment maps , and invariants of hamiltonian group actions _ , int.math.res.not .
( 2000 ) , no.16 , 831 - 882 .
a.floer , _ proof of the arnold conjecture for surfaces and generalizations to certain khler manifolds _ , duke math.j * 53 * ( 1986 ) , 132 .
a.floer , _ morse theory for lagrangian intersections _ , j.diff.geom .
* 28 * ( 1988 ) , 513547 .
a.floer , _ the unregularized gradient flow of the symplectic action _ , comm.pure appl.math . * 41 * ( 1988 ) , 775813 .
a.floer , _ symplectic fixed points and holomorphic spheres _ , comm.math.phys .
* 120 * ( 1989 ) , 575611 .
u.frauenfelder , _ the arnold - givental conjecture and moment floer homology _ , int.math.res . not .
( 2004 ) , no 42 , 21792269 .
j.hong , y.kim , p.pac , _ multivortex solutions of the abelian chern - simons theory _ , phys.rev.lett . *
64 * , 22302233 ( 1990 ) .
r.jackiw , e.weinberg , _ self - dual chern - simons vortices _ , phys.rev.lett .
* 64 * , 22342237 ( 1990 ) .
a.jaffe , c.taubes , _ vortices and monopoles _ , progress in physics * 2 * ( 1980 ) .
b.schroers , _ the spectrum of bogomolnyi solitons in gauged linear sigma models _ , nucl.phys.b * 475 * , 440468 ( 1996 ) .
y.yang , _ solitons in field theory and nonlinear analysis _ , springer , new york ( 2001 ) . | in this paper we use floer homology to prove existence of solutions of a generalized abelian higgs equation introduced by schroers . |
transport of inertial particles is involved in several fields of science , e.g. droplets growth in clouds , @xcite , planetary formations , @xcite , or plankton accumulation in the ocean , @xcite . as far as technological applications are concerned , inertial particle dynamics is crucial for solid or liquid fuelled rockets , injection systems of internal combustion engines or for sediments accumulation in pipelines , e.g. @xcite .
inertial particles differ from perfectly lagrangian tracers due to inertia which prevents them from following the flow trajectories .
the main effect consists of `` preferential accumulation '' , see for instance @xcite . in homogeneous isotropic conditions it amounts to the small - scale clustering discussed in a number of recent papers , @xcite
. in presence of inhomogeneity new features emerge leading to the so - called turbophoresis as a preferential accumulation near the boundary in wall turbulence @xcite . under appropriate conditions
, particles may achieve extremely large concentrations at the wall with a substantial reduction of mobility .
this turbulence - induced transport and the issuing preferential accumulation as been addressed in a number of papers dealing with a variety of configurations , from boundary layers to planar channels and pipes , attacked both from the experimental @xcite and the numerical @xcite side .
though a complete understanding of the phenomenon is still lacking , the advection of the particles by the coherent motions in the wall - layer is certainly essential , as discussed by @xcite . in other words
the structures responsible for particle accumulation at the wall are the same which sustain turbulence kinetic energy production in the buffer layer , see also @xcite .
turbophoresis and small scale clustering are different aspects of the same inertial particle dynamics .
both phenomena are induced by non trivial phase relationships due to quasi - coherent vortical structures .
the main difference is provided by the characteristic scales , associated with the kolmogorov time unit @xmath1 in one case@xmath2 being kolmogorov length and @xmath3 the kinematic viscosity and with the larger energy producing time scale in the other .
inhomogeneity is essential to have spatial segregation .
for instance , in the kinetic model presented in @xcite , the spatial transport of particle concentration presents , beside a fick - like gradient type diffusion component , a contribution associated with the spatial variations of turbulence intensity .
however anisotropy is probably a key ingredient of the process , see e.g. the preferential direction of the trajectories of particles approaching the wall @xcite .
the two features are strongly entangled in wall bounded flows . a special flow exists however the homogeneous shear flow in a confined box which retains most of the anisotropic dynamics of wall bounded flows still preserving , on average , spatial homogeneity
. the flow is bound by a computational box of fixed extension and its integral scale grows initially to eventually saturate due to confinement .
target of the analysis is the statically steady state with time independent ensemble averages .
similar features are found in the the experimental data by @xcite . by using an active grid to generate a flow with integral scale close to the transversal dimension of the apparatus ,
the authors were able to achieve confinement from the outstart .
as a consequence , the integral scale did not increase downstream , see also the discussion in @xcite .
our flow shares with the wall - layer streamwise vortices and turbulent kinetic energy production mechanisms . in the numerical experiment
, this corresponds to pseudo - cyclic fluctuations associated with the regeneration of streamwise vortical structures .
velocity fluctuations are strongly anisotropic at the large scales driven by production while , for smaller separations , the classical energy transfer mechanisms become effective in inducing re - isotropization .
this classical issue , see e.g. @xcite , has been recently revisited by more complete diagnostic tools ( e.g. so(3 ) decomposition of turbulent fluctuations ) able to quantify on a scale - by - scale basis the amount of anisotropy in the carrier fluid , as discussed both experimentally @xcite and numerically @xcite ( see @xcite for a review ) .
despite anisotropy of the velocity field is now well understood and the carrier fluid shows tendency towards isotropy recovery below the shear scale , @xmath4 with @xmath5 the average turbulent kinetic energy dissipation rate per unit mass and @xmath6 the average shear rate , the behavior of particle distributions is still not fully explored .
anisotropic transport of inertial particles has been recently addressed by @xcite , who analyzed numerically the initial transient of the homogeneous shear flow , i.e. before saturation of the integral scale occurs , with the purpose of modeling unconfined conditions .
the focus was mainly on the comparison of different particle dynamics models .
however , by considering particle configurations in orthogonal planes , the authors also discussed the anisotropy of particle clusters concluding that particles are most concentrated in the streamwise and least concentrated in the cross - stream direction .
the same flow was dealt with by @xcite to investigate issues such as turbulence modulation in the two way coupling regime or heat transfer induced by the disperse phase .
purpose of the present paper is the quantitative evaluation of the shear induced anisotropy in particles clustering .
it is now well known that particles respond to the fluid velocity fluctuations in a certain range of scales which is determined by their stokes time .
the relevant parameter is the stokes number , ratio of particle stokes time and flow time scale . in order to work with well defined conditions
, one needs a shear flow whose characteristic time and length scales are constant in time .
the best candidate , is the statistically steady homogeneous shear flow in a confined box we have described above . in this flow , below the shear scale
, the velocity fluctuations tend to recover isotropy .
the question is then what happens to particle clusters .
do they become isotropic in the smallest scales ?
for given velocity field , how is the geometry of the clusters affected by the relaxation time of the particles ? in fact , the main contribution of the present study is the quantitative evidence that particle distributions do not necessarily reduce their anisotropy at small scales , despite the isotropy recovery occurring in the velocity field .
rather clusters anisotropy may even grow below the kolmogorov length where the velocity field is smooth and almost isotropic . as a matter of fact ,
inertia manifest itself in a rather peculiar and unexpected way , and leads , under certain coupling conditions , to singular particle distributions which viscosity can not regularize @xcite . after introducing a suitable observable the angular distribution function its spherical decomposition is used to evaluate the relative importance of its different components .
the scaling exponents of the respective singularities show that , under appropriate conditions , anisotropy is a leading order effect which may easily persist down to vanishing separations .
the data offer preliminary evidence of the anisotropic geometry of the fractal support of inertial particle distributions under shear flows , thus non - trivially extending results recently achieved for isotropic transport .
concerning the carrier fluid , the velocity field @xmath7 is decomposed into a mean flow @xmath8 and a fluctuation @xmath9 , see figure [ fig : sketch ] for notations .
rogallo s technique is employed to rewrite the navier - stokes equations for velocity fluctuations in a deforming coordinate system convected by the mean flow according to the transformation of variables @xmath10 , @xcite .
the resulting system is numerically integrated by a pseudo - spectral method combined with a fourth order runge - kutta scheme for temporal evolution , see @xcite .
the two parameters controlling the homogeneous shear flow are the taylor - reynolds number @xmath11 and the shear strength @xmath12 . for the simulations discussed below they are @xmath13 and @xmath14 , corresponding to a ratio of shear to kolmogorov scale @xmath15 .
navier - stokes equations are integrated in a @xmath16 periodic box with a resolution of @xmath17 fourier modes corresponding to @xmath18 collocation points in physical space due to the @xmath19 dealiasing rule .
the kolmogorov scale is @xmath20 which correspond to @xmath21 ensuring sufficient resoltion at small scales .
actually a well resolved velocity field is crucial to minimize numerical errors associated with the interpolation of the fluid velocity at particle positions which is necessary to advect the particles in the present mixed eulerian - lagrangian formulation , see e.g. @xcite .
the disperse phase consists of diluted particles with mass density @xmath22 much larger than the carrier fluid @xmath23 , assumed small enough to be modeled as material points . at this dilution self - interactions and the back reaction on the fluid
are negligible , leaving the stokes drag in the relative motion with the fluid as the only relevant force on each particle , @xcite .
accordingly , the equations for particles position @xmath24 and velocity @xmath25 read @xmath26 \end{array}\ ] ] where @xmath27 is the instantaneous fluid velocity evaluated at @xmath24 and @xmath28 is the stokes relaxation time ( @xmath29 denotes the particle diameter ) .
the particle dynamics is controlled by the ratio of @xmath30 to a characteristic flow time scale , typically the kolmogorov time scale @xmath31 , i.e. the relevant control parameter is the stokes number @xmath32 .
particle velocities are decomposed as @xmath33 + u_i^p$ ] where @xmath34 denotes the particle velocity deviation with respect to the local mean flow of the carrier fluid .
( [ eq_part_phys ] ) can be written as @xmath35 - \frac{d u_i}{d t } \
, , \end{array}\ ] ] to be finally rearranged in rogallo s computational space as ( see also @xcite ) @xmath36 - s u^p_2 \delta_{i1 } \ . \end{array}\ ] ] the particle equations are integrated by the same fourth order runge - kutta scheme used for the navier - stokes equations , with fluid velocity at particle positions evaluated by tri - linear interpolation
. the accuracy of the interpolation scheme may be an issue . to assess its effect on the numerical results , we have preliminarily run two different simulations at half the resolution and smaller reynolds number of the cases to be discussed in the main body of the paper .
the two simulations employ two different interpolation schemes , namely a linear and a quadratic one . as always in turbulence
, comparisons need to be made in terms of the relevant statistical observables .
we anticipate that here we deal with the different projections of the angular distribution function , to be introduced in [ results ] . in figure [ interp_comp ]
the solid lines denote results obtained with the linear interpolation scheme while symbols correspond to the quadratic lagrange polynomials .
the difference can not be appreciated on the scale of the diagram , and is always below the statistical accuracy of the data . starting from an
already fully developed fluid velocity field in statistically steady conditions , five different populations of @xmath37 particles each , with stokes numbers @xmath38 , are initialized with random and homogeneous positions and velocities matching that of the local fluid .
samples for particle statistics corresponding to @xmath39 independent snapshots separated in time by @xmath40 are collected after an initial transient of @xmath41 .
discarding the initial transient is crucial to have results independent from the rather arbitrary initial state used to initialize the particles .
we observe that , also under this respect , a statisticaly steady flow is mandatory to have well definite experimental conditions , especially in cases where the response of particle populations with different relaxation times are compared .
a visual impression of instantaneous particles configurations is provided in figure [ fig : part_pos ] , where slices of the domain in selected coordinate planes are displayed for three different stokes numbers .
the typical particle distribution exhibits many voids , strongly correlated with high enstrophy , @xcite , intertwined with thin `` stretched '' regions where particles concentrate .
clustering is specially manifest near @xmath42 , see the mid panel of the figure in comparison with top and bottom ones which show more even spreading .
the typical void dimension , as caught by the eye , is larger at our largest stokes number , @xmath43 .
the ballistic limit , where particles follow their trajectories with no significant influence from the fluid and the expected spatial distribution is homogeneous , is apparently still far away . in the opposite extreme case ,
passive tracers are recovered for vanishing @xmath44 , particles follow the fluid path , and , again , homogeneity is eventually restored .
in fact , clustering still takes place , though at smaller scales , for the smallest stokes number we have considered , consistently with theoretical arguments @xcite and numerical simulations @xcite for homogeneous and isotropic turbulence aimed at explaining droplets growth in clouds . despite of the mentioned similarities with isotropic flows ,
ours manifests specific features compelled by large scale anisotropy .
the shear - induced orientation is apparent from the bottom - left / top - right alignment of particles sheets in the shear plane @xmath45 , see the right panels of figure [ fig : part_pos ] .
this behavior , clearly visible at @xmath46 , is still discernible in the other two cases . since , apparently , the effect is strong and persistent , we are interested in putting forward suitable tools to evaluate the anisotropy of clustering .
as we shall see , this is best done by extending in due form a line of analysis proved successful for the isotropic case .
the main statistical tool is the radial distribution function ( rdf ) of particle pairs @xmath47 which is a function of radial distance @xmath48 , see e.g. @xcite where the rdf is dealt with for isotropic flows .
the rdf , sometimes called correlation function , is defined as @xmath49 where @xmath50 is the density of pairs in the whole volume @xmath51 and @xmath52 is the number of pairs in a ball @xmath53 of radius @xmath48 .
the concept is easily extended to anisotropic cases by considering the number of pairs @xmath54 contained in a spherical cone of radius @xmath48 , with axis along the direction @xmath55 and solid angle @xmath56 , see the sketch in the right panel of figure [ fig : sketch ] . by this definition
the number of pairs in the ball @xmath53 is @xmath57 , hence @xmath58 .
we define the angular distribution function ( adf ) as @xmath59 which retains information on the angular dependence of the distribution .
the rdf is the spherical average of the adf @xmath60 and it is shown in figure [ fig : radial_pdf ] for a few particle populations . the behavior of the rdf near the origin , @xmath61 , is related to important geometrical features of the spatial distribution .
specifically , @xmath62 is the so - called correlation dimension of the multi - fractal measure associated with the particle density , @xcite .
a positive @xmath63 indicates the occurrence of small scale clustering .
its value is inferred from the slope near the origin in the log - log plots shown in figure [ fig : radial_pdf ] , see the scaling behavior apparent in the range @xmath64 $ ] . from the figure , particles with @xmath65 exhibit maximum accumulation ,
i.e. the rdf diverges at a faster rate as @xmath48 is decreased , see also @xcite .
the solid lines superimposed to the present data correspond to the scaling laws extracted at matching stokes number by @xcite in isotropic conditions .
the agreement between our data and the isotropic ones is remarkable , showing that , even under strong shear , certain features of the clustering process may be universal .
small scale clustering is not shared by heaviest particles .
they show instead the saturation of the correlation function to a constant value @xmath66 below a critical scale @xmath67 , see e.g. the open squares in figure [ fig : radial_pdf ] with @xmath68 .
this means that the number of pairs below @xmath67 is proportional to volume with an effective density @xmath69 larger than its overall mean value @xmath70 . by inspection of figure
[ fig : radial_pdf ] , for the heaviest particles ( squares ) , the saturation occurs at @xmath71 .
the interpretation is that eddies with a time scale sufficiently smaller then particle stokes time do not influence the clustering process .
the saturation scale should then correspond to the size @xmath72 of the smallest eddies able to aggregate particles . as an order of magnitude estimate , our data are consistent with the results given by @xcite for the inertial range of isotropic turbulence , where the authors find @xmath73 with @xmath74 .
our data match this estimate also for the run at @xmath75 , where the saturation occurs close to the kolmogorov scale . for lighter particles , clustering keeps on going below the kolmogorov length and its lower limit can not be interpreted by arguments which , tuned by experiment , are nevertheless taken from inertial range theory . concerning the large scale behavior ,
heavier particles apparently begin to show accumulation at larger scales .
each population is uniform with @xmath76 at very large @xmath48 .
this trend is followed by the lightest particles down to @xmath77 below which they begin to follow the power - law .
the rdf starts deviating from the uniform distribution@xmath76 at a scale @xmath78 which increases monotonically with the stokes number , see .
e.g. the range @xmath79 $ ] .
this is consistent with intuition , since the particle stokes time progressively matches the eddy - turnover time of larger turbulent eddies .
the data again agree reasonably well with the results of @xcite , which identify the range of time scales relevant to clustering as @xmath80 where @xmath81 is the eddy turn over time at scale @xmath82 . in terms of lengths , clustering starts to occur at @xmath83 where @xmath84 is the integral scale and @xmath85 ( @xmath86 ) . using this estimate , our heaviest particles
are expected to begin accumulating at the integral scale @xmath87 while , e.g. , the lightest ones at @xmath88 , which is not too far from the value @xmath77 we infer from the plot .
the rdf quantitatively confirms the overall impression gained from the visualizations of figure [ fig : part_pos ] .
the strong anisotropy apparent in those plots , however , needs a description in terms of the more complete adf shown in figure [ fig : ball_vera ] as a contour plot on the unit sphere for two separations @xmath48 and particles with @xmath42 .
the adf has been normalized with its average on the unit sphere , @xmath47 , in order to compare the relative anisotropy content of different scales . from figure
[ fig : ball_vera ] , as separation decreases ( i.e. moving from the right , @xmath89 , to the left panel , @xmath90 ) , the normalized adf exhibits preferential clustering in directions consistent with those observed in the visualizations of figure [ fig : part_pos ] .
the adf allows for a systematic evaluation of anisotropy in particle clustering . for given separation @xmath48 ,
its angular dependence can be resolved in terms of spherical harmonics , @xmath91 in this notation , the classical rdf @xmath47 is the projection of the adf on the isotropic sector @xmath92 , namely @xmath93 this decomposition amounts in projecting our function on orthogonal sub - spaces invariant under rotations .
each sub - space is labeled by the index @xmath94 and it is spanned by @xmath95 base elements @xmath96 . growing levels of anisotropy are checked by increasing @xmath94 , consistently with its geometrical meaning as number of zero crossings of @xmath96 .
figure [ fig : y_ij ] shows the normalized projections on sectors @xmath97 , i.e. the normalized amplitudes @xmath98 , for the particle population with @xmath99 . for given @xmath94
most modes are negligible and the figure reports only those with significant signal level .
the @xmath100 mode provides the most significant contribution to the anisotropic component of @xmath101 .
the corresponding spherical harmonics , @xmath102 , roughly selects the intensity of the signal along the principal direction of the mean deformation tensor which corresponds to maximum straining and is inclined @xmath103 in the mean flow plane .
@xmath102 is negative in the first and third quadrant and positive in the others , thus explaining the negative sign of @xmath104 in figure [ fig : y_ij ] .
this description well captures the alignment of most thin particle clusters observed in the mid panel of figure [ fig : part_pos ] .
the signal content rapidly decreases with the order of the sector @xmath94 and the adf is satisfactorily reconstructed
see decomposition ( [ so3_dec])using only the first few sectors with @xmath105 as shown in bottom panels of figure [ fig : ball_vera ] .
we conclude that the probability of finding two particles at separation @xmath106 along the main straining direction is @xmath107 higher than in the perpendicular direction in the mean flow plane .
as discussed , the adf provides a quantitative account of the anisotropy induced by the fluid velocity field on the disperse phase .
as shown below , it can be effectively used to parameterize the level of anisotropy through the scales in terms of the stokes number .
the plot of figure [ fig : y_2_2 ] gives the normalized amplitude of the most energetic anisotropic mode in absolute value@xmath108for our set of stokes numbers @xmath109 ranging from heavy to light particles .
focusing on the heaviest particles , @xmath110 , the relative amplitude of the strongest anisotropic mode first increases towards the small scales to reach a maximum at @xmath111 .
below this scale the anisotropy level decreases , until the very small scales become essentially isotropic .
connecting this result with the previous discussion concerning the saturation of the rdf @xmath112 , we conclude that the heaviest particles show a regular concentration at scales smaller than @xmath67 where the distribution recovers isotropy .
particles with smaller stokes numbers behave in an entirely different way .
the anisotropy , as measured by the ratio @xmath113 , substantially increases to saturate at small scales close to kolmogorov length .
it keeps an almost constant value below @xmath2 . in other words ,
the clustering process maintains its anisotropic features even below the dissipative scale for sufficiently small stokes number particles .
remind that the overall clustering process described by @xmath112 is here characterized by a singular exponent @xmath63 .
the saturation observed on the ratio @xmath113 implies that the dominating anisotropic contribution inherits the same behavior , @xmath114 .
we have provided evidence that large scale shear induces preferential orientation on the patterns a disperse phase of small inertial particles forms in turbulence .
the effect is indirect : the shear imprints anisotropy on velocity fluctuations which , in turn , arrange particle configurations in directionally biased clusters .
recently it became increasingly clear that the multi - scale nature of the velocity field is crucial in explaining most features of particles distributions .
typically , a range of eddies exists able to break spatial homogeneity of particles configurations .
the particles segregate and their pattern shows local concentrations and voids on length scales correlated with those of the inducing eddies .
the range of scales of the inducing eddies is determined by the stokes time of the particles , and moves from integral down to kolmogorov length with reducing the relaxation time . for light particles ,
clustering reaches down kolmogorov scale , leading to a singularity in the radial distribution function .
the exponent of the singularity is a measure of its intensity .
all these features , originally found in experiments and numerical simulations of isotropic turbulence , are also present in our anisotropic fields .
they are thus generic aspects of the clustering process which seem independent of the specific geometry of the forcing . at a qualitative level , the specific characteristics of clustering under anisotropic advection consists of preferential orientation of the particle patterns . in order to quantify the new scale - dependent features added by the non - trivial geometry of the forcing
, we have introduced the concept of angular distribution function .
it can be understood as a generalization of the previous radial distribution function , to which it reduces by averaging on the unit sphere , i.e. by performing the isotropic projection .
this quantitative tool has led to our most unexpected finding : the advecting field anisotropy , known to be confined to the large scales , affects the singular , small scale , clustering process .
in fact , anisotropy results into a strong directionality of the probability to find a couple of particles at viscous scale separation , with @xmath0 variations on the solid angle easily observed .
technically , for stokes number order unity , the anisotropic component of the angular distribution function diverges at small scales with singularity exponent comparable to that found in the isotropic projection ( rdf ) .
for very small stokes number , we can not even exclude that the singularity exponent of the strongest anisotropic sector may even exceed that of the radial distribution function .
conversely , heavy particles appear to preferentially concentrate on finite sized patches endowed with a range of multi - scale and shear oriented features , with finest scales more or less evenly and isotropically distributed .
the geometrical properties of patterns of inertial particles differ considerably from that one could naively guess from velocity fluctuations .
recent findings , extending somehow a number of previous results on shear induced anisotropy @xcite , show that velocity fluctuations manifest two neatly distinct ranges , one dominated by the production of turbulent kinetic energy above the shear scale @xmath115 , the other , corresponding to the classical inertial range of kolmogorov theory below .
in the two ranges , velocity fluctuations display different isotropy recovery rates , a smaller one in the production range , a larger one in the inertial transfer range between @xmath115 and the viscous scale @xmath2 , @xcite .
actually , concerning the velocity field , isotropy is matter of fact always recovered at dissipative scales , provided the scale separation @xmath116 is large enough , i.e. the local reynolds number ( taylor - reynolds number , or , equivalently for wall bounded flows the wall normal distance in inner units , @xmath117 , with @xmath118 the wall shear stress ) is sufficiently large .
this condition is violated in an essential way only very close to solid walls in wall - bounded turbulence ( small @xmath119 ) .
the anisotropy in the particle configurations depends strongly on the properties of the advecting velocity field .
however , despite of the isotropy recovery of the velocity field , isotropy may never be recovered in the small scales of the clusters , as it happens at small stokes number .
actually , in the range from @xmath115 to @xmath2 clustering of light inertial particles shows a substantial increase of directionality . from our results at moderate reynolds number one can conjecture the behavior of the clusters at high reynolds numbers . in principle
the intermittency of the turbulent field may induce a dependence on reynolds number .
in fact , as shown in @xcite for isotropic flows , the fractal properties of particle distributions depend at most weakly on reynolds number and strongly on the stokes number . for given geometry of the external forcing
, i.e. fixing the integral and the shear scale , and at given particle stokes time , the increase of the reynolds number corresponds to increasing the stokes number based on kolmogorov time , @xmath44 .
given the weak dependence on reynolds number , this is somehow equivalent to reading figs .
[ fig : radial_pdf ] and [ fig : y_2_2 ] by successively moving from light to heavy particles . along the process , we infer the saturation at small scales of the radial distribution function ( @xmath120 , fig . [
fig : radial_pdf ] ) and the small scale isotropy of the clusters ( fig .
[ fig : y_2_2 ] ) .
clustering is confined to the intermediate scales where it shows high levels of anisotropy .
this means that , in the limit of large reynolds number , any particles population should be organized in anisotropic finite - sized patches which are eventually uniform and isotropic in their finest scales .
we stress once more that an intermediate range of scales always exists , however , where a multi - scale aggregation process takes place with a substantial directionality of the clusters .
on the other hand , at finite reynolds number , sufficiently small particles small relaxation time will always show small scale clustering , in the sense of a singularity in the radial distribution function . in this case
anisotropy may persist below kolmogorov scale , as described by the angular distribution function at small separations for the lightest populations in our simulations .
this finite reynolds number effect becomes extremely important in the near wall region of turbulent wall bounded flows , where the local reynolds number constructed with the distance from the wall is not huge and the velocity field is meanwhile strongly anisotropic .
the issuing anisotropy of the fine scales of particles clusters will then have a significant impact on phenomena of collision , aggregation of dusts into larger particles , evaporation / condensation rate of droplets in pipe lines and a number of other physically and technologically significant contexts . a final comment concerns the extension of the present results to the near wall region of wall bounded flows .
as shown by recent results on the scale - by - scale statistics of the velocity field , e.g. energy transfer across scales , spectral distribution of turbulent kinetic energy production , intermittency and anisotropy , the homogeneous shear flow reproduces the essential features observed in the wall region , despite of significant differences in the large scale geometry of the two systems .
however , particles in the wall region are strongly affected by turbophoresis which is a predominant effect associated with inhomogeneity . in wall
flows particle segregation is controlled by the two concurrent processes of small scale anisotropic clustering and of accumulation at the wall . clearly , the focus of the present paper is on the former one , leaving the combined analysis of the two effects for future investigations .
, in the @xmath122 direction , is a function of @xmath123 , with @xmath124 . for a linear mean profile ,
the shear rate @xmath125 is constant .
right : sketch of the spherical cone of amplitude @xmath56 in direction @xmath126 .
[ fig : sketch],title="fig : " ] , in the @xmath122 direction , is a function of @xmath123 , with @xmath124 . for a linear mean profile ,
the shear rate @xmath125 is constant .
right : sketch of the spherical cone of amplitude @xmath56 in direction @xmath126 .
[ fig : sketch],title="fig : " ] + , @xmath127 , @xmath42 ) , in a @xmath128 computational box with @xmath129 collocations points corresponding to @xmath130 .
the disperse phase is computed by using two different schemes to interpolate the fluid velocity at particle positions , namely linear interpolation ( solid line ) and quadratic lagrange polynomials ( symbols ) .
the statistical observables shown in the plots are defined in [ results ] .
left panel : @xmath131 , projection of the adf in the isotropic sector .
right panel : @xmath108 , normalized most energetic anisotropic component of the adf .
[ interp_comp],title="fig : " ] , @xmath127 , @xmath42 ) , in a @xmath128 computational box with @xmath129 collocations points corresponding to @xmath130 .
the disperse phase is computed by using two different schemes to interpolate the fluid velocity at particle positions , namely linear interpolation ( solid line ) and quadratic lagrange polynomials ( symbols ) .
the statistical observables shown in the plots are defined in [ results ] .
left panel : @xmath131 , projection of the adf in the isotropic sector .
right panel : @xmath108 , normalized most energetic anisotropic component of the adf .
[ interp_comp],title="fig : " ] . left panel adf computed at separation @xmath133 ; right panel @xmath134 .
data for @xmath42 .
bottom panels : estimate of adf by using only the isotropic sector and the @xmath135 sector .
[ fig : ball_vera ] , title="fig : " ] . left panel adf computed at separation @xmath133 ; right panel @xmath134 .
data for @xmath42 .
bottom panels : estimate of adf by using only the isotropic sector and the @xmath135 sector .
[ fig : ball_vera ] , title="fig : " ] + . left panel adf computed at separation @xmath133 ; right panel @xmath134 .
data for @xmath42 .
bottom panels : estimate of adf by using only the isotropic sector and the @xmath135 sector .
[ fig : ball_vera ] , title="fig : " ] . left panel adf computed at separation @xmath133 ; right panel @xmath134 .
data for @xmath42 .
bottom panels : estimate of adf by using only the isotropic sector and the @xmath135 sector .
[ fig : ball_vera ] , title="fig : " ] | recently , clustering of inertial particles in turbulence has been thoroughly analyzed for statistically homogeneous isotropic flows .
phenomenologically , spatial homogeneity of particles configurations is broken by the advection of a range of eddies determined by the stokes relaxation time of the particles which results in a multi - scale distribution of local concentrations and voids .
much less is known concerning anisotropic flows . here , by addressing direct numerical simulations ( dns ) of a statistically steady particle - laden homogeneous shear flow ,
we provide evidence that the mean shear preferentially orients particle patterns . by imprinting anisotropy on large scales velocity fluctuations
, the shear indirectly affects the geometry of the clusters .
quantitative evaluation is provided by a purposely designed tool , the angular distribution function of particle pairs ( adf ) , which allows to address the anisotropy content of particles aggregates on a scale by scale basis .
the data provide evidence that , depending on the stokes relaxation time of the particles , anisotropic clustering may occur even in the range of scales where the carrier phase velocity field is already recovering isotropy .
the strength of the singularity in the anisotropic component of the adf quantifies the level of fine scale anisotropy , which may even reach values of more than @xmath0 direction - dependent variation in the probability to find two close - by particles at viscous scale separation . |
neural models aim at elucidating the grounding architecture that regulates information processing in biological nervous systems @xcite .
the level of abstraction that is accommodated for in the chosen mathematical scheme reflects the specific topic under investigations .
detailed single neurons models can be devised , which account for the sharp changes in the action potential .
the number of neurons in the human cortex is extraordinarily large and for many practical aspects a bottom - up approach that moves from a detailed representation of each individual unit is virtually impracticable .
the physiology of neurons is also quite intricate .
a single neuron receives input from thousand of axon terminals coming from other neurons , but the inherent degree of redundancy contributes to drastically reduce the effective connections to a limited subset .
almost @xmath0 of the neurons are excitatory , whereas the remaining @xmath1 belong to the inhibitory type .
irrespectively of their intimate characteristics , neurons can be pictured as complex switching devices which organize into two macroscopic families of homologous constituents , the activators and inhibitors .
this enables in turn to climb the hierarchy of complexity , and elaborate , to a suitable level of coarse graining , on the interlaced dynamics of homogenous populations of binary neurons .
the celebrated wilson - cowan ( wc ) model @xcite and its further extensions provide an interesting arena to shed light onto the complex dynamics of a multi - species neuronal system .
the model displays elementary limit cycle behavior , mimicking in silico neural oscillations .
stimulus - dependent evoked responses are also predicted , as well as the existence of multiple stable states .
the original wc model is deterministic in nature and the average concentration of active / inactive excitatory / inhibitory neurons appear to be self - consistently ruled by a pair of coupled ordinary differential equations .
spatially extended versions of the model can be straightforwardly designed which in turn amount to operate with partial differential equations @xcite .
deterministic systems can be analyzed by resorting to a gallery of well - developed tools from the theory of dynamical systems and involving concepts such as trajectories in phase space and attractors of various kinds @xcite .
stochastic systems exhibit an equivalently rich plethora of attributes , ranging from stochastic fluctuations about the deterministic trajectories , rare events which prompt the transitions from one attractor of the deterministic dynamics to another , stationary probability distributions , to mention a few @xcite .
the stochasticity can be externally imposed and hence impact on the deterministic scenario , as an exogenous source of disturbance .
more intriguing is instead the role played by intrinsic perturbation , stemming from the intimate discreteness of the system under scrutiny .
individual - based effects should be unavoidably accommodated for in any sensible model of natural phenomena and yield an endogenous stochastic contribution , which lacks instead under the idealized continuum representation . at first sight
it might appear surprising that stochastic effects are important when the interacting population consist of a large number of individual constituents .
demographic fluctuations can however amplify through resonant effects @xcite and drive the spontaneous emergence of collective macroscopic patterns , both in time @xcite and in space @xcite , marking a clear distinction between stochastic and deterministic viewpoints .
endogenous noise is certainly relevant to neural systems @xcite .
channel noise arising from the variability in the opening and closing of a finite number of ion channels , is a microscopic source of intrinsic disturbance . at the population level ,
the instantaneous photograph of the system is drawn by looking at the number of active neurons , which sits on a given patch .
state transition can be assigned to monitor the evolution of the inspected system via a suitably defined master equation , the mean field deterministic picture being eventually recovered in the thermodynamic limit . following these lines of reasoning ,
the wc model has been recently revised under the generalized stochastic angle , so to cast the accent on the specific traits which ultimately emanates from the endogenous component of the noise @xcite .
our paper is positioned in this context @xcite .
we will in particular reformulate the wc dynamics as a simple birth and death stochastic process , with non linear transition rates .
the obtained master equation will be expanded in powers of the inverse population size so to yield an approximate stochastic description for the density of the interacting species .
more specifically , the fluctuating hydrodynamics for the stochastic wc model is represented by a pair of non linear langevin equations , with multiplicative noise .
these latter are shown to provide an adequate description of the original stochastic process , the analysis being specifically targeted to the domain of dynamical bistability . to gain further insight
we will confine ourselves to inspecting the special setting where the bistability solely interests one the two species , the excitatory neurons , while the concentration of the inhibitors converges to a stationary stable fixed point , in the limit of diverging population . working in this setting
, one can effectively reduce the system to just one langevin equation for the excitatory species : an additional contribution appears in the expression for the multiplicative noise , which exemplifies the action exerted by the inhibitory neuron .
the population of inhibitors acts as a source of intrinsic noise , shaking the discrete ensemble of excitators from the _
inside_. interestingly , the perturbation magnifies as @xmath2 , the relative ratio of inhibitory versus excitatory neurons , gets reduced . for actual physiological conditions ( @xmath3 ) ,
the mechanism here identified is hence potentially relevant .
finite size noise and its inhibitory based component causes an extension of the region deputed to the excitatory bistability , as compared to deterministic predictions .
this contribute in turn to broaden the plasticity of the system and elevate its degree of inherent flexibility .
the analytical tools developed in the early part of the paper are subsequently generalized to the relevant setting where both species can undergo bistable dynamics . a recursive perturbative scheme targeted to such a general scenario
is successfully implemented and tested versus direct numerical simulation .
the paper is organized as follows . in the next section
we will introduce the stochastic version of the wilson - cowan model and carry out the analysis that yields to the fluctuating hydrodynamical picture . in section 3
, we will silence the retroaction of the activators on the inhibitors , and assign the parameters so to fall in the region of deterministic bistability for the excitatory species . by linearizing around the fixed point of the inhibitors
, one can reduce the overall dynamics to a non - linear langevin equation for the density of activators : the multiplicative noise incorporates the effect which stems from the population of inhibitors . working in this framework
, one can analytically determine the probability distribution for both species and quantitatively elaborate on their mutualist interdependence .
noise - induced bistability is uncovered , the effect being progressively more evident the smaller the population ratio @xmath2 . building on these premises , in section 4
we generalize the study to the setting where the retroaction of the activators is restored . finally in section 5 we sum up and draw our conclusions .
the wilson - cowan ( wc ) model describes the self - consistent interactions between distinct populations of excitatory and inhibitory neurons . we shall hereafter provide a plausible microscopic formulation of the dynamics that eventually converges to the sought wc model in the deterministic limit . as mentioned earlier , by providing a microscopic characterization of the scrutinized dynamics we aim at shedding light onto the role played by finite size fluctuations .
in particular , working under the kramers - moyal approximation we will derive an effective description of the stochastic model in terms of fluctuating hydrodynamical equations .
the model deals with a spatially homogeneous distribution of the interacting entities .
extensions of the proposed scheme which assume the constituents dislocated on the nodes of a complex graph , so to probe the heterogeneity in the emerging distribution , are certainly conceivable . in the following
, individual excitatory and inhibitory neurons will be respectively identified with the symbols @xmath4 and @xmath5 .
both excitatory and inhibitory neurons can manifest into two distinct states : activated ( @xmath6 ) or quiescent ( @xmath7 ) . to identify to which
subpopulation a neuron belongs , the quantities @xmath4 and @xmath5 will be decorated with the index @xmath6 or @xmath7 .
more concretely , we will denote by @xmath8 ( resp .
@xmath9 ) an activated neuron of the excitatory type ( resp
. inhibitory ) , whereas @xmath10 ( resp .
@xmath11 ) identifies a quiescent excitatory ( resp .
inhibitory ) neuron .
the stochastic dynamics of the system is ruled by a minimal set of birth and death reactions , as summarized in the following : @xmath12}{\longrightarrow } x_a \nonumber \\
x_a&&\overset{\alpha_e}{\longrightarrow } x_q \nonumber \\
y_q&&\overset{f_i[s_i]}{\longrightarrow } y_a \nonumber \\
y_a&&\overset{\alpha_i}{\longrightarrow } y_q\end{aligned}\ ] ] where @xmath13 are constant rate function ; @xmath14 are non linear firing rate , sigmoid in shape , function of the currents @xmath15 , and bound to the interval @xmath16 $ ] . before completing the description of the elements that define the stochastic model , we remark that , by definition , the number of excitatory ( resp .
inhibitory ) neurons , @xmath17 ( resp .
@xmath18 ) , is an invariant quantity .
label with @xmath19 the number of neurons of the type @xmath8 , and with @xmath20 the number of elements @xmath10 . clearly , by virtue of equations ( [ chemical_eqs ] ) , @xmath21 .
similarly , and with an obvious meaning of the involved notation @xmath22 .
simply stated , the conservation laws that we have here identified enable one to reduce to a total of two the number of independent variables that uniquely lay out the dynamics of the system .
we then set : @xmath23 where the weights @xmath24 are positive defined .
the positive constants @xmath25 and @xmath26 encode the interaction of the examined families of neurons with the surrounding environment .
the specific form of the sigmoid function @xmath27 , with @xmath28 is not essential for the forthcoming discussion . to proceed in the analysis we introduce the number density @xmath29 and @xmath30 .
we further label with @xmath31 the probability of seeing the system in the state photographed by @xmath32 , at time @xmath33 .
the chemical equations introduced above define a markov process , that is ruled by the following master equation : @xmath34\end{aligned}\ ] ] where @xmath35 are the transition rates . before making explicit the definition of the transition rates , we write the master equation in a compact form involving the so called step operators @xmath36 and @xmath37 defined as : @xmath38 making use of the above operators , one can re - write the master equation in the form : @xmath39p(x , y , t)\end{aligned}\ ] ] where : @xmath40 the master equation ( [ me_eps ] ) provides an exact description of the markov dynamics .
in particular , it contains information on both the ideal mean field dynamics ( formally recovered in the limit of diverging system size ) and the associated finite size corrections . to elaborate on those aspects ,
we seek to approximate the exact master equation , via the kramers - moyal perturbative recipe @xcite . to this end
we taylor expand the step operators ( assuming @xmath41 and @xmath42 small ) and get : @xmath43 then , inserting in the master equation , one eventually obtains : @xmath44p(x , y , t)\end{aligned}\ ] ] and performing the obvious time rescaling @xmath45 : @xmath46p(x , y , t)\end{aligned}\ ] ] where @xmath47 measures the relative population of inhibitors over activators .
this is a two dimensional fokker - planck equation , which we shall write in a more transparent form by introducing the vector @xmath48 . in formulae : @xmath49 + \frac{1}{2 n}\sum_{i , j}\frac{\partial}{\partial x_i}\frac{\partial}{\partial x_j } \left [ b_{ij}({\bf x})p({\bf x},t ) \right]\ ] ] where : @xmath50\ ] ] and @xmath51\ ] ] the above fokker - planck equation is equivalent to the following non linear langevin equations @xcite : @xmath52 where the noise have zero mean and @xmath53 , with @xmath54 .
notice that the factor @xmath2 has been incorporated in the definition of matrix @xmath55 .
for this reason the noise term in the second of equations ( [ lang0 ] ) scales as @xmath56 .
we now make the transformation @xmath57 , where @xmath58 is such that @xmath59 and @xmath60 . in our case : @xmath61\ ] ] a straightforward calculation yields : @xmath62 we have hence obtained a system of non linear langevin equations that constitutes the fluctuating hydrodynamics approximation of the exact markov dynamics .
notice that the noise is multiplicative : the amplitude of the stochastic perturbation is self - consistently adjusted , as function of the density of the simultaneously evolving species . performing the thermodynamics limit @xmath63 ( at constant @xmath2 )
, the noise term fades away and one recovers a slightly modified version of the classical wc model . as compared to the original wc formulation , the first ( resp .
second ) equation of ( [ mult_lang ] ) displays an additional factor @xmath64 ( resp .
@xmath65 ) , which multiplies the sigmoid function @xmath66 ( resp .
@xmath67 ) .
this term reflects the finite carrying capacity of the system , as imposed at the microscopic level : working under diluted condition , i.e. assuming @xmath68 one readily converge to the wc model in its primitive conception @xcite .
as already mentioned the deterministic wc model presents a large plethora of interesting dynamical behaviors . depending on the assigned parameters ,
the system can evolve toward an isolated fixed point , present multiple fixed points with their associated basins of attraction or even exhibits an oscillatory periodic attractor . all along this paper
we will be primarily interested in a parameters setting that yields bistability .
as illustrated in figure [ fig1 ] , the system possesses two attractive fixed points ( green dots ) , which are separated by an unstable saddle point ( not displayed ) .
the ( red ) arrows stand for the stable and unstable manifolds associated to the saddle point .
the stable direction marks the transition between the adjacent basins of attractions of the two linearly stable fixed points .
deterministic wc trajectories are asymptotically attracted towards one of the stable equilibria , depending on their initial condition .
more interesting is instead the dynamics of the wc model in its stochastic representation : finite size , hence endogenous , noise drives seemingly erratic transitions between the two , deterministically stable , fixed points .
this picture is exemplified in figure [ fig2 ] , where the time evolution of species @xmath69 is plotted .
this is an individual trajectory obtained upon integration of the markov wc model via the usual gillespie monte carlo scheme . where @xmath70 is the gain parameter .
the quantities @xmath71 and @xmath72 enters the definition of the sigmoid non linear function and allow for a swift control of the location of the fixed points .
the offset @xmath71 sets in particular the degree of residual activity when @xmath73 .
the parameters here employed read @xmath74 , @xmath75 , @xmath76 , @xmath77 @xmath78 , @xmath79 , @xmath80 , @xmath81 , @xmath82 , @xmath83 , @xmath84 , @xmath85 , @xmath86 , @xmath87 , @xmath88 , @xmath88 . ]
is plotted versus time .
the time series is obtained as an individual realization of the stochastic wc model , integrated via the gillespie monte carlo scheme .
the parameters are set as in figure [ fig1 ] . ] to test the adequacy of the approximate langevin description ( [ mult_lang ] ) , we conducted a series of numerical simulations .
the distribution @xmath89 ( resp .
@xmath90 ) as obtained via the gillespie @xcite algorithm , hence drawn in accordance with the exact governing master equation , are displayed in figure [ fig3 ] , with ( cyan ) symbols .
the homologous quantities computed via a direct integration of the langevin equation is depicted with ( light purple ) diamonds .
the agreement is satisfying and points to the validity of the approximate langevin picture .
the multiplicative component of the noise is essential to reach this level of correspondence at moderate system sizes .
notice that @xmath89 shows the typical bimodal distribution , signature of a bistable dynamics .
conversely , at this noise level the two bumps of @xmath90 merge together in a unique isolated crest .
[ cols="^,^ " , ]
endogenous noise stemming from finite size fluctuations can play a role of paramount importance in shaping the self - emerging macroscopic dynamics of a multi - species model .
demographic perturbations resulting from the inherent discreteness of the inspected medium can be rationalized by resorting to a linear noise treatment , which proves perfectly adequate when the underlying deterministic dynamics admits a trivial fixed point . in general ,
non linearities are present and may eventually yield more complex dynamical behaviors , including multistability . in this case
the linear noise machinery can be punctually employed to resolve the local dynamics of the system around each of the mutually competing equilibria .
to quantify the statistics of the rare events that materialize in the transitions from one attractor of the deterministic dynamics to another , and so reconstruct the underlying stationary probability distributions , requires extending beyond the domain of application of the linear noise approximation .
working in this framework , we have here considered a stochastic version of the wilson cowan ( wc ) model , which accomodates for a finite population of excitatory and inhibitory neurons .
the model assumes a finite carrying capacity , the population census summing up to a constant .
the relative ratio of inhibitory over excitatory neurons acts as a control parameter of the model .
the master equation that rules the dynamics of the stochastic wc model is expanded in powers of the inverse population size : a fluctuating hydrodynamics picture is consequently obtained , which consists in a coupled pair of non linear langevin equations , with multiplicative noise .
this latter equations are solved numerically and shown to return a statistical description of the stochastic dynamics , which adheres quantitatively with that displayed by the exact markov model .
our analysis is specifically targeted to the region of bistability , a phenomenon that can a priori interest both the interacting families . to gain analytical insight into the investigated process
we however specialize on a simplified scenario , assigning the parameters so to have just one stationary stable equilibrium for the population of inhibitors .
the excitatory neurons can instead undergo bistable behavior . working in this setting ,
we operate a substantial reduction in complexity . in particular , the dynamics of the excitatory species is ruled by an independent , non linear langevin equation : the indirect imprint of the inhibitors is reflected in a modified noise term , while the deterministic contribution is evaluated at the fixed point of species @xmath91 .
the population of inhibitors can be ideally schematized as an additional noise source .
this latter contribution gets more pronounced the smaller the relative ratio of inhibitory versus excitatory neurons , broadening in turn the region where a bistable - like behavior is detected .
analytical expressions for the distributions of , respectively , excitatory and inhibitory neurons are obtained , which show a satisfying degree of agreement with the simulated data .
the methods are subsequently generalized to the relevant setting where both species display multiple fixed points .
an iterative scheme is in fact devised that builds on the zeroth order approximation described above , and allows us to explain the modifications displayed in the recorded probability distribution function , when a modest feedback of the excitatory on inhibitory neurons is accommodated for . in conclusion we have contributed to the understanding of the stochastic wc model , focusing on a dynamical regime for which the linear noise approximation is manifestly inadequate .
the non linear langevin equations which represent the fluctuating hydrodynamic limit of the original model , can be analytically characterized under suitable operating conditions , so contributing to shed light on the subtle dynamical interplay between distinct families of interacting neurons .
99 e. r. kandel , , j. h. schwartz , and t. m. jessell , `` principles of neural science '' fourth edition mcgraw - hill ( 2000 ) .
wilson and j.d .
cowan biophys .
j. , 12 1 - 24 ( 1972 ) .
wilson and j.d .
cowan kybernetik 13 55 - 80 ( 1973 ) .
z. p. kilpatrick `` wilson - cowan model '' , encyclopedia of computational neuroscience springer science business media , new york ( 2013 ) .
s. strogatz , `` non linear dynamics and chaos : with applications to physics , biology , chemistry and engineering '' ( perseus book group , cambridge , ma , 2001 ) . c. w. gardiner , `` handbook of stochastic methods '' , 3rd ed .
( springer - verlag , berlin , 2004 ) .
h. risken , `` the fokker - planck equation '' , 2nd ed .
( springer - verlag , berlin , 1989 ) . m. s. bartlett , j. r. stat .
a ( gen . ) 120 , 48 ( 1957 ) . a. j. mckane and t. j. newman , phys .
94 , 218102 ( 2005 ) .
t. dauxois , f. di patti , d. fanelli , and a. j. mckane , phys .
e 79 , 036112 ( 2009 ) . c. a. lugo and a. j. mckane , phys .
e 78 , 051911 ( 2008 ) .
t. butler and n. goldenfeld , phys .
e 80 , 030902(r ) ( 2009 ) .
de anna , f. di patti , d. fanelli , a. mckane , and t. dauxois , phys .
e 056110 ( 2010 ) .
t. biancalani , d. fanelli , and f. di patti , phys .
e 81 , 046215 ( 2010 ) .
t. e. woolley , r. e. baker , e. a. gaffney , and philip k. maini phys .
e 84 , 021915 ( 2011 ) .
bressloff phys .
rev e 82 051903 ( 2010 ) .
i. goychuk and a. goychuk new journal of physics 17 4 045029 ( 2015 ) .
e. negahbani , d.a .
steyn - ross , m. l. steyn - ross , m. t. wilson and j. w. sleigh j math neurosci .
5 1 9 ( 2015 ) .
j. d. cowan , j. neuman , w. van drongelen , j math neurosci .
t. c. butlera , m. benayoun , e. wallace , w. van drongelen , 6 1 1 ( 2016 ) . n. goldenfeld and j. cowan pnas 109 606 - 609 ( 2012 ) . n. g. van kampen , `` stochastic processes in physics and chemistry '' , 3rd ed .
( elsevier , amsterdam , 2007 ) .
mckane , t. biancalani , t. rogers bulletin of mathematical biology 76 ( 4 ) , 895 - 921 ( 2014 ) t biancalani , l dyson , aj mckane phys .
112 ( 3 ) , 038101(2015 ) .
f jafarpour , t biancalani , n goldenfeld phys . rev .
115 ( 15 ) , 158101 ( 2015 ) .
t biancalani , m assaf phys .
115 ( 20 ) , 208101 ( 2015 ) .
d. t. gillespie , j. phys .
81 , 2340 ( 1977 ) .
kloeden and e. platen `` numerical solution of stochastic differential equations '' .
springer , berlin ( 1992 ) . | we consider a stochastic version of the wilson - cowan model which accommodates for discrete populations of excitatory and inhibitory neurons .
the model assumes a finite carrying capacity with the two populations being constant in size .
the master equation that governs the dynamics of the stochastic model is analyzed by an expansion in powers of the inverse population size , yielding a coupled pair of non - linear langevin equations with multiplicative noise .
gillespie simulations show the validity of the obtained approximation , for the parameter region where the system exhibits dynamical bistability .
we report analytical progress by silencing the retroaction of the activators on the inhibitors , while still assigning the parameters so to fall in the region of deterministic bistability for the excitatory species .
the proposed approach forms the basis of a perturbative generalization which applies to the case where a modest degree of coupling is restored . |
thanks to its flexibility to simulate the different characteristics of matter from perfect fluid to dark energy and its ability to describe the different stages of the evolution of the universe , spinor field has become quite popular among the cosmologists @xcite .
but some recent study @xcite suggests that flexible though it is , the existence of non - diagonal components of the energy - momentum tensor of the spinor field imposes very severe restrictions on the geometry of the universe as well as on the spinor field , thus justifying our previous claim that spinor field is very sensitive to the gravitational one @xcite . in some recent papers @xcite within the scope of bianchi type - i cosmological model
the role of spinor field in the evolution of the universe has been studied .
it is found that due to the spinor affine connections the energy momentum tensor of the spinor field becomes non - diagonal , whereas the einstein tensor is diagonal .
this non - triviality of non - diagonal components of the energy - momentum tensor imposes some severe restrictions either on the spinor field or on the metric functions or on both of them . in case
if the restrictions are imposed on the components of spinor field only , it becomes massless and invariants constructed from bilinear spinor forms also become trivial . imposing restriction wholly on metric functions one obtains frw model , while if the restrictions are imposed both on metric functions and spinor field components , the initially bi model becomes locally rotationally symmetric .
these results motivated us to consider the other bianchi models and study the influence of spacetime geometry on the spinor field and vice versa .
a bianchi type-@xmath0 model describes an anisotropic spacetime and generates particular interest among physicists .
weaver @xcite , ib@xmath1ez et al .
@xcite , socorro and medina @xcite , and bali et al .
@xcite have studied b-@xmath2 spacetime in connection with massive strings .
recently , belinchon @xcite studied several cosmological models with b-@xmath2 & b - iii symmetries under the self similar approach
. a spinor description of dark energy within the scope of a b-@xmath0 model was given in @xcite .
bianchi type @xmath2 spacetime filled with dark energy was investigated in @xcite . in this paper
we study the self - consistent system of nonlinear spinor field and gravitational one given by the bianchi type @xmath2 spacetime in order to clarify the role of non - diagonal components of the energy - momentum tensor of spinor field in the evolution of the universe .
let us consider the case when the anisotropic space - time is filled with nonlinear spinor field .
the corresponding action can be given by @xmath3 with @xmath4 here @xmath5 corresponds to the gravitational field @xmath6 where @xmath7 is the scalar curvature , @xmath8 , with g being newton s gravitational constant and @xmath9 is the spinor field lagrangian .
the gravitational field in our case is given by a bianchi type-@xmath0 anisotropic space time : @xmath10 with @xmath11 and @xmath12 being the functions of time only and @xmath13 is some arbitrary constant .
the nontrivial christoffel symbols for are @xmath14 the nonzero components of the einstein tensor corresponding to the metric are [ et ] @xmath15 for a spinor field @xmath16 , the symmetry between @xmath16 and @xmath17 appears to demand that one should choose the symmetrized lagrangian @xcite . keeping this in mind we choose the spinor field lagrangian as @xcite : @xmath18 - m_{\rm sp } { \bar \psi}\psi - f , \label{lspin}\ ] ] where the nonlinear term @xmath19 describes the self - interaction of a spinor field and
can be presented as some arbitrary functions of invariants generated from the real bilinear forms of a spinor field .
since @xmath20 and @xmath21 ( complex conjugate of @xmath20 ) have four component each , one can construct @xmath22 independent bilinear combinations .
they are [ bf ] @xmath23 where @xmath24 $ ] .
invariants , corresponding to the bilinear forms , are [ invariants ] @xmath25 according to the fierz identity , among the five invariants only @xmath26 and @xmath27 are independent as all others can be expressed by them : @xmath28 and @xmath29 therefore , we choose the nonlinear term @xmath19 to be the function of @xmath26 and @xmath27 only , i.e. , @xmath30 , thus claiming that it describes the nonlinearity in its most general form .
indeed , without losing generality we can choose @xmath31 , with @xmath32 taking any of the following expressions @xmath33 . here
@xmath34 is the covariant derivative of spinor field : @xmath35 with @xmath36 being the spinor affine connection . in @xmath37 s
are the dirac matrices in curve space - time and obey the following algebra @xmath38 and are connected with the flat space - time dirac matrices @xmath39 in the following way @xmath40 where @xmath41 is a set of tetrad 4-vectors . for the metric
we choose the tetrad as follows : @xmath42 the dirac matrices @xmath43 of bianchi type-@xmath0 spacetime are connected with those of minkowski one as follows : @xmath44 @xmath45 with @xmath46 where @xmath47 are the pauli matrices : @xmath48 note that the @xmath39 and the @xmath49 matrices obey the following properties : @xmath50 where @xmath51 is the diagonal matrix , @xmath52 is the kronekar symbol and @xmath53 is the totally antisymmetric matrix with @xmath54 .
the spinor affine connection matrices @xmath55 are uniquely determined up to an additive multiple of the unit matrix by the equation @xmath56 with the solution @xmath57 from the bianchi type - vi metric one finds the following expressions for spinor affine connections : [ sac123 ] @xmath58 variation of with respect to the metric function @xmath59 gives the einstein field equation @xmath60 where @xmath61 and @xmath7 are the ricci tensor and ricci scalar , respectively . here
@xmath62 is the energy momentum tensor of the spinor field .
varying with respect to @xmath63 one finds the spinor field equations : [ speq ] @xmath64 where we denote @xmath65 and @xmath66 , with @xmath67 , @xmath68 and @xmath69 in view of , eq .
can be rewritten as @xmath70 - m_{\rm sp } { \bar \psi}\psi - f(i,\,j ) \nonumber \\ & = & \frac{\imath}{2 } { \bar \psi}[\gamma^{\mu } \nabla_{\mu } \psi - m_{\rm sp } \psi ] - \frac{\imath}{2}[\nabla_{\mu } \bar \psi \gamma^{\mu } + m_{\rm sp } { \bar \psi } ] \psi - f(i,\,j),\nonumber \\ & = & 2 ( i f_i + j
f_j ) - f = 2 k f_k - f(k ) .
\label{lspin01}\end{aligned}\ ] ] in what follows we consider the case when the spinor field depends on @xmath71 only , i.e. @xmath72 .
the energy - momentum tensor of the spinor field is given by @xmath73 then in view of and the energy - momentum tensor of the spinor field can be written as @xmath74 as is seen from , in the case where , for a given metric @xmath36 s are different , there arise nontrivial non - diagonal components of the energy momentum tensor .
after a little manipulations from one finds the following components of the energy momentum tensor : [ ttot ] @xmath75 \nonumber \\ & = & \frac{1}{4 } \frac{a_2}{a_1 } e^{2 m x_3 } \biggl[\biggl(\frac{\dot a_1}{a_1 } - \frac{\dot a_2}{a_2}\biggr ) a^3 - \frac{2m}{a_3}a^0\biggr ] , \label{emt12}\\ t_3 ^ 1 & = & \frac{\imath}{4 } \frac{a_3}{a_1 } e^{m x_3 } \biggl(\frac{\dot a_3}{a_3 } - \frac{\dot a_1}{a_1}\biggr ) { \bar \psi}{\bar \gamma}^3 { \bar \gamma}^1 { \bar \gamma}^0 \psi = \frac{1}{4 } \frac{a_3}{a_1 } e^{m x_3 } \biggl(\frac{\dot a_3}{a_3 } - \frac{\dot a_1}{a_1}\biggr ) a^2 \label{emt13}\\ t_3 ^ 2 & = & \frac{\imath}{4 } \frac{a_3}{a_2 } e^{-m x_3 } \biggl(\frac{\dot a_2}{a_2 } - \frac{\dot a_3}{a_3}\biggr ) { \bar \psi}{\bar \gamma}^2 { \bar \gamma}^3 { \bar \gamma}^0 \psi = \frac{1}{4 } \frac{a_3}{a_2 } e^{-m x_3 } \biggl(\frac{\dot a_2}{a_2 } - \frac{\dot a_3}{a_3}\biggr)a^1 . \label{emt23}\end{aligned}\ ] ] as one sees from the spinor field possesses non - trivial of non - diagonal components of the energy momentum tensor .
in this section we solve the self consistent system of spinor and gravitational field equations .
we begin with the spinor field equations and then solve the gravitational field equations .
finally we study the influence of the non - diagonal components of the energy momentum tensor on the components of the spinor field and metric functions .
let us begin with the spinor field equations .
in view of and the spinor field equations take the form [ sf1 ] @xmath76 where we define the volume scale @xmath77 as we have already mentioned , @xmath20 is a function of @xmath71 only .
we consider the 4-component spinor field given by @xmath78 denoting @xmath79 from for the spinor field we find we find [ speq1pfg ] @xmath80 the foregoing system of equations can be written in the form : @xmath81 with @xmath82 and @xmath83 it can be easily found that @xmath84 the solution to the equation can be written in the form @xmath85 where @xmath86 and @xmath87 is the solution at @xmath88 , with @xmath89 being quite large , so that the volume scale @xmath90 , hence the expanding universe becomes large enough . as it will be shown later , @xmath91 for @xmath32 taking one of the following expressions @xmath92 with trivial spinor - mass and @xmath93 for @xmath94 for any spinor - mass .
since our universe is expanding , the quantities @xmath95 and @xmath96 become trivial at large @xmath88 . hence in case of @xmath94 with non - trivial spinor - mass one can assume @xmath97 , whereas for other cases with trivial spinor - mass we have @xmath98 with @xmath99 being some constants . here
we have used the fact that @xmath100 the other way to solve the system is given in @xcite .
it can be shown that bilinear spinor forms the obey the following system of equations : [ inv ] @xmath101 where we denote @xmath102 , @xmath103 , @xmath104 , @xmath105 and @xmath106 .
combining these equations together and taking the first integral one gets [ inv0 ] @xmath107 now let us consider the gravitational field equations .
in view of and with find the following system of einstein equations [ eebvi0 ] @xmath108 with the additional constraints [ ac ] @xmath109 = 0 , \label{ac12}\\ t_2 ^ 1 & = & \frac{1}{4 } \frac{a_3}{a_1 } e^{m x_3 } \biggl(\frac{\dot a_3}{a_3 } - \frac{\dot a_1}{a_1}\biggr ) a^2 = 0 , \label{ac13}\\ t_3 ^ 2 & = & \frac{1}{4 } \frac{a_3}{a_2 } e^{-m x_3 } \biggl(\frac{\dot a_2}{a_2 } - \frac{\dot a_3}{a_3}\biggr)a^1 = 0 . \label{ac23}\end{aligned}\ ] ] from one dully finds @xmath110 let us now find expansion and shear for bianchi type-@xmath0 metric .
the expansion is given by @xmath111 and the shear is given by @xmath112 with @xmath113 - \frac{1}{3 } \vartheta p_{\mu\nu } , \label{shearcomp}\ ] ] where the projection vector @xmath114 : @xmath115 in comoving system we have @xmath116 .
in this case one finds @xmath117 and [ shearcomps ] @xmath118 one then finds @xmath119 = \frac{1}{2}\biggl[\sum_{i=1}^3 h_i^2 - \frac{1}{3}\vartheta^2\biggr ] .
\label{sheargen}\ ] ] inserting into and we find @xmath120 and [ shearcomps0 ] @xmath121 as it was found in previous papers , due to explicit presence of @xmath12 in the einstein equations , one needs some additional conditions . in an early work we propose two different situation , namely , set @xmath122 and @xmath123 which allowed us to obtain exact solutions for the metric functions . in a recent paper we imposed the proportionality condition , widely used in literature .
demanding that the expansion is proportion to a component of the shear tensor , namely @xmath124 the motivation behind assuming this condition is explained with reference to thorne @xcite .
the observations of the velocity - red - shift relation for extragalactic sources suggest that hubble expansion of the universe is isotropic today within @xmath125 per cent @xcite . to put more precisely ,
red - shift studies place the limit @xmath126 on the ratio of shear @xmath49 to hubble constant @xmath127 in the neighborhood of our galaxy today .
collins et al .
@xcite have pointed out that for spatially homogeneous metric , the normal congruence to the homogeneous hypersurfaces satisfies the condition : @xmath128 under this proportionality condition it was also found that the energy - momentum distribution of the model is strictly isotropic , which is absolutely true for our case .
further on account of we finally find @xmath129^{\frac{1}{3}- \frac{1}{2n_3 } } , \quad a_2 = x_0 \bigl[\frac{1}{x_0 x_1 } v\bigr]^{\frac{1}{3}- \frac{1}{2n_3}},\quad a_3 = x_1 \bigl[\frac{1}{x_0 x_1 } v\bigr]^{\frac{1}{3}+ \frac{1}{n_3}}. \label{metf}\end{aligned}\ ] ] as it is obvious from the isotropization of the spacetime can take place only for large value of @xmath130 .
the equation for @xmath90 can be found from the einstein equation which after some manipulation looks @xmath131 v , \label{vdefein}\ ] ] with @xmath132 in order to solve we have to know the relation between @xmath32 and @xmath90 . recalling that @xmath32 takes one of the following expressions @xmath33 , with @xmath65 and @xmath133
let us first find those relations for different @xmath32 . in case of @xmath94 , i.e. @xmath134 from
we find @xmath135 with the solution @xmath136 in this case spinor field can be either massive or massless . in the cases where @xmath32 takes any of the following expressions @xmath92 that gives @xmath137 , we consider a massless spinor field . in case of @xmath138 , @xmath139 .
then from we have @xmath140 with the solution @xmath141 in case of @xmath142 the equations and can be rewritten as @xmath143 which can be rearranged as @xmath144 with the solution @xmath145 it should be noticed that in this case one can use the following parametrization for @xmath146 and @xmath114 : @xmath147 here we like to note that for the case in concern one can consider the massive spinor field as well . in that case
we have @xmath148 which can be rearranged as @xmath149 from follows @xmath150 further setting @xmath151 and @xmath152 eq .
can be written as @xmath153 with the solution @xmath154 as one sees for massless spinor field from follows @xmath155 , which is equivalent to for @xmath156 . moreover , given the fact that @xmath157 $ ] for the massive spinor field @xmath32 comes out to be a time varying quantity that has the range @xmath158.$ ] in our purpose we consider here only the massless spinor field .
finally , for @xmath159 the equations and can be rewritten as @xmath160 which can be rearranged as @xmath161 with the solution @xmath162 in this case one can use the following parametrization for @xmath146 and @xmath114 : @xmath163 thus we see that @xmath32 is a function of @xmath90 .
for the cases considered here we established that @xmath93 .
so one can easily consider the case when @xmath164 . in that case
it is possible to study both massive and massless spinor field to clarify the role of spinor mass . further inserting @xmath165 into one
finds the expression for @xmath90 .
one can further study the behavior of @xmath90 numerically for different @xmath19 .
but before that let us first see what happens to the result obtained if the additional conditions are taken into account .
thus , until this point we have only used the einstein system of equations without the additional conditions . in what follows we turn to them and see how these conditions effect our solutions . from and
one dully finds @xmath166 in view of the relations and fulfill even without imposing restrictions on the metric functions . on account of from one
finds @xmath167 the equalities and can be rewritten in terms of spinor field components as follows : @xmath168 on the other hand , in view of and from the equality @xmath169 we have either @xmath170 or @xmath171 in case of @xmath172 we find @xmath173 . then taking into account that @xmath174 we ultimately find @xcite @xmath175
thus we see that in the case considered here the initially massive , nonlinear spinor field becomes linear and massless as a result of special geometry of the bianchi type-@xmath0 spacetime , which is equivalent to solving the corresponding einstein equation in vacuum . in this case for volume scale @xmath90
we find @xmath176 with the solution in quadrature @xmath177 with @xmath178 and @xmath179 are being some arbitrary constants . in fig .
[ vbv0lin ] we have plotted the evolution of volume scale . for simplicity we have set @xmath180 , @xmath181 , @xmath182 , @xmath183 and @xmath184 .
the initial value of volume scale is taken to be @xmath185 and @xmath186 is calculated using .
as far as spinor field is concerned the matrix @xmath187 in in this case becomes trivial and the components of the spinor field can be written as @xmath188 with @xmath189 s being the constant of integration obeying @xmath190 the second possibility is to consider with @xmath191 . in this case the nonlinear term as well as the massive term do not vanish . in
what follows , we will consider the case for @xmath94 , setting @xmath192 as far as other cases are concerned we can revive them setting spinor mass @xmath193 . then inserting into and taking into account that in this case @xmath194 we find @xmath195 , \label{detvbvi0new}\end{aligned}\ ] ] with the solution in quadrature @xmath196 + c_1 } , \label{vbv0nonlinsol}\ ] ] with @xmath178 and @xmath197 are being some arbitrary constants .
it can be shown that the metric functions and the components of the spinor field as well as the invariants constructed form them are inverse function of @xmath90 of some degree , hence at any spacetime point where @xmath198 it is a singularity .
so we assume that at the beginning @xmath90 was small but non - zero . from we
see that at initial stage the nonlinear term prevails if @xmath199 such that @xmath200 and @xmath201 , whereas for the nonlinearity to become dominant for large value of @xmath90 one should have @xmath202 such that @xmath203 and @xmath204 . in figs .
[ vbv0pos ] and [ vbv0neg ] we have plotted the evolution of volume scale for positive and negative coupling constants @xmath205 and @xmath206 , respectively . for simplicity
we set @xmath180 , @xmath181 , @xmath182 , @xmath184 , @xmath207 , @xmath208 , @xmath209 , @xmath210 , @xmath211 , @xmath212 and @xmath213 . in case of positive coupling constants @xmath214 and @xmath215 the model describes an expanding universe , while for negative coupling constants @xmath216 and @xmath217 we have a cyclic universe that expands to some maximum and then contracts to minimum only to expand again . the initial value of volume scale is taken to be @xmath185 and @xmath186 is calculated using . and @xmath215.,title="fig:",height=264 ]
+ . and @xmath218.,title="fig:",height=264 ] + .
let us also see what happens to deceleration parameter for positive coupling constants . using the definition @xmath219 it can be shown that in this case our universe is expanding with acceleration .
taking into account the discussion about the value of @xmath220 we can rewrite @xmath221}{\frac{6n_3 \bar x}{4 n_3 - 6 } v^{4/3 - 2/n_3 } + 3\kappa \left[m_{\rm sp}\,v_0 v + \lambda_1 v_0^{2n_1 } v^{2(1 - n_1 ) } + \lambda_2 v_0^{2n_2 } v^{2(1 - n_2)}\right ] + c_1}. \label{decparext}\ ] ] as it was mentioned earlier , at large @xmath71 , hence for large @xmath90 prevails the term with @xmath222 . taking this into account we find @xmath223 and @xmath215.,title="fig:",height=264 ] + . in fig .
[ vbv0dec ] we have illustrated the evolution of the the deceleration parameter . as we see that the spinor field nonlinearity leads to the late time accelerated expansion of the universe .
within the scope of bianchi type-@xmath0 spacetime we study the role of spinor field on the evolution of the universe . in this case
we consider the spinor field that depends only on time @xmath71 . even in this case
the spinor field possesses non - zero non - diagonal components of energy - momentum tensor thanks to its specific relation with gravitational field .
this fact plays vital role on the evolution of the universe . due to the specific behavior of the spinor field we have two different scenarios . in one case the bilinear forms constructed from it becomes trivial , thus giving rise to a massless and linear spinor field lagrangian .
this case is equivalent to the vacuum solution of the bianchi type-@xmath0 spacetime .
the second case allows non - vanishing massive and nonlinear terms and depending on the sign of coupling constants gives rise to expanding mode of expansion or the one that after obtaining some maximum value contracts and ends in big crunch generating spacetime singularity .
this result once again shows the sensitivity of spinor field to the gravitational one . *
acknowledgments * + this work is supported in part by a joint romanian - lit , jinr , dubna research project , theme no .
05 - 6 - 1119 - 2014/2016 .
i would also like to thank the referee for some valuable suggestions that helped me to improve the ms . | within the scope of bianchi type-@xmath0 spacetime we study the role of spinor field on the evolution of the universe .
it is found that the presence of nontrivial non - diagonal components of energy - momentum tensor of the spinor field plays vital role on the evolution of the universe . as a result of their mutual influence there
occur two different scenarios . in one case
the invariants constructed from the bilinear forms of the spinor field become trivial , thus giving rise to a massless and linear spinor field lagrangian . according to the second scenario massive and nonlinear terms
do not vanish and depending on the sign of coupling constants we have either an expanding mode of expansion or the one that after obtaining some maximum value contracts and ends in big crunch generating spacetime singularity .
this result shows that the spinor field is highly sensitive to the gravitational one . -24pt |
in ref . @xcite an @xmath7 conformally symmetric model was proposed for strong interactions at low energies , based on the observation , published in 1919 by h. weyl in ref .
@xcite , that the dynamical equations of gauge theories retain their flat - space - time form when subject to a conformally - flat metrical field , instead of the usual minkowski background .
confinement of quarks and gluons is then described through the introduction of two scalar fields which spontaneously break the @xmath7 symmetry down to @xmath8 and @xmath9 symmetry , respectively . moreover ,
a symmetric second - order tensor field is defined that serves as the metric for flat space - time , coupling to electromagnetism .
quarks and gluons , which to lowest order do not couple to this tensor field , are confined to an anti - de - sitter ( ads ) universe @xcite , having a finite radius in the flat space - time .
this way , the model describes quarks and gluons that oscillate with a universal frequency , independent of the flavor mass , inside a closed universe , as well as photons which freely travel through flat space - time . the fields in the model of ref .
@xcite comprise one real scalar field @xmath10 and one complex scalar field @xmath11 .
their dynamical equations were solved in ref .
@xcite for the case that the respective vacuum expectation values , given by @xmath12 and @xmath13 , satisfy the relation @xmath14 a solution for @xmath12 of particular interest leads to ads confinement , via the associated conformally flat metric given by @xmath15 . the only quadratic term in the lagrangian of ref . @xcite is proportional to @xmath16 hence , under the condition of relation ( [ slvacua ] ) , one obtains , after choosing vacuum expectation values , a light @xmath10 field , associated with confinement , and a very heavy complex @xmath11 field , associated with electromagnetism .
weak interactions were not contemplated in ref .
@xcite , but one may read electroweak for electromagnetism . here , we will study the supposedly light mass of the scalar field that gives rise to confinement . the conformally symmetric model of ref .
@xcite in itself does not easily allow for interactions between hadrons , as each hadron is described by a closed universe .
hence , in order to compare the properties of this model with the actually measured cross sections and branching ratios , the model has been further simplified , such that only its main property survives , namely its flavor - independent oscillations . this way the full ads spectrum is , via light - quark - pair creation , coupled to the channels of two or more hadronic decay products for which scattering amplitudes can be measured .
the ads spectrum reveals itself through the structures observed in hadronic mass distributions .
however , as we have shown in the past ( see ref .
@xcite and references therein ) , there exists no simple relation between enhancements in the experimental cross sections and the ads spectrum .
it had been studied in parallel , for mesons , in a coupled - channel model in which quarks are confined by a flavor - independent harmonic oscillator @xcite .
empirically , based on numerous data on mesonic resonances measured by a large variety of experimental collaborations , it was found @xcite that an ads oscillation frequency of @xmath17 agrees well with the observed results for meson - meson scattering and meson - pair production in the light @xcite , heavy - light @xcite , and heavy @xcite flavor sectors , thus reinforcing the strategy proposed in ref .
@xcite .
another ingredient of the model for the description of non - exotic quarkonia , namely the coupling of quark - antiquark components to real and virtual two - meson decay channels @xcite via @xmath18 quark - pair creation , gives us a clue about the size of the mass of the @xmath10 field .
for such a coupling it was found that the average radius @xmath19 for light - quark - pair creation in quarkonia could be described by an flavor - independent mass scale , given by @xmath20 where @xmath21 is the effective reduced quarkonium mass . in earlier work ,
the value @xmath22 @xcite was used , which results in @xmath23 mev for the corresponding mass scale .
however , the quarkonium spectrum is not very sensitive to the precise value of the radius @xmath19 , in contrast with the resonance widths . in more recent work @xcite ,
slightly larger transition radii have been applied , corresponding to values around 40 mev for @xmath24 .
nevertheless , values of 3040 mev for the flavor - independent mass @xmath24 do not seem to bear any relation to an observed quantity for strong interactions .
however , we will next present experimental evidence for the possible existence of a quantum with a mass of about 38 mev , which in the light of its relation to the @xmath18 mechanism we suppose to mediate quark - pair creation .
moreover , its scalar properties make it a perfect candidate for the quantum associated with the above - discussed scalar field for confinement .
@xcite , we made notice of an apparent interference effect around the @xmath25 threshold in the invariant - mass distribution of @xmath26 events , which we observed in preliminary radiation data of the babar collaboration @xcite .
the effect , with a periodicity of about 74 mev , could be due to interference between the typical oscillation frequency of 190 mev of the @xmath27 pair and that of the gluon cloud . [ cols="^ " , ]
thus , a signal with the shape of a narrow breit - wigner resonance seems to be visible on the slope of the @xmath28 resonance , though with little more than 2@xmath10 relevance . nevertheless , by coincidence or not , it comes out exactly in the expected place , namely at @xmath29 mev .
unfortunately , the data @xcite do not have enough statistics to pinpoint possible higher excitations as well .
so we can not relate , to a minimum degree of accuracy , the other observed deficit enhancements to the possible existence of hybrid states .
from the fact that the @xmath30 has not been observed before , one must conclude that it probably does not interact at least to leading order through electroweak forces , but instead couples exclusively to quarks and gluons .
the interference effect we discussed in sec .
[ oscillations ] might well be explained by @xmath30 exchange between the quarks , which interferes with the natural quark oscillations . moreover , since the interference effect is smaller for light quarks than for heavier ones , it is likely that their coupling to the @xmath30 is proportional to flavor mass , as one expects from theory .
the @xmath30 could very well be just a light scalar glueball , albeit much lighter than found in refs .
its low mass precludes decay into hadrons , while the absence of electroweak couplings does not allow it to decay into leptons either , at least to lowest order .
it may decay , though , into photons via virtual quark loops , and through photons , eventually , into @xmath0 pairs .
however , the probability for such reactions to occur is extremely remote , since the coupling between the quarks and the light scalar is proportional to the quark mass @xcite . in the case of bottom quarks , we found events of the order of or less than one percent of the total . for light quarks ,
their mass ratio with respect to bottom quarks reduces this rate to @xmath31@xmath32 .
hadrons will certainly interact with a light scalar ball of glue .
for example , a proton struck by such a scalar particle may absorp it and then emit photons , or decay into a neutron and a lepton - neutrino pair . yet
another possibility is that , being closed universes themselves , these scalar particles mainly collide elastically with hadronic matter . in that case , depending on their linear momentum , they may remove light nuclei from atoms .
nevertheless , we do not see such processes happening around us
. therefore , these light scalars are probably not abundantly present near us . however , in the early universe they may have existed , most probably inflated to hadrons under collisions . on the other hand , interactions with hadrons might have been observed in bubble - chamber experiments , where isolated protons could be the result of collisions with a light scalar particle emerging from one of the interaction vertices .
light higgs fields have been considered in supersymmetric extensions of the standard model @xcite .
furthermore , axions , which appear in models motivated by astrophysical observations , are assumed to have higgs - like couplings @xcite .
model predictions for the branching fraction of @xmath33higgs decays , for higgs masses below @xmath34 @xcite , range from @xmath32 @xcite to @xmath35 @xcite .
furthermore , the three anomalous events observed in the hypercp experiment @xcite were interpreted as the production of a scalar boson with a mass of 214.3 mev , decaying into a pair of muons @xcite . however , in ref .
@xcite the babar collaboration found no evidence for dimuon decays of a light scalar particle in radiative decays of @xmath28 and @xmath36 mesons .
the babar limits for dimuon decays of a light scalar particle rule out much of the parameter space allowed by the light - higgs @xcite and axion @xcite models . nonetheless , in ref .
@xcite y .- j .
zhang and h .- s .
shao , pointed out that the transitions @xmath37higgs are not yet excluded by the lepton - universality test in @xmath38 decays studied by babar in refs .
@xcite .
the light scalar glueball we have discussed here seems to correspond to the lowest - order empty - universe solution of ref .
@xcite for strong interactions .
it has similar properties as the electroweak higgs , but now for strong interactions .
quarks couple to it with an intensity which is proportional to their mass , in the same way that mass couples to gravity @xcite . in the standard model @xcite , the higgs boson and the graviton are the only particles yet to be observed , and no higgs particle for strong interactions is anticipated .
however , n. trnqvist recently proposed @xcite the light scalar - meson nonet @xcite as the higgs bosons of strong interactions , while in ref .
@xcite he obtained a nonzero pion mass by means of a small breaking of a relative symmetry between the electroweak and the strong interactions .
a relation between the lightest scalar - meson nonet and glueballs has often been advocated by p. minkowski and w. ochs ( see e.g. ref .
@xcite ) . in our view
though , the light scalar mesons are dynamically generated through @xmath39 pair creation / annihilation , which mixes the quark - antiquark and dimeson sectors @xcite .
in sec . [ intro ]
we discussed why we expect an additional scalar particle to exist , besides the higgs boson for the electroweak sector .
furthermore , we have estimated its mass based on the average radius for @xmath18 quark - pair creation , which had been extracted over the past three decades from numerous data on mesonic resonances ( see ref .
@xcite and references therein ) . in sec .
[ oscillations ] we recalled our results on an apparent interference effect in annihilation data , and stressed the possibility that it may stem from some internal oscillation with a frequency of about 38 mev . in sec .
[ phenomenon ] we showed that missing data in the reactions @xmath0
@xmath1 @xmath40 @xmath1 @xmath3 and @xmath0 @xmath1 @xmath40 @xmath1 @xmath4 exhibit maxima at @xmath24 , @xmath41 , and @xmath42 , for @xmath43 mev . each of the results , viz.the interference effect observed in ref .
@xcite , the small flavor - independent oscillations in electron - positron and proton - antiproton annihilation data , observed in ref . @xcite and summarized in fig .
[ interference ] , the excess signals visible in the @xmath5 mass distributions of @xmath28 @xmath1 @xmath40 @xmath1 @xmath4 ( fig . [ mumuall2s ] ) , in @xmath36 @xmath1 @xmath40 @xmath1 @xmath4 ( fig . [ alldiff]a ) , in @xmath36 @xmath1 @xmath44 @xmath1 @xmath4 ( fig . [ alldiff]b ) , and in the @xmath0 mass distributions of @xmath0 @xmath1 @xmath3 ( fig .
[ alldiff]c ) , and finally the resonance signal shown in fig . [ hybrid ] , is much too small to make firm claims .
however , we observe here that all points in the same direction . indeed
, the probability must be close to zero that one accidentally finds the same oscillations in four different sets of data ( refs .
@xcite ) involving different flavors , statistical fluctuations at @xmath45 mev in yet another four sets of different data ( figs . [ mumuall2s ] and [ alldiff ] ) , and moreover a resonance - like fluctuation at @xmath46 mev in a further set of data ( fig . [ hybrid ] ) .
furthermore , the related mass comes where predicted by our analyses in mesonic spectroscopy ( see ref .
@xcite and references therein ) . in sec .
[ gluons ] we discussed that , most probably , the missing signal is due to the emission of an as yet unobserved light scalar particle , while part of the excess data corroborates such an interpretation .
since the corresponding particle has all the right properties , we conclude that we found first indications , of the possible existence of a higgs - like particle , namely the scalar boson related to confinement .
furthermore , the data also suggest the existence of two replicas of the @xmath30 with masses that are two and three times heavier than the @xmath30 .
in addition , we believe that this 38 mev boson , which we designate by @xmath30 , consists of a mini - universe filled with glue , thus forming a very light scalar glueball . furthermore , we have pinpointed the masses of possible @xmath6 hybrids , one of which shows up as an enhancement in the invariant - mass distribution of babar data , albeit with a 2@xmath10 significance at most .
finally , we urge the babar collaboration to inspect their larger data set in order to settle , with higher statistics , the possible existence of the @xmath30 and the related @xmath6 hybrid spectrum .
we are grateful for the precise measurements and data analyses of the babar , cdf , and cmd-2 collaborations , which made the present analysis possible .
one of us ( evb ) wishes to thank drs .
b. hiller , a. h. blin , and a. a. osipov for useful discussions .
this work was supported in part by the _ fundao para a cincia e a tecnologia _ of the _ ministrio da cincia , tecnologia e ensino superior _ of portugal , under contract cern / fp/ 109307/2009 .
m. dohse , arxiv:0706.1887 .
e. van beveren and g. rupp , pos hql2010:003 ( 2010 ) [ arxiv:1011.2360 ] .
e. van beveren , c. dullemond , and g. rupp , , 772 ( 1980 ) [ erratum - ibid .
d * 22 * , 787 ( 1980 ) ] .
e. van beveren , g. rupp , t. a. rijken , and c. dullemond , , 1527 ( 1983 ) .
e. van beveren and g. rupp , arxiv:1009.1778 .
e. van beveren , t. a. rijken , k. metzger , c. dullemond , g. rupp and j. e. ribeiro , , 615 ( 1986 ) [ arxiv:0710.4067 ] .
e. van beveren and g. rupp , , 012003 ( 2003 ) [ arxiv : hep - ph/0305035 ] .
e. van beveren and g. rupp , arxiv:1009.4097 .
e. van beveren , and g. rupp , , 1620 ( 2009 ) [ arxiv:0809.1149 ] .
e. van beveren and g. rupp , arxiv : hep - ph/0201006 .
e. van beveren , d. v. bugg , f. kleefeld and g. rupp , , 265 ( 2006 ) [ arxiv : hep - ph/0606022 ] .
e. van beveren and g. rupp , , 111501r ( 2009 ) [ arxiv:0905.1595 ] .
b. aubert [ babar collaboration ] , slac - pub-13360 , babar - conf-08 - 004 , arxiv:0808.1543 .
v. m. aulchenko _ et al . _
[ cmd-2 collaboration ] , , 743 ( 2005 ) [ pisma zh .
fiz . * 82 * , 841 ( 2005 ) ] [ arxiv : hep - ex/0603021 ] .
t. aaltonen _
[ cdf collaboration ] , , 152001 ( 2009 ) [ arxiv:0906.5218 ]
. b. aubert _ et al . _ [ babar collaboration ] , , 112002 ( 2008 ) [ arxiv:0807.2014 ] .
e. van beveren and g. rupp , arxiv:1009.5191 .
del amo sanchez _ et al .
_ [ the babar collaboration ] , , 191801 ( 2010 ) [ arxiv:1002.4358 ] .
e. guido [ babar collaboration ] , arxiv:0910.0423 .
b. aubert _ et al . _ [ babar collaboration ] , , 081803 ( 2009 ) [ arxiv:0905.4539 [ hep - ex ] ] .
h. p. nilles , m. srednicki and d. wyler , , 346 ( 1983 ) .
j. m. frere , d. r. t. jones and s. raby , , 11 ( 1983 ) .
j. p. derendinger and c. a. savoy , , 307 ( 1983 ) .
j. r. ellis , j. f. gunion , h. e. haber , l. roszkowski and f. zwirner , , 844 ( 1989 ) .
x. g. he , j. tandean and g. valencia , , 081802 ( 2007 ) [ arxiv : hep - ph/0610362 ] .
j. p. lees _ et al .
_ [ the babar collaboration ] , , 151802 ( 2010 ) [ arxiv:1001.1883 [ hep - ex ] ] .
j. a. bicudo , j. e. f. ribeiro and a. v. nefediev , , 085026 ( 2002 ) [ arxiv : hep - ph/0201173 ] .
p. hasenfratz and j. kuti , 75 ( 1978 ) .
s. ono , , 307 ( 1984 ) .
e. klempt and a. zaitsev , , 1 ( 2007 ) [ arxiv:0708.4016 ] .
f. iddir and l. semlala , arxiv : hep - ph/0211289 . m. r. pennington , arxiv : hep - ph/0309228 .
p. minkowski and w. ochs , arxiv : hep - ph/0209225 . m. boglione and m. r. pennington , , 1998 ( 1997 ) [ arxiv : hep - ph/9703257 ] .
p. minkowski and w. ochs , arxiv : hep - ph/9905250 . c. j. morningstar and m. j. peardon , , 034509 ( 1999 ) [ arxiv : hep - lat/9901004 ] .
e. b. gregory , a. c. irving , c. c. mcneile , s. miller and z. sroczynski , pos * ( lat2005 ) * , 027 ( 2006 ) [ arxiv : hep - lat/0510066 ] .
b. a. li , , 034019 ( 2006 ) [ arxiv : hep - ph/0510093 ] .
m. r. pennington , e - conf .
c070910:106 ( 2007 ) [ arxiv:0711.1435 ] .
r. dermisek and j. f. gunion , , 041801 ( 2005 ) [ arxiv : hep - ph/0502105 ]
. y. nomura and j. thaler , , 075008 ( 2009 ) [ arxiv:0810.5397 [ hep - ph ] ] .
r. dermisek and j. f. gunion , , 111701r ( 2006 ) [ arxiv : hep - ph/0510322 ] .
r. dermisek , j. f. gunion and b. mcelrath , , 051105r ( 2007 ) [ arxiv : hep - ph/0612031 ] .
h. park _ et al .
_ [ hypercp collaboration ] , , 021801 ( 2005 ) [ arxiv : hep - ex/0501014 ] . m. l. mangano and p. nason , , 1373 ( 2007 ) [ arxiv:0704.1719 [ hep - ph ] ] .
y. j. zhang and h. s. shao , arxiv:0911.1766 [ hep - ph ] .
a. einstein , , 844 ( 1915 ) .
s. l. glashow , nucl .
* 22 * , 579 ( 1961 ) .
s. weinberg , , 1264 ( 1967 ) .
a. salam , in elementary particle theory , edited by n. svartholm ( almquist and wiksells , stockholm , 1969 ) , p. 367 .
s. l. glashow , j. iliopoulos and l. maiani , , 1285 ( 1970 ) .
n. a. trnqvist , arxiv : hep - ph/0204215 .
n. a. trnqvist , , 145 ( 2005 ) [ arxiv : hep - ph/0504204 ] .
p. minkowski and w. ochs , , 283 ( 1999 ) [ arxiv : hep - ph/9811518 ] . | we present evidence for the existence of a light scalar particle that most probably couples exclusively to gluons and quarks .
theoretical and phenomenological arguments are presented to support the existence of a light scalar boson for confinement and quark - pair creation .
previously observed interference effects allow to set a narrow window for the scalar s mass and also for its flavor - mass - dependent coupling to quarks . here , in order to find a direct signal indicating its production , we study published babar data on leptonic bottomonium decays , viz.the reactions @xmath0
@xmath1 @xmath2 @xmath1 @xmath3 ( and @xmath4 ) .
we observe a clear excess signal in the invariant - mass projections of @xmath0 and @xmath5 , which may be due to the emission of a so far unobserved scalar particle with a mass of about 38 mev . in the process of our analysis
, we also find an indication of the existence of a @xmath6 hybrid state at about 10.061 gev .
further signals could be interpreted as replicas with masses two and three times as large as the lightest scalar particle . |
* figure 1 : * reduction factor @xmath54 for the excitation probability as a function of the characteristic parameter @xmath26 defined in eq .
[ zdef ] for the two model wave functions ( i : , ii : - - - ) . + * figure 2 : * reduction @xmath102 of the total cross section for the excitation of the @xmath11 first excited state from the @xmath103 ground state of @xmath0be as a function of the projectile velocity @xmath17 for the two model wave functions ( i : , ii : - - - ) . | we study the corrections of first order electromagnetic excitation due to higher order electromagnetic interactions .
an effective operator is introduced which takes these effects into account in the sudden approximation .
evaluating the matrix - elements of this operator between the relevant states corrections to the first order result are obtained in a simple way . as an example we discuss the excitation of the first excited state in @xmath0be .
it tends to improve the agreement between experiment and theory .
425.70 in 0.80 in 0.20 in electromagnetic excitation in the energy domain of several tens of mev / u up to relativistic energies is a growing field of study .
the cross - section can become large and irreducible nuclear effects can be kept under control . with increasing beam energy
the equivalent photon spectrum becomes harder , and also particle - unstable states can be reached .
the coulomb dissociation @xmath0li @xmath1 @xmath2li + 2n and @xmath3o @xmath1 @xmath4n + p , which is also astrophysically relevant , are examples @xcite . recently ,
bound states were also excited and their ( doppler shifted ) de - excitation @xmath5-rays were measured .
a large deformation of the neutron - rich nucleus @xmath6 mg was recently deduced from a measurement of the @xmath7 885 kev transition to the ground state after medium energy electromagnetic excitation @xcite .
the 320 kev @xmath8 @xmath5-transition in @xmath0be was recently observed .
the measured cross - section for the @xmath0be ( @xmath9 ) coulomb excitation was found to be noticeably less than expected from the known lifetime and @xmath10 order pure coulomb excitation @xcite .
apart from possible nuclear and coulomb - nuclear interference effects , a possible reason for this discrepancy is the influence of higher order electromagnetic interaction .
it is the purpose of this letter to describe a framework suitable for fast projectiles .
e.g. the rather loosely bound @xmath0be in its @xmath10 excited @xmath11 state could easily be excited electromagnetically into the continuum in a second step @xcite .
electromagnetic excitation is mainly characterized by two parameters , the adiabaticity parameter @xmath12 and the strength parameter @xmath13 the excitation energy is given by @xmath14 , the impact parameter in a straight - line approximation is denoted by @xmath15 , and @xmath16 , where @xmath17 is the projectile velocity .
the target charge number is denoted by @xmath18 , and @xmath19 the electric multipole operator . in coulomb excitation below the barrier ,
multiple electromagnetic excitation is usually treated in a coupled channels approach using the relevant states from appropriate nuclear models , like the harmonic vibrator or rigid rotor . for a review see ref .
@xcite .
the situation for coulomb excitation above the barrier becomes simpler , because the excitations tend to be sudden .
while @xmath14 is restricted to a few mev , the adiabaticity parameter @xmath20 is typically less than 1 , for the important range of impact parameters @xmath21 , where @xmath22 and @xmath23 are the nuclear radii of projectile and target .
thus fast collisions become the domain of the sudden approximation @xcite , or of a recently developed low-@xmath20 approximation @xcite . in this case
it can be advantageous to construct operators which take into account the influence of intermediate states . for simplicity ,
let us use the straight - line approximation and the dipole approximation .
the first order excitation amplitude is given by @xcite @xmath24 with @xmath25 where the projectile moves in the @xmath26-direction and the impact parameter points to the @xmath27-direction .
the dipole effective charge is given by @xmath28 using a model of two pointlike inert clusters @xmath15 and @xmath29 with charge numbers @xmath30 , @xmath31 and masses @xmath32 and @xmath33 .
@xmath34 and @xmath35 are the modified bessel functions .
for @xmath36 we get the classical coulomb push @xmath37 in the limit @xmath38 , the excitation amplitude can be evaluated easily to all orders in the sudden approximation .
it is given by @xmath39 this could be generalized to higher multipolarities and trajectories corrected for coulomb deflection . by comparing @xmath40 and
@xmath41 the influence of multiple electromagnetic excitation can be assessed .
this is a remarkably simple procedure , all intermediate states are included .
the reduction of the excitation probability due to higher order effects is given by @xmath42 for loosely bound states , e.g. in @xmath0be = @xmath43be + n , we use simple model wave functions to reveal the characteristic parameters .
we choose two models , with the correct asymptotic behaviour .
their differences give a feeling about the model dependence .
for the wave functions we make the ansatz for the radial part in the initial state ( with orbital angular momentum @xmath44 ) @xmath45 which corresponds to the solution of the schrdinger equation with a @xmath46-like potential .
for the final state ( @xmath47 ) we choose @xmath48 and the more extended wave function @xmath49 respectively .
the constants @xmath50 ( @xmath51 ) are calculated from the binding energies @xmath52 . with these wave functions
the calculated mean lifetimes of the @xmath11 state are 110 fs and 97 fs , respectively .
they are smaller than the experimental value of ( @xmath53 ) fs @xcite . despite this difference ,
the ratio of higher order to first order effects can be given with some confidence , where , e.g. , common spectroscopic factors cancel out .
the absolute value depends on more sophisticated details of the nuclear model ( see , e.g. , ref .
@xcite ) .
the reduction factor @xmath54 can be calculated analytically .
we get @xmath55 and @xmath56 resp . , where we have introduced the parameter @xmath57 the parameter @xmath26 is directly related to @xmath58 ( eq . [ chidef ] ) .
it describes the ratio of the strength of the coulomb push @xmath59 and the `` looseliness '' of the system , and is a measure of the importance of higher order effects .
the quantity @xmath54 is plotted in fig . 1 for the two model wave - functions described above . for large @xmath15 the sudden approximation fails ( @xmath60 ) , on the other hand , higher order effects diminish due to the decrease of the strength parameter @xmath58 .
the product @xmath61 is a very small number for low excitation energies @xmath14 and high projectile velocities @xmath17 . the ranges of validity for the first order calculation ( @xmath62 small , @xmath20 arbitrary ) and the sudden approximation ( @xmath20 small , @xmath62 arbitrary ) overlap . in a convenient and accurate interpolation procedure
we calculate the total cross section in the following way @xmath63 which should be an accurate expression for all values of b in the integrand . for small impact parameters we have @xmath64 and the first order approximation cancels out in the calculation of @xmath65 .
we compare it with the total cross section in the first order calculation @xmath66 both cross sections depend on the value of the minimum impact parameter @xmath67 .
the change in the two cross sections will be similar so that their ratio is less affected by a change in @xmath67 .
if the sudden approximation is not well enough fulfilled , one could use the low-@xmath20 approximation @xcite which takes second order electromagnetic effects into account . for the @xmath0be coulomb excitation a reduction of the cross section from @xmath68 mb ( expected from the first order calculation ) to the measured @xmath69 mb
was recently found in an experiment at ganil @xcite .
indeed , the parameter @xmath26 can become substantial in this case , and higher order effects are not negligible . for a collision with @xmath70 fm and an energy of @xmath71 mev ( @xmath72 )
we have @xmath73 and from fig .
1 we see a substantial reduction of the excitation probability .
in a recent experiment at riken @xcite , which is currently evaluated , an energy of about @xmath74 mev ( @xmath75 ) for the @xmath0be was used .
this leads to a value of @xmath76 for the same impact parameter corresponding to a smaller reduction .
the product @xmath77 takes on the small values @xmath78 and @xmath79 , respectively , for the two models and the ganil conditions . at @xmath80 fm
we get an excitation probability of 3.4% and 3.9% in the first order calculation , decreasing with @xmath81 . for impact parameters larger than @xmath82 fm
the excitation becomes adiabatic and the excitation probability drops off exponentially .
the excitation probability is small compared to 1 and the non - conservation of unitarity in the first order approximation will not affect the calculation of the cross section .
the apparent reduction of the b(e1)-value ( due to higher order effects ) is given by @xmath83 which is plotted in fig . 2 as a function of the projectile velocity @xmath17 for the two model wave functions described above .
we assume a minimum impact parameter @xmath84 fm corresponding to a grazing collision of the projectile and target .
this will give an estimate of the largest possible effect to be expected from the higher order contributions .
we obtain a reduction , depending on the particular model chosen , of @xmath85 or @xmath86 for the ganil energy and @xmath87 or @xmath88 for the riken energy .
the range of the reduction r in the two models gives a feeling of the reliabiliy of the results .
the use of more realistic wavefunctions is outside the scope of the present work .
the reduction of the excitation probability for the @xmath11 state by higher order effects will be accompanied by an increase of the cross section in the breakup channel . in the dipole approximation ,
the lowest order correction was of @xmath89 order .
e1-e2 excitation contributes already in second order .
nevertheless , it is smaller than the third order e1-e1-e1 correction .
for low @xmath20 we can estimate the ratio of the two amplitudes as @xmath90 for @xmath91 , @xmath92 , and @xmath93 at the ganil energy .
thus the dipole approximation is reasonable , at least for a first exploration .
possible higher order effects in the @xmath6 mg intermediate energy coulomb excitation @xcite can also be estimated .
assuming a rigid rotor model , high energy coulomb excitation was calculated in the sudden approximation in ref .
@xcite .
the characteristic strength parameter is @xmath94 using the b(e2)-value found in ref .
@xcite we have ( see eq . 6 of ref .
@xcite , where a factor @xmath95 is missing on the rhs . ) @xmath96 and @xmath97 with @xmath98 fm for an energy of @xmath99 mev . from fig .
1a of ref .
@xcite it can be seen that higher order effects are negligible for the value of the strength parameter @xmath100 .
this is in agreement with the result found in the coupled channel calculation of ref .
@xcite . in conclusion
, we provide a framework to apply corrections to 1@xmath101 order electromagnetic excitation .
it is appropriate for fast collisions .
we constructed operators which take the influence of intermediate states into account .
this can lead to a great simplification as compared , e.g. , to the coupled channels approach . in this approach ,
a set of states , considered to be relevant , has to be chosen with known electromagnetic matrix - elements . in the present approach ,
of course , the model dependence can not be altogether avoided ; it enters when the corresponding matrix - elements of the operator has to be calculated , or at least estimated . as an example we studied the excitation of the @xmath11 state in @xmath0be . the importance of higher order effects for this case of an extremely loosely bound nucleus
was established .
the estimate for the reduction of the cross section can only partly explain the observed reduction of the b(e1)-value in the ganil experiment .
however , for a final analysis , more accurate calculations with improved wave functions , including e2 and nuclear effects in the excitation , should be performed .
we wish to thank p. g. hansen and m. g. saint - laurent for stimulating discussions .
note added in revision : in the meantime we got to know about a coupled channel study of the @xmath0be coulomb excitation by c. a. bertulani , l. f. canto , and m. s. hussein .
they get very similar conclusions as compared to our findings .
99 see e.g. g. baur and s. typel , in : proceedings of the 6th international conference on clusters in nuclear structure and dynamics , 6 9 sept .
1994 , ed . by f.
haas , ( strasbourg , 1995 ) and further references given there . t. motobayashi et al . , phys .
lett .
b 346 ( 1995 ) 9 .
r. anne et al . , ganil preprint p-94 - 35 ( 1994 ) , z. phys .
a in print .
p. g. hansen , private communication .
k. alder and a. winther , electromagnetic excitation ( north - holland , amsterdam , 1975 ) . s. typel and g. baur , nucl .
phys . a 573 ( 1994 ) 486 .
s. typel and g. baur , phys .
rev .
c 50 ( 1994 ) 2104 .
d. j. millener , j. w. olness , and e. k. warburton , phys .
rev .
c 28 ( 1983 ) 497 .
t. motobayashi , private communication .
g. baur , z. phys .
a 332 ( 1989 ) 203 . |
homodyne detection is an experimental method that is used to reconstruct quantum states of coherent light by repeatedly measuring a discrete set of field quadratures @xcite .
usually , a very high detection efficiency and ad - hoc designed apparatuses with low electronic noise are required @xcite .
new methods capable of discriminating between different quantum states of light , even with low detection efficiencies , will pave the road to the application of quantum homodyne detection for studying different physical systems embedded in a high noise environment @xcite .
for this purpose , specific quantum statistical methods , based on minimax and adaptive estimation of the wigner function , have been developed in @xcite .
these approaches allow for the efficient reconstruction of the wigner function under any noise condition , at the price of acquiring larger amounts of data .
hence , they overcome the limits of more conventional pattern function quantum tomography @xcite . the important consequence of this novel statistical approach is that the @xmath0 detection efficiency threshold can be overcome and quantum tomography is still practicable when the signals are measured with appropriate statistics .
the scope of this paper is to report the results of this method tested by performing numerical experiments .
indeed , we consider a linear superposition of two coherent states and numerically generate homodyne data according to the corresponding probability distribution distorted by an independent gaussian noise simulating efficiencies lower than @xmath0 . by properly expanding the set of numerically generated data ,
we are able to reconstruct the wigner function of the linear superposition within errors that are compatible with the theoretical bounds .
our results support the theoretical indications that homodyne reconstruction of linear superposition of quantum states is indeed possible also at efficiencies lower than 0.5 .
let us consider a quantum system with one degree of freedom described by the hilbert space @xmath1 of square integrable functions @xmath2 over the real line .
the most general states of such a system are density matrices @xmath3 , namely convex combinations of projectors @xmath4 onto normalised vector states @xmath5 any density matrix @xmath3 can be completely characterised by the associated wigner function @xmath6 on the phase - space @xmath7 ; namely , by the non positive - definite ( pseudo ) distribution defined by @xmath8=\frac{1}{2\pi}\int_{\mathbb{r}}{\rm d}u\,{\rm e}^{i\,u\,p}\,\left < q - v/2\vert\hat{\rho}\vert q+v/2\right>\ . \label{wigner}\ ] ] here @xmath9 and @xmath10 are the position and momentum operators obeying the commutation relations @xmath11=i$ ] , @xmath12 , and @xmath13 are eigenstates of @xmath14 : @xmath15 .
notice that @xmath6 is a square integrable function : @xmath16 among the advantages of such a representation , is the possibility of expressing the mean value of any operator @xmath17 with respect to a state @xmath3 as a pseudo - expectation with respect to @xmath6 of an associated function @xmath18 over the phase - space , where @xmath19\ .
\label{wigner1}\ ] ] indeed , by direct inspection one finds @xmath20 in homodyne detection , a monochromatic signal photon state is mixed with a coherent reference state , a so - called local oscillator , by a @xmath21 beam splitter . the output
is collected by two photodiodes and the difference photocurrent is measured .
it can be proved that , when the local oscillator is significantly more intense than the signal , the homodyne photocurrent is proportional to the signal quadrature @xcite . denoting by @xmath22 and @xmath23 the single mode annihilation and creation operators associated with the signal ,
the quadrature operator is defined as @xmath24 where @xmath25 is the relative phase between signal and local oscillator .
the continuum set of quadratures with @xmath26 $ ] provides a complete characterization of the signal state . using the annihilation and creation operators @xmath27 one constructs position and momentum - like operators , @xmath28 and @xmath29 . with respect to the latter ,
the quadrature operator reads : @xmath30 quadrature operators have continuous spectrum extending over the whole real line , @xmath31 ; given a generic one - mode photon state associated with a density matrix @xmath3 , its diagonal elements with respect to the ( pseudo ) eigenvectors @xmath32 represent the probability distribution over the quadrature spectrum . in homodyne detection experiments the collected data consist of @xmath33 pairs of quadrature amplitudes and phases @xmath34 :
these can be considered as independent , identically distributed stochastic variables .
given the probability density @xmath35 , one could reconstruct the wigner function by substituting the integration with a sum over the pairs for a sufficiently large number of data .
however , the measured values @xmath36 are typically not the eigenvalues of @xmath37 , rather those of @xmath38 where @xmath39 is a normally distributed random variable describing the possible noise that may affect the homodyne detection data and @xmath40 parametrizes the detection efficiency that increases from @xmath41 to @xmath42 with @xmath40 increasing from @xmath41 to @xmath43 @xcite
. the noise can safely be considered gaussian and independent from the statistical properties of the quantum state , that is , @xmath39 can be considered as independent from @xmath37 .
as briefly summarised in appendix a , the wigner function is reconstructed from a given set of @xmath33 measured homodyne pairs @xmath34 , @xmath44 , by means of an estimator of the form @xcite @xmath45-\frac{x_\ell}{\sqrt{\eta}}\right)\ , \\ \label{estimator2 } & & k_h^\eta\left([(q , p);\phi_\ell]-\frac{x_\ell}{\sqrt{\eta}}\right)=\int_{-1/h}^{1/h}{\rm d}\xi\,\frac{|\xi|}{4\pi}\ , { \rm e}^{i\xi(q\cos\phi_\ell+p\sin\phi_\ell - x_\ell/\sqrt{\eta})}\ , { \rm e}^{\gamma\xi^2}\ .\end{aligned}\ ] ] this expression is an approximation of the wigner function in by a sum over @xmath33 homodyne pairs @xmath34 .
the parameter @xmath46 serves to control the divergent factor @xmath47 , while @xmath48 , through the characteristic function @xmath49 of a circle @xmath50 of radius @xmath48 around the origin , restricts the reconstruction to the points @xmath51 such that @xmath52 .
both parameters have to be chosen in order to minimise the reconstruction error which is conveniently measured @xcite by the @xmath53-distance between the true wigner function and the reconstructed one , @xmath54 . since such a distance is a function of the data through @xmath55 , the @xmath53-norm has to be averaged over different sets , @xmath56 , of quadrature data : @xmath57 \equiv e\left[\int_{\mathbb{r}^2}{\rm d}q{\rm d}p\ , \left|w_\rho(q , p)-w^{\eta , r}_{h , n}(q , p)\right|^2\right]\ , \ ] ] where @xmath58 denotes the average over the @xmath56 data samples , each sample consisting of @xmath33 quadrature pairs @xmath34 corresponding to measured values of @xmath59 with @xmath26 $ ] . in @xcite , an optimal dependence of the parameters @xmath48 and @xmath46 upon the number of data , @xmath33 ,
is obtained by minimizing an upper bound to @xmath60 .
the functional relation between the parameters @xmath46 and @xmath48 on @xmath33 also depends on an auxiliary parameter @xmath61 .
this was introduced in @xcite to characterise the localisation properties on @xmath62 of the fourier transforms of the wigner functions of the following class of density matrices addressed in that context : @xmath63(q , p)\right|^2\,{\rm e}^{2\beta(w_1 ^ 2+w_2 ^ 2)^{s/2}}\,\leq\,(2\pi)^2l \right\}$ ] .
] the quadrature probability distribution can be conveniently related to the wigner function by passing to polar coordinates @xmath113 , @xmath114 , such that @xmath115 and @xmath116 : @xmath117\\ \nonumber & = & \int_0^\pi{\rm d}\phi\int_{-\infty}^{+\infty}{\rm d}\xi\frac{|\xi|}{(2\pi)^2}\int_{-\infty}^{+\infty}{\rm d}x\ , { \rm e}^{i\xi(q\cos\phi+p\sin\phi - x)}\,p_\rho(x,\phi)\\ \label{fourier } & = & \int_0^\pi{\rm d}\phi\int_{-\infty}^{+\infty}{\rm d}\xi\frac{|\xi|}{(2\pi)^2}\int_{-\infty}^{+\infty}{\rm d}x\ , { \rm e}^{i\xi(q\cos\phi+p\sin\phi)}\,f[p_\rho(x,\phi)](\xi)\ , \end{aligned}\ ] ] where @xmath118(\xi)$ ] denotes the fourier transform with respect to @xmath36 of the probability distribution : @xmath119(\xi)=\int_{-\infty}^{+\infty}{\rm d}x\,{\rm e}^{-ix\xi}\,p_\rho(x,\phi)\ .\ ] ] since @xmath39 can be considered a normally distributed random variable independent of @xmath37 , the noise affected distribution of the eigenvalues of @xmath37 in is given by the following convolution : @xmath120 its fourier transform is connected with that of @xmath35 according to @xmath121(\xi)={\rm e}^{\gamma\xi^2}\,f[p^\eta_\rho(\,x\,,\phi)](\xi/\sqrt{\eta})\ , \quad\hbox{with}\quad \gamma:=\frac{1-\eta}{4\eta}\ .\ ] ] by inserting @xmath122(\xi)$ ] into , one can finally write the wigner function in terms of the noisy probability distribution @xmath72 : @xmath123
homodyne reconstruction is particularly useful to expose quantum interference effects that typically spoil positivity of the wigner function : it is exactly these effects that are claimed not to be accessible by homodyne reconstruction in presence of efficiency lower than @xmath64 , namely when @xmath40 in is smaller than @xmath65 @xcite .
however , in @xcite it is theoretically shown that @xmath66 only requires increasingly larger data sets for achieving small reconstruction errors .
however , this claim was not put to test in those studies as the values of @xmath40 in the considered numerical experiments were close to @xmath43 . instead , we here consider values @xmath66 and reconstruct the wigner function of the following superposition of coherent states @xmath67 with @xmath68 any complex number @xmath69 .
the wigner function corresponding to the pure state @xmath70 is shown in figure [ true ] for @xmath71 .
its general expression together with that of its fourier transform and of the probability distributions @xmath35 and @xmath72 are given in appendix b. ( @xmath73 ; @xmath74).,scaledwidth=45.0% ] in appendix c , a derivation is provided of the @xmath53-errors and of the optimal dependence of @xmath46 and @xmath48 on the number of data @xmath33 and on a parameter @xmath75 that takes into account the fast decay of both the wigner function and its fourier transform for large values of their arguments .
the following upper bound to the mean square error in is derived : @xmath76 with @xmath77 and @xmath78 as given in . as explained in appendix c ,
the quantities @xmath79 do not depend on @xmath46 , @xmath48 and @xmath33 . by taking the derivatives with respect to @xmath48 and @xmath46 ,
one finds that the upper bound to the mean square deviation is minimised , for large @xmath33 , by choosing @xmath80 + we generated @xmath81 samples of @xmath82 quadrature data @xmath34 distributed according to the noisy probability density @xmath83 explicitly given in of appendix b , considering an efficiency lower than @xmath64 ( @xmath84 ) .
starting from each set of simulated quadrature data we reconstructed the associated wigner function by means of and .
the averaged reconstructed wigner functions @xmath85 $ ] for @xmath86 are shown in figure [ 045 ] for two different values of the parameter @xmath75 .
$ ] over @xmath81 samples of @xmath82 noisy quadrature data ( efficiency @xmath86 ) .
two different values of @xmath75 are considered.,scaledwidth=65.0% ] the mean square error of the reconstructed wigner functions has been computed as in and compared with the mathematically predicted upper bounds @xmath87 .
the dependence of the upper bound reconstruction error on the parameter @xmath75 is discussed at the end in appendix c. in table [ tab2 ] , we compare the reconstruction errors @xmath60 with their upper bound @xmath87 for two significant values of @xmath75 .
.calculated @xmath60 for @xmath81 samples of noisy quadrature data ( @xmath84 ) for two different values of @xmath75 .
comparison with the mathematical prediction of the upper bound @xmath87 .
[ cols="^,^,^",options="header " , ] + despite common belief , the interference features clearly appear in the reconstructed wigner function also for efficiencies lower than @xmath64 and the reconstruction errors are compatible with the theoretical predictions . in the next section ,
we make a quantitative study of the visibility of the interference effects .
the interference effects in the state @xmath88 can be witnessed by an observable @xmath89 of the form @xmath90 with respect to an incoherent mixture of the two coherent states , @xmath91 its mean value is given by @xmath92 therefore , from it follows that the phase - space function @xmath93 associated to @xmath94 is @xmath95 for details see and in appendix b. let us denote by @xmath96 , the estimated wigner function @xmath97 in for the @xmath98-th set of collected quadrature data .
it yields a reconstructed mean value @xmath99 of which one can compute mean , @xmath100 , and standard deviation , @xmath101 , with respect to the @xmath56 sets of collected data : @xmath102 we computed @xmath103 and @xmath101 with @xmath81 simulated sets of noisy data with @xmath84 for two different numbers of simulated quadrature data ( see figure [ ovsbeta ] ) .
we repeated the procedure for different values of the parameter @xmath75 .
the results are presented in figure [ ovsbeta ] , where the error bars represent the computed @xmath104 . as a function of @xmath75 .
the error bars represent @xmath105 . for each @xmath75 , @xmath81 set of @xmath33 noisy quadrature data have been considered .
the square markers refer to @xmath86 ( @xmath106 blue marker and @xmath107 green markers ) while the round ones refer to @xmath108 ( @xmath106 ) .
the error bars for @xmath108 have been multiplied by @xmath109 in order to make them more visible.,scaledwidth=60.0% ] + in order to be compatible with the interference term present in @xmath88 , the reconstructed wigner functions should yield an average incompatible with the incoherent mean value in , namely such that @xmath110 we thus see that the condition in is verified for @xmath86 , that is the reconstructed wigner functions are not compatible with incoherent superpositions of coherent states , if enough data are considered .
we also notice that the same behavior is valid for the high efficiency @xmath111 .
the dependence of the errors on @xmath75 can be understood as follows : when @xmath75 decreases the integration interval in becomes larger and approaches the exact interval @xmath112 $ ] .
nevertheless this occurs at the price of increasing the reconstruction error .
this can be noted both in figure [ ovsbeta ] ( larger error bars ) and in figure [ 045 ] ( increasingly noisy effects in the reconstructed wigner function ) .
this problem can be overcome with a larger number of data samples @xmath56 , that reduce the reconstruction noise and compensate for the effect of decreasing @xmath75 .
we simulated quadrature data corresponding to high electronic noise and detection efficiencies lower than @xmath0 . under these operating conditions
the wigner function of a linear superposition of two coherent states could be reconstructed using the tomographic techniques developed in @xcite . moreover , by taking into account the decay properties of the wigner function along with those of its fourier transform , we have checked that the numerical reconstruction errors are compatible with the theoretical error bounds computed there .
furthermore , the reconstruction of the quantum interference pattern of the wigner function has been supported by a numerical study of the variance of an operatorial interference witness excluding that the reconstructed interferences might be an artefact of the reconstruction algorithm .
we have thus confirmed 1 ) that , as theoretically predicted in @xcite , a @xmath0 detection efficiency is not , as often stated in the quantum optics literature , an absolute threshold below which homodyne quantum state reconstruction is generically impossible and 2 ) that , instead , by suitably enlarging the size of the set of collected quadrature data , and using alternative techniques different from standard pattern function quantum tomography , one may indeed access quantum features even in low efficiency conditions .
these results also provide the tools for quantum state reconstruction in these operating conditions and set the boundaries of the applicability of the novel statistical approach to homodyne quantum state reconstruction by checking the increase of the number of quadrature data necessary for faithful reconstruction with decreasing detector efficiency .
the authors are grateful to francesca giusti for insightful discussions and critical reading of the paper and thank referee number 2 of the manuscript in reference @xcite for stimulating the discussions that led to this work .
this work has been supported by a grant from the university of trieste ( fra 2013 ) .
the wigner function corresponding to the pure state @xmath70 and its fourier transform read @xmath124(w_1,w_2)&=&\frac{1}{2\left(1+{\rm e}^{-2|\alpha|^2}\right)}\left ( { \rm e}^{-\frac{(w_1 + 2\sqrt{2}\alpha_2)^2+(w_2 - 2\sqrt{2}\alpha_1)^2)}{4}}\,+\ , { \rm e}^{-\frac{(w_1 - 2\sqrt{2}\alpha_2)^2+(w_2 + 2\sqrt{2}\alpha_1)^2)}{4}}\right.\\ \label{awig2 } & & \hskip 1 cm \left.+2\,{\rm e}^{-\frac{w_1 ^ 2+w_2 ^ 2}{4}}\,\cos\left(\sqrt{2}(w_1\alpha_1+w_2\alpha_2)\right)\right ) \ .\end{aligned}\ ] ] for a generic wigner function @xmath6 one computes the quadrature probability density @xmath35 in by means of the so - called radon transform : @xmath125 it follows that the probability density @xmath35 and the noise - affected probability density @xmath72 in are given by : @xmath126 where @xmath127
here we derive an upper bound to the mean square error of the reconstructed wigner function .
this analysis is necessary in order to find an optimal functional relation between the free parameters in the reconstruction algorithm such to minimise the reconstruction error . for this purpose
we follow the techniques developed in @xcite adapting them to the case on a linear superposition of coherent states . using and ,
one starts by rewriting the error in as the sum of three contributions : @xmath128-\left| e\left[w^{\eta}_{h , n}(q , p)\right]\right|^2\right)\\ \label{error2 } & + & \int_{c^c_r(0)}{\rm d}q{\rm d}p\ , \left|w_\rho(q , p)\right|^2\ \\ \label{error3 } & + & \int_{c_r(0)}{\rm d}q{\rm d}p\,\left|e\left[w^{\eta}_{h , n}(q , p)\right]-w_\rho(q , p)\right|^2 , \,\end{aligned}\ ] ] @xmath129 denoting the region outside the circle @xmath50 , where @xmath130 .
the first and the third term correspond to the variance and bias of the reconstructed wigner function respectively , while the second term is the error due to restricting the reconstruction to the circle @xmath50 .
+ given a density matrix @xmath3 , the second term can be directly calculated .
this is true also of the bias ; indeed , because of the hypothesis that the pairs @xmath34 are independent identically distributed stochastic variables , it turns out that @xmath131&=&\frac{1}{\pi\,n}\sum_{\ell=1}^n\,e\left[k^\eta_h\left([(q , p);\phi_\ell]-\frac{x_\ell}{\sqrt{\eta}}\right)\right]= \frac{1}{\pi}\,e\left[k^\eta_h\left([(q , p);\phi]-\frac{x}{\sqrt{\eta}}\right)\right]\\ \label{avrec } & = & \int_0^\pi{\rm d}\phi\int_{-1/h}^{1/h}{\rm d}\xi\frac{|\xi|}{(2\pi)^2}\int_{-\infty}^{+\infty}{\rm d}x\ , { \rm e}^{i\xi(q\cos\phi+p\sin\phi - x/\sqrt{\eta})}\,{\rm e}^{\gamma\xi^2}\,p_\rho(x,\phi)\end{aligned}\ ] ] differs from the true wigner function @xmath6 in by the integration over @xmath132 being restricted to the interval @xmath133 $ ]
. moreover , its fourier transform reads @xmath134\right](w)&=&\int_{-\infty}^{+\infty}{\rm
d}q\int_{-\infty}^{+\infty}{\rm d}p\,{\rm e}^{-i(qw_1+pw_2)}\,e\left[w^{\eta}_{h , n}(q , p)\right]\\ \label{favrec } & = & \chi_{[-1/h,1/h]}(\|w\|)\,f\left[w_\rho\right](w)\ , \qquad w=(w_1,w_2)\ , \end{aligned}\ ] ] where @xmath135}(\|w\|)$ ] is the characteristic function of the interval @xmath133 $ ]
. then , by means of plancherel equality , one gets @xmath136-w_\rho(q , p)\right|^2\leq \int_{\mathbb{r}^2}{\rm d}q{\rm d}p\,\left| e\left[w^{\eta}_{h , n}(q , p)\right]-w_\rho(q , p)\right|^2\\ \nonumber & & \hskip 0.5cm=\big\|e\left[w^\eta_{h , n}\right]-w_\rho\big\|_2 ^ 2=\frac{1}{4\pi^2 } \big\|f\left[e\left[w^\eta_{h , n}\right]\right]-f\left[w_\rho\right]\big\|_2 ^ 2\\ \label{error4 } & & \hskip 0.5 cm = \frac{1}{4\pi^2}\big\|f\left[w_\rho\right]\,\chi_{[-1/h,1/h]}-f\left[w_\rho\right]\big\|_2 ^ 2=\frac{1}{4\pi^2}\int_{\|w\|\geq1/h}{\rm d}w\ , \big|f\left[w_\rho\right](w)\big|^2\ .\end{aligned}\ ] ] the variance contribution can be estimated as follows : firstly , by using and , one recasts it as @xmath137-\left| e\left[w^{\eta}_{h , n}(q , p)\right]\right|^2\right)= \nonumber\\ & & \hskip-2.8 cm = \frac{1}{\pi^2\,n}\left\{e\left[\left\|k^{\eta}_{h}\left([\,(q , p)\,;\phi]-\frac{x } { \sqrt{\eta}}\right)\,\chi_r(q , p)\right\|^2\right ] \,-\,\left\|e\left[k^{\eta}_{h}\left([\,(q , p)\,;\phi]-\frac{x}{\sqrt{\eta}}\right)\,\chi_r(q , p)\right]\right\|^2\right\ } \ . \label{varrec}\end{aligned}\ ] ] then , a direct computation of the first contribution yields the upper bound @xmath138-\frac{x } { \sqrt{\eta}}\right)\,\chi_r(q , p)\right\|^2\right]\leq \sqrt{\frac{\pi}{\gamma}}\frac{r^2}{16\,h}\,{\rm e}^{\frac{2\gamma}{h^2}}\,(1+o(1))\ , \ \gamma:=\frac{1-\eta}{4\eta}\ , \ ] ] with @xmath139 denoting a quantity which vanishes as @xmath46 when @xmath140 . on the other hand ,
the second contribution can be estimated by extending the integration over the whole plane @xmath7 and using together with : @xmath141-\frac{x}{\sqrt{\eta}}\right)\,\chi_r(q ,
p)\right]\right\|^2\leq\frac{1}{4\pi^2}\left\|f\left[w_\rho\right]\right\|^2= \left\|w_\rho\right\|^2\leq \frac{1}{2\pi}\ .\ ] ] let us consider now the specific case of @xmath142 , the superposition of coherent states defined in .
the auxiliary parameter @xmath75 labelling the class of density matrices @xmath143 in footnote [ foot1 ] with @xmath144 can be used to further optimize the reconstruction error @xmath145 . in particular , since @xmath146 , we get the upper bounds @xmath147(w_1,w_2)\right|\leq 2 \\
\nonumber \left|w_\alpha(q , p)\right|^2&\leq&\frac{3}{4\pi^2}\left ( { \rm e}^{-2(q-\sqrt{2}\alpha_1)^2 - 2(p-\sqrt{2}\alpha_2)^2}\,+\ , { \rm e}^{-2(q+\sqrt{2}\alpha_1)^2 - 2(p+\sqrt{2}\alpha_2)^2}\right.\\ \label{bound2 } & & \left .
+ 4\,{\rm e}^{-2(q^2+p^2)}\right)\leq \frac{3}{2\pi^2}\left({\rm e}^{-(\sqrt{2}r-|\alpha|)^2}+2{\rm e}^{-2r^2}\right ) \ , \\
\nonumber \left|f\left[w_\alpha\right](w_1,w_2)\right|^2&\leq & \frac{3}{4}\left ( { \rm e}^{-\frac{(w_1 + 2\sqrt{2}\alpha_2)^2+(w_2 - 2\sqrt{2}\alpha_1)^2)}{2}}\,+\ , { \rm e}^{-\frac{(w_1 - 2\sqrt{2}\alpha_2)^2+(w_2 + 2\sqrt{2}\alpha_1)^2)}{2}}\right .
\\ \label{bound3 } & & \left .
+ 4\,{\rm e}^{-\frac{w_1 ^ 2+w_2 ^ 2}{2}}\right)\leq\frac{3}{2}\left({\rm e}^{-(s/\sqrt{2}-2|\alpha|)^2}+2{\rm e}^{-s^2/2}\right)\ , \end{aligned}\ ] ] where @xmath148 in and @xmath149 in .
then , one derives the upper bounds @xmath150 @xmath151(w_1,w_2)\right|^2\ , { \rm e}^{2\beta(w_1 ^ 2+w_2 ^ 2)}\\ & & \hskip+2.5 cm \leq\frac{6\pi\left(1+{\rm e}^{16\beta|\alpha|^2/(1 - 4\beta)}\left(1+\frac{2\sqrt{\pi}|\alpha|}{\sqrt{1 - 4\beta}}-{\rm e}^{-4|\alpha|^2/(1 - 4\beta)}\right)\right)}{1 - 4\beta}\ , \end{aligned}\ ] ] which simultaneously hold for @xmath77 .
then , by means of the cauchy - schwartz inequality , one can estimate the contribution to the error , @xmath152 and similary for , @xmath153(w_1,w_2)\right|^2\leq{\rm e}^{-\beta / h^2}\,\delta_3(h)\ , \end{aligned}\ ] ] where @xmath154 if @xmath155 , @xmath156 otherwise , and @xmath157 altogether , the previous estimates provide the following upper bound to the mean square error in - : @xmath158 where @xmath79 do not depend on @xmath46 , @xmath48 and @xmath33 and @xmath159 is the leading order term in . by setting the derivatives with respect to @xmath48 and @xmath46 of the right hand side equal to @xmath41 , one finds @xmath160 whenever @xmath75 is such that the arguments of the logarithms are much smaller than the number of data @xmath33 , to leading order in @xmath33 the upper bound to the mean square deviation is minimised by @xmath161 the range of possible values of @xmath75 is @xmath162 .
however , the upper bound @xmath87 becomes loose when @xmath163 and @xmath164 . in the first case , it is the quantity @xmath165 which diverges , in the second one , it is the variance contribution which diverges as the logarithm of the number of data .
it thus follows that the range of values @xmath166 $ ] where the numerical errors @xmath167 are comparable with their upper bounds @xmath87 is roughly between @xmath168 and @xmath169 for @xmath86 as indicated by the following figure [ upperbound ] .
smithey , d. t. and beck , m. and raymer , m. g. and faridani , a. , measurement of the wigner distribution and the density matrix of a light mode using optical homodyne tomography : application to squeezed states and the vacuum , " _ phys .
lett . _ * 70 * , 1244 ( 1993 ) n. b. grosse , n. owschimikow , r. aust , b. lingnau , a. koltchanov , m. kolarczik , k. ldge , and u. woggon , pump - probe quantum state tomography in a semiconductor optical amplifier , " _ opt .
express _ * 22 * , 32520 ( 2014 ) m. esposito , f. benatti , r. floreanini , s. olivares , f. randi , k. titimbo , m. pividori , f. novelli , f. cilento , f. parmigiani , and d. fausti , pulsed homodyne gaussian quantum tomography with low detection efficiency " , _
new j. phys .
_ , * 16 * , 043004 ( 2014 ) . | standard quantum state reconstruction techniques indicate that a detection efficiency of @xmath0 is an absolute threshold below which quantum interferences can not be measured
. however , alternative statistical techniques suggest that this threshold can be overcome at the price of increasing the statistics used for the reconstruction . in the following we present numerical experiments proving that quantum interferences can be measured even with a detection efficiency smaller than @xmath0 . at the same time we provide a guideline for handling the tomographic reconstruction of quantum states based on homodyne data collected by low efficiency detectors . |
the term ` stochastic resonance ' was introduced in the early 80s ( see @xcite and @xcite ) in the study of periodic advance of glaciers on earth .
the stochastic resonance is the effect of nonmonotone dependence of the response of a system on the noise when this noise ( for instance the temperature ) is added to a periodic input signal ( see e.g. @xcite , in which the author explains also differences and similarities with the notion of stochastic filtering ) .
an extensive review on stochastic resonance and its presence in different fields of applications can be found in @xcite .
following @xcite , as stochastic resonance we intend the phenomenon in which the transmission of a signal can be improved ( in terms of statistical quantities ) by the addition of noise . from the statistical point of view
the problem is to estimate a signal @xmath0 transmitted through a channel .
this signal has to be detected by a receiver that can reveal signals louder than a threshold @xmath1 .
if @xmath2 is bounded from above by @xmath1 , the signal is not observable and the problem has not a solution .
but , if some noise @xmath3 is added to the signal , the perturbed signal @xmath4 may be observable and inference can be done on @xmath2 .
too few noise is not sufficient to give good estimates and too much noise deteriorates excessively the signal .
the optimal in some sense level of the noise will be called stochastic resonance in this framework .
usually ( see @xcite ) the criterion applied to measure optimality of estimators are the shannon mutual information or the kullback divergence .
more recently the fisher information quantity have been also proposed ( see @xcite and @xcite ) .
here we are concerned with the fisher information quantity .
it happens that this quantity , as a function of the noise , can be maximized for certain noise structures . if there is only one global maximum , the corresponding noise level is the value for which we have stochastic resonance , if several local maxima are present , the phenomenon is called stochastic multi - resonance . in this paper
we study the problem of estimation and hypotheses testing for the following model : we suppose to have a threshold @xmath5 and a subthreshold constant and non negative signal @xmath6 , @xmath7 .
we add , in continuous time , a noise that is a trajectory of a diffusion process @xmath8 and we observe the perturbed signal @xmath9 where @xmath10 is the level of the noise .
we propose two schemes of observations : _
i ) _ we observe only the proportion of time spent by the perturbed signal over the threshold @xmath1 and _ ii ) _ we measure the energy of the perturbed signal when it is above the threshold .
the asymptotic is considered as time goes to infinity .
this approach differs from the ones in the current statistical literature basically for two reasons : the noise structure is an ergodic diffusion process and not a sequence of independent and identically distributed random variables and data are collected in continuous time .
this second aspect is a substantial difference but it is not a problem from the point of view of applications for the two schemes of observations proposed if one thinks at analogical devices .
we propose two different estimators for the schemes and we study their asymptotic properties .
we present an example where , in both cases , it emerges the phenomenon of stochastic resonance . for the same model we also solve the problem of testing the simple hypothesis @xmath11 against the simple alternative @xmath12 by applying the bayesian maximum a posterior probability criterion .
it emerges that the overall probability of error is nonmonotonically dependent on @xmath13 .
we show again that there exists a non trivial local minimum of this probability that is again the effect of stochastic resonance .
the presence of stochastic resonance in this context is noted for the first time here . the paper is organized as follows . in section
[ sec : model ] we set up the regularity assumptions of the model . in sections
[ sec : time ] and [ sec : energy ] we prove some asymptotic properties estimators for the two schemes and we calculate numerically the points where the fisher information quantity attains its maximum for both models .
it turns out that the estimators proposed are asymptotically equivalent to the maximum likelihood estimators .
section [ sec : test ] is devoted to the problem of hypotheses testing .
all the figures are collected at the end of the paper .
let @xmath1 be the threshold and @xmath6 a constant signal . taking @xmath14 will not influence the calculations that follows but
may improve the exposition , so we use this assumption .
let @xmath15 be a given diffusion process solution to the following stochastic differential equation @xmath16 with non random initial value @xmath17 .
the process @xmath15 is supposed to have the ergodic property with invariant measure @xmath18 and invariant distribution function @xmath19)$ ] as @xmath20 .
the functions @xmath21 and @xmath22 satisfy the global lipschitz condition @xmath23 where @xmath24 is the lipschitz constant . under condition [ cond : c1 ]
, equation has a unique strong solution ( see e.g. @xcite ) but any equivalent condition to [ cond : c1 ] can be assumed because we do not use explicitly it in the sequel .
the following conditions are needed to ensure the ergodicity of the process @xmath15 .
if @xmath25 and @xmath26 then there exists the stationary distribution function @xmath27 and it takes the following form @xmath28 again , any other couple of conditions that imply the existence of @xmath27 can be used instead of [ cond : c3 ] and [ cond : c3 ] .
we perturb the signal @xmath6 by adding , proportionally to some level @xmath10 , the trajectory diffusion process @xmath15 into the channel
. the result will be the perturbed signal @xmath29 .
this new signal will be detectable only when it is above the threshold @xmath1 .
moreover , @xmath30 is still ergodic with trend and diffusion coefficients respectively @xmath31 and @xmath32 and initial value @xmath33 , but we will not use directly this process . we denote by @xmath34 the observable part of the trajectory of @xmath30 , being @xmath35 the indicator function of the set @xmath36 .
we consider two possibile schemes of observation : * we observe only the proportion of time spent by @xmath30 over the threshold @xmath1 @xmath37 * we measure the energy of the signal @xmath38
@xmath39 in the next sections , for the two models we establish asymptotic properties of estimators given by the generalized method of moments . in @xcite different properties of the generalized method of moments for ergodic diffusion processes are studied . in this note
we follows the lines given in the paper of@xcite for the i.i.d . setting .
these results are interesting in themselves independently from the problem of stochastic resonance .
we give an example of stochastic resonance based on the process where the phenomenon of stochastic resonance appears pronounced and in which results in a closed form can be written down .
the random variable @xmath40 can be rewritten in terms of the process @xmath15 as @xmath41 by the ergodic property of @xmath15 we have that @xmath42 where @xmath43 has @xmath27 as distribution function . from it
derives that @xmath44 so that @xmath6 is a one - to - one continuous function of @xmath45 . from the glivenko - cantelli theorem ( see e.g. @xcite ) for the empirical distribution function ( edf ) defined by @xmath46 follows directly that @xmath40 is a @xmath47-consistent estimator of @xmath45 thus also @xmath48 is a @xmath49-consistent estimator for @xmath6 .
we can calculate the asymptotic variance of this estimator .
it is known that ( see @xcite and @xcite ) the edf is asymptotically gaussian and in particular @xmath50 where @xmath51 is the inverse of the analogue of the fisher information quantity in the problem of distribution function estimation : @xmath52 where @xmath53 and @xmath54 .
the quantity @xmath55 is also the minimax asymptotic lower bound for the quadratic risk associated to the estimation of @xmath56 , so that @xmath57 is asymptotically efficient in this sense .
the asymptotic variance @xmath58 of @xmath59 can be derived by means of the so - called @xmath60-method ( see e.g. @xcite ) : @xmath61 thus @xmath62 where @xmath2 is the density of @xmath27 .
the quantity @xmath58 can also be derived from the asymptotic minimal variance @xmath63 of the edf estimator .
in fact , with little abuse of notation , by putting @xmath64 and @xmath65 we have that @xmath66 we now show that @xmath40 also maximizes the approximate likelihood of the model .
in fact , for the central limit theorem for the edf we have @xmath67 where @xmath68 is a standard gaussian random variable .
thus @xmath69 where @xmath70 is the distribution function of @xmath68 .
we approximate the likelihood function of @xmath40 by @xmath71 that is maximal when @xmath72 thus , the maximum likelihood estimator of @xmath6 ( constructed on the approximated likelihood ) reads @xmath73 so if the approximation above is acceptable , one can infer the optimality property of @xmath40 of having minimum variance from being also the maximum likelihood estimator . to view the effect of stochastic resonance on the fisher information we consider a particular example . by setting @xmath74 and
@xmath75 the noise become a standard process solution to the stochastic differential equation @xmath76 in such a case , the ergodic distribution function @xmath27 is the gaussian law with zero mean and variance 1/2 .
the asymptotic variance @xmath58 assumes the following form @xmath77 where @xmath78 is the classical error function .
in figure [ fig1 ] it is shown that for this model there exists the phenomenon of stochastic resonance . for a fixed level of noise @xmath13 the fisher information increases as the signal @xmath6 is closer to the threshold @xmath1 .
for a fixed value of the signal @xmath6 , the fisher information , as a function of @xmath13 , has a single maximum , that is the optimal level of noise .
for example , if @xmath79 then the optimal level is @xmath80 and for @xmath81 it is @xmath82 .
suppose that it is possibile to observe not only the time when the perturbed process is over the threshold but also its trajectory above @xmath1 , say @xmath83 , @xmath84 .
we now show how it is possibile to estimate the unknown signal @xmath6 from the equivalent of the energy of the signal for @xmath85 : literally from the quantity @xmath86 we use the following general result from @xcite on the estimation of functionals of the invariant distribution functions for ergodic diffusion processes . _
let @xmath87 and @xmath88 be such that @xmath89 .
then @xmath90 is a @xmath49-consistent estimator for @xmath91 where @xmath43 is distributed according to @xmath27_. in our case @xmath92 and @xmath93 .
the estimator @xmath94 can be rewritten as @xmath95 and it converges to the quantity @xmath96 that is a continuous and increasing function of @xmath6 , @xmath97 .
its inverse @xmath98 allows us to have again @xmath99 . by applying the @xmath60-method once again
, we can obtain the asymptotic variance of @xmath100 from the asymptotic variance of @xmath94 .
@xmath101 where @xmath102 the asymptotic variance of @xmath94 is given by ( see @xcite ) @xmath103 where @xmath104 and its inverse is also the minimal asymptotic variance in the problem of estimation of functionals for ergodic diffusion .
thus , the asymptotic variance of @xmath100 is given by @xmath105 * remark : * _ by the asymptotic normality of @xmath94 follows that @xmath100 is also the value that maximizes the approximate likelihood function .
in fact , as in the previous example , if we approximate the density function of @xmath94 with @xmath106 it is clear that @xmath100 is its maximum . _ as before , we put in evidence the phenomenon of stochastic resonance by using the process as noise .
the quantities involved ( @xmath107 and @xmath108 ) transform into the following @xmath109 ( from which it appears that @xmath107 is an increasing function of @xmath6 ) and @xmath110 in figure [ fig1 ] it is plotted the fisher information of the model as a function of @xmath6 and @xmath13 .
also in this case there is evidence of stochastic resonance . for a fixed value of @xmath6
is then possibile to find the optimal noise level @xmath13 .
for example , taking @xmath111 then we have stochastic resonance at @xmath112 and for @xmath113 , @xmath114 .
we now study a problem of testing two simple hypotheses for the model discussed in the previous section . as in @xcite , we apply the maximum _ a posteriori _ probability ( map ) criterion .
we will see that the decision rules for our model are similar to the one proposed by chapeau - blondeau in the i.i.d . setting .
given the observation @xmath40 we want to verify the null hypothesis that the unknown constant signal is @xmath115 against the simple alternative @xmath116 , with @xmath117 : @xmath118 suppose that , before observing @xmath40 , we have a _
prior _ information on the parameter , that is @xmath119 and @xmath120 .
the map criterion uses the following likelihood ratio @xmath121 and the decision rule is to accept @xmath122 whenever @xmath123 ( decision @xmath124 ) or refuse it otherwise ( decision @xmath125 ) .
the overall probability of error is @xmath126 let now be @xmath127 then , the likelihood @xmath128 appears as @xmath129 to write explicitly the decision rule and then study the effect of stochastic resonance we have to distinguish three cases : @xmath130 , @xmath131 and @xmath132 . 1 .
let it be @xmath130 , then put @xmath133 then , if @xmath134 accept @xmath122 and @xmath135 .
if @xmath136 , then if @xmath137 reject @xmath122 otherwise accept it . in both cases @xmath138 2 .
let it be @xmath131 , then put @xmath139 then , if @xmath134 reject @xmath122 and @xmath140 .
if @xmath136 , then if @xmath137 accept @xmath122 otherwise reject it . in both cases
@xmath141 3 .
let it be @xmath132 , then put @xmath142 then , if @xmath143 reject @xmath122 otherwise accept it . in both cases
@xmath144 as before , we apply this method to the model . in this case
the variance @xmath145 , for a fixed threshold @xmath1 and noise level @xmath13 , is a non decreasing function of @xmath6 being @xmath146 only a scale factor ( figure [ fig2 ] gives a numerical representation of this statement ) .
thus , the @xmath147 is , in general , given by formula .
what is amazing is the behavior of @xmath147 . in figure [ fig3 ]
it is reported the graph of @xmath147 as a function of @xmath13 and @xmath116 given @xmath148 and @xmath149 . for @xmath116 around 1/2
the @xmath147 shows the effect of stochastic resonance .
so it appears that in some cases the noise level @xmath13 can reduce sensibly the overall probability of making the wrong decision .
this kind of behavior is non outlined in the work of @xcite .
* remark : * _ following the same scheme , similar results can be obtained for the model _
ii ) _ when we observe the energy @xmath94 . in this case
it is sufficient to replace in the values of @xmath150 and @xmath151 with the quantities @xmath152 and @xmath40 with @xmath94 in the decision rule . _
the use of ergodic diffusions as noise in the problem of stochastic resonance seems quite powerful .
characterizations of classes of ergodic process that enhance the stochastic resonance can be done ( see e.g. @xcite ) but not in a simple way as in the i.i.d .
case as calculations are always cumbersome .
the problem of a parametric non constant signal can also be treated while the full nonparametric non constant signal requires more attention and will be object for further investigations . for i.i.d .
observations , @xcite and @xcite considered the problem of non parametric estimation for regression models of the form @xmath153 , @xmath154 . their approach can be applied in this context
. other criterion of optimality than the fisher information quantity can be used as it is usually done in information theory ( e.g. shannon mutual information or kullback divergence ) .
the analysis of the overall probability of error seems to put in evidence something new with respect to the current literature ( see e.g. @xcite ) .
it is worth noting that in a recent paper @xcite models driven by ergodic diffusions have also been used but the effect of stochastic resonance is not used to estimate parameters .
barndorff - nielsen , o. , cox , d.r . , _
asymptotic techniques for use in statistics_. chapman & hall , 1989 , bristol .
benzi , r. , sutera , a. , vulpiani , a. , the mechanism of stochastic resonance .
_ j. phys .
a _ , v * 14 * , ( 1981 ) , l453-l458 .
berglund , n. , gentz , b. , a sample - path approach to noise - induced synchronization : stochastic resonance in a double - well potential .
wias preprint no .
627 , ( 2000 ) .
chapeau - blondeau , f. , nonlinear test statistic to improve signal detection in non - gaussian noise .
_ ieee sig .
_ , v * 7 * , n 7 , ( 2000 ) , 205 - 207 .
chapeau - blondeau , f. , rojas - varela , j. , information - theoretic measures improved by noise in nonlinear systems .
14th int . conf . on math .
theory of networks and systems _ , perpignane , france , ( 2000 ) , 79 - 82 .
greenwood , p.e . ,
ward , l.m . , wefelmeyer , w. , statistical analysis of stochastic resonance in a simple setting .
_ phys . rev .
, v * 60 * , n 4 , ( 2000 ) , 4687 - 4695 .
gammaitoni , l. , hnggi , p. , jung , p. , marchesoni , f. , stochastic resonance .
_ rev . of modern phys .
_ , 70 , ( 1998 ) , 223 - 288 .
klimontovich , yu.l .
, what are stochastic filtering and stochastic resonance ?
_ physics - uspecky _ , v * 42 * , n 1 , ( 1999 ) , 37 - 44 . kutoyants , yu.a . ,
efficiency of the empirical distribution function for ergodic diffusion .
_ bernoulli _ , v * 3 * , n 4 , ( 1997a ) , 445 - 456 . kutoyants , yu.a . , on semiparametric estimation for ergodic diffusion .
a. razmadze math .
_ , 115 , ( 1997b ) , 45 - 58 .
kutoyants , yu.a . , _ statistical inference for ergodic diffusion processes_. monography to appear ( 2000 ) .
ibragimov , i.a . ,
khasminskii , r.z . , _ statistical estimation ( asymptotic theory)_. springer , 1981 , new york .
liptser , r.s . ,
shiryayev , a.n .
, _ statistics of random processes _ , part .
springer - verlag , 1977 , new york .
mller , u. u. , nonparametric regression for threshold data .
_ canadian j. statist _ , * 28 * , ( 2000 ) , to appear .
mller , u. u. , ward , l. m. , stochastic resonance in a statistical model of a time - integrating detector _
e ( 3 ) _ , * 61 * , ( 2000 ) , to appear .
negri , i. ( 1998 ) .
stationary distribution function estimation for ergodic diffusion processes . _ statistical inference for stochastic processes _ , v * 1 * , n 1 , 61 - 84 .
nicolis , c. , stochastic aspects of climatic transitions
response to a periodic forcing .
_ tellus _ , * 34 * , n 1 , ( 1982 ) , 1 - 9 .
[ cols="^ " , ] as a function of @xmath13 for different values of @xmath116 given @xmath148 and @xmath149 . from left - top
to right - bottom : @xmath116 = 0.1 , 0.25 , 0.5 , 0.75 , 0.9 , 0.95 .
the bottom plot is to show the effect of stochastic resonance for @xmath116 = 0.5 . ] | a subthreshold signal is transmitted through a channel and may be detected when some noise with known structure and proportional to some level is added to the data .
there is an optimal noise level , called stochastic resonance , that corresponds to the highest fisher information in the problem of estimation of the signal . as noise we consider an ergodic diffusion process and the asymptotic
is considered as time goes to infinity .
we propose consistent estimators of the subthreshold signal and we solve further a problem of hypotheses testing .
we also discuss evidence of stochastic resonance for both estimation and hypotheses testing problems via examples .
* keywords : * stochastic resonance , diffusion processes , unobservable signal detection , maximum a posterior probability .
* msc : * 93e10 , 62m99 . |
after neutrino oscillation experiments have shown that neutrinos possess a non - zero rest mass the question of the absolute mass scale of neutrinos has become important for particle physics and cosmology ( see e.g. @xcite ) .
the goal of the karlsruhe tritium neutrino experiment ( katrin , @xcite ) is to search for the mass of the electron antineutrino with a sensitivity of @xmath0{ev / c^2}$ ] .
this will probe most of the mass range in which the three neutrino flavours have nearly degenerate masses and where neutrinos are of cosmological importance . to this end a precision measurement of the endpoint region of the @xmath3-decay of tritium will be performed with katrin .
the shape of this spectrum depends sensitively on the neutrino mass .
this will be done by using a windowless gaseous tritium source for small systematic uncertainties , and an electrostatic energy filter of mac - e filter type ( electrostatic filter with magnetic adiabatic collimation , @xcite ) for the analysis of the electron energy with high luminosity and high resolution .
this is the same method as used at the mainz @xcite and troitzk neutrino mass experiments @xcite , which have set the until now best upper limits of @xmath4{ev / c^2 } ( 95\%)$ ] on the neutrino mass .
figure [ fig : setup ] gives an overview of the katrin set - up .
it consists of a source section with the windowless gaseous tritium source , the wgts ( a ) , a transport section with a differential- and a cryopumping section ( b ) , a pre - spectrometer ( c ) , the main spectrometer ( d ) and a detection section with the focal plane detector ( e ) .
the katrin set - up ( explanation see text ) ] the wgts provides a source of electrons from tritium @xmath3-decay with high and uniform intensity .
it basically consists of a @xmath5{m}$ ] long beam tube of @xmath6{mm}$ ] diameter through which tritium is circulated in a loop with a flow of @xmath7{ci / s}$ ] and a maximum pressure of @xmath8{mbar}$ ] .
this results in a column density of @xmath9{t_2/cm^2}$ ] and @xmath10 @xmath3-particles per second accepted by the spectrometer . in order to keep density fluctuations at @xmath11 or less the wgts is temperature stabilized to @xmath12{mk}$ ] at a temperature of @xmath12{k}$ ] by two - phase neon cooling .
the purity of the @xmath13 is maintained at @xmath14 by continous purification of the gas in the tritium loop .
these stringent stability requirements result in a highly complex cryostat system @xcite .
the technologies used will be tested using a demonstrator set - up , which will be available in spring 2010 . in the final set - up the properties of the source
will be checked regularly with an electron source , which is under development @xcite , by laser raman spectroscopy , which already has been deployed successfully for test measurements @xcite , and a penning trap in the beamline to the spectrometers using ft - icr , which has recently been tested successfully @xcite .
the transport section serves to guide the electrons from tritium @xmath3-decay to the spectrometers without any losses and to prevent the transport of tritium at the same time , with a reduction factor for tritium of @xmath15 and a final flow of @xmath16mbar - l / s .
this is achieved by a differential pumping section , which recently arrived at karlsruhe for testing and commissioning , followed by a cryogenic pumping section . here
the remaining tritium gets cryosorbed on a layer of argon frost on the walls of the stainless steel vessel at a temperature of @xmath17{k}$ ] .
this method has been tested with the trap experiment successfully @xcite .
the pre - spectrometer is a mac - e filter like the main spectrometer .
it rejects all electrons with energy @xmath18{ev}$ ] below the endpoint , which are of no interest for the neutrino mass measurement and which even may increase the main spectrometer background .
it has been successfully tested and is working at design parameters with an intrinsic background of @xmath19(@xmath20 counts per second ( cps ) ) .
for now it is being used as a test case for the technologies and methods used for the main spectrometer , such as the vacuum system ( see e.g. @xcite ) and the electromagnetic design . for these tests
it is presently equipped with a 64-pixel pin diode @xcite .
the endpoint region of the @xmath3-spectrum is measured with the main spectrometer .
this consists of a stainless steel vacuum vessel of @xmath21{m}$ ] length and @xmath5{m}$ ] diameter and will reach a vacuum of @xmath22{mbar}$ ] using turbomolecular pumps and non - evaporating getter material .
the magnetic field at the entrance of this mac - e filter is @xmath23{t}$ ] and at its exit @xmath24{t}$ ] while the field in the central analysis plane will be @xmath25{t}$ ] resulting in an energy resolution of @xmath26{ev}$ ] .
the retardation voltage is applied directly to the spectrometer hull .
for fine shaping of the field and background suppression a modular inner wire electrode system will be used .
the main spectrometer has been set - up at karlsruhe and an outgassing rate of @xmath27mbar - l / s has been achieved after bakeout .
high demands on the quality of the electric potential and therefore for the precision of the machining of the wire modules required stringent quality control ( see e.g. @xcite ) .
the installation of the wire modules is in progress . the high voltage ( hv ) that creates the retardation potential in the analysis plane has to be highly stable .
to reach the sensitivity of katrin , fluctuations of the hv of @xmath28{ppm}$ ] , i.e. of @xmath29{mv}$ ] , have to be detected .
for this purpose a precision hv - divider has been developed @xcite to measure the applied voltage with this precision . in conjunction with calibration sources based on conversion electrons from the decay of @xmath30 this
will ensure the required short and long term stability of the high voltage .
electrons with sufficient energy to pass the retardation potential are counted with the focal plane detector .
this is a 148-fold segmented pin diode with surroundings made of low - activity materials to reduce the intrinsic background of the detector .
simulations show that a background of @xmath31{cps}$ ] can be achieved in the signal region
. the detector will be commissioned at karlsruhe in spring 2010 . the katrin background count rate necessary to reach the intended sensitivity is @xmath32{cps}$ ] .
by shifting the signal in the detector via a dc - voltage the intrinsic detector background can be reduced to @xmath33{cps}$ ] . however , other sources of background have to be controlled as well .
two major sources are background from the tank walls due to muon interactions and radioactive decays , and background from incidential penning traps due to the magnetic and electric fields employed in the mac - e filter .
the former is suppressed by the axial magnetic field , which guides most electrons from the wall back to the wall .
the remaining wall electrons will be suppressed by the wire electrode system , which is slightly more negative ( o(100v ) ) than the tank wall .
the background due to penning traps has been investigated in detail at the pre - spectrometer and can be avoided in large part by proper shaping of the electrodes . however , some unavoidable penning traps still exist , like the trap formed by the pre- and the main spectrometer .
investigation of these traps has started and first results are promising @xcite .
the katrin experiment searching for the neutrino mass is under construction .
the major components are being assembled and tested , and integration of the components will start soon .
installation of the inner electrode system of the main spectrometer is ongoing . in 2010
the properties of the main spectrometer will be investigated using calibration sources .
data taking using tritium will start in 2012 and will proceed for five years .
the work by the author is supported by the german bundesministerium fr bildung und forschung .
9 e. w. otten and c. weinheimer , rep . prog . phys .
* 71 * ( 2008 ) 086201 j. lesgourgues and s. pastor , phys . rep .
* 429 * ( 2006 ) 307 the katrin collaboration ( j. angrik _ et al .
_ ) , _ katrin design report 2004 _ , fzka scientific report 7090 , 2005 , available online at http://bibliothek.fzk.de/zb/berichte/fzka7090.pdf a. picard
_ et al . _ ,
b63 * ( 1992 ) 345 v. m. lobashev and p. e. spivak , nucl .
instr . meth . *
a240 * ( 1985 ) 305 ch .
et al . _ ,
j. * c 40 * ( 2005 ) 447 v. m. lobashev _
* b 460 * ( 1999 ) 227 s. grohmann , cryogenics * 49 * ( 2009 ) 413 b. bornschein , prog .
* 57 * ( 2006 ) 38 k. valerius _
et al . _ ,
new journal of physics * 11 * ( 2009 ) 063018 m. sturm _
, to be published in las .
* 20(2 ) * , 2010 m. ubieto - daz _ et al .
_ , int . j. mass spectr .
* 288 * ( 2009 ) 1 o. kazachenko _
et al . _ , nucl instr . and meth .
* a 587 * ( 2008 ) 136 f. eichelhardt _ et al .
_ , fusion science and technology * 54 * ( 2008 ) 615 x. luo _
et al . _ ,
vacuum * 81 * ( 2008 ) 777 j. wolf , journal of the vacuum society of japan * 52 * ( 2009 ) 278 s. wuestling _
et al . _ ,
meth * a 568 * ( 2006 ) 382 m. prall _ et al .
_ , submitted to ieee transactions on nuclear science , arxiv:0907.5490 th .
thmmler , r. marx and ch .
weinheimer , new journal of physics * 11 * ( 2009 ) 103007 m. beck , submitted to european physical journal * a * ( 2009 ) , arxiv:0909.3337 | the karlsruhe tritium neutrino mass experiment , katrin , aims to search for the mass of the electron neutrino with a sensitivity of @xmath0{ev / c^2}$ ] ( 90% c.l . ) and a detection limit of @xmath1{ev / c^2}$ ] ( @xmath2 ) .
both a positive or a negative result will have far reaching implications for cosmology and the standard model of particle physics and will give new input for astroparticle physics and cosmology .
the major components of katrin are being set up at the karlsruhe institut of technology in karlsruhe , germany , and test measurements of the individual components have started .
data taking with tritium is scheduled to start in 2012 . |
while most readers will be familiar with the notion of _ feedback control _
, for completeness we begin by defining this term .
feedback control is the process of monitoring a physical system , and using this information as it is being obtained ( in real time ) to apply forces to the system so as to control its dynamics .
this process , which is depicted in figure [ fig1 ] , is useful if , for example , the system is subject to noise . since quantum mechanical systems , including those which are continually observed , are dynamical systems , in a broad sense the theory of feedback control developed for classical dynamical systems applies directly to quantum systems . however , there are two important caveats to this statement .
the first is that most of the exact results which the theory of feedback control provides , especially those regarding the optimality and robustness of control algorithms , apply only to special subclasses of dynamical systems . in particular , most apply to linear systems driven by gaussian noise @xcite . since observed quantum systems in general obey a non - linear dynamics , an important question that arises is whether exact results regarding optimal control algorithms can be derived for special classes of quantum systems .
in addition to the need to derive results regarding optimality which are specific to classes of quantum systems , there is a property that sets feedback control in quantum systems apart from that in other systems .
this is the fact that in general the act of measuring a quantum system will alter it .
that is , measurement induces dynamics in a quantum system , and this dynamics is noisy as a result of the randomness of the measurement results .
thus , when considering the design of feedback control algorithms for quantum systems , the design of the algorithm is not independent of the measurement process . in general different ways of measuring
the system will introduce different amounts of noise , so that the search for an optimal feedback algorithm must involve an optimization over the manner of measurement . in what follows
we will discuss a number of explicit examples of feedback control in a variety of quantum systems , and this will allow us to give specific examples of the dynamics induced by measurement . before we examine such examples however , it is worth presenting the general equations which describe feedback control in quantum systems , in analogy to those for classical systems . in classical systems ,
the state - of - knowledge of someone observing the system is given by a probability density over the dynamical variables ( a phase - space probability density ) .
let us consider for simplicity a single particle , whose dynamical variables are its position , @xmath0 and momentum , @xmath1 .
if the observer is continually monitoring the position of the particle , then her stream of measurement results , @xmath2 is usually described well by @xmath3 where in each time interval @xmath4 , @xmath5 is a gaussian random variable with variance @xmath4 and a mean of zero .
such a gaussian noise process is called wiener noise .
the constant @xmath6 determines the relative size of the noise , and thus also the _ rate _ at which the measurement extracts information about @xmath0 ; when @xmath6 is increased , the noise decreases , and it therefore takes the observer less time to obtain an accurate measurement of @xmath0 .
as the observer obtains information , her state - of - knowledge regarding the system , @xmath7 , evolves .
the evolution is given by the kushner - stratonovich ( k - s ) equation .
this is @xmath8 p dt + \sqrt{\gamma } ( x - \langle x(t ) \rangle ) p dw , \ ] ] where @xmath9 is the mass of the particle , @xmath10 is the force on the particle , @xmath11 is the expectation value of @xmath0 at time @xmath12 , and @xmath13 turns out to be a wiener noise , uncorrelated with the probability density @xmath14 . because of this we can alternatively write the stream of measurement results as @xmath15 the above will no doubt be familiar to the majority of the readership . for linear dynamical systems the k - s equation reduces to the equations of the well - known kalman - bucy filter @xcite .
the k - s equation is the essential tool for describing feedback control ; it tells us what the observer knows about the system at each point in time , and thus the information that he or she can use to determine the feedback forces at each point in time .
in addition , when we include these forces in the system dynamics , the resulting k - s equation , in telling us the observer s state - of - knowledge is also telling us how effective is our feedback control : the variance of this state - of - knowledge , and the fluctuations of its mean ( note that these are two separate things ) tell us the remaining uncertainty in the system .
the k - s equation thus allows us to design and evaluate feedback algorithms .
the description of dynamics and continuous measurement in quantum mechanics is closely analogous to the classical case described above . in quantum mechanics , however , the observer s state - of - knowledge
must be represented by a matrix , rather than a probability density .
this matrix is called the _ density matrix _ , and usually denoted by @xmath16 .
the dynamical variables are also represented by matrices .
if the position is represented by the matrix @xmath17 , then the expectation value of the particle s position at time @xmath12 is given by @xmath18 $ ] @xcite . while the notion that a state - of - knowledge is described by a matrix will appear very strange to most of the readership , do nt let this put you off
when we consider feedback control we will always discuss it in terms of standard physical quantities such as the expectation values , variances , or probability densities for the dynamical variables .
the reason we speak of the density matrix as representing the observer s state - of - knowledge is because all these quantities can be obtained directly from the density matrix .
the dynamics of an unobserved quantum system may be written as @xmath19 \equiv h\rho - \rho h$ ] for a given matrix @xmath20 called the hamiltonian ( @xmath21 is planck s constant ) . if an observer makes a continuous measurement of a particle s position , then the full dynamics of the observer s state - of - knowledge is given by the quantum equivalent of the kushner - stratonovich equation .
this is @xmath22dt - ( \gamma/8 ) [ x , [ x , \rho]]dt \nonumber\\ & & + \frac{\sqrt{\gamma}}{2 } ( ( x - \langle x(t ) \rangle)\rho + \rho ( x - \langle x(t ) \rangle ) ) dw , \label{sme}\end{aligned}\ ] ] where the observer s stream of measurement results is @xmath23 this is referred to as the _ stochastic master equation _ (
sme ) , and was first derived by belavkin @xcite .
this is very similar to the k - s equation , but has the extra term @xmath24 $ ] that describes the noisy dynamics ( or _ quantum back - action _ ) which is introduced by the measurement . in the quantum mechanics
literature the measurement rate @xmath6 is often referred to as the _ measurement strength_. the sme is usually derived directly from quantum measurement theory @xcite without using the mathematical machinery of filtering theory . a recent derivation for people familiar with filtering theory
may be found in @xcite .
armed with the quantum equivalent of the k - s equation , we can proceed to consider feedback control in quantum systems . interpreting `` applications of feedback control in quantum systems '' in a broad sense
, it would appear that one can break most such applications into three general classes . while these classes may be somewhat artificial , they are a useful pedagogical tool , and we will focus on specific examples from each group in turn in the following three sections .
the first group is the application of results and techniques from classical control theory to the general theory of the control of quantum systems .
this includes the application control theory to obtain optimal control algorithms for special classes of quantum systems .
an example of this is the realization that classical lqg control theory can be applied directly to obtain optimal control algorithms for observed linear quantum systems @xcite .
we will discuss this and other examples in section [ systems ] .
it is worth noting at this point that while the direct application of classical control theory to quantum systems is very useful , it is not the only approach to understanding the design of feedback algorithms in quantum systems .
another approach is to try to gain insights into the relationship between measurement ( information extraction ) and disturbance in quantum mechanics which are relevant to feedback control .
such questions are also of fundamental interest to quantum theorists since they help to elucidate the information theoretic structure of quantum mechanics .
references @xcite and @xcite take this approach , elucidating the information - disturbance trade - off relations for a two - state system ( the simplest non - linear quantum system ) , and exploring the effects of this on the design of feedback algorithms .
the second group is the application of feedback control to classes of control _ problems _ which arise in quantum systems , some of which are analogous to those in classical systems , and some of which are peculiar to quantum systems .
a primary example of this is in adaptive measurement , where feedback control is used during the measurement process to change the properties of the measurement .
this is usually for the purpose of increasing the information which the measurement obtains about specific quantities , or increasing the rate at which information is obtained .
we will discuss such applications in section [ problems ] .
the third group is the design and application of feedback algorithms to control specific quantum systems .
examples are applications to the cooling of a nano - mechanical resonator and the cooling of a single atom trapped in an optical cavity .
we will discuss these in section [ physical ] .
most readers of this article will certainly be familiar with the classical theory of optimal linear - quadratic - gaussian ( lqg ) control .
this provides optimal feedback algorithms for linear systems driven by gaussian noise , and in which the observer monitors some linear combination of the dynamical variables . in lqg control ,
the control objective is the minimization of a quadratic function of the dynamical variables ( such as the energy ) .
it turns out that for a restricted class of observed quantum systems ( those which are linear in a sense to be defined below ) , this optimal control theory can be applied directly .
this was first realized by belavkin in @xcite , and later independently by yanagisawa and kimura @xcite and doherty and jacobs @xcite .
quantum mechanical systems whose hamiltonians are no more than quadratic in the dynamical variables are referred to as linear quantum systems , since the equations of motion for the matrices representing the dynamical variables are linear .
further , these linear equations are precisely the same as those for a classical system subject to the same forces .
the simplest example is the harmonic oscillator . if we denote the matrices for position and momentum as @xmath17 and @xmath14 respectively , then the hamiltonian is @xmath25 where @xmath9 is the mass of the particle , and @xmath26 is the angular frequency with which it oscillates .
the resulting equations for @xmath17 and @xmath14 are @xmath27 which are of course identical to the classical equations for the dynamical variables @xmath0 and @xmath1 in a classical harmonic oscillator .
further , it turns out that if an observer makes a continuous measurement of any linear combination of the position and momentum , then the sme for the observer s state of knowledge of the quantum system reduces to the kushner - stratonovich equation , which in this case , because the system is linear , is simply the kalman - bucy filter equations .
however , there is one twist .
the kalman - bucy equations one obtains for the quantum system are those for a classical harmonic oscillator driven by gaussian noise of strength @xmath28 .
this comes from the extra term in the sme which describes the `` quantum back - action '' noise generated by the measurement .
since the observer s state of knowledge evolves in precisely the same way as for the equivalent linear classical system , albeit driven by noise , we can apply classical lqg theory to these quantum systems .
if we apply linear feedback forces , then , for a fixed measurement strength @xmath6 , the quantum mechanics will tell us how much noise the system is subject to , and lqg theory will tell us the resulting optimal feedback algorithm for a given quadratic control objective @xcite .
note however , that since the noise driving the system depends upon the strength of the measurement , then the performance of the feedback algorithm will also depend upon the strength of the measurement .
moreover , the performance of the algorithm will be influenced by two competing effects : as the measurement gets stronger , we can expect the algorithm to do better as a result of the fact that the observer is more rapidly obtaining information .
however , as the measurement gets stronger , the induced noise also increases , which will reduce the effectiveness of the algorithm .
we can therefore expect that there will be an optimal measurement strength at which the feedback is most effective . in a linear quantum system one
therefore must first find the optimal feedback algorithm using lqg control theory , and then perform a second optimization over the measurement strength .
this is not the case in classical control .
an explicit example of optimizing measurement strength may be found in @xcite .
the application of lqg theory to linear quantum systems will be useful when we examine the control of a nanomechanical resonator in section [ physical ] . the close connection between linear quantum and classical systems allows one to apply other results from classical control theory for linear systems to quantum linear systems .
transfer function techniques have been applied to linear quantum systems by yanagisawa and kimura @xcite , and dhelon and james have elucidated how the small gain theorem can be applied to linear quantum optical networks @xcite .
finally , it is possible to obtain exact results for the control of linear quantum systems for at least one case beyond lqg theory : james has extended the theory of risk - sensitive control to linear quantum systems @xcite . for nonlinear quantum systems ,
naturally many of the approaches developed for classical non - linear systems can be expected to be useful .
a few specific applications of methods developed for classical nonlinear systems have been explored to date .
one example is the use of linearization to obtain control algorithms @xcite , and another is the application of a classical guidance algorithm to the control of a quantum system @xcite .
a third example is the application of the projection filter technique to obtain approximate filters for continuous state - estimation of nonlinear quantum optical systems
the bellman equation has also been investigated for a two - state quantum system in @xcite ; it is not possible to obtain a general analytic solution to this equation for such a system , and as yet no - one has attempted to solve this problem numerically .
as quantum systems become increasingly important in the development of technologies , no doubt many more techniques and results from non - linear control theory will be applied in such systems .
the objective of lqg control is to use feedback to minimize some quadratic function of the dynamical variables , and this is natural if one wishes to maintain a desired behavior in the presence of noise .
while stabilization and noise reduction in dynamical systems is a very important application of feedback control , it is by no means the only application .
another important class of applications is _ adaptive measurement_. an adaptive measurement is one in which the measurement is altered as information is obtained .
that is , a process of feedback is used to alter the measurement as it proceeds , rather than altering the system the primary distinction between adaptive measurement and more traditional control objectives however is that the goal of the former is usually to optimize some property of the information obtained in the measurement process rather than to control the dynamics of the system .
adaptive measurement has , even at this relatively early stage in the development of the field of quantum feedback control , found many potential applications in quantum systems .
the reason for this is due to the interplay of the following things .
the first is that unlike classical states the majority of quantum states are not fully distinguishable from each other , even in theory , and only carefully chosen measurements will optimally distinguish between a given set of states .
the second is that because quantum measurements generally disturb the system being measured , one must also choose one s measurements very carefully in order to extract the maximal information about a given quantity ( lest the measurement disturb this quantity ) . combining these two things
with the fact that one is usually limited in the kinds of measurements one can perform , due to the available physical interactions between a given system and measuring devices , it is frequently impossible to implement optimal measurements in quantum systems .
the use of adaptive measurement increases the range of possible measurements one can make in a given physical situation , and in some cases allows optimal measurements to be constructed where they could not be otherwise .
as far as the author is aware , the first application of quantum adaptive measurement was introduced by dolinar in 1973 @xcite . here
the problem involves communicating with a laser beam , where each bit is encoded by the presence or absence of a pulse of laser light .
quantum effects become important when the average number of photons in each pulse is small ( e.g. @xmath29 ) .
in fact , it is not possible to completely distinguish between the presence or absence of a pulse .
the reason for this is that the quantum nature of the pulse of laser light is such that there is always a finite probability that there are no photons in the pulse .
it turns out that the optimal way of distinguishing the two states is by mixing the pulse with another laser beam at a beam splitter , and detecting the resulting combined beam . in this case
both input states will produce photon clicks .
the optimal procedure is to vary the amplitude of the mixing beam with time , and in particular to use a process of feedback to change this amplitude after the detection of each photon .
this feedback procedure distinguishes the states maximally well within the limits imposed by quantum mechanics ( referred to as the helstrom bound @xcite ) .
the objective of dolinar s adaptive measurement scheme is to discriminate maximally _ well _ between two states .
alternatively one may wish to discriminate maximally _
fast_. to put this another way , one may wish to maximize the amount of information which is obtained in a specified time , even if it is not possible to obtain all the information in that time .
such considerations can be potentially useful in optimizing information transmission rates when the time taken to prepare the states is significant . in @xcite it
is shown that adaptive measurement can be used to increase the speed of state - discrimination .
of particular interest in quantum control theory are situations which reveal differences between measurements on quantum systems and those on their classical counterparts .
the rapid - discrimination adaptive measurement scheme of @xcite is one such example .
the reason for this is that in the case considered in @xcite it is only possible to use an adaptive algorithm to increase the speed of discrimination if quantum mechanics forbids perfect discrimination between them . since all classical states are completely distinguishable ( at least in theory ) ,
this adaptive measurement is only applicable to quantum systems .
the question of whether or not there are more general situations which provide classical analogues of this adaptive measurement is an open question , however .
the problem of quantum state - discrimination is a special case of _
parameter estimation_. in parameter estimation , the possible states that a system could have been prepared in ( or alternatively the possible hamiltonians that may describe the dynamics of a system ) are parametrized in some way . the observer then tries to determine the value of the parameter by measuring the system .
when discriminating two states the parameter has only two discrete values : in the above case it is the amplitude of the laser pulse , which is either zero or non - zero . in a more general case one wishes to determine a continuous parameter .
an example of this is the detection of a force . in this case
a simple system which feels the force , such as a quantum harmonic oscillator , is monitored , and the force is determined from the observed dynamics .
two examples of this are the atomic force microscope ( afm ) @xcite and the detection of gravitational waves @xcite .
adaptive measurement is useful in parameter estimation because measurements which are not precisely tailored will disturb the system so as to degrade information about the parameter . to the author s knowledge
no - one has yet investigated adaptive measurement in force estimation , although the subject is discussed briefly in @xcite .
however , the use of adaptive measurement for the estimation of a magnetic field using a cold atomic cloud has been investigated in @xcite , and for the estimation of a phase shift imparted on a light beam has been explored in references @xcite . in @xcite the authors show that adaptive measurement outperforms other known kinds of measurements for estimation of phase shifts on continuous beams of light .
it turns out that there are certain kinds of measurements which it is very difficult to make because the necessary interactions between the measurement device and the system are not easily engineered .
one such example in quantum mechanics is a measurement of what is referred to as the `` pegg - barnett '' or `` canonical '' phase @xcite .
there are subtleties in defining what one means by the phase of a quantum mechanical light beam ( or , more precisely , in obtaining a definition with all the desired properties ) .
astonishingly , the question was not resolved until 1988 , when pegg and barnett constructed a definition which has the desired properties for all practical purposes .
this is called the _ canonical phase _ of a light beam .
now , the most practical method of measuring light is to use a photon counter .
however , it turns out that it is not possible to use a photon counter , even indirectly , to make a measurement which measures precisely canonical phase nevetheless , in @xcite ( see also @xcite ) wiseman showed that the use of an adaptive measurement process allows one to more closely approximate a canonical phase measurement .
for a light pulse which has at most one photon , this adaptive phase measurement measures precisely canonical phase .
wiseman s adaptive phase measurement has now been realized experimentally @xcite .
one application of feedback control is in preparing quantum systems in well - defined states . due to noise from the environment ,
the state of quantum systems which have been left to their own devices for an appreciable time will contain considerable uncertainty .
many quantum devices , for example those proposed for information processing @xcite , require that the quantum system be prepared in a state which is specified to high precision . naturally one can prepare such states by using measurement followed by control - that is , using a measurement to determine the state to high precision , and then applying a control field to move the system to the desired point in phase space .
since the rate at which a given measuring device will extract information is always finite , one can ask whether it is possible to increase the rate at which information is extracted by using a process of adaptive measurement .
that is , to adaptively change the measurement as the observer s state - of - knowledge changes , so that the uncertainty ( e.g. the entropy ) of the observer s state - of - knowledge reduces faster .
we will refer to the process of reducing the entropy as _ purifying _ the qubit ( a term taken from the jargon of quantum mechanics ) .
it turns out that it is indeed possible to increase the rate of purification using adaptive measurement .
for a two - state quantum system ( often referred to as a _ qubit _ ) , an adaptive measurement is presented in @xcite that will speed - up the rate of purification in a certain sense .
specifically , it will increase the rate of reduction of the _ average entropy _ of the qubit ( where the average is taken over the possible realizations of the measurement - e.g. over measurements on many identical qubits ) by a factor of two .
this adaptive measurement has two further interesting properties .
the first is that there is no analogous adaptive scheme for the equivalent measurement on a classic two - state system ( a single bit ) ; the speed - up in the reduction of the average entropy is a quantum mechanical effect .
the second property has to do with the statistics of the entropy reduction . for a fixed measurement , while the entropy decreases with time _ on average _ , on any given realization of the measurement the entropy fluctuates randomly as measurement proceeds . for the adaptive measurement however , in the limit of strong feedback ,
the entropy reduces _
deterministically_. for a finite feedback force , there will always remain some residual stochasticity in the entropy reduction , but this will be reduced over that for the fixed measurement @xcite . thus if one is preparing many qubits in parallel , this adaptive measurement will reduce the spread in the time it takes the qubits to be prepared .
the above adaptive algorithm does not speed up purification in every sense of the word , however : wiseman and ralph @xcite have recently shown that while the above adaptive measurement reduces the _ average entropy _ of an ensemble of qubits more quickly , quite surprisingly it actually _ increases _ the _ average time _
it takes to prepare a given value of the entropy by a factor of two ! in this sense , therefore , the above measurement strategy does not , in fact , speed up preparation .
the reason for this is that when one considers an ensemble of spins , the majority purify quickly , whereas the average value of the entropy across the ensemble is increased considerably by a small number of straggling qubits that purify slowly .
thus , the adaptive measurement works by decreasing the time taken for the stragglers , but _ increasing _ the time taken for the majority , so that , for strong feedback , all qubits take the same time to reach a given purity .
thus the feedback algorithm constructed in @xcite does not speed up the average time a given qubit will take to reach a target entropy , and this will be the important quantity if one is preparing qubits in sequence .
application of the above results to rapid purification of superconducting qubits is analyzed in @xcite .
is it possible , therefore , to use adaptive measurement to speed up the average preparation time ? while the answer for a single qubit is almost certainly no , the answer is almost certainly yes in general : in @xcite the authors show that , for a quantum system with @xmath30 states it is possible to speed up the rate at which the average entropy is reduced by a factor proportional to @xmath30 . while it is not shown directly that this algorithm also decreases the _ average time _ to reach a given target purity , it is fairly clear that this will be the case , although more work remains to be done before all the answers are in .
another application of feedback control involving purification of states has been explored in @xcite . in this case , rather than the rate of purification , the authors are concerned with using feedback control during the measurement to obtain a specific final state , and in particular when the control fields are restricted . in this section
we have been considering the application of feedback in quantum systems to problems which lie outside the traditional applications of noise reduction and stabilization .
most of these fall under the category of adaptive measurement , and these we have discussed above .
one that does not is quantum error correction .
the goal of such a process is noise reduction , but with the twist that the state of the system must be encoded in such a way that the controller does not disturb the information in the system .
while we will not discuss this further here , the interested reader is referred to @xcite and references therein .
we now present examples of feedback control applied to two specific quantum systems .
the first is a nano - mechanical resonator , and the second is a single atom trapped in an optical cavity . in both cases
the goal of the feedback algorithm will be to reduce the entropy of the system and prepare it in its ground state .
a nano - mechanical resonator is a thin , fairly ridged bridge , perhaps 200 nm wide and a few microns long .
such a bridge is formed on a layered wafer by etching out the layer beneath .
if one places a conducting strip along the bridge and passes a current through it , the bridge can be made to vibrate like a guitar string by driving it with a magnetic field .
so long as the amplitude of the oscillation is relatively small , the dynamics is essentially that of a harmonic oscillator @xcite .
one of the primary goals of research in this area is to observe quantum behavior in these oscillators @xcite .
the first step in such a process is to reduce the thermal noise which the oscillator is subject to in order to bring it close to its ground state .
typical nanomechanical resonators have frequencies of the order of tens of megahertz .
this means that to cool the resonator so that its average energy corresponds to its first excited state requires a temperature of a few millikelvin .
dilution refrigerators can obtain temperatures of a few hundred millikelvin , but to reduce the temperature further requires something else . in reference
@xcite the authors show that , at least in theory , feedback control could be used to obtain the required temperatures . .
on the near side of the resonator is a single electron transistor ( set ) , whose central island is formed by the two junctions marked `` j '' . on the far side is a t - shaped electrode or `` gate '' .
the voltage on this gate is varied and this results in a varying force on the resonator .
image courtesy of keith schwab.,width=268 ] to perform feedback control one must have a means of monitoring the position of the resonator and applying a feedback force .
the position can be monitored using a single electron transistor , and this has recently been achieved experimentally @xcite .
a feedback force can be applied by varying the voltage on a gate placed adjacent to the resonator .
this configuration is depicted in figure [ fig2 ] .
since the oscillator is harmonic , classical lqg theory can be used to obtain an optimal feedback algorithm for minimizing the energy of the resonator , so long as one takes into account the quantum back - action noise caused by the measurement as described in @xcite .
the details involved in obtaining the optimal feedback algorithm and calculating the optimal measurement strength are given in @xcite .
it has further been shown in @xcite that adaptive measurement and feedback can be used to prepare the resonator in a squeezed state .
the second example of feedback control we consider is that of cooling an atom trapped in an optical cavity .
an optical cavity consists of two parallel mirrors with a single laser beam bouncing back and forward between them .
the laser beam forms a standing wave between the mirrors , and if the laser frequency is chosen appropriately , a single atom inside the cavity will feel a sinusoidal potential due its interaction with the standing wave . it is therefore possible to trap an atom in one of the wells of this potential .
it turns out that information regarding the position of the atom can be obtained by monitoring the phase of the light which leaks out one of the mirrors .
specifically , the phase of the output light tells the observer how far up the side of a potential well the atom is .
in addition , by changing the intensity of the laser beam that is driving the cavity , one changes the height of the standing wave , and thus the height of the potential wells . in this system
we therefore have a means to monitor the atom and to apply a feedback force . in @xcite
the authors present a feedback algorithm which can be used to cool the atom to its ground state . actually , the algorithm will prepare the atom either in its ground state , or its first excited state , each with a probability of @xmath31 .
however , from the measurement record the observer know which one , and can take appropriate action if the resulting state is not the desired one .
if the location of the atom was known very accurately , then we could use the following feedback algorithm to reduce its energy : increase the height of the potential when the atom is climbing up the side of a well , and reduce it when the atom is falling down towards the centre . this way the energy of the atom is reduced on each oscillation , and the atom will eventually be stationary at the centre of the well . however , it turns out that this algorithm is not effective , either classically or quantum mechanically , when the variance of atom in phase space is appreciable .
the reason is that the cyclic process of raising and lowering the potential , which reduces the energy of the atom s mean position and momentum , actually increases the variance of the phase - space probability density .
classically we can eliminate this problem by observing the atom with sufficient accuracy , but quantum mechanically heisenberg s uncertainty relation prevents us from reducing the variance sufficiently . as a result , an alternative algorithm is required . if turns out that one can obtain an effective cooling algorithm by calculating the derivative of the total motional energy of the atom with respect to changes in the height of the potential . in doing
so one finds that the energy change is maximal and minimal at a certain points in the oscillatory motion of the atom . as a result
, one can use a bang - bang algorithm to switch the potential high when the energy reduction is maximal , and switch it low when the resulting energy increase is minimal , in a similar fashion to the classical algorithm described above .
the result is that the atom will lose motional energy on each cycle . the curious effect whereby the atom will cool to the ground state only half the time is due to the symmetry of the system , and the fact that the feedback algorithm respects this symmetry .
specifically , the feedback process can not change the average parity of the initial probability density .
since the ground state has even parity , and the first excited state odd parity , if the initial density of the atom has no particular parity ( a reasonable assumption ) , then to preserve this on average the process must pick even and odd final states equally often . full details regarding the feedback algorithm and the resulting dynamics of the atom is given in @xcite .
although we do not have the space to describe them here , applications of feedback control have been proposed in a variety of other quantum systems .
three of these are cooling the motion of a cavity mirror by modulating the light in the cavity @xcite , controlling the motion of quantum - dot qubits @xcite , and preparing spin - squeezed states in atomic clouds @xcite .
in addition , feedback control has now been experimentally demonstrated in a number of quantum systems .
namely in optics @xcite , cold atom clouds @xcite , and trapped ions @xcite .
to summarize , feedback control has a wide variety of applications in quantum systems , particularly in the areas of noise reduction , stabilization , cooling and precision measurement .
such applications can be expected to grow more numerous as quantum systems become important as the basis of new technologies .
while exact results from modern control theory can be used to obtain optimal control algorithms for some quantum systems , this is not true for the majority of quantum control problems due to their inherent non - linearity . as a result many techniques developed for the control of nonlinear classical systems are of considerable use in designing algorithms for quantum feedback control , and the efficacy of many of these techniques still remain to be explored in quantum systems .
hopefully as further quantum control problems arise in specific systems , and effective control algorithms are developed , rules of thumb will emerge for the control of classes of quantum devices .
this work was supported by the hearne institute for theoretical physics , the national security agency , the army research office and the disruptive technologies office .
v. p. belavkin , `` non - demolition measurement and control in quantum dynamical systems , '' in _ information , complexity and control in quantum physics _ , a. blaquiere , s. diner , and g. lochak , eds.1em plus 0.5em minus 0.4emspringer - verlag , new york , 1987 .
v. b. braginsky , m. l. gorodetsky , f. y. khalili , a. b. matsko , k. s. thorne , and s. p. vyatchanin , `` noise in gravitational - wave detectors and other classical - force measurements is not influenced by test - mass quantization , '' _ phys .
d _ , vol .
67 , no . 8 , p. 082001
, 2003 .
d. t. pope , h. m. wiseman , and n. k. langford , `` adaptive phase estimation is more accurate than nonadaptive phase estimation for continuous beams of light , '' _ physical review a _ , vol .
70 , p. 043812
, 2004 .
j. f. ralph , e. j. griffith , and c. d. h. a. d. clark , `` rapid purification of a solid state charge qubit , '' e. j. donkor , a. r. pirich , and h. e. brandt , eds .
6244.1em plus 0.5em minus 0.4emspie , 2006 , p. ( in press ) .
w. p. smith , j. e. reiner , l. a. orozco , s. kuhr , and h. m. wiseman , `` capture and release of a conditional state of a cavity qed system by quantum feedback , '' _ phys .
_ , vol . 89 , p. 133601
, 2002 .
p. bushev , d. rotter , a. wilson , f. dubin , c. becher , j. eschner , r. blatt , v. steixner , p. rabl , and p. zoller , `` feedback cooling of a single trapped ion , '' _ phys .
_ , vol .
96 , p. 043003 | we give an introduction to feedback control in quantum systems , as well as an overview of the variety of applications which have been explored to date .
this introductory review is aimed primarily at control theorists unfamiliar with quantum mechanics , but should also be useful to quantum physicists interested in applications of feedback control .
we explain how feedback in quantum systems differs from that in traditional classical systems , and how in certain cases the results from modern optimal control theory can be applied directly to quantum systems .
in addition to noise reduction and stabilization , an important application of feedback in quantum systems is adaptive measurement , and we discuss the various applications of adaptive measurements .
we finish by describing specific examples of the application of feedback control to cooling and state - preparation in nano - electro - mechanical systems and single trapped atoms . |
t tauri stars are thought to be young stars that are still accreting gas through a disk . @xcite
discovered t tauri star kh 15d ( k7 v , @xmath2 pc ) , which shows an eclipse for 1/3 of its period , implying that it is obscured by faint circumstellar material rather than a stellar companion ; earlier data showed the star rebrighten at mid - eclipse , although this has weakened in recent data @xcite .
this indicates that the obscuring material subtends a large angle , which may herald the presence a low mass stellar companion to shepherd the material via resonant interaction or to create a distorted disk @xcite .
in addition , the duration of the eclipse trough is lengthening over time , changing from 16 days to 18 days over 5 years , and data from 50 - 90 years ago showed no evidence for eclipses @xcite , which means that the obscuring material is rapidly evolving . during eclipse , residual flux is still present at a level of @xmath31/20 of the flux outside of eclipse indicating that either ( a ) the obscuring material is porous ( covering factor is @xmath4 ) , ( b ) the eclipse is total , but scattered light `` fills in '' the eclipse , creating some residual flux , ( c ) some flux is scattered and some transmitted . the scattering interpretation may be favored as the star becomes slightly bluer during eclipse as one would expect from scattered light , although the fluctuations in color are comparable to the color difference @xcite . in addition , the equivalent width of h@xmath5 increases dramatically from 2 to @xmath6 indicating a change from an unobscured `` weak - line '' t tauri star to a classical t tauri star ! the eclipse provides a natural `` coronagraph '' during which the suppression of stellar light allows any scattering by surrounding material to become more prominent .
we attempted to distinguish the above possibilities by carrying out spectropolarimetry inside and outside of the eclipse .
since the unpolarized flux of the star dilutes the percentage of polarization , we expected that the percentage polarization due to scattered light would increase by a factor of at least 20 during the eclipse if the scattering material lies exterior to the occulting material and if the starlight is intrinsically unpolarized .
this scenario is analogous to type 1 active galactic nuclei ( agn ) which are unobscured and show low polarization versus type 2 agn which are obscured and show much higher polarization @xcite .
some t tauri stars show an anti - correlation between total flux and polarization ( e.g. * ? ? ?
* ; * ? ? ?
* ) , which may be exaggerated in kh 15d as the flux variations are much more extreme . on the other hand ,
if the residual flux is only due to incomplete coverage by the occulting material , then the polarization will not change .
interstellar foreground polarization , which is measured to be 0.10.4% in the v - band for ngc 2264 @xcite , will not change during eclipse . before our observations , the optical polarization of kh 15d had not yet been measured , but during eclipse it shows the characteristics of classical t tauri stars which have a mean polarization of 1.2% with a standard deviation of 0.8% and maximum polarization of 2.5% @xcite .
in 2 we summarize the spectropolarimetric observations , in 3 we give the results of the observations , in 4 we discuss the implications of these results for models of this t tauri star , and in 5 we review our conclusions .
the primary observational goal was to measure the difference in polarization between eclipse and non - eclipse .
outside eclipse , the star is sufficiently bright to observe at the palomar 200-inch telescope ( @xmath7 ) , but during the eclipse the star is a factor of @xmath8 fainter which required using keck to obtain a similar number of photons in a reasonable observing time .
the palomar observations were taken with the double spectrograph spectropolarimeter @xcite , while the keck i observations were taken with the low resolution imaging spectrograph polarimeter @xcite . both instruments
add dual - beam polarimetry optics to a double spectrograph incorporating dichroics to separate the blue and red beams .
the observations and reductions were done using standard spectropolarimetry techniques @xcite .
the observations of kh 15d were performed in sequences of four 15-minute exposures at four different positions of a rotating half - wave plate .
each four - exposure sequence allows measurement of the stokes parameters @xmath9 and @xmath10 . for each hour of observations , the position angle of the spectrograph slit
was aligned with the parallactic angle for the midpoint of the hour .
the instrumental setup is described in table 1 , while the observational summary is given in table 2 .
null , polarized , and flux standards were observed during each night . since instrumental polarization can be present in the palomar 200-inch @xcite , we observed four null standards each at two perpendicular slit position angles .
we used the measured polarization of the null standards as an estimate of the mean and uncertainty of the instrumental polarization , and we subtracted the mean instrumental stokes parameters @xmath9 and @xmath10 from the measured stokes parameters of kh 15d to correct for the instrumental polarization . at palomar
the instrumental polarization was found to be @xmath11% over @xmath12 and @xmath13% over @xmath14 . at keck
the instrumental polarization was negligible , @xmath15% .
at palomar the observations were obtained through clouds , while the keck nights were mostly clear , but not photometric .
ccc dbsp , red & 5620 - 8120 & 2.45 dbsp , blue & 3350 - 5750 & 3.38 lrisp , red & 3200 - 5850 & 2.0 lrisp , blue & 5650 - 9400 & 2.0 llccl 2003 jan 25 & palomar & 3 & 2@xmath160 & partly cloudy2003 feb 23 & keck & 3 & 1@xmath165 & partly cloudy2003 feb 24 & keck & 1 & 1@xmath160 & clear2003 feb 25 & keck & 2 & 1@xmath165 & clear a 7th magnitude b2iii star , hd 47887 , is located 39@xmath17 away from kh 15d . in the keck observations a diffraction spike from this bright star periodically swept through the slit , passing within several arcseconds of kh 15d in several exposures . in these exposures
sky subtraction was difficult , making it impossible to accurately measure the polarization and also adversely affecting the color of the extracted spectra for two of the six hours .
figure 1 depicts the phase of our observations with respect to the 48.36 day period of kh 15d .
[ lightcurve ]
the continuum polarization in the keck and palomar data was measured over wavelength bins of width 1000 .
we averaged the 3 measurements from palomar and subtracted off the mean polarization measured from observations of 4 null - standard stars .
we computed the mean of 4 measurements from keck and the error bars were computed from the standard deviation of these measurements plus the statistical error of the individual measurements .
the palomar data show a small enough polarization and large enough errors that the data are consistent with being unpolarized at the @xmath3 1 @xmath18 level ( see figure 2 ) .
however , the keck data showed a much higher polarization , and are inconsistent with the palomar data at a high significance .
tables 3 and 4 summarize the polarization measurements of the eclipsed and uneclipsed data .
the eclipsed data show a slight increase in polarization from the blue ( 4600 - 5600 ) to the red ( 7000 - 8000 ) at 3 @xmath18 significance .
most of this change is due to an increase in the @xmath9 stokes parameter ( figure 2 ) , while @xmath10 remains constant ; however , the errors are large enough that a change in the polarization angle is only significant at 1.5 @xmath18 .
lccc 46005600 & [email protected] & 31@xmath1927 & [email protected] & [email protected]&-19@xmath1935 & 170008000 & [email protected]&-22@xmath1941 & [email protected] lcc 40004600 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] figure 3 shows the total flux spectrum in and out of eclipse .
the spectral shape appears nearly identical in and out of eclipse , despite the change in total flux of @xmath20 magnitudes .
we measured the mean of the ratio of the uneclipsed to the eclipsed data in each of the three bands ( 46005600 , 6000 - 7000 , and 7000 - 8000 ) , excluding the emission lines , and computed the standard deviation of the ratio within each band , which is reported in table 3 .
the flux ratios differ between the bluest ( 4600 - 5600 ) and reddest ( 7000 - 8000 ) bands by only 1 @xmath18 , which is statistically consistent with no color change .
the absolute value of the uneclipsed to eclipsed flux ratios are not to be trusted as the nights were not photometric , so we report only the ratio of each band to the 60007000 band . for comparison with the photometric data of @xcite we also computed synthetic photometry of our spectra which gave @xmath21 mag and @xmath22 mag during eclipse and @xmath23 mag outside of eclipse , which are in agreement with the range of values reported by @xcite .
we estimate the uncertainty on these colors to be @xmath24 mag as there appear to be differences between individual spectra on scales of @xmath25 which may be due to fluctuations in the flat - fields , difficulty of sky subtraction , overall systematic uncertainties in flux calibration , or other instrumental variations .
there are a large number of stellar photospheric features which are seen in both spectra and are indicative of a k6v spectral type with @xmath26 k , somewhat hotter than the k7v type estimated by @xcite which may be due to the fact that our spectrum covers bluer wavelengths .
almost all of these features show the same equivalent widths in both spectra , indicating that the eclipsed spectrum is a scaled version of the uneclipsed spectrum . we have measured the lick indices @xcite for both the eclipsed and uneclipsed spectra , plotted in figure 4 .
the h@xmath27 indices are contaminated by the [ sii ] emission line in the keck data .
the na d equivalent width is smaller during the eclipse , which we attribute to unobscured na d emission filling in the absorption line .
the measured indices for kh 15d agree best with those of a k6v star ; the agreement is good enough that there is no evidence for optical veiling .
the h@xmath5 emission line has a much higher flux in the eclipsed spectrum than in the uneclipsed spectrum by a factor of @xmath28 ( since our data were not photometric , this factor is probably @xmath29 )
. this may be partly due to variability and partly due to partial eclipse of the h@xmath5 line as @xcite found that the flux of h@xmath5 could decrease or increase during eclipse and the shape of h@xmath5 varied significantly as well .
they measured an equivalent width of 2 outside eclipse and 40 - 70 in eclipse , compared with our measurements of 3 and 30 , respectively .
in addition , four emission lines appear in the eclipsed spectrum that are not apparent in the uneclipsed spectrum .
the fluxes and equivalent widths of the emission lines are given in table 5 , which we have computed by subtracting off a continuum template , with errors estimated from the standard deviation of the residuals .
included is the na d absorption line at @xmath30 which is stronger in the uneclipsed spectrum than in the eclipsed spectrum ( we have used the uneclipsed spectrum as the continuum template for this line ) .
lcc ( ) & ( @xmath31 erg @xmath32 s@xmath33 & ( ) @xmath34 6563 & 5.88 @xmath190.71 & [email protected] @xmath34 6563 & 1.74 @xmath190.19 & [email protected] @xmath34 4069 & [email protected] & [email protected] @xmath34 5890 & [email protected] & [email protected]@xmath35 6716,6731 & [email protected] & [email protected]@xmath34 6300 & [email protected] & [email protected] @xmath34 6364 & [email protected] & [email protected]
the variation of the polarization between eclipse and non - eclipse indicates that the _ scattering region is not completely obscured by the occulting material_. if the occulter completely blocks all the light from the star , if the scattering is nearly isotropic , and if the scattering region is nearly spherically symmetric , then the residual flux , @xmath36% , indicates that the optical depth of the unobscured scattering material is @xmath37 .
since the spectra inside and outside eclipse are consistent with no change in the spectral shape , and since the spectrum within eclipse is consistent with very weak wavelength dependence of the polarization , we conclude that the scattering mechanism is nearly achromatic , implying that the scatterers may be electrons or large grains .
we discuss each of these possibilities in turn .
if the scattering mechanism were electrons , then the low measured polarization of 2% indicates that the scattering region need be only slightly asymmetric .
for example , in the model of @xcite , the polarization caused by an axisymmetric thomson scattering envelope is @xmath38 , where @xmath5 is a `` shape factor '' which depends on the precise geometry of the scatterers .
since the eclipsing region is likely in the equatorial plane of the circumstellar material , we can assume that @xmath39 .
thus , @xmath40 for @xmath41% . for a scattering geometry which consists of a slightly oblate or prolate spheroid of axis ratio @xmath42 ,
then @xmath43 , so @xmath44 to create the observed polarization , i.e. only a 20% deviation from spherical symmetry is required to give the measured polarization . note that this example is simply illustrative since we have no preferred model for the scattering region .
if we assume that all of the light during eclipse is due to electron scattering , then we can estimate the bound - free continuum emission expected from this gas .
consider a nearly symmetric region of uniform electron density and temperature surrounding the star of size @xmath45 .
we assume that the eclipsing material is located near the keplerian radius , @xmath46 , for a 48 day period .
the low polarization implies that the electron scattering region must be at a larger radius , @xmath47 au to avoid becoming too asymmetric due to obscuration .
since the material is nearly symmetric , the mean optical depth of the scattering region is @xmath48 , which implies an electron number density of @xmath49 @xmath50 ( assuming a uniform density ) .
the bound - free emission expected at , say , 6000 must be less than @xmath51 erg @xmath32 s@xmath52 to avoid changing the strength of the lick indices . computing the emission from an optically - thin spherical region of plasma in lte , we find that @xmath53 k to be consistent with this upper limit . note that this is a strong lower limit since it is likely that the scattering region need be larger to give the small observed polarization .
in addition , one can place a constraint on the temperature based on the strength of balmer line emission ; however , the interpretation is complicated by the unknown optical depth of the balmer lines .
even so , the lower limit on the temperature based on the lack of detected bound - free emission is four times the virial temperature at @xmath46 , meaning that this gas would be strongly unbound and will outflow in less than a dynamical timescale , implying a mass loss rate of @xmath54 yr @xmath55 , which is rather large for a weak - line t tauri star .
thus , it is more likely that the scattering is not due to electrons , but may be due to large dust grains , which we discuss next .
if the scattering were due to dust , then the grain sizes would have to be fairly large to allow the weak wavelength dependence of the polarization _ and _ flux .
this is true for the zodiacal light which reflects an optical spectrum that is identical to that of the sun @xcite .
based on the achromaticity of the zodiacal scattering , the scatterers are thought to be large dust grains , 10 - 100 @xmath56 m in size , and there is further evidence that these grains might be quite `` fluffy '' ( i.e. low filling factor ) with densities as low as @xmath57 g @xmath50 . for the estimation of the wavelength dependence of the scattering and polarization by dust particles we approximate the particles as spheres ( mie scattering ) and compute the polarized differential scattering cross section by single particles using a standard approach @xcite.swolf / miex - web / miex.htm - or contact the authors .
] we have chosen the distribution of refractive indices of the materials from @xcite : 62.5% astronomical silicate and 37.5% graphite .
since graphite is crystalline , it was considered as two `` different '' materials in the computation : the first with a refractive index as measured perpendicular to the crystal rotation axis , 25% , and the second parallel , 12.5% .
we assumed that distribution of the dust size , @xmath58 , follows a power - law @xmath59 for @xmath60 , and derived the scattering properties of the ensemble of particle species and sizes ( * ? ? ? * wolf & voshchinnikov , in prep . ) .
we averaged the scattered polarized flux over the entire @xmath61 steradians divided by the total scattered flux to estimate the mean polarization as a function of wavelength for each dust model .
we then compared this to the relative wavelength dependence of the measured polarization ( the absolute polarization depends on the exact geometry of the scattering region , which is unknown ; see the next section for one particular model ) .
we also compared the relative wavelength dependence of the scattered flux by dividing the eclipsed spectrum by the uneclipsed spectrum .
given the assumptions that ( a ) the wavelength dependence of the flux and polarization does not depend on the geometry of the scatterers , and ( b ) the scatterers can be described by a single power - law , we can then place a constraint on the parameters of the dust model .
we find that the lower limit , @xmath62 , does not change the fit much , so we simply vary @xmath63m@xmath64 m and @xmath65 .
the minimum @xmath66 is 12 for 5 data points for @xmath67 m and @xmath68 , with the confidence contours plotted in figure 5 . at 90% confidence , @xmath69 and @xmath70m@xmath71 m , which indicates that the scattering is dominated by large dust grains .
unfortunately , the minimum reduced @xmath72 , which is rather large .
this is due to the fact that the measured polarization increases toward longer wavelengths , while the dust scattering models predict a polarization that is nearly constant , or slightly decreasing , with wavelength .
the opacity of large dust grains scales as @xmath73 @xmath74 g@xmath55 , where @xmath75 m .
thus , the eddington limit for the smaller of the large dust grains scales as @xmath76 erg s@xmath55 , which is similar to the luminosity of kh 15d , @xmath77 erg s@xmath55 .
thus , smaller dust grains will be removed from the system by radiation forces . in addition
, radiation forces on the larger grains may be significant enough to modify the structure of the system .
@xcite have proposed a model for the occulter which consists of a semi - infinite edge crossing the star .
this model is attractive as the finite size of the star and limb darkening can account for the shape of the ingress and egress , and , if the occulter is at the keplerian radius , then it must be inclined to the orbital plane by 15@xmath78 to give the correct duration of ingress and egress .
however , we have found that the ingress and egress of the eclipse ( figure 1 , near phase of 0.3 and 0.7 ) show a more gradual rolloff than expected for a semi - infinite occulter .
this can be fit with an optical depth of the occulter which scales with position as @xmath79 , where @xmath80 is the coordinate perpendicular to the edge of the occulter , which gives an excellent fit to the phased lightcurve data ( solid line in figure 1 ) .
this material causing extinction will scatter as well , and in the case of large grains there will be significant forward scattering @xcite .
in fact , this may account for the scattered flux present during the eclipse ; however , to compute the scattered light requires a knowledge of the three - dimensional distribution of scattering and occulting material , as well as a knowledge of the scattering properties .
we construct such a model by assuming that ( a ) the occulter is located at the keplerian radius and ( b ) the 15@xmath78 tilt of the occulter is due to a warped disk .
we model the material as a succession of tilted circular annuli , creating a warped disk , with the tilt angle given by @xmath81},\ ] ] where @xmath82 is the radius of maximum tilt , @xmath83 is the maximum tilt , @xmath84 is the radius in the disk , and @xmath85 is the width of the tilted region , and we assume that the nodes of the warped annuli are aligned .
the observer we place at an angle @xmath86 to the equatorial plane , and we rotate the warped disk as a solid body with a period equal to that of kh 15d . in addition , we cut the disk off at @xmath87 and @xmath88 .
the opacity of the scattering material we distribute as @xmath89 ^ 2 $ ] , to produce the observed roll - off in intensity , where @xmath90 is the angular scale height , @xmath91 is the polar angle , and @xmath92 is a fiducial opacity . we find a good fit to the lightcurve for @xmath93 , @xmath94 , @xmath95 , @xmath96 au ( this is the keplerian radius ) , @xmath97 .
when taking into account only extinction , this distribution produces the solid lightcurve in figure 1 after the addition of 4.5% of the uneclipsed flux during eclipse .
however , part of the extinction is due to scattering , so we also have computed the singly - scattered component of this density distribution assuming an albedo of unity and using the scattering function for zodiacal light derived by @xcite which is strongly forward scattering and has a maximum polarization of 22% .
we find that the polarized flux of the scattered light during eclipse is 0.1% of the unobscured total flux , which is the same as the observed polarized flux during eclipse of kh 15d ( figure 6 ) .
figure 7 shows the appearance of the scattered light in this model at mid - eclipse ( an animated version of this figure which shows the appearance as a function of phase is available on - line ) .
one prediction of this model is that the polarization angle should vary by @xmath98 during the eclipse .
however , the total model flux during mid - eclipse is only 2.5% of the uneclipsed flux which is about half the observed mid - eclipse flux , indicating that there may be another component of scattered light which has low polarization , such as diffracted light , which is not taken into account in the model of @xcite , or back - scattered light . in a future study
we may carry out more sophisticated modeling , varying the radius of the warp as well as the symmetry of the warp , and including multiple scattering and varying the dust grain model , to see how the scattered flux varies with dust composition and with wavelength of the observed light , and we will see whether diffraction , multiple - scattering , or back - scattering can explain the brightening at mid eclipse . there is evidence that the ingress and egress differ slightly in time between the even and odd eclipses , which may imply that the warp is an @xmath99 mode rather than @xmath100 as in our model .
if so , the back warp will be partly eclipsed by the front warp , while the edges will be visible during mid - eclipse , which may explain the rebrightening .
alternatively , the symmetry during mid - eclipse of the front warp may cause a brightening due to diffraction , which is likely to be wavelength dependent .
if the albedo is less than unity , then mid - infrared observations may point toward an @xmath99 configuration if they also show a peak at mid - eclipse since the infrared light is emitted isotropically , similar to the back - scattered light .
a summary of the results : \1 ) near mid - eclipse the polarization is @xmath32% , and is consistent with zero outside of eclipse ( [email protected]% ) .
combined with other photometric data , the polarized flux near mid - eclipse is @xmath24% of the flux outside of eclipse .
this indicates that there is an unobscured scattering region surrounding the t tauri star which creates the polarization ; the polarization can not be solely due to foreground transmission effects ( not associated with the star ) as there would no variation .
\2 ) the data are nearly consistent with no wavelength dependence of the polarization during eclipse .
this indicates that the scatterers are approximately grey .
in addition , electron scattering is unlikely to cause the observed polarization .
\3 ) the spectrum within the eclipse is consistent with being a scaled version of the spectrum outside of eclipse with no wavelength dependence of the ratio except for the presence of stronger emission lines which also argues for grey scattering ; rayleigh scattering is ruled out .
our mie dust scattering model implies @xmath101 and @xmath70m@xmath71 m at 90% confidence ; however , the best fit model is not an adequate fit .
\4 ) the qualitative features of the eclipse can be reproduced by a model of a warped disk that orbits in front of the star , periodically eclipsing the star , having an extended atmosphere which causes a decrease in light before and after eclipse as well as scattered light within eclipse , which is consistent with the observed polarized flux .
the polarization adds an extra clue that the eclipsing material completely occults the star since the maximum observed polarization of classical t tauri stars is @xmath102% @xcite and the kh 15d spectrum looks like that of a classical t tauri star during eclipse .
this adds to the evidence that the eclipse is complete based on ( a ) the steep egress and ingress ; ( b ) the flat bottom of the trough of the eclipse ; ( c ) the fine tuning that would be required to arrange 95% covering .
it is interesting that the trough of the eclipse does show some slight variations , namely an increase near mid - eclipse , which may indicate that the scattering region is partially eclipsed and this eclipse varies with time .
it would be interesting to measure the polarization as a function of time within the eclipse to look for variation .
we have modeled the eclipse with a warped disk since a discrete cloud would be sheared in ten orbits , while the eclipses have repeated with similar durations for @xmath3 50 orbits . however , this does not mean that the warped disk model is without its own problems : if the warped model is correct , what drives the warp at such a short period ?
the interaction of the stellar magnetosphere is proposed to cause a similar occulting behavior in the t tauri star aa tauri @xcite ; however , this star has a much short period between occultations , 8 days , while the longer period of kh 15d is much longer than the spin period of any known t tauri star .
one can place a lower limit on the warp radius of @xmath103 since inwards of this radius the warp would have to be at @xmath104 to create the correct duration of ingress due to the finite size of the star . at @xmath105 , the keplerian period is 0.8 days , while a warp precession timescale is typically much longer than the keplerian timescale
. it may be that a planet excites a wave in the disk @xcite or that an elongated dusty vortex causes obscuration @xcite ; each of these models might be consistent with the observed polarization profile , which remains to be quantified with a scattering calculation similar to the one presented here .
future observations of interest would be to ( 1 ) measure the magnitude and angle of polarization throughout ingress and/or egress to test the warp model ; ( 2 ) measure the polarization at higher spectral resolution within the h@xmath5 emission line to see whether it is unpolarized ; ( 3 ) look for variations of the flux during eclipse to determine whether the stellar variability is washed out by the time delays from scattering ( although this requires a rather extended scattering region ) .
in addition , near- and mid - infrared observations may show fluctuations if the occulting region has a warped geometry , as we have suggested .
we thank catrina hamilton , lynne hillenbrand , patrick ogle , and russel white for useful discussions and marshall cohen for exchange of observing time .
we thank bill herbst for providing published photometric data .
some of the data presented herein were obtained at the w.m .
keck observatory , which is operated as a scientific partnership among the california institute of technology , the university of california and the national aeronautics and space administration .
the observatory was made possible by the generous financial support of the w.m .
keck foundation .
the authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of mauna kea has always had within the indigenous hawaiian community .
we are most fortunate to have the opportunity to conduct observations from this mountain .
support for e. a. was provided by the national aeronautics and space administration through chandra postdoctoral fellowship award pf0 - 10013 issued by the chandra x - ray observatory center , which is operated by the smithsonian astrophysical observatory for and on behalf of the national aeronautics space administration under contract nas 8 - 39073 .
s. wolf was supported by nasa through grant nag 5 - 11465 .
antonucci , r. , 1993 , ara&a , 31 , 473 barge , p. & viton , m. , 2003 , apj , 593 , l117 bohren c.f . ,
huffman d.r . , 1983 ,
_ absorption and scattering of light by small particles , _ john wiley & sons , new york bouvier , j. et al .
, 1999 , a&a , 349 , 619 breger , m. & dyck , h. m. , 1972 , apj , 175 , 127 brown , j. c. ; mclean , i. s. , 1977 , a&a , 57 , 141 bryden , g. , r@xmath106yczka , m. , lin , d. n. c. & bodenheimer , p. , 2000
, apj , 540 , 1091 goodrich , r. , 1991 , pasp , 103 , 1314 hamilton , c. m. , herbst , w. , shih , c. & ferro , a. j. , 2001 , apj , 554 , l201 hamilton , c. m. , herbst , w. , mundt , r. , bailer - jones , c. a. l. , johns - krull , c. m. , 2003 , astro - ph/0305477 herbst , w. et al . , 2003 ,
pasp , 114 , 1167 hong , s. s. , 1985 , a&a , 146 , 67 kartje , j. , 1995 , apj , 452 , 565 kearns , k. e. & herbst , w. , 1998 , apj , 116 , 261 leinert , ch .
et al . , 1998 ,
a&as , 127 , 1 martin p.g .
, 1978 , cosmic dust .
its impact on astronomy , clarendon press , oxford mnard , f. , 1991 , in _ iau colloq .
129 _ , ed .
bertout c. , collin - souffrin , s. , & lasota , j. p. ( gif - sur - yvette : editions frontieres ) , p. 59
mnard , f. , bouvier , j. , dougados , c. , melnikov , s. y. , & konstantin , n. g. , 2003 , a&a , in press , astro - ph/0306552 miller , j. s. , robinson , l. b. & goodrich , r. w. , 1988 , in _ instrumentation for ground - based optical astronomy _ ,
l. b. robinson ( new york : springer ) , p. 157
mekkaden , m. v. , 1999 , a&a , 344 , 111 ogle , p. m. , cohen , m. h. , miller , j. s. , tran , h. d. , goodrich , r. w. & martel , a. r. , 1999 , apjs , 125 , 1o oke , j. b. et al . , 1995 ,
pasp , 107 , 375 oke , j. b. & gunn , j. e. , 1982 , pasp , 94 , 586 prugniel , ph . & soubiran , c. , 2001 , a&a , 369 , 1048 van de hulst , h. c. , 1981 , _ light scattering by small particles _
, dover : new york voshchinnikov n.v . , 2002 , astrophys .
space phys .
, 12 , 1 weingartner , j. c. & draine , b. t. , 2001 , apj , 548 , 296 winn , j. n. , garnavich , p. m. , stanek , k. z. , sasselov , d. d. , 2003 , apj , 593 , l121 wiscombe w.j . , 1980 , appl .
optics , 19 , 1505 wolf , s. , 2000 , computer physics communications , 150 , 99 wolf , s. , henning , th . & stecklum b. , 1999 , a&a , 349 , 839 worthey , g. & ottaviani , d. l. , 1997 , apjs , 111 , 377 | kh 15d is a strongly variable t tauri star in the young star cluster ngc 2264 that shows a decrease in flux of 3.5 magnitudes lasting for 18 days and repeating every 48 days .
the eclipsing material is likely due to orbiting dust or rocky bodies in a partial ring or warped disk that periodically occults the star .
we measured the polarized spectrum in and out of eclipse at the keck and palomar observatories .
outside of the eclipse , the star exhibited low polarization consistent with zero . during eclipse ,
the polarization increased dramatically to @xmath0% across the optical spectrum , while the spectrum had the same continuum shape as outside of eclipse and exhibited emission lines of much larger equivalent width , as previously seen . from the data , we conclude that ( a ) the scattering region is uneclipsed ; ( b ) the scattering is nearly achromatic ; ( c ) the star is likely completely eclipsed so that the flux during eclipse is entirely due to scattered light , a conclusion also argued for by the shape of the ingress and egress .
we argue that the scattering is not due to electrons , but may be due to large dust grains of size @xmath1 m , similar to the interplanetary grains which scatter the zodiacal light .
we construct a warped - disk model with an extended dusty atmosphere which reproduces the main features of the lightcurve , namely ( a ) a gradual decrease before ingress due to extinction in the atmosphere ( similar for egress ) ; ( b ) a sharper decrease within ingress due to the optically - thick base of the atmosphere ; ( c ) a polarized flux during eclipse which is 0.1% of the total flux outside of eclipse , which requires no fine - tuning of the model .
the inclination of the warp is set by the duration of ingress and egress assuming that the warp is located at the keplerian radius , and the inclination of the observer is then determined by the duration of the eclipse . |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.