article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
the very metal - poor stars , presently found in the halo of the galaxy , are believed to have formed at the earliest times , shortly after it became possible for the universe to make stars with sufficiently long main - sequence lifetimes ( i.e. , masses @xmath11 ) to survive for @xmath12 gyr .
the chemical compositions of these stars are thus expected to reflect a quite small number of nucleosynthesis processes , possibly as small as one , while the compositions of more metal - rich stars like the sun reflect the cumulative ( hence quite complex ) results of the various processes that have been in operation during the entire history of galactic chemical evolution .
recent abundance analyses for extremely metal - poor stars have provided quite valuable information on the individual nucleosynthesis processes involved ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
in particular , detailed abundance studies of heavy metals in such stars have led to important progress in the understanding of the origin of neutron - capture elements in the early galaxy .
sneden et al .
( 1996 , 2000 , 2003 ) have studied the abundances of the extremely metal - poor star cs 22892052 , the first example of a growing class of metal - poor stars that exhibit very large excesses of r - process elements relative to iron ( [ r - process / fe ] @xmath13 ) .
even from the first analysis , it was apparent that the relative abundance pattern of the heavy neutron - capture elements ( 56 @xmath6 _ z _ @xmath6 76 ) in this star was identical ( within observational errors ) to that of the ( inferred ) solar system r - process component , which has been strengthened as more and better data have been acquired .
this striking similarity may be surprising , considering that the abundance pattern of the solar system has certainly been influenced by the integrated yields from a variety of nucleosynthesis sites .
it should be pointed out that , in contrast , the abundance pattern of the lighter neutron - capture elements ( @xmath14 ) in cs 22892052 exhibit clear deviations from that of the solar - system r - process component @xcite .
recent analyses of a small number of additional very metal - poor stars with moderate excesses of r - process elements ( @xmath15 [ r - process / fe ] @xmath16 ) ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) have obtained similar results , that is , the abundance pattern of heavy neutron - capture elements ( @xmath17 ) for each star is quite similar to that of the r - process component in solar - system material .
these results imply that the neutron - capture elements in these very metal - poor stars were produced by the r - process , which produces quite similar abundance patterns at least for the range of @xmath18 ( schatz et al .
2002 ; wanajo et al .
2002 ; otsuki , mathews , & kajino 2003 ) . by way of contrast , the abundance patterns of the lighter neutron - capture elements ( @xmath19 ) are significantly different from that of the r - process component in the solar system .
this observational fact is interpreted as a result of the existence of ( at least ) two distinct classes of r - process events ( e.g. , * ? ? ?
* and references therein ) .
the search for , and subsequent detailed abundance studies of , r - process - enhanced , very metal - poor stars enable one to make estimates of the ages of these objects by use of the radioactive species that can be identified in them .
the long - lived radioactive r - process elements , in particular th and u , whose half lives ( 14 gyr for @xmath20th ; 4.5 gyr for @xmath21u ) are comparable to or shorter than the cosmic age ( @xmath22 15 gyr ) , provide a powerful tool for determination of a lower limit on the age of the galaxy , and hence of the universe . if th and/or u is detected in a very metal - poor star , we can estimate the lapse of time from the era of the nucleosynthesis process that created these elements to the present .
this is accomplished by measuring the present abundance ratios , either of u / th , when they are both detected ( which only applies to two stars thus far ) , or by measurement of the u or th abundance ratio as compared to stable r - process elements , such as eu , and comparing them to predictions of the initial production ratios from models ( both site - independent , e.g. , schatz et al .
2002 , and site - dependent , e.g. , wanajo et al . 2002 , 2003 ; otsuki , mathews , & kajino 2003 ) . the cosmochronometry technique ( in stars ) was pioneered by butcher ( 1987 ) , who unfortunately only had roughly solar - abundance stars to work with , hence he had to contend with the difficulties arising from complex continua and line - blending problems , which are not so severe for very metal - poor stars .
one clear advantage of the cosmochronometric method is that the resulting age estimate is free from a host of uncertainties encountered by other estimates of the cosmic age , such as calibration of distance scales , detailed understanding of stellar evolution , etc .
furthermore , since it is thought that the large overabundances of r - process elements are likely to have been associated with a single production event , quite likely a type ii supernova explosion in the early galaxy , one is not forced to model the entire complex history of galactic chemical evolution in order to estimate the age .
the difficulties in this method , aside from the rarity of the very metal - poor stars with detectable th and u ( presently estimated to be no more than @xmath23% of giants with [ fe / h ] @xmath24@xmath25 and log@xmath26 , for elements a and b. also , the term `` metallicity '' will be assumed here to be equivalent to the stellar [ fe / h ] value . ] ; beers , private communication ) , arise from the need to accurately predict the initial production ratios , which in turn depend on having a detailed understanding of the nucleosynthesis pathways , nuclear mass models , and cross - sections for species that have not been measured adequately at present ( see schatz et al .
2002 for details ) . in the seminal study by franois , spite , & spite ( 1993 ) ,
th abundances were reported for the first time for stars with metallicities as low as [ fe / h ] @xmath27 2.6 .
sneden et al .
( 1996 ) first succeeded in obtaining a clear detection of th , along with other r - process elements , in cs 22892052 ( see above ) , a relatively bright , extremely metal - poor giant ( [ fe / h ] = 3.1 ) discovered during the hk survey of beers and collaborators ( see beers et al . 1992 ) . from these authors measurement of the th / eu ratio
, the age of this star was estimated to be 15.2 @xmath28 3.7 gyr . in their analysis
, they assumed the initial abundance ratio between th and stable elements to be the same as that of the calculated initial solar - system abundance ratio .
subsequently , westin et al .
( 2000 ) estimated the age of another r - process - enhanced metal - poor star , hd 115444 , to be 14.2 @xmath28 4 gyr from the th / eu ratio .
johnson & bolte ( 2001 ) also measured th abundances for five very metal - poor stars .
most recently , cowan et al .
( 2002 ) studied the moderately r - process - enhanced , very metal - poor star , bd+17@xmath293248 .
this star , with [ fe / h ] = 2.1 , and [ eu / fe ] = + 0.9 , has the distinction of being among the most metal - rich stars in which the r - process - enhancement phenomenon has been observed .
these authors , based on an average of a number of chronometer pairs involving u and th , obtained an age estimate of @xmath30 gyr .
sneden et al .
( 2003 ) have assembled a definitive high - s / n set of spectra for cs 22892052 , drawing on space - based and ground - based observations .
their analysis has led to a revision of the th / eu age estimate for this star , to @xmath31 gyr .
these results are consistent with the very recent estimate of the age of the universe by wmap ( bennett et al .
2003 ) to within the reported errors .
recently , cayrel et al . ( 2001 ) and hill et al . (
2002 ) reported the detection of u , as well as th , in a high - quality vlt / uves spectrum of the extremely metal - poor star cs 31082001 , a bright [ fe / h ] = 2.9 giant discovered in the hk survey .
they also found that the abundance pattern of neutron - capture elements with @xmath32 closely mimics that of the r - process component in the solar system .
however , it is of interest that the situation seems to be different for nuclei near the heaviest elements .
th ( @xmath33 ) and u ( @xmath34 ) show larger deviations from their expected levels , if we adopt @xmath12 gyr as the age of this object , as was estimated from the u / th ratio by @xcite .
this result suggests that either cs 31082001 is peculiar in some way that has affected the surface abundances of the actinides , or that the production ratios of th and u , as compared to the stable elements ( e.g. , eu ) exhibit some real dispersion amongst very metal - poor stars .
further observational work is clearly required to answer this important question .
the primary purpose of this study is to obtain measurements of the abundance patterns of neutron - capture elements for very metal - poor stars , and to examine the age estimation of stars using the th / eu chronometer . in honda
et al . ( 2003 ; hereafter paper i ) , we reported equivalent width measurements of absorption lines in high - quality spectra of 22 very metal - poor stars obtained with the subaru telescope high dispersion spectrograph ( hds , * ? ? ?
this sample is as large as those in previous studies by mcwilliam et al .
( 1995 ) , ryan , norris , & beers ( 1996 ) , and johnson & bolte ( 2001 ) , and the quality of the spectra is quite high ( @xmath35 per resolution element ) in the blue region .
in particular , the sample was selected to include as many objects with excesses of neutron - capture elements as possible , excluding the objects affected by the s - process , to investigate the nature of r - process nucleosynthesis in the early galaxy . in this paper , we present the abundance analyses for the neutron - capture elements , and discuss the observed abundance distributions in very metal - poor stars with excesses of r - process elements . in
2 we describe the determination of atmospheric parameters for our program stars .
the abundance analysis is described in 3 , where error estimates in our derived abundances , and comparisons with previous studies , are also discussed . in
4 we discuss the patterns in the abundance ratios of neutron - capture elements produced by r - process nucleosynthesis , based on the results of our analysis . in this same section ,
the abundance patterns of other elements , including the actinide th , are considered , as is the suitability of th - based chronometers for age estimates .
in order to perform an abundance analysis using model stellar atmospheres , we first need to determine the atmospheric parameters , i.e. , the effective temperatures , @xmath36 , the surface gravities , @xmath37 , the microturbulent velocities , @xmath38 , and the metallicities , [ fe / h ] , for each program star , based on the available photometric data and our high - resolution spectra .
these are considered in turn in the following subsections .
we estimate @xmath36 from the available photometric data , employing the empirical @xmath36 scale obtained by @xcite for the @xmath39 color indices , and that by @xcite for the @xmath40 index .
the broadband @xmath40 color has been frequently used for the determination of @xmath36 , because this color has been measured for the majority of the metal - poor stars discovered to date . however , the @xmath40 color of a giant star depends not only on @xmath36 , but also on metallicity and the presence , or not , of molecular absorption features , in particular those arising from carbon . hence , we give preference to the use of @xmath39 .
the photometry data used for the @xmath36 determination are provided in table [ tab : photo ] .
the @xmath41 data are taken from the two micron all sky survey point source catalog ( 2mass , skrutskie et al .
the @xmath40 data are taken from the simbad database for bright stars , and from the list of beers et al .
( 2003 , in preparation ) for the fainter stars .
since @xcite provides a @xmath36 scale for photometric data measured on the tcs system @xcite , the 2mass infrared photometry data are first transformed to those of the tcs system @xcite .
we found that the difference between the two photometric systems is negligible for our purposes . before applying the photometric data to the @xmath36 scale ,
we first need to correct for the effects of reddening on the measured colors .
schlegel , finkbeiner , & davis ( 1998 ) constructed a full - sky map of the distribution of galactic dust based on far - infrared data observed with the infrared astronomical satellite ( iras ) and the diffuse infrared background experiment ( dirbe ) instrument onboard the cosmic background explorer ( cobe ) satellite .
we make use of the reddening estimates they obtained from this map .
an exception is the star hd 140283 , for which a rather high value of @xmath42 was derived from the schlegel , finkbeiner , & davis ( 1998 ) map .
it should be kept in mind , however , that this applies to the entire line of sight , which may not be appropriate for bright , nearby stars .
for this star , which is only some 50 pc away , we instead adopted a reddening of @xmath43 , as estimated by ryan et al .
we note that arce and goodman ( 1999 ) cautioned that the map of schlegel , finkbeiner , & davis ( 1998 ) may overestimate the reddening values when the color excess @xmath44 exceeds about 0.15 mag , but the color excesses of our stars are lower than this threshold .
the extinction at @xmath45 is evaluated based on the relation @xmath46 .
we assume that the extinction at @xmath47 is negligible .
the @xmath36 estimated from @xmath39 and @xmath40 are listed in table 2 .
these were derived based on the latest photometric data we could obtain , including the recent 2mass release .
our abundance analyses were completed before some of these photometric data became available , and we have not made a re - analysis if the difference of the @xmath36 estimated from the updated photometry and that adopted by the analysis is less than 100 k. as a result , the @xmath36 adopted in the present analysis , also given in table 3 , is sometimes slightly different ( by 80 k maximum ) from that derived based on @xmath39 .
the @xmath36 estimated from @xmath40 for cs 30306132 is different by more than 400 k from that obtained from @xmath39 .
this object turned out to be moderately carbon - rich in our analysis , and this discrepancy should be , at least partially , due to the effect of ch molecular absorption in the b band . except for this object ,
the agreement between the two estimates of @xmath36 is fairly good : the average of the difference ( @xmath48 ) for the 21 objects other than cs 30306132 is 90 k , with a standard deviation of 64 k. other parameters ( @xmath37 , @xmath38 , [ fe / h ] ) are determined by the analysis of fe lines .
only lines with equivalent widths less than 100 m are used in this analysis .
we exclude stronger lines because they are severely affected by pressure broadening , which may not be well - estimated by our calculations .
the abundance of fe is evaluated for each fe i line using kurucz model atmospheres ( kurucz 1993 ) . in order to determine the @xmath38 of our program stars
, we use a subset of the fe i lines listed in table 2 of paper i. the @xmath38 is set by the value for which iron lines with a variety of strengths yield a consistent iron abundance , changing the values of this parameter with steps of 0.1 km s@xmath49 .
the @xmath38 estimated for our objects fall within the range 1 to 2.5 km s@xmath49 .
we determined the surface gravity , @xmath37 , for each star by demanding that the iron abundance determined from lines is equal to that derived from lines . in the first step , abundances are estimated from and lines by assuming a @xmath37 value .
we repeat the calculation by changing the @xmath37 value until the iron abundances derived from both ionization stages are in agreement .
the @xmath38 is then re - determined using the derived value of @xmath37 .
the @xmath37 value is then determined again on the basis of the newly obtained @xmath38 , and the process is iterated until convergence .
figure [ fig : tg ] shows the correlation between the estimated @xmath37 and @xmath36 obtained for our program stars .
the scatter is small , indicating that no extraordinary star is included in our sample , and that the derived @xmath50 is typical of red giants .
exceptions are the star hd 140283 , which we regard as a main - sequence turnoff star , and bs 17583100 , which is regarded as a subgiant .
our final set of derived atmospheric parameters are summarized in table [ tab : param ] .
we take the abundance as the indicator of metallicity of each star .
the atmospheric parameters adopted in the present work are generally in good agreement with those derived by other recent abundance studies . for example
, our model parameters ( @xmath36 = 4570 k , @xmath37 = 1.1 , and [ fe / h ] @xmath51 ) for the bright , well - studied giant hd 122563 are in good agreement with those reported by westin et al .
( 2000 ) ( @xmath36 = 4500 k , @xmath37 = 1.3 , and [ fe / h ] @xmath52 ) on the basis of a high s / n spectrum of this star .
the parameters of another metal - poor star , hd 115444 , derived by the present work ( @xmath36 = 4720 k , @xmath37 = 1.5 , and [ fe / h ] @xmath53 ) also show good agreement with those obtained for this star by @xcite ( @xmath36 = 4650 k , @xmath37 = 1.5 , and [ fe / h ] @xmath54 ) .
sneden et al .
( 2000 ) derived model parameters for cs 22892052 ( @xmath36 = 4710 k , @xmath37 = 1.5 , and [ fe / h ] = 3.1 ) ; these are in good agreement with our values ( @xmath36 = 4790 k , @xmath37 = 1.8 , and [ fe / h ] @xmath55 ) .
the small discrepancy in the derived [ fe / h ] values is primarily due to the higher @xmath36 and @xmath37 adopted by this work ( see subsection 3.6 ) .
hill et al .
( 2002 ) derived model parameters for cs 31082001 ( @xmath36 = 4825 k , @xmath37 = 1.5 , and [ fe / h]@xmath56 ) .
our estimates for this object ( @xmath36 = 4790 k , @xmath50 = 1.8 , and [ fe / h]@xmath57 ) agree quite well with theirs .
giridhar et al .
( 2001 ) derived model parameters for cs 22169035 ( @xmath36 = 4500 k , @xmath50 = 1.0 , and [ fe / h]@xmath58 ) and bs 16085050 ( @xmath36 = 4750 k , @xmath37 = 1.0 , and [ fe / h]@xmath59 ) .
our estimates for these objects are @xmath36 = 4670 k , @xmath37 = 1.3 , and [ fe / h]@xmath60 for cs 22169035 , and @xmath36 = 4950 k , @xmath37 = 1.8 , and [ fe / h]@xmath56 for bs 16085050 .
giridhar et al .
( 2001 ) determined the @xmath36 based on the requirement that the fe abundance derived from lines be independent of excitation potential and equivalent width .
these differences may arise because of the alternate methods used to obtain @xmath36 .
differences in @xmath37 and [ fe / h ] propagate from that in @xmath36 .
for our quantitative abundance measurements we make use of the analysis program sptool developed by y. takeda ( private communication ) , based on kurucz s width6 .
sptool calculates synthetic spectra and equivalent widths of lines on the basis of the given atmospheric parameters , line data , and chemical composition , under the assumption of lte .
the abundance analyses were made for 33 elements from c to th , where we have used the solar - system abundances obtained by @xcite .
the derived abundances for our program stars are listed in table [ tab : res1][tab : res6 ] .
carbon abundances provide an important tool for understanding not only the evolutionary stage of a given star , but also the nucleosynthesis history of the early galaxy . however , since the primary purpose of this work is to investigate the abundance patterns of the heavy elements , we have selected stars with rather weak ch g - bands , indicative of low carbon abundances , in order to avoid the difficulties of line contamination due to ch and cn molecular lines . even in our present sample ,
however , possible contamination from these molecular lines must be considered .
for instance , it is known that contamination from @xmath62ch lines can affect the region surrounding the 4019 line ( norris , ryan , & beers 1997 ) . to ascertain the possible effects of the molecular lines on our analysis , in this section we obtain estimates of the abundances of the carbon isotopes ( @xmath63c and @xmath62c ) .
the carbon ( @xmath63c ) abundance is measured using the spectrum synthesis technique for the ch 4323 band , adopting the ch line data from @xcite .
we have confirmed that the ch band of the solar spectrum is well - reproduced by spectrum synthesis using this line list , and the kurucz model atmosphere of the sun @xcite .
examples of the observed spectra of the ch band for four of our objects are shown in figure [ fig : ch ] . in this figure
we also show synthetic spectra for three carbon abundances with differences of 0.3 dex .
it is seen that one can certainly specify the best - fit features to better than 0.3 dex .
the ch band is too weak in bs 16085050 and bs 16920017 to measure their carbon abundances , hence for these two objects we have only estimated upper limits .
the derived carbon abundances for our program stars are listed in tables [ tab : res1][tab : res6 ] .
we estimate the carbon isotope ratios ( @xmath63c/@xmath62c ) from the ch lines around 4200 @xcite .
examples of the observed spectra for the same four stars discussed above , along with synthetic spectra for three different isotope ratios , are shown in figure [ fig:13ch ] .
the @xmath62ch line appears redward of the @xmath63ch line in each panel . note that
, in general , the @xmath62ch lines are very weak in our objects . as a result , we are able to estimate the isotope ratios for six of our program stars , but can only obtain upper limits for 13 objects . since no @xmath63ch line appears in this wavelength region for the other three stars in our sample ( cs 22952015 , bs 16085050 , and bs 16920017 ) , the isotope ratio is not derived for these objects .
the results of this exercise are also provided in tables [ tab : res1][tab : res6 ] .
the @xmath63c/@xmath62c ratios derived by this analysis fall in the range @xmath64 , which are typical values for red giants .
the abundances for elements with @xmath66 are derived with a standard analysis using the equivalent widths reported in paper i. the transition probabilities of atomic lines employed in this analysis are adopted from @xcite , as mentioned in paper i. the results show the behavior of typical halo stars for most of our objects ( e.g. , * ? ? ?
* ) , with a few particularly interesting exceptions .
giridhar et al .
( 2001 ) found that bs 16085050 is @xmath3-enhanced , and that cs 22169035 is @xmath3-poor , results which our new data also confirm .
we plan to discuss these results in more detail in a future paper ( honda et al . , in preparation ) .
the cr abundances for our sample stars also show higher values , as compared with previous studies ( e.g. , * ? ? ?
this is because we use lines with high excitation potentials .
the zn abundances derived from the two lines at 4722 and 4810 ( see paper i ) will be discussed separately , along with the abundances of other iron - peak elements , in forthcoming paper .
we apply a standard analysis to most lines of the neutron - capture elements using the equivalent widths reported in paper i. for some strong lines , as well as lines contaminated by other absorption features , a spectral synthesis technique is required . for most lines ,
the transition probabilities compiled by @xcite and @xcite are adopted .
line data compiled by @xcite are used for the analysis of . for , and
, we adopt the recent measurements of transition probabilities by @xcite , respectively .
studies of atomic data have progressed substantially lately , therefore we can also make use of the newest data for and , reported by @xcite and c. sneden ( private communication ) , respectively . the effects of hyperfine splitting and isotope shifts
are included in the analysis of the , , and lines .
we assume the isotope ratios of the r - process fraction in the solar system in the analysis of our very metal - poor stars , a convention that has been adopted by others ( e.g. , * ? ? ?
this assumption is justified by the result of our analysis that the abundance patterns of stars in our sample for which both ba and eu are detected are in good agreement with that of the r - process component in the solar system .
the effect of hyperfine splitting in some la lines is quite significant . in a previous stage of our analysis ,
older line data were used , and the effect of hyperfine splitting was not included .
we find , however , that our re - analysis , which includes this effect and the line data by @xcite , dramatically reduces the scatter of the derived abundances obtained from individual la lines .
we also find that the scatter of nd abundances derived from individual lines significantly decrease by using the line list recently provided by @xcite , as compared with previous line data .
our analysis underscores the importance of obtaining accurate line data , including hyperfine splitting , for use in abundance studies of these heavy elements .
we detect the yb ii 3694.2 line in seven objects , and obtain yb abundances by adopting the line data used by sneden et al .
( 1996 ) without including hyperfine splitting .
however , the yb ii 3694.2 line may indeed be affected by hyperfine splitting , as pointed out by aoki et al .
( 2002a ) and @xcite , because two of the seven isotopes of this element have odd mass numbers ( @xmath68yb and @xmath69yb ) .
this indicates that the yb abundances may be over - estimated by a single line approximation . however , from a re - analysis based on new line data ( sneden , private communication ) , the yb abundances appear to be better determined .
we detect the elements sr and ba , both of which have strong resonance lines in the blue region , for all objects in our sample .
y and zr are also detected in most objects .
in addition to these four elements , we detect eu in 11 objects .
several elements heavier than eu , including th ( see next subsection ) , are detected in seven objects , all of which exhibit overabundances of neutron - capture elements .
we discuss the abundance patterns of these stars in more detail in 4 .
we detect the th ii line ( 4019.1 ) in seven program stars , including four stars for which th abundances have already been reported in the literature .
figures [ fig : th1 ] and [ fig : th2 ] show the observed spectra of this region in these seven stars , as well as the spectrum of hd 122563 , a star with no neutron - capture enhancements , for comparison .
the 4019 line is known to potentially suffer from blends with lines of other elements ; the impact of the blending of course depends on the relative abundances of the competing species .
the ce ii line , for example , is important in stars with super - solar [ ce / fe ] ratios .
sneden et al .
( 1996 ) demonstrated the importance of this line for improving their spectral syntheses of the 4019 line of cs 22892052 .
norris et al .
( 1997 ) pointed out the significant contribution of @xmath62ch lines at this wavelength in @xmath62c - rich objects .
johnson & bolte ( 2001 ) compiled a line list for the analysis of the 4019 line , including the above contaminating species , which we employ in the present analysis . before carrying out our spectral analysis of the th line , the abundances of fe , ni , nd , co , and ce are obtained first , and fixed , so that only the abundance of the th is altered in the final synthesis step .
note that we adopt the partition function of th from @xcite . in the calculation of the spectra , the contamination of @xmath62ch lines at 4019.00 and 4019.17 @xcite is also included , adopting the carbon isotope ratios derived in subsection [ sec : ab6 ] . for cs 31082001 and cs 30306132 , for which only an upper - limit of the @xmath62c abundance was derived
, we use the upper limit as a substitute for the @xmath62c abundance in the calculation .
we find that the effect of @xmath62ch is negligible in the spectra of these two stars .
the detection of the 4019 line has already been reported by previous authors for cs 22892052 , hd 115444 , hd 186478 , and cs 30182001 @xcite .
we have confirmed this detection in the spectra obtained with hds for these four objects .
in addition to these four objects , the th line is newly detected in three additional stars , hd 6268 , hd 110184 , and cs 30306132 .
the radioactive species uranium is a key element for precision estimation of stellar ages from abundance studies such as ours , owing primarily to its relatively `` short '' half - life of 4.5 gyrs .
errors in its measurement for a given spectrum lead to smaller errors in the estimated decay age than , for example , measurements of thorium , with its much longer half - life .
however , detection of u in stars is quite difficult , because of the weakness of its spectral lines ( e.g. , 3859 ) , and its low present abundance in @xmath22 1215 gyr old stars ( note that these ages are sufficiently long that only 10%15% of the originally produced u will survive ) .
since we found no evidence of absorption by this element in our spectra , we can only estimate upper limits on its abundance for the seven stars in which th is detected ( tables [ tab : res1][tab : res6 ] ) . though @xcite reported the detection of u in their very high s / n ( @xmath71 ) spectrum of cs 31082001 , we could not find u in this object , probably because of the lower quality of our spectrum .
the upper limits we are able to derive from our present spectra are rather high , hence no meaningful information is provided for the age estimates of these stars .
higher quality spectra for these very metal - poor stars are obviously required , at least if one hopes to detect u. even without a detection , improved upper limits to the abundance of u can be used to provide interesting limits on the ages of these stars ( see the discussion in truran et al .
2002 ) .
we now estimate the uncertainties in our abundance analysis arising from two sources .
the first source is random error , which might be caused , for example , by uncertainties in the adopted _ gf _ values and errors in the equivalent width measurements .
the second is the systematic error arising from uncertainties in our adopted atmospheric parameters .
the size of the random errors are estimated from the standard deviation ( 1-@xmath72 ) of the abundances derived from individual lines for elements that had three or more lines available to include in the analysis . for elements based only on one or two lines ,
we employ the mean of the random errors estimated from those elements with multiple lines available .
the derived standard deviations are shown in table [ tab : res1][tab : res6 ] .
thorium is a particular case , since the th abundance is determined from the one detected absorption line at 4019 using the spectrum synthesis technique .
the random error associated with this measurement might be estimated from errors of the fits themselves ( e.g. , a @xmath73 analysis ) , but we find that this approach is not useful for our limited quality spectra . in the present analysis ,
we have chosen to simply estimate the maximum fitting errors `` by eye '' . in figures [ fig : th1 ] and [ fig : th2 ] , synthetic spectra for the adopted th abundance and two other possible th abundances with 0.1 dex or 0.15 dex differences with respect to our adopted fit are shown ( see the figure captions ) .
we estimate the fitting error in the best cases ( for hd 6268 and cs 31082001 ) to be about 0.10 dex , and for those in other cases to be 0.15 dex .
we adopt these fitting errors as estimates of the random error associated with our derived th abundance .
note that we neglect the error due to the uncertainty of the transition probability , which is very difficult to estimate . in any case
, the following discussion pertains primarily to the relative abundances among our sample stars . any ambiguity in the estimated atmospheric parameters for a given star can result in an inappropriate choice of model atmospheres for use in the abundance analysis .
in order to estimate the typical errors due to this ambiguity , we evaluate the effects of changes in adopted atmospheric parameters for the case of hd 115444 .
the error in @xmath36 is assumed to be 100 k. this should be a reasonable assumption , because the errors in the photometry data and the reddening correction is 0.030.05 mags , which corresponds to an uncertainty of about 50 k ( for @xmath36 estimated from @xmath39 ) .
the difference between the @xmath36 obtained from @xmath40 and that from @xmath39 is also about 100 k on average ( see subsection 2.1 ) .
errors in @xmath37 , metallicity , and @xmath38 are assumed to be 0.3 dex , 0.5 dex , and 0.5 km s@xmath49 , respectively .
errors in the resulting abundance estimates due to these uncertainties are evaluated by varying the individual parameters .
changes in the final abundances ( @xmath74 ) caused by the above - noted parameter changes are listed for hd 115444 in table [ tab : error ] . as found in previous studies , a higher @xmath36 results in higher derived abundances , in general .
the effect of the difference in metallicity assumed in the model atmosphere is relatively small .
the assumption of a larger @xmath38 results in lower derived abundances .
this effect is important in the abundance analysis for , , , , and , which are based on rather strong lines .
total systematic errors are derived by the root - sum - square ( rss ) of the uncertainties contributed by the four parameters ( @xmath75 k , @xmath76 , @xmath77 [ fe / h ] = + 0.5 , and @xmath78 km s@xmath49 ) . the typical error due to the uncertainties of the stellar atmospheric parameters is about 0.14 dex for the neutron - capture elements , with the exception of sr and ba , which are influenced primarily by the adopted @xmath38 .
it should be noted that the systematic errors in the abundance analysis for neutron - capture elements are small ( @xmath79 dex ) , and the abundances are affected by the atmospheric parameters in a similar way .
therefore , the errors in the _ relative abundances _ among these elements are smaller than the random errors estimated above .
the elemental abundances for several of the stars in our sample have already been studied by previous authors . in this subsection
, we compare our results with those of the previous studies .
we focus here on the neutron - capture elements of the four stars in which th was previously detected .
the abundances of many neutron - capture elements in cs 22892052 have been investigated by @xcite .
we compare our results with those of the comprehensive study for this object carried out to date @xcite .
our derived abundances for most elements agree well with their results ; figure [ fig : de052 ] shows the difference between our results and theirs .
the th abundance of cs 22892052 obtained by our present analysis is log@xmath80(th ) @xmath81 , which is 0.15 dex higher than the th abundance derived by their latest analyses ( @xmath82 ) .
this small difference is probably due to the differences in adopted @xmath36 and @xmath50 ( see subsection 2.3 and table 7 ) .
we note that the log(th / eu ) of this object by our analysis is @xmath83 , which is in good agreement with their result ( log(th / eu ) = @xmath84 ) .
the extremely r - process - enhanced , very metal - poor star cs 31082001 has been studied in detail by @xcite and @xcite .
our results for the abundances of the neutron - capture elements show good agreement with these previous studies within the stated errors ( figure [ fig : de052 ] ) .
although the species sm and eu show small differences , this may be due primarily to measurement errors , since our hds spectrum is not of the same quality as the extremely high s / n spectrum obtained by @xcite .
cs 31082001 is known to exhibit unusually strong absorption lines of .
the th abundance derived by our analysis is log@xmath80(th ) @xmath85 ( log(th / eu ) @xmath86 ) .
these values agrees very well with those ( log@xmath80(th ) @xmath87 0.05 ; log(th / eu ) @xmath88 ) reported by hill et al .
( 2002 ) .
@xcite derived the abundances of a number of neutron - capture elements in hd 115444 from a high - quality spectrum of this star .
the abundances of most neutron - capture elements derived by our analysis agree with theirs within the errors .
an exception is the case of tm ( figure [ fig : de052 ] ) .
although we investigated a number of possible sources of difficulty , we could not resolve the likely reason for the difference .
we note , however , that our present result is based on two lines , which agree well with one another , while the @xcite analysis is based on only a single line .
our estimate of the th abundance in hd 115444 , log@xmath80(th ) @xmath89 , appears to be somewhat higher than the value obtained by westin et al .
( 2000 ) , log@xmath80(th ) @xmath90 .
the difference between the two measurements is at a similar level as the uncertainties in the analysis , so perhaps this should not be of concern .
however , inspection of the synthetic spectra used in this analysis indicates that this discrepancy could arise from differences in the line lists that were adopted . in our synthetic spectrum _ without _ the line ( figures [ fig : th1 ] ) the absorption feature at 4019.1 , where exists , is considerably weaker than that shown in the synthetic spectrum of @xcite ( their figure 7 ) . in order to identify the reason for this difference
, we also compare the synthetic spectra for hd 122563 , which was studied by both teams .
this star is a very metal - poor giant with similar atmospheric parameters as hd 115444 , but exhibits no th feature .
we find that our synthetic spectrum of hd 122563 , calculated with the same line list as used for hd 115444 , reproduces very well the observed spectrum at 4019.1 . on the other hand ,
the absorption feature at 4019.1 in the synthetic spectrum of @xcite for this star is deeper than that in the observed spectrum .
although the discrepancy between the observed and synthetic spectra seems to be similar to the noise level of their observed spectra , the spectral feature at 4019.1 was also observed in our very high - quality spectrum of this star .
for this reason , we suggest that the discrepancy between the observed and synthetic spectra found in figure 7 ( lower panel ) of @xcite is _ not _ due to random errors in the observed spectrum , but rather arises from an overestimate of the absorption at 4019.1 in their calculation . if the absorption at this wavelength is also overestimated for hd 115444 , the contribution of the is underestimated .
this may explain the difference of 0.2 dex in the th abundances between the results of @xcite and our present analysis .
@xcite derived the elemental abundances of hd 115444 and hd 186478 .
the fe abundances derived by these authors are lower by 0.3 dex and 0.1 dex than ours for hd 115444 and hd 186478 , respectively ( figure [ fig : de052 ] ) .
these differences are likely due to differences in the atmospheric parameters adopted in the two analyses .
their @xmath36 and @xmath37 for these objects are lower than ours by about 200 k and 0.8 dex , respectively .
the th abundances derived by @xcite for hd 115444 and hd 186478 are log@xmath80(th ) @xmath91 and log@xmath80(th ) @xmath92 , respectively .
these values are lower by 0.39 dex and 0.41 dex than our results , respectively .
these differences are well explained by the differences in the adopted @xmath36 and @xmath37 .
note that differences of this size do not affect the relative abundances of neutron - capture elements very significantly .
for example , the differences in the value of log(th / eu ) become smaller , 0.21 and 0.19 for hd 115444 and hd 186478 , respectively , roughly the same size as the value of the estimated random errors .
the present analysis has shown that our objects indeed possess very low metallicities ( @xmath93 [ fe / h ] @xmath2 ) .
the iron abundances derived by the present analysis are similar to the values which were estimated by the previous studies from lower dispersion spectroscopy ( beers et al .
1992 ; bonifacio et al .
2000 ; allende - prieto et al .
one exception is the star cs 22952015 , whose metallicity was estimated to be [ fe / h ] @xmath94 by beers et al .
( 1992 ) , while our estimate is rather higher , [ fe / h ] @xmath95 .
another is cs 30306132 , whose metallicity was estimated to be [ fe / h]@xmath96 from lower - dispersion spectroscopy by beers et al .
( 2003 , in preparation ) , while our result is [ fe / h]@xmath97 .
this discrepancy may be due to the fact that this star has strong ch molecular bands .
the metallicity range of the stars in our sample should be kept in mind , as it is known that a large scatter in the abundance ratios in many neutron - capture elements appears when the metallicity drops to [ fe / h ] = 2.5 or below ( e.g. , * ? ? ?
* ) . in this section ,
we discuss the relative abundances of the neutron - capture elements , the dispersion of the observed abundance ratios , and the origin of these elements ( subsection 4.1 ) .
we discuss in detail the abundance pattern of neutron - capture elements for the seven stars in our sample for which th is detected , and the impact of our new results on the cosmochronology technique based on th ( subsection 4.2 ) .
the abundances of the @xmath3- and iron - peak elements are discussed only for comparison purposes .
details on the behavior of these elements will be presented separately in a future paper in this series ( honda et al .
, in preparation ) .
figure [ fig : ba ] shows the values of [ ba / fe ] as a function of [ fe / h ] for all of the stars in our sample ( filled circles with error bars ) , as well as the results reported by previous authors for comparison ( open circles ) .
ba abundances have been reported for stars with lower metallicity than the abundances of other neutron - capture elements such as eu .
this is because ba has strong ba ii resonance lines ( 4554 and 4934 ) , which remain detectable even as the overall level of metallicity decreases .
previous studies have shown that [ ba / fe ] drops below the solar ratio , on average ( e.g. , ryan et al . 1996 ; mcwilliam 1998 ) , at the lowest metallicities .
this trend is thought to arise because of a change in the primary nucleosynthesis sources for ba between stars with [ fe / h ] @xmath99 and those with [ fe / h ] @xmath100 .
ba in the more metal - rich stars is believed to originate primarily from the ( main ) s - process in low - mass or intermediate - mass stars ( 1 - 8 @xmath101 ) with a comparatively small contribution by the r - process .
the production of ba at the lowest metallicities , however , is most likely due to the r - process alone , occurring in type ii supernova explosions of massive and hence rapidly - evolving stars .
the contribution of the s - process to ba in the early galaxy is small , because the time - scale for the evolution of lower mass stars is long , and the ejecta from these stars contribute only to stars which formed somewhat later , with [ fe / h ] @xmath102 ( e.g. , * ? ? ?
* and references therein ) .
our results , shown in figure [ fig : ba ] , confirm the existence of an extremely large scatter ( a factor of @xmath221000 over the entire range ) in the ba abundances of the most metal - poor stars , even larger than reported in previous studies .
for instance , the standard deviation of the [ ba / fe ] values of our sample is 0.82 dex , while that found by @xcite for 24 objects with [ fe / h ] @xmath0 is 0.59 dex .
one clear reason for this larger dispersion is that the detection limit for ba lines in our work is lower than in previous programs , thanks to the high quality of the subaru / hds spectra .
the other reason is likely due to our selection of candidate neutron - capture - enhanced stars , as mentioned in paper i. in spite of this selection bias , the large dispersion in ba abundances that exists in the metallicity range of @xmath103 [ fe / h ] @xmath99 , and the possibly decreased scatter at lower iron abundances , provides an important clue to the sites and mechanisms of astrophysical neutron - capture processes @xcite .
roughly 95% of the eu in solar - system material is believed to be produced by the r - process ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , hence this element is particularly suitable for investigating the characteristics of the r - process in the early galaxy .
however , measurements of eu abundances in extremely metal - poor stars ( [ fe / h ] @xmath104 ) are still quite limited , because there is no strong eu line , unlike those of sr and ba .
we have detected eu in 11 objects in our sample ; the derived [ eu / fe ] ratios for our sample are shown in figure [ fig : eu ] as a function of [ fe / h ] ( filled circles ) , along with the results obtained by previous studies ( open circles ) .
our results show a large scatter also in [ eu / fe ] in very metal - poor stars , as has been reported by previous authors ( e.g. , * ? ? ?
* ; * ? ? ?
the ratio [ ba / eu ] is useful for distinguishing the contributions of the r- and s - processes , because the expected ratios from the two processes are quite different .
figure [ fig : baeu ] shows [ ba / eu ] as a function of [ fe / h ] for the 11 stars in our sample for which eu has been detected .
the dotted line indicates the value of [ ba / eu ] of the solar - system r - process component ( [ ba / eu ] @xmath105 , arlandini et al . 1999 ) , while the dashed line indicates that of the s - process component ( [ ba / eu ] @xmath106 , arlandini et al . 1999 ) . the [ ba / eu ] ratios exhibited by our stars are clearly associated with the r - process , rather than the s - process .
similar results have been reported by previous authors .
mcwilliam ( 1998 ) found that the mean of the [ ba / eu ] ratios for 10 stars with [ fe / h ] @xmath6 2.4 is 0.69 , consistent with the value expected from pure r - process nucleosynthesis within the measurement uncertainties .
here we find that the ratio of [ ba / eu ] matches that of the solar - system r - process component for many stars at low metallicity .
these results suggest that ba , as well as eu , is primarily produced by the r - process during the early history of the galaxy , and that ba can be used as a powerful tool to investigate the behavior of r - process elements in the early galaxy , as it remains detectable in stars of metallicities that are far lower than those in which eu is detectable .. though most of the stars shown in figure [ fig : eu ] exhibit [ eu / fe ] @xmath107 , we expect that the [ eu / fe ] values of the stars with low ba abundances are lower than zero , and in reality will , once detected , exhibit a similarly large dispersion as seen in [ ba / fe ] .
this prediction should be confirmed by studies of eu lines using higher quality spectra for objects with [ fe / h ] @xmath108 ( e.g. , ishimaru et al . 2004 ) .
large levels of scatter are also found in the abundance ratios of almost all neutron - capture elements with _ z _ @xmath109 56 .
figure [ fig : scatter ] shows the average of the elemental abundances relative to iron [ x / fe ] , and the standard deviation as a measure of the scatter in the abundances , as a function of atomic number .
the number of objects in which the species is detected depends upon the element : i.e. , neutron - capture elements heavier than ba are detected only in stars with excesses of neutron - capture elements .
the dispersion of the abundances for elements which are detected in less than 12 objects is shown by the thin bars .
even though this limitation will make the dispersion of the abundances of heavy neutron - capture elements ( @xmath110 ) smaller , the dispersion of the abundances of neutron - capture elements is much larger than found for the @xmath3- and iron - peak elements .
the scatter in the abundances of heavy neutron - capture elements relative to iron found in these very metal - poor stars means that the nucleosynthesis site of iron - peak elements and the r - process elements are quite different , and that the mixing of the yields from the first supernovae into the ism is incomplete in the early stages of the galaxy . the large scatter of r - process elements appearing at [ fe / h ] @xmath111 should provide a constraint on the dominant site of r - process nucleosynthesis ( e.g. , *
* ; * ? ? ?
we have confirmed that cs 22892052 and cs 31082001 are extremely r - process - rich objects , and that hd 6268 , hd 115444 and hd 186478 are moderately r - process - rich objects .
in addition , we have found two new r - process - enhanced objects , cs 22183031 and cs 30306 - 132 .
cs 22183031 exhibits a large excess of r - process elements ( e.g. , [ eu / fe ] @xmath112 ) .
unfortunately , we could not detect many lines because of the relatively low s / n in our spectrum of cs 22183 - 031 .
we are able to detect numerous lines in cs 30306 - 132 , and found that this object has a moderate enhancement of r - process elements ( [ eu / fe ] @xmath22 + 0.8 ) . in
4.2 we describe this object in detail .
figure [ fig : lightr ] shows the abundance ratios of [ sr / fe ] , [ y / fe ] , and [ zr / fe ] as a function of [ fe / h ] .
the [ sr / fe ] ratios in our program stars are distributed over a very wide range , from @xmath114 to @xmath115 , confirming the large dispersions in this ratio found by previous studies ( e.g. , * ? ? ?
* ; * ? ? ?
the stars in our sample appear to exhibit a rather higher mean [ sr / fe ] ratio than those of previous studies , but this is likely because of the sample selection , as discussed in the previous subsection .
the scatter that appears in the [ y / fe ] and [ zr / fe ] ratios is smaller than that in [ sr / fe ] .
however , y and zr do not have such strong spectral lines as the resonance lines , hence these two elements are not detected in several objects in our sample .
this may account for their smaller dispersion ; clearly , this should be investigated by obtaining higher - quality spectra of stars that presently have only upper limits on their y and zr abundances . in order to investigate the reason of the large dispersion in the sr abundances , in figure [ fig : srbafe ]
we plot [ sr / ba ] as a function of [ fe / h ] . as discussed previously by @xcite ,
although the dispersion in [ sr / ba ] at very low metallicity is rather smaller than that of [ sr / fe ] , the range is still almost 2 dex .
this stands in stark contrast to the range of [ ba / eu ] exhibited by the stars in our sample , all of which have quite similar values .
this result suggests that either ( a ) the process that contributed significant amounts of sr in these metal - deficient stars did not yield similar amounts of ba , or ( b ) the process that produced ba at very low metallicity yielded a variety of sr / ba ratios . to investigate the correlation between sr and ba , figure [ fig : srbabafe ] shows the ratio [ sr / ba ] for the stars in our sample as a function of [ ba / fe ] .
such a diagram was was also shown , for a different set of stars , by @xcite and sneden , preston , & cowan ( 2003 ) . the sample of sneden et al .
( 2003 ) includes s - process - rich stars , in which pb is detected . by way of contrast ,
we have excluded stars known to exhibit large excesses of s - process elements and stars with [ fe / h ] @xmath116 from this figure , in order to avoid possible contamination by s - process nucleosynthesis .
( a possible contribution of the so - called weak s - process is discussed below . )
one clear result found in this figure is that the dispersion in [ sr / ba ] decreases with increasing ba abundance .
this correlation is much clearer in our figure than in figure 10 of @xcite , presumably because their sample includes several rather metal - rich objects , which could well be affected by the contribution of s - process nucleosynthesis .
moreover , our new sample of stars has added a number of objects at the high end of the [ ba / fe ] range , hence this contributes to making the correlation in this range clearer .
figure [ fig : srbares ] shows the abundances of sr and ba for very metal - poor stars . for the stars in common between our study and those of others , we have adopted the abundances derived in the present study . as a result , a total of 46 stars are shown in this figure .
it is obvious that the dispersion in the sr abundance decreases with increasing ba abundance . to demonstrate this quantitatively , we divided the sample into three groups on the ba abundance ( @xmath117(ba)@xmath118 : 16 stars , @xmath119(ba)@xmath120 : 19 stars , and @xmath121(ba ) : 11 stars ) , then measured the standard deviation of the sr abundances for each group .
the results are 0.25 dex , 0.38 dex , and 0.71 dex , respectively
. clearly , the dispersion increases with decreasing ba abundances .
while the standard deviation in the range of @xmath117(ba)@xmath118 is of a similar level as the typical observational errors , in the range of @xmath122(ba)@xmath123 the dispersion is significantly larger than the measurement errors .
we also plot in figure [ fig : srbares ] the values of sr and ba in solar - system material ( grevesse , noels , & sauval 1996 , the dotted circle ) , as well as the solar - system r - process component ( the filled square ) .
unfortunately , estimation of the s - process component for sr ( knowledge of which is required in order to obtain the r - process residual value ) is quite uncertain , because of possible contamination from the so - called weak component of the s - process @xcite . for the r - process component of sr
, we adopted the average of the values derived by @xcite and by arlandini et al .
( 1999 , the estimate from their classical model ) , and show the difference by an error bar . a simple model can be constructed for the enrichment of sr and ba , assuming initial abundances of sr and ba ( @xmath124(sr ) and @xmath124(ba ) ) and a constant sr / ba ratio ( ( sr / ba)@xmath125 ) in the yields of the r - process , i.e. , @xmath80(sr)=@xmath124(sr)@xmath126(sr / ba)@xmath127 @xmath80(ba)=@xmath124(ba)@xmath128 .
the two solid lines in figure [ fig : srbares ] show cases for different initial sr abundances : @xmath124(sr ) @xmath129 and 5 , respectively .
@xmath124(ba)@xmath130 and ( sr / ba)@xmath131 is assumed for both cases .
a glance at this figure shows that the observational data fill the range between the two lines .
hence , the distribution of the observed sr and ba abundances are quite naturally explained by the simple assumptions of a large dispersion of sr abundances at @xmath122(ba)@xmath132 and enrichment of ba and sr with a constant sr / ba ratio . moreover ,
if we extend the line representing the enrichment of sr and ba , the sr and ba abundance of the r - process component in solar - system material is also explained .
a similar scenario for sr and ba enrichment has already been proposed by previous studies ( e.g. , * ? ? ?
* ; * ? ? ?
in particular , @xcite selected several metal - deficient stars with high and low ba abundances , and concluded that ba - poor stars show high sr / ba ratios , while the sr / ba ratios of ba - rich stars are similar to that of the r - process component in the solar system .
these authors suggested the existence of two processes ( sites ) that produce sr .
our interpretation for the sr and ba abundance distributions is essentially the same as theirs .
however , the enrichment of sr and ba is seen much more clearly in the present study , as the result of adding new measurements and excluding stars of higher metallicity . the enrichment of ba in the metallicity range around [ fe / h]@xmath111 is sometimes referred to as the main r - process , which produces heavy ( a @xmath133 130 ) neutron - capture elements @xcite . our interpretation for the variation of the sr and ba abundances results in the sr / ba ratio produced by this process , ( sr / ba)@xmath125 , to be about 1.5 .
the absence of stars with sr / ba @xmath134 in figure [ fig : srbares ] means that ( sr / ba)@xmath125 is not significantly smaller than unity .
this value is much higher than the sr / ba predicted by recent models of r - process nucleosynthesis ( e.g. , @xmath135 : otsuki , mathews , & kajino 2003 ) .
the value derived by these models should be , however , a lower limit of the sr / ba ratio produced by a single site , as these models deal with only one condition which can produce heavy neutron - capture elements like ba .
the yields in a real r - process site , such as a type ii supernova explosion , should be an integration of the results of a variety of conditions including the cases with low neutron - to - seed ratios , and produce light neutron - capture elements like sr .
if our interpretation for the correlation between sr and ba abundances is correct , this places a quite strong constraint on the sr / ba ratio produced by the main r - process .
in contrast , the nucleosynthesis process that is responsible for the production of light neutron - capture elements ( e.g. , sr ) without producing the heavier elements ( e.g. , ba ) is as yet unclear .
one possibility is the existence of an independent nucleosynthesis process which dominantly produces lighter neutron - capture elements , probably prior to the main r - process which provides heavier ones ( e.g. , truran et al .
for instance , the above observational results are qualitatively explained by the assumption that different mass ranges of supernovae are responsible for the light and heavy neutron - capture elements ( e.g. , qian & wasserburg 2000 ) .
two different explosion mechanisms , i.e. , prompt and delayed explosion , are proposed for low - mass and high - mass supernovae , respectively , and numerical simulations of r - process nucleosynthesis have been made ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
possible r - process nucleosynthesis in neutron - star mergers may also play a role in the enrichment of the light neutron - capture elements ( e.g. , * ? ? ?
* ; * ? ? ?
* ) . on the other hand , another explanation by r - process nucleosynthesis in a single site may be possible if incomplete mixing of supernova ejecta with interstellar matter is assumed .
for instance , some neutrino - heated wind models predict significantly large overproduction of light neutron - capture elements in the early phase ( @xmath136 second ) of the explosion prior to the r - process which produces the heavier elements ( e.g. , woosley et al .
cameron ( 2001 ) suggested that r - process nucleosynthesis may take place in the jets associated with gamma - ray bursts , which would be a possible mechanism for production of an inhomogeneous r - process in a single event .
stars with high sr and low ba abundances may form from the interstellar medium polluted by the ejecta enriched in light neutron - capture elements , if the ejecta is not well mixed with the matter ejected in the later phase of the r - process .
an alternative explanation for the enrichment of sr without ba may be the so - called weak s - process , which proposes a neutron - capture process proposed to occur in core he - burning massive stars ( e.g. , * ? ? ? * ) .
this process is expected to be unimportant in metal - deficient stars , because @xmath137ne is believed to be the neutron source . however
, if this process _ can _ operate , it may contribute to light neutron - capture elements in metal - deficient stars .
further systematic studies of stars with low ba abundances , especially of the abundance patterns for elements around sr , will provide important information to constrain the nucleosynthesis process responsible for their production , and their ejection dynamics as well .
one important discovery in recent abundance studies of the neutron - capture elements in metal - deficient stars is that the abundance pattern of heavy neutron - capture elements are very similar , essentially an exact match , within observational errors , to that of the r - process component in the solar system . to examine details of r - process nucleosynthesis at low metallicity ,
it is essential to detect as many elements as possible over the entire range of atomic numbers ( _ z _ = 31 @xmath22 92 ) .
such a study is possible for very metal - poor stars with overabundances of neutron - capture elements ( [ x / fe ] @xmath138 + 0.5 ) . in such stars
the absorption lines of neutron - capture elements are relatively strong , and the blending from lines of other lighter elements is comparatively weak , due to the overall low metallicity of the star .
we examine the abundance pattern of neutron - capture elements in detail for the seven very metal - poor stars in which the th absorption line ( 4019 ) was detected ( see figures [ fig : th1 ] and [ fig : th2 ] ) .
all of these stars exhibit enhancements of their neutron - capture elements . among them , cs 31082001 and cs 22892052 , which have already been studied by hill et al .
( 2002 ) and sneden et al . (
2003 ) , are the most extreme cases , showing large enhancements of the neutron - capture elements ( e.g. , [ eu / fe]=+1.7 and + 1.5 , respectively ) .
the star hd 110184 does not exhibit a remarkable enhancement of neutron - capture elements relative to iron , but our high s / n spectrum makes it possible to study th and other r - process elements in this star as well . abundance studies of r - process - enhanced , metal - poor stars to date have shown that the abundance patterns of the neutron - capture elements with 56 @xmath6 _ z _ @xmath6 70 agree very well with that of the solar - system r - process pattern ( e.g. , sneden et al . 2003 ) . in order to investigate this phenomenon further , using our extended sample , we now compare the scaled abundance patterns of our stars with the abundance distributions in solar - system material . for each of the seven stars with detectable r - process elements , we use the solar - system r - process abundance pattern as a template to compare the heavy - element abundances of the stars on a common scale ( figure [ fig : pattern ] ) .
we scale the elemental abundances of our objects to match the solar - system abundances for elements between ba and yb ( @xmath139 ) .
the logarithmic values of the scaling factor ( @xmath140 ) for individual stars are listed in the columns labeled s.s .
r - process in table [ tab : scale ] . for this analysis
, we adopt the r - process fraction in the solar system given by burris et al .
the total solar - system abundances were taken from @xcite .
figure [ fig : pattern ] shows that the abundance patterns of the elements with 56 @xmath5 70 for these very metal - poor stars agree very well with that of the solar - system r - process component .
we also attempted to compare the abundances of our metal - poor stars with the solar - system s - process distribution in the same manner .
take cs 30306132 , for example ( figure [ fig : pattern132 ] ) . as expected from the [ ba / eu ] ratios of these objects , the agreement between the abundance patterns of our objects and the s - process pattern is poor .
we also compare the abundances of our subset of metal - poor stars with r - process enhancements with the total elemental abundance distribution of the solar system ( figure [ fig : pattern132 ] ) .
the logarithmic values of the scale factor are given in the column labeled s.s .
total in table [ tab : scale ] .
the agreement is clearly better than that in the case for the abundance pattern of the s - process component alone .
this is because the r - process fraction dominates in the total solar - system abundances of elements with @xmath141 .
the scaled abundances of ba , la , and ce in our stars , shown in figures [ fig : pattern ] , are less than the solar - system total abundances of these elements .
presumably , this arises because the s - process contribution to the abundances of ba , la , and ce in the solar system is larger than for other elements like eu ( burris et al .
2000 ) . from these comparisons ,
we conclude that the abundance patterns of heavy neutron - capture elements in our objects agree best with the r - process component in the solar system . in table
[ tab : scale ] we also provide the standard deviation ( 1-@xmath72 ) of the difference between the scaled abundances of each star and the total solar - system abundances , as well as its r - process component .
these standard deviations can be taken as indicators of the level of agreement between the abundance pattern of each star and the pattern of the solar - system abundances .
the average of the observational errors for elements with @xmath139 ( @xmath142 ) are also provided in table 8 .
comparisons of the observed 1-@xmath72 values with @xmath142 indicates good agreement between the scaled abundance pattern of our objects with that of the solar - system r - process component .
we note that the 1-@xmath72 value for hd 110184 is slightly larger than those for other objects , and than its @xmath142 .
one reason for this result may be that the r - process enhancement of this star ( e.g. , [ eu / fe ] = 0.06 ) is small , and a small contribution by the s - process may affect the abundance pattern of this star . the excellent agreement between the abundance pattern of the heavy neutron - capture elements in very metal - deficient stars and that of the solar - system r - process component has already been reported for several very metal - poor stars by previous studies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , and is sometimes referred to as the universality of r - process nucleosynthesis . the apparent universality is ascribed to the fact that the predicted abundance patterns of elements with @xmath139 appear to be rather insensitive to variations in the parameters of the current r - process models ( e.g. , the entropy / baryon ratio ) .
the nucleosynthesis paths in this mass range ( i.e. , @xmath143 150 ) are quite similar among the r - process models that are used to predict the production of the actinide nuclei , even though the abundances of the actinides show an apparent variation in some stars ( e.g. , * ? ? ?
* ; * ? ? ?
as mentioned in 1 , the actinide th is heavier than the elements at the third abundance peak produced by the r - process ( os , ir , pb , etc . ) , and is a key element for the understanding of this nucleosynthesis process .
application of the abundances of this element to cosmochronometry has also been discussed extensively in recent studies of very metal - deficient stars .
the abundances of th measured for the seven stars in our sample are presented in figure [ fig : pattern ] . for the th abundance
, the solid line here shows the _ initial _ abundance of this radioactive element , as estimated by cowan et al .
( 1999 ) , rather than the _ present _ th abundance , which is shown by the dashed line in this figure .
since these very metal - poor stars are believed to be born in the early galaxy , the th abundances of our sample are expected to be lower than the value shown by the dashed line , if we assume that the abundance patterns of heavy neutron - capture elements , including th , produced by the r - process , is indeed universal .
this is in fact found for the stars hd 110184 , hd 115444 , hd 186478 , and cs 22892052 .
however , the th abundances of the other three stars are _ higher _ than would be expected from this logic . in order to investigate this issue more clearly
, we show the abundance ratios between th and the stable r - process element eu ( @xmath144(th / eu ) = @xmath122(th ) @xmath145(eu ) ) in figure [ fig : theu ] , where we plot our results , and the results of previous studies @xcite .
the average and standard deviation of the values of our seven stars are @xmath146 dex and 0.17 dex , respectively .
the standard deviation is as large as , or slightly larger than , the typical observation errors , which were estimated from the root - sum - square of the random errors of th and eu abundances ( error bars in figure [ fig : theu ] ) .
we have found that cs 31082001 and cs 30306132 have clearly higher th / eu ratios than the well studied star cs 22892052 ( figure [ fig : theu ] ) . in figure
[ fig : th2 ] , the observed spectra around the line are shown .
the 4018.6 line exists blueward of the line .
since the atmospheric parameters in these giants are quite similar , the ratio of the line strengths between and directly reflects the th / nd abundance ratio . in cs 22892052 ,
the line is as strong as the line .
the ratios of the line strengths in hd 6268 and hd 115444 are rather similar to that of cs 22892052 .
in contrast , in cs 31082001 the line is significantly stronger than the line .
in addition , in cs 30306132 , the is clearly detected , while there is no evidence for the presence of the line .
these results indicate that the th / nd ratios in cs 31082001 and cs 30306132 are significantly higher than in the other three stars . as discussed in detail in subsection 4.2.1 ,
nd in these objects is regarded as a product of the r - process , and can be taken as representative of the stable r - process elements . therefore , the above inspection suggests the existence of some dispersion in the abundance ratios between th and the other neutron - capture elements ( @xmath147 ) .
a similar result was already reported for cs 31082001 by @xcite and @xcite , as mentioned in 1 .
our present study shows that this object is not unique , but that there is at least one other similar object that shares the same `` actinide boost , '' cs 30306132 , though its enhancement factor of r - process elements is much smaller than that of cs 31082001 .
we might expect that , in the near future , as additional r - process - enhanced metal - poor stars are identified , such behaviors will be seen in additional stars .
if the conventional th / eu chronometer ( cowan et al , 1999 ) is simply applied to the th / eu abundance ratios , the ages of a few stars , those exhibiting actinide boosts , are estimated to be shorter than that of the present age of the sun .
in particular , the derived ages of cs 31082001 and cs 30306132 are as low as zero . even with a level of uncertainty of as much as 5 gyr in the age estimation , the ages derived from the above method appear unrealistic .
we note that the average of the th / eu ratios of our seven stars ( @xmath148(th / eu ) @xmath149 ) is quite similar to the value of solar - system material ( @xmath144(th / eu ) @xmath150 @xcite ) .
that is , the average of the ages derived from application of the th / eu chronometer is similar to the age of the sun , and hence is also unrealistic .
the observations seem to suggest that some very metal - poor stars had _ higher _ initial th abundances than expected from the solar - system r - process abundance pattern . in other words , even though the abundance pattern of the elements with @xmath139 agrees with the abundance pattern of the solar system r - process component , the initial abundance ratios of the heaviest elements , like th , to the lighter ones ( e.g. , eu ) are not necessarily the same as those expected from the r - process component in the solar system . in order to apply the abundance ratios between th and other stable elements as chronometers , estimates of the initial abundance ratios for these elements are essential
, hence a deeper understanding of the r - process nucleosynthesis for wider mass ranges is necessary .
one possible alternative cosmochronometer is the u / th ratio , recently applied by cayrel et al .
( 2001 ) and hill et al .
( 2002 ) for the extremely r - process - enhanced , metal - poor star cs 31082001 .
since @xmath20th and @xmath21u are neighboring actinide nuclei , their production rate is expected to be quite similar .
this justifies the assumption that the initial abundance ratio is the same as in the initial solar - system abundance ratio , as shown by recent theoretical calculations ( e.g. , * ? ? ?
* ; * ? ? ?
at present , u / th is expected to be the best available chronometer .
we would like to point out that , even though we conclude from our analysis that there exists a real scatter in the abundance ratios between th and other neutron - capture elements with @xmath151 , the level of this scatter is at most a factor of three , much smaller than the ratios between light ( @xmath152 ) and heavy ( @xmath151 ) neutron capture elements , which are as large as a factor of 10 ( see figure [ fig : pattern ] ) .
recent models of the r - process showed that the abundance ratios between the elements at the second and the third r - process peaks are quite sensitive to the parameters in the calculation ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) .
the small dispersion of th / eu ratios found in our stars , as well as in other previously studied stars , should be an important constraint on modeling the r - process , as we are beginning to place limits on the possible range over within which the initial production ratio can fall . in order to derive a clear conclusion ,
it is required to systematically study the th abundances for a larger sample of very metal - poor stars , based on high s / n spectra , which would enable one to detect the th line even in stars with lower th abundances .
we have conducted detailed abundance analyses for 22 very metal - poor stars based on high - resolution , high s / n near - uv - blue spectra obtained with subaru / hds .
our sample of stars covers the metallicity range @xmath93 [ fe / h ] @xmath2 .
this paper reports the results of abundance analyses concentrating in particular the r - process elements .
the main conclusions of this study are as follows : \(1 ) we have identified a new highly r - process element - enhanced , metal - poor star , cs 22183031 , a giant with [ fe / h ] = @xmath7 and [ eu / fe ] = + 1.2 . the lower s / n of its spectrum , however , prevented th from being clearly detected .
we also identified a new , moderately r - process - enhanced , metal - poor star , cs 30306132 , a giant with [ fe / h ] = @xmath8 and [ eu / fe ] = + 0.85 , in which th was detected .
\(2 ) we have confirmed the large star - to - star scatter in the abundances of neutron - capture elements relative to iron observed for stars with [ fe / h ] @xmath24 . the abundance pattern of the heavy neutron - capture elements ( @xmath153 ) in seven r - process - enhanced , metal - poor stars are , however , quite similar to that of the r - process component in solar - system material .
the ba to eu ratios in 11 metal - poor stars , including these seven objects , are nearly equal to that of the solar system r - process component .
these results prove that heavy neutron - capture elements in very metal - poor stars are primarily synthesized by the r - process .
\(3 ) we have investigated the correlation between sr and ba abundances in our sample , and in other very metal - poor stars ( [ fe / h ] @xmath0 ) studied by previous authors , and conclude that the dispersion of the sr abundances clearly decreases with increasing ba abundance .
this trend suggests the existence of two nucleosynthesis processes , one of which produces sr with very small production of ba , and the other which produces sr and ba with sr / ba @xmath154 .
the so - called main r - process might be associated with the latter process , while that responsible for the former is yet unidentified .
\(4 ) the th / eu abundance ratios ( log(th / eu ) ) measured for the seven r - process - enhanced stars range from @xmath9 to @xmath10 .
we have confirmed the high th / eu ratio found for the extremely r - process - enhanced star cs 31082001 @xcite ; the newly discovered moderately r - process - enhanced star cs 30306 - 132 exhibits a similar ( high ) th / eu ratio as cs 31082001 .
since these very metal - poor stars are believed to be formed in the early galaxy , this result means that a small dispersion appears in the abundance ratios between th and rare - earth elements , such as eu , in very metal - poor stars . in order to apply the th / eu ratios to estimates of stellar ages , further understanding of th production by r - process nucleosynthesis
is required .
clearly , important information on the nature of r - process nucleosynthesis is being obtained from abundance studies of very metal - poor stars , especially for those with enhancements of r - process elements .
further observational studies of these objects , based on spectra of even higher quality , will provide strong constraints on element production via r - process nucleosynthesis in the early galaxy . in this study
we have discovered a new , moderately r - process - enhanced , metal - poor star , cs 30306132 , which has a similar ( high ) th / eu ratio to cs 31082001 . we have also identified a new highly r - process element - enhanced , metal - poor star , cs 22183031 , a giant with [ fe / h ] = @xmath7 and [ eu / fe ] = + 1.2 .
we would like to thank yoichi takeda for development of the sptool software .
we would like to thank chris sneden for providing useful atomic line lists .
we also thank an anonymous referee for many helpful comments .
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 .
we have also made use of the simbad database , maintained by the cds , strasbourg , france .
this work has supported in part by grants - in - aid for scientific research 12047233 and 13640313 of the ministry of education , science , sports and culture of japan . t.c.b .
acknowledges partial support for this work from grants ast 00 - 98508 and ast 00 - 98549 awarded by the u.s .
national science foundation .
allende prieto , c. , rebolo , r. , garc@xmath155a , l@xmath156pez , r. , serra - ricart , m. , beers , t. c. , rossi , s. , bonifacio , p. , & molaro , p. 2000 , aj , 120 , 1516 alonso , a. , arribas , s. , & martinez - roger , c. 1994 , a@xmath157a , 282 , 684 alonso , a. , arribas , s. , & martinez - roger , c. 1996 , a@xmath157as , 313 , 873 alonso , a. , arribas , s. , & martinez - roger , c. 1998 , a@xmath157as , 131 , 209 alonso , a. , arribas , s. , & martinez - roger , c. 1999 , a@xmath157as , 140 , 261 aoki , w. , et al .
2001 , apj , 561 , 346 aoki , w. et al . 2002a , , 54 , 427 aoki , w. , norris , j.e . ,
ryan , s.g . ,
beers , t.c . , ando , h. , & tsangarides , s. 2002b , , 580 , 1149 arlandini , c. , kppeler , f. , wisshak , k. , gallino , r. , lugaro , m. , busso , m. , & straniero , o. , 1999 , , 525 , 886 arce , h. g. , & goodman , a. a. 1999 , apj , 512 , l135 beers , t. c. , preston , g. w. , & shectman , s. a. 1992 , aj , 103 , 1987 bennet , c. l. , et al .
2003 , apjs , 148 , 29 bonifacio , p. , monai , s. , & beers , t. c. 2000 , aj , 120 ,
2065 burris , d. l. , pilachowski , c. a. , armandroff , t. e. , sneden , c. , cowan , j. j. , & roe , h 2000 , apj , 544 , 302 butcher , h. r. 1987 , nature , 328 , 127 cameron , a. g. w. 2001 , apj , 562 , 456 cayrel , r.,et al .
2001 , nature , 409 , 691 cayrel , r. , et al .
2004 , a&a , in press cowan , j. j. , pfeiffer , b. , kratz , k .-
, thielemann , f .- k . , sneden , c. , burles , s , tytler , d , & beers , t. c. 1999 , apj , 521 , 194 cowan , j. j. , et al .
2002 , apj , 572 , 861 den hartog , e. a. , lawler , j. e. , sneden , c. & cowan , j. j. 2003 , apjs , 148 , 543 franois , p. , spite , m. , & spite , f. 1993 , a@xmath157a , 274 , 821 freiburghaus , c. , rosswog , s. , & thielemann , f .- k . 1999 ,
apj , 525 , l121 giridhar , s. , lambert , d. l. , gonzales , g. , & pandey , g. 2001 , pasp , 113 , 519 grevesse , n. , noels , a. , & sauval , a.j .
1996 , cosmic abundances , ed .
s. s. holt , @xmath157 g. sonnebord , asp conf .
ser . , 99 , 117 hill , v. , et al .
2002 , a&a , 387 , 560 hillebrandt , w. , nomoto , k. , & wolff , r. g. 1984 , a&a , 133 , 175 hoffman , r. d. , woosley , s. e. & qian , y .- z .
1997 , , 482 , 951 hoffman , r. d. , woosley , s. e. & weaver , t. a. 2001 , , 549 , 1085 honda , s. , et al .
2003 apjs submitted ( paper i ) ishimaru , y. , & wanajo , s. 1999 , apj , 511 , l33 ishimaru , y. , wanajo , s. , aoki , w. , & ryan , s.g .
2004 , apj , 600 , l47 johnson , j. a. , & bolte , m. 2001 , apj , 554 , 888 johnson , j. a. , & bolte , m. 2002 , apj , 579 , 616 kppeler , f. , beer , h. , & wisshak , k. , 1989 , rep .
, 52 , 945 kurucz , r. l. 1993 , kurucz cd - rom , no.13 ( harvard - smithonian center for astrophysics ) lawler , j. e. , bonvallet , g. , & sneden , c. 2001a , apj , 556 , 452 lawler , j. e. , wickliffe , m. e. , den hartog , e. a. , & sneden , c. 2001b , apj , 563 , 1075 lawler , j. e. , wickliffe , m. e. , cowley , c. r. , & sneden , c. 2001c , apjs , 137 , 341 meyer , b. s. , mathews , g. j. , howard , w. m. , woosley , s. e. , & hoffman , r. d. 1992 , apj , 399 , 656 mcwilliam , a , preston , g. w. , sneden , c , & searle , l 1995 , aj , 109 , 27 mcwilliam , a 1998 , aj , 115 , 1640 morell , o. , kallander , d. , butcher , h. r. 1992 , a@xmath157a , 259 , 543 noguchi , k.,et al .
2002 , pasj , 54 , 855 norris , j. e. , ryan , s. g. , & beers , t. c. 1997 , apj , 489 , l169 norris , j. e. , ryan , s. g. , & beers , t. c. 2001 , apj , 561 , 1034 otsuki , k. , tagoshi , h. , kajino , t. , & wanajo , s. 2000 , apj , 533 , 424 otsuki , k. , mathews , g. j. , & kajino , t. 2003 , new astronomy , 8 , 767 qian , y .- z . , &
wasserburg , g. j. 2000 , phys.rep . , 333 , 77 rosswog , s. , liebendorfer , m. , thielemann , f .- k . , davies , m. b. , benz , w. , & piran , t. 1999 , a&a , 341 , 499 ryan , s. g. , norris , j. e. , beers , t. c. 1996 , apj , 471 , 254 schatz , h. , toenjes , r. , pfeiffer , b. , beers , t. c. , cowan , j. j. , hill , v. , & kratz , k .-
2002 , apj , 579 , 626 schlegel , d. j. , finkbeiner , d. p. , & davis , m. 1998 , apj , 500 , 525 skrutskie , m.f .
et al . , 1997 , in the impact of large scale near - ir sky surveys , ed .
f. garzon et al .
( dordrecht : kluwer ) , p. 187 sneden , c. , mcwilliam , a. , preston , g. w. , cowan , j. j. , burris , d. l. , & armosky , b. j. 1996 , apj , 467 , 819 sneden , c , cowan , j. j. , ivans , i. i. , fuller , g. m. , burles , s , beers , t. c. , & lawler , j. e. 2000 , apj , 533 , l139 sneden , c. , cowan , j. j. , beers , t. c. , truran , j. w. , lawler , j. e. , & fuller , g. 2001 astrophysical ages and times scales , asp conference series vol .
245,235 sneden , c. , et al .
2003 , apj , 591 , 936 sneden , c. , preston , g. w. , & cowan , j. j. 2003 , apj , 592 , 504 sumiyoshi , k. , terasawa , m. , mathews , g. j. , kajino , t. , yamada , s. , suzuki , h. 2001 , apj , 562 , 880 truran , j. w. , cowan , j. j. , pilachowski , c. a. , & sneden , c. 2002 , pasp , 114 , 1293 tsujimoto , t. , shigeyama , t. , & yoshii , y. 2000 , apj , 531 , 33 wanajo , s. , itoh , n. , ishimaru , y. , nozawa , s. & beers , t.c .
2002 , , 577 , 853 wanajo , s. , tamamura , m. , itoh , n. , nomoto , k. , ishimaru , y. , beers , t. c. , & nozawa , s. 2003 , apj , 593 , 968 wasserburg , g. j. , & qian , y. -z .
2000 , apj , 529 , l24 westin , j. , sneden , c. , gustafsson , b. , & cowan , j. j. 2000 , apj , 530 , 783 woosley , s.e . , wilson , j.r . , mathews , g.j . , hoffman , r.d .
, & meyer , b.s .
1994 , , 433 , 229 lp7mmp7mmp7mmp7mmp7mmp7mmcrrrrrr object & @xmath158 & @xmath45 & @xmath159 & @xmath160 & @xmath161 & @xmath47 & @xmath40 & @xmath44 & @xmath162 & @xmath163 + hd 4306 & 9.71 & 9.08 & 9.08 & 7.42 & 6.98 & 6.82 & 0.63 & 0.00 & 0.63 & 2.26 + hd 6268 & 8.89 & 8.10 & 8.10 & 6.34 & 5.84 & 5.71 & 0.79 & 0.00 & 0.79 & 2.39 + hd 88609 & 9.52 & 8.59 & 8.59 & 6.67 & 6.13 & 6.01 & 0.93 & 0.00 & 0.93 & 2.58 + hd 110184 & 9.48 & 8.31 & 8.31 & 6.13 & 5.51 & 5.35 & 1.17 & 0.00 & 1.17 & 2.96 + hd 115444 & 9.75 & 8.97 & 8.97 & 7.16 & 6.70 & 6.61 & 0.78 & 0.00 & 0.78 & 2.36 + hd 122563 & 7.11 & 6.20 & 6.20 & 4.79 & 4.03 & 3.73 & 0.91 & 0.00 & 0.91 & 2.47 + hd 126587 & 9.88 & 9.15 & 8.84 & 7.26 & 6.78 & 6.67 & 0.73 & 0.10 & 0.63 & 2.17 + hd 140283 & 7.70 & 7.21 & 7.18 & 6.01 & 5.70 & 5.59 & 0.49 & 0.01 & 0.48 & 1.59 + hd 186478 & 10.08 & 9.18 & 8.81 & 7.12 & 6.60 & 6.44 & 0.90 & 0.12 & 0.78 & 2.37 + bs
16082129 & 14.22 & 13.55 & 13.49 & 11.86 & 11.40 & 11.31 & 0.67 & 0.02 & 0.65 & 2.18 + bs 16085050 & 12.89 & 12.15 & 12.09 & 10.47 & 10.01 & 9.94 & 0.74 & 0.02 & 0.72 & 2.15 + bs 16469075 & 14.19 & 13.42 & 13.36 & 11.75 & 11.26 & 11.20 & 0.77 & 0.02 & 0.75 & 2.16 + bs 16920017 & 14.64 & 13.88 & 13.85 & 12.17 & 11.68 & 11.59 & 0.76 & 0.01 & 0.75 & 2.26 + bs 16928053 & 14.32 & 13.47 & 13.44 & 11.66 & 11.14 & 11.04 & 0.85 & 0.01 & 0.84 & 2.40 + bs 16929005 & 14.23 & 13.61 & 13.58 & 12.17 & 11.75 & 11.67 & 0.62 & 0.01 & 0.61 & 1.91 + bs 17583100 & 12.88 & 12.37 & 12.06 & 11.10 & 10.69 & 10.66 & 0.51 & 0.10 & 0.41 & 1.40 + cs 22169035 & 13.80 & 12.88 & 12.76 & 11.00 & 10.48 & 10.35 & 0.92 & 0.04 & 0.88 & 2.40 + cs 22183031 & 14.27 & 13.62 & 13.50 & 12.11 & 11.65 & 11.58 & 0.65 & 0.04 & 0.61 & 1.91 + cs 22892052 & 13.96 & 13.18 & 13.18 & 11.30 & 10.85 & 10.93 & 0.78 & 0.00 & 0.78 & 2.25 + cs 22952015 & 14.05 & 13.27 & 13.15 & 11.49 & 11.02 & 10.92 & 0.78 & 0.04 & 0.74 & 2.23 + cs 30306132 & 13.61 & 12.81 & 12.78 & 11.52 & 11.00 & 10.75 & 0.80 & 0.01 & 0.79 & 2.03 + cs 31082001 & 12.44 & 11.67 & 11.67 & 10.05 & 9.61 & 9.46 & 0.77 & 0.00 & 0.77 & 2.21 + lccccccccc object & @xmath164 & @xmath165 & @xmath36(adopted ) & @xmath38 & @xmath37 & [ fe / h ] & @xmath72 & n + hd 4306 & 4814 & 5058 & 4810 & 1.6 & 1.8 & 2.89 & 0.09 & 81 + hd 6268 & 4679 & 4683 & 4600 & 2.1 & 1.0 & 2.63 & 0.11 & 82 + hd 88609 & 4512 & 4467 & 4550 & 2.4 & 1.1 & 3.07 & 0.20 & 70 + hd 110184 & 4220 & 4225 & 4240 & 2.1 & 0.3 & 2.52 & 0.11 & 47 + hd 115444 & 4707 & 4702 & 4720 & 1.7 & 1.5 & 2.85 & 0.15 & 76 + hd 122563 & 4599 & 4493 & 4570 & 2.2 & 1.1 & 2.77 & 0.19 & 84 + hd 126587 & 4919 & 5058 & 4960 & 1.8 & 2.1 & 2.78 & 0.12 & 80 + hd 140283 & 5633 & 5585 & 5630 & 1.4 & 3.5 & 2.53 & 0.08 & 78 + hd 186478 & 4700 & 4702 & 4720 & 2.2 & 1.6 & 2.50 & 0.12 & 78 + bs 16082129 & 4898 & 5002 & 4900 & 1.6 & 1.8 & 2.86 & 0.15 & 76 + bs 16085050 & 4943 & 4827 & 4950 & 1.8 & 1.8 & 2.91 & 0.10 & 82 + bs 16469075 & 4950 & 4762 & 4880 & 1.4 & 2.0 & 3.03 & 0.14 & 79 + bs 16920017 & 4839 & 4762 & 4760 & 1.4 & 1.2 & 3.12 & 0.23 & 69 +
bs 16928053 & 4681 & 4596 & 4590 & 1.6 & 0.9 & 2.91 & 0.14 & 80 + bs 16929005 & 5267 & 5118 & 5270 & 1.3 & 2.7 & 3.09 & 0.15 & 63 + bs 17583100 & 5934 & 5902 & 5930 & 1.4 & 4.0 & 2.42 & 0.10 & 55 + cs 22169035 & 4667 & 4535 & 4670 & 1.9 & 1.3 & 2.72 & 0.17 & 35 + cs 22183031 & 5247 & 5118 & 5270 & 1.2 & 2.8 & 2.93 & 0.20 & 49 + cs 22952015 & 4879 & 4783 & 4850 & 1.5 & 1.5 & 2.94 & 0.26 & 57 + cs 30306132 & 5105 & 4683 & 5110 & 1.8 & 2.5 & 2.42 & 0.13 & 93 + cs 22892052 & 4778 & 4702 & 4790 & 1.8 & 1.6 & 2.92 & 0.14 & 82 + cs 31082001 & 4825 & 4721 & 4790 & 1.9 & 1.8 & 2.81 & 0.12 & 69 + @c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c & & & & & + & & & & & & & & & & & & & & & & & + & fe / h & 4.62 & 2.89 & 0.09 & 81 & 4.88 & 2.63 & 0.11 & 82 & 4.44 & 3.07 & 0.2 & 70 & 4.99 & 2.52 & 0.11 & 47 + & @xmath63c/@xmath62c & & @xmath13320 & & & & 4 & & & & @xmath1333 & & & & 5 & & + 6 & c & 5.78 & 0.11 & & & 5.26 & 0.67 & & & 4.98 & 0.51 & & & 5.37 & 0.67 & & + 12 & mg & 5.25 & 0.56 & 0.17 & 4 & 5.44 & 0.49 & 0.25 & 4 & 4.94 & 0.43 & 0.28 & 4 & 5.25 & 0.19 & 0.16 & 2 + 13 & al & 3.02 & 0.56 & 0.09 & 1 & 3.27 & 0.57 & 0.17 & 1 & 2.65 & 0.75 & 0.17 & 1 & 3.11 & 0.84 & 0.14 & 1 + 14 & si & 5.26 & 0.60 & 0.09 & 1 & 5.46 & 0.54 & 0.17 & 1 & 5.06 & 0.58 & 0.17 & 1 & 5.52 & 0.49 & 0.14 & 1 + 20 & ca & 3.97 & 0.50 & 0.01 & 3 & 4.09 & 0.36 & 0.11 & 3 & 3.66 & 0.37 & 0.08 & 4 & 4.13 & 0.29 & 0.08 & 2 + 21 & sc & 0.45 & 0.17 & 0.03 & 3 & 0.51 & 0.03 & 0.03 & 3 & 0.14 & 0.04 & 0.07 & 3 & 0.59 & 0.06 & 0.08 & 2 + 22 & ti & 2.55 & 0.42 & 0.09 & 21 & 2.62 & 0.23 & 0.10 & 12 & 2.22 & 0.27 & 0.12 & 17 & 2.58 & 0.08 & 0.12 & 21 + 23 & v & 1.24 & 0.13 & 0.17 & 3 & 1.42 & 0.05 & 0.22 & 3 & 1.34 & 0.41 & 0.17 & 2 & & & & + 24 & cr & 2.97 & 0.19 & 0.09 & 2 & 3.20 & 0.16 & 0.17 & 2 & 2.73 & 0.13 & 0.17 & 2 & 3.03 & 0.12 & 0.14 & 1 + 25 & mn & 2.08 & 0.42 & 0.13 & 6 & 2.49 & 0.27 & 0.17 & 6 & 1.91 & 0.41 & 0.25 & 6 & 2.67 & 0.20 & 0.19 & 4 + 27 & co & 2.32 & 0.29 & 0.09 & 2 & 2.48 & 0.19 & 0.17 & 2 & 1.77 & 0.08 & 0.17 & 2 & 2.28 & 0.12 & 0.14 & 2 + 28 & ni & 3.33 & 0.03 & 0.09 & 2 & 3.40 & 0.22 & 0.17 & 2 & 2.75 & 0.43 & 0.17 & 2 & 3.54 & 0.19 & 0.14 & 2 + 29 & cu & & & & & & & & & 0.39 & 0.75 & 0.17 & 1 & 1.02 & 0.67 & 0.14 & 1 + 38 & sr & 0.08 & 0.11 & 0.09 & 1 & 0.36 & 0.07 & 0.33 & 3 & 0.35 & 0.20 & 0.17 & 2 & 0.34 & 0.06 & 0.14 & 2 + 39 & y & 0.99 & 0.33 & 0.18 & 6 & 0.57 & 0.17 & 0.06 & 8 & 0.89 & 0.05 & 0.17 & 5 & 0.40 & 0.11 & 0.16 & 5 + 40 & zr & 0.22 & 0.06 & 0.09 & 2 & 0.10 & 0.12 & 0.20 & 4 & 0.15 & 0.31 & 0.17 & 3 & 0.25 & 0.16 & 0.10 & 3 + 45 & ru & & & & & & & & & & & & & & & & + 46 & pd & & & & & 0.98 & 0.05 & 0.17 & 1 & & & & & & & & + 56 & ba & 1.84 & 1.17 & 0.09 & 2 & 0.45 & 0.04 & 0.17 & 2 & 1.90 & 1.05 & 0.17 & 2 & 0.82 & 0.52 & 0.14 & 2 + 57 & la & 2.55 & 0.88 & 0.09 & 1 & 1.32 & 0.09 & 0.05 & 5 & & & & & 1.76 & 0.46 & 0.14 & 5 + 58 & ce & & & & & 0.88 & 0.12 & 0.11 & 8 & & & & & 1.45 & 0.56 & 0.14 & 7 + 59 & pr & & & & & 1.54 & 0.29 & 0.17 & 2 & & & & & 2.32 & 0.60 & 0.14 & 1 + 60 & nd & & & & & 0.89 & 0.25 & 0.08 & 8 & & & & & 1.33 & 0.30 & 0.14 & 7 + 62 & sm & & & & & 1.27 & 0.38 & 0.21 & 3 & & & & & 1.63 & 0.09 & 0.23 & 4 + 63 & eu & 2.96 & 0.62 & 0.09 & 2 & 1.56 & 0.52 & 0.03 & 3 & & & & & 1.91 & 0.06 & 0.05 & 3 + 64 & gd & & & & & 1.01 & 0.53 & 0.17 & 2 & & & & & 1.85 & 0.42 & 0.14 & 1 + 65 & tb & & & & & 2.15 & 0.13 & 0.17 & 1 & & & & & 2.72 & 0.55 & 0.14 & 2 + 66 & dy & & & & & 1.00 & 0.46 & 0.14 & 5 & & & & & 1.44 & 0.09 & 0.24 & 4 + 68 & er & & & & & 1.20 & 0.46 & 0.13 & 3 & & & & & 1.75 & 0.20 & 0.08 & 3 + 69 & tm & & & & & 2.22 & 0.26 & 0.07 & 3 & & & & & 2.72 & 0.35 & 0.14 & 2 + 70 & yb & & & & & 1.33 & 0.34 & 0.17 & 1 & & & & & 1.70 & 0.14 & 0.14 & 1 + 76 & os & & & & & & & & & & & & & & & & + 77 & ir & & & & & 0.87 & 0.39 & 0.17 & 1 & & & & & & & & + 90 & th & & & & & 1.93 & 0.61 & 0.10 & 1 & & & & & 2.50 & 0.07 & 0.15 & 1 + 92 & u & & & & & 2.63 & @xmath1660.50 & & 1 & & & & & 2.52 & @xmath1660.50 & & 1 + + [ fe / h ] and @xmath63c/@xmath62c are given for the first and second lines , respectively @c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c & & & & & + & & & & & & & & & & & & & & & & & + & fe / h & 4.66 & 2.85 & 0.15 & 76 & 4.74 & 2.77 & 0.19 & 84 & & 2.78 & 0.12 & 80 & 4.98 & 2.53 & 0.08 & 78 + & @xmath63c/@xmath62c & & 7 & & & & 5 & & & & @xmath13320 & & & & @xmath13320 & & + 6 & c & 5.30 & 0.41 & & & 5.38 & 0.41 & & & 5.97 & 0.19 & & & 6.31 & 0.28 & & + 12 & mg & 5.12 & 0.39 & 0.03 & 4 & 5.35 & 0.54 & 0.06 & 6 & 5.22 & 0.42 & 0.1 & 6 & 5.26 & 0.21 & 0.13 & 5 + 13 & al & 3.14 & 0.48 & 0.11 & 1 & 3.30 & 0.40 & 0.12 & 1 & 2.98 & 0.71 & 0.12 & 1 & 2.92 & 1.02 & 0.09 & 1 + 14 & si & 5.05 & 0.35 & 0.11 & 1 & 5.22 & 0.44 & 0.12 & 1 & 5.33 & 0.56 & 0.12 & 1 & 5.26 & 0.24 & 0.09 & 1 + 20 & ca & 3.78 & 0.27 & 0.06 & 4 & 3.89 & 0.30 & 0.10 & 4 & 3.94 & 0.36 & 0.11 & 4 & 4.08 & 0.25 & 0.04 & 4 + 21 & sc & 0.30 & 0.02 & 0.00 & 3 & 0.40 & 0.00 & 0.07 & 3 & 0.52 & 0.13 & 0.03 & 2 & 0.66 & 0.02 & 0.04 & 3 + 22 & ti & 2.47 & 0.30 & 0.10 & 26 & 2.45 & 0.20 & 0.12 & 27 & & 0.41 & 0.29 & 26 & 2.72 & 0.23 & 0.08 & 22 + 23 & v & 1.10 & 0.05 & 0.15 & 4 & 1.28 & 0.05 & 0.13 & 4 & 1.45 & 0.23 & 0.12 & 2 & 1.55 & 0.07 & 0.08 & 3 + 24 & cr & 2.88 & 0.06 & 0.17 & 3 & 3.05 & 0.15 & 0.15 & 3 & 3.36 & 0.47 & 0.15 & 3 & 3.35 & 0.21 & 0.15 & 3 + 25 & mn & 2.01 & 0.53 & 0.12 & 6 & 2.46 & 0.16 & 0.09 & 6 & 2.21 & 0.40 & 0.11 & 6 & 2.59 & 0.27 & 0.10 & 6 + 27 & co & 2.26 & 0.19 & 0.12 & 3 & 2.33 & 0.18 & 0.16 & 3 & 2.44 & 0.3 & 0.13 & 3 & 2.65 & 0.26 & 0.03 & 4 + 28 & ni & 3.29 & 0.11 & 0.11 & 2 & 3.33 & 0.15 & 0.12 & 2 & 3.35 & 0.12 & 0.12 & 2 & 3.76 & 0.04 & 0.09 & 2 + 29 & cu & & & & & 0.66 & 0.78 & 0.12 & 1 & & & & & & & & + 38 & sr & 0.11 & 0.04 & 0.11 & 2 & 0.04 & 0.11 & 0.12 & 1 & 0.21 & 0.07 & 0.12 & 1 & 0.03 & 0.42 & 0.09 & 1 + 39 & y & 0.78 & 0.16 & 0.08 & 7 & 0.87 & 0.33 & 0.12 & 6 & 0.65 & 0.10 & 0.1 & 5 & 0.84 & 0.54 & 0.18 & 3 + 40 & zr & 0.06 & 0.18 & 0.17 & 4 & 0.19 & 0.03 & 0.11 & 3 & 0.16 & 0.33 & 0.12 & 2 & 0.15 & 0.23 & 0.09 & 2 + 45 & ru & & & & & & & & & & & & & & & & + 46 & pd & 1.06 & 0.09 & 0.11 & 1 & & & & & & & & & & & & + 56 & ba & 0.49 & 0.14 & 0.11 & 2 & 1.76 & 1.21 & 0.12 & 2 & 0.66 & 0.10 & 0.12 & 2 & 1.37 & 1.06 & 0.09 & 2 + 57 & la & 1.53 & 0.10 & 0.05 & 5 & 2.44 & 0.89 & 0.12 & 5 & 1.59 & 0.03 & 0.07 & 4 & & & & + 58 & ce & 1.10 & 0.12 & 0.10 & 8 & & & & & 1.08 & 0.07 & 0.12 & 5 & & & & + 59 & pr & 1.49 & 0.56 & 0.11 & 2 & & & & & 1.29 & 0.69 & 0.12 & 1 & & & & + 60 & nd & 1.06 & 0.30 & 0.06 & 7 & 1.95 & 0.67 & 0.12 & 1 & 1.07 & 0.22 & 0.19 & 6 & & & & + 62 & sm & 1.24 & 0.63 & 0.19 & 4 & & & & & 1.27 & 0.53 & 0.12 & 2 & & & & + 63 & eu & 1.64 & 0.66 & 0.03 & 3 & 2.60 & 0.38 & 0.12 & 1 & 1.70 & 0.53 & 0.04 & 3 & & & & + 64 & gd & 1.10 & 0.66 & 0.21 & 3 & 2.31 & 0.63 & 0.12 & 1 & 1.30 & 0.39 & 0.12 & 1 & & & & + 65 & tb & 2.29 & 0.21 & 0.11 & 1 & & & & & & & & & & & & + 66 & dy & 1.07 & 0.61 & 0.09 & 5 & & & & & 1.19 & 0.42 & 0.12 & 2 & & & & + 68 & er & 1.26 & 0.62 & 0.09 & 4 & & & & & & & & & & & & + 69 & tm & 2.18 & 0.52 & 0.11 & 2 & & & & & & & & & & & & + 70 & yb & 1.46 & 0.43 & 0.11 & 1 & & & & & & & & & & & & + 76 & os & & & & & & & & & & & & & & & & + 77 & ir & 0.89 & 0.59 & 0.11 & 1 & & & & & & & & & & & & + 90 & th & 1.97 & 0.79 & 0.15 & 1 & & & & & & & & & & & & + 92 & u & 2.35 & @xmath1661.00 & & 1 & & & & & & & & & & & & + + @c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c & & & & & + & & & & & & & & & & & & & & & & & + & fe / h & 5.01 & 2.50 & 0.12 & 78 & 4.65 & 2.86 & 0.15 & 76 & 4.60 & 2.91 & 0.10 & 82 & 4.48 & 3.03 & 0.14 & 79 + & @xmath63c/@xmath62c & & 6 & & & & @xmath1337 & & & & @xmath1333 & & & & @xmath1335 & & + 6 & c & 5.81 & 0.25 & & & 5.99 & 0.29 & & & & @xmath1660.52 & & & 5.74 & 0.21 & & + 12 & mg & 5.50 & 0.42 & 0.15 & 4 & 5.00 & 0.28 & 0.15 & 7 & 5.28 & 0.61 & 0.17 & 7 & 4.91 & 0.36 & 0.24 & 4 + 13 & al & 3.36 & 0.61 & 0.14 & 1 & 2.67 & 0.94 & 0.16 & 1 & 3.03 & 0.53 & 0.10 & 1 & 2.67 & 0.77 & 0.12 & 1 + 14 & si & 5.48 & 0.43 & 0.14 & 1 & 5.05 & 0.36 & 0.16 & 1 & 5.57 & 0.93 & 0.10 & 1 & 4.92 & 0.40 & 0.12 & 1 + 20 & ca & 4.25 & 0.39 & 0.13 & 4 & 3.81 & 0.31 & 0.04 & 4 & 3.83 & 0.38 & 0.03 & 4 & 3.63 & 0.30 & 0.16 & 4 + 21 & sc & 0.70 & 0.03 & 0.09 & 3 & 0.30 & 0.01 & 0.06 & 2 & 0.56 & 0.30 & 0.07 & 2 & 0.18 & 0.04 & 0.05 & 2 + 22 & ti & 2.81 & 0.29 & 0.12 & 24 & 2.40 & 0.24 & 0.12 & 22 & 2.33 & 0.22 & 0.09 & 24 & 2.22 & 0.23 & 0.10 & 21 + 23 & v & 1.64 & 0.14 & 0.15 & 3 & 1.16 & 0.02 & 0.16 & 2 & 1.34 & 0.25 & 0.10 & 2 & 1.04 & 0.07 & 0.12 & 2 + 24 & cr & 3.34 & 0.17 & 0.14 & 2 & 2.81 & 0.00 & 0.16 & 3 & 2.82 & 0.06 & 0.10 & 2 & & & & + 25 & mn & 2.65 & 0.24 & 0.11 & 6 & 2.22 & 0.31 & 0.10 & 6 & 2.47 & 0.01 & 0.15 & 6 & 1.83 & 0.53 & 0.15 & 4 + 27 & co & 2.50 & 0.08 & 0.14 & 2 & 2.38 & 0.32 & 0.13 & 3 & 2.29 & 0.28 & 0.09 & 3 & 2.23 & 0.34 & 0.07 & 3 + 28 & ni & 3.42 & 0.33 & 0.14 & 2 & 3.20 & 0.19 & 0.16 & 2 & 3.54 & 0.20 & 0.10 & 2 & 3.24 & 0.02 & 0.12 & 2 + 29 & cu & 1.03 & 0.68 & 0.14 & 1 & & & & & & & & & & & & + 38 & sr & 0.37 & 0.05 & 0.14 & 2 & 0.67 & 0.73 & 0.16 & 1 & 1.70 & 1.71 & 0.10 & 1 & 0.12 & 0.23 & 0.12 & 1 + 39 & y & 0.36 & 0.09 & 0.08 & 5 & 1.19 & 0.56 & 0.25 & 4 & & & & & 0.88 & 0.08 & 0.00 & 4 + 40 & zr & 0.39 & 0.28 & 0.23 & 3 & 0.26 & 0.01 & 0.36 & 4 & & & & & 0.12 & 0.30 & 0.12 & 2 + 45 & ru & & & & & & & & & & & & & & & & + 46 & pd & 0.73 & 0.07 & 0.14 & 1 & & & & & & & & & & & & + 56 & ba & 0.51 & 0.23 & 0.14 & 2 & 1.61 & 0.97 & 0.16 & 2 & & 1.56 & 0.10 & 2 & 1.93 & 1.12 & 0.12 & 2 + 57 & la & 1.21 & 0.07 & 0.06 & 5 & & & & & & & & & & & & + 58 & ce & 0.75 & 0.12 & 0.12 & 8 & & & & & & & & & & & & + 59 & pr & 1.45 & 0.25 & 0.14 & 1 & & & & & & & & & & & & + 60 & nd & 0.75 & 0.26 & 0.09 & 6 & & & & & & & & & & & & + 62 & sm & 0.98 & 0.54 & 0.20 & 4 & & & & & & & & & & & & + 63 & eu & 1.34 & 0.61 & 0.06 & 3 & & & & & & & & & & & & + 64 & gd & 0.88 & 0.53 & 0.14 & 2 & & & & & & & & & & & & + 65 & tb & & & & & & & & & & & & & & & & + 66 & dy & 0.78 & 0.55 & 0.02 & 4 & & & & & & & & & & & & + 68 & er & 1.03 & 0.50 & 0.22 & 4 & & & & & & & & & & & & + 69 & tm & 1.92 & 0.43 & 0.12 & 3 & & & & & & & & & & & & + 70 & yb & 1.27 & 0.27 & 0.14 & 1 & & & & & & & & & & & & + 76 & os & & & & & & & & & & & & & & & & + 77 & ir & & & & & & & & & & & & & & & & + 90 & th & 1.85 & 0.56 & 0.15 & 1 & & & & & & & & & & & & + 92 & u & 2.00 & @xmath1661.00 & & 1 & & & & & & & & & & & & + + @c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c & & & & & + & & & & & & & & & & & & & & & & & + & fe / h & 4.39 & 3.12 & 0.23 & 69 & 4.60 & 2.91 & 0.14 & 80 & 4.52 & 3.09 & 0.15 & 63 & 5.09 & 2.42 & 0.10 & 55 + & @xmath63c/@xmath62c & & @xmath1333 & & & & @xmath1335 & & & & @xmath1337 & & & & @xmath1333 & & + 6 & c & & @xmath1660.07 & & & 5.42 & 0.23 & & & 6.39 & 0.92 & & & 6.67 & 0.53 & & + 12 & mg & 4.67 & 0.21 & 0.28 & 6 & 5.06 & 0.39 & 0.35 & 4 & 4.87 & 0.38 & 0.32 & 4 & 5.43 & 0.27 & 0.26 & 5 + 13 & al & 2.64 & 0.71 & 0.19 & 1 & 2.74 & 0.83 & 0.15 & 1 & 2.53 & 0.85 & 0.16 & 1 & 3.13 & 0.92 & 0.09 & 1 + 14 & si & 4.42 & 0.01 & 0.19 & 1 & 4.93 & 0.29 & 0.15 & 1 & 4.84 & 0.38 & 0.16 & 1 & 5.45 & 0.32 & 0.09 & 1 + 20 & ca & 3.21 & 0.03 & 0.18 & 4 & 3.76 & 0.31 & 0.07 & 4 & 3.73 & 0.46 & 0.17 & 4 & 4.22 & 0.28 & 0.07 & 6 + 21 & sc & 0.23 & 0.28 & 0.07 & 2 & 0.04 & 0.30 & 0.13 & 2 & 0.45 & 0.53 & 0.16 & 1 & 0.92 & 0.17 & 0.07 & 2 + 22 & ti & 2.24 & 0.34 & 0.13 & 23 & 2.24 & 0.13 & 0.09 & 23 & 2.35 & 0.42 & 0.12 & 16 & 2.99 & 0.39 & 0.07 & 17 + 23 & v & 1.14 & 0.26 & 0.19 & 1 & 0.93 & 0.16 & 0.15 & 2 & & & & & 1.92 & 0.34 & 0.09 & 1 + 24 & cr & 2.53 & 0.02 & 0.19 & 2 & 2.68 & 0.08 & 0.15 & 2 & & & & & 3.02 & 0.23 & 0.09 & + 25 & mn & 2.50 & 0.23 & 0.20 & 6 & 2.24 & 0.24 & 0.09 & 6 & 1.56 & 0.74 & 0.16 & 2 & 2.49 & 0.48 & 0.06 & 3 + 27 & co & 2.07 & 0.27 & 0.23 & 3 & 2.14 & 0.13 & 0.11 & 3 & 2.19 & 0.36 & 0.06 & 3 & 2.89 & 0.39 & 0.04 & 3 + 28 & ni & 3.51 & 0.38 & 0.19 & 2 & 3.09 & 0.25 & 0.15 & 2 & 3.07 & 0.09 & 0.16 & 2 & 3.88 & 0.05 & 0.09 & 2 + 29 & cu & & & & & & & & & & & & & & & & + 38 & sr & 0.63 & 0.43 & 0.19 & 1 & 0.22 & 0.23 & 0.15 & 1 & 0.11 & 0.28 & 0.16 & 2 & 0.66 & 0.16 & 0.09 & 1 + 39 & y & & & & & 1.15 & 0.47 & 0.13 & 5 & 0.70 & 0.16 & 0.16 & 2 & 0.02 & 0.17 & 0.12 & 3 + 40 & zr & & & & & 0.55 & 0.25 & 0.25 & 3 & & & & & & & & + 45 & ru & & & & & & & & & & & & & & & & + 46 & pd & & & & & & & & & & & & & & & & + 56 & ba & 2.73 & 1.83 & 0.19 & 2 & 1.85 & 1.16 & 0.15 & 2 & 1.46 & 0.59 & 0.16 & 2 & 0.53 & 0.33 & 0.09 & 2 + + @c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c & & & & & + & & & & & & & & & & & & & & & & & + & fe / h & 4.79 & 2.72 & 0.17 & 35 & 4.58 & 2.93 & 0.20 & 49 & 4.57 & 2.94 & 0.26 & & 5.09 & 2.42 & 0.13 & + & @xmath63c/@xmath62c & & @xmath1333 & & & & @xmath1333 & & & & & & & & @xmath13320 & & + 6 & c & 5.59 & 0.25 & & & 6.05 & 0.42 & & & 5.40 & 0.22 & & & 6.48 & 0.34 & & + 12 & mg & 4.80 & 0.06 & 0.50 & 4 & 5.07 & 0.42 & 0.32 & 5 & 4.45 & 0.19 & 0.21 & 5 & 5.49 & 0.33 & 0.19 & 7 + 13 & al & 2.64 & 1.11 & 0.19 & 1 & 2.98 & 0.56 & 0.15 & 1 & 2.85 & 0.68 & 0.19 & 1 & 3.41 & 0.64 & 0.14 & 1 + 14 & si & 5.15 & 0.32 & 0.19 & 1 & 5.36 & 0.74 & 0.15 & 1 & 5.00 & 0.39 & 0.19 & 1 & 5.64 & 0.51 & 0.14 & 1 + 20 & ca & & & & & 3.72 & 0.29 & 0.15 & 4 & 3.45 & 0.03 & 0.16 & 2 & 4.28 & 0.34 & 0.03 & 3 + 21 & sc & 0.12 & 0.33 & 0.05 & 2 & 0.54 & 0.30 & 0.05 & 2 & 0.13 & 0.36 & 0.06 & 2 & 0.87 & 0.12 & 0.14 & 2 + 22 & ti & 2.17 & 0.13 & 0.06 & 13 & 2.47 & 0.38 & 0.03 & 16 & 1.96 & 0.12 & 0.06 & 13 & 2.95 & 0.35 & 0.03 & 25 + 23 & v & & & & & 1.75 & 0.68 & 0.15 & 1 & 1.03 & 0.03 & 0.19 & 1 & 1.79 & 0.21 & 0.14 & 2 + 24 & cr & & & & & & & & & & & & & 3.31 & 0.06 & 0.14 & 2 + 25 & mn & 2.76 & 0.09 & 0.19 & 2 & 2.01 & 0.45 & 0.27 & 3 & 2.11 & 0.34 & 0.39 & 3 & 2.84 & 0.13 & 0.23 & 6 + 27 & co & 1.98 & 0.22 & 0.15 & 3 & 2.36 & 0.37 & 0.11 & 4 & 2.04 & 0.06 & 0.19 & 2 & 2.74 & 0.24 & 0.14 & 3 + 28 & ni & 3.12 & 0.41 & 0.19 & 2 & 3.44 & 0.12 & 0.15 & 2 & 3.20 & 0.11 & 0.19 & 2 & 3.57 & 0.26 & 0.14 & 2 + 29 & cu & & & & & & & & & & & & & & & & + 38 & sr & 0.64 & 0.84 & 0.19 & 2 & 0.09 & 0.10 & 0.15 & 1 & 0.73 & 0.71 & 0.19 & 2 & 0.64 & 0.14 & 0.14 & 2 + 39 & y & & & & & 0.49 & 0.21 & 0.15 & 2 & & & & & 0.07 & 0.12 & 0.12 & 6 + 40 & zr & & & & & & & & & & & & & 0.65 & 0.46 & 0.14 & 4 + 45 & ru & & & & & & & & & & & & & & & & + 46 & pd & & & & & & & & & & & & & & & & + 56 & ba & 2.21 & 1.71 & 0.19 & 2 & 0.33 & 0.38 & 0.15 & 2 & 2.49 & 1.77 & 0.19 & 2 & 0.02 & 0.22 & 0.14 & 2 + 57 & la & & & & & & & & & & & & & 0.78 & 0.42 & 0.06 & 5 + 58 & ce & & & & & & & & & & & & & 0.31 & 0.48 & 0.26 & 5 + 59 & pr & & & & & & & & & & & & & 0.65 & 0.97 & 0.14 & 1 + 60 & nd & & & & & & & & & & & & & 0.38 & 0.55 & 0.12 & 8 + 62 & sm & & & & & & & & & & & & & 0.82 & 0.62 & 0.14 & 2 + 63 & eu & & & & & 1.22 & 1.16 & 0.08 & 3 & & & & & 1.02 & 0.85 & 0.05 & 3 + 64 & gd & & & & & & & & & & & & & 0.37 & 0.96 & 0.14 & 2 + 65 & tb & & & & & & & & & & & & & & & & + 66 & dy & & & & & & & & & & & & & 0.43 & 0.82 & 0.11 & 3 + 68 & er & & & & & & & & & & & & & 0.62 & 0.83 & 0.22 & 4 + 69 & tm & & & & & & & & & & & & & 1.38 & 0.89 & 0.14 & 2 + 70 & yb & & & & & & & & & & & & & 0.81 & 0.65 & 0.14 & 1 + 76 & os & & & & & & & & & & & & & & & & + 77 & ir & & & & & & & & & & & & & & & & + 90 & th & & & & & & & & & & & & & 1.12 & 1.21 & 0.15 & 1 + 92 & u & & & & & & & & & & & & & 1.42 & @xmath1661.50 & & 1 + + @c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c@c & & & + & & & & & & & & & + & fe / h & 4.70 & 2.81 & 0.12 & & 4.59 & 2.92 & 0.14 & + & @xmath63c/@xmath62c & & @xmath13310 & & & & 20 & & + 6 & c & 5.84 & 0.09 & & & 6.55 & 0.91 & & + 12 & mg & 5.43 & 0.66 & 0.11 & 4 & 4.93 & 0.27 & 0.17 & 7 + 13 & al & 3.05 & 0.61 & 0.19 & 1 & 2.94 & 0.61 & 0.15 & 1 + 14 & si & 5.36 & 0.62 & 0.19 & 1 & 4.87 & 0.24 & 0.15 & 1 + 20 & ca & 3.97 & 0.42 & 0.19 & 1 & 3.70 & 0.26 & 0.09 & 4 + 21 & sc & 0.19 & 0.17 & 0.19 & 1 & 0.17 & 0.08 & 0.01 & 2 + 22 & ti & 2.55 & 0.34 & 0.05 & 15 & 2.18 & 0.08 & 0.03 & 19 + 23 & v & 1.47 & 0.28 & 0.19 & 2 & 1.07 & 0.01 & 0.15 & 1 + 24 & cr & & & & & & & & + 25 & mn & 2.41 & 0.17 & 0.24 & 5 & 2.14 & 0.33 & 0.12 & 4 + 27 & co & 2.49 & 0.38 & 0.13 & 3 & 2.12 & 0.12 & 0.05 & 3 + 28 & ni & 3.64 & 0.20 & 0.19 & 1 & 3.05 & 0.28 & 0.15 & 2 + 29 & cu & & & & & & & & + 38 & sr & 0.58 & 0.47 & 0.19 & 2 & 0.44 & 0.44 & 0.15 & 2 + 39 & y & 0.22 & 0.36 & 0.46 & 6 & 0.23 & 0.46 & 0.22 & 7 + 40 & zr & 0.52 & 0.72 & 0.33 & 4 & 0.29 & 0.60 & 0.18 & 3 + 45 & ru & & & & & 0.13 & 1.22 & 0.15 & 1 + 46 & pd & & & & & & & & + 56 & ba & 0.43 & 1.02 & 0.19 & 2 & 0.22 & 0.92 & 0.15 & 2 + 57 & la & 0.48 & 1.11 & 0.06 & 6 & 0.84 & 0.86 & 0.08 & 5 + 58 & ce & 0.19 & 0.99 & 0.09 & 7 & 0.38 & 0.91 & 0.08 & 7 + 59 & pr & 0.63 & 1.38 & 0.19 & 1 & 1.09 & 1.03 & 0.15 & 1 + 60 & nd & 0.06 & 1.26 & 0.12 & 9 & 0.34 & 1.09 & 0.12 & 9 + 62 & sm & 0.33 & 1.50 & 0.12 & 4 & 0.65 & 1.29 & 0.19 & 4 + 63 & eu & 0.59 & 1.67 & 0.01 & 3 & 0.86 & 1.51 & 0.02 & 3 + 64 & gd & 0.18 & 1.54 & 0.19 & 2 & 0.40 & 1.43 & 0.34 & 3 + 65 & tb & 1.16 & 1.30 & 0.19 & 2 & 1.31 & 1.26 & 0.32 & 3 + 66 & dy & 0.00 & 1.64 & 0.13 & 5 & 0.21 & 1.54 & 0.14 & 5 + 68 & er & 0.22 & 1.62 & 0.26 & 4 & 0.45 & 1.50 & 0.14 & 5 + 69 & tm & 1.20 & 1.46 & 0.19 & 3 & 1.41 & 1.36 & 0.05 & 4 + 70 & yb & 0.32 & 1.54 & 0.19 & 1 & 0.67 & 1.29 & 0.15 & 1 + 76 & os & 0.46 & 1.90 & 0.19 & 1 & & & & + 77 & ir & & & & & & & & + 90 & th & 0.92 & 1.80 & 0.10 & 1 & 1.42 & 1.41 & 0.15 & 1 + 92 & u & 1.96 & @xmath1661.35 & & 1 & 1.92 & @xmath1661.50 & & 1 + + lcccccccccccc species & & & & & & & & r.s.s . + & @xmath167 & @xmath168 & & @xmath169 & @xmath170 & & @xmath115 & @xmath171 & & @xmath171 & @xmath115 & + & + 0.12 & @xmath1720.13 & & + 0.08 & @xmath1720.08 & & @xmath1720.03 & + 0.01 & & 0.07 & @xmath1720.09 & 0.17 + & + 0.11 & @xmath1720.13 & & + 0.06 & @xmath1720.06 & & @xmath1720.07 & + 0.03 & & 0.31 & @xmath1720.33 & 0.36 + & + 0.12 & @xmath1720.12 & & + 0.02 & @xmath1720.01 & & @xmath1720.03 & + 0.01 & & 0.12 & @xmath1720.09 & 0.15 + & + 0.08 & @xmath1720.08 & & + 0.02 & @xmath1720.01 & & @xmath1720.02 & + 0.01 & & + 0.11 & @xmath1720.08 & 0.11 + & + 0.06 & @xmath1720.06 & & @xmath1720.09 & + 0.10 & & + 0.02 & + 0.00 & & + 0.12 & @xmath1720.07 & 0.13 + & + 0.13 & @xmath1720.14 & & + 0.02 & @xmath1720.02 & & @xmath1720.03 & + 0.01 & & + 0.08 & @xmath1720.05 & 0.14 + & + 0.05 & @xmath1720.05 & & @xmath1720.08 & + 0.09 & & + 0.00 & + 0.01 & & + 0.26 & @xmath1720.19 & 0.21 + & + 0.14 & @xmath1720.14 & & + 0.03 & @xmath1720.01 & & @xmath1720.02 & + 0.00 & & + 0.02 & @xmath1720.01 & 0.14 + & + 0.05 & @xmath1720.04 & & @xmath1720.09 & + 0.10 & & + 0.02 & @xmath1720.01 & & + 0.03 & @xmath1720.02 & 0.12 + & + 0.14 & @xmath1720.16 & & + 0.04 & @xmath1720.04 & & @xmath1720.07 & + 0.03 & & + 0.44 & @xmath1720.31 & 0.35 + & + 0.00 & + 0.01 & & @xmath1720.11 & + 0.11 & & + 0.01 & 0.00 & & + 0.01 & @xmath1720.01 & 0.11 + & + 0.13 & @xmath1720.15 & & + 0.03 & @xmath1720.03 & & @xmath1720.05 & + 0.02 & & + 0.23 & @xmath1720.17 & 0.22 + & + 0.12 & @xmath1720.14 & & + 0.05 & @xmath1720.04 & & @xmath1720.04 & + 0.02 & & + 0.17 & @xmath1720.16 & 0.21 + & + 0.13 & @xmath1720.15 & & + 0.03 & @xmath1720.03 & & @xmath1720.05 & + 0.01 & & + 0.36 & @xmath1720.20 & 0.25 + & + 0.14 & @xmath1720.18 & & + 0.05 & @xmath1720.05 & & @xmath1720.09 & + 0.03 & & + 0.46 & @xmath1720.42 & 0.46 + & + 0.05 & @xmath1720.07 & & @xmath1720.06 & + 0.05 & & @xmath1720.01 & + 0.02 & & + 0.16 & @xmath1720.22 & 0.23 + & + 0.07 & @xmath1720.07 & & @xmath1720.09 & + 0.09 & & + 0.02 & 0.00 & & + 0.15 & @xmath1720.08 & 0.14 + & + 0.06 & @xmath1720.07 & & @xmath1720.09 & + 0.09 & & + 0.02 & @xmath1720.01 & & + 0.05 & @xmath1720.03 & 0.12 + & + 0.11 & @xmath1720.12 & & @xmath1720.07 & + 0.06 & & @xmath1720.03 & + 0.04 & & + 0.46 & @xmath1720.37 & 0.39 + & + 0.08 & @xmath1720.08 & & @xmath1720.09 & + 0.10 & & + 0.02 & @xmath1720.01 & & + 0.04 & @xmath1720.02 & 0.13 + & + 0.08 & @xmath1720.08 & & @xmath1720.09 & + 0.10 & & + 0.03 & @xmath1720.01 & & + 0.02 & @xmath1720.01 & 0.13 + & + 0.08 & @xmath1720.08 & & @xmath1720.09 & + 0.10 & & + 0.03 & @xmath1720.01 & & + 0.03 & @xmath1720.02 & 0.13 + & + 0.09 & @xmath1720.08 & & @xmath1720.09 & + 0.11 & & + 0.03 & @xmath1720.01 & & + 0.01 & + 0.00 & 0.14 + & + 0.08 & @xmath1720.09 & & @xmath1720.09 & + 0.09 & & + 0.02 & @xmath1720.02 & & + 0.01 & @xmath1720.01 & 0.14 + & + 0.08 & @xmath1720.09 & & @xmath1720.09 & + 0.10 & & + 0.02 & @xmath1720.01 & & + 0.05 & @xmath1720.03 & 0.13 + & + 0.08 & @xmath1720.08 & & @xmath1720.09 & + 0.10 & & + 0.03 & @xmath1720.01 & & + 0.02 & @xmath1720.01 & 0.13 + & + 0.10 & @xmath1720.09 & & @xmath1720.10 & + 0.10 & & + 0.00 & @xmath1720.02 & & + 0.00 & + 0.00 & 0.14 + lcclrclr object name & & & & @xmath142 + & @xmath140 & @xmath72 & & @xmath140 & @xmath72 & + hd 6268 & 2.08 & 0.18 & & 2.34 & 0.18 & 0.13 + hd 110184 & 2.58 & 0.24 & & 2.84 & 0.22 & 0.15 + hd 115444 & 2.16 & 0.18 & & 2.42 & 0.22 & 0.11 + hd 186478 & 1.89 & 0.10 & & 2.17 & 0.25 & 0.12 + cs 22892052 & 1.43 & 0.13 & & 1.69 & 0.24 & 0.15 + cs 30306132 & 1.45 & 0.15 & & 1.73 & 0.24 & 0.14 + cs 31082001 & 1.16 & 0.12 & & 1.42 & 0.23 & 0.15 + ch lines ( 4223.9 ) exist redward of @xmath63ch lines .
synthetic spectra are shown for three isotope ratios : @xmath63c/@xmath62c = 2 , 4 , and 6 for hd 6268 , @xmath63c/@xmath62c = 3 , 5 , and 7 for hd 122563 , and @xmath63c/@xmath62c = 10 , 20 , and 50 for cs 22892052 and cs 30306132.,width=529 ] ( th ) @xmath175 .
three dashed lines indicate the synthetic spectra with three different th abundances with a step of 0.1 dex ( hd 6268 ) or 0.15 dex ( hd 110184 , hd 115444 , and hd 122563 ) . the adopted th
abundances are log@xmath176(th ) @xmath177 for hd 6268 , and log@xmath176(th ) @xmath178 for hd 110184 , log@xmath176(th ) @xmath179 for hd 115444 . in the spectrum of hd 122563
, the th line is not detected , and the dashed lines show synthetic spectra with log@xmath176(th ) @xmath180 and @xmath181.,width=453 ] values as a function of [ fe / h ] .
our results are shown by filled circles with error bars , while results by previous studies ( compilation of literature data taken from norris et al .
2001 , and mcwilliam 1998 , burris et al .
2000 ) are shown by open circles .
the error bars shown here are random and systematic errors for [ fe / h ] , and rss of random errors [ ba / h ] and [ fe / h ] for [ ba / fe ] . , width=302 ] as a function of [ fe / h ] for our sample .
the solid line indicates mean value of our [ ba / eu ] .
the dotted line indicates the [ ba / eu ] of the r - process component in the solar system , while the dashed line means that of the s - process - component in the solar system.,width=302 ] ) for our objects as a function of atomic number ( dots ) . the standard deviation of the abundances are shown by bars .
the thin bars indicate the standard deviation of the abundances for the element which is detected in less than 12 objects . , width=302 ] ( upper panel ) , [ y / fe ] ( middle panel ) , and [ zr / fe ] ( lower panel).,title="fig:",width=188 ] ( upper panel ) , [ y / fe ] ( middle panel ) , and [ zr / fe ] ( lower panel).,title="fig:",width=188 ] ( upper panel ) , [ y / fe ] ( middle panel ) , and [ zr / fe ] ( lower panel).,title="fig:",width=188 ] @xmath186 from @xcite ( open circles ) , @xcite ( squares ) , @xcite ( triangles ) , and the present work ( filled circles ) . for the object which was studied by both our study and others , we adopted the result by ours .
the objects which are known to show s - process abundance pattern are excluded from the sample .
the solar values are also plotted .
the filled square indicates the values of the r - process component of solar - system material .
solid lines show the enrichment of sr and ba assuming the initial abundances of sr and ba and a constant sr / ba ratio in the yields of the main r - process .
see text for details.,width=302 ] | we have obtained high - resolution , high s / n near - uv - blue spectra of 22 very metal - poor stars ( [ fe / h ] @xmath0 ) with subaru / hds , and measured the abundances of elements from c to th . the metallicity range of the observed stars is @xmath1 [ fe / h ] @xmath2 .
as found by previous studies , the star - to - star scatter in the measured abundances of neutron - capture elements in these stars is very large , much greater than could be assigned to observational errors , and in comparison with the relatively small scatter in the @xmath3- and iron - peak elements . in spite of the large scatter in the ratios of the neutron - capture elements relative to iron , the abundance _
patterns _ of heavy neutron - capture elements ( @xmath4 ) are quite similar within our sample stars . the ba / eu ratios in the 11 very metal - poor stars in our sample in which both elements have been detected are nearly equal to that of the solar system r - process component .
moreover , the abundance patterns of the heavy neutron - capture elements ( 56 @xmath5 70 ) in seven objects with clear enhancements of the neutron - capture elements are similar to that of the solar system r - process component .
these results prove that heavy neutron - capture elements in these objects are primarily synthesized by the r - process .
in contrast , the abundance ratios of the light neutron - capture elements ( 38 @xmath6 _ z _ @xmath6 46 ) relative to the heavier ones ( 56 @xmath6 _ z _ @xmath6 70 ) exhibit a large dispersion .
our inspection of the correlation between sr and ba abundances in very metal - poor stars reveals that the dispersion of the sr abundances clearly decreases with increasing ba abundance .
this trend is naturally explained by hypothesizing the existence of two processes , one that produces sr without ba , and another that produces sr and ba in similar proportions .
this result should provide a strong constraint on the origin of the light neutron - capture elements at low metallicity .
we have identified a new highly r - process element - enhanced , metal - poor star , cs 22183031 , a giant with [ fe / h ] = @xmath7 and [ eu / fe ] = + 1.2 .
we also identified a new , moderately r - process - enhanced , metal - poor star , cs 30306132 , a giant with [ fe / h ] = @xmath8 and [ eu / fe ] = + 0.85 .
the abundance ratio of the radioactive element th ( _ z _ = 90 ) relative to the stable rare - earth elements ( e.g. , eu ) in very metal - poor stars has been used as a cosmochronometer by a number of previous authors .
thorium is detected in seven stars in our sample , including four objects for which the detection of th has already been reported .
new detections of thorium have been made for the stars hd 6268 , hd 110184 , and cs 30306132 .
the th / eu abundance ratios ( log(th / eu ) ) , are distributed over the range @xmath9 to @xmath10 , with typical errors of 0.10 to 0.15 dex . in particular , the ratios in two stars , cs 31082001 and cs 30306132 , are significantly higher than the ratio in the well - studied object cs 22892052 and those of other moderately r - process - enhanced metal - poor stars previously reported .
since these very metal - poor stars are believed to be formed in the early galaxy , this result suggests that the abundance ratios between th and stable rare - earth elements such as eu , both of which are presumably produced by r - process nucleosynthesis , may exhibit real star - to - star scatter , with implications for ( a ) the astrophysical sites of the r - process , and ( b ) the use of th / eu as a cosmochronometer . |
vs. @xmath2 topologies for sd@xmath3single diffraction , dd@xmath3double diffraction , dpe@xmath3double pomeron exchange , sdd@xmath3sd@xmath4dpe , and exclusive final state production . the hatched areas in the @xmath5-@xmath2 plots represent regions where particle production occurs.,scaledwidth=70.0% ] the phenomenal success of the standard model is tainted by our inability to analytically calculate `` soft '' processes , where the strong coupling constant is large and perturbative techniques fail . yet
, a fundamental understanding of soft processes may help reveal the underlying mechanism of confinement and explain the intricacies of hadron structure .
experimentally , diffractive processes may be used as tools for new discoveries , as they provide low background environments for certain production channels , among which most notable is the exclusive production of higgs bosons ( see sec .
[ sec : dijets ] ) .
the cdf collaboration has reported several results on soft and hard diffraction processes from @xmath6 collisions at the fermilab tevatron using rapidity gaps and / or a leading antiproton as a signature for diffraction ( fig . [
fig : diffraction ] ) .
these results have revealed regularities in the data that point to a qcd picture of diffraction as an exchange of a spin zero color - singlet combination of gluons and / or quarks carrying the quantum numbers of the vacuum ( see review in @xcite ) .
one result that has attracted widespread attention is the observation of a breakdown of qcd factorization in hard diffractive processes , which is expressed as a suppression by a factor of @xmath7(10 ) of the production cross section relative to theoretical expectations .
however , of equal importance is the finding of a breakdown of regge factorization in soft diffraction by a factor of the same magnitude @xcite
. combined , these two results strongly support the hypothesis that the breakdown of factorization is due to a saturation of the probability of forming a rapidity gap by an exchange of a color - neutral construct of the underlying parton distribution function ( pdf ) of the proton , which is historically referred to as `` pomeron '' .
renormalizing the `` gap probability '' to unity over all @xmath8 phase space corrects for the unphysical effect of overlapping diffractive rapidity gaps and leads to an agreement between theory and experiment ( see @xcite and references therein ) .
the gap probability renormalization model is further supported by the following soft - diffraction results obtained by cdf @xcite : * double - diffraction ( central gap ) : same suppression factor as in single - diffraction ; * multi - gap diffraction : double - gap to single - gap ratio non - suppressed ; * energy independence : @xmath9 constant as @xmath10 ; * pomeron intercept and slope : they are related
! @xcite .
similar results are found for hard - diffraction . in this paper
, we review the results obtained in run ii . the cdf ii detector is shown schematically in fig .
[ fig : detector ] ( from ref .
the components of the main detector @xcite used in the diffractive program are the tracking system , the central ( ccal ) , plug ( pcal ) , and forward ( fcal ) calorimeters , and the erenkov luminosity counters ( clc ) .
the diffractive program benefited from dedicated triggers and a system of special forward detectors .
the following forward detectors were employed @xcite : * rps ( roman pot spectrometer ) - detects leading @xmath11 s at @xmath12 ; * mpcal ( miniplug calorimeters ) - measure @xmath13 and @xmath14 at @xmath15 ; * bsc ( beam shower counters ) - identify rapidity gaps at @xmath16 .
there are three classes of results obtained thus far from run ii data : * exclusive dilepton and diphoton production ( see talk by m. albrow ) . * rapidity gaps between jets ( see talk by c. mesropian ) .
* diffractive / exclusive dijet and w / z production ( this talk ) .
diffractive dijet production at the tevatron is suppressed by a factor of @xmath7(10 ) relative to expectations based on the proton pdf extracted from diffractive deep inelastic scattering ( ddis ) at the desy @xmath17 collider hera ( see ref . @xcite . while no ddis suppression is expected in certain models ( see e.g. @xcite ) , the primary exchange in ddis
is a @xmath18 pair , while dijets are mainly produced by a @xmath19 exchange .
the dijet rates at the tevaytron are calculated using a gluon pdf extracted from ddis .
a more direct comparison could be made by measuring the dsf in diffractive @xmath20 production at the tevatron , which is dominated by a @xmath18 exchange as in ddis . in run
i , only the overall diffractive @xmath20 fraction was measured @xcite . in run
ii , cdf measured both the @xmath20 and @xmath21 diffractive fractions and also lpans to attempt to measure the dsf .
production diagrams.,scaledwidth=100.0% ] figure [ wz_diagrams ] shows schematic fynman diagrams for diffractive @xmath22 production . in leading order , the @xmath22
is produced by a quark in the pomeron ( left ) , while production by a gluon ( right ) is suppressed by a factor of @xmath23 and can be distinguished from quark production by an associated jet @xcite .
the data analysis is based on events with rps tracking from a data sample of approximately @xmath24 fb@xmath25 .
in addition to the @xmath22 selection requirements ( see below ) , a hit in the rps trigger counters and a rps reconstructed track with @xmath26 and @xmath27 are required . a novel feature of the analysis is the determination of the full kinematics of the @xmath28 decay , which is made possible by obtaining the neutrino @xmath29 from the missing @xmath13 , as usual , and @xmath30 from the formula @xmath31 $ ] , where @xmath32 $ ] .
mass from the diffractive data sample and a gaussian fit.,scaledwidth=50.0% ] the cdf @xmath22 selection requirements are @xmath33 gev , @xmath34 gev , @xmath35 gev , and vertex @xmath36-coordinate @xmath37 cm . the @xmath20 mass distribution for events with @xmath38
is shown in fig .
[ fig : w - mass ] along with a gaussian fit .
the obtained value of @xmath39 gev is in good agreement with the world average @xmath20-mass of @xmath40 gev @xcite .
figure [ fig : xi_cal ] shows the @xmath41 distributions of the @xmath22 events satisfying different selection requirements . in the @xmath20 case , the requirement of @xmath42 is very effective in removing the overlap events in the region of @xmath43 , while a mass cut of @xmath44 gev has the same effect . in the @xmath21 case ,
the @xmath41 distribution of all @xmath21 events is used and normalized to the rp - track distribution in the region of @xmath45 ( @xmath46 ) to obtain the nd background in the diffractive region of @xmath43 .
accounting for the rps acceptance of @xmath47 % , the trigger counter efficiency of @xmath48 % , the track reconstruction efficiency of @xmath49 % , multiplying by two to include the production by @xmath50 , and correcting the number of nd events for the effect of overlaps due to multiple interactions by multiplying it by a factor of @xmath51 % , allows the calculation of the diffractive fraction of @xmath22 events as @xmath52 , which yields the results : 1em @xmath53}\%,$ ] @xmath54}\%.$ ] 1em the @xmath55 value is consistent with the run i result of : @xmath56~\%\;(run~i)$ ] , obtained from the measured value of @xmath57\%$ ] @xcite , which is multiplied by a factor of 0.85 tto account for the reduced ( @xmath58-@xmath59 ) range in run ii .
distribution for various @xmath20 ( left ) and @xmath21 ( right ) event samples.,title="fig:",scaledwidth=40.0% ] distribution for various @xmath20 ( left ) and @xmath21 ( right ) event samples.,title="fig:",scaledwidth=41.5% ]
[ [ diffractive - dijet - production ] ] diffractive dijet production + + + + + + + + + + + + + + + + + + + + + + + + + + + + preliminary resluts on the @xmath60 , @xmath61 , and @xmath59 dependence of the diffractive structure function from dijet production at @xmath62 gev were reported at `` diffraction-2006 '' @xcite . the measured run ii @xmath60 rates confirm the factorization breakdown observed in run i. the @xmath61 and @xmath59 dependence results are shown in fig [ fig : xbjq2 ] .
( momentum fraction of parton in @xmath11 ) for different value s of @xmath63 ; _ ( right ) _
@xmath64 slope vs. @xmath61 .
, scaledwidth=51.0% ] ( momentum fraction of parton in @xmath11 ) for different value s of @xmath63 ; _ ( right ) _
@xmath64 slope vs. @xmath61 . , scaledwidth=49.0% ] in the range @xmath65 gev@xmath66 , where the inclusive @xmath13 distribution falls by a factor of @xmath67 , the ratio of the sd / nd distribution increases by only a factor of @xmath68 .
the above results indicate that the @xmath61 evolution in diffractive interactions is similar to that in nd interactions .
* @xmath59-dependence . * the slope parameter @xmath69 of an exponential fit to @xmath59 distributions near @xmath70 shows no @xmath61 dependence in the range @xmath71 .
the above results support the picture of a composite pomeron formed from color - singlet combinations of the underlying parton densities of the nucleon @xcite .
[ [ exclusive - dijet - production ] ] exclusive dijet production + + + + + + + + + + + + + + + + + + + + + + + + + + the process of exclusive dijet production is important for testing and / or calibrating models for exclusive higgs production at the lhc .
the cdf collaboration has made the first observation of this process and their main result is shown in fig .
[ fig : xsec_vs_mjj ] .
details can be found in ref .
this result favors the model of ref .
@xcite , which is implemented in the monte carlo simulation exhume @xcite .
normalized to measured @xmath72 values .
the solid curve is the cross section predicted by exhume .
, width=302 ]
we review cdf ii results from @xmath0 collisions at @xmath11.96 tev and compare them with theoretical expectations .
we concentrate on diffractive w / z and dijet production , a comparison of which allows the determination of the quark to gluon ratio of the diffractive exchange .
we also discuss the recently published result of exclusive dijet production , which is used to check / calibrate theoretical predictions for exclusive higgs boson production at the lhc .
results from cdf ii on rapidity gaps between jets and on exclusive diphoton and dilepton production have also been presented at this conference .
combined , these results and the analysis methods used in obtaining them provide a powerful launching board for searches at the lhc aimed at exploiting diffractive and exclusive production to discover new physics .
99 k. goulianos , `` diffraction and exclusive ( higgs ? ) production from cdf to lhc '' ? in _
les rencontres de physique de la vallee daoste : results and perspectives in particle physics , la thuile , italy , 4 - 10 march 2007 , pp .
177 - 200 _ , frascati physics series , special 44 issuee , edited by mario greco . k. goulisnos , `` diffraction at cdf , '' in proceedings of _ 12@xmath73 international conference on elastic and diffractive scattering ( blois workshop ) - forward physics and qcd _ , desy , hambyrg , germany 21 - 25 may 2007 , arxiv:0712.3633v2[hep - ph]3 jun 2008 , pp . 137 - 144 , http://arxiv.org/abs/0712.3633 ; ib .
_ pomeron intercept and slope : the qcd connection _ , pp . 248 - 253
. t. aaltonen _ et al . _
( cdf collaboration ) , phys .
d * 77 * , 052004 ( 2008 ) .
d. acosta _ et al .
_ ( cdf collaboration ) , phys .
d * 71 * , 032001 ( 2005 ) .
j. collins , j. phys .
g * 28 * , 1069 ( 2002 ) ; arxiv : hep - ph/0107252 .
abe et al .
( cdf collaboration ) , phys .
lett . * 78 * , 2698 - 2703 ( 1997 ) .
yao et al . , journ .
g * 33 * , 1 ( 2006 ) and 2007 partial update for 2008 ; http:\pdg.lbl.gov .
k. goulianos , proceedings of science , pos(diff2006)016 .
v. a. khoze , a. d. martin , and m. g. ryskin , eur .
j. c * 19 * , 477 ( 2001 ) [ erratum - ibid .
c * 20 * , 599 ( 2001 ) ] .
j. monk and a. pilkington , comput .
175 * , 232 ( 2006 ) ; arxiv : hep - ph/0502077 . | results on diffraction from the fermilab tevatron collider obtained by the cdf ii collaboration using data from @xmath0 collisions at @xmath11.96 tev are reviewed and compared with theoretical expectations .
implications for predictions of exclusive higgs boson production rates at the large hadron collider are discussed .
address = the rockefeller university , 1230 york avenue , new york , ny 10065 , usa |
the experimental results on deep inelastic charm production obtained by the h1 @xcite and zeus @xcite collaborations at hera provide a strong impetus for further theoretical studies .
this process is truly semihard because of the presence of two larges scales : the virtuality of the exchanged photon ( @xmath3 ) and the charm mass ( @xmath4 ) , both being much larger than @xmath5 but much smaller than @xmath6 .
therefore , in the present note , we focus on the semihard approach @xcite ( sha ) , which we had applied earlier to the @xmath1 meson photoproduction @xcite in a similar manner .
it is known that the resummation @xcite of the terms @xmath7^n,\;$ ] @xmath8^n\;$ ] and @xmath9^n$ ] in sha results in the so called unintegrated gluon distribution + @xmath10 , which determines the probability to find a gluon carrying the longitudinal momentum fraction @xmath11 and transverse momentum @xmath12 at the probing scale @xmath13 .
it obeys the bfkl equation @xcite and reduces to the conventional gluon density @xmath14 once the @xmath15 dependence is integrated out : @xmath16 the factorization scale @xmath13 ( such that @xmath17 @xmath18 ) indicates the scale of the nonperturbative input distribution .
the ccfm evolution equation @xcite includes coherence effects via parton angular ordering and reproduces the bfkl evolution equation in the small @xmath11 limit .
therefore the ccfm unintegrated parton distribution @xmath19 ( unlike the function @xmath20 ) depends also on the maximum angle allowed for any emission corresponding to @xmath21 . in the small @xmath11 limit it reduces to @xmath22 @xcite .
when calculating the spin average of the matrix element squared , we substitute the full lepton tensor for the photon polarization matrix : @xmath23/(q^2)^2 \label{epsq}\ ] ] ( including also the photon propagator factor ) .
the virtual gluon polarization matrix is taken in the form @xcite : @xmath24 where @xmath25 and @xmath26 are the 4-momenta of the incoming electron and proton .
to parametrise the unintegrated structure functions , we use the prescriptions of ref .
the proposed method lies upon a strai ghtforward perturbative solution of the bfkl equation where the collinear gluon density @xmath27 is used as the boundary condition in the integral form ( 1 ) .
technically , the unintegrated gluon density is calculated as a convolution of the collinear gluon density with universal weight factors : @xmath28 @xmath29 @xmath30 where @xmath31 and @xmath32 stand for bessel functions ( of real and imaginary arguments , respectively ) , and @xmath33 .
the latter parameter is connected with the pomeron trajectory intercept : @xmath34 in the lo , and @xmath35 in the nlo approximations , respectively , where @xmath36 @xcite . in the previous work @xcite
we used the standard grv parametrization @xcite for the collinear gluon density , from which the unintegrated gluon distribution was developed according to eqs .
( 4 ) - ( 6 ) .
some other essential parameters were chosen as follows : the charm quark mass @xmath37 gev , the peterson fragmentation parameter @xmath38 , the overall @xmath39 fragmentation probability 0.26 .
the pomeron intercept @xmath2 was regarded as free parameter , and then the value @xmath40 has been extracted from a fit to the experimental @xmath41 spectrum measured by the zeus collaboration @xcite .
in this paper we also use an unintegrated gluon density coming from a solution of the ccfm evolution equation ( see @xcite ) .
it was shown in @xcite that a good description of the inclusive structure function @xmath42 and the production of forward jets in dis , which are believed to be a prominent signature of small @xmath11 parton dynamics , can be obtained using the ccfm unintegrated gluon distribution .
in @xcite we have used the hadron level monte carlo program cascade described in @xcite to predict the cross section for @xmath43 photoproduction at hera energies .
it was shown that in this case the description of the differential cross section @xmath44 for @xmath45 1 gev@xmath46 for different regions of @xmath41 is improved .
the theoretical predictions on the differential cross sections of deep inelastic @xmath47 production are shown in figure [ fig : fig1 ] for the zeus kinematical region : @xmath48 gev@xmath46 , @xmath49 gev and @xmath50 .
different curves in figure [ fig : fig1 ] correspond to different values of the pomeron intercept parameter : @xmath2 = 0.166 ( dotted ) , 0.35 ( solid ) and 0.53 ( dashed ) . we see that the theoretical curves with @xmath51 describe all zeus experimental data @xcite on the differential cross sections , except the @xmath44 distribution . the sha calculations with bfkl unintegrated gluon distribution show some shift to negative @xmath52 with respect to the data .
this descrepancy between the data and the sha prediction could result from the use of the peterson fragmentation function .
another reason may be connected with the bfkl lo unitegrated gluon density .
, ccfm scheme with peterson fragmentation.[fig : fig2 ] ] , ccfm sheme with jetset fragmentation.[fig : fig3 ] ] the use of ccfm unintegrated gluon density ( with angular ordering ) from mc generator cascade @xcite with the peterson fragmentation function for @xmath53 transition gives some positive shift to the @xmath44 distribution ( figure [ fig : fig2 ] ) compared to bfkl curve ( see figure [ fig : fig1]e ) . finally , when we use the ccfm unintegrated gluon distribution with more realistic jetset based fragmentation function @xcite implemented in @xcite we obtain good agreement between our theoretical results and the zeus experimental data @xcite for @xmath44 ( figure [ fig : fig3 ] ) .
one of us ( n.z . ) would like to thank organizing committee of the workshop `` diffraction 2000 '' for the invitation and financial support and r. fiore for exiting scientific atmosphere .
this talk is based on the work supported in part by the royal swedish academy of sciences .
99 h1 collab .
, c. adloff et al .
, z. phys .
c72 ( 1996 ) 593 ; nucl .
b545 ( 1999 ) 21 .
zeus collab . ,
j. breitweg et al . , phys .
b407 ( 1997 ) 402 .
zeus collab . ,
j. breitweg et al . , eur .
j. c12 ( 2000 ) 35 .
gribov , e.m .
levin , m.g .
ryskin , phys .
100 ( 1983 ) 1 .
levin , m.g .
ryskin , yu.m .
shabelski , a.g .
shuvaev , sov .
53 ( 1991 ) 657 .
baranov , n.p .
zotov , phys .
b458 ( 1999 ) 389 .
baranov , h. jung , n.p .
zotov , in proc . of the workshop on monte carlo generators , ed . by a.
doyle , g. grindhammer , g. ingelman , h. jung , desy , hamburg ( 1999 ) , p. 484 ; hep - ph/9910210 .
s. catani , m. ciafaloni , f. hautmann , nucl .
b366 ( 1991 ) 135 .
collins , r.k .
ellis , nucl . phys
. b360 ( 1991 ) 3 .
e. kuraev , l. lipatov , v. fadin , sov .
jetp 44 ( 1976 ) 443 ; 45 ( 1977 ) 199 ; y. balitskii , l. lipatov , sov .
( 1978 ) 822 .
m. ciafaloni , nucl .
phys . b296
( 1998 ) 49 ; s. catani , f. fiorani , g. marchesini , phys .
b234 ( 1990 ) 339 , nucl .
b336 ( 1990 ) 18 ; g. marchesini , nucl .
b445 ( 1995 ) 49 .
j. blumlein , desy 95 - 121 .
ross , phys .
lett . b431
( 1998 ) 161 .
m. gluck , e. reya , a. vogt , z. phys .
c67 ( 1995 ) 433 .
zeus collab . ,
j. breitweg et al . , eur .
j. c6 ( 1999 ) 67 .
h. jung , in proc . of the workshop on monte carlo generators , ed . by a.
doyle , g. grindhammer , g. ingelman , h. jung desy , hamburg ( 1999 ) , p. 75 ; hep - ph/9908497 .
t. sjstrand , comp .
82 ( 1994 ) 74 . | in the framework of semihard ( k@xmath0 factorization ) qcd approach , we consider the differential cross sections of @xmath1 meson production at hera . the consideration is based on bfkl and ccfm gluon distributions .
we find that in the case of bfkl lo gluon distribution the theoretical results are sensitive to the pomeron intercept parameter @xmath2 .
we present a comparison of the theoretical results with available zeus experimental data . |
in multicarrier systems like ofdm , subcarriers allocated to a user may face interference due to a variety of reasons .
these include the activity of other users and allocation decisions of neighbouring base stations in a cellular network .
predicting the presence or absence of interference in a particular subcarrier may not be feasible at a transmitter in such uncoordinated networks .
nevertheless , it is practical to assume that a subcarrier allocated to a user does not face interference in every channel instantiation .
thus , there is a scope for harnessing such _ bursty _
interference in multicarrier systems and exploring the possibility of opportunistic rate increments .
the following toy example , based on parallel linear deterministic channels , captures the intuition behind our problem formulation .
consider @xmath0 transmitters ( @xmath2 and @xmath3 ) and @xmath0 receivers ( @xmath4 and @xmath5 ) .
for @xmath6 , @xmath7 has messages for @xmath8 and at discrete time index @xmath9 , @xmath7 can transmit @xmath0 bits @xmath10 $ ] .
the @xmath0 bits correspond to @xmath0 subcarriers ( parallel channels ) allocated to each transmitter - receiver pair .
depending on the interference channel realization ( stays constant for @xmath9 ) , @xmath8 receives one of the three possibilities : @xmath11 $ ] , @xmath12 $ ] and @xmath13 $ ] ( shown in figure [ fig : toy_example_wireless ] ) , where @xmath14 and @xmath15 .
the first possibility corresponds to the interference free case ( for @xmath8 ) and the remaining two possibilities correspond to interference from @xmath16 ( only one of the subcarriers of @xmath8 gets interfered ) .
hence , there are @xmath17 distinct possibilities for the pair of received values at @xmath4 and @xmath5 over time duration @xmath18 . in the toy example .
the `` @xmath19 '' operator denotes modulo @xmath0 addition and indicates the presence of interference . as shown above , interference is not present in all channel realizations for @xmath8 ( hence bursty ) ; but whenever it is present , it is limited to just @xmath20 out of the @xmath0 transmitted bits . ] the crucial constraint in this setup is that the transmitters do not know _ a priori _ the interference channel realization . the channel is used @xmath18 times ( time index @xmath9 ) and we have the following ( symmetric ) rate requirement : ensure base rate @xmath21 at a receiver when _ any _ one of the subcarriers ( of the receiver ) gets interfered and ensure rate @xmath22 at a receiver when both subcarriers ( of the receiver ) are interference free ( _ i.e. , _ opportunistically deliver incremental rate @xmath23 , in addition to @xmath21 , whenever a receiver is interference free ) . in this setup , we are interested in characterizing the rate region @xmath24 as the performance metric .
clearly , @xmath25 ( a maximum of @xmath0 bits per time index can be sent by a transmitter ) and corner point @xmath26 is easily achievable .
also , the corner point @xmath27 can be easily achieved by using a repetition code across the @xmath0 subcarriers ( _ i.e. , _
@xmath28 and @xmath29 ) .
the repetition code ensures decodability of the message ( of rate @xmath21 ) irrespective of which subcarrier gets interfered . using time sharing between corner points @xmath30 and @xmath31
, we can achieve @xmath32 .
intuitively this looks like the best we can do , and indeed it can be shown to be tight using entropy inequalities .
the problem pursued in this paper is a generalization of this example through parallel linear deterministic interference channels ( leading to a rate region with more than two non - trivial corner points in most cases ) . in @xcite and @xcite ,
the problem of harnessing bursty interference was studied for a single carrier scenario using a degraded message set approach .
this approach guarantees a base rate when the carrier faces interference .
in addition to the base rate , an incremental rate is provided whenever the carrier is interference free . in the multicarrier version considered in this paper , every user ( receiver )
is allocated @xmath1 subcarriers ( parallel channels ) and we extend the degraded message set approach for a rate tuple @xmath33 as follows : ( a ) when all @xmath1 subcarriers of a user get interfered , the user achieves rate @xmath34 ( b ) when any @xmath35 out of @xmath1 subcarriers get interfered , the user achieves rate @xmath36 and ( c ) when all @xmath1 subcarriers are interference free , the user achieves rate @xmath37 .
thus , the user experiences opportunistic rate increments as the number of interfered subcarriers decreases . maintaining low message complexity
is the practical idea behind considering the number of interfered subcarriers rather than the specific set of subcarriers interfered .
the problem formulation has some similarity with symmetric multilevel diversity coding @xcite and our results demonstrate that similar tools ( subset entropy inequalities ) as in @xcite can be used in this context .
our main contributions in this paper are : * inner bounds for @xmath38 and @xmath39 setups using erasure coding across subcarriers ( employed for specific interfered _ levels _ in a subcarrier ) , signal - scale alignment @xcite and han - kobayashi scheme .
* develop outer bounds using techniques inspired by multilevel diversity codes .
* the inner and outer bounds coincide for several regimes .
the remainder of this paper is organized as follows .
section [ sec : model ] formalizes the setup and rate requirements .
section [ sec : main_results ] states the main results .
inner bounds and outer bounds are discussed in sections [ sec : inner_bounds ] and [ sec : outer_bounds ] respectively .
we conclude the paper with a short discussion in section [ sec : discussion ] .
we consider a system with two base stations ( transmitters ) @xmath2 and @xmath3 and two users ( receivers ) @xmath4 and @xmath5 . for @xmath40 , user @xmath8
is allocated @xmath1 subcarriers @xmath41 by the base station @xmath7 .
the transmit signals of base stations @xmath2 and @xmath3 are assumed to be independent .
the channel is modeled by a @xmath0-user multicarrier ( parallel ) linear deterministic interference channel @xcite where , similar to @xcite , interfering links in each subcarrier may or may not be active ( unknown to the transmitters ) . at discrete time index @xmath42 , the transmit signal on subcarrier @xmath43 is @xmath44 where @xmath45 is a finite field .
the received signals on subcarrier @xmath43 of @xmath8 when @xmath43 faces interference from @xmath46 ( corresponding to user @xmath47 ) and when it is interference free are described below as ( [ eq : interfered ] ) and ( [ eq : interference_free ] ) respectively , @xmath48 where @xmath49 is a @xmath50 shift matrix in the terminology of deterministic channel models @xcite and @xmath51 denotes the transmit signal on subcarrier @xmath46 for user @xmath52 .
all operations above are in @xmath53 .
similar to @xcite , the transmitters are assumed to have prior knowledge of parameters @xmath54 and @xmath55 ( direct and interfering channel strengths ) , and the presence ( or absence ) of interference in a subcarrier is assumed to be constant throughout the channel usage duration . without loss of generality , we assume @xmath56 . let @xmath57 denote the normalized strength of the interfering signal . since interference free capacity for a single carrier can be achieved when @xmath58 @xcite , we focus on @xmath59 . for every time instant ,
it is convenient to consider a subcarrier as indexed levels of bit pipes .
each bit pipe can carry a symbol from @xmath45 . + let @xmath60 denote the interfering signal for @xmath8 on subcarrier @xmath43 .
we use @xmath61 $ ] to denote the transmit signals sent during @xmath18 time slots on @xmath43 and @xmath62 is defined similarly from @xmath63 .
also , we define @xmath64 $ ] .
the rate requirements for both the users are constrained to be symmetric . for @xmath8 , messages @xmath65 corresponding to rate tuple @xmath66 are encoded in @xmath67 .
based on the number of interfered subcarriers for @xmath8 , we have the following requirements for the desired messages : 1 .
@xmath8 decodes @xmath68 when all @xmath1 subcarriers of @xmath8 get interfered . 2 .
@xmath8 decodes @xmath69 when any @xmath35 out of @xmath1 subcarriers of @xmath8 get interfered .
3 . @xmath8 decodes @xmath70 when all @xmath1 subcarriers of @xmath8 are interference free .
a rate tuple is considered achievable if the probability of decoding error is vanishingly small as @xmath71 . to simplify our analysis
, we consider two setups : @xmath72-setup and @xmath73-setup . in the @xmath72-setup
, @xmath34 is assumed to be zero and in the @xmath73-setup @xmath23 is assumed to be zero .
the rate regions for these two setups are analyzed separately in this paper .
depending on whether @xmath74 or @xmath75 , we have different results for @xmath72-setup and @xmath73-setup .
we have a tight characterization of capacity in this case .
[ thm : capacity_l_0_less_l ] for @xmath74 , the capacity region for @xmath72-setup is as follows .
@xmath76 in this case , we have a tight characterization in certain regimes .
[ thm : inner_l_0_more_l]for @xmath75 , consider the following rate inequalities : @xmath77 inequalities ( [ eq : ob_3 ] ) , ( [ eq : ob_4 ] ) and ( [ eq : ob_5 ] ) are inner bounds ; ( [ eq : ob_3 ] ) and ( [ eq : ob_4 ] ) are outer bounds .
[ cor:1 ] we have a tight characterization for the regime @xmath78 in the @xmath79-setup .
this follows from the observation that ( [ eq : ob_5 ] ) is not active in presence of ( [ eq : ob_3 ] ) and ( [ eq : ob_4 ] ) for @xmath80 ( see appendix for detailed proof ) .
[ conj : ob_5 ] for the @xmath72-setup with @xmath75 , ( [ eq : ob_5 ] ) is an outer bound .
if conjecture [ conj : ob_5 ] holds , we have a tight characterization for @xmath72-setup when @xmath75 . in this case
, we have a tight characterization in certain regimes .
[ thm : inner_m_l_less_l ] for @xmath74 , consider the following rate inequalities : @xmath81 inequalities ( [ eq : ob_6 ] ) , ( [ eq : ob_7 ] ) and ( [ eq : ob_8 ] ) are inner bounds ; ( [ eq : ob_6 ] ) and ( [ eq : ob_7 ] ) are outer bounds .
[ cor:2 ] we have a tight characterization for the regime @xmath82 in the @xmath83-setup .
this follows from the observation that ( [ eq : ob_8 ] ) is not active in presence of ( [ eq : ob_6 ] ) and ( [ eq : ob_7 ] ) for @xmath84 ( see appendix for detailed proof ) .
[ conj : ob_8 ] for the @xmath73-setup with @xmath74 , ( [ eq : ob_8 ] ) is an outer bound .
if conjecture [ conj : ob_8 ] holds , we have a tight characterization for @xmath73-setup when @xmath74 . in this case
, we have a tight characterization in certain regimes .
[ thm : inner_m_l_more_l]for @xmath75 , consider the following rate inequalities:@xmath85 inequalities ( [ eq : ob_9 ] ) , ( [ eq : ob_10 ] ) and ( [ eq : ob_11 ] ) are inner bounds ; ( [ eq : ob_9 ] ) and ( [ eq : ob_10 ] ) are outer bounds .
[ cor:3 ] we have a tight characterization for the regime @xmath86 in the @xmath83-setup .
this follows from the observation that ( [ eq : ob_11 ] ) is not active in presence of ( [ eq : ob_9 ] ) and ( [ eq : ob_10 ] ) for @xmath86 ( see appendix for detailed proof ) .
[ conj : ob_11 ] we conjecture that ( [ eq : ob_11 ] ) is an outer bound for @xmath73-setup when @xmath75 .
if conjecture [ conj : ob_11 ] holds , we have a tight characterization for @xmath73-setup when @xmath75 .
figure [ fig : inner_bounds ] summarizes the inner bounds for different regimes depending on values of @xmath87 , @xmath1 and @xmath35 .
the inner bound rate region is obtained from achievable corner points ( shown in figure [ fig : inner_bounds ] ) using time - sharing .
achievability schemes for corner points shown in figure [ fig : inner_bounds ] can be described as follows . * _ @xmath89 _ : this appears in cases ( 1)-(5 ) in figure [ fig : inner_bounds ] .
it can be achieved by using the top @xmath54 levels in all the @xmath1 subcarriers for message @xmath90 . * _ @xmath91 _ : this corner point is achievable for @xmath92 and appears in cases ( 1)-(3 ) in figure [ fig : inner_bounds ] . to achieve this , the top @xmath93 levels of each subcarrier
are used for @xmath94 and the bottom @xmath95 levels are used for @xmath96 . since the top @xmath97 levels of a subcarrier are always interference free , using @xmath1 subcarriers we achieve @xmath91 . *
_ @xmath98 _ : this corner point is achievable for @xmath92 and appears in case ( 1 ) in figure [ fig : inner_bounds ] .
since any @xmath35 out of @xmath1 subcarriers get interfered , an erasure code ( across @xmath1 subcarriers ) can recover symbols at rate @xmath99 from the bottom @xmath100 levels of @xmath1 subcarriers .
also , an additive rate of @xmath101 can be obtained by using the top @xmath93 levels of @xmath1 subcarriers . adding the contributions from the bottom @xmath100 levels and top @xmath93 levels of all @xmath1 subcarriers , we achieve @xmath102 .
* _ @xmath103 and @xmath104 _ : these appear in case ( 2 ) in figure [ fig : inner_bounds ] and are achievable for @xmath105 using the following signal - scale alignment technique @xcite .
the @xmath54 levels in a subcarrier @xmath43 are divided into @xmath106 bands @xmath107,@xmath108,@xmath109 and @xmath110 as shown in figure [ fig : alignment ] . for @xmath111 , when subcarrier
@xmath43 faces interference , only @xmath107 of @xmath46 interferes with @xmath108 and @xmath109 of @xmath43 .
also , only @xmath108 of @xmath46 interferes with @xmath110 of @xmath112 .
given this structure , the trick will be to not transmit any information in band @xmath108 .
this keeps @xmath110 interference free as shown in figure [ fig : alignment ] . using @xmath107 and @xmath110 of @xmath1 subcarriers for @xmath113
, we achieve @xmath114 . using only @xmath109 of @xmath1 subcarriers for @xmath96
we achieve @xmath115 .
hence @xmath116 is achievable . for @xmath104 ,
the same signal - scale alignment trick is used in addition to a rate @xmath117 erasure code across @xmath1 subcarriers for @xmath109 .
+ , width=345,height=148 ] * _ @xmath118 _ : this appears in case ( 3 ) in figure [ fig : inner_bounds ] and is achievable for @xmath119 .
han - kobayashi scheme @xcite can achieve rate @xmath120 for a single interfered subcarrier when @xmath119 .
this scheme is used for each of the @xmath1 subcarriers to achieve this corner point . * _ @xmath121 and @xmath122 _ : these are achievable for @xmath123 .
the corner point @xmath121 appears in cases ( 4 ) and ( 5 ) in figure [ fig : inner_bounds ] and is achievable using the following signal - scale alignment strategy .
the top @xmath124 levels of a subcarrier are used for @xmath96 .
the next @xmath125 levels are used for @xmath94 .
this ensures that the levels used for @xmath94 are always interference free . using @xmath1 subcarriers we achieve , @xmath121 . to achieve @xmath122 ( which appears in case ( 4 ) in figure [ fig : inner_bounds ] ) , a similar scheme is used with a rate @xmath117 erasure code ( across @xmath1 subcarriers ) for the top @xmath124 levels of a subcarrier . *
_ @xmath126 _ : this appears in case ( 5 ) in figure [ fig : inner_bounds ] and is achievable for @xmath123 . for the classical two user interference channel ( single carrier ) with @xmath123 , rate @xmath127 is easily achievable . using this single carrier scheme for @xmath1 subcarriers ,
we achieve @xmath126 .
* _ @xmath129 _ : this appears in cases ( 6 ) , ( 7 ) and ( 9 ) in figure [ fig : inner_bounds ] and is achievable for @xmath130 . using the top @xmath93 levels of @xmath1 subcarriers for @xmath68
, we achieve @xmath131 . for @xmath94 ,
a rate @xmath117 erasure code is used for the bottom @xmath100 levels across @xmath1 subcarriers to obtain @xmath132 . * _ @xmath133 _ : this appears in cases ( 6 ) , ( 7 ) and ( 9 ) in figure [ fig : inner_bounds ] and is achievable for @xmath134 .
the achievability is same as that of @xmath135 in the @xmath72-setup . *
_ @xmath136 _ : this appears in cases ( 7)-(8 ) in figure [ fig : inner_bounds ] and is achievable for @xmath137 . a signal - scale alignment technique similar to the one in figure [ fig : alignment ]
is used to achieve @xmath138 .
additionally , a rate @xmath117 erasure code across @xmath1 subcarriers for @xmath109 is used to achieve @xmath139 . * _ @xmath140 _ : this appears in case ( 8) in figure [ fig : inner_bounds ] and is achievable for @xmath137 .
the achievability is same as that of @xmath141 in the @xmath72-setup . * _ @xmath142 and @xmath118 _ : the corner point @xmath118 appears in cases ( 9 ) and ( 10 ) while @xmath142 appears in case ( 10 ) in figure [ fig : inner_bounds ] .
both corner points are achievable for @xmath119 . to achieve @xmath142
, we use the scheme for achieving @xmath143 in the @xmath72-setup ( _ i.e. _ , han - kobayashi scheme is used for all the @xmath1 subcarriers ) .
also , by using @xmath68 instead of @xmath94 , the above scheme achieves corner point @xmath118 in the @xmath73-setup . * _ @xmath144 and @xmath145 _ : these are achievable for @xmath123 .
the corner point @xmath146 appears in case ( 11 ) in figure [ fig : inner_bounds ] and is achievable using the following signal - scale alignment strategy .
the top @xmath124 levels of a subcarrier are used for @xmath94 with a rate @xmath117 erasure code across @xmath1 subcarriers .
the next @xmath125 levels are used for @xmath68 .
this ensures that the levels used for @xmath68 are always interference free . using @xmath1 subcarriers we achieve , @xmath146
. to achieve @xmath147 ( which appears in case ( 11 ) in figure [ fig : inner_bounds ] )
, we use the same scheme as that for @xmath148 in the @xmath72-setup . * _ @xmath149 and @xmath150 _ : the corner point @xmath150 appears in cases ( 11 ) and ( 12 ) while @xmath151 appears in case ( 12 ) in figure [ fig : inner_bounds ] .
both corner points are achievable for @xmath152 .
to achieve @xmath149 , we use the scheme for achieving @xmath153 in the @xmath72-setup ( case(5 ) in figure [ fig : inner_bounds ] ) .
also , by using @xmath68 instead of @xmath94 , the above scheme achieves the corner point @xmath150 in the @xmath73-setup .
in this section , we first define additional notation for outer bound proofs .
this is followed by outer bound proofs for @xmath72-setup ( which use techniques @xcite from multilevel diversity coding ) and outer bound proofs for @xmath73-setup .
there are @xmath154 ways in which any @xmath35 out of @xmath1 subcarriers get interfered .
every such choice is a receiver configuration for a user .
we use additional notation for a special set of receiver configurations described below . consider a circulant matrix @xmath155 of dimension @xmath1 with the first row consisting of @xmath156 consecutive ones followed by @xmath35 zeros .
the other rows are cyclic right shifts of the first row . as an example
, @xmath157 is shown below .
@xmath158 + we use @xmath155 to list a specific set of receiver configurations in the following manner .
each row corresponds to a receiver configuration with @xmath1 subcarriers indexed by the columns . in each row
, @xmath20 denotes an interference free subcarrier and @xmath159 denotes an interfered subcarrier .
hence , out of @xmath154 choices , @xmath155 lists only @xmath1 receiver configurations . for example , the third row in @xmath157 shown above indicates a situation for @xmath8 where only subcarrier @xmath160 gets interfered .
the structure of @xmath161 corresponds to the choice of receiver configurations we use in some of our outer bound proofs .
this structure enables the use of sliding window subset inequality @xcite in such proofs .
we now describe additional notation related to receiver configurations of a user .
when @xmath8 is in receiver configuration indicated by row @xmath162 of @xmath155 , we use @xmath163 to denote the received signal on @xmath1 subcarriers ( over @xmath18 time slots ) . in the same spirit , we define @xmath164 as the interfering signal over all @xmath1 subcarriers for @xmath8 in this receiver configuration . the received signal in interference free subcarriers in @xmath163 is denoted by @xmath165 and the received signal in interfered subcarriers in @xmath163 is denoted by @xmath166 . when all @xmath1 subcarriers of @xmath8 are interference free , the received signal is denoted by @xmath167 .
now , a direct consequence of the sliding window subset inequality @xcite in our setting can be stated as follows .
@xmath168 we prove outer bound ( [ eq : ob_1 ] ) using a careful choice of receiver configurations represented by rows of @xmath155 . the high level idea is to divide the received signal into interfered and interference free terms followed by the use of ( [ eq : sliding_window ] ) on the interference free terms .
the proof can be described as follows .
using fano s inequality for @xmath8 @xmath6 , for any @xmath169 there exists a large enough @xmath18 such that , @xmath170 ( a ) follows from ( [ eq : sliding_window ] ) and ( b ) follows from @xmath171 and ( [ eq : sliding_window ] ) . substituting @xmath172 and @xmath173 in ( [ eq : a_1 ] ) ,
we can obtain two inequalities corresponding to different users . on adding these two inequalities , @xmath174 where ( a ) follows from the structure of @xmath155 .
the inequality ( [ eq : a_00 ] ) in the proof of outer bound ( [ eq : ob_1 ] ) also holds for @xmath75 .
hence , @xmath175 substituting @xmath172 and @xmath173 in ( [ eq : a_10 ] ) , we can obtain two inequalities corresponding to different users . on adding these two inequalities , @xmath176 where ( a ) follows from @xmath75 and the structure of @xmath155 .
outer bounds ( [ eq : ob_6 ] ) , ( [ eq : ob_7 ] ) , ( [ eq : ob_9 ] ) and ( [ eq : ob_10 ] ) can be shown by using the el gamal - costa injective interference channel bounds @xcite as follows . for this outer bound proof , we consider two receiver configurations with no interfered subcarriers in common and apply the injective channel bound @xcite as shown below
. for any @xmath169 there exists a large enough @xmath18 such that , @xmath177 where ( a ) follows from the injective channel bound @xcite .
outer bounds ( [ eq : ob_7 ] ) and ( [ eq : ob_10 ] ) have the same proof . for the proof , we consider the receiver configuration with all @xmath1 subcarriers interfered and apply the injective channel bound @xcite as shown below
. for any @xmath169 there exists a large enough @xmath18 such that , @xmath178 where @xmath179 corresponds to the receiver configuration with all @xmath1 subcarriers interfered and ( a ) follows from the injective channel bound @xcite . for this outer bound proof , we consider two receiver configurations with minimum number of interfered subcarriers ( _ i.e. , _
@xmath180 ) in common and apply the injective channel bound @xcite as shown below . for any @xmath169 there exists a large enough @xmath18 such that , @xmath181 where ( a ) follows from the injective channel bound @xcite .
it is optimal to treat interference as noise in the regimes where erasure coding across subcarriers leads to tight inner bounds .
however , outer bound conjectures on ( [ eq : ob_5 ] ) and ( [ eq : ob_11 ] ) ( _ i.e. , _ conjectures [ conj : ob_5 ] and [ conj : ob_11 ] ) suggest that this may not be the case for all regimes . for @xmath182 ( and @xmath183 ) , both imply @xmath184 ; this can be simply achieved by dividing the @xmath1 subcarriers between the two users .
an erasure coding scheme in this case will lead to @xmath185 .
hence , erasure coding across subcarriers may not be optimal in all regimes .
the work of s. mishra and s. diggavi was supported in part by nsf award 1136174 and muri award afosr fa9550 - 09 - 064 . the work of i .- h . wang was supported by eu project conect fp7-ict-2009 - 257616 . 1 n. khude , v. prabhakaran and p. viswanath , `` opportunistic interference management , '' in proc . ieee int .
inf . theory ( isit ) , 2009 .
n. khude , v. prabhakaran and p. viswanath , `` harnessing bursty interference , '' in proc .
information theory workshop ( itw ) , 2009 .
j. r. roche , r. w. yeung and k. p. hau , `` symmetrical multilevel diversity coding , '' ieee trans .
theory 43 , pp .
1059 - 1064 , may 1997 .
a. s. avestimehr , s. diggavi and d. tse , `` wireless network information flow , '' allerton conf . on comm . , control , and computing , 2007
g. bresler , a. parekh and d. tse , `` the approximate capacity of the many - to - one and one - to - many gaussian interference channels , '' ieee trans .
inf . theory 56(9 ) , pp .
4566 - 4592 , 2010 .
a. e. gamal and y. kim , `` network information theory , '' cambridge university press , 2011 .
j. jiang , n. marukala and t. liu , `` symmetrical multilevel diversity coding with an all - access encoder , '' in proc .
inf . theory ( isit ) , pp .
1662 - 1666 , 2012 .
a. b. carleial , `` a case where interference does not reduce capacity , '' ieee trans .
inf . theory 21(5 ) , pp .
569 - 570 , sep .
1975 . in this regime , inequalities ( [ eq :
ob_3 ] ) , ( [ eq : ob_4 ] ) and ( [ eq : ob_5 ] ) can be rewritten ( shown below ) as ( [ eq : simp_ob_3 ] ) , ( [ eq : simp_ob_4 ] ) and ( [ eq : simp_ob_5 ] ) respectively .
@xmath186 figure [ fig : cor_1 ] shows the situation in this regime , @xmath187 ; it is clear that ( [ eq : simp_ob_5 ] ) ( dashed red line in figure [ fig : cor_1 ] ) is not active in presence of ( [ eq : simp_ob_3 ] ) and ( [ eq : simp_ob_4 ] ) ( solid green lines in figure [ fig : cor_1 ] ) .
since inequalities ( [ eq : simp_ob_3 ] ) and ( [ eq : simp_ob_4 ] ) are inner bounds as well as outer bounds in this regime , we have a tight characterization . ) for the regime @xmath78 in the @xmath72-setup .
inequality ( [ eq : simp_ob_5 ] ) ( dashed red line ) is not active in presence of ( [ eq : simp_ob_3 ] ) and ( [ eq : simp_ob_4 ] ) ( solid green lines).,width=288,height=244 ] in this regime , inequalities ( [ eq : ob_6 ] ) , ( [ eq : ob_7 ] ) and ( [ eq : ob_8 ] ) can be rewritten ( shown below ) as ( [ eq : simp_ob_6 ] ) , ( [ eq : simp_ob_7 ] ) and ( [ eq : simp_ob_8 ] ) respectively .
@xmath189 figure [ fig : cor_2 ] shows the situation in this regime ; it is clear that ( [ eq : simp_ob_8 ] ) ( dashed red line in figure [ fig : cor_2 ] ) is not active in presence of ( [ eq : simp_ob_6 ] ) and ( [ eq : simp_ob_7 ] ) ( solid green lines in figure [ fig : cor_2 ] ) .
since inequalities ( [ eq : simp_ob_6 ] ) and ( [ eq : simp_ob_7 ] ) are inner bounds as well as outer bounds in this regime , we have a tight characterization . ) for the regime @xmath188 in the @xmath73-setup .
inequality ( [ eq : simp_ob_8 ] ) ( dashed red line ) is not active in presence of ( [ eq : simp_ob_6 ] ) and ( [ eq : simp_ob_7 ] ) ( solid green lines).,width=288,height=233 ] for this regime , inequalities ( [ eq : ob_9 ] ) , ( [ eq : ob_10 ] ) and ( [ eq : ob_11 ] ) can be rewritten ( shown below ) as ( [ eq : simp_ob_9 ] ) , ( [ eq : simp_ob_10 ] ) and ( [ eq : simp_ob_11 ] ) respectively .
@xmath190 to show ( [ eq : simp_ob_11 ] ) is not active in presence of ( [ eq : simp_ob_9 ] ) and ( [ eq : simp_ob_10 ] ) , it is sufficient to prove ( [ eq : simp_ob_9 ] ) _ _ dominates _ _ ] ( [ eq : simp_ob_11 ] ) in this regime .
we prove this in two steps as shown below ( analysis for @xmath191 followed by analysis for @xmath192 ) .
as shown above , ( [ eq : simp_ob_9 ] ) dominates ( [ eq : simp_ob_11 ] ) for both @xmath195 and @xmath200 . since inequalities ( [ eq : simp_ob_9 ] ) and ( [ eq : simp_ob_10 ] ) are inner bounds as well as outer bounds
, we have a tight characterization in the regime @xmath201 . | we study opportunistic interference management when there is bursty interference in parallel @xmath0-user linear deterministic interference channels .
a degraded message set communication problem is formulated to exploit the burstiness of interference in @xmath1 subcarriers allocated to each user .
we focus on symmetric rate requirements based on the number of interfered subcarriers rather than the exact set of interfered subcarriers .
inner bounds are obtained using erasure coding , signal - scale alignment and han - kobayashi coding strategy .
tight outer bounds for a variety of regimes are obtained using the el gamal - costa injective interference channel bounds and a sliding window subset entropy inequality @xcite .
the result demonstrates an application of techniques from multilevel diversity coding to interference channels .
we also conjecture outer bounds indicating the sub - optimality of erasure coding across subcarriers in certain regimes . |
biological evolution offers a number of important , unsolved problems that are well suited for investigation by methods from statistical physics .
many of these can be studied using complex , interacting model systems far from equilibrium @xcite .
areas that have generated exceptional interest among physicists are those of coevolution and speciation .
a large class of coevolution models have been inspired by one introduced by bak and sneppen @xcite , in which species with different levels of fitness " compete , and the least fit species and those that interact with it are regularly mutated " and replaced by new species with different , randomly chosen fitness .
models in this class exhibit avalanches of extinctions and appear to evolve towards a self - organized critical state @xcite .
although such models may be said to incorporate darwin s principle of survival of the fittest , " they are artificial in the sense that mutation and selection are assumed to act collectively on entire species , rather than on individual members of their populations . in reality , mutations are changes in the genotypes of _ individual organisms _ that are introduced or passed on during reproduction .
these changes in the genotype affect the phenotype ( physical and behavioral characteristics of the organism and its interactions with other organisms ) , and it is on the level of the phenotypes of individuals that competition and selection act . a number of evolution models ( see , e.g. , ref .
@xcite for a review from a physicist s point of view ) therefore take as their basis a genome in the form of a string of letters " as in eigen s quasi - species model of molecular evolution @xcite .
depending on the level of the modeling , the letters of the genomic `` alphabet '' can be a ( possibly large ) number of alleles at a gene locus , or the four nucleotides of a dna or rna sequence @xcite .
however , the size of the alphabet is not of great importance in principle , and it is common in models to use a binary alphabet with the two letters 0 and 1 ( or @xmath3 ) @xcite .
we believe a fruitful approach to the study of coevolution is one in which selection is provided by interspecies interactions along lines commonly considered in community ecology , while genetic variation is provided by random mutations in the genomes of individual organisms .
an early attempt in this direction is the coupled @xmath4 model with population dynamics introduced by kauffman and johnsen @xcite , but thus far not many similar models have been studied . recently ,
hall , christensen , and collaborators @xcite introduced a model they called the `` tangled - nature '' model , in which each individual lives in a dynamically evolving fitness landscape " created by the populations of all the other species .
here we consider a simplified version of this model , in which no individual is allowed to live through more than one reproduction cycle .
this restriction to nonoverlapping generations enables us to both study the model in detail by linear stability analysis and to perform very long monte carlo simulations of its evolutionary behavior . in short , the model consists of populations of different species , on which selection acts through asexual reproduction rates that depend on the populations of all the other species via a constant , random interaction matrix that allows mutualistic , competitive , and predator / prey relations .
in addition to these direct interactions , all individuals interact indirectly through competition for a shared resource .
the competition keeps the total population from diverging .
genetic variety is provided by a low mutation rate that acts on the genomes of individual organisms during reproduction , inducing the populations to move through genotype space .
the resulting motion is intermittent : long periods of stasis with only minor fluctuations are interrupted by bursts of significant change that is rapid on a macroscopic time scale , reminiscent of what is known in evolutionary biology as punctuated equilibria @xcite .
a preliminary report on some aspects of the work reported here is given in ref .
@xcite .
the main foci of the present paper are the structure and stability of communities and the statistical properties of the dynamical behavior on very long ( `` geological '' ) time scales .
the rest of the paper is organized as follows . in sec .
[ sec : mod ] we describe our model in detail , including the detailed monte carlo algorithm used for our simulations . in sec .
[ sec : lsa ] we discuss the properties of the fixed points of the population for the mutation - free version of the model .
many of these properties can be understood analytically within a simple mean - field approach .
a full , probabilistic description of our model is also possible , although the mathematics is somewhat involved . in appendix
[ sec : ab ] we provide , for the sake of completeness , the key equations of this approach . in sec .
[ sec : mc ] we give a detailed report on our large - scale monte carlo simulations and the numerical results , together with a discussion of their relations to the fossil record and to other theoretical models . finally , in sec .
[ sec : concl ] , we summarize our results and give our conclusions and suggestions for future studies .
as mentioned in sec . [ sec : int ] , we use a simplified version of the tangled - nature model introduced by hall , christensen , and collaborators @xcite .
it consists of a population of individuals with a genome of @xmath5 genes , each of which can take one of the two values 0 or 1 .
thus , the total number of different genotypes is @xmath6 .
we consider each different genotype as a separate species , and we shall in this paper use the two terms interchangeably .
this is justified by the idea that in the relatively short genomes we can consider computationally , each binary `` gene '' actually represents a group of genes in a coarse - grained sense . in our version of the model
, the population evolves in discrete , nonoverlapping generations ( as in , e.g. , many insects ) , and the number of individuals of genotype @xmath7 in generation @xmath8 is @xmath9 .
the total population is @xmath10 . in each generation , the probability that an individual of genotype @xmath7 produces a litter of @xmath11 offspring before it dies is @xmath12 , while the probability that it dies without offspring is @xmath13 . although the fecundity @xmath11 could be quite complex in reality ( e.g. , a function of @xmath7 on the average , but random both in time and for each individual ) , we here take a simplistic approach and assume it is a constant , independent of @xmath7 and @xmath8 .
the main difference from the model of refs .
@xcite is that the generations are nonoverlapping in our model , while individuals in the original model may live through several successive reproduction cycles .
this simplification facilitates theoretical analysis and enables us to make significantly longer simulations than those reported in refs .
@xcite . in the `` opposite '' direction
, the earlier model could be generalized to include non - trivial `` age structures , '' since individuals living through several cycles can be assigned an age , with age - dependent survival and reproductive properties @xcite . as in the original model @xcite , the reproduction probability @xmath14 is taken as @xmath15 } \;. \label{eq : p}\ ] ] here , the verhulst factor @xmath16 @xcite represents an environmental `` carrying capacity '' that might be due to limitations on shared resources , such as space , light , or water .
it prevents the total population from indefinite growth , stabilizing it at @xmath17 .
the interaction matrix @xmath18 expresses pair interactions between different species such that the element @xmath19 gives the effect of the population density of species @xmath20 on species @xmath7 .
thus , a mutualistic relationship is represented by both @xmath19 and @xmath21 being positive , while both being negative models a competitive relationship .
if they are of opposite signs , we have a predator - prey situation . to concentrate attention on the effects of interspecies interactions , we follow refs .
@xcite in setting the self - interactions @xmath22 .
the offdiagonal elements of @xmath19 are randomly and uniformly distributed on @xmath23 $ ] as in ref .
the interaction matrix is set up at the beginning of the simulation and is not changed later , a situation that corresponds to quenched disorder in spin - glass models @xcite .
we note that we have not attempted to give @xmath18 a particularly biologically realistic form .
some possible modifications are discussed in sec .
[ sec : concl ] . in this model
there is a one - to - one correspondence between genotype ( the @xmath7th specific bit string ) and phenotype ( the @xmath7th row and column of @xmath18 ) .
thus , the phenotype specifies both how the @xmath7th species influences the other species that are either actually or only potentially present in the community ( the @xmath7th column ) and how it is influenced by others ( the @xmath7th row ) .
the reproduction probability @xmath14 provides selection of the `` most fit '' phenotypes according to these `` traits '' ( matrix elements ) and the populations of the other species present in the community .
the effect ( or lack thereof ) of a particular trait depends on the community in which the species exists , just as a cheetah s superior speed is only relevant to survival if fast - moving prey is available . in each generation , the genomes of the individual offspring organisms are subjected to mutation with probability @xmath24 per gene and individual .
mutated offspring are re - assigned to their new genotypes before the start of the next generation .
this provides the genetic variability necessary for evolution to proceed . the analytic form of @xmath12 , eq .
( [ eq : p ] ) , is by no means unique .
for instance , one could use a `` soft dynamic '' @xcite in which the effects of the interactions and the verhulst factor factorize in @xmath14 .
alternatively , one could dispense with the exponential and use linear or bilinear relationships instead , as is most common in population biology @xcite .
investigations of the effects of such modifications are left for the future .
our evolution algorithm proceeds in three layers of nested loops .
* loop over generations @xmath8 . *
loop over the @xmath25 populated genotypes @xmath7 in generation @xmath8 .
* loop of length @xmath9 over the individuals of genotype @xmath7 .
each individual produces @xmath11 offspring for generation
@xmath26 with probability @xmath12 , or dies without offspring with probability @xmath13 . in either case ,
no individual survives from generation @xmath8 to generation @xmath26 .
* loop of length equal to the total number of offspring of genotype @xmath7 generated in loop 3a , attempting to mutate each gene of each individual offspring with probability @xmath24 and moving mutated offspring to their respective new genotypes for generation @xmath26 .
though neither our model nor the simulations are deterministic , a number of their gross properties can be understood in terms of a mean - field approximation that ignores statistical fluctuations and correlations .
the time evolution of the populations is then given by the set of difference equations , @xmath27+(\mu /l)f\sum_{k(i)}n_{k(i)}(t)p_{k(i)}(\{n_j(t)\})+o(\mu ^2)\ ; , \label{eq : eom}\ ] ] where @xmath28 runs over the @xmath5 genotypes @xmath29 that differ from @xmath7 by one single mutation ( i.e. , the hamming distance @xcite @xmath30 ) .
the corrections of @xmath31 correspond to multiple mutations in single individuals . naturally , a full investigation of this equation is highly non - trivial , even in the absence of random noise .
in particular , eq . ( [ eq : eom ] ) can be regarded as a logistic map in a @xmath32-dimensional space .
logistic maps are known to admit , in general , both fixed points and cycles of non - trivial periods @xcite . to keep both analysis of simulations and theoretical investigations manageable , we choose parameters in such a way that we can focus our attention on fixed points .
of course , what we simulate are stochastic processes .
however , as long as the mutation rate is well below the error threshold for mutational melt - down @xcite , new successful mutations can become fixed in the population before another successful mutant arises .
this results in a separation of time scales between the ecological scale of fluctuations within fixed - point communities and the evolutional scale of the durations of such communities @xcite , which are known as quasi - steady states ( qss ) or quasi evolutionarily stable strategies ( q - ess ) @xcite . by ignoring mutations
, we can make some analytical progress and gain some insight into the nature and stability properties of these qss . a fixed point of eq .
( [ eq : eom ] ) , defined by @xmath34 is characterized by having only @xmath33 ( @xmath35 ) non - vanishing @xmath36 .
( henceforth , an asterisk will signify a quantity as a fixed - point value . ) following ref .
@xcite , we denote a fixed point as _ feasible _ if @xmath37 for all @xmath33 values of @xmath7 . corresponding to the coexistence of @xmath33 genotypes , such a point will be simply referred to as an `` @xmath33-species fixed point . ''
when mutations are ignored , eq .
( [ eq : eom ] ) reduces to @xmath38 specializing to the form of @xmath14 given by eq .
( [ eq : p ] ) , we see that all single - species fixed points are trivially `` identical , '' with @xmath39 .
this somewhat unrealistic result is just a consequence of our choice of @xmath22 and can be avoided by lifting this restriction .
however , the absence of self - interactions places no restriction on the main purpose of our work the exploration of the effects of random but time independent interspecies interactions . proceeding to the @xmath40 cases
, the existence of an @xmath33-species fixed point depends critically on the submatrix @xmath41 , with matrix elements @xmath19 in which both @xmath7 and @xmath20 are among the @xmath33 species in question .
( we shall use the tilde to emphasize that a quantity corresponds to an @xmath33-species subspace , rather than to the full , @xmath32-dimensional genotype space . ) in particular , if @xmath41 is non - singular , then a unique fixed point exists , as we show below . on the other hand , singular @xmath41 s may result in a variety of `` degenerate '' cases .
we provide just two examples to illustrate the mathematical richness of singular @xmath41 s .
if all elements vanish , then the behavior in this subspace is highly degenerate , with the _ total _ population given again by @xmath42 , regardless of @xmath33 .
however , the fractions of each species , @xmath43 are completely undetermined , an understandable consequence of having dynamically indistinguishable species .
another example of a singular interaction submatrix is @xmath44 in an @xmath45 subspace .
then , eq . ( [ eq : mu=0 ] ) will drive one of the species to extinction , so that no fixed point with both @xmath46 ( stable or unstable ) can exist in this two - dimensional subspace , and the system collapses to @xmath47 . for the remainder of this article
, we shall study analytically only the dynamical behavior of interacting species with nonsingular @xmath41 s . to proceed , we insert eq .
( [ eq : fp ] ) into eq .
( [ eq : mu=0 ] ) and arrive at the @xmath33 equations @xmath48 with our choice of @xmath14 , these lead to @xmath49 ( the analysis in this section remains valid even if @xmath50 , although @xmath22 is the case explicitly considered elsewhere in this paper . )
note that the right - hand side of eq .
( [ eq : fix2 ] ) is _ independent _ of @xmath7 .
since @xmath41 is non - singular , we define its inverse by @xmath51 , with elements @xmath52 : @xmath53 then eq . ( [ eq : fix2 ] ) can be inverted to give @xmath54 \sum_j\tilde{w}_{ij } \,\ , .
\label{eq : fix3}\ ] ] the only unknown in this equation , @xmath55 , can now be found via the normalization condition , @xmath56 , which leads to @xmath57 where @xmath58 putting these into eq .
( [ eq : fix3 ] ) , we have the explicit form of the fixed - point populations : @xmath59 \frac{\sum_j\tilde{w}_{ij}}{\tilde{\sigma } } \;. \label{eq : fix5}\ ] ] although eq .
( [ eq : fix5 ] ) appears to provide fixed - point values for _ any _ choice of control parameters , we emphasize that it is applicable only for a limited range of @xmath11 and @xmath41 .
the subtlety lies in its stability properties within the @xmath33-species subspace .
first , we remind the reader that , even in the case of @xmath60 , there are a variety of behaviors .
solutions may have a stable fixed point with either monotonic or oscillatory decay of small deviations , or they may show bifurcations , period doubling , and chaos @xcite . in the present study , we are interested in the effects of the interspecies interactions , and we therefore choose to focus on systems with monotonically decaying fluctuations in the noninteracting limit .
this means that small fluctuations about the fixed point must decay as @xmath61 . for the single - species and @xmath62 cases ,
this restriction is easily translated to the condition @xmath63 for the fecundity .
we use @xmath64 in all of our simulations .
next , through a straightforward linear stability analysis around the @xmath33-species fixed point , we obtain a condition that represents decay of all small perturbations , i.e. , all the eigenvalues of the matrix @xmath65 must have real parts , lying between @xmath66 and 1 .
carrying out the differentiation and using eq .
( [ eq : mu=0 ] ) to simplify the result , we find this matrix to be of the form @xmath67 , where @xmath68 is the @xmath33-dimensional unit matrix and the elements of @xmath69 are given by @xmath70 our criterion translates to the requirement that all the eigenvalues of @xmath69 must have real parts that lie in @xmath71 .
fixed points with this property will be called `` internally stable . ''
unfortunately , this criterion can not be made more explicit . given @xmath11 and a set of @xmath19
, @xmath69 must be constructed using eqs .
( [ eq : w])([eq : sigma ] ) and diagonalized .
the matrix @xmath69 is recognized as what is known in ecology as the _ community matrix _
@xcite of the @xmath33-species fixed point for the discrete - time dynamic defined by eq .
( [ eq : mu=0 ] ) .
it has been shown , both numerically @xcite and analytically @xcite , that the proportion of large , random matrices for which all eigenvalues have a negative real part vanishes as the proportion of nonzero matrix elements increases .
this has been used as an argument that highly connected ecosystems are intrinsically unstable @xcite , contrary to ecological intuition .
however , the matrix that must be studied to determine the internal stability of a fixed point is the community matrix @xmath69 , which has a complicated relationship to the interaction matrix @xmath18 and should not be expected to have a simple element distribution .
in fact , for some bilinear population dynamics models there is numerical evidence that most _ feasible _ fixed points are also internally stable @xcite .
however , the relations between connectivity and stability have not yet been fully clarified and are still being discussed @xcite . of course , issues of internal stability are somewhat academic for simulations . in practice
an internally unstable fixed point could not be observed for more than a brief time , especially since the populations would be driven away from such fixed points by the noise due to both the birth / death process and the mutations .
since mutations are essential for the long - time evolutionary behavior , the `` external stability '' properties of the fixed - point communities are important .
even if the population corresponds to an internally stable @xmath33-species fixed point , mutations will generate small populations of `` invader '' species , i.e. , genotypes outside the resident @xmath33-species community . denoting such invaders by the subscript @xmath72 and linearizing eq .
( [ eq : mu=0 ] ) about the @xmath33-species fixed point , we see that the important quantity is the multiplication rate of the invader species in the limit of vanishing @xmath73 , @xmath74 } \;. \label{eq : other}\ ] ] to obtain this result we exploited eqs .
( [ eq : fix4 ] ) and ( [ eq : fix5 ] ) .
explicitly , the condition for stability against the invader , @xmath75 , reduces to the requirement that the argument of the exponential function in eq .
( [ eq : other ] ) must be positive .
we also note that , if @xmath76 for all @xmath20 in the resident community , then the multiplication rate of the invader equals unity .
in population biology the lyapunov exponent @xmath77 $ ] is known as the _ invasion fitness _ of the mutant with respect to the resident community @xcite . from eq .
( [ eq : other ] ) it becomes clear that the success of an invader depends not only on its direct interactions @xmath78 with each of the resident species , but also on the interactions between the resident species through the inverse interaction submatrix @xmath79 .
the model described in sec .
[ sec : mod ] was studied by monte carlo simulations with the following parameters : @xmath80 , @xmath64 , @xmath81 , and @xmath82 .
the random matrix @xmath18 ( with zero diagonal and other elements randomly distributed on @xmath23 $ ] ) was chosen at the beginning of each simulation run and then kept constant for the duration of the run . in this regime both
the number of populated species @xmath25 and the total population @xmath83 are substantially smaller than the number of possible species , @xmath84 .
this appears biologically reasonable in view of the enormous number of different possible genotypes in nature .
( but see further discussion in sec.[sec : stat ] and appendix [ sec : ac ] . ) in a qss , the average number of mutant offspring of species @xmath7 in generation @xmath26 is approximately @xmath85 .
thus , with the mutation rate used here , each of the dominant species will produce of the order of one mutant organism per generation . as shown in sec .
[ sec : stab ] , during a qss most of these mutants become extinct after one generation .
however , due to the small genome size , the same mutant of a species with a large population will be re - generated repeatedly by mutation from the parent species . a high - population genotype with its `` cloud '' of closely related low - population mutants
could be considered a quasi - species in the sense of eigen , with the high - population genotype as the `` wildtype '' @xcite .
an alternative interpretation of the model is therefore as one of the coevolution of quasi - species @xcite , in which the successful invasion of a resident community by a mutant represents a speciation event in the lineage of the parent genotype .
most of our simulations were started with a small population of 100 individuals of a randomly selected single genotype , corresponding to the entry into an empty ecological niche by a small group of identical individuals
. however , runs starting from a random number of small populations of different species give essentially the same results .
generally , the initial species are not likely to be stable against mutants , and they usually become extinct within 100 generations . to eliminate any short initial transients , most of the quantitative analyses presented here are based on time series from which the first 4096 generations were removed .
the simulated quantities were recorded every 16 generations .
this time resolution was chosen to be just larger than the average time it would take for the descendants of a single individual of a mutant genotype @xmath72 to completely replace a resident genotype @xmath20 of @xmath86 individuals in the case of a maximally aggressive mutant , @xmath87 and @xmath88 .
this time , which is obtained by numerical solution of eq .
( [ eq : other ] ) , is about 15 generations and represents a minimum growth time " for the model .
sampling at shorter intervals would mostly add random noise to the results , while sampling on a much coarser scale could miss important evolutionary events .
it is easy to show , by solving eq .
( [ eq : other ] ) analytically for short times , that the growth time and thus the optimal sampling interval increases logarithmically with @xmath16 .
most quantitative results in this paper are based on averages over 16 independent simulation runs of @xmath89 generations each @xcite .
we define the _ diversity _ of the population as the number of species with significant populations , thus excluding small populations of mainly unsuccessful mutants of the major genotypes .
operationally we define the diversity as @xmath90 $ ] , where @xmath91 is the information - theoretical entropy @xcite , @xmath92 with @xmath93 this measure of diversity is known in ecology as the shannon - wiener index @xcite .
it is different from the definition of diversity as the number of populated species ( known in ecology as the _ species richness _
@xcite ) that was used in refs . @xcite .
the entropy - based measure significantly reduces the noise during qss , caused by unsuccessful mutations .
the shannon - wiener diversity index is shown in fig .
[ fig : timeser](a ) for a simulation of @xmath94 generations , together with the normalized total population , @xmath95 $ ] .
we see quiet periods during which @xmath96 is constant except for small fluctuations , separated by periods during which it fluctuates wildly . the total population @xmath97 is enhanced relative to its noninteracting fixed - point value during the quiet periods , while it decreases toward the vicinity of this value during the active periods . to verify that different quiet periods indeed correspond to different resident communities @xmath98 , we show in fig .
[ fig : spec ] the genotype labels @xmath7 ( integers between 0 and @xmath99 , corresponding to the decimal representation of the genotype bit string ) versus time for the populated species .
the populations of the different species are indicated by the grayscale ( by color online ) .
we see that , in general , the community is completely rearranged during the active periods , so that the quiet periods can be identified with the qss of the ecology .
an exception is afforded by the event near @xmath100 generations in figs .
[ fig : timeser](a ) and [ fig : spec](a ) . in this instance
the populations of the dominant species decrease , and the total population is for a brief time spread over a larger number of species .
however , the dominant species regain their footing , " and the original resident community continues for approximately another 50000 generations .
this situation is reminiscent of rare events in nucleation theory @xcite , in which a fluctuation of the order of a critical or even a supercritical droplet nevertheless may decay back to the metastable state . to investigate the stability properties of individual communities against invaders
, we chose from a particular simulation run of @xmath94 generations ten different qss of durations longer than 20000 generations .
the genotype labels and fixed - point populations that characterize these qss are given in the first four columns of table [ table1 ] .
the population - weighted average of the hamming distances @xmath101 between pairs of genotypes in the community , @xmath102 and the corresponding standard deviation , @xmath103 are shown in the fifth and sixth column , respectively . even though the community as a whole moves far through the @xmath32-dimensional population space , the different qss communities are seen to retain the property that they consist of relatively close relatives ( which of course are all descendants of the single , initial genotype ) . in sec .
[ sec : stat ] we demonstrate that @xmath104 and @xmath105 remain in the range shown in table [ table1 ] , even during very long simulations .
the averages and standard deviations of the offdiagonal interaction - matrix elements @xmath106 between the community members are shown in the seventh and eighth column of table [ table1 ] , respectively .
they show that the qss communities are strongly mutualistic , as was also observed for the stable states in ref . @xcite .
in contrast , the matrix elements of 22 feasible , but otherwise randomly chosen , communities @xcite were found to be approximately uniformly distributed on @xmath23 $ ] . in fig .
[ fig : eigenval ] we show histograms of the multiplication rates eq .
( [ eq : other ] ) ( i.e. , the exponential function of the invasion fitness ) of mutants that differ from the resident species by one single mutation ( nearest - neighbor species , @xmath107 ) , and by two mutations ( next - nearest - neighbor species , @xmath108 ) @xcite .
only a very small proportion of the nearest - neighbor mutants have a multiplication rate above unity [ fig . [
fig : eigenval](a ) ] .
( in our sample , this very small proportion came from just one of the ten qss considered . ) among the next - nearest neighbors , on the other hand , a not insignificant proportion may be successful invaders [ fig .
[ fig : eigenval](b ) ] .
the picture for third - nearest neighbors ( not shown ) is essentially the same as for the next - nearest neighbors .
we also considered the multiplication rates for neighbors of the two most long - lived qss observed in our simulations , which lasted about @xmath109 and @xmath110 generations , respectively .
they both had no nearest neighbors with multiplication rates above unity , and the proportions for second - and third - neighbors were significantly smaller than those included in fig .
[ fig : eigenval ] . in both parts of fig .
[ fig : eigenval ] are also shown the multiplication rates for neighbors of the 22 randomly chosen , feasible communities introduced in the previous paragraph . in that case , approximately half of the neighbors at any hamming distance have multiplication rates above unity . summarizing the results from table [ table1 ] and fig .
[ fig : eigenval ] , we see that a long - lived qss community is characterized by strongly mutualistic interactions and is surrounded by a `` protection zone '' of closely related genotypes that are very unlikely to successfully invade the resident community .
the evidence from the two most long - lived qss indicates that the average lifetime of a qss is positively correlated with the extent of the protection zone .
by contrast , a randomly chosen , feasible community has an approximately uniform distribution of @xmath106 over @xmath23 $ ] .
it also typically has no protection zone and so is more vulnerable to invasion than a qss .
it is clear from the fact that none of our randomly chosen feasible communities turned out to be a qss that qss are relatively rare in this model , even among feasible communities .
we believe this may be a favorable condition for continuing evolution , as the ecology can move rapidly from qss to qss through series of unstable communities .
the protection zone surrounding a qss community acts much like the free - energy barrier that separates a metastable state in a physical system from the stable state or other metastable states . in both cases a sequence of improbable mutations or fluctuations
is required to reach a critical state that will lead to major rearrangement of the system .
it is thus natural to investigate in detail the lifetime statistics of individual qss in the way common in the study of metastable decay @xcite . to this effect
we started simulations from each of the ten qss communities discussed in sec .
[ sec : stab ] and ran until the overlap function with the initial community , @xmath111 became less than 0.5 , at which time the system was considered to have escaped from the initial state .
( we note that the overlap function defined here is simply the cosine of the angle between two unit vectors in the space of population vectors @xmath98 . )
the precise value of the cutoff used is not important , as long as it is low enough to exclude fluctuations during the qss @xcite .
the composition of the population at this time ( the `` exit community '' ) was then recorded and the run terminated .
each individual initial qss was simulated for 300 independent escapes .
the means and standard deviations of the individual lifetime distributions are shown in the last two columns of table [ table1 ] .
the mean lifetimes were found to range over about one order of magnitude , from approximately 10000 to 80000 generations . in most cases
the individual lifetime distributions were exponential ( for which the standard deviation equals the mean ) , or they had an exponential component that described the behavior in the short - time end of the range of observed lifetimes . in about half of the qss studied , a tail of very long lifetimes beyond the exponential part ( indicated by a standard deviation significantly larger than the mean ) was also observed .
we further found that an initial qss does not always escape via the same exit community . typically , the genotypes in the exit community that are not present in the initial qss differ from those in the original community by a hamming distance of two or three .
this picture is quite consistent with the stability properties of the qss discussed in sec .
[ sec : stab ] .
the range of the lifetimes of the ten qss discussed above was limited : for a period to be identified by eye as a qss , its lifetime could not be too short , while the lifetimes were bounded above by the length of the simulation run .
the variation of a factor of ten within these limits indicates that the dynamical behavior of the system may display a very wide range of time scales .
another indication to this effect is provided by the details of the activity within periods that appear as high - activity in figs . [ fig : timeser](a ) and [ fig : spec](a ) .
such detail is shown in figs .
[ fig : timeser](b ) and [ fig : spec](b ) .
statistically , the picture in these expanded figures is similar to the one seen on the larger scale , with shorter quiet periods punctuated by even shorter bursts of activity .
these observations , which are similar to refs .
@xcite , suggest statistical self - similarity at least over some range of time scales .
indications of self similarity have also been seen in fossil diversity records @xcite . the suggestion of self - similarity that emerges from figs .
[ fig : timeser ] and [ fig : spec ] makes it natural to investigate the statistics of the durations of active periods and qss over a wider range of time scales than that represented by the ten qss included in table [ table1 ] . from fig .
[ fig : timeser ] we see that a reasonable method to make this distinction is to observe the magnitude of the entropy changes , @xmath112 , and to consider the system to be in the active state if this quantity is above some suitably chosen cutoff .
this cutoff can best be determined from a histogram of @xmath113 , such as shown in fig .
[ fig : periods](a ) .
it shows that the probability density of the entropy derivative consists of two additive parts : a near - gaussian one corresponding to population fluctuations caused mostly by the birth / death process in the qss , and a second one corresponding to large changes during the active periods . from this figure we chose the cutoff as 0.015 , which was used to classify every sampled time point as either active or quiet .
normalized histograms for the durations of the active and quiet periods are shown in fig .
[ fig : periods](b ) . about 97.4% of the time
is spent in qss .
the active periods are seen to be relatively short , and their probability density is fitted well by an exponential distribution @xcite . on the other hand , the lifetimes of the qss show a very broad distribution @xcite possibly a power law with an exponent near @xmath1142 for durations longer than about 200 generations .
we note that there is some evidence of power laws with exponents near @xmath1142 in the distributions of several quantities extracted from the fossil record , such as the life spans of genera or families , and the sizes of extinctions @xcite . however , the fossil data are sparse and extend over no more than one or two decades in time , and they can be fitted almost as well by exponential distributions @xcite .
although our data extend over a much wider range of time scales than the paleontological data , indisputable evidence of a power law remains to be established .
though a few qss of the order of one million generations will certainly appear in every run of @xmath115 generations , more definite conclusions about the statistics of such long qss must await simulations an order of magnitude longer .
another feature in fig .
[ fig : periods](b ) is that the distribution of qss durations changes to a smaller slope below about 200 generations .
a similar effect has been observed in the fossil record of lifetimes of families @xcite . from the discussion in sec .
[ sec : life ] it is clear that by using a cutoff for the intensity of some variable which is large in the qss and small in the active periods ( but in both cases with fluctuations of unknown distribution ) , it is impossible to classify the periods unambiguously . for example , by increasing the cutoff one can make appear as a single qss period what previously appeared as two successive , shorter qss periods separated by a short active period . since the probability of encountering an extremely large fluctuation in a qss increases proportionally to the length of the qss , this effect will affect the longest periods most severely . the problem with determining a suitable cutoff can be avoided by , instead of concentrating on period statistics , calculating the power spectral density ( psd ) @xcite of variables such as the shannon - wiener diversity index or the total population .
psds for these two quantities are shown in fig .
[ fig : psd ] .
the psd for the diversity @xmath96 , shown in fig .
[ fig : psd](a ) , appears inversely proportional to the frequency @xmath116 ( @xmath1 noise or flicker noise @xcite ) for @xmath117 , goes through a crossover regime of about one decade in @xmath116 where it is @xmath118 with @xmath119 , and then appears to return to @xmath120 for @xmath121 . for @xmath122 our data are insufficient to determine the psd unambiguously , and much longer simulations would be necessary @xcite . in the psd for @xmath97 , shown in fig . [
fig : psd](b ) , the substantial population fluctuations due to the birth / death process during the qss periods produce a large noise background which interferes with the interpretation at high frequencies .
however , for @xmath123 the behavior is also consistent with @xmath1 noise . on time
scales much longer than the mean duration of an active period , the time series for the diversity can be approximated by constants during the qss , interrupted by delta - function like spikes corresponding to the active periods ( see fig . [
fig : timeser ] ) . in this limit
, the relation between a very wide distribution of the qss durations @xmath124 , described by a long - time power - law dependence of the probability density @xmath125 , and the low - frequency behavior of the psd is known analytically @xcite : @xmath126 thus , the approximate @xmath1 behavior of the psds in fig .
[ fig : psd ] is consistent with the approximate @xmath127 behavior of the probability density for the qss durations , shown in fig .
[ fig : periods](b ) . as well as
power - law distributions , psds that go as @xmath128 with @xmath129 have been extracted from the fossil record @xcite . however ,
such observed psds extend only over one to two decades in frequency ( corresponding to power - law probability densities extending only over one or two decades in time ) , and more recent work indicates that the @xmath1 spectra obtained in ref .
@xcite may be artifacts of the analysis method @xcite . although @xmath1 noise , at least in some frequency interval , is a property of our model and is possibly also seen in other models of macroevolution @xcite ( but see refs .
@xcite for conflicting opinions on its presence in the bak - sneppen model ) , whether or not it is really present in the fossil record remains an open question .
an important issue in evolutionary biology is whether or not the evolving ecosystem is stationary in a statistical sense . in fig .
[ fig : runave ] we show several diversity - related measures , averaged over a moving time window and independent simulation runs .
these quantities are : the species richness @xmath25 , the number @xmath130 of genotypes with @xmath131 , the shannon - wiener index @xmath96 , the total population @xmath97 , and the average hamming distance @xmath104 between genotypes in the community and its standard deviation @xmath105 .
as seen in the figure , none of these quantities show any signs of a long - time trend over @xmath115 generations .
this is consistent with fossil evidence for constant diversity @xcite .
however , it is in disagreement with simulations of the original version of the tangled - nature model @xcite , in which a slow growth in species richness was observed .
this discrepancy might possibly result from the different forms of the interaction matrix used in the two studies , but it may also be due to the relatively short time series used in ref .
@xcite . for our model ,
a simple , combinatorial phase - space argument ( containing neither the individual population sizes nor the mutation rate ) indicates that the most probable number of major genotypes should indeed be limited .
the argument goes as follows .
a community of @xmath33 genotypes can be chosen in @xmath132 different ways and is influenced by @xmath133 different pairs of @xmath19 and @xmath21 .
we let the probability that the pair of interactions is suitable to forming a stable community be @xmath134 ( if the requirement is simply that both @xmath19 and @xmath21 should be positive , then @xmath135 with our choice of @xmath18 ) .
thus , a stable @xmath33-species community can be formed in approximately @xmath136 ways , which obeys the recursion relation @xmath137 with @xmath138 .
this recursion relation can either be used to find the most probable value @xmath139 numerically , or one can easily obtain an estimate valid for @xmath140 : @xmath141 .
the value numerically obtained with @xmath135 is @xmath142 , but the close correspondence to the results shown in fig .
[ fig : runave ] may be fortuitous since the dependence of @xmath143 on @xmath134 is only logarithmic .
another question of interest is whether the limited size of the genome leads to `` revisiting '' of genotypes and communities .
the answer is affirmative and indicates the importance of further studies of the effects of the genome size on the long - time dynamical behavior . for genotypes
the question of revisits is easy to answer , and the results for a single run of @xmath115 generations are shown in fig .
[ fig : specvis ] . as we see , in less than @xmath144 generations ,
at least two individuals of every genotype have appeared simultaneously ( curve labeled @xmath145 ) . by the end of the run ,
almost every genotype has enjoyed being a major species with @xmath146 at least once .
plots on a log versus linear scale ( not shown ) indicate that the curves are reasonably well approximated by exponential approaches to @xmath147 , indicating that genotypes appear to be nearly randomly visited and revisited at a constant rate dependent on @xmath148 . for communities ,
the question is more difficult .
if a community is simply defined as an internally stable @xmath149-species community , then an estimate for their total number can be obtained from @xmath150 of eq .
( [ eq : omega ] ) as described in appendix [ sec : ac ] .
the result with the parameter values used in this paper is of the order of @xmath151 .
however , unstable communities are also briefly visited , especially during active periods , and a more inclusive way of counting communities would be a coarse - graining procedure based on the overlap function defined in eq .
( [ eq : over ] ) .
the nonzero populations are recorded at @xmath152 , and the overlap with this initial community is monitored until at some @xmath153 it falls below a suitably chosen cutoff @xmath154 .
the set @xmath155 is then recorded as a new community with @xmath156 as its starting time .
the process is repeated , now comparing with the newly recorded community .
this procedure creates a list of communities and their starting times . to address the issue of revisits
, we next scan this list to extract those communities that represent revisits to a previously visited community , or `` prototype community . ''
we compare each community @xmath157 in the original list with the previously found communities @xmath158 , sequentially in order of increasing @xmath159 .
if none of the overlaps is greater than @xmath154 , @xmath157 is added to the list of prototypes @xmath160 with @xmath161 .
if an overlap greater than @xmath154 is found at some @xmath162 , we stop the comparisons and add this community to the list of revisits @xmath163 .
the starting time of this revisited community , @xmath164 , is associated in the list @xmath163 with the unique @xmath165 , the starting time of the associated prototype .
as the system evolves , the number of items in each list increases monotonically .
clearly , the cutoffs should be sufficiently small that communities that differ only by fluctuations inside a qss are not considered different , but sufficiently large to avoid counting significantly different communities as identical . from inspection of the overlap fluctuations in some of the qss
included in table [ table1 ] , we found that @xmath154 in the range of 0.90 to 0.95 is optimal .
estimates of the total numbers of communities ( stable and unstable ) that can in principle be distinguished by this method are obtained in appendix [ sec : ac ] .
depending on details of the assumptions , the estimates vary between @xmath166 and @xmath167 for these cutoffs .
the results of the procedure described above with @xmath168 and @xmath169 are shown in fig .
[ fig : commvis ] for a single run of @xmath115 generations .
the upper pair of curves in fig .
[ fig : commvis](a ) corresponds to @xmath170 , and the lower pair to @xmath171 . in each pair ,
the upper curve shows the total number of communities in the lists @xmath172 and @xmath173 , while the lower curve shows just the number of prototypes in @xmath174 . for @xmath175 ,
about 40% of the communities are seen to be revisits , while the proportion is about 30% for @xmath168 . in both cases ,
the curve for the number of prototypes remains approximately linear , indicating that the supply of previously unvisited communities is nowhere near to being exhausted , even for such a long run . in view of the enormous numbers of available communities estimated above , this result is reasonable .
qss appear as plateaus in the curves showing the numbers of prototypes .
figure [ fig : commvis](b ) provides a different perspective ( for @xmath170 only ) .
each revisit is represented by a point with @xmath176 as abscissa and @xmath177 as ordinate .
thus , every point lies strictly below the diagonal .
long - lived qss appear as large gaps and horizontal segments ( e.g. , the one near @xmath178 generations ) . the inset shows a detail of @xmath94 generations near the diagonal around @xmath179 .
as we see , there are many points _ just below _ the diagonal , representing the fluctuations around the ( many short - lived ) qss . by contrast
, points far below the diagonal represent `` throwbacks '' to the vicinity of earlier prototype communities .
note that the density of points is much higher just below the diagonal , implying that a large portion of the revisited communities are `` fluctuation related . '' to highlight these differences , we show the cumulative probability of the ratio @xmath180 in fig .
[ fig : commvis](c ) , in which the lower curve corresponds to the data in fig .
[ fig : commvis](b ) . for the case of @xmath181 ( lower curve ) , we see that about 60% of the revisits can be regarded as `` fluctuation related '' ( with @xmath182 ) , while the rest are `` throwbacks . ''
roughly , the latter component appears to be distributed uniformly over all earlier times .
the upper curve shows the corresponding result for @xmath168 .
corresponding to using a more coarse - grained covering of state space , it naturally displays a larger proportion of throwbacks .
not surprisingly , there is no sharp distinction between these two components of the revisited communities since even a rough partition depends on the details of coarse graining .
nevertheless , we can conclude that the dynamics produces a steady stream of essentially new communities drawn from the vast supply of possibilities .
in this work we have studied , by linear stability analysis and large - scale monte carlo simulations , a simplified version of the tangled - nature model of biological coevolution , recently introduced by hall , christensen , and collaborators @xcite .
selection is provided by interspecies interactions through the reproduction probability @xmath14 [ eq . ( [ eq : p ] ) ] , which corresponds to a nonlinear population - dynamics model of the community ecology , while the genetic variability necessary for evolution is provided by a low rate of mutations that act on individual organisms during reproduction . at the low mutation rate studied here ,
the model provides an intermittent , statistically self - similar behavior , characterized by periods of relative calm , interrupted by bursts of rapid turnover in genotype space . during the quiet periods , or quasi - steady states ( qss ) , the population consists of a community of a relatively small number of mutualistically interacting genotypes .
the populations of the individual genotypes , @xmath9 , fluctuate near a stable fixed point of a deterministic mean - field , mutation - free version of the model [ eq .
( [ eq : mu=0 ] ) ] . during the active periods ,
the system moves through genotype space at a rate that is rapid on a macroscopic ( `` geological '' ) time scale , although of course finite on a microscopic ( `` ecological '' ) scale . these periods of rapid change are characterized by large fluctuations in the diversity and an overall reduction of the total population . in long simulation runs of @xmath115 generations , the ecosystem spends about 97.4% of the time in qss . the time series produced by the model
are statistically stationary , and there is no evidence that any particular quantity is being optimized as the system moves through genotype space . in that sense , the dynamical behavior can probably best be described as a neutral drift @xcite .
overall , the dynamical behavior of this model resembles closely the punctuated - equilibria mode of evolution , proposed by gould and eldredge @xcite .
investigation of duration statistics for the quiet periods shows a very wide distribution with a power - law like long - time behavior characterized by an exponent near @xmath0 .
consistent with this result , the power spectral density of the diversity shows @xmath1 noise .
while there are claims that similar statistics characterize the fossil record @xcite , this is still a contested issue @xcite . at best
, observations of power - law distributions and @xmath1 noise in the fossil record extend over no more than one or two decades in time or frequency , and it must remain an open question whether this is the optimal interpretation of the scarce data available . due to the absence of sexual reproduction , our model can at best be applied to the evolution of asexual , haploid organisms such as bacteria .
it should also be noted that no specific , biologically relevant information has been included in the interaction matrix @xmath18 .
in particular , this fact may be responsible for our qss being strongly dominated by mutualistic relationships .
the absence of biologically motivated detail in @xmath18 is both a strength and a weakness of the model .
its strength lies in reinforcing the notion of universality in macroevolution models , e.g. , power law behaviors and @xmath1 noise . by the same token
, its weakness lies in the lack of biological detail , to the point that comparison with specific observational or experimental data is difficult .
clearly , the detailed effects of interspecies interactions on the macroevolutionary behavior in models similar to the one studied here represents an important field of future research .
examples include the importance of the connectivity of the interaction matrix , correlated interspecies interactions @xcite , and interaction structures corresponding to food webs with distinct trophic levels . despite all these caveats , we find it encouraging that such a simple model of coevolution with individual - based dynamics can produce punctuated equilibria , power - law distributions , and @xmath1 noise consistent with current theories of biological macroevolution .
we believe future research should proceed in the direction pointed out by this and similar models .
this entails combining stochastic models from community ecology with models of mutations and sexual reproduction at the level of individual organisms , and investigating the consequences of more biologically realistic interspecies interactions .
we are grateful for many helpful comments on the manuscript by j. a. travis , and we acknowledge useful conversations and correspondence with i. abou hamad , g. brown , t. f. hansen , n. ito , h. j. jensen , a. kolakowska , s. j. mitchell , m. a. novotny , h. l. richards , b. schmittmann , v. sevim , u. tuber , and j. c. wilgenbusch . p. a. r. appreciates the hospitality of the department of physics , virginia polytechnic institute and state university , and the department of physics and astronomy and erc center for computational sciences , mississippi state university .
the research was supported in part by national science foundation grant nos .
dmr-9981815 , dmr-0088451 , dmr-0120310 , and dmr-0240078 , and by florida state university through the center for materials research and technology and the school of computational science and information technology .
a complete description of the stochastic process can be given in terms of a master equation , which specifies the evolution of @xmath183 , the probability that the system is found with composition @xmath184 at time @xmath8 .
here , @xmath185 , and @xmath6 is the number of individuals of species @xmath7 . in our case , @xmath80 and @xmath186 .
similar to the main text , we define @xmath187 and let @xmath16 be the carrying capacity .
we write the probability for an individual of species @xmath7 to survive to reproduce as @xmath188 \right\ } ^{-1 } .
\label{eq : b1}\ ] ] the main difference between this expression and eq .
( [ eq : p ] ) lies in the interpretation . here
, @xmath184 is a `` coordinate variable '' in the @xmath147-dimensional space , in contrast to @xmath189 being just a point in this space .
to proceed , we define the symbol @xmath190 m_i}$ ] as the rate for survival ( from @xmath148 individuals to @xmath191 `` mothers '' ) .
since each individual is given a chance to survive according to @xmath14 , we have @xmath192 m_i } = \frac{n_i!}{m_i!\left ( n_i - m_i\right ) ! } \left ( p_i\right ) ^{m_i}\left ( 1-p_i\right ) ^{n_i - m_i } , \label{eq : b2}\ ] ] which has the form of a binomial distribution .
next , each mother gives rise to @xmath11 offspring .
however , due to mutations , not every offspring is of the same genotype as the mother .
although it is possible to have mutants with a genotype differing from the mother by more than one bit , we restrict ourselves here to a simpler version , namely mutant genotypes that can differ only by one bit . since our simulations typically involve @xmath193 , this restriction should not lead to serious difficulties . given that only one bit may be flipped , there are @xmath194 possible varieties of offspring for each maternal genotype .
we introduce the notation @xmath195 we now define the multinomial - like symbol @xmath196 b_{j,0},b_{j,1}, ... ,b_{j , l } } } = \frac{\left ( fm_j\right ) ! } { \left ( b_{j,0}\right ) ! } \left ( 1-\mu \right ) ^{b_{j,0}}\prod_{k=1}^l\frac 1{\left ( b_{j , k}\right ) ! } \left ( \frac \mu l \right ) ^{b_{j , k } } , \label{eq : b3}\ ] ] which is the probability that the @xmath197 offspring are distributed into the specific set @xmath198 .
the last ingredient needed is the connection matrix @xmath199 so that the number of offspring born _ into _ species @xmath200 due to mutations is @xmath201 the final result is @xmath202 b_{j,0},b_{j,1}, ... ,b_{j , l } } } \prod_i { { n_i \atopwithdelims [ ] m_i } } { \cal p}\left ( \vec{n},t\right ) \;. \label{eq : b6}\ ] ] in principle , once the dynamics of @xmath203 is found , the time dependence of various quantities ( e.g. , averages and correlations ) can be computed .
for instance , @xmath204 is the average number of individuals of species @xmath7 at time @xmath8 . from this very detailed description ,
we see that the evolution of various quantities can be derived .
however , these evolution equations are extremely complex in general and progress is usually possible only by making a `` mean - field '' approximation , in which all correlations are neglected .
thus , averages of products of @xmath184 are replaced by the products of the averages . as a result , a non - linear evolution equation for @xmath205 results . in a subsequent paper @xcite
, we shall show that eq .
( [ eq : eom ] ) is precisely the result of such an approach ( with @xmath206 denoted by @xmath207 ) .
we shall also demonstrate that , for a qss , @xmath208 is well approximated by a gaussian ( for @xmath209 ) , the center of which is just the fixed point given by eq .
( [ eq : fix5 ] ) .
the width of this gaussian is governed in part by the community matrix @xmath210 of eq .
( [ eq : a2 ] ) .
in addition to this ( systematic , `` drift '' ) part , there is a part governing the noise correlations .
between the two , we can compute distributions of step sizes ( @xmath211 ) . for a long - lived qss ,
it is easy to compile data so that quantitative comparisons with these predictions can be made .
in this appendix we obtain estimates for the numbers of different communities that could in principle be observed in an infinitely long simulation .
even our most conservative estimate is vastly larger than the numbers observed in a simulation run of @xmath115 generations in sec .
[ sec : stat ] .
if a community is simply defined as an internally stable @xmath33-species community , then an estimate for their total number can be obtained from @xmath212 of eq .
( [ eq : omega ] ) .
since @xmath212 is quite sharply peaked around @xmath139 , we can estimate that @xmath213 with @xmath214 . for @xmath135 and @xmath80 ,
this gives @xmath215 , while direct summation of @xmath216 with the same parameters gives @xmath217 with 46% due to the contribution from @xmath218 and convergence by @xmath219 . as it does not include unstable communities
, this would serve as our most conservative estimate of the total number of communities that could be visited in an infinitely long simulation . to estimate the total number of different communities that can be distinguished by the overlap cutoff method described in sec .
[ sec : stat ] , we note that the overlap threshold @xmath154 is simply the cosine of the angle between two vectors in population space , @xmath220 . for convenience ,
we define @xmath221 , such that @xmath222 is small ( @xmath223 and @xmath224 ) .
then , @xmath225 .
the fact that the number of populated species at any one time is near @xmath226 , means that the direction of the population vector @xmath98 to a reasonable approximation lies within the first hyper - octant of one of @xmath227 @xmath226-dimensional hyperspheres .
the total number of communities that can be distinguished with a cutoff @xmath228 is the ratio of the total @xmath226-dimensional solid angle covered by the hyper - octants to that of a hyper - cone subtended by @xmath229 .
the solid angle of the hyper - octants is @xmath230 where @xmath231 is the surface area of a @xmath232-dimensional unit sphere .
the solid angle of the hyper - cone is @xmath233 for small @xmath234 . using stirling s approximation for the gamma functions in @xmath235 , @xmath236 and the identity @xmath237
, we estimate the total number of different communities to be @xmath238 this estimate is reasonable only as long as @xmath234 is large enough to exclude minor fluctuations within qss .
observation of overlap fluctuations in some of the qss included in table [ table1 ] indicates that @xmath228 in the range of 0.90 to 0.95 is optimal . with the parameters used in this work , eq .
( [ eq : ac3 ] ) yields @xmath239 for @xmath240 and @xmath241 for 0.95 .
if , instead of using @xmath242 we assume that the number of genotypes at any time is near 8 , as indicated by the top curve in fig .
[ fig : runave ] , these estimates are increased by six orders of magnitude . p. a. rikvold and r. k. p.
computer simulation studies in condensed matter physics xvi _ , edited by d. p. landau , s. p. lewis , and h .- b .
schttler ( springer - verlag , berlin , in press ) , e - print arxiv : nlin.ao/0303010 .
we constructed the arbitrary , feasible communities by the same method described in ref .
@xcite : from a larger set of randomly selected genotypes we eliminated those with the most negative @xmath243 until all remaining @xmath37 .
the resulting communities have @xmath244 $ ] with a mean of 2.5 .
we define an @xmath245th neighbor of a community of @xmath33 resident species as a genotype that can be created from one or more of the resident genotypes by flipping @xmath245 bits ( hamming distance @xmath245 ) , and which is not also a member of the original community or a neighbor of order @xmath246 .
if an @xmath245th neighbor can be created by flipping @xmath245 bits in more than one of the resident genotypes , it is counted with the corresponding multiplicity in the histograms shown in fig .
[ fig : eigenval ] . since the time scale of the durations of the active periods turned out to be so short compared with the sampling interval of 16 generations , only the general _ shape _ of the duration distribution marked by @xmath247 in fig .
[ fig : periods](b ) can be taken seriously . in order to resolve the short - time dynamical behavior better ,
we therefore performed three independent simulations of length @xmath248 generations with sampling every generation . from a histogram of @xmath249 analogous to fig .
[ fig : periods](a ) we determined a cutoff of @xmath250 ( @xmath251 times the cutoff used with sampling every 16 generations ) .
the distribution of active periods was found to be extremely narrow and exponential - like with an average of only about 1.6 generations .
the distribution of the qss durations was not significantly affected by the change in sampling rate .
the psd of a variable @xmath252 , where @xmath8 takes @xmath253 discrete values with an increment @xmath254 , is here defined as @xmath255 \exp ( 2 \pi i f t ) |^2 $ ] , where @xmath256 is the average of @xmath252 and @xmath257 with @xmath258 .
it was calculated with subroutine spctrm from ref .
@xcite , using a welch window and variance reduction as described there . our simulations with sampling every generation showed , as expected , that the psd does not have any simple shape for frequencies between the nyquist frequencies of 1/32 , corresponding to sampling every 16 generations , and 1/2 , corresponding to sampling every generation .
composition and lifetime statistics of ten qss that lasted at least 20000 generations in a particular simulation run of @xmath94 generations .
the qss are listed in order of increasing mean lifetime .
columns one through four give the genotype labels and , in parentheses , the initial populations [ the fixed - point populations given by eq .
( [ eq : fix5 ] ) ] for the @xmath33 genotypes in each qss .
columns five and six give the population - weighted mean and standard deviation of the @xmath259 hamming distances between the genotypes in the initial community , eqs .
( [ eq : avhamm ] ) and ( [ eq : devhamm ] ) , respectively . columns seven and eight give the mean and standard deviations of the @xmath260 offdiagonal interaction matrix elements @xmath19 between the genotypes in the initial community , respectively .
columns nine and ten give the mean and standard deviations of the lifetimes , obtained from 300 independent escapes for each qss .
see discussion in secs .
[ sec : stab ] and [ sec : life ] . [ cols=">,>,>,>,>,>,>,>,^,^,^,^,^,^ " , ] generations .
the model parameters , which are the same in the subsequent figures , are : mutation rate @xmath82 per individual , carrying capacity @xmath261 , fecundity @xmath64 , and genome length @xmath80 .
top curve ( black ) : shannon - wiener diversity index @xmath262 $ ] .
bottom curve ( gray , red online ) : normalized total population @xmath263 $ ] . ( a ) the whole time series , showing intervals of quasi - steady states ( qss ) separated by periods of high activity . ( b )
part of one of the periods of high activity , shown on a time scale expanded twenty times .
several shorter qss are resolved between bursts of high activity .
comparison with ( a ) suggests statistical self - similarity .
see discussion in secs .
[ sec : gen ] and [ sec : life ] .
, title="fig : " ] generations .
the model parameters , which are the same in the subsequent figures , are : mutation rate @xmath82 per individual , carrying capacity @xmath261 , fecundity @xmath64 , and genome length @xmath80 .
top curve ( black ) : shannon - wiener diversity index @xmath262 $ ] .
bottom curve ( gray , red online ) : normalized total population @xmath263 $ ] .
( a ) the whole time series , showing intervals of quasi - steady states ( qss ) separated by periods of high activity .
( b ) part of one of the periods of high activity , shown on a time scale expanded twenty times .
several shorter qss are resolved between bursts of high activity .
comparison with ( a ) suggests statistical self - similarity .
see discussion in secs .
[ sec : gen ] and [ sec : life ] .
, title="fig : " ] versus time for the same simulation run shown in fig .
[ fig : timeser ] . in order of decreasing darkness from black to very light gray
the symbols indicate @xmath146 , @xmath264 $ ] , @xmath265 $ ] , @xmath266 $ ] , and @xmath267 .
( online the colors are black , blue , red , green , and yellow in the same order . )
note that the difference between the label for two species bears no simple relation to their hamming distance .
each qss is composed of a different set of species , punctuated by periods during which the population moves rapidly through genotype space .
( a ) corresponding to fig .
[ fig : timeser](a ) , sampled every 960 generations to facilitate plotting .
( b ) corresponding to fig .
[ fig : timeser](b ) , sampled every 48 generations .
see discussion in secs .
[ sec : gen ] and [ sec : life ] .
, title="fig : " ] versus time for the same simulation run shown in fig .
[ fig : timeser ] . in order of decreasing darkness from black to very light gray
the symbols indicate @xmath146 , @xmath264 $ ] , @xmath265 $ ] , @xmath266 $ ] , and @xmath267 .
( online the colors are black , blue , red , green , and yellow in the same order . )
note that the difference between the label for two species bears no simple relation to their hamming distance .
each qss is composed of a different set of species , punctuated by periods during which the population moves rapidly through genotype space . (
a ) corresponding to fig .
[ fig : timeser](a ) , sampled every 960 generations to facilitate plotting .
( b ) corresponding to fig .
[ fig : timeser](b ) , sampled every 48 generations .
see discussion in secs .
[ sec : gen ] and [ sec : life ] .
, title="fig : " ] against each of ten specific qss ( thick , black lines ) , compared with the same quantity against randomly chosen , feasible communities [ thin , gray lines ( green on line ) ] .
the multiplication rates are calculated from eq .
( [ eq : other ] ) . ( a ) when the mutant species differ from the resident community by a hamming distance of one ( nearest neighbors @xcite ) .
( b ) when the mutant species differ from the resident community by a hamming distance of two ( next - nearest neighbors ) .
see discussion in sec .
[ sec : stab ] .
, title="fig : " ] against each of ten specific qss ( thick , black lines ) , compared with the same quantity against randomly chosen , feasible communities [ thin , gray lines ( green on line ) ] .
the multiplication rates are calculated from eq .
( [ eq : other ] ) . ( a ) when the mutant species differ from the resident community by a hamming distance of one ( nearest neighbors @xcite ) .
( b ) when the mutant species differ from the resident community by a hamming distance of two ( next - nearest neighbors ) .
see discussion in sec .
[ sec : stab ] .
, title="fig : " ] generations each .
the near - gaussian central peak corresponds to the qss , while the near - exponential wings correspond to the active periods .
also shown by a thin line is a histogram based on birth / death fluctuations in ten specific qss communities with zero mutation rate .
the latter is re - normalized so that the maxima of the two histograms coincide .
based on these histograms , a cutoff of @xmath268/16 = 0.015 $ ] was used to distinguish between active periods and qss .
( b ) normalized histograms for the length of active periods ( @xmath269 ) and qss [ solid , light gray circles ( cyan online ) and solid , dark gray squares ( red online ) ] . based on 16 independent runs of @xmath115 generations each .
two of the histograms ( @xmath247 and circles ) use a constant bin width of 16 generations . in order to capture the information for large durations , the data for the qss were also analyzed with exponentially increasing bin size ( squares ) .
error bars showing standard error based on the spread between the individual runs are shown only where they are larger than the symbol size .
the black curve through the points for the active periods is a least - squares fit of an exponential distribution to the data .
the straight line is a guide to the eye , corresponding to @xmath270 behavior for the qss data .
see sec .
[ sec : life ] for details .
, title="fig : " ] generations each .
the near - gaussian central peak corresponds to the qss , while the near - exponential wings correspond to the active periods .
also shown by a thin line is a histogram based on birth / death fluctuations in ten specific qss communities with zero mutation rate .
the latter is re - normalized so that the maxima of the two histograms coincide .
based on these histograms , a cutoff of @xmath268/16 = 0.015 $ ] was used to distinguish between active periods and qss .
( b ) normalized histograms for the length of active periods ( @xmath269 ) and qss [ solid , light gray circles ( cyan online ) and solid , dark gray squares ( red online ) ] . based on 16 independent runs of @xmath115 generations each .
two of the histograms ( @xmath247 and circles ) use a constant bin width of 16 generations . in order to capture the information for large durations , the data for the qss were also analyzed with exponentially increasing bin size ( squares ) .
error bars showing standard error based on the spread between the individual runs are shown only where they are larger than the symbol size .
the black curve through the points for the active periods is a least - squares fit of an exponential distribution to the data .
the straight line is a guide to the eye , corresponding to @xmath270 behavior for the qss data .
see sec .
[ sec : life ] for details .
, title="fig : " ] generations , sampled every 16 generations and averaged over 16 independent runs .
black curve : no variance reduction .
gray curve ( red online ) : 16-fold variance reduction .
error bars shown for the eight lowest frequencies in each curve are standard errors , based on the spread between the individual runs .
the straight line is a guide to the eye , corresponding to @xmath1 behavior .
( a ) shannon - wiener diversity index @xmath262 $ ] .
( b ) normalized total population @xmath263 $ ] .
see discussion in sec .
[ sec : psd ] .
, title="fig : " ] generations , sampled every 16 generations and averaged over 16 independent runs .
black curve : no variance reduction .
gray curve ( red online ) : 16-fold variance reduction .
error bars shown for the eight lowest frequencies in each curve are standard errors , based on the spread between the individual runs .
the straight line is a guide to the eye , corresponding to @xmath1 behavior .
( a ) shannon - wiener diversity index @xmath262 $ ] .
( b ) normalized total population @xmath263 $ ] .
see discussion in sec .
[ sec : psd ] .
, title="fig : " ] generations and over 16 ( circles ) or seven ( triangles ) independent runs of @xmath115 generations each . shown
are , from above to below , the total number of species @xmath25 , the number @xmath130 of species with @xmath131 , the shannon - wiener index @xmath96 , the average hamming distance @xmath104 between genotypes in the community , the normalized total population @xmath263 $ ] , and the standard deviation @xmath105 of the hamming distances .
these data indicate the absence of any systematic long - time trends in the dynamical behavior .
see discussion in sec .
[ sec : stat ] . ] generations .
a new community is counted whenever the overlap function falls below @xmath228 .
the upper pair of curves corresponds to @xmath273 , and the lower pair to 0.90 .
the top curve in each pair ( thin , gray curve , red online ) counts both prototype communities and revisits , while the bottom curve ( heavy , black curve ) excludes revisits .
( b ) for @xmath273 only , the abscissa @xmath176 gives the starting time of each revisit , while the ordinate @xmath177 gives the starting time of the associated prototype community .
inset : detail for @xmath182 over @xmath94 generations .
( c ) lower curve : cumulative histogram for @xmath180 from part ( b ) .
about 60 % of the revisits are to other recently visited communities , while the rest are approximately uniformly distributed over all previously visited communities .
upper curve : corresponding result for @xmath240 .
see discussion in sec .
[ sec : stat ] .
, title="fig : " ] generations .
a new community is counted whenever the overlap function falls below @xmath228 .
the upper pair of curves corresponds to @xmath273 , and the lower pair to 0.90 .
the top curve in each pair ( thin , gray curve , red online ) counts both prototype communities and revisits , while the bottom curve ( heavy , black curve ) excludes revisits .
( b ) for @xmath273 only , the abscissa @xmath176 gives the starting time of each revisit , while the ordinate @xmath177 gives the starting time of the associated prototype community .
inset : detail for @xmath182 over @xmath94 generations .
( c ) lower curve : cumulative histogram for @xmath180 from part ( b ) .
about 60 % of the revisits are to other recently visited communities , while the rest are approximately uniformly distributed over all previously visited communities .
upper curve : corresponding result for @xmath240 .
see discussion in sec .
[ sec : stat ] .
, title="fig : " ] + generations .
a new community is counted whenever the overlap function falls below @xmath228 .
the upper pair of curves corresponds to @xmath273 , and the lower pair to 0.90 .
the top curve in each pair ( thin , gray curve , red online ) counts both prototype communities and revisits , while the bottom curve ( heavy , black curve ) excludes revisits .
( b ) for @xmath273 only , the abscissa @xmath176 gives the starting time of each revisit , while the ordinate @xmath177 gives the starting time of the associated prototype community .
inset : detail for @xmath182 over @xmath94 generations .
( c ) lower curve : cumulative histogram for @xmath180 from part ( b ) .
about 60 % of the revisits are to other recently visited communities , while the rest are approximately uniformly distributed over all previously visited communities .
upper curve : corresponding result for @xmath240 .
see discussion in sec .
[ sec : stat ] .
, title="fig : " ] | we present a study by linear stability analysis and large - scale monte carlo simulations of a simple model of biological coevolution .
selection is provided through a reproduction probability that contains quenched , random interspecies interactions , while genetic variation is provided through a low mutation rate . both selection and mutation act on individual organisms .
consistent with some current theories of macroevolutionary dynamics , the model displays intermittent , statistically self - similar behavior with punctuated equilibria . the probability density for the lifetimes of ecological communities
is well approximated by a power law with exponent near @xmath0 , and the corresponding power spectral densities show @xmath1 noise ( flicker noise ) over several decades .
the long - lived communities ( quasi - steady states ) consist of a relatively small number of mutualistically interacting species , and they are surrounded by a `` protection zone '' of closely related genotypes that have a very low probability of invading the resident community . the extent of the protection zone affects the stability of the community in a way analogous to the height of the free - energy barrier surrounding a metastable state in a physical system .
measures of biological diversity are on average stationary with no discernible trends , even over our very long simulation runs of approximately @xmath2 generations . |
ferromagnetic systems have attracted much interest for a long history because of the intriguing physics and applications.@xcite as the development of information technology , the research on magnetization dynamics in micromagnets has become an active field.@xcite great efforts have been devoted to this field by aiming to manipulate magnetization more efficiently.@xcite for theoretical simulation , the magnetization dynamics is usually described by the landau - lifshitz - gilbert ( llg ) equation,@xcite @xmath3 where @xmath4 represents the direction of the magnetization . @xmath5 in the first term on the right - hand side of eq.([eq1 ] ) is the effective magnetic field which drives the magnetization procession and determines the domain structure in the equilibrium states . up to date
, different sources of effective magnetic field have been identified , e.g. , the external magnetic field , the crystal anisotropy induced by the spin - orbit coupling , the exchange energy due to the spatial inhomogeneity , and also the demagnetization field from dipole - dipole interaction.@xcite the second term , the gilbert damping torque , describes the magnetization relaxation to the effective field axis on the time scale of @xmath6.@xcite the last one with first - order gradient of the magnetization is the spin torque induced by the transport of the spin polarized itinerant electrons , where @xmath7 is proportional to the spin current density.@xcite the @xmath2-term,@xcite first proposed by zhang and li,@xcite was demonstrated to be critically important to the current - driven domain wall motion , which overwhelms the threshold current due to pinning force and transverse anisotropy for domain wall motion@xcite and results in the steady domain wall velocity @xmath8 in the absence of the external magnetic field .
therefore , the determination of the effective magnetic field and the parameters in the llg equation , such as @xmath9 and @xmath2 , is an important issue for magnetization dynamics study.@xcite previously , we have derived the gilbert damping ( @xmath9 ) term based on the kinetic spin bloch equation ( ksbe ) approach@xcite in homogeneous ferromagnetic systems.@xcite in the present work , we derive the whole llg equation from the @xmath0-@xmath1 model in inhomogeneous ferromagnetic systems based on the same approach .
from the first - order gradient of the magnetization inhomogeneity , we obtain the current - induced torques which are consistent to the above llg equation . within the second - order gradient
, we find that the llg equation should be modified and written as @xmath10 the second - order gradient introduces two contributions to the effective magnetic field . the one given by @xmath11
is identified as the spin stiffness field discussed in previous works.@xcite the other term , in the form of @xmath12 , has never be referred in the literature . in the present paper ,
we call it `` vertical spin stiffness '' in the sense of the fact that this new field is vertical to the plane defined by the magnetization and the normal spin stiffness @xmath13 .
interestingly , this vertical spin stiffness can not be written in terms of the free energy , and therefore , it can not be derived from the functional derivative of the free energy with respect to the local magnetization @xmath14/{{\bf m}_d}$ ] previously.@xcite we find that this vertical spin stiffness results in the tilt of the magnetization . the new term can significantly change the domain - wall structure in ferromagnetic semiconductors .
since the magnitude of this field is proportional to the factor of @xmath2 , the proposed effect is expected to be important in ferromagnetic semiconductors where @xmath2 is large due to the strong spin - orbit interaction.@xcite this paper is organized as follows : in sec.ii , we setup our model and derive the ksbes for the itinerant electrons in the inhomogeneous ferromagnetic system .
we calculate the spin torque by solving the ksbes in sec.iii and discuss the results in sec.iv .
finally , we briefly summarize in sec.v .
we use the exchange interaction hamiltonian density @xmath15 with @xmath16 denoting the coupling constant . following ref .
, we here assume that the ferromagnetic interaction exists among @xmath17 and show later that this assumption is justified after integrating out the conduction electrons ( giving rise to spin stiffness ) . the pauli matrices @xmath18 are used to describe the itinerant electrons .
in contrast to the previous work on the gilbert damping,@xcite we introduce the inhomogeneity by considering the position dependence of the magnetization direction @xmath19 .
@xmath20 is the uniform saturate magnetization . for the strong exchange coupling in ferromagnetic system ,
the rotation framework@xcite is employed here . that is , the local spinor operators of the itinerant electrons are defined as @xmath21 , with @xmath22 ( @xmath23 ) labeling the spin orientation parallel ( anti - parallel ) to @xmath17 . therefore , one has @xmath24 .
the spinor operators @xmath25 are connected to the ones defined in the lattice coordinate system @xmath26 via the unitary transformation @xmath27 .
the transformation matrices is given by @xmath28 with @xmath29 for @xmath30.@xcite in the rotation framework , the kinetic hamiltonian should be expressed as @xmath31 , where the gauge field introduced by the coordinate transformation is given by @xmath32 .
then , one obtains the hamiltonian density associated with the gauge field,@xcite @xmath33+a^\dag \frac{a^2}{2 m } a. \label{eq3}\ ] ] to derive the ksbes of the itinerant electrons , we follow the nonequilibrium green function approach.@xcite the dyson equation of the contour - ordered green function can be expressed as @xmath34 where the single particle contour - ordered green function between two space - time points @xmath35 and @xmath36 on the contour @xmath37 is defined as @xmath38\rangle$].@xcite @xmath39 describes the local electric potential energy , whereas @xmath40 represents the self - energy correction due to interactions , such as the electron - impurity , electron - phonon , and electron - electron interactions .
@xmath41 stands for the free - particle green function . the left- and right - inverses of @xmath42
are given by @xmath43 with @xmath44 . by multiplying them to eq.([eq4 ] ) , one obtains @xmath45g(1,2)\\ \nonumber & & \mbox{}-g(1,2)[h_0(-\stackrel{\leftarrow}{\bf p}_2,{\bf r}_2)+u_e(2)]\\ & & \mbox{}+\int_c d3 [ \sigma(1,3)g(3,2)-g(1,3)\sigma(3,2 ) ] , \label{eq7}\end{aligned}\ ] ] with @xmath46 corresponding to the momentum operators @xmath47 . to transform the above equation to the center -
of - mass and relative variables @xmath48 one rewrites the following green functions as @xmath49 the self - energy can be written in the same way .
similarly , one obtains @xmath50 the hamiltonian can be written as @xmath51 where @xmath52 and @xmath53 represent the momentum operators respect to the center - of - mass and relative variables . according to eq.([eq3 ] ) , the left- and right - operators of the kinetic hamiltonian should be different and can be written as @xmath54,\\ { h}_{\rm k}(-\stackrel{\leftarrow}{\bf p},{\bf r})&=&\tfrac{1}{2m}[\stackrel{\leftarrow}{\bf p}^2+i(\nabla \cdot { \bf a})-2\stackrel{\leftarrow}{\bf p}\cdot { \bf a}+{\bf a}^2 ] , \label{eq22216 - 17}\end{aligned}\ ] ] separately . by assuming the magnetization varies smoothly respect to the spatial coordinates , we only keep the spatial gradient up to the second order .
since the gauge field is already the first - order gradient , both @xmath55 and @xmath56 are second - order gradient terms .
we include both of them and neglect the higher - order ones .
the gradient term of the gauge field can be written in the center - of - mass coordinate system as @xmath57 similar calculation gives @xmath58 moreover , one can easily show that @xmath59 with @xmath60 representing the unit matrix . by substituting all these equations into eq.([eq7 ] ) and doing fourier transformation respect to the relative coordinate @xmath61 ,
one obtains @xmath62 .
\label{eq20}\end{aligned}\ ] ] the details can be found in appendixa .
we then perform the gradient expansion up to the first order and obtain @xmath63-\tfrac{i}{2}\{\partial_{\bf k}h_0 , \partial_{\bf r } g\ } + \tfrac{i}{2}\{\partial_{\bf r}\partial_{\bf k}h_0,g\}\\ & & \hspace{-0.2cm}\mbox{}+\tfrac{i}{2}\{\partial_{\bf r}(h_0+u_e),\partial_{\bf k}g\ } + \int_c dt_3(\sigma g - g\sigma ) , \label{eq21}\end{aligned}\ ] ] where all the quantities are defined at @xmath64 and @xmath65 .
we should point out that the commutator notation @xmath66 $ ] is still used although the left- and right - operators of @xmath67 are in different expressions [ see eqs.(16 ) and ( 17 ) ] . by taking the isochronous condition , i.e. , @xmath68 , one has@xcite @xmath69^<\\ \nonumber & & = \int_{-\infty}^td\tau [ \sigma^>(t,\tau)g^<(\tau , t ) -\sigma^<(t,\tau)g^>(\tau , t)\\ & & \hspace{0.2cm}\mbox{}-g^>(t,\tau)\sigma^<(\tau , t ) + g^<(t,\tau)\sigma^>(\tau , t ) ] , \label{eq22}\end{aligned}\ ] ] where the lesser and greater green functions are defined by @xmath70 and @xmath71 .
therefore , the correlation function @xmath72 can be written as @xmath73-\tfrac{i}{2}\{\partial_{\bf k}h_0 , \partial_{\bf r } g^<\}. \label{eq23}\end{aligned}\ ] ] within the generalized kadanoff - baym ansatz,@xcite we have @xmath74 where @xmath75 is the local density matrix of the itinerant electrons with momentum @xmath76 located at @xmath64 .
therefore , we write the general form of the ksbes of the itinerant electrons as @xmath77+\tfrac{1}{2}\ { \nabla_{\bf k } h_0 , \nabla_{\bf r}\rho_{\bf k}\}-\nabla_{\bf r}u_e\cdot \nabla_{\bf k}\rho_{\bf k}\\ \nonumber & & \mbox{}-\tfrac{1}{2}\{\nabla_{\bf r}h_0,\nabla_{\bf k}\rho_{\bf k}\}-\tfrac{1}{2}\{\nabla_{\bf r}\cdot\nabla_{\bf k}h_0 , \rho_{\bf k}\}\\ & & = \partial_t\rho_{\bf k}|^c_{\rm scat}+\partial_t\rho_{\bf k}|^f_{\rm scat}. \label{eq24}\end{aligned}\ ] ] on the right - hand side of above equations , @xmath78 and @xmath79 from the integral term in eq.([eq23 ] ) represent the spin - conserving and spin - flip scatterings .
the details of these terms can be found in ref .. we specify the hamiltonian @xmath80/(2m)$ ] with the upper ( lower ) sign representing the left ( right ) operator case .
the electric potential energy is given by @xmath81 ( @xmath82 ) .
the final form of the ksbes is given by @xmath83+i\tfrac{k_i}{m}[a_i,\rho_{\bf k}]+\tfrac{1}{2m}\{{\bf a},\nabla_{\bf
r}\rho_{\bf k}\}\\ \nonumber & & \mbox{}+\tfrac{1}{m}{\bf k}\cdot \nabla_{\bf r}\rho_{\bf k}-e{\bf e}\cdot \nabla_{\bf k}\rho_{\bf k } -\tfrac{k_i}{2m}\{\nabla_{\bf k}\rho_{\bf k},\nabla_{\bf r}a_i\}\\ & & = \partial_t\rho_{\bf k}|^c_{\rm scat}+\partial_t\rho_{\bf k}|^f_{\rm scat}. \label{eq25}\end{aligned}\ ] ] interestingly , we find that the contribution from @xmath84 is completely canceled by the gauge field gradient term from @xmath85 $ ] and the @xmath86-term is irrelevant .
in general cases , the ksbes are too complicated to solve analytically and the numerical scheme should be employed . however , the analytical solution can still be expected within some simplification of the ksbes . in the following ,
we assume : ( _ i _ ) the spatial dependence is weak in the rotation coordinate systems , hence @xmath87 ; ( _ ii _ ) the scattering is strong enough to set up the steady - state condition @xmath88 . without loss of generality , we take the magnetization gradient along arbitrary direction . the external electric field is applied on the purpose of producing current - induced magnetization dynamics .
therefore , the ksbes become @xmath89-\tfrac{k_i}{2m}\{\nabla_{\bf k } \rho_{\bf k},\nabla_{\bf r}a_i\ } \\ & & = \partial_t\rho_{\bf k}|^c_{\rm scat}+\partial_t\rho_{\bf k}|^f_{\rm scat}. \label{eq26}\end{aligned}\ ] ] for the steady - state situation with a small current due to a static electric field , we assume that the distribution of the itinerant electrons is not far away from the fermi distribution .
the scattering effect of the spin - conserving process is introduced by the relaxation time approximation with the average momentum relaxation time @xmath90 .
therefore , one can linearly expand the density matrices by considering the drift effect , @xmath91 , where the isotropic density matrices @xmath92 representing the spin polarized fermi distribution in the absence of the electric field . by substituting these density matrices into eq.([eq26 ] ) , the driving term and the spin - conserving scattering term cancel out .
further , one introduces the average spin relaxation time @xmath93 and rewrite the spin - flip scattering as @xmath94 .
here , @xmath95 describes the equilibrium spin polarization due to the spin - splitted band structure in the ferromagnetic system . under the above procedures , one finally obtains the equations of the steady - state spin polarization @xmath96 by using the relation @xmath97 . here
, we denotes @xmath98 .
then , the equation of the total spin polarization @xmath99 can be obtained by summing eq.([eq27 ] ) over the @xmath65-space @xmath100 for the lowest order approximation , one substitutes @xmath101 into eq.([eq28 ] ) . by considering @xmath102 , one obtains @xmath103 in which the relation @xmath104 is used .
the quantity @xmath105 stands for the density of the itinerant electrons .
the spin polarization can then be found @xmath106 where @xmath107 and @xmath108 .
the equilibrium spin polarization @xmath109 with the value of the spin polarizability along the magnetization direction @xmath110 is negative , because @xmath111 is anti - parallel to @xmath112 and @xmath113 . with the current defined as @xmath114 , the transverse spin polarization is given by @xmath115 here , we have neglected the gauge field in the denominator . we should point out that the subscript @xmath116 refers to the lattice coordinate axis .
in contrast , the gauge field is defined in the rotation frame , hence one needs to transform the relevant terms back to the lattice coordinate system .
the rotation transformations are given by @xmath117 and @xmath118.@xcite one then obtains the transverse spin polarization in the lattice coordinate @xmath119 .
\label{eq32}\end{aligned}\ ] ] since the @xmath0-@xmath1 exchange interaction can be equivalently written as @xmath120 , the spin procession field induced by the @xmath0-@xmath1 exchange interaction @xmath121 is @xmath122 , \label{eq33}\end{aligned}\ ] ] with @xmath123 .
we finally obtain the llg equation with gilbert damping torque as eq.([eq2 ] ) , where @xmath124}$ ] and @xmath125 .
\label{eq34}\ ] ] here , the effective magnetic field @xmath126 includes the isotropic and demagnetization sources@xcite as well as the external magnetic field .
we discuss our results based on eq.([eq2 ] ) .
one notices that the third term on the right - hand side of the equation is the current - induced torque obtained in the previous works.@xcite the fourth term is independent of the current but associated to the second - order gradient of the magnetization .
this term contains two contributions .
the one in the form @xmath127 is identified as the effective spin stiffness.@xcite in the limit @xmath128 or @xmath129 , from eq .
( [ eq34 ] ) one has @xmath130 , which is consistent with the previous result.@xcite however , @xmath131 should be modified for finite @xmath2 .
this stiffness is widely used in the study on the domain wall@xcite and demonstrated to be critical to determine the width of the domain wall.@xcite one finds that the spin stiffness increases with increasing the exchange coupling strength @xmath16 , which agrees with the previous computation.@xcite by taking @xmath132 @xmath133 , @xmath134 ( ref . ) , and @xmath135 with @xmath136 representing the free electron mass , one estimates the spin stiffness @xmath137 pj / m in gamnas.@xcite however , the other effective field in the form @xmath138 has not been studied yet .
it is obvious that this torque prevents the magnetization from varying in a plane and induces transverse component instead .
one notices that this term has no contribution to the free energy , because it is always perpendicular to the magnetization .
therefore , it can not be derived from the variation of the free energy with respect to the magnetization , which can explain the reason for missing this term in previous works . in the following ,
we focus on the steady domain - wall solution of the llg equation in the absence of the current to illustrate the effect of this new effective magnetic field due to the vertical spin stiffness . in the ferromagnetic thin film or nanowire structures
, one takes the total effective magnetic field , @xmath139,@xcite with @xmath140 and @xmath141 representing the anisotropy constant and demagnetization field , respectively . here
, we have added stiffness constant @xmath142 arising from the non - itinerant - electron origin ( such as dipole - dipole interaction between the localized @xmath1 electrons ) to describe general systems , @xmath143 and @xmath144 .
the magnetization direction is given by @xmath145 .
obviously , the ground state is the homogeneous configuration with the magnetization pointing to the easy axis , i.e. , the @xmath146-axis .
however , the inhomogeneous configurations can also stably exist , for example the domain wall structure . to discuss the formation of the inhomogeneous magnetization structure ,
we first write the equation of motion for @xmath147 and @xmath148 in the absence of the current , @xmath149,\\ \nonumber -\alpha\dot\theta+{\sin}\theta\dot\varphi&=&\gamma k_\perp { \sin}\theta\cos\theta\sin^2\varphi+\gamma{k\sin\theta\cos\theta}\\ \nonumber & & \mbox{}-\gamma a_{\rm eff}[\partial_x^2\theta-\sin\theta \cos\theta(\partial_x\varphi)^2]\\ & & \mbox{}-\tfrac{\gamma { \beta^\prime } a_{\rm eff}}{\sin\theta}\partial_x(\sin^2\theta\partial_x\varphi ) .
\label{eq35 - 36}\end{aligned}\ ] ] for the steady state , one obtains @xmath150/{(1+{\beta^\prime}^2)},\\ \nonumber \partial_x^2\varphi&= & \big[\tfrac{{\beta^\prime } k}{a_{\rm eff}}{\cos}\theta + \tfrac{k_\perp}{a_{\rm eff}}{\sin}\varphi ( { \beta^\prime}{\cos}\theta{\sin}\varphi+{\cos}\varphi)\\ & & \mbox { } -(1+{\beta^\prime}^2)2{\cot}\theta\partial_x\theta\partial_x\varphi \big]/{(1+{\beta^\prime}^2)}. \label{eq37 - 38}\end{aligned}\ ] ] at @xmath151 , one obtains a single wall solution in the @xmath152-@xmath153 plane ( located at @xmath154 ) , @xmath155 , @xmath156 with @xmath157 .
when @xmath158 , the magnetization can not vary in a fixed plane since @xmath147 and @xmath148 are coupled . unfortunately , eqs.(36 ) and
( 37 ) can not be solved analytically in general . however , in the absence of the demagnetization field ( @xmath159 ) , there is a solution where the gradient of @xmath148 is a constant , i.e. , @xmath160 . in this case
, the equations can be written as @xmath161,\\ 0&=&\tfrac{1}{{(1+{\beta^\prime}^2 ) } } \big[\tfrac{{\beta^\prime } k}{a_{\rm eff}}{\cos}\theta -\lambda(1+{\beta^\prime}^2)2{\cot}\theta\partial_x\theta \big ] .
\label{eq39 - 40}\end{aligned}\ ] ] obviously , both equations give the solution in the same form @xmath162 which is just the domain wall solution with the corresponding width @xmath163}$ ] and @xmath164 , respectively .
the self - consistent condition @xmath165 determines the value of @xmath166 as @xmath167^{\tfrac{1}{2}}.\end{aligned}\ ] ] the domain wall thickness is enhanced by @xmath168 as @xmath169 -axis is plotted as : ( a ) @xmath170 , @xmath171 ( or @xmath172 ) ; ( b ) @xmath173 , @xmath171 ; and ( c ) @xmath173 and @xmath174 . by denoting the magnetization direction @xmath145 , the position dependences of @xmath147 and @xmath148
are given in ( d)-(f ) .
@xmath157.,title="fig:",width=283 ] -axis is plotted as : ( a ) @xmath170 , @xmath171 ( or @xmath172 ) ; ( b ) @xmath173 , @xmath171 ; and ( c ) @xmath173 and @xmath174 . by denoting the magnetization direction @xmath145 , the position dependences of @xmath147 and @xmath148
are given in ( d)-(f ) .
@xmath157.,title="fig:",width=283 ] -axis is plotted as : ( a ) @xmath170 , @xmath171 ( or @xmath172 ) ; ( b ) @xmath173 , @xmath171 ; and ( c ) @xmath173 and @xmath174 . by denoting the magnetization direction @xmath145 , the position dependences of @xmath147 and @xmath148
are given in ( d)-(f ) .
@xmath157.,title="fig:",width=321,height=113 ] -axis is plotted as : ( a ) @xmath170 , @xmath171 ( or @xmath172 ) ; ( b ) @xmath173 , @xmath171 ; and ( c ) @xmath173 and @xmath174 . by denoting the magnetization direction @xmath145
, the position dependences of @xmath147 and @xmath148 are given in ( d)-(f ) .
@xmath157.,title="fig:",width=321 ] this solution indicates that the magnetization rotates along the easy axis inside the domain wall .
one can calculate the change of @xmath148 across the domain wall to obtain @xmath175 . when @xmath176 , one finds that @xmath177 , which indicates that @xmath148 can be approximated as a constant within a domain wall as shown in fig.[fig1](a ) , where the magetization always lies in the @xmath178-@xmath179 plane [ see also fig.[fig1 ] ( d ) ] .
the solution , therefore , returns to the nel wall case.@xcite in contrast , when @xmath180 , @xmath148 oscillates within a wall and the wall structure becomes more like a one - dimensional vortex with strong correlation between @xmath147 and @xmath148 [ see fig.[fig1](b ) and ( e ) ] .
in ferromagnetic metals , @xmath2 ( @xmath181-@xmath182 ) is small@xcite and @xmath183 is large due to the strong dipole - dipole interaction , hence the vertical spin stiffness is unimportant .
however , in ferromagnetic semiconductors , e.g. , gamnas , @xmath142 would be small@xcite and @xmath2 can be large , @xmath184,@xcite therefore , the domain wall structure can behave like a vortex .
when @xmath185 , the behavior of @xmath148 in a steady solution is determined by the competition between the hard - axis anisotropy , @xmath186 , and the transverse stiffness is proportional to @xmath168 .
then @xmath148 tends to be locked to @xmath187 ( @xmath188 is integer ) and deviates from being a linear function of position , forming soliton - like structure or a staircase behavior [ fig.[fig1](c ) and ( f ) ] . in a domain wall ,
modulation of @xmath147 occurs when @xmath189 is close to @xmath190 , but @xmath147 tends to be close to @xmath191 ( @xmath192 is integer ) when @xmath148 changes in order to lower the energy cost due to the hard axis anisotropy .
thus a domain wall has an oscillating structure as seen in fig.[fig1](c ) and ( f ) .
the number of oscillation increases as @xmath168 is enhanced . as far as we find numerically
, the number of oscillation also depends on the initial condition of @xmath193 and @xmath194 at the boundary of the wall .
finally , we should point out that the domain - wall solution for @xmath195 is the same as that for @xmath171 in the absence of the vertical spin stiffness [ see fig.[fig1](a ) and ( d)].@xcite as is well - known,@xcite the dynamics is strongly affected by the structure .
for instance , vortex walls are easier to move than planar domain walls in the current - driven case , because of the perpendicular component near the vortex core .
we may therefore expect that the wall for finite @xmath180 , e.g. , in gamnas,@xcite would have even lower threshold current due to the structure change arising from the transverse exchange torque .
inclusion of the transverse stiffness in the micromagnetic simulations is thus crucially important in systems with strong spin - orbit interaction .
in summary , we have derived the ksbes in ferromagnetic systems based on the @xmath0-@xmath1 model in the presence of the inhomogeneity of the magnetization .
we analytically solved the ksbes and derived the spin torque due to the spin polarization of the itinerant electrons .
the current - induced spin torque from the first - order magnetization gradient is consistent with the previous works .
we found that the second - order gradient of the magnetization inhomogeneity gives rise to an effective magnetic field that is perpendicular to the spin stiffness field .
this vertical spin stiffness is proportional to the nonadiabatic parameter @xmath2 .
we showed that the new term modifies the domain wall structure and causes magnetization rotation along the easy axis .
the vertical spin stiffness is expected to be crucially important in ferromagnetic semiconductors , and needs to be included in the llg equation in numerical simulations on the magnetization dynamics .
this work was supported by the natural science foundation of china under grant no .
one of the authors ( g.t . )
thanks support by a grant - in - aid for scientific research in priority areas , `` creation and control of spin current '' ( grant no .
1948027 ) , the kurata memorial hitachi science and technology foundation , and the sumitomo foundation .
the fourier transformations of the functions in eq.([eq7 ] ) respect to the relative coordinate are calculated as follows . for simplification ,
we omit all the temporal coordinates without leading to any ambiguity . the green function at momentum @xmath65 is defined as @xmath196
then , the electric potential energy term can be written as @xmath197 terms with hamiltonian read @xmath198 and @xmath199\\ \nonumber & & = \int d{\bf r } e^{-i{\bf k}\cdot { \bf r } } e^{-\frac{\bf r}{2}\partial_{\bf r}^{h_0}}[g({\bf r},{\bf r})h_0(-\tfrac{1}{2}\stackrel{\leftarrow}{\bf p}_{\bf r}+\stackrel{\leftarrow}{\bf p},{\bf r})]\\ \nonumber & & = e^{-i\frac{1}{2}\partial_{\bf k}\partial_{\bf r}^{h_0 } } \int d{\bf r } e^{-i\bf k\cdot r}[g({\bf r},{\bf r})h_0(-\tfrac{1}{2}\stackrel{\leftarrow}{\bf p}_{\bf r}+\stackrel{\leftarrow}{\bf p},{\bf r})]\\ \nonumber & & = e^{-i\frac{1}{2}\partial_{\bf k}\partial_{\bf r}^{h_0 } } g({\bf r},{\bf k})h_0(-\tfrac{1}{2}\stackrel{\leftarrow}{\bf p}_{\bf r}+{\bf k},{\bf r})\\ & & = e^{-i\frac{1}{2}(\partial_{\bf k}\partial_{\bf r}^{h_0}-\partial_{\bf r}^g\partial_{\bf k}^{h_0})}g({\bf r},{\bf k})h_0({\bf k},{\bf r}).\end{aligned}\ ] ] for the integral terms , one has @xmath200 g({\bf r},{\bf k}^\prime)\sigma({\bf r},{\bf k}^{\prime\prime})\\ & & = e^{\tfrac{i}{2}(\partial^g_{\bf k}\partial^\sigma_{\bf r}-\partial_{{\bf k}}^\sigma\partial^g_{\bf r } ) } \sigma({\bf r},{\bf k})g({\bf r},{\bf k}).\end{aligned}\ ] ] here , the time integral @xmath201 is omitted for simplification . similarly , one can show @xmath202 in these equations , we use the notation @xmath203 . 0 c. kittel , rev .
mod . phys . *
21 * , 541 ( 1949 ) .
a. hubert and r. schfer , _ magnetic domains _ ( springer - verlag , berlin , 1998 )
. y. tserkovnyak , a. brataas , g. e. w. bauer , and b. i. halperin , rev .
mod . phys . * 77 * , 1375 ( 2005 ) .
g. tatara , h. kohno , and j. shibata , phys . rep . * 468 * , 213 ( 2008 ) .
d. a. allwood , g. xiong , c. c. faulkner , d. atkinson , d. petit , and r. p. cowburn , science * 309 * , 1688 ( 2005 ) .
l. thomas , m. hayashi , x. jiang , r. moriya , c. rettner , and s. s. p. parkin , nature ( london ) * 443 * , 197 ( 2006 ) .
s. s. p. parkin , m. hayashi , and l. thomas , science * 320 * , 190 ( 2008 ) .
t. l. gilbert , phys . rev . * 100 * , 1243 ( 1955 ) . l. d. landau , e. m. lifshitz , and l. p. pitaevski , _ statistical physics _ ,
part 2 , 3rd ed .
( pergamon , oxford , 1980 ) .
y. tserkovnyak , a. brataas , and g. e. w. bauer , j. magn .
mater . * 320 * , 1282 ( 2008 ) .
l. berger , phys .
b * 33 * , 1572 ( 1986 ) .
s. zhang and z. li , phys .
* 93 * , 127204 ( 2004 ) .
a. thiaville , y. nakatani , j. miltat , and y. suzuki , europhys . lett . * 69 * , 990 ( 2005 ) .
h. kohno , g. tatara , and j. shibata , j. phys .
. jpn . * 75 * , 113706 ( 2006 ) .
h. kohno and j. shibata , j. phys .
* 76 * , 063710 ( 2007 ) .
y. tserkovnyak , h. j. skadsem , a. brataas , and g. e. w. bauer , phys .
b * 74 * , 144405 ( 2006 ) .
z. li and s. zhang , phys .
b * 70 * , 024417 ( 2004 ) . g. tatara and h. kohno , phys . rev . lett . * 92 * , 086601 ( 2004 ) .
g. tatara , t. takayama , h. kohno , j. shibata , y. nakatani , and h. fukuyama , j. phys .
. jpn . * 75 * , 064708 ( 2006 ) .
m. w. wu , j. h. jiang , and m. q. weng , phys .
rep . * 493 * , 61 ( 2010 ) , and references therein .
c. m. jaworshi , j. yang , s. mack , d. d. awschalom , j. p. heremans , and r. c. myers , nature mater . * 9 * , 898 ( 2010 ) .
e. l. nagaev , _ physics of magnetic semiconductors _
( mir , moscow , 1983 ) .
j. knig , h. h. lin , and a. h. macdonald , phys .
rev . lett . * 84 * , 5628 ( 2000 ) .
j. knig , t. jungwirth , and a. h. macdonald , phys .
b * 64 * , 184423 ( 2001 ) .
h. haug and a .- p .
jauho , _ quantum kinetics in transport and optics of semiconductors _ ( springer - verlag , berlin , 1996 ) .
m. w. wu and h. metiu , phys .
b * 61 * , 2945 ( 2000 )
. n. l. schryer and l. r. walker , j. appl . phys . * 45 * , 5406 ( 1974 ) .
d. c. ralph and m. d. stiles , j. magn .
magn . mater . *
320 * , 1190 ( 2008 ) .
t. dietl , h. ohno , and f. matsukura , phys .
b * 63 * , 195205 ( 2001 ) .
l. cywiski and l. j. sham , phys .
b * 76 * , 045205 ( 2007 ) . | we calculate the magnetization torque due to the spin polarization of the itinerant electrons by deriving the kinetic spin bloch equations based on the @xmath0-@xmath1 model .
we find that the first - order gradient of the magnetization inhomogeneity gives rise to the current - induced torques , which are consistent to the previous works . at the second - order gradient
, we find an effective magnetic field perpendicular to the spin stiffness filed .
this field is proportional to the nonadiabatic parameter @xmath2 .
we show that this vertical spin stiffness term can significantly modify the domain - wall structure in ferromagnetic semiconductors and hence should be included in the landau - lifshitz - gilbert equation in studying the magnetization dynamics . |
it is by now relatively well known that gravitational collapse can produce rich structure even within highly simplified systems such as spherical symmetry . in particular , near the threshold of black hole formation , the strong field dynamics of general relativity exhibits critical phenomena . the pioneering work demonstrating this critical behavior in the collapse of a single massless
scalar field @xcite has been supplemented by investigations of gravitational waves @xcite , a perfect fluid @xcite and a variety of other matter models , all of which exhibit the same general characteristics .
indeed , to our knowledge no system which has been studied in this context has been shown _ not _ to exhibit this critical phenomena . at this point in the subject s development ,
dynamical evolutions ( solutions of the full partial differential equations of motion ) , in tandem with analytic and perturbative calculations have given us a reasonable understanding of many of the phenomenological details of critical behavior in collapse .
( see @xcite for an excellent review of the subject . ) in light of this , one of the more interesting discoveries in some of the more recently studied models @xcite is the presence of two distinct types of behavior at the threshold of black hole formation .
specifically , in these models , certain regions of parameter space ( initial - data space ) are found to yield near - critical collapsing configurations which display self - similarity and , in the super - critical regime , a scaling law which is continuous in the black hole mass . by analogy with the theory of phase transitions ,
this is called a type ii transition .
however , it is found that other regions of parameter space lead to critical collapse which has a static ( or periodic ) solution as an intermediate attractor
this results in a black hole transition with a nonzero mass gap .
again , in analogy with the discontinuous behavior in order parameter which frequently accompanies first order phase transitions , this is called type i behavior . for the case of the einstein - yang - mills ( eym )
model studied in @xcite , the static solution appearing in type i collapse is the well - known @xmath0 bartnik - mckinnon solution@xcite , while for the massive scalar model considered in @xcite , the type i threshold solutions are apparently unstable members of the family of `` oscillating soliton stars '' which have previously been constructed by seidel and suen @xcite , albeit in a different context . finally , in the case of einstein - skyrme ( es ) collapse considered in @xcite ,
a static type i solution was observed , which , as in the eym case , had previously been constructed and studied within a purely static _ ansatz _ @xcite .
as noted above , each of these three models also exhibits type ii behavior in the case of massive scalar collapse @xcite , the type ii critical solution is the same one originally observed in massless scalar collapse @xcite , and , interestingly , in the es model , the type ii solution is evidently identical @xcite to that observed in the eym model @xcite .
heuristically , one expects type ii behavior in _ any _ collapse model where the initial configuration can be made sufficiently `` kinetic - energy - dominated '' ( ultra - relativistic ) , whereas type i behavior is expected only in those models which have an intrinsic length scale ( or equivalently , mass scale ) , and which have some type of self - interaction which can `` balance '' the attractive gravitational interaction . as noted in the concluding remarks of @xcite , hints of further interesting phenomenology in the eym model
have been seen in the super - critical regime where all evolutions are characterized by black hole formation . in this paper
we study this regime in more detail and present evidence for a new type of critical transition in which the intermediate attractors are the `` minimally unstable '' ( one unstable mode in perturbation theory ) colored black holes discovered by bizon @xcite , and independently by volkov and galtsov @xcite .
this result is , of course , analogous to the discovery that the @xmath0 bartnik - mckinnon solution is the intermediate attractor for type i collapse .
crucially , in order to accurately model the dynamics of super - critical solutions for long times after the formation of an event horizon , we use black - hole excision techniques .
such methods were first successfully employed in a dynamical context by seidel and suen @xcite , and have subsequently been studied and implemented by many other authors ( see @xcite and references therein for a more extensive discussion ) . however , to our knowledge , this is the first time that excision has been used to study critical collapse , and we feel our results highlight the power and potential of the strategy to elucidate issues relating to the formation and long - time evolution of black holes . our adoption of excising techniques necessitates the use of a different coordinate system than that used in @xcite that work used polar slicing and areal spatial coordinates , a system which generalizes the usual schwarzschild coordinates to time - dependent spacetimes . as is well - known , the @xmath1constant slices in the polar / areal system can not penetrate apparent horizons thus , for all practical purposes , the slices remain outside of event horizons and therefore can not be used in conjunction with excision .
we therefore retain areal spatial coordinates , but adopt maximal slicing in this case the slices _ do _ cross apparent and event horizons , and excision techniques _ can _ be used .
the outline of the remainder of the paper is as follows .
the next section describes the eym model and the equations of motion we subsequently solve numerically .
we pay particular attention to gauge and coordinate choices , as well as to regularity , boundary and initial conditions for the fields . in section [ sec : excision ] we describe our numerical scheme , focusing on our specific coordinate choices and on some details of our black - hole excising technique . in sections [ sec : results ] and
[ sec : discussion ] we describe our results and conclusions .
we are interested in investigating the gravitational collapse of a self - gravitating yang - mills gauge field .
to begin , let us consider the action for an eym theory s = d^4x where @xmath2 is the yang - mills field strength tensor . on varying the action with respect to the metric , @xmath3 , and the gauge connection , @xmath4 , we get the general equations of motion .
we simplify these further by making some additional assumptions .
in particular , we choose the gauge group to be @xmath5 and focus on spherically symmetric gravitational collapse .
this places restrictions on both the spacetime metric and the form of the gauge connection .
even so , the equations we derive have a rather general form as can be seen in equations ( [ eom_first][eom_last ] )
. subsets of these equations have been considered in a variety of different contexts .
in particular @xcite evolved a version of the einstein - yang - mills equations in polar ( @xmath6 ) , areal ( @xmath7 ) coordinates with the additional assumption on the yang - mills field that the connection @xmath8 was purely magnetic .
since our interest here is to consider the same model studied in @xcite , but to penetrate into the super - critical regime of the phase space , we will likewise make the magnetic _ ansatz_. " as described in the appendix , together with appropriate gauge choices , this _ ansatz _ effectively sets all but one of the components of the gauge connection to zero .
in addition to making these gauge choices we must also choose a coordinate system . as mentioned in the introduction , in order to evolve the system for long times to the future of black hole formation
, we choose maximal time slices and areal ( or radial ) spatial coordinates .
as detailed in the appendix , the equations can now be written in the following form .
the evolution equations are [ eq : evpi ] & = &
+ ar^2 w ( 1-w^2 ) + & = & + [ eq : evw ] & = & a + w and the constraint equations are w & = & + [ eq : alpha ] & = & ( aa - 2r ) + 2r^2 ( a^2 - 1 +
2r aa ) + 4 g ( s - 3 ) + [ eq : a ] a & = & a 1-a^2 2r + 32 ra^3 k_^^2 + 4 g r a + [ eq : ktt ] k_^ & = & -3r k_^ + 4 g ( g^2 a r^2 ) where the matter stress - energy terms are given by s - 3 & = & a^2 ( 1-w^2)^2 2 g^2 r^4 + 1 g^2 r^2 ( ^2 + ^2 ) + & = & a^2 ( 1-w^2)^2 4 g^2 r^4 + 1 2 g^2 r^2 ( ^2 + ^2 ) .
we also note that we have an algebraic relation for the sole component , @xmath9 , of the shift vector , @xmath10 : = r k_^ .
in addition to the equations of motion , we need boundary conditions on the fields .
these are determined by demanding regularity at the origin of spherical symmetry , and by enforcing an outgoing condition on the radiation fields at large radius .
this latter condition assumes that there is no radiation coming in from outside our finite mesh .
this is not completely true , as in general there will be backscattering of the propagating fields off regions of high curvature .
however , if our domain of integration is large enough , the contributions from this scattering are dynamically negligible .
the demand for a regular origin results in boundary conditions on the fields at @xmath11 . from equations ( [ eq : alpha ] ) and ( [ eq : a ] ) it can be seen that @xmath12 , as well as @xmath13 .
in addition , @xmath14 .
for the matter fields , @xmath15 and @xmath16 , so that the auxiliary variables @xmath17 and @xmath18 are both zero at the origin .
we note that these conditions constrain the yang - mills field , @xmath19 , to be in a vaccum state at @xmath11 . for initial data ,
we choose a time - symmetric kink for the gauge potential which was previously used in @xcite .
this pulse is given by [ eq : kink ] w(0,r ) & = & ( x - r s ) + ( 0,r ) & = & 0 .
with the parameters @xmath20 and @xmath21 chosen such that @xmath22 and @xmath23 .
the two parameters @xmath24 and @xmath25 define the center and width of the pulse respectively .
they also serve as the two parameters which we will vary in order to explore the phase space .
we note that the implementation of this data in @xcite was incorrect but ultimately had no effect on the overall conclusions of that work .
we have fixed the implementation of the kink data here and note a minor improvement in the conservation of energy .
it should be emphasized that we do not incorporate any form of adaptive mesh refinement into our numerical approach .
instead , we use a fixed uniform grid with a mesh spacing which is sufficiently fine to uncover the new critical behavior .
the drawback to this approach , of course , is that we are unable to fully resolve the discretely self similar solutions which arise near the type ii black hole threshold .
however , our primary interest here is in the supercritical regime , and we are satisfied that previous work has established the nature of the type ii transition .
what _ is _ important in this work is the use of black hole excision techniques which allow us to evolve well beyond the formation of the black hole .
we discretize the evolution equations ( [ eq : evpi][eq : evw ] ) using two time levels with centered time differences , and angled spatial differences as described in @xcite .
the constraint equations are then integrated outward from the origin .
equation ( [ eq : alpha ] ) , the slicing equation , is rewritten in first order form & = & ( aa - 2r ) + 2r^2 ( a^2 - 1 +
2r aa ) + 4 g ( s - 3 ) + & = & . the apparent horizon equation in maximal - areal coordinates is simply a r k^_= 1 .
thus , we can compute a characteristic function @xmath26 such that @xmath27 where @xmath28 is a threshold value , ostensibly unity , but set to a slightly larger value in practice typically , @xmath29to ensure that the inner boundary lies strictly within the apparent horizon . at any time
, the characteristic function tells us which grid points are deemed to be inside ( @xmath30 ) or outside ( @xmath31 ) the horizon . as the system evolves , we recompute @xmath26 at each time step and simply `` discard '' any grid points for which @xmath32 .
the radial locations of these grid points will then generally satisfy @xmath33 , where @xmath34 is the instantaneous inner boundary of the computational domain , and roughly coincides with the location of the apparent horizon .
we need _ no _ boundary conditions for the functions evolved with equations ( [ eq : evpi][eq : evw ] ) since the characteristic directions at @xmath35 are such that only events with @xmath36 are in the past domain of dependence .
we can therefore use the discrete form of the evolution equations up to _ and including _ the inner boundary point .
however , the constraint equations ( [ eq : alpha][eq : ktt ] ) must have an inner boundary condition . until a horizon is formed
, we can simply use conditions derived from regularity .
once a horizon forms , however , we need an alternative .
if the horizon initially forms at time @xmath37 and radial coordinate @xmath38 , then for a function @xmath39 , we use @xmath40 as the initial boundary value for @xmath41 .
we then evolve @xmath41 along this boundary using the evolution equation for @xmath41 .
the evolution equations that we use are as follows : & = & + r a k^ _ - 4 r a j_r , + & = & 3 k^ _ ( 12 k^_- r ) + 4 ( s^r_r - j_r ) + ( a^2 - 1 ) - . since the lapse @xmath42 does not have an evolution equation , we simply leave @xmath42 and @xmath43 `` frozen '' at whatever specific values they have when excision at that particular radius , @xmath34 , begins .
finally , we note that the programs which we used to generate the results that we will describe below were written in rnpl ( rapid numerical prototyping language ) @xcite .
this language has been specifically designed to facilitate the differencing and subsequent numerical solution of partial differential equations .
on evolution of the equations , we verify previously examined aspects of this system as discussed in @xcite . in particular
, we find that in different regions of the parameter space , there are both types of critical behavior at the threshold of black hole formation .
we observe type i behavior , wherein black holes turn on at finite mass and the intermediate attractor is the @xmath0 bartnik - mckinnon solution . we also confirm aspects of type ii collapse in the appropriate region of parameter space .
this includes such things as echoing and the scaling relation for the black hole mass .
however , because our code does not employ mesh - refinement , we are unable to fully reproduce the results of @xcite . however , as discussed previously , our main emphasis here is to address the outstanding question of what happens in the super - critical region and to the future of black hole formation . on evolving into the supercritical ( black hole ) regime , we find that the dynamics of the resulting spacetimes exhibit two distinct types of behavior which we describe below . it turns out that these can be characterized by the asymptotic vacuum state of the yang - mills potential @xmath19 .
these two behaviors are related to the type i and type ii critical collapses identified previously .
the original type i collapse is described , in part , by a black hole which forms with finite mass .
slightly super - critical type i evolutions , now continued well to the future of black hole formation confirm this behavior , as one would expect .
in addition , after the black hole is formed , part of the remaining yang - mills field outside the horizon radiates to infinity and some of it falls down the black hole causing the black hole to grow .
how much larger the initial black hole becomes through the infall of additional matter depends , in part , on where in the parameter space the evolution begins . eventually , however , and in accordance with what might be expected from the no - hair theorems , the exterior spacetime settles down to a schwarzschild black hole with the yang - mills field in its vacuum state . in general , for slightly super - critical collapse ( either type i or type ii ) , as the remaining mass - energy radiates away the gauge potential approaches one of its two vacuum values : @xmath44 .
for the case of slightly super - critical type i collapse , in the asymptotic regime , the gauge potential always picks the @xmath45 vacuum state ( recall that our initial data family ( [ eq : kink ] ) has @xmath46 and @xmath47 ) .
this turns out to be true even for collapsing configurations which are no longer just slightly super - critical but are , in fact , well into the black hole regime .
we can thus characterize a sort of generalized " super - critical type i collapse by the asymptotic value of the gauge potential @xmath19 .
when type ii collapse is considered , the overall picture is broadly the same .
again , after the original black hole forms ( at _ finite _ mass since we are slightly super - critical ) some of the matter remaining in the region exterior to the black hole falls into the black hole causing it to grow and the rest is radiated off to infinity .
the resulting exterior spacetime agains settle down to schwarzschild with the yang - mills field in its vacuum state .
however , the gauge potential in the slightly super - critical type ii collapse approaches its other vacuum value : @xmath48 . similarly , further into the type ii super - critical regime ( where we should emphasize the discretely self - similar nature of the critical solution is no longer clear and where the black holes which form all have finite mass ) we find
generalized " super - critical type ii collapse which can be characterized by an asymptotic value of the gauge potential of @xmath49 .
it is worth commenting here that our terminology has become a bit loose .
we have been speaking of type i and type ii collapse well into the super - critical regime when in fact these terms originally referred to types of critical behavior in a very small neighborhood of the exact critical solutions .
for instance , as one moves away from the type ii threshold , the discrete self - similarity of the exactly critical solution is no longer clearly exhibited by configurations which collapse to a black hole .
in addition , a significant fraction of the mass present will end up in the black hole .
this is to be contrasted with the tiny black holes that can be formed in near - critical type ii collapse .
it should also be made clear that type i and type ii do not refer to different _ types _ of black holes .
the end - state exterior geometry in all of these evolutions is a schwarzschild black hole .
however , we are , in a sense , generalizing the concepts of type i and type ii collapse to distinguish two kinds of collapse _ dynamics _ in the super - critical regime .
it quickly becomes clear that these regions are well defined in parameter space and it becomes natural to try and find the boundary separating them .
indeed , the problem reduces to tuning a single parameter across the threshold between these two types of collapse dynamics .
what we find is that at the threshold between these two types of dynamics a static solution emerges as the intermediate attractor for this new near - critical collapse .
the type ii side of this threshold exhibits the following near - critical evolution .
as the field collapses , a finite mass black hole is formed .
the remaining field exterior to the black hole then approaches a static , non - trivial configuration .
after some time the field then departs from the static configuration and disperses to infinity leaving the original black hole virtually unchanged .
the gauge potential in this case approaches the vacuum value of @xmath50 and thus this side of the threshold is part of the generalized type ii collapse dynamics .
the type i side of this threshold has , of course , a very similar development initially .
again , a black hole of finite mass is formed exactly as before and is followed by the approach of the exterior yang - mills field to the static solution .
however , as the field leaves the static solution , most of the matter collapses into the black hole with only a small portion radiating off to infinity .
the gauge potential , in this case , approaches the vacuum state of @xmath45 .
figures [ fig : showevolve ] and [ fig : showevolve_log ] show the development of these two cases overlayed with the static , intermediate attractor which is the exactly critical solution .
since we are in the super - critical regime , these new , exactly critical solutions are themselves black holes with yang - mills hair .
indeed , these attractors are the colored black holes discovered by bizon and independently by volkov and galtsov @xcite shortly after the discovery of the bartnik - mckinnon solutions . in general , these colored , static black hole solutions are described by their horizon radius ( or , relatedly , their mass ) @xmath51 and the number @xmath52 of zero crossings of the yang - mills potential @xmath19 .
the solutions which serve here as intermediate attractors in the dynamical collapse are the lowest lying colored black holes with @xmath0 .
subsequent to their discovery , it was shown that these black holes are unstable .
in particular , the colored black holes with label @xmath52 have @xmath53 unstable modes @xcite . of these @xmath53 unstable modes ,
@xmath52 arise from spherically symmetric perturbations in the gravitational sector away from staticity and another @xmath52 unstable modes arise from perturbations in the sphaleron sector , _
i.e. _ away from the magnetic _ ansatz _ for the yang - mills fields .
since we are solving the nonlinear evolution equations while remaining within the magnetic _ ansatz _ we would expect our data to excite all the unstable modes within the gravitational sector while exciting none of the unstable modes in the sphaleron sector . as a result
, we would expect the @xmath0 family of solutions to have a single unstable mode , thus making it relevant to critical collapse .
the picture of this new critical solution in the black hole regime having a single unstable mode thus accords perfectly with previous understanding of both type i and type ii collapse .
this new critical behavior within the black hole regime is , in many ways , qualitatively similar to that of the type i collapse .
solutions in one of these new interpolating families asymptote to one of the static , colored black hole solutions , @xmath54 .
initial data for collapsing configurations close to one of the critical solutions will approach @xmath54 , and remain in its vicinity for some amount of time @xmath55 , as measured , for example , by an observer at infinity .
the evolution will then peel off the static solution with the remaining field either dispersing to infinity or collapsing and adding a finite amount of mass to the already present black hole .
one can quantify the amount of time spent near the static solution in the same way as with the type i collapse .
the time , as measured by an asymptotic observer , is @xmath56 , where the coefficient @xmath57 is the characteristic time scale for the collapse of the unstable solution @xmath54 or , in other words , the inverse lyapounov exponent of the single unstable mode . with our evolutions , we confirm this relation and can calculate the value of @xmath57 in the manner described in @xcite . for figures [ fig : showevolve ] and [ fig : showevolve_log ] ,
where the static solution has a black hole radius of @xmath58 , we confirm the linear relation and find @xmath59 as shown in figure [ fig : time_exponent ] .
once we have established the existence of these new , colored critical solutions , it then becomes straightforward to map out parameter space for our choice of initial data .
figure [ fig : phase_space ] does this for the main region of interest .
figure [ fig : phase_space ] should be compared to figure 1 in @xcite .
there , it was suggested that there was a coexistence between black holes exhibiting both types of critical behavior .
however , that suggestion was based on evolutions which did not actually continue to the future of the resulting black hole .
thus , up to this point , only the region marked dispersion " had been well explored .
evolutions had been performed in the black hole regime , but detailed understanding of what happened in collapse was largely limited to times prior to the formation of the black hole .
now , with our use of apparent horizon boundary conditions , we can explore more completely this super - critical region .
thus , what is actually present is two well - demarcated regions in the black hole regime separated by a critical line , as described above .
= 7.5 cm = 7.5 cm
we have presented evidence for the existence of a new type of critical phenomena within the black hole regime of the spherically symmetric einstein - yang - mills model .
previous work on critical behavior in this model has established that both type i and type ii critical solutions emerge in sufficiently general families of initial data . however , that earlier work was unable to accurately evolve to the future of black hole formation . using a different coordinate system and horizon excision techniques , we are able to evolve well into the black hole regime .
we find that these two different types of critical collapse can be generalized to types of collapsing spacetimes with a distinguishing characteristic being the asymptotic vacuum value of the yang - mills potential .
these two phases " of black hole formation are separated by a critical line in phase space such that dynamically forming black hole solutions exactly at criticality are the static , @xmath0 colored black holes @xcite .
this , of course , fits nicely within our conceptual framework that critical solutions are intermediate attractors of co - dimension one .
it is worth emphasizing a few observations at this point .
first , it is important to note that the mass of the final schwarzschild black hole in the evolution will be discontinuous across this new critical line in figure [ fig : phase_space ] .
( in fact , indications of jumps in the black hole mass - spectrum in the super - critical regime were present in earlier calculations @xcite . ) naively , this might be somewhat surprising since it means that nearby configurations in the space of initial data can have , in principle , very different sized black holes as their end - states .
further , we note that when the colored black holes were originally investigated , it was found that the @xmath52th static colored black hole is parameterized by its horizon radius , @xmath51 .
in addition , as @xmath60 , the black hole solution approaches the corresponding @xmath52th bartnik - mckinnon solution . in the dynamical context discussed here for the @xmath0 solutions , one notices an interesting manifestation of this result .
the colored black hole attractors separating the type i and type ii regions in figure [ fig : phase_space ] , are likewise parameterized by their horizon radius in such a way that as the
triple point " is approached , @xmath60 , and the bartnik - mckinnon solution emerges as the relevant attractor separating type i collapse from dispersion .
finally , it is natural to ask if other collapsing systems might exhibit similar behavior .
indeed , one can simply look in the literature on static , hairy black holes and point to a number of systems that one would conjecture should have a phase diagram " similar to figure [ fig : phase_space ] .
such theories should include einstein - yang - mills coupled to a dilaton ( eymd ) and the einstein - skyrme system . however , there are other examples of systems which exhibit both type i and type ii critical behavior .
these include a massive scalar field and ( it is conjectured ) a massive , charged scalar field @xcite .
these latter systems are particularly intriguing in light of the results presented here because though they do exhibit type i and type ii critical behavior , no - hair theorems suggest that these systems do not have static black hole solutions .
it should thus be of interest to investigate the nature of the super - critical regime for these , and possibly other , systems .
this research has been supported in part by nsf grants phy93 - 18152 , phy94 - 07194 and phy97 - 22068 .
mwc would like to thank p. bizon and t. chmaj for useful communications @xcite , including descriptions of unpublished results which indicated that the parameter - space regime studied here might lead to colored black holes .
we collect here some of the details involved in deriving the equations of motion used in the paper , and to present a somewhat more general framework for future work on the einstein - yang - mills system ( eym ) .
the action for our model is s = d^4x where @xmath2 is the yang - mills field strength tensor . as usual ,
greek indices range over the four spacetime dimensions and latin indices indicate group indices .
spacetime indices are raised and lowered by the metric @xmath3 while we will not lower group indices .
thus , repeated upper group ( latin ) indices will be summed over .
the equations of motion are found by varying the action with respect to the fields .
varying with respect to the metric yields the einstein equations g _ & = & t _ + & = & 1 g^2(2 f^a _ f^a_^ - 12 g _ f^a _
f^a ) . varying with respect to the connection @xmath4 gives d _
f^a = _ f^a + ^abc a_^b f^c = 0 . with the general equations in hand we are in a position to make some simplifying assumptions .
to begin , we assume that we have an @xmath5 gauge field .
in addition we will assume spherical symmetry .
the most general parameterization of the spacetime metric is then ds^2 = ( -^2 + a^2 ^2 ) dt^2 + 2 a^2dt dr + a^2 dr^2 + b^2 r^2 d^2 where the metric coefficients @xmath61 and @xmath21 will depend on the temporal and radial coordinates @xmath62 and @xmath63 and @xmath64 is the usual metric on the unit two - sphere .
likewise , the most general parameterization of the gauge connection is now @xmath65 where the coefficients @xmath66 and @xmath67 will all depend on @xmath62 and @xmath63 .
the @xmath68 are the spherical projection of the pauli spin matrices and form an anti - hermitian basis for the group @xmath5 satisfying @xmath69 = \epsilon^{abc}\tau^{c}$ ] with @xmath70 .
we note that this connection is invariant under a gauge transformation of the form @xmath71 .
the field strength derived from this connection is f & = & ^r ( - u ) dtdr + & & + ( ^ d+ ^ d ) + & & + ( ^ d- ^ d ) + & & - ( 1 - w^2 -^2 ) ^r dd with these assumptions , the equations of motion for the yang - mills fields can now be written in first order ( in time ) form as [ eom_first ] & = & + up - v ( p + a q ) + ab^2 r^2 w ( 1-w^2-^2 ) + & = & - u + v ( + a ) + ab^2 r^2 ( 1-w^2-^2 ) y & = & - wp + & = & a(- wq ) + ( - wp ) where we have used the definitions [ eq : defpi ] & = & a + [ eq : defphi ] & = & w - v + p & = & a + [ eq : defq ] q & = & + vw + [ eq : defy ] y & = & b^2 r^2 2a ( - u ) .
we also have evolution equations for @xmath18 and @xmath72 : & = & + uq - v - 2 a b^2 r^2 y + & = & - u + v + w 2 a b^2 r^2 y , which follow from differentiation of ( [ eq : defphi ] ) and ( [ eq : defq ] ) with respect to time , and elimination of explicit time derivatives using the definitions ( [ eq : defpi]-[eq : defy ] ) . finally , the relevant einstein equations are the evolution equations for the metric components and the components of the extrinsic curvature , together with the hamiltonian and momentum constraints , as follows : & = & - a k^r_r + ( a ) + [ eq : bdot ] & = & - b k^ _ + r ( rb) ^r_r & = & k^r_r
+ k^r_r k - 1 a( a) - 2arb + 4 g + ^ _ & = & k^ _
+ k^ _ k + ( rb)^2 - 1 a(rb)^2 ( r b a(rb)) + 4 g -2 arb+ 4 k^r_r k^ _ + 2 k^_^2 = 16g[eom_last ] - ( rb) rb ( k^ _ - k^r_r)- k^_ = 4 g j_r . here , @xmath73 is the energy density , @xmath74 is the momentum density and @xmath75 , @xmath76 and @xmath77 are the stress energy components projected onto our spacelike hypersurface .
explicitly , we have = 14g^2 \ { 4 y^2 b^4 r^4 + ( 1-w^2-^2)^2 b^4 r^4 + 2 b^2 r^2 a^2 } s^r_r = 14g^2 \ { - 4 y^2 b^4 r^4 - ( 1-w^2-^2)^2 b^4 r^4 + 2 b^2 r^2 a^2 } s^_= 14g^2 \ { 4 y^2 b^4 r^4 + ( 1-w^2-^2)^2 b^4 r^4 } s^_= s^_j_r = - 1 g^2 a b^2 r^2 ( + pq ) .
as can be seen , we have chosen to write things in a fairly general form .
indeed , this is the most general eym theory we could consider in spherical symmetry .
we can thus consider more general matter configurations , together with a greater variety of coordinate systems than have been considered to date .
subsets of these equations have been investigated before in different contexts .
for example , static versions of these equations have been studied in order to understand particle - like solutions and colored black holes .
in addition , @xcite evolved a version of these equations in polar ( @xmath6 ) , areal ( @xmath7 ) coordinates with the additional assumption on the yang - mills field that the connection @xmath8 was purely magnetic . for our work here
, we will also make this
more specifically , we will assume that the electric charge density , @xmath78 , is identically zero .
in addition to this _ ansatz _ , we also make the following gauge choice : @xmath79 .
making this gauge choice leaves the connection invariant under a gauge transformation of the form @xmath80 . from the definition for @xmath78 in equation ( [ eq : defy ] ) we see that the function @xmath81 , by these assumptions , depends only on @xmath62
however , the residual gauge freedom implies that @xmath81 is arbitrary up to a function of @xmath62
. we can thus completely fix the remaining gauge freedom by choosing @xmath82 as well .
a consequence of these choices is that the remaining coefficients of the gauge connection will no longer be independent and one of them ( we choose @xmath67 ) can be set to zero without loss of generality .
thus , in the context of this particular problem , the magnetic _ ansatz _ has effectively set all the fields but @xmath19 to zero . as discussed in the body of the paper ,
we want to be able to evolve the system for long times to the future of any black hole formation . to this end
, we choose a coordinate system with maximal slicing : @xmath83 .
we will retain the choice of areal , or radial , coordinates ( @xmath7 ) so that the coordinate @xmath63 is immediately related to the area of origin - centered two - spheres .
given all these assumptions , the equations which we must solve simplify considerably .
the evolution equations are & = &
+ ar^2 w ( 1-w^2 ) + & = & + & = & a + w and the constraint equations are w & = & + & = & ( aa - 2r ) + 2r^2 ( a^2 - 1 +
2r aa ) + 4 g ( s - 3 ) + a & = & a 1-a^2 2r + 32 ra^3 k_^^2 + 4 g r a + k_^ & = & -3r k_^ + 4 g ( g^2 a r^2 ) where the matter stress - energy terms are given by s - 3 & = & a^2 ( 1-w^2)^2 2 g^2 r^4 + 1 g^2 r^2 ( ^2 + ^2 ) + & = & a^2 ( 1-w^2)^2 4 g^2 r^4 + 1 2 g^2 r^2 ( ^2 + ^2 ) .
finally , we have an algebraic condition for the non - vanishing component of the shift vector = r k_^ , which follows from ( [ eq : bdot ] ) with @xmath84 . | we extend the investigation of the gravitational collapse of a spherically symmetric yang - mills field in einstein gravity and show that , within the black hole regime , a new kind of critical behavior arises which separates black holes formed via type i collapse from black holes formed through type ii collapse .
further , we provide evidence that these new attracting critical solutions are in fact the previously discovered colored black holes with a single unstable mode .
i#1 # 1 epsf |
in a serie of papers kempf et al .
@xcite introduced a deformed quantum mecahnics based on generalized uncertainty principle ( gup ) .
other similar issues leading to the same gup have been also initiated by some authors @xcite .
one major consequence of the gup is the appearance of minimal uncertainties in position and/or momentum leading to an uv / ir mixing , which allows to probe short distance physics ( uv ) from long distance one ( ir ) .
a minimal uncertainty in position or minimal length has appeared in different context like string theory ven , groos , amati , loop quantum gravity @xcite and non - commutative field theories @xcite .
the minimal length defines a scale below it the physics becomes inaccessible , leading to a natural cut - off which prevents from the usual uv divergencies .
recently the schrodinger equation in momentum space for the harmonic oscillator with minimal lengths in arbitrary dimensions has been solved kempf0,kempf1,minic .
the cosmological constant problem and the classical limit of the physics with minimal lengths have been also investigated chang , chang01 . on the other hand
the effect of the minimal length on the properties of the 3d coulomb potential has been studied in refs .
brau , akhoury , benzik and of the 3d dirac oscillator using supersymmetric quantum mechanics @xcite .
the casimir force for the electromagnetic field in the presence of the minimal length has been also computed nouicer , harbach . in this paper
we are interested by the effect of the presence of minimal lengths on the energy eigenvalues and eigenfunctions of one dimensional quantum mechanical systems in singular potentials , like the attractive dirac potential and coulomb potential . in section 2 , after implementing the minimal length via generalized position and momentum operators , we derive fundamental relations for our caulculations . in sections 3 and 4 ,
the one dimensional attractive dirac potential and coulomb potential are considered in great details .
the last section is left for concluding remarks .
following @xcite we consider the following one dimensional realization of the position and momuntum operators @xmath0where @xmath1 is a parameter assumed to be large .
the representation @xmath2leads to the following generalized commutator and generalized uncertainty principle ( gup ) @xmath3 = i\hbar \exp ( p^{2}/\mu ^{2}),\quad \delta x\delta p\geq \frac{\hbar } { 2}\left [ 1+\frac{(\delta p)^{2}}{\mu ^{2}}\right ] .
\label{gup}\]]a consequence of this relation , is the appearance of a minimal length given by @xmath4which means a lost of localized states in the @xmath5-space since we can not probe the coordinates space with a resolution less than the minimal length . in this situation
we use the momentum space representation and maximally localized states to define a quasi position representation @xcite .
then the existence of a one parameter familly of position basis related to the minimal length allows us to write @xmath6 where the vectors @xmath7 represent maximal localisation states in the sence that @xmath8 .
they are normalized states unlike the ones of ordinary quantum mechanics . in the following we derive neccessary relations for our calculation taking in mind that we must recover the usual quantum mechanics in the limit @xmath9 . using the maximally localized vectors we derive the following maximally localized wavefunction @xmath10these states are far from being the well known plane waves .
however in the limit @xmath9 and choosing the normalization constant @xmath11 as @xmath12 we recover the usual planes waves of ordinary quantum mechanics.on the other hand the states defined by eq.([plane ] ) are physical ones since @xmath13using the completness relation @xmath14 and @xmath15 we obtain @xmath16with the aid of the relation @xmath17 , where @xmath18 are the roots of @xmath19 , we finally get@xmath20from this equation we have the usual completeness relation for the eigenstates @xmath21 @xmath22let us finally show that the maximally localized states @xmath23 like the coherent states , do not form an orthogonal set .
indeed we have@xmath24using the variable @xmath25 defined by @xmath26 we have @xmath27 the right hand is a well behaved function unlike the dirac distribution of ordinary quantum mechanics .
it is clear that the limit @xmath28 restores the usual normalization @xmath29 .
consider now the momentum representation of the one dimensional schrdinger equation for a particle of mass @xmath30 in the potential @xmath31 @xmath32 let us project on the momentum states @xmath21 and inserting the closure relations for the maximally localized states and the momentum states respectivelly we have @xmath33then using the expression of the quasi - position eigenvectors we obtain the following equation @xmath34 which can be written as @xmath35 where the potential @xmath36 is the generalized fourier transform of the potential @xmath37 @xmath38
let us begin our investigation of the effect of minimal length on eigenfunctions and eigenenergies by considering the one dimensional dirac delta potential @xmath39 this potential is known to shares some common properties with the coulomb potential in the limit of strong coupling . using
the generalized fourier transform @xmath40 we obtain the usual expression @xmath41 substituting in schrdinger equation @xmath42 we have @xmath43 this equation can be written in the following integral form @xmath44 with @xmath45 and @xmath46 a separable kernel given by @xmath47 where @xmath48 writing the wave function in the following form@xmath49we deduce the equation @xmath50 = 0\ ] ] which admits the obvious solution @xmath51 this equation is the spectral condition from which we extract the possible energy levels . using the expressions of @xmath52 and @xmath53 and performing the integral we obtain @xmath54 e^{\frac{p_{0}^{2}}{\mu ^{2}}}-1=0 , \label{int}\ ] ] where we set @xmath55 = -@xmath56 for bound states@xmath57 an asymptotic expansion in @xmath58 gives the following equation @xmath59 using the fact that @xmath58 is a small parameter , the solution of @xmath60 gives the following energy eigenvalue @xmath61 an asymptotic expansion in @xmath1 gives @xmath62 using the normalization condition @xmath63 and the the following formula @xcite @xmath64 to calculate the constant @xmath65 , the normalized wave function in momentum space is finally given by @xmath66 ^{-1/2}\frac{e^{-\frac{p_{0}^{2}}{4\mu ^{2}}}e^{-\frac{1}{2\mu ^{2}}p^{2}}}{\left ( p^{2}+p_{0}^{2}\right ) } .\ ] ] the maximally localized wave function @xmath67 are obtained from the generalized inverse fourier transform of the momentum space wave function given by @xmath68 thus using eq.(28 ) we obtain @xmath69 ^{-1/2}e^{-\frac{p_{0}^{2}}{4\mu ^{2}}}\int_{-\infty } ^{+\infty } dpe^{-\frac{p^{2}}{\mu ^{2}}}\frac{\exp \left ( \frac{ix}{2\hbar } \mu \sqrt{\pi } \hbox{erf}\left ( \frac{p}{\mu } \right ) \right ) } { \left ( p^{2}+p_{0}^{2}\right ) } .\ ] ] using the residus theorem we show that the second integral vanishes and we obtain @xmath70 ^{-1/2}e^{\frac{3p_{0}^{2}}{4\mu ^{2}}}\left [ e^{-\frac{\left\vert x\right\vert } { 2\hbar } \mu \sqrt{\pi } \text{erfi}\left ( \frac{p_{0}}{\mu } \right ) } \right ] , \label{c}\ ] ] where erfi@xmath71 is the imaginary error function .
let us now define @xmath72 and rewrite schrdinger equation with an effective potential , in the manner of the theory without the minimal length , @xmath73 where @xmath74 using the inverse fourier transform we obtain the following regularized dirac potential in one dimension ( see figure 1 . )
@xmath75 the results in the case without the minimal length are obtained by observing , that for large @xmath76 we have @xcite @xmath77 and then the eigenfunctions @xmath78 and @xmath79 reduce respectively to the momentum and coordinate sapce eigenfunctions of the one dimentional dirac potantial without the minimal length
. @xmath80
finally we consider the one dimensional coulomb potential @xmath81 in the presence of a minimal length .
this potential without the minimal length , which seems to have applications in semiconductors or insulators , has been solved with different results @xcite . in the presence of the minimal length of kempf et al .
@xcite , this potential has been recently solved but with some incorrectness @xcite . using the following formula @xmath82 where * p * means the cauchy principal value of the integral , the coulomb potential in momentum space is then given by@xmath83 .\]]substituting in schrodinger s equation ( [ schro ] ) , we obtain @xmath84 \psi ( p^{\prime } ) = 0 , \label{c1}\]]where the last term @xmath85 is the signature of the singularity of coulomb potential in one dimension in coordinates space .
deriving @xmath86with respect to @xmath87 gives@xmath88the solution to this equation is given by @xmath89 with the normalization constant , calculated in section 3 , given by @xmath90 ^{-1/2}.\ ] ] expanding the last term in @xmath91 and integrating over @xmath87 we obtain @xmath92 the requirement that the wave function must be single valued gives the spectral condition @xmath93 solving this equation and using @xmath94 we obatin @xmath95 an asymtotic expansion in @xmath1 gives @xmath96 and@xmath97the first term in @xmath98 is the spectrum of the usual one dimensional coulomb potential @xcite , while the remaining ones are the corrections due to the perturbation of the space by the presence of the minimal length . the correction term in @xmath98
is in concordance with the ones to the energy spectrum of coulomb potential in three dimensions derived in @xcite . in @xcite
an additional correction term proportional to @xmath99 @xmath100 has been taken into account .
however as noted in @xcite , this term like e@xmath101 in eq.@xmath102 will affect the outgoing scattered waves .
the wave functions associated with the energy levels given by @xmath103 are physically acceptable ones unless @xmath1 is large but finite . in the two cases
we can extract a bound for the minimal length .
in fact requiring that @xmath104 for bound states we obtain @xmath105 setting @xmath106 and the fact that bohr radius is @xmath107 m we obtain @xmath108 let us finally discuss about the original singularity of coulomb potential in one dimension expressed by @xmath109 . as noted above
, @xmath110 is an account of the discontinuity at the origin of the wave function in the coordinates space@xmath57with the aid of the wavefunctions given by @xmath111 we show that @xmath112 as it is obvious from @xmath113 the singularity is completely removed by the presence of the minimal length .
this is a natural consequence since the minimal length play the role of a cut - off which suppress the contribution of high momentum .
this was not the situation in @xcite .
the regularizing effect of the minimal length is best seen by computing the effective coulomb potential induced by the presence of the minimal . indeed reappeting the same calculations leading to @xmath114 along with @xmath115 and taking the inverse fourier sine
transform we obtain @xmath116 we note that the usual divergence is removed and the presence of the minimal length regularizes the theory ( see figure 2 below ) .
similar effective coulomb potential in three dimensions induced by the non commutativity of the space - time coordinates have been recently computed @xcite
in this paper we have solved , in the framework of quantum mechanics in the presence of minimal lengths , the problem of a particle moving in simple singular one dimensional potentials namely , the dirac delta potential and the coulomb potential .
our calculations are based on using physical maximal localisation states . in the two cases
we have calculated the effective potential induced by the presence of the minimal length and as a consequence we have shown that the singularity at the origin , present in the case without the minimal length , is now completely removed . for a particle in the coulomb potential the corrections to the energy spectrum are in concordance with that of refs.@xcite . | in this paper we address the problem of a particle moving in singular one dimensional potentials in the framework of quantum mechanics with minimal length . using the momentum space representation
we solve exactly the schrodinger equation for the dirac delta potential and coulomb potential .
the effect of the minimal length is revealed by a computation of effective generalized dirac delta potential and coulomb potential .
pacs numbers : 02.40.gh,03.65.ge |
before 1995 , only 24 quasars with redshifts larger than four had been published ; most had been found in multicolor or grism surveys for high - redshift quasars ( e.g. , warren et al .
1987 ; schmidt , schneider , & gunn 1987 ) . in the past eight years
the number of known @xmath6 quasars has increased to more than 400 , largely due to the discoveries from the sloan digital sky survey ( sdss ; 2000 ) and the digitized palomar sky survey ( kennefick , djorgovski , & meylan 1996 ) .
the sdss to date has published the discovery of over 200 quasars with redshifts greater than 3.6 ( fan et al .
2001b ; anderson et al . 2001
; schneider et al . 2001 ; schneider et al .
in addition , ( 2000 ) , ( 2001a ) , and ( 2003 ) have discovered the seven highest - redshift quasars known , all with @xmath7 and four of which are the subject of this paper .
this large sample provides us with new opportunities to study high - redshift quasars ; it has already been used to determine the luminosity function of quasars and its evolution with redshift from @xmath8 to @xmath9 ( fan et al . 2001c , 2003 ) .
the luminosity function is well - fit with a power - law , while the comoving number density of quasars drops exponentially with redshift over this range .
one of the goals of this paper is to determine the extent to which the observed luminosity function might be biased by gravitational lensing . understanding the ( intrinsic ) quasar luminosity function and evolution thereof at high redshifts is important for several reasons .
we observe that supermassive , quiescent black holes are ubiquitous at low redshift ( magorrian et al .
1998 ) ; comparison of the present - day mass function with the high - redshift luminosity function can constrain models for the duty cycle of black holes and their feeding mechanisms .
there is increasing evidence for a population of optically faint agn at high redshift , which appear in deep x - ray images with the _ chandra _ satellite ( mushotzky et al .
2000 ; brandt et al .
2001 ; alexander et al .
2001 ) ; understanding their nature requires that we understand the nature of the optically luminous objects as well .
finally , the optical ( rest - frame uv ) luminosity function puts important constraints on the ultraviolet ionizing background ( fan et al .
2001c ) , which is especially important now that we appear to be observing the first hints that the universe became completely reionized around redshift @xmath10 ( e.g. , becker et al . 2001
; djorgovski et al .
2001 ; white et al .
2003 ) .
the continuum luminosities of the @xmath7 quasars are all very high , with absolute magnitudes at rest frame 1450@xmath11 between @xmath12 and @xmath13 for the wmap cosmology that we use throughout this paper ( a flat @xmath14 cdm model with @xmath15 and @xmath16 ; 2003 ) .
this yields a lower limit to the masses of their black holes of order @xmath17 by the eddington argument ( fan et al .
it is a challenge to explain in standard cosmologies how such massive black holes could form at a cosmic epoch less than a billion years after the big bang ( e.g. , efstathiou & rees 1988 ; turner 1991 ; haiman & loeb 2001 ) .
all these results assume that the apparent magnitudes that we measure for these quasars are intrinsic to the quasar themselves , and represent isotropic fluxes
. however , gravitational lensing can systematically brighten the images of quasars by a significant amount .
this is a particularly important effect for flux - limited surveys like sdss , which are sensitive to the most luminous quasars at any given redshift ( e.g. , maoz & rix 1993 ) . the resulting magnification bias ( e.g.
, pei 1995 , and references therein ) can systematically change the quasar luminosity function , causing an overestimation of the black hole masses powering these objects .
the expected fraction of multiply imaged quasars at a given redshift depends both on the cosmological model ( which sets the distribution and profiles of halos as a function of redshift , as well as the dependence of cosmological pathlength on redshift ) , and on the luminosity function of quasars , which determines the effect of magnification bias . in the wmap cosmology ( using a halo distribution taken from the large n - body simulations of 2001 ) , the fraction of random lines of sight at @xmath18 that produce multiple images at all splitting angles ( see the discussion of lensing models in 3 below ) is of order 0.2% ; this fraction rises to 0.4% at @xmath10 .
however , magnification bias greatly increases the lensing rate in a flux - limited sample . at @xmath18 , using the canonical cosmology , and a luminosity function @xmath19 ( fan et al .
2001b ) with a break at @xmath20 ( see 3 below for further discussion ) , we expect about 1% of quasars in our sample to be lensed ( specifically , multiply imaged with magnification of at least two ) .
this fraction increases to 8% if we assume a steeper slope of @xmath21 , and to @xmath22 if we keep the slope of @xmath23 , but assume a break in @xmath24 at a 10 times lower luminosity ( @xmath25 ) . at @xmath10
, magnification bias can result in lensing probabilities approaching 100% if the luminosity function is particularly steep ( comerford , haiman , & schaye 2002 ; wyithe & loeb 2002a , 2002b ) .
however , the detailed shape of @xmath24 is not well - measured , especially at higher redshifts and luminosities .
such quasars represent the highest peaks in the density field at high redshift , where the mass function of massive dark halos in which bright quasars likely reside is expected to be very steep ( fan et al .
2001c ) .
the median expected splitting of gravitationally lensed quasars is predicted to be somewhat less than @xmath26 ( turner , ostriker , & gott 1984 ) , with flux ratios as large as 20:1 ; this is in broad agreement with the observed splitting separation found by the cosmic lens all sky survey ( class ; 2003 ) , but see the discussion in 3 below on class implications for the details of the lens profiles .
the sdss images themselves have psf widths of order @xmath27 ( abazajian et al .
2003 ) , making them inadequate to look for all but the very rare wide - separation ( @xmath28 ) lenses ( pindor et al .
we are thus carrying out a snapshot imaging survey of 250 sdss - discovered quasars with @xmath29 using the advanced camera for surveys ( acs ; 2003 ) on the _ hubble space telescope ( hst ) _ , which can easily resolve pairs separated by as little as @xmath30 with flux ratios as large as 10:1 .
our observations should therefore be sensitive to essentially all expected lenses ( turner et al .
1984 ; hinshaw & krauss 1987 ) .
( 2003 ) presented some preliminary results based on some of these ( and other , non-_hst _ ) data ; here we give a full analysis of the _ hst _ observations for all four of the @xmath7 quasars in our sample . throughout this paper ,
we adopt the wmap cosmology mentioned above , with an rms mass fluctuation within a sphere of radius @xmath31 mpc of @xmath32 , and power law index @xmath33 for the power spectrum of density fluctuations ( spergel et al . 2003 ) .
we also adopt the cosmological transfer function from & hu ( 1999 ) .
conversions between @xmath34 and @xmath35 assume @xmath36 ( schmidt , schneider , & gunn 1995 ) with spectral index @xmath37 .
the full sample of @xmath29 quasars included in our hst snapshot program will be presented in future papers in this series . in this paper , we present results of imaging the four quasars with @xmath38 discovered by ( 2000 , 2001a ) ; see table [ tab : tab1 ] .
these four objects were selected from the sdss imaging database based on their extremely red @xmath39 color , absence of detection in @xmath40 and @xmath41 , and relatively blue @xmath42 color ( which distinguishes the @xmath43 quasars from brown dwarfs ) .
the sdss images of these quasars have an image scale of @xmath44 and @xmath45-band seeing fwhm between @xmath46 and @xmath47 . in order to test the ability of the sdss photometric pipeline to identify quasar pairs
, we created simulated sdss images of pairs of point sources ( see 2003 for details ) in these observing conditions and at the appropriate signal - to - noise ratio .
we find that the sdss star - galaxy separator would have identified pairs of point sources as resolved objects for image separations greater than @xmath48 and for flux ratios less than @xmath49 5:1 .
thus the sdss data are suitable for exploring only a fraction of the parameter space of lensing ( in terms of separations and flux ratios ) that is of interest ; higher resolution _ hst _ observations were needed to probe the rest of that parameter space .
we did not restrict our sample to point sources since the sdss star - galaxy separation algorithm ( see 2002 ) breaks down at these faint magnitudes .
the only object of the four that was flagged as an extended source was sdssp j104433.04 - 012502.2 .
we used the high resolution camera ( hrc ) on acs to image the four @xmath7 quasars presented herein . for these four objects , the ly@xmath50 forest lies in the sdss @xmath51-band ; we therefore observed them in the sdss @xmath45-band , f850lp , despite its lower sensitivity .
the wide - field camera on acs has higher sensitivity in @xmath51 and @xmath45 than does the hrc , but has has substantially higher overhead and is mildly undersampled .
the exposure times were 1200 seconds for each object , 600 seconds in each of two exposures to help in cosmic ray rejection .
we did not dither the images , and our cosmic ray rejection is far from perfect ( 2.2 ) ; however , our analysis indicates that imperfect cosmic ray removal is not critical for our search for gravitational lenses .
the raw images were calibrated by the calacs package in iraf as part of on - the - fly - reprocessing ( otfr ) at the time of download .
for the most part , we find these initial reductions to be sufficient for our purposes .
the images that we present are the `` cosmic ray rejected '' ( crj ) images that are output by the otfr algorithms at stsci .
these files were re - requested a number of weeks after the observations were completed to ensure that we are using the latest `` reference '' files and that the otfr reprocessing is current .
the crj files have all been reduced in the standard manner , including having been overscan- , bias- and dark - corrected , flat - fielded and photometrically calibrated , in addition to having bad pixels masked and cosmic rays removed ( see the acs manual for more details ) .
the plate scale on the hrc of the acs is not the same in the @xmath52 and @xmath53 directions and the sky axes are not perpendicular on the ccd ; these crj images have _ not _ been corrected for this distortion .
a @xmath54 cutout of each of the acs images is given in figures [ fig : fig1]-[fig : fig4 ] .
with the exception of cosmic ray rejection and hot pixel identification , the images output by the otfr algorithm are acceptable for our purposes .
the calacs rejects pixels that are significantly different in the two images .
this process successfully rejects the majority of the cosmic rays in our images . however , many of the cosmic rays occupy multiple pixels , and we have only two exposures , meaning that some cosmic rays remain .
in addition , hot pixels can remain in the corrected images if they have not been properly identified in the masks that are constructed for each period of observations between the annealing procedure that helps to repair hot pixels .
fortunately , the morphologies of the residual cosmic rays and hot pixels are considerably different from that of the acs psf ( the psf is spread out over many tens of pixels ) , so spurious sources due to cosmic rays and most other defects are easy to recognize . the point spread function is narrow enough that any lens with a separation greater than @xmath55 will be obvious by visual inspection ( see [ sec : multiple ] ) . to search for faint secondary images at separations smaller than this
requires that we fit and subtract off a model for the point spread function of each image .
we have done this using the v6.0 ( pre - launch of acs ) version of the tiny tim software ( krist 1995 ) , which produces a model psf for the instruments on _ hst _ given the object s sed and position in the focal plane , the filter curve , and knowledge of the optics of the instrument .
such model psfs are necessary because it is not always possible to do psf subtraction using an empirical template ( another point source in the field - of - view ) .
that is , given the positional dependence of the psf across the field - of - view , the tiny tim models are often better than using an empirical template . for our acs / hrc images ,
the model correctly takes into account the different scales and non - orthogonality of the @xmath52 and @xmath53 axes .
we fit the psf model to each crj image , allowing the location on the ccd and normalization to vary , and using sinc interpolation when shifting the model psf by fractional pixels .
inputs to tiny tim were the camera used ( acs / hrc ) , the location on the ccd of each quasar to the nearest pixel ( as measured by imexam in iraf ) and a keck spectrum ( becker et al . 2001 ) of each quasar ( for the wavelength dependence of the psf ) .
we minimized the sum of the square of residuals ( since the formal errors per pixel over the image are close to constant ) .
the results are shown in figures [ fig : fig1][fig : fig4 ] .
each of the four images show the familiar first airy ring ; on a very hard stretch , the second airy ring is faintly visible .
note that _ hst _ psfs are not symmetric as a result of the three primary mirror supports , the signature of which can be seen in the psf subtracted images .
given inevitably somewhat imperfect psf subtraction and the expected asymmetry of the residual images , in none of the four cases is there an obvious secondary image ; none of them are obviously gravitationally lensed .
the acs / hrc image of this quasar is shown in the left panel of figure [ fig : fig1 ] ; the right panel shows the residuals after psf subtraction .
the sdss @xmath45-band magnitude of this quasar is 19.3 ; the _ hst _ @xmath45-band ( st ) magnitude is 19.91 , see also table [ tab : tab1 ] .
the offset between the sdss and st @xmath45-band magnitude is systematic ( see the other objects below ) with the st magnitudes always being fainter .
this is probably due to the differences in the filter curves ( the sdss curve having less red response due to atmospheric features ) and the differences between a vega - based ( st ) and an ab - based ( sdss ) photometric system given that vega has an ab @xmath45-band magnitude of 0.57 ( fukugita et al .
we find no offset between the sdss and st @xmath51-band magnitudes ( in the _ hst _ images of our lower redshift quasars ) , where the filter curves appear to be much more similar .
\(2000 ) describe a @xmath57 image of this object taken with the keck 10 m telescope in @xmath58 seeing , which was consistent with a point source .
that is consistent with the hst image ; there is no apparent splitting in this object .
\(2002 ) found a faint galaxy ( @xmath59 ) @xmath60 south - west of this quasar .
our 20-minute exposure shows this object faintly after rebinning .
( 2002 ) suggest that this galaxy could amplify the flux of the quasar by up to a factor of two .
the lack of a secondary image of the quasar restricts any possible magnification to be no larger than this in the absence of microlensing and external shear ( wyithe & loeb 2002a ) .
the acs / hrc image of this quasar and residuals after psf subtraction are shown in figure [ fig : fig2 ] .
the sdss @xmath45-band magnitude of this quasar is 18.79 ; the _ hst _ @xmath45-band ( st ) magnitude is 19.49 , see also table [ tab : tab1 ] .
one other point source and two other extended sources are in the acs / hrc field of this quasar , but all are more than @xmath62 away ( out of the field - of - view of fig .
[ fig : fig2 ] ) , and are therefore unlikely to lens ( or significantly magnify ) the quasar .
the black dots in the field - of - view of figure [ fig : fig2 ] are residual cosmic rays that were not properly cleaned by the otfr reductions .
the acs / hrc image of this quasar is shown in figure [ fig : fig3 ] .
the sdss @xmath45-band magnitude of this quasar is 19.46 ; the _ hst _ @xmath45-band ( st ) magnitude is 20.23 , see also table [ tab : tab1 ] .
there is some evidence for a few very faint extended sources in the acs / hrc field ( off the image in fig .
[ fig : fig3 ] ) , but as with figure [ fig : fig2 ] , all of the black dots in figure [ fig : fig3 ] are residual cosmic rays .
( 2003 ) have taken deep images to @xmath64 in moderate seeing with subaru in this field , and have found no foreground galaxies close enough to the line of sight to cause gravitational lensing .
the acs / hrc image of this quasar is shown in figure [ fig : fig4 ] .
the sdss @xmath45-band magnitude of this quasar is 20.02 ; the _ hst _ @xmath45-band ( st ) magnitude is 20.54 , see also table [ tab : tab1 ] .
there is one other relatively bright point source in the field , nearly due west of the target quasar , but it is at the edge of the field ( off the image in fig . [
fig : fig4 ] ) and is not of interest in terms of lensing . as in the previous object , a deep image by ( 2003 )
shows no obvious lensing galaxies in the foreground .
however , ( 2003 ) have taken deep vla images at 20 cm in this field , and found a statistically significant overdensity of four radio sources within @xmath66 of the quasar , which may indicate the possibility of a magnifying cluster along the line of sight .
our acs images have an image scale of @xmath67 , and we can visually rule out any multiple imaging at scales larger than about @xmath68 . to make this more quantitative ,
we have carried out simulations of quasar pairs at various separations and flux ratios , by adding a scaled - down and shifted version of our quasar images to itself .
this naive procedure produces a slightly incorrect sky level , but the inaccuracy is negligible for our present purposes .
figure [ fig : sim3 ] shows simulated observations of sdss j0836 + 0054 with an additional point source 2.5 and 5.0 mag fainter than the quasar and separated from the quasar by @xmath69 .
the secondary image is obvious in both cases even without psf subtraction .
figure [ fig : sim1 ] shows the result of a @xmath30 offset . with a flux ratio of 10:1 , the secondary object is visible as an enhancement in the first airy ring of the primary object ; it becomes clear upon psf subtraction ( figure [ fig : sim1resid ] ) . with a flux ratio of 100:1
, we can not discern a pair with @xmath30 separation even after subtraction of the psf ; however , even if we were able to do so , it would not significantly improve the constraints that we derive below since most split images will have larger image separations .
the shape of the high - redshift quasar luminosity function is not tightly constrained because of the limited size and luminosity range of the available statistical samples .
( 1995 ) , using a set of 90 quasars with @xmath70 , found a power - law luminosity function slope ( @xmath71 ) of @xmath72 .
fan et al .
( 2001c ) measured @xmath73 from 39 quasars in the redshift interval 3.6 to 5.0 .
these slopes apply to the high - luminosity ( @xmath74 ) part of the luminosity function . at these high redshifts ,
direct observations currently yield only poor constraints on the slope of the quasar luminosity function ( lf ) at @xmath75 , and essentially no constraints at @xmath76 , where the luminosities of the known quasars all lie within @xmath77 magnitudes of the detection threshold of sdss . from the number counts of @xmath2 quasars , ( 2001c )
place a 3-@xmath78 lower limit on the slope of @xmath79 .
gravitational lensing tends to flatten the intrinsic luminosity function , producing an apparent slope of @xmath80 in the limit of an arbitrarily steep intrinsic lf down to a fixed low
luminosity cutoff ( e.g. , schneider , ehlers , & falco 1992 ) . as a result , the relatively shallow apparent slope at @xmath75 does not necessarily reflect the slope of the underlying lf , which could be much steeper , if most quasars were lensed and magnified .
one of the primary goals of this work is to constrain the slope of the quasar lf at high redshift , using the lack of evidence for strong lensing among the four quasars in the hst images ( comerford et al .
2002 ) . in line with previous work
, we describe the intrinsic ( not necessarily observed ) quasar luminosity function as a broken power law ( e.g. , boyle , shanks , & peterson 1988 ; pei 1995 ) : @xmath81 the lf is described by four parameters : the normalization @xmath82 , the faint - end slope @xmath83 , the bright - end slope @xmath84 , and the characteristic luminosity @xmath85 at which the lf steepens .
the lensing probability is very sensitive to the last two parameters , @xmath84 and @xmath85 , while it is only weakly dependent on the faint
end slope below the break , and it is strictly independent of the normalization @xmath82 . here
we apply the lensing model from ( 2002 ) , in which lenses are associated with dark matter halos , to compute the total lensing probability , including the effect of magnification bias . in this model , the abundance of halos as a function of potential well depth is adopted from the simulations of ( 2001 ) .
while it would be simplest to assume that the halos have a common density profile , such as the singular isothermal sphere ( sis ) , recent work has suggested that the typical density profile of massive lenses is shallower than that of the sis ( porciani & madau 2000 ; kochanek & white 2001 ; li & ostriker 2002 ; huterer & ma 2003 ) . the observed dearth of large separation lenses can not be explained if all lenses have the same density profile .
a prescription that fits the observed steep splitting angle distribution is to assume that all halos below @xmath86 have sis profiles ( adopting a standard conversion between circular velocity and halo mass ) , while all halos above this mass follow the dark matter density profile suggested by , frenk , & white ( 1997 , hereafter nfw ) .
nfw profiles are much less efficient lenses than are sis profiles .
this prescription is physically motivated in & white ( 2001 ) based on the efficiency of gas cooling , and we adopt it in this paper . in our calculations , this is essentially equivalent to ignoring all lenses above a halo mass of @xmath87 , as the massive halos do not contribute to lensing at small separations . as an example
, we first fix the characteristic luminosity ( `` knee '' ) of the quasar lf at a fiducial value of @xmath20 , and derive constraints on the bright - end slope , @xmath84 .
the choice for this value is somewhat arbitrary : the highest redshift where the quasar lf is measured to faint enough magnitudes to define a reliable break is at @xmath88 ( boyle et al .
2000 ) , where @xmath89 . in pure luminosity evolution models , the break at @xmath90
would then be at around @xmath20 ( e.g. madau , haardt , & rees 1999 ; wyithe & loeb 2002a ) .
having fixed the value of the break , we next vary the slope @xmath84 of the lf , and require that the expected probability for no lensing among the four highest redshift quasars is p=0.0026 ( @xmath91 ) or greater . assuming that we can rule out a second image with a flux ratio of 100 down to a splitting angle of @xmath69 , we find the constraint on the slope @xmath92 .
we find that resolving high contrast lensing configurations , with a flux ratio @xmath93 , or resolving angular separations smaller than @xmath69 does not significantly improve this constraint , since in any model , the fraction of lenses with separations less than @xmath69 is small .
for example , ruling out flux ratios of 10 down to @xmath30 would only improve the constraint to @xmath94 .
it is more informative to illustrate joint constraints on the parameters @xmath95 . in figure
[ fig : contours ] , we show contours for the likelihood at @xmath96 , 0.05 , 0.01 , and 0.001 that none of the four @xmath7 quasars are lensed ( again , to a separation of 0.3 arcseconds , and to flux ratios of 100 ) .
the figure shows , as in ( 2003 ) , that the derived limit on the slope @xmath84 depends strongly on the assumed break .
overall , the constraint shown in this figure is similar to that derived from the sample of seven high redshift quasars in ( 2003 ) , although the present constraints are somewhat weaker , due to the smaller number of sources .
the recent results of ( 2003 ) are consistent with ours under the assumption of no lensing . as shown in ( 2003 ) , a stronger constraint is available if two of the known @xmath97 quasars are indeed magnified , as suggested by ( 2002 ) and ( 2003 ) .
the lack of lenses among the four high redshift quasars allows us to place constraints on the quasar lf at high redshift , shown in figure [ fig : contours ] .
it is useful to keep in mind that only models with extreme magnification bias , predicting significant lensing probabilities , can be constrained by the _ lack _ of lensing events .
for example , for the range of lf slopes considered by & loeb ( 2002a ) , @xmath98 , the lensing fraction for the @xmath2 sdss sample would be 7 - 30% .
the chance of finding no lenses among the four quasars for their steepest slope of 3.43 is @xmath99 , which translates to a significance of only @xmath100 for ruling out this slope ( consistent with our results shown in figure [ fig : contours ] at the break of @xmath101 ) . accordingly
, as figure [ fig : contours ] shows , the constraints obtained here still allow very steep lf slopes , especially if we assume that the break in the lf lies at relatively bright magnitudes .
nevertheless , these constraints are fairly direct , and are among the most stringent limits we currently have on the shape of the high - redshift quasar lf .
we expect that the full @xmath6 sample will further tighten the constraints ( or yield a measurement should a lens be found ) .
table [ tab : tab2 ] shows that the lensing probability ( for a slope of @xmath102 and break @xmath101 ) decreases from 8.2% at @xmath10 to 1.2% at @xmath18 .
this implies that seeing no lensing with a single @xmath10 quasar provides a constraint that is equivalent to about seven @xmath18 quasars , from the equation @xmath103 .
the lensing probability is a steep function of @xmath84 and @xmath104 , but since our full sample will include @xmath105 times more sources than utilized here , we expect that we will be able to significantly tighten the constraints in the @xmath106 plane with this sample
. given our assumed shape of the quasar lf , the constraints obtained in the @xmath106 plane translate directly to the global ionizing emissivity of the @xmath107 quasar population , @xmath108 .
the emissivity is of great interest , since a related quantity the mean background flux can be constrained independently from the level of resonant neutral hydrogen absorption detected in high resolution spectra of the sdss quasars .
specifically , recent upper limits on this background were obtained from the gunn peterson ( 1965 ) trough of the @xmath65 quasar ( e.g. , mcdonald & miralda - escud ' e 2001 ; cen & mcdonald 2002 ; lidz et al .
2002 ; fan et al .
2002 ) , requiring that @xmath109 ionizations per hydrogen atom per second at @xmath10 .
the conversion to the ionizing emissivity @xmath110 is @xmath111 where we have assumed @xmath112 , @xmath113 is the hydrogen ionization cross section at the lyman limit ( and we assumed @xmath114 above the threshold ) , @xmath115 is the planck constant , @xmath116hz is the frequency of a lyman limit photon , and @xmath117 is the mean free path of ionizing photons ( see the appendix in 2002 ) .
the last quantity unfortunately has a very large uncertainty .
nevertheless , for a crude estimate , we here assume that the effective mean free path corresponds to a redshift interval of @xmath118 at @xmath119 ( see haardt & madau 1996 ; steidel , pettini , & adelberger 2001 ) , and scales with redshift as @xmath120 towards higher redshift ( cen & mcdonald 2002 ) . in this case
, the model with @xmath20 and @xmath121 corresponds to @xmath122 . in other words ,
the constraint we derive here from the absence of lensing is approximately at the same level as that implied by the upper limit on the ionizing background .
note that the constraints we obtain apply to the total emissivity of quasars alone , but the upper limits on the background from the gunn - peterson trough constrains the total emissivity ( including galaxies ) .
future improvements of the lensing constraint can yield tighter limits on the ionizing emissivity of the quasar population than hitherto possible subject to the uncertainties in ( 1 ) the mean evolution of the mean free path with redshift , ( 2 ) the faint
end slope @xmath83 , which does affect the total emissivity @xmath123 , and ( 3 ) the escape fraction of ionizing photons at @xmath107 .
the fact that none of the four high redshift quasars appears to be strongly magnified also implies that the masses inferred for the central black holes from the eddington limit argument are real .
if a quasar is strongly beamed towards us , the bolometric luminosity would be overestimated , but , mclure , & jarvis ( 2003 ) , following & cen ( 2002 ) argued that emission line equivalent widths being normal mitigates the beaming hypothesis .
the inferred black hole masses in these four quasars are very high , @xmath124 ; how black holes can grow so large so early in the universe requires explanation ( efstathiou & rees 1988 ; turner 1991 ) . & loeb ( 2001 ) considered the presence of the @xmath125 black hole for the @xmath126 quasar in simple models for black hole growth .
they found that it could be marginally accommodated in simple models of black hole growth by mergers and accretion in an lcdm universe .
these considerations depend strongly on the redshift , and on the typical radiative efficiency of black holes .
the highest redshift quasars therefore represent a bigger challenge to models , especially if their radiative efficiency during their growth was as high as 20% , the typical value inferred recently for the brightest quasars at lower redshifts ( yu & tremaine 2002 ) .
this is twice the fiducial value of @xmath127 adopted in & loeb ( 2001 ) and most other previous work on black hole growth , doubling the expected growth timescale .
finally , the lack of strong lensing appears consistent with the large apparent sizes of ionized regions around the @xmath107 quasars .
& cen ( 2002 ) argue that the @xmath65 source can not be magnified by more than a factor of a few , since an intrinsically faint source would not be able to ionize a neutral igm out to the observed distance away from the source of @xmath128 ( comoving ) mpc the spectrum shows transmission at wavelengths corresponding to this distance .
similar conclusions were drawn for the @xmath129 quasar ( fan et al .
2003 ) recently by ( 2003 ) .
we have obtained high resolution _ hst _ images of the four highest redshift quasars known prior to 2003 to look for the signature of gravitational lensing .
we have found no evidence of multiple images , significantly limiting the amount by which these quasars can be magnified by foreground mass concentrations ( in the absence of microlensing ) .
thus the masses of these quasars as derived from the eddington argument can not be significantly inflated by magnification .
the lack of any strong lenses puts a @xmath130 constraint on the bright end slope of the @xmath131 luminosity function of @xmath92 .
this limit on the slope of the luminosity function is also consistent with previous limits on the ionizing emissivity of @xmath2 quasars of @xmath122 ionizations per hydrogen atom per second based on the gunn - peterson trough in the highest redshift quasar studied herein .
the next paper in the series will present an analysis of a much larger sample of @xmath6 quasar observations with _
hst _ and will discuss the constraints that we can place upon the quasar luminosity function from that larger ( but lower redshift ) sample .
furthermore , there are many more @xmath6 ( and even a few @xmath7 ) quasars known now than were known at the time that these _ hst _ observations were approved .
higher resolution images should be taken of as many of these as possible ; much of the parameter space can even be covered by ground - based observations since _ hst _ resolution splittings are only needed to discern the smallest expected splittings .
funding for the creation and distribution of the sdss archive has been provided by the alfred p. sloan foundation , the participating institutions , the national aeronautics and space administration , the national science foundation , the u.s .
department of energy , the japanese monbukagakusho , and the max planck society .
the sdss web site is http://www.sdss.org/. the sdss is managed by the astrophysical research consortium ( arc ) for the participating institutions .
the participating institutions are the university of chicago , fermilab , the institute for advanced study , the japan participation group , the johns hopkins university , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , university of pittsburgh , princeton university , the united states naval observatory , and the university of washington .
this work was supported in part by hst grant hst - go-09472.01-a ( g. t. r. and m. a. s. ) , national science foundation grants ast-9900703 ( d. p. s. ) , ast-0071091 ( m. a. s. ) , ast-0307384 ( x. f. ) , ast-0307291 ( z. h. ) and ast-0307200 ( z. h. ) .
x. f. and d. j. e. acknowledge alfred p. sloan research fellowships . ,
h. c. , clampin , m. , hartig , g. f. , illingworth , g. d. , sirianni , m. , martel , a. r. , meurer , g. r. , mccann , w. j. , et al .
2003 , in future euv / uv and visible space astrophysics missions and instrumentation . edited by j. chris blades , oswald h. w. siegmund
. proceedings of the spie , volume 4854 , pp .
81 - 94 ( 2003 ) .
, 81 lllll [email protected] & 5.74 & 19.3 & 19.91 & 1 + 083643.85 + 005453.3 & 5.82 & 18.79 & 19.49 & 2 + 130608.26 + 035626.3 & 5.99 & 19.46 & 20.23 & 2 + 103027.10 + 052455.0 & 6.30 & 20.02 & 20.54 & 2 + | over the last few years , the sloan digital sky survey ( sdss ) has discovered several hundred quasars with redshift between 4.0 and 6.4 . including the effects of magnification bias , one expects a priori that an appreciable fraction of these objects are gravitationally lensed .
we have used the advanced camera for surveys on the _ hubble space telescope _ to carry out a snapshot imaging survey of high - redshift sdss quasars to search for gravitationally split lenses .
this paper , the first in a series reporting the results of the survey , describes snapshot observations of four quasars at @xmath0 and 6.30 , respectively .
we find that none of these objects has a lensed companion within 5 magnitudes with a separation larger than @xmath1 arcseconds ; within 2.5 magnitudes , we can rule out companions within 0.1 arcseconds .
based on the non - detection of strong lensing in these four systems , we constrain the @xmath2 luminosity function to a slope of @xmath3 @xmath4 , assuming a break in the quasar luminosity function at @xmath5 .
we discuss the implications of this constraint on the ionizing background due to quasars in the early universe . given that these quasars are not highly magnified , estimates of the masses of their central engines by the eddington argument must be taken seriously , possibly challenging models of black hole formation . |
solar flares are powerful events resulting from the sudden conversion of stored magnetic energy to particle acceleration and excess radiation in the corona , and releasing up to @xmath0 total erg of energy @xcite .
solar flares can have significant detrimental effects on earth , including disruption and damage to satellites , power grids , and telecommunications infrastructure @xcite .
since coronal magnetic field measurements are not currently available , much research has focused on the inference of coronal magnetic field structure from the photospheric magnetic field . in this , it is assumed that turbulent motions in the photosphere create complex photospheric magnetic configurations @xcite , and this will then be reflected in complex coronal magnetic configurations @xcite .
complex coronal field configurations store the energy that is released as a solar flare .
several researchers have studied the use of poisson statistics to relate flare rate of different active region ( ar ) classes ( e.g. , mcintosh or mount wilson classes ) to probability of observing a c- , m- , or x - class flare @xcite .
other researchers have studied the relationship between quantitative image parameters and the probability of observing a c- , m- , or x - class flare @xcite .
several researchers have also studied the relation between quantitative image parameters and flare size @xcite or flare index @xcite .
to the best of our knowledge , no researchers have studied the direct _ prediction _ of flare size on the basis of quantitative image parameters as we study here .
several researchers have performed a qualitative study of the time evolution of flare parameters by plotting parameters versus time @xcite .
other researchers have studied flare prediction in light of different predictive time windows @xcite . related to flare time properties ,
several researchers have studied the relation between quantitative image parameters and flare productivity and flare rates @xcite . to the best of our knowledge , no researchers have studied the direct prediction of time - to - flare on the basis of quantitative image parameters as we study here . in this paper , we use support vector machine regression ( svr ) to predict solar flare size and time - to - flare from quantitative measures of magnetogram complexity . the remainder of this paper is organized as follows . in section
[ sec : svm ] , we briefly introduce regression in general , svr in particular , and other applications of svr to time series analysis . in section [ sec : data_experiment ] , we explain our data set , experimental procedures , and error measures . in section [ sec : results ] we present our results of solar flare size prediction and solar flare time regression . in section [
sec : conclusions ] , we discuss our conclusions and future work .
in the general form of regression , we wish to predict some continuous target variable @xmath1 given some @xmath2-dimensional input vector @xmath3 @xcite . in this work
, we wish to predict the flare size ( `` size regression '' ) or time to flare occurrence ( `` time regression '' ) given a set of features describing magnetic complexity . the equation @xmath4 , which maps input data to the target variable , is determined through optimization of some criterion based on training data .
regression commonly takes the general form @xcite @xmath5 where @xmath6 is a function whose output indicates the target value @xmath1 for a given input vector @xmath7^\top$ ] ; @xmath8^\top$ ] is a vector of weights applied to the basis functions @xmath9 ; basis functions @xmath9 are some ( linear or nonlinear ) function of input data @xmath3 ; and @xmath10^\top$ ] . in this work
, the 38-dimensional input vector @xmath3 corresponds to the 38 magnetic complexity features described in @xcite and summarized in table [ tab : features ] , the weight parameters @xmath11 are chosen to optimize some criterion ( minimization of an @xmath12-insensitive error function for the case of svr ) , and the output @xmath13 indicates the prediction of flare size or time - to - flare .
the function @xmath13 is linearly related to @xmath11 but the transformation @xmath14 allows for introduction of a nonlinear relationship to the input data @xmath3 .
support vector machines ( svms ) @xcite have become popular and are considered state - of - the - art for regression problems @xcite .
an important property of svms is that they can be formulated as a convex optimization problem , assuring that any local optimum is also a global optimum @xcite .
support vector regression minimizes the regularized error function @xmath15 where @xmath16 is the inverse regularization parameter , @xmath17 is the total number of training points , @xmath18 is an @xmath12-insensitive error function , @xmath19 is the output for the @xmath20-th training point , @xmath21 is the target for the @xmath20-th training point , and @xmath22 is the euclidean norm .
the use of an @xmath12-insensitive error function returns a value of 0 if the absolute difference @xmath23 and encourages sparse solutions @xcite . in this work
we use the matlab svm - km toolbox @xcite and the transformation @xmath24 implicitly defined by the radial basis ( gaussian ) kernel function @xmath25 .
the parameter @xmath12 is chosen based on the specific problem ( size or time regression ) and we optimize the values of @xmath16 and @xmath26 by grid search .
the optimal value of weight vector @xmath11 is determined by minimizing the error function in equation ( [ eq : svr_error ] ) .
this minimization is conducted using a training dataset which consists of @xmath17 pairs of feature vectors and labels @xmath27 ; the performance of the regressor is then determined on the basis of a test dataset consisting of pairs of feature vectors and labels disjoint from the training data . the use of a disjoint test dataset is important since the optimization can overfit a solution to the training data , resulting in an inflated performance when evaluated on the training data . since the parameters @xmath16 and @xmath26 must also be optimized , an additional dataset must be provided for this optimization ; this dataset is commonly called the tuning dataset ( since it is used to tune the parameters of the regressor ) and is disjoint from both the training and testing datasets . support vector regression has been shown to outperform other nonlinear methods , including multi - layer perceptrons , for time series prediction @xcite .
time series prediction algorithms have been used in many practical applications such as financial market prediction ( e.g. , @xcite ) , electric utility load forecasting ( e.g. , @xcite ) , weather and environmental prediction ( e.g. , @xcite ) , biomedical signal processing for estimation of missing microarray data ( e.g. , @xcite ) , and travel time prediction ( e.g. , @xcite ) . in the aforementioned references , there is no significant feature extraction and the focus is generally on prediction using one - dimensional time series data .
image deblurring is studied in @xcite in which the image regions are reformulated as one - dimensional signals prior to application of svr .
to the best of our knowledge , there are no studies of svr applied to the problem of estimating a continuous label vector from a high - dimensional image feature vector , nor of svr applied to the prediction of solar flare characteristics .
in our previous work @xcite , we used a large set of line - of - sight michelson doppler imager ( mdi ) magnetograms from 2000 through 2010 .
this included noaa ars 880910933 , and some 594,000 total images .
we describe the complexity of each ar by extracting a large set of features of postulated importance for the build up of magnetic energy and compare these to the onset of solar flares over a range of time periods following each magnetogram .
we consider three general categories of features : ( 1 ) snapshots in space and time , encompassing total magnetic flux , magnetic gradients , and neutral lines , ( 2 ) evolution in time , encompassing flux evolution , and ( 3 ) structures at multiple size scales , encompassing wavelet analysis . in total ,
38 different spatial features are extracted from each magnetogram image as summarized in table [ tab : features ] and as described in detail in @xcite . in this work ,
we predict the solar flare sizes or predict the time to flare occurrence ( when in the future a flare is expected to occur ) by redefining the label vector to be continuous rather than discrete .
these continuous label vectors have continuous numbers related to the flare size or time of flare at each flaring region . here
we describe and define the error measures we use for quantifying the performance of the regression models . in the following equations
, @xmath1 denotes the actual target value , @xmath28 is the target value estimated by regression , and @xmath17 is the total number of data points . for notational simplicity
, we drop the index over @xmath29 and interpret any summations as being over this range .
the root mean square error ( rmse ) measures the square root of the mean square error : @xmath30 the mean absolute error ( mae ) is given by : @xmath31 both the rmse and mae are expressed in the same units as the original target variable @xmath1 and give a measure of the unsigned error .
the relative mean error ( rme ) is defined as : @xmath32 the rme is a signed error which can give a measure of the bias of the regressor , i.e. , whether we consistently over- or under - estimate the target variable @xmath1 .
positive values of rme indicate a systematic underestimation while negative value indicate a systematic overestimation of target variables .
an rme of zero indicates no systematic bias in error .
the pearson s correlation coefficient is given by : @xmath33 where @xmath34 is the mean of @xmath1 over the @xmath17 data points , and @xmath35 is the mean of the estimates @xmath28 over the @xmath17 data points .
@xmath36 measures the linear correlation between the actual flare sizes and predicted ( regressed ) flare sizes , giving a value of @xmath37 if the actual and predicted values are perfectly correlated , 0 if they are completely uncorrelated , and @xmath38 if they are perfectly negatively correlated . in our size regression experiments ,
the goal is to predict the size of a future solar flare using svr .
we consider two different experiments for size regression . in experiment 1
we apply svr only to those regions which ultimately flare , ignoring the non - flaring regions
. this will give the most accurate results of the discriminative potential of magnetogram complexity features for determining solar flare size .
we note that in an operational scenario , an automated classification algorithm ( such as that presented in @xcite ) could be used to automatically separate flaring from non - flaring regions prior to size regression . in experiment 2
we apply svr to both flaring and non - flaring regions , as well as consider a thresholding of the regression results for flare prediction .
solar flares are categorized as a- , b- , c- , m- or x - class according to the peak flux ( in watts per square meter , w m@xmath39 ) of 100800 pm x - rays near earth , as measured by the _ geostationary operational environmental satellite _ ( _ goes _ ) . as reference ,
table [ tab : flare_sizes ] shows the solar flare classes and their corresponding peak fluxes .
these _ goes _ classes are the standard for measuring absolute size of flares ; however , due to the exponential nature of the class definitions , goes classes are not convenient for measuring the error of a size estimate .
for example an error of @xmath40 w m@xmath39 has a very different significance when interpreted in light of a c - class flare than an x - class flare .
as such , we seek a continuous label corresponding to flare size for which a given error can be interpreted independent of the underlying flare class .
this would imply , for example , that an error of 1 arbitrary units would indicate a prediction of a c2.0 rather than a c1.0 _ or _
an x2.0 rather than an x1.0 . rather than an exponentially spaced label vector
, we define a linearly spaced continuous - valued label vector for flare size as follows and as summarized in table [ tab : flare_sizes ] .
flare magnitudes are offset by 0 for b - class , 9 for c - class , 18 for m - class , and 27 for x - class .
for example , a c3.3 flare would be assigned value @xmath41 and an m5.7 would be assigned value @xmath42 .
we note that the two a - class flares in our dataset are assigned fractional values close to 1 and are aggregated into the b - class for data balancing ( discussed below ) .
we define this linearly spaced continuous - valued label vector for flare size for a range of different predictive time windows .
thus , for a predictive time window of @xmath43 hours , if an ar flares within @xmath43 hours of the present time , we define the label as the linearly spaced flare size .
since the predictive time window is larger than the nominal cadence of the mdi data ( 96 minutes ) , a single flare event will be regressed independently by multiple feature vectors .
furthermore , it is possible that more than one flare occurs in the given predictive time window .
in such a case , we define the corresponding entry in the label vector to be the largest flare size which occurs in that time window .
we repeat this experiment for @xmath43 in the range @xmath44 $ ] hours in steps of 2 hr .
the error measures are computed to compare the estimated flare sizes with the real flare sizes at each time @xmath43 in the time window .
we further note that the _ goes _ background level will have an effect on the accuracy of the regression for experiment 2 when non - flaring regions are included .
if a flare is predicted to be smaller than the concurrent _ goes _ background flux , the label computed from the _ goes _ flare catalog will indicate no flare and we will see an apparent error equal to the prediction flare size ; it is possible , however , that a flare actually did occur but was missed by the _ goes _ instrument due to the high background .
since the distribution of flares is unbalanced across the classes ( more smaller flares than larger ) , we balance the data by randomly subsampling the larger classes to the size of the smallest class to build a more accurate regression model
. the class with smallest number of data samples will be the x - class flares in our dataset .
we use 100-fold cross validation for each time @xmath45 $ ] .
in other words , we randomly subsample the data 100 times and train a regressor on each of these 100 data subsets ; furthermore , each of these 100 data subsets is balanced such that there is equal representation among the flare classes . since the data balancing inherently removes a large portion of the training data , cross validation in this scenario allows us to better utilize our entire training dataset by considering multiple randomly chosen subsets and averaging the results .
we choose 100-fold cross validation since the number of x - class flares is approximately 1% of the total number of flares in the worst case ; this means that , statistically , we will use each data sample at least once in the training process .
we choose a value of @xmath46 which will penalize errors greater than a tenth of a magnitude ; we choose this value on the basis of the fact that the _ goes _ magnitudes are specified only to a significant digit of a tenth .
the svr inverse regularization parameter @xmath16 controls the smoothness of the decision boundary and the kernel parameter @xmath26 controls the smoothing of the input data ; both @xmath16 and @xmath26 are optimized for @xmath47 hr , 10-fold cross validation , and a tuning dataset consisting of a randomly chosen fourth of the dataset . a time window of @xmath47 hr is used since this is the dataset with the largest number of x - class flares , and 10-fold cross validation is used rather than 100-fold cross validation for computational reasons .
no noticeable difference in average performance was noted with 10-fold versus 100-fold cross validation simulations . in our time regression experiments ,
the goal is to predict how many hours in the future a flare will occur using svr . in experiment 3
we apply svr only to those regions which ultimately flare . for this experiment
we define a label vector containing the number of hours to the next flare in an ar , as measured from the current time .
since there is no distinguishing between flare sizes in this experiment , there is no need for data balancing .
we choose a value of @xmath48 which corresponds to the 96 minute cadence of the mdi data expressed in hours .
the svr inverse regularization parameter @xmath16 controls the smoothness of the decision boundary and the kernel parameter @xmath26 controls the smoothing of the input data ; both @xmath16 and @xmath26 are optimized using 10-fold cross validation and a * tuning dataset consisting of a * randomly chosen fourth of the data .
in this section , we present results for two different definitions of the basic flare size label vector as defined in table [ tab : flare_sizes ] . in experiment 1
we use the definitions for the flare label vector directly as summarized in table [ tab : flare_sizes ] .
experiment 1 is designed to study the discriminatory potential of the 38 magnetogram complexity measures for determining flare size ; we note that in an operational scenario , the ar could first be classified as flaring by a separate classifier ( e.g. , @xcite ) . in experiment 2
we additionally include non - flaring regions and assign those regions a flare size of 0 .
experiment 2 is designed to combine flare prediction ( i.e. , will a flare of any size occur ) with flare size prediction .
parameters @xmath16 and @xmath26 are optimized via grid search using a tuning dataset consisting of a randomly chosen fourth of the ars . for each of the 10-fold cross validation runs , this tuning dataset is randomly subsampled to balance the classes by choosing @xmath49 data points from each class for tuning training , where @xmath50 is the size of the smallest class ( x - class flares in all cases ) .
all remaining unbalanced data unseen in the tuning training is used as tuning test data . in other words , we vary parameters @xmath51 and @xmath52 and train 10 regressors for each combination of @xmath16 and @xmath26 using a balanced subset of the tuning dataset while testing performance using the tuning data which is disjoint from that used to train the regressor .
rmse results are computed over the tuning test data and averaged over the 10-fold cross validations ; we find a minimum rmse of approximately 4.9 for @xmath53 and @xmath54 which we use for subsequent size regression simulations in experiment 1 .
we now use the remaining @xmath55 of the ars unseen in tuning to cross validate performance of the size regression .
similar to the @xmath16 and @xmath26 optimization , training data are balanced and all data unused in the balanced training are used as test data for each of the 100 folds .
we plot the rmse , mae , rme , and correlation coefficient @xmath36 for size regression in fig .
[ fig : size_errors ] ; these errors are averaged over the test data ( unseen in training and tuning ) for the 100-fold cross validation runs .
we see an average rmse ( fig .
[ fig : size_errors](a ) ) ranging between 5.6 and 6.4 , with smaller errors for larger predictive time windows . with
our linearly spaced flare size label vector , recall that an error of 1 indicates an error of one magnitude increment ( e.g. , m1.0 versus m2.0 ) and an error of 10 indicates an error of one class increment ( e.g. , m1.0 versus x1.0 ) .
we thus find an rmse of slightly more than half a class .
similarly , we find an average mae ( fig .
[ fig : size_errors](b ) ) between 4.5 and 5.5 , again with smaller errors for larger predictive time windows and again indicating an error of approximately half a class .
looking at the rme in fig .
[ fig : size_errors](c ) , we see that our estimates of flare size are biased toward overestimating by 2.34.3 mag increments with larger bias for smaller predictive time windows . the correlation coefficient in fig .
[ fig : size_errors](d ) ranges from 0.46 to 0.51 , with larger correlation for longer predictive time windows . in fig . [
fig : size_errors ] we have included the standard deviation across the 100-fold cross validation runs as error bars .
we note that the smaller time windows have a larger standard deviation ; this is a direct result of the smaller sample sizes for the shorter time windows .
the error bars for all but the 2-hr time window are within the average values for larger time windows in rmse and mae .
thus the trend of decreasing error with increasing time window may be more of an artifact of small sample statistics than underlying physics .
even so , the consistency in error across predictive time windows may indicate a presence of magnetic complexity features correlated to eventual flare size which persist long before flaring .
it also appears that these more persistent magnetic complexity features may allow for better prediction of flare size since the bias shown in rme in fig .
[ fig : size_errors](c ) is smaller with increased window length and r in fig .
[ fig : size_errors](d ) is larger with increased window length . in fig .
[ fig : size_error_hist](a ) we plot a histogram of the error @xmath56 , aggregated over the 12 predictive time windows and the 100-fold cross validation runs .
we see that the distribution is approximately normal with a mean skewed to -2.7 ( as would be expected from the rme plot in fig .
[ fig : size_errors](c ) ) and a standard deviation of 4.9 . as reference ,
the green line in fig .
[ fig : size_error_hist ] is a sample normal distribution with mean -2.7 , standard deviation 4.9 , and the same total counts as the error histogram .
we additionally plot the histograms of the error @xmath56 for each flare class in fig .
[ fig : size_error_hist](b)(e ) , again aggregated over the 12 predictive time windows and 100-fold cross validation runs .
we see that the mean error is different for the different flare classes .
we find an average error of approximately 3/4 a _ goes _ class for the ab- and x - class flares , albeit overestimating the ab flares and underestimating the x flares .
we find a similar error of approximately 1/3 a _ goes _ class for the c- and m - class flares , again overestimating the smaller ( c - class ) flares and underestimating the larger flares .
parameters @xmath16 and @xmath26 are optimized via grid search using the same tuning dataset and procedure as for experiment 1 , except now including non - flaring regions in the dataset and subsampling procedures .
we find a minimum rmse of approximately 7.2 for @xmath57 and @xmath58 which we use for subsequent size regression simulations in experiment 2 .
we note that the rmse resulting from optimization of @xmath16 and @xmath26 already indicates a higher average error in experiment 2 as compared to experiment 1 .
we now use the remaining @xmath55 of the ars unseen in tuning to cross validate performance of the size regression , using the same procedure as in experiment 1 .
we plot the rmse , mae , rme , and correlation coefficient @xmath36 for size regression in fig .
[ fig : size_errors2 ] .
we see an average rmse ( fig .
[ fig : size_errors2](a ) ) ranging between 6.5 and 7.7 , with smaller errors for larger predictive time windows , corresponding to an rmse of approximately three quarters of a class .
similarly , we find an average mae ( fig . [ fig : size_errors2](b ) ) between 4.8 and 5.6 , again with smaller errors for larger predictive time windows and indicating an error of approximately half a class .
looking at the rme in fig .
[ fig : size_errors2](c ) , we see that our estimates of flare size are biased toward overestimating by 3.54.7 mag increments with larger bias for smaller predictive time windows . the correlation coefficient in fig .
[ fig : size_errors2](d ) ranges from 0.25 to 0.55 , with larger correlation for longer predictive time windows . in figure
[ fig : size_errors2 ] we have included the standard deviation across the 100-fold cross validation runs as error bars .
we note that again the smaller time windows have a larger standard deviation due to small sample statistics . in fig .
[ fig : size_error_hist2](a ) we plot a histogram of the error @xmath56 , aggregated over the 12 predictive time windows and the 100-fold cross validation runs . here
, we see a distribution that deviates significantly from normal . as reference , the green line in fig .
[ fig : size_error_hist2](a ) is a sample normal distribution with the sample mean -3.9 , sample standard deviation 5.6 , and the same total counts as the error histogram .
we separately plot the error histogram for non - flaring regions and the different flare classes in fig .
[ fig : size_error_hist2](b)-(f ) .
we see that the distribution of error for non - flaring regions ( fig . [
fig : size_error_hist2](b ) ) is the source of the non - normal distribution .
furthermore , we find that most of the non - flaring regions were overestimated by the regressor , i.e. , we predicted a flare when the _ goes _ catalog does not indicate one . some of this error can be attributed to the _ goes _ background flux as previously discussed .
overall , we again find different average errors for different flare classes , with a tendency to overestimate small flares and underestimate large flares .
next , we look at the flare prediction accuracy by thresholding the regression results . we define a non - flaring ar to be one with a predicted flare size of @xmath59 and flaring ar to be one with a predicted flare size of @xmath60 where @xmath61 is the chosen threshold
. metrics of flare prediction accuracy will be computed from the entries of the confusion matrix shown in table [ tab : confusion_matrix ] . in table
[ tab : confusion_matrix ] tp is the true positive ( correct flare prediction ) , fn false negative ( incorrect no - flare prediction ) , fp false positive ( incorrect flare prediction ) , and tn true negative ( correct no - flare prediction ) .
since flares are rare events , the overall prediction accuracy @xmath62 can be misleading .
as such , we also compute the percentage of correctly classified flaring regions ( the true positive rate or tpr , also known as the sensitivity ) @xcite @xmath63 the percentage of correctly classified non - flaring regions ( the true negative rate or tnr , also known as the specificity ) @xcite @xmath64 the heidke skill score @xcite @xmath65}{(tp+fn)(fn+tn)+(tp+fp)(fp+tn)},\ ] ] and the true skill score ( tss ) @xcite @xmath66 we vary the threshold @xmath67 $ ] in steps of 0.5 .
note that a threshold of @xmath68 corresponds to all flaring regions in this experiment , a threshold of @xmath69 corresponds to all flares @xmath70c1.0 , a threshold of @xmath71 corresponds to all flares @xmath70m1.0 and a threshold of @xmath72 corresponds to all flares @xmath70x1.0 .
we plot the tpr and tnr versus @xmath61 in fig .
[ fig : prediction_accuracy](a ) and the tss and hss versus @xmath61 in fig .
[ fig : prediction_accuracy](b ) . as reference , using the same 38 features , @xcite obtained tpr @xmath73 0.8 , tnr @xmath73 0.7 , tss @xmath73 0.5 , and hss @xmath73 0.4 for flares @xmath70c1.0 . at a comparable threshold @xmath69
, we find performance in this work to be tpr @xmath74 , tnr @xmath75 , tss @xmath76 , and hss @xmath77 .
we thus find a comparable performance with a slightly lower tpr , higher tnr , and higher tss .
this is not surprising in that we are using the same 38 features as in the flare prediction work of @xcite and are instead usng a more general continuous - valued label vector rather than a discrete two - class label vector . in this section , in experiment 3 , we use a label vector of the time - to - flare in hours for flaring regions .
experiment 3 is designed to study the discriminatory potential of the 38 magnetogram complexity measures for determining time - to - flare ; we note that in an operational scenario , the ar could first be classified as flaring by a separate classifier ( e.g. , @xcite ) .
parameters @xmath16 and @xmath26 are optimized via grid search using the same tuning dataset as experiments 1 and 2 but no data balancing as discussed in section [ sec : experiment ] .
instead , we use 500 randomly selected data points ( approximately 10% chosen based on computational considerations ) from the tuning dataset for training and the remaining for testing .
we find a minimum rmse of approximately 39.5 hr for @xmath78 and @xmath79 and use these values for subsequent time regression simulations .
we implement 100-fold cross validation and train over the remaining @xmath55 of the data . for each fold , we randomly choose 1000 data points to serve as the training data and the remainder for the test data .
the choice of 1000 training points is motivated by computational considerations and by the approximate size of the balanced training set in experiments 1 and 2 .
averaged across the 100 cross validation runs , we find an rmse of [email protected] hr , mae of [email protected] hr , rme of [email protected] hr , and @xmath36 of [email protected] .
figure [ fig : time_error_hist ] shows the histogram of the time prediction error .
we see that the distribution is appears to be a double - sided exponential with a more heavy tail on the positive side with a sample mean of @xmath81 and a sample standard deviation of @xmath82 .
the positive value of mae and the heavy right tail on the error distribution indicate a bias toward underestimating time to flare .
overall , we find the time regression problem to be more error prone than the size regression problem .
this is not surprising given the consistency of the flare size regression results across the different time windows .
it appears ( see discussion in section [ sec : experiment1 ] ) that the magnetic complexity features are stable across a large window of time , indicating that these magnetic complexity features may not be particularly discriminative for time - based estimates of flaring .
we have studied the prediction of solar flare size and time - to - flare using 38 features describing magnetic complexity of the photospheric magnetic field from a large set of mdi data .
the 38 magnetic complexity features were previously shown to have discriminative potential for flare prediction in @xcite and were studied here for their discriminative potential in determining the size of a future flare and when a future flare will occur . to the best of our knowledge , these regression problems have not been studied before .
this work uses support vector regression ( svr ) , considered state - of - the - art for time series prediction , to formulate a mapping from the 38-dimensional feature space to a continuous - valued label vector representing flare size or time - to - flare . in the study of size prediction , we have defined a linearly spaced label vector such that an error of 1 arbitrary unit can be interpreted independently of the underlying flare class . in the study of time - to - flare prediction
, we use a label vector with the number of hours to flare . in size regression ,
when we consider flaring regions only , we find an average error in estimating flare size of approximately half a _ goes _ class . when we additionally consider non - flaring regions , we find an increased average error of approximately 3/4 a _ goes _ class . in both of these experiments , we find a different error distribution for different flare classes , with a tendency to overestimate small flares and underestimate large flares .
we also consider thresholding the regressed flare size for the experiment containing both flaring and non - flaring regions and find a true positive rate of 0.69 and a true negative rate of 0.86 for flare prediction .
these results are consistent with previous studies of flare prediction using these magnetic complexity features @xcite , with a slightly lower true positive rate and slightly higher true negative rate . the results for both of these size regression experiments are consistent across a wide range of predictive time windows , indicating that the magnetic complexity features may be persistent in appearance long before flare activity . this conjecture of persistence in appearance across time is supported by our larger error rates of some 40 hr in the time - to - flare regression problem . the 38 magnetic complexity features considered here appear to have discriminative potential for flare size , but their persistence in time makes them less discriminative for the time - to - flare problem . future scientific work will focus on more study of the discriminative potential of different magnetic complexity features for prediction of flare size as well as the study of the persistence of these features across time .
algorithmically , we also plan to implement other regression methods , namely relevance vector regression , to avoid the computational overhead of optimizing kernel parameters .
additionally , relevance vector regression will allow for a study of the underlying confidence in the regressed values .
other methods of dealing with unbalanced data , e.g. , by penalizing different errors differently @xcite , may alleviate the differences in error across the flare classes .
finally , the use of more recent hmi data will allow for a larger and richer dataset for training and testing these regression methods .
abramenko , v. i. 2005 , , 228 , 29 abramenko , v. i. , yurchyshyn , v. b. , wang , h. , spirock , t. j. , & goode , p. r. 2003 , , 597 , 1135 al - ghraibah , a. , boucheron , l. e. , & mcateer , r. t. j. 2015 , , 579 , a64 barnes , g. , & leka , k. d. 2006 , , 646 , 1303 bishop , c. m. 2006 , _ pattern recognition and machine learning _ ( new york , ny : springer ) bloomfield , d. s. , higgins , p. a. , mcateer , r. t. j. , & gallagher , p. t. 2012 , , 747 , l41 brown , m. p. s. , grundy , w. n. , lin , d. , cristianini , n. , sugnet , c. w. , furey , t. s. , ares , jr . , m. , & haussler , d. 2000 , pnas , 97 , 262 burges , c. j. c. , knowledge discovery and data mining , 2 , 121 canu , s. , grandvalet , y. , guigue , v. , & rakotomamonjy , a. 2005 , _ svm and kernel methods matlab toolbox _ , ( perception systemes et information , insa de rouen : rouen , france ) colak , t. , & qahwaji , r. 2009 , space weather , 7 , s06001 conlon , p. a. , gallagher , p. t. , mcateer , r. t. j. , ireland , j. , young , c. a. , kestener , p. , hewett , r. j. , & maguire , k. 2008 , , 248 , 297 cui , y. , li , r. , zhang , l. he , y. , & wang , h. 2006 , , 237 , 45 elattar , e. e. , goulermas , j. , & wu , q. h. 2010 , ieee transactions on systems , man , and cybernetics
part c , 40 , 438 falconer , d. , barghouty a. f. , khazanov , i. , & moore , r. 2011 , space weather , 9 , s04003 fletcher , l. , dennis , b. r. , hudson , h. s. , krucker , s. , phillips , k. , veronig , a. , battaglia , m. , bone , l. , caspi , a. , chen , q. , gallagher , p. , grigis , p. t. , ji , h. , liu , w. , milligan , r. o. , & temmer , m. 2011 , , 159 , 19 georgoulis , m. k. , & rust , d. m. , 2007 , , 661 , l109 guo , j. , zhang , h. , chumak , o. v. , & liu , y. 2006 , , 237 , 25 hagyard , m. j. , stark , b. a. , & venkatakrishnan , p. 1999
, , 184 , 133 hewett , r. j. , gallagher , p. t. , mcateer , r. t. j. , young , c. a. , ireland , j. , conlon , p. a. , & maguire , k. 2008 , , 248 , 311 jing , j. , song , h. , abramenko , v. , tan , c. , & wang , h. 2006 , , 644 , 1273 jing , j. , tan , c. , yuan , y. , wang , b. , wiegelmann , t. , xu , y. , & wang , h. 2010 , , 713 , 440 koskinen , h. , tanskanen , e. , pirjola , r. , pulkkinen , a. , dyer , c. , rodgers , d. , cannon , p. , mandeville , j .- c .
, & boscher , d. 2001 , _ esa tech report esws - rmi - rp-0001 _ , finnish meteorological institute leka , k. d. , & barnes , g. 2003a , , 595 , 1277 li , d. , mersereau , r. m. , & simske , s. 2007 , ieee transactions on neural networks , 18 , 931 mason , j. p. , & hoeksema , j. t. 2010 , , 723 , 634 mcateer , r. t. j. , gallagher , p. t. , & ireland , j. 2005 , , 631 , 628 mcateer , r. t. j. , gallagher , p. t. , & conlon , p.a . , 2010 ,
asr , 45 , 1067 mcateer , r. t. j. , 2015 , , 290 , 7,1897 mcateer , r. t. j. , aschwanden , m.j . ,
dimitropoulou , m. , georgoulis , m.k . , pruessner , g. , morales , l. , ireland , j. , abramenko , v. , 2015 , space sciences review , 10.1007/s11214 - 015 - 0158 - 7 meunier , n. 2004 , , 420 , 333 mohandes , m. a. , halawani , t. o. , rehman , s. , & hussain , a. a. ( 2004 ) renewable energy , 29 , 939 qahwaji , r. , & colak , t. 2007 , , 241 , 195 sammis , i. , tang , f. , & zirin , h. 2000 , , 540 , 583 sapankevych , n. i. , & sankar , r. 2009 , ieee computational intelligence magazine , 4 , 24 schrijver , c. j. 2007 , , 655 , 117 schrijver , c. j. 2014 , space weather , 12 , 487 smola , a. j. , & schlkopf , b. 2004 , statistics and computing , 14 , 199 song , h. , tan , c. , jing , j. , wang , h. , yurchyshyn , v. , & abramenko , v. 2014 , , 254 , 101 tay , f. e. h. , & cao , l. 2001 , omega , 29 , 309 vapnik , v. n. 1998 , _ statisical learning theory _
( new york : wiley ) wang , h. 2006 , , 649 , 490 wang , w. , xu , z. , & lu , j. w. 2003 , engineering computations , 20 , 192 welsch , b. t. , li , y. , schuck , p. w. , & fisher , g. h. 2009 , , 705 , 821 wheatland , m. s. 2000 , , 532 , 1209 wheatland , m. s. 2005 , space weather , 3 , s07003 wu , c .- h .
, ho , j .- m . , & lee , d. t. 2004 , ieee transactions on intelligent transportation systems , 5 , 276 wu , c. l. , chau k. w. , & li , y. s. 2008 , journal of hydrology , 358 , 96 yu , p .- s . , chen , s .-
, & chang , i .-
2006 , journal of hydrology , 328 , 704 lll * gradient features * & * flux evolution features * + gradient mean & fe sum + gradient std & fe absolute sum + gradient median & fe gradient sum + gradient min & fe 3@xmath26 area + gradient max & fe mean + gradient skewness & fe std + gradient kurtosis & fe median + * neutral line features * & fe min + nl length & fe max + nl no . fragments & * wavelet features * + nl gradient - weighted length & wavelet energy level 1 + nl curvature mean & wavelet energy level 2 + nl curvature std & wavelet energy level 3 + nl curvature median & wavelet energy level 4 + nl curvature min & wavelet energy level 5 + nl curvature max & * flux features * + nl bending energy mean & total unsigned flux + nl bending energy std & total signed flux + nl bending energy median & total negative flux + nl bending energy min & total positive flux + nl bending energy max & lll class & peak flux ( w m@xmath83 ) & regression target value + a & @xmath84 & 0.11 + b & @xmath85@xmath86 & 19 + c & @xmath86@xmath40 & 1018 + m & @xmath40@xmath87 & 1927 + x & @xmath88 & 2836 + | we study the prediction of solar flare size and time - to - flare using 38 features describing magnetic complexity of the photospheric magnetic field . this work uses support vector regression to formulate a mapping from the 38-dimensional feature space to a continuous - valued label vector representing flare size or time - to - flare . when we consider flaring regions only , we find an average error in estimating flare size of approximately half a _ geostationary operational environmental satellite _ ( _ goes _ ) class .
when we additionally consider non - flaring regions , we find an increased average error of approximately 3/4 a _ goes _ class .
we also consider thresholding the regressed flare size for the experiment containing both flaring and non - flaring regions and find a true positive rate of 0.69 and a true negative rate of 0.86 for flare prediction .
the results for both of these size regression experiments are consistent across a wide range of predictive time windows , indicating that the magnetic complexity features may be persistent in appearance long before flare activity .
this is supported by our larger error rates of some 40 hr in the time - to - flare regression problem .
the 38 magnetic complexity features considered here appear to have discriminative potential for flare size , but their persistence in time makes them less discriminative for the time - to - flare problem . |
the @xmath0 noise is a random process described by the power spectral density ( psd ) , @xmath1 , roughly proportional to the reciprocal frequency , @xmath0 , i.e. , @xmath2 , with @xmath3 close to @xmath4 .
it was observed first as an excess low - frequency noise in vacuum tubes @xcite , later in condensed matter @xcite and other systems @xcite . the general nature of @xmath0 noise ( named also `` flicker noise '' and `` @xmath0 fluctuations '' ) is up to now the subject of several discussions and investigations , see @xcite for review .
many models have been proposed to explain the origin of @xmath0 noise .
a short discussion about the models and theories of @xmath0 noise is available in the introduction of paper @xcite .
widely used model of @xmath0 noise interprets the spectrum as a superposition of lorentzians with a wide range distribution of relaxation times @xcite . another possibility to model signals and processes featuring @xmath5 noise
is a representation of the signals as consisting of the renewal pulses or events with the power - law distribution of the inter - event time @xcite .
a class of models of @xmath0 noise relevant for driven nonequilibrium systems involves the self - organized criticality ( soc ) @xcite .
soc refers to the tendency of nonequilibrium systems driven by slow constant energy input to organize themselves into a correlated state where all scales are relevant @xcite . in @xcite a simple driven automaton model of sandpiles that reaches a state characterized by power - law time and
space correlations has been introduced . however , the mechanism of self - organized criticality not necessarily results in @xmath6 fluctuations with @xmath3 close to @xmath4 @xcite . the @xmath0 noise in the fluctuations of a mass
was first seen in a sandpile model with threshold dissipation , proposed in @xcite .
in addition , the exponent @xmath3 is exactly @xmath4 in the spectrum of fluctuations of mass in a one - dimensional directed model of sandpiles @xcite . in most cases
the @xmath0 noise is a gaussian process @xcite , however sometimes @xmath0 fluctuations are non - gaussian @xcite .
processes with the power - law distributions of the signal characteristics can be modeled by presuming that the time between the adjacent pulses experience slow ( the change from one inter - pulse duration to the next much smaller than the duration itself ) brownian - like motion @xcite .
moreover , the nonlinear stochastic differential equations ( sdes ) generating @xmath5 noise have been obtained and analyzed @xcite starting from this point process model .
sde generating @xmath0 noise should necessarily be nonlinear , because systems of linear sdes do not generate signals with @xmath0 spectrum .
such nonlinear sdes have been applied to describe signals in socio - economical systems @xcite . in the signal consisting of a sequence of pulses
the pulse number is a progressively increasing quantity and it can be understood as an internal time of the process . the purpose of this paper is to investigate the distinction between the internal time of the system and the physical time in connection with @xmath0 noise .
we intend to generalize the mechanism leading to @xmath0 noise in the point process model , proposed in @xcite . instead of a sequence of pulses we start from an sde describing a brownian - like motion .
we compose a new equation by interpreting the time in the sde as an internal parameter and adding an additional equation relating the internal time to the physical time .
we demonstrate that the relation between the internal time and the external time , depending on the intensity of the signal , can lead to @xmath0 noise in a wide interval of frequencies .
a process @xmath7 obtained by randomizing the time clock of a random process @xmath8 using a new clock @xmath9 , where @xmath9 is a random process with non - negative increments , is called the subordinated process @xcite .
the process @xmath9 is referred to as directing process , randomized time or operational time . in physics
the time - subordinated equations have been applied to describe anomalous diffusion .
fogedby @xcite introduced a class of coupled langevin equations consisting of a langevin process @xmath10 in a coordinate @xmath11 and a lvy process representing a stochastic relation @xmath12 .
this class of coupled langevin equations has been further investigated in @xcite , where @xmath13-time joint probability distributions have been analyzed .
properties of the inverse @xmath14-stable subordinator have been considered in @xcite .
it has been shown @xcite that the description of anomalous diffusion by a markovian dynamics governed by an ordinary langevin equation but proceeding in an auxiliary , operational time instead of the physical time is equivalent to a fractional fokker - planck equation .
numerical simulation of subordinated equations has been explored in @xcite .
in contrast to the description of the anomalous diffusion , in this paper we consider the situation when small increments of the physical time are proportional to the increments of the operational time , with the the coefficient of proportionality that depends on the stochastic variable @xmath15 representing the signal intensity .
thus , in our case the randomness of the operational time comes from the randomness of @xmath15 .
the paper is organized as follows : in we briefly present the point process model of @xmath0 noise and obtain the psd of the signal by a new method . in we generalize the mechanism leading to @xmath0 noise presented in .
we introduce the difference between the physical and the internal time and consider time - subordinated langevin equations . in
we examine several stochastic processes and , introducing the internal and external times , we check whether @xmath0 noise can be obtained . in
we discuss a way of solving highly non - linear sdes by introducing suitably chosen internal time and the variable step of integration .
summarizes our findings .
one of the models of @xmath0 noise has been presented in @xcite . in this model
a signal consist of pulses with the time between adjacent pulses undergoing a brownian - like motion .
it has been shown that this brownian - like motion of the inter - pulse durations can yield @xmath0 noise . in this section
we briefly present this model and obtain the psd of the signal using a different method than the method used in @xcite .
the new method allows us better estimate the frequency range where the psd has @xmath0 behavior .
let us consider a signal consisting of a pulse sequence having correlated inter - pulse durations .
we assume that : ( i ) the pulse sequences are stationary and ergodic ; ( ii ) all pulses are described by the same shape function @xmath16 .
the general form of the signal can be written as @xmath17 where the functions @xmath16 determine the shape of the individual pulse and time moments @xmath18 determine when the pulse occurs .
inter - pulse duration is @xmath19 .
such a pulse sequence is schematically shown in .
.,scaledwidth=60.0% ] the psd of such a signal is given by the equation @xmath20 where @xmath21 is the observation time and the brackets @xmath22 denote the averaging over realizations of the pulse sequence .
note that in equation we consider one - sided psd , thus we have multiplier @xmath23 in it . introducing the fourier
transform @xmath24 of the pulse shape function @xmath16 , we can write equation as @xmath25 here @xmath26 .
if the pulses are narrow and we are considering low frequencies then the fourier transform @xmath24 of the pulse shape is almost constant . in this case
we can replace the actual pulses with @xmath27-functions and drop @xmath24 in the equations .
the psd can be decomposed into two parts , @xmath28 the first term can be written as @xmath29 where @xmath30 is the mean number of pulses per unit time . by changing @xmath31 into @xmath32
in the second part of the psd one sees that it can be expressed as @xmath33 where the time difference @xmath34 is @xmath35 thus , equation becomes @xmath36 assuming that the joint probability @xmath37 exist we can write the average in the above equation as @xmath38 if the inter - pulse durations follow the markov process then the conditional probabilities depend only on the previous value of the inter - pulse duration , @xmath39 . in this case
@xmath40 let us consider a situation when the probability density function ( pdf ) of inter - pulse durations @xmath41 is significant only for @xmath42 in some range @xmath43 and is very small for @xmath42 outside this range .
in addition , we will assume that the conditional probability @xmath44 has the following properties : the average equal to the previous value of inter - pulse duration @xmath45 and the dispersion @xmath46 is much smaller than the dispersion of inter - pulse durations @xmath47 these assumptions denote that the average difference between the neighboring inter - pulse durations is small , i.e. , the increments and decrements of the inter - event duration are small in comparison to the inter - event time it self . when @xmath48 then the dispersion of inter - pulse durations is @xmath49 .
thus , we assume that @xmath50 . when the assumptions and @xmath51 hold , we can approximate the conditional probability @xmath44 by a @xmath27-function : @xmath52 .
the approximation in equation is valid only for sufficient small @xmath53 , smaller than some maximum value @xmath54 , because the error grows with the number of terms . using in equation the approximation of the conditional probability by @xmath27-function
we obtain @xmath55 where @xmath56 is the characteristic function of inter - pulse durations .
we can estimate the value of @xmath54 as follows : the approximation of the conditional probability @xmath44 by @xmath27-function is not applicable when the dispersion of @xmath57 for a given @xmath58 becomes comparable with the dispersion @xmath59 .
assuming that the dispersion of @xmath60 , for a given @xmath58 , grows linearly with @xmath61 ( as would be the case for a random walk ) we require that @xmath62 and , therefore , @xmath63 for high enough frequency , when @xmath64 the characteristic functions @xmath65 corresponding to large @xmath66 are small and we can neglect in equation the terms with @xmath67 . including only the terms with @xmath68 we get the expression for the psd @xmath69 after the summation in equation we obtain @xmath70 we have dropped @xmath71 in @xmath72 because @xmath54 is large , @xmath73 .
in addition , for small frequencies @xmath74 we approximated @xmath75 in the denominator as @xmath76 .
the function @xmath77 has a sharp peak of the width @xmath78 at @xmath79 and decreases at larger @xmath80 . if @xmath81 then this peak is much narrower that the width of the pdf @xmath82 .
in addition , the peak of the function @xmath77 has a significant overlap with @xmath82 when @xmath83 . in this case
we obtain the following approximate expression for the psd : @xmath84 this equation shows that we get @xmath0 spectrum .
summing up the assumptions made above , the range of the frequencies where this expression for psd holds is @xmath85 when @xmath86 the upper limit of the frequency range is determined by @xmath87 . in this case
the ratio of upper and lower limiting frequencies is @xmath88 . for larger @xmath89
the ratio of upper and lower limiting frequencies is @xmath90 .
the parameters used are @xmath91 , @xmath92 , @xmath93.,scaledwidth=60.0% ] as an example , let us consider the point process where the inter - pulse durations perform a random walk and are related via the equation @xmath94 here each sign occurs with probability @xmath71 .
in addition , we have reflections from the minimum inter - pulse duration @xmath91 and from the maximum inter - pulse duration @xmath87 .
numerically obtained psd of such a signal is shown in .
we see a power - law part in the psd with the slope @xmath95 in a broad range of frequencies from @xmath96 to @xmath97 .
this range of frequencies agrees with the estimation .
psd of the power - law form with the exponents different from @xmath95 can be obtained by including in equation an additional drift term . in @xcite
it has been shown that the drift term of the power - law form @xmath98 and power - law pdf of the inter - pulse duration @xmath99 lead to the power - law psd @xmath100 with @xmath101 . as a process generating the power - law probability distribution function for @xmath60 a multiplicative stochastic process @xmath102 has been suggested . here
@xmath103 are normally distributed uncorrelated random variables with a zero expectation and unit variance . for this process @xmath104 and @xmath105 . has been used for modeling of the internote interval sequences of the musical rhythms @xcite .
in this section we generalize the model presented in the previous section .
we do this by noticing that in the pulse sequence there are two strictly increasing sequences of numbers : the physical time @xmath106 and the pulse number @xmath31 .
the pulse number can be interpreted as an internal time of the pulse sequence .
the relation between the physical time and the internal time is not deterministic because the inter - pulse durations are random .
thus we propose the introduction of the difference between the physical and the internal , operational , time as a way to obtain @xmath0 noise also for other stochastic processes . to do this we start with a stochastic process and interpret the time as an internal parameter .
in addition to this stochastic process we need to include an additional relation between the physical time and the internal time . in order to maintain a similarity to the point process described in the previous section , the increments of the physical time
should be a power - law function of the magnitude of the signal . in this section as an initial stochastic process we take a process described by a stochastic differential equation .
langevin equation coupled to an additional equation for the physical time have been introduced to describe anomalous diffusion @xcite .
in particular , position - dependent time subordinator has been investigated in @xcite .
let us start with the langevin equation describing the diffusion of the particle subjected to an external force @xmath107 here @xmath108 and @xmath109 are drift and diffusion coefficients and @xmath110 is a standard wiener process .
for generality we assume that both coefficients @xmath111 and @xmath112 can depend on the stochastic variable @xmath15 . in case when the diffusion coefficient @xmath112 in equation depends on @xmath15 we assume it interpretation . in equation we replace the physical time @xmath106 by the operational time @xmath113 , @xmath114 the pdf @xmath115 of the stochastic variable @xmath15 as a function of the operational time @xmath113 obeys the fokker - planck equation corresponding to it sde @xcite @xmath116 primarily we consider the situation when the small increments of the physical time are deterministic and are proportional to the increments of the operational time .
thus , the physical time @xmath106 is related to the operational time @xmath113 via the equation @xmath117 here the positive function @xmath118 is the intensity of random time that depends on the intensity of the signal @xmath15 . if we interpret equation as describing the diffusion of a particle in non - homogeneous medium , the function @xmath118 models the position of structures responsible for either trapping or accelerating the particle @xcite .
large values of @xmath118 corresponds to trapping of the particle , whereas small @xmath118 leads to the acceleration of diffusion . for fixed particle position @xmath15
the coefficient @xmath118 in equation is constant and from equation follows the relationship @xmath119 for the pdf @xmath120 of the physical time @xmath106 as a function of the operational time @xmath113 .
equations and together define the subordinated process . however , now the processes @xmath121 and @xmath122 are not independent .
let us derive the langevin equation for the stochastic variable @xmath15 in the physical time @xmath106 . to do this
, we consider the joint pdf @xmath123 of the stochastic variables @xmath15 and @xmath106 . equations and yield the two - dimensional fokker - planck equation
@xmath124 this equation is a combination of equations and .
the zero of the physical time @xmath106 coincides with the zero of the operational time @xmath113 , therefore , the initial condition for equation is @xmath125 . coinciding zeros of @xmath106 and @xmath113 lead also to the boundary condition @xmath126 for @xmath127 , because @xmath106 and @xmath113 are strictly increasing . instead of @xmath15 and @xmath106
we can consider @xmath15 and @xmath113 as stochastic variables .
the stochastic variable @xmath106 is related to the operational time @xmath113 via equation , therefore , the joint pdf @xmath128 of the stochastic variables @xmath15 and @xmath113 is related to the pdf @xmath123 according to the equation @xmath129 this equation can be obtained by noticing that the last term in equation contains derivative @xmath130 and thus should be equal to @xmath131 . using equations and we get @xmath132 the pdf @xmath133 has the initial condition @xmath134 and the boundary condition @xmath135 for @xmath136 .
the pdf of the subordinated random process @xmath137 is @xmath138 . integrating both sides of equation we obtain @xmath139 thus , position - dependent trapping leads to the position - dependent coefficients in the fokker - planck equation , even if the initial sde has constant coefficients .
corresponds to the single equation in the physical time with the multiplicative noise , @xmath140 in fact , the fokker - planck equation can be obtained from the coupled equations and @xmath141 the relationship between the physical time @xmath106 and the operational time @xmath113 can be not necessarily deterministic , equation can have a stochastic term . if the fluctuations of this stochastic term are much faster than the fluctuations of the stochastic variable @xmath15 , we can approximate them by the average value . in this case @xmath118
describes the average increment of the physical time .
if this average is positive , the derivation presented above is still valid and equation holds .
in this section we consider several stochastic processes and , introducing the internal and external times , we check whether @xmath0 noise can be obtained . in a signal consisting of pulses
the internal time is just the pulse number and the increment of the physical time is equal to the inter - pulse duration .
the intensity of such a signal is inversely proportional to the inter - pulse duration . in order to obtain @xmath0 noise similarly as for the signal consisting of pulses we choose the function @xmath118 in equation as a power - law function of @xmath15 , @xmath142 , where @xmath143 is the power - law exponent .
let us start from a simple brownian motion @xmath144 in order to keep the stochastic variable @xmath15 always positive we include reflective boundary at @xmath145 .
we consider equation together with the relation @xmath146 between the physical time @xmath106 and internal time @xmath113 .
according to the resulting equation in the physical time is @xmath147 more generally , the initial equation can include a position - dependent force . if we take the equation describing the bessel process @xmath148 together with equation ,
then the resulting equation in the physical time becomes @xmath149 here the parameter @xmath150 gives the power - law exponent of the steady - state pdf .
the same equation in physical time arises starting from the geometric brownian motion , @xmath151 and the relation between the internal time and the physical time @xmath152 nonlinear sde for generating signals with @xmath6 spectrum has been proposed in @xcite . as has been shown in @xcite ,
the reason for the appearance of @xmath0 spectrum is the scaling properties of the signal : the change of the magnitude of the variable @xmath153 is equivalent to the change of the time scale @xmath154 .
connection of the power - law exponent @xmath3 in the psd with the parameters of equation is given by the equation @xcite @xmath155 analysis @xcite of sde shows that equation is valid only for the values of the parameters @xmath143 and @xmath150 yielding @xmath156 .
nonlinear sde leads to the stationary process and non - diverging steady state pdf only when the diffusion of stochastic variable @xmath15 is restricted .
the simplest choice of the restriction is the reflective boundary conditions at @xmath157 and @xmath158 .
the presence of the restrictions of diffusion makes the scaling properties of equation only approximate and limits the power - law part of the psd to a finite range of frequencies .
this range of frequencies has been qualitatively estimated as @xcite @xmath159 by increasing the ratio @xmath160 one can get an arbitrarily wide range of the frequencies where the psd has @xmath6 behavior . and @xmath161 ( solid red line ) together with the corresponding internal time ( dashed green line ) .
reflective boundaries at @xmath162 and @xmath163 have been used .
( b ) psd of the generate signal . dashed green line shows the slope @xmath0.,title="fig:",scaledwidth=45.0% ] and @xmath161 ( solid red line ) together with the corresponding internal time ( dashed green line ) .
reflective boundaries at @xmath162 and @xmath163 have been used .
( b ) psd of the generate signal . dashed green line shows the slope @xmath0.,title="fig:",scaledwidth=45.0% ] an example of a signal generated by equation together with the internal time @xmath113 is shown in ( a ) .
we used the parameters @xmath164 , @xmath161 and reflective boundaries at @xmath162 and @xmath163 .
the method of numerical solution is discussed in the next section .
we see that internal time @xmath113 increases rapidly when the signal @xmath15 acquires large values and @xmath113 changes slowly when @xmath15 is small . according to equation
the choice of @xmath161 should result in @xmath0 behavior of the psd .
the corresponding power spectral density @xmath1 is shown in ( b ) .
the numerical solution of the equation confirms a presence of a wide region of frequencies where the spectrum has @xmath0 behavior .
when the stochastic variable @xmath15 can acquire both positive and negative values , the function @xmath118 can not be just a simple power - law , because @xmath118 becomes unbounded or equal to zero when @xmath165 . in order to avoid this problem
we require that function @xmath118 should have power - law behavior only asymptotically , for large vales of @xmath166
. one of the possible choices is @xmath167 here we added a constant @xmath168 that corrects the behavior of the function @xmath118 at @xmath169 .
the power - law behavior is preserved when @xmath170 .
the stochastic variable @xmath15 can acquire both positive and negative values if we start from the ornstein - uhlenbeck process @xmath171 here the parameter @xmath172 is the relaxation rate .
we consider equation together with the relation @xmath173 between the physical time @xmath106 and internal time @xmath113 . according to , equations and leads to sde @xmath174 in the physical time @xmath106
can be written as @xmath175 where @xmath176 defines a cut - off position at large values of @xmath15 .
another interesting equation describing the evolution in internal time is @xmath177 in this equation the relaxation rate depends on the magnitude of the signal . if @xmath178 we get the equation of ornstein - uhlenbeck type , whereas for large values of @xmath166 the relaxation decreases with increasing @xmath166 .
together with result in the following equation in the physical time : @xmath179 finally , the combination of equations and , @xmath180 together with leads to a more general equation in the physical time @xmath181 nonlinear sde has been investigated in @xcite .
it has been shown that sde generates a signal with the steady - state pdf described by the @xmath53-gaussian distribution featuring in the non - extensive statistical mechanics .
in addition , the spectrum of the generated signal has @xmath6 behavior in a wide range of frequencies , with the power - law exponent @xmath3 given by equation . ,
@xmath161 and @xmath182 ( solid red line ) together with the corresponding internal time ( dashed green line ) .
( b ) psd of the generate signal . dashed green line shows the slope @xmath0.,title="fig:",scaledwidth=45.0% ] , @xmath161 and @xmath182 ( solid red line ) together with the corresponding internal time ( dashed green line ) .
( b ) psd of the generate signal . dashed green line shows the slope @xmath0.,title="fig:",scaledwidth=45.0% ] an example of a signal generated by equation together with the internal time is shown in ( a ) .
we used the parameters @xmath164 , @xmath161 and @xmath182 .
we see that the internal time @xmath113 increases rapidly when the absolute value of the signal @xmath15 is large and @xmath113 changes slowly when the absolute value of @xmath15 is small .
the internal time @xmath113 increases both for positive and negative values of @xmath15 .
the psd of a signal generated by equation is shown in ( b ) .
the numerical solution confirms a presence of a region where the spectrum behaves as @xmath0 .
thus the introduction of negative values of @xmath15 does not destroy @xmath0 spectrum .
introduction of the internal time can be an effective technique for solution of highly non - linear sdes . for numerical solution of nonlinear equations
the solution schemes involving a fixed time step @xmath183 can be inefficient .
for example , in equation with @xmath184 large values of stochastic variable @xmath15 lead to large coefficients and thus require a very small time step .
the numerical solution scheme can by improved by introducing the internal time @xmath113 that is different from the real , physical , time @xmath106 .
let us consider equation with the noise multiplicativity exponent @xmath184 .
we can introduce internal time @xmath113 using the equation @xmath185 then , according to equations and , sde is equivalent to coupled equations @xmath186 now , equation is much simpler than the initial equation .
discretizing the internal time @xmath113 with the step @xmath187 and using the euler - marujama approximation for the sde we get @xmath188 here @xmath103 are normally distributed uncorrelated random variables . equations and provide the numerical method for solving equation .
one can interpret equations , as an euler - marujama scheme with a variable time step @xmath189 that adapts to the coefficients in the equation .
the cost of the introduction of the internal time is the randomness of the increments of the real , physical time @xmath106 . to get the discretization of time with fixed steps the signal generated in such a way
should be interpolated .
another possible choice is to introduce the internal time @xmath113 by the equation @xmath190 in this case we obtain a different pair of equations @xmath191 note , that now the internal time @xmath113 is dimensionless even if @xmath15 and @xmath106 are not . discretizing the internal time @xmath113 with the step @xmath187 and using the euler - marujama approximation for the sde we obtain @xmath192 this method of solution has been proposed in @xcite . on the other hand , using milstein approximation for the sde we have @xmath193 note , that the last term in equation differs from the corresponding term in the equation obtained just by using a variable time step @xmath194 in the milstein approximation for equation .
numerical simulation of subordinated equations using fixed step of operational time and random increment of physical time has been discussed in @xcite .
variable time step makes numerical simulation in @xcite similar to the method proposed in this section .
the main difference of our method from previous discussions of subordinated equations lies in the depence of the increment of the physical time on the magnitude of the signal @xmath15 .
in summary , we have demonstrated that starting from a random process described by a sde and introducing the difference between the internal time and the physical time @xmath0 behavior of the psd can be obtained .
one of the physical situations where the difference between the internal and physical time can arise is transport in an inhomogeneous medium . impurities and regular structures in the medium can cause transport of variable speed , the particle may be trapped for some time or accelerated .
nonhomogeneous systems exhibit not only subdiffusion related to traps , but also enhanced diffusion as a result of the disorder . for example , movement of particles between two neighboring lattice sites in an interacting particle system is superdiffusive due to the disorder and subdiffusive without the disorder @xcite .
the dynamics in a medium with traps is described by the continuous time random walk theory ( ctrw ) @xcite . in a description equivalent to ctrw the dynamics of
the particle is markovian and governed by the langevin equation in an auxiliary , operational , time instead of the physical time .
this markovian process is subordinated to the process yielding the physical time . in the case of subdiffusion
the psd of the signals generated by subordinated langevin equations has power - law behavior @xmath195 as @xmath196 @xcite , where @xmath14 is the power - law exponent in the time dependence of the mean square displacement . since for subdiffusion @xmath197 , the power - law exponent @xmath3 in the psd is smaller than @xmath4 .
the results obtained in this paper suggest that @xmath0 noise in subdiffusion should occur in heterogeneous medium , where trapping time depends on the position @xcite .
the traditional ctrw provides a homogeneous description of the medium .
more complex situation is the diffusion in nonhomogeneous media , for example , diffusion on fractals and multifractals @xcite .
heterogeneous medium with steep gradients of the diffusivity can be created via a local variation of the temperature in thermophoresis experiments @xcite .
spatial heterogeneities are also present in the case of anomalous diffusion in subsurface hydrology @xcite . in the random walk description
spatially varying diffusivity can be translated into a local dependence of the waiting time for a jump event . in the heterogeneous medium
the properties of a trap can reflect the medium structure , thus in the description of transport the waiting time should explicitly depend on the position of the particle @xcite .
a method to include position dependent waiting time is a consideration of the position - dependent time subordinator @xcite . in general , the trapping time can depend not on the position of the particle but on some other quantity .
then in the dynamics of this quantity the difference between the physical and operational time also arises , with the relationship between the times dependent on the intensity of the signal . in socio - economical systems the internal time can reflect fluctuating human activity @xcite .
for example , in finance the long - range correlations in volatility arise due to fluctuations of the trading activity @xcite .
we have shown that @xmath0 noise occurs when the internal time and the physical time are related via the power - law function of the signal intensity , for example , via equation or .
although we have considered only random processes described by a sde , we expect that the mechanism of the appearance of @xmath0 noise presented here is quite general and should work also for other random processes .
we anticipate that the present model can be useful for explaining @xmath0 noise in different complex systems .
in addition , we suggested a way of solving highly non - linear sdes by introducing suitably chosen internal time and variable step of integration .
10 url # 1#1urlprefix[2][]#2 johnson j b 1925 _ phys . rev . _ * 26 * 71 | internal mechanism leading to the emergence of the widely occurring @xmath0 noise still remains an open issue . in this paper
we investigate the distinction between internal time of the system and the physical time as a source of @xmath0 noise . after demonstrating the appearance of @xmath0 noise in the earlier proposed point process model , we generalize it starting from a stochastic differential equation which describes a brownian - like motion in the internal ( operational ) time .
we consider this equation together with an additional equation relating the internal time to the external ( physical ) time .
we show that the relation between the internal time and the physical time that depends on the intensity of the signal can lead to @xmath0 noise in a wide interval of frequencies .
the present model can be useful for the explanation of the appearance of @xmath0 noise in different systems .
_ keywords _ : stochastic processes ( theory ) , stochastic processes , current fluctuations , driven diffusive systems ( theory ) |
the study of entanglement , which is one of the most fundamental aspects of quantum mechanics , has lead to and is still leading to much important progress and applications in different fields of modern physics such as quantum information@xcite , condensed matter physics@xcite , etc .. to name a few , it has lead to better undertanding of density matrix renormalization group ( dmrg)@xcite ; it has also been proposed to be a tool for the characterization of certain topological phases@xcite . among various ways of quantifying entanglement , in condensed matter or many - body physics efforts
have mainly focused on the bipartite block entanglement entropy ( von neumann entropy ) and its generalizations ( rnyi or tsallis entropy ) .
it has become increasingly useful in characterizing phases@xcite and phase transitions@xcite .
the area law@xcite is one of the most important results on entanglement entropy : it states that the entanglement entropy is proportional to the area of the surface separating two subsystems .
however , thus far there are two important classes of systems that violate the area law : in gapless one dimensional ( 1d ) systems , a logarithmic divergence@xcite is found where according to the area law the entanglement entropy should saturate as the size of the subsystem grows ; in higher dimensions , for free fermions the area law is found to be corrected by a similar logarithmic factor @xmath1@xcite , where @xmath2 is the linear dimension of the subsystem . in this work ,
we first show that the scaling behavior of the entanglement entropy for systems with a fermi surface is the same as that of 1d systems with fermi points@xcite .
we then seek for a generalization of the latter to interacting fermions in the fermi liquid phase .
we first develop an intuitive understanding via a toy model , showing that in this model the entanglement entropy has the same form as that given in by gioev and klich ( gk ) in ref .
we then develop a more general and formal treatment using the method of high - dimensional bosonization@xcite .
this approach will not only lead to a reproduction of the result for free fermions obtained by gk based on widom s conjecture@xcite , but will also lend itself to the inclusion and subsequent treatment of fermi liquid type ( forward scattering ) interactions .
this paper is organized as follows . in sec .
[ sec : toy_model ] , we describe the toy model for which the entanglement entropy can be written in the same form as the gk result . then in sec .
[ sec : bosonization ] , we briefly introduce the tool box of multi - dimensional bosonization , and apply it to free fermions to reproduce the gk formula .
the main results of this work are presented in sec .
[ sec : ee ] in which we calculate the entanglement entropy of a fermi liquid for a special geometry using a combination of multi - dimensional bosonization and the replica trick .
we subsequently summarize and discuss our results .
some technical details are discussed in two appendices .
consider a set of decoupled parallel 1d chains of non - interacting spinless fermions with spacing @xmath3 as shown in fig .
[ fig : realspace ] . here
we only consider @xmath4 for simplicity , but this toy model is viable in general @xmath5 dimensions .
the asymptotic behavior of entanglement entropy in large @xmath2 limit of a convex subsystem @xmath6 of this model can be obtained by simply counting the number of chains that intersect @xmath6 , and each segment contributes a @xmath7 where @xmath2 is the linear dimension of the subsystem@xcite . due to the logarithm , different shapes only lead to differences at the area law level . since each segment must have two intersections , we can count the intersections instead , which also automatically takes care of non - convex geometries .
although there is an additional correction for multiple intervals on a single chain@xcite , as long as only the @xmath1 behavior is concerned , that contribution is negligible . for @xmath2 large enough ,
we can write the number of these intersections as an integral over the surface of @xmath6 projected onto the direction perpendicular to the chains times one half of the chain density , @xmath8 . to make contact with the gk result , we note this model also has fermi surfaces as shown in fig .
[ fig : fermisurface ] with a total `` area '' of @xmath9 .
this enables us to replace the density of chains by an integral over the fermi surfaces of the system @xmath10 where @xmath11 indicates the occupied area in momentum space so its boundary @xmath12 is the fermi surface(s ) .
therefore we can write the entanglement entropy as [ eq : toy_model ] s(_a ) & = l _ a ds_x + & = l _ a _ _ x ds_k , where @xmath13 is the direction along the chains which is also normal to the fermi surface , and an overall factor of @xmath14 accounts for the double counting of chain segments . in eq .
( [ eq : toy_model ] ) we recover the gk formula in this special case but written in a slightly different way . in ref .
[ ] the entanglement entropy is given as : s = _ a _ _ _ ds _ ds _ , where the real space surface integral is carried out over the subsystem whose volume is normalized to @xmath15 .
the surface area is factored out as @xmath16 . however , in our formula the surface area @xmath17 is implicitly included in the integral over the surface of the subsystem .
we note that the model discussed in ref.@xcite is equivalent with our toy model , but motivated from a different perspective . in ref.@xcite , models are constructed from the momentum space , either with fermi surfaces as our toy model , or a square fermi surface , and a boxlike and a spherical geometry are discussed .
in contrast , our toy model is constructed from a real space perspective , and general single connected geometries are discussed . motivated by the toy model ,
in this work we extend this intuitive understanding of gk s result to generic free fermi systems and generalize it to include fermi liquid interactions in two dimensions ( 2d ) via high dimensional bosonization . using the method of multi - dimensional bosonization
, the fermi liquid theory can be written as a tensor product of low - energy effective theories of quasi-1d systems similar to this toy model , along all directions .
this provides us with a tool to treat the entanglement entropy of fermions in high dimensions , even in the presence of _
interactions_. at this point , we could also include forward scattering for each chain , and from 1d bosonization we know that for spinless fermions this only leads to renormalization of the fermi velocity , thus does not change the logarithmic scaling of the entanglement entropy for this toy model .
this hints that the same conclusion might hold for fermi liquids , as we can include fermi liquid interactions in a similar way via high dimensional bosonization .
although as we show later , this is indeed true at the leading order , the situation is more delicate than it seems to be .
the fermi liquid interactions couple a family of `` toy models '' aligned along different directions in the language of high dimensional bosonization , and lead to a correction to the entanglement entropy @xmath18 .
the scheme of multi - dimensional bosonization was first introduced by haldane@xcite , followed by others@xcite .
the basic idea is to start with a low energy effective hamiltonian ( obtained through a renormalization group ( rg ) approach ) restricted to within a thin shell of thickness @xmath19 around the fermi surface , @xmath20 .
then one divides this thin shell into @xmath21 patches with dimensionality @xmath22 as shown in fig.(2 ) in such a way that @xmath23 and @xmath24 , where @xmath25 is the space dimension , @xmath26 is the linear dimension of the tangential extent of each patch . the condition @xmath27 minimizes inter - patch scattering ; @xmath28 and @xmath24 together makes the curvature of the fermi surface negligible . in the end
we shall take the limit @xmath29 , so that the sum over all patches can be converted to an integral over the fermi surface . in this work
, we treat the free theory in general @xmath5 dimensions , but shall restrict ourselves to @xmath4 when interactions are included . [
fig : patching ] for an arbitrary patch @xmath30 , labeled by the fermi momentum @xmath31 at the center of the patch , we introduce the patch fermion field operator ( ; ) = e^i_ _ ( ; ) e^i(-_ ) _ , where @xmath32 is the usual fermion field in momentum space , @xmath33 the effective fermi liquid hamiltonian can be written as & h[^ , ] = d^dx _ ^( ; ) ( ) ( ; ) + & + d^dx d^dy _ , v(,;- ) ^( ; ) ( ; ) + & ^ ( ; ) ( ;) , [ eq : h ] with @xmath34 being the effective mass , @xmath35 the effective interaction in the forward scattering channels .
even though this model is restricted to special interactions of this form , forward scattering is known to be the only marginal interaction in rg analysis@xcite . as the leading order contribution of the entanglement entropy
is dominated by the low energy modes around the fermi surface , it is sufficient to consider this model .
similar to the 1d case , the bosonic degrees of freedom are the density modes of the system , in this case defined within each patch of the fermi surface : [ eq : current_operator ] j( ; ) = _ ( ; - ) ( ; ) \{^_- _ - ^d_,0 } .
though @xmath36 is not explicitly bounded in the above definition of the patch density operator , its transverse components @xmath37 ( those parallel to the fermi surface ) are limited @xmath38 due to the patch confinement . their commutation relation is [ eq : j_commutator_1 ] & & [ j( ; ) , j ( ; ) ] _ , _ + , 0^d _ k ( ; ) + & n
_ + + [ eq : j_commutator ] & = _ , _ + , 0^d ( _ ) _
2(_ ) + ( / ) , where _ 2(_ ) = _ ( 1 - q_^()/ ) , @xmath39^d$ ] , @xmath40 is the occupation number of state with momentum @xmath41 , @xmath42 is the outward normal direction of patch @xmath30 , @xmath43 represents all other component(s ) of @xmath36 that are perpendicular to @xmath42 , and @xmath44 is the linear dimension of the entire system .
the appearance of @xmath45 is a result of momentum conservation .
the calculation of the commutator is reduced to computing the difference of occupied states , i.e. the area difference below the fermi surface , between the two @xmath46 functions @xmath47 as indicated by eq .
( [ eq : j_commutator_1 ] ) .
this is similar to 1d bosonization .
if we consider both @xmath41 and @xmath36 to be 1d momenta , eq .
( [ eq : j_commutator_1 ] ) would give us the 1d bosonization commutator .
the 2d result eq .
( [ eq : j_commutator ] ) is similar , because the fermi surface confined within the patch is essentially flat thus the dispersion is 1d .
that leads to the @xmath48 dependence of the commutator as that of the 1d case , even for @xmath49 .
the difference is that , as illustrated in fig .
( 3 ) , due to the patch confinement on the transverse direction(s ) , when @xmath43 increases @xmath50 would increasingly find itself outside the patch thus not contributing to the commutator . according to fig .
( 3 ) , one can see that this gives rise to the factor @xmath51 , which diminishes the commutator at large @xmath43 .
[ fig : commutator ] . as shown in eq .
( [ eq : j_commutator ] ) , the commutator is reduced to computing the difference of occupied states , i.e. the area difference below the fermi surface , between the two @xmath46 functions @xmath52 .
the solid box indicates the original patch , or @xmath53 .
the red line shows the fermi surface .
both @xmath54 and @xmath55 are denoted by dashed boxes .
the occupied part in @xmath56 is denoted by blue , that of @xmath57 is denoted by red , and the overlapping region is denoted by yellow . subtracting the remaining blue area from the red , we obtain that @xmath57 occupies @xmath58 more states , which gives us the commutator.,title="fig:",width=302 ] it is usually neglected in literature because the long wavelength limit is taken@xcite .
however , as it is important in the present context to correctly count the number of total degrees of freedom , this @xmath51 factor can not be neglected because it comes from counting the transverse degrees of freedom . to simplify things ,
we replace @xmath59 by _ 2(_ ) = 1 and we also limit @xmath43 to this range .
this approximation makes it easier to do fourier transform while keeping the total degrees of freedom intact . to see that , it is sufficient to consider one direction , comparing the area enclosed by the two different functions : @xmath60 over the range @xmath61 and @xmath62 over the range @xmath63 .
both functions enclose the same area thus the same number of states .
this approximation can also be interpreted as relaxation of the hard wall cutoff in eq .
( [ eq : current_operator ] ) , softening of the step function @xmath64 . in eq .
( [ eq : current_operator ] ) , @xmath36 is not bounded while @xmath41 is bounded by @xmath64 .
if we relax the restriction on @xmath41 on the transverse direction , allowing @xmath41 with @xmath65 in the summation , but require @xmath66 to be bounded within the patch , we would obtain the alternative @xmath67 . using from now on the above approximation
, we construct the local bosonic degrees of freedom @xmath68 as j( ; ) = _ x_ ( ; x_ , _
) , where @xmath69 , @xmath70 , and @xmath71 .
the commutation relations for the @xmath72 s are then [ eq : bcom ] & [ _ x_ ( ;) , (; ) ] = i 2_ , ( x_ - y_ ) + & _ = 1^d-1 ( ) + which is the bosonic commutation relation we are looking for .
the factor @xmath73 arising from transverse directions must be treated with care in different circumstances . in most literature , the focus is the physics at large length scale @xmath74 ; therefore , this factor is usually approximated by @xmath75 which is good in that limit without further discussion .
this is also what we shall do for most of the time unless noted otherwise : & [ _ x_ ( ;) , (; ) ] + & i 2_,^d-1 ( x_ - y_ ) ^d-1 ( _ - _ ) .
however , the more accurate expression is useful for us to understand how to count the transverse degrees of freedom correctly .
it tells us that the transverse degrees are not independent on the short length scale @xmath76 .
more importantly , later on we need to consider the limit @xmath77 ; without eq .
( [ eq : bcom ] ) , this limit would be ill - defined . with the above , the hamiltonian @xmath78 $ ]
is found to be quadratic in terms of these @xmath79 s : [ eq : bosonized_h ] h[^ , ] = _ , ; j(;- ) j ( ; ) + + v( , ; q ) j(;- ) j ( ; ) , where @xmath80 is the fourier transform of @xmath81 .
so it is also quadratic in the bosonic fields associated with the @xmath79 s . the kinetic energy part of eq .
( [ eq : h ] ) or its bosonized version eq .
( [ eq : bosonized_h ] ) can be written in terms of the boson fields constructed above as : [ eq : bh_free ] h_0 & = _ ; j(;- ) j( ; ) + & = _ d^2x ( _ x_ ( ; ) ) ^2 .
we see that there is no coupling between different patches .
the theory is thus formally a tensor product of many independent theories , one for each patch .
we can therefore calculate the entanglement entropy patch by patch and sum up contributions from each patch in the end . within a single patch
there is no dynamics in the perpendicular direction as dictated by the hamiltonian , and the problem is reduced to a _
one dimensional _ problem ! note that transverse degrees of freedom are not completely independent . according to eq .
( [ eq : bcom ] ) , the commutator is non - vanishing for @xmath82 up to a length scale @xmath83 .
this is a consequence of restricting @xmath43 to within the range @xmath84 $ ] .
physically one can view this as _ discretization _ along the transverse direction due to a restricted momentum range , similar to the relation between a lattice and its brillouin zone .
in this view the single patch problem is reduced to a 1d problem with a chain density of @xmath85 .
therefore , the hamiltonian ( [ eq : bh_free ] ) becomes h_0 = _ ; _ dx_ ( _ x_ ( ; ) ) ^2 .
note that the bosonized theory of a single patch is chiral . to directly make use of our toy model
, we need to consider two patches having opposite @xmath42 simultaneously .
this is because for a 1d fermion model at non - zero filling , there are two fermi points .
both need to be considered to construct well - defined local degrees of freedom .
once we consider such two patches together , it is more convenient to combine the two chiral theories into a non - chiral theories .
this is also what we will do for the rest of this work .
introduce the non - chiral fields ( ;) = ( ( ;) - ( - ; ) ) , + ( ;) = ( ( ;) + ( - ; ) ) , where @xmath86 indicates the patch with normal direction opposite to that of patch @xmath30 : @xmath87 .
one finds that @xmath88 and @xmath89 are mutually dual fields with @xmath30 restricted to one hemisphere , but @xmath90 and @xmath89 now commute while @xmath88 and @xmath89 have a non - trivial commutator : & [ ( ; ) , _ y_( ; ) ] = [ ( ; ) , _ y_( ; ) ] = 0 , + & [ _ x_( ; ) , ( ;) ] = [ _ x_( ; ) , ( ;) ] + & = 2 i _ , ( x_ - y_ ) ^d-1(_ - _ )
. therefore , two patches with opposite @xmath42 are equivalent to a set of ordinary 1d boson fields . throughout the rest of this work
, we shall assume this chiral - to - nonchiral transformation is done , and when we refer to patches we always refer to the two companion patches that form a non - chiral patch together .
for the non - chiral boson theory , it is known that the entanglement entropy of a single interval ( with two end points ) is @xmath7 .
before we proceed further , we note that the relation between boson fields and the original fermion fields is not completely local .
however , the underlying physical quantity that matters is not the fields , but the fermion density , or in other words , the fermion number basis one chooses to expand the hilbert space of the problem .
this physical basis is also what one uses to do the partial trace .
it is known that the fermion density operator obeys a locally one - to - one corresponding relation to the boson fields .
thus we argue that in 1d the nonlocal relation between the fermion and boson fields does not affect the partial trace operation , so as the calculation of entanglement entropy . by referring to our result for the toy model , the contribution from a single patch
is readily given s( ) = l _ a _
d_x ( ) ^d-1 , where an additional factor of @xmath91 has been introduced in order to count only once each pair of patches forming a non - chiral theory . identifying @xmath92 as the surface element at the fermi surface @xmath93 and taking the @xmath94 limit
, the total entanglement entropy is [ eq : gk ] s & = l _ a _ _ k d_x .
so we recover the gk result for generic free fermions .
when fermi liquid interactions ( forward scattering ) are included , the full hamiltonian will no longer be diagonal in the patch index @xmath30 .
but it is still quadratic in terms of the patch density operators , i.e. the bosonic degrees of freedom , and can be diagonalized by a bogoliubov transformation . according to eq .
( [ eq : j_commutator ] ) and ignoring terms of @xmath95 , one can define a set of boson creation / annihilation operators @xmath96 as follows : @xmath97 it can be shown that the full hamiltonian is diagonal in @xmath36 , and it can be diagonalized by a bogoliubov transformation@xcite independently for each @xmath36 sector . in ref .
[ ] , only a hubbard-@xmath98 like interaction is considered for practical reasons .
but in principle , such a bogoliubov transformation also applies to general interactions : & _ i ( ) = _ j u_ij _ j ( ) + v_ij ^_j ( ) , + & _ i ( ) = _ j u_ij _ j ( ) + v_ij^_j ( ) , where both @xmath99 and @xmath100 refer to the patch index , @xmath101 and @xmath102 are the bogoliubov bosonic annihilation operators that diagonalize the hamiltonian . with proper choice of @xmath103 s and @xmath104 s
, the hamiltonian is readily diagonalized . ref .
[ ] solves the hubbard-@xmath98 like interaction and provides a successful description of fermi liquids , even in the strong @xmath98 limit .
however , even for this simple case in which @xmath98 has no @xmath36-dependence , the bogoliubov transformation still depends on @xmath36 . to be more precise , @xmath105 and @xmath106 will depend only on the angle between the patch normal direction @xmath42 and @xmath36 , leading to discontinuities in the derivatives at @xmath107 .
consequently , the real space fields constructed from the bogoliubov operators @xmath101 and @xmath102 are _ no longer _ local with respect to the original boson fields .
the real space bogoliubov fields are constructed in a manner similar to eq .
( [ eq : cna_operator ] ) : ( ; ) = i _ , _ > 0 .
then one can show that the original local degrees of freedom @xmath108 can be expressed in terms of above bogoliubov fields as [ eq : nonlocality ] ( ; ) & = ( ; ) + d_l f(,l;- ) ( l ; ) , where @xmath109 is typically long - range , even for the short - range hubbard-@xmath98 interaction . for more general cases , with further @xmath36-dependence in the interaction , the non - locality would only be enhanced .
the loss of locality prevents us from calculating the entanglement entropy directly using those eigen modes , since it is difficult to implement the partial trace using those non - local degrees of freedoms .
therefore , although the bogoliubov fields have a local core as we would expect for fermi liquids from adiabaticity , they _ do _ acquire a nonlocal dressing due to interaction .
though in principle the partial trace can be done with those bogoliubov fields , such nonlocality makes it difficult and we have not been able to do it , which further renders calculating the entanglement entropy impossible .
this is very different from the 1d theory , where for local interactions the eigen fields remain local , since there are only two fermi points .
there the transformation can never involve such angular @xmath36-dependence due to limited dimensionality . despite these technical difficulties ,
the non - locality may suggest possible corrections to the entanglement entropy .
this is indeed the case as revealed by our later calculation for fermi liquid interactions , although in this case such extra contributions are only of @xmath110 which is of @xmath111 comparing to the leading term .
this shows that the mode - counting argument in ref .
[ ] , though correctly suggesting the @xmath1 violation to the area law for fermi liquids , does not always fully account for all sources of entanglement entropy .
in order to preserve locality , we need to work with the original local degrees of freedom . to do that
, we adopt the approach used by calabrese and cardy@xcite ( cc ) on calculating the entanglement entropy of a free massive @xmath112 bosonic field theory .
the calculation is done in terms of the green s function by applying the replica trick . in our case
, we find that the cc approach can be generalized in a special geometry for solving the interacting theory which is quadratic after bosonization . in this way
, we avoid diagonalizing the hamiltonian and thus the nonlocality issue . however , we do have to regularize the theory by adding a mass term by hand . in the end
we shall take the small mass limit , and replace the divergent correlation length @xmath113 by the subsystem size @xmath2 .
the regularization procedure facilitates the calculation , but also strictly restricts us to computing the entanglement entropy only at the @xmath1 level . in this section , by using the replica trick we convey the calculation of entanglement entropy into computing the green s function on an @xmath114-sheeted replica manifold .
we first demonstrate the method by applying it to free fermion theory in @xmath5-dimensional space ; then based on it we compute the entanglement entropy perturbatively for a simple fermi liquid theory in powers of the interaction strength up to the second order . in this part
, we briefly describe the replica trick in @xmath115 space - time dimensions ( @xmath116 ) so that later on we can straightforwardly generalize it to @xmath117 space - time dimensions ( @xmath118 ) accordingly for our problem .
the replica trick makes use of the following identity : s_a = -(_a _ a ) = -_n 1 _ a^n .
[ eq : replica]to compute @xmath119 , cc use path integral to express the density matrix @xmath120 in terms of the boson fields ( \{(x)}|\{(x) } ) = z^-1 , where @xmath121 is the partition function , @xmath122 is the inverse temperature , and @xmath123 are the corresponding eigenstates of @xmath124 : @xmath125 .
@xmath120 can be expressed as a ( euclidean ) path integral : & = z^-1 _ x((x,0)-(x) ) + & _
x((x , ) - ( x ) ) e^-s_e , where @xmath126 , with @xmath127 being the euclidean lagrangian .
the normalization factor @xmath128 , i.e. the partition function is found by setting @xmath129 and integrating over these variables .
this has the effect of sewing together the edges along @xmath130 and @xmath131 to form a cylinder of circumference @xmath122 as illustrated in fig .
( [ fig : rdm ] ) ( left panel ) .
the reduced density matrix of an interval @xmath132 can be obtained by sewing together only those points which are not in the interval @xmath6 .
this has the effect of leaving an open cut along the line @xmath130 which is shown in fig .
[ fig : rdm ] ( right panel ) . to compute @xmath133
, we make @xmath114 copies of above set - up labeled by an integer @xmath134 with @xmath135 , and sew them together cyclically along the open cut so that @xmath136 $ ] for all @xmath137 . in fig .
[ fig : n_rdm ] we show the case @xmath138 .
let us denote the path integral on this @xmath114-sheeted structure ( known as @xmath114-sheeted riemann surface ) by @xmath139 . then _
a^n = , so that s_a = -_n 1 .
together for all @xmath140 s , we get the partition function @xmath128 .
right : when only sew @xmath141 together , we get @xmath142.,width=302 ] [ fig : replica ] if we consider the theory as that of one field living on this complex @xmath143sheeted riemann surface instead of a theory of @xmath114 copies , it is possible to remove the replica index @xmath114 from the fields , and instead consider a problem defined on such an @xmath114-sheeted riemann surface which can be realized by imposing proper boundary conditions . in ref .
[ ] , cc consider the entanglement entropy between the two semi - infinite 1d system ( i.e. cutting an infinite chain into two halves at @xmath144 ) for free massive boson fields . for such geometry , as illustrated in fig .
[ fig : riemann_surface ] , the @xmath143sheeted riemann surface constraint is realized by imposing a @xmath145 periodicity on the angular variable of the polar coordinates of the @xmath116 plane instead of the usual @xmath146 one . in this way
, the @xmath116 variable @xmath147 acquires @xmath114 branches @xmath148 , and each branch corresponds to one copy of @xmath72 .
notation - wise this corresponds to ( x,)_k ( _ k ) ( ) , and the sewing conditions @xmath149 simply becomes the continuity condition for @xmath150 across its consecutive branches .
here we use a generalized polar coordinate : @xmath151 with @xmath152 , and @xmath153 . the massive free boson theory considered by cc
is defined by the following action @xmath154 the @xmath116 bosonic green s function @xmath155 on the @xmath143sheeted riemann surface satisfies the differential equation @xmath156 to compute the partition function , one can make use of the identity z_n = - d^d+1 x g^(n ) ( , ) .
note that here the integration is over the entire @xmath143sheeted space .
the above is applicable to general quadratic theories of bosons , and will be applied by us later to bosonized theories of interacting fermions . here
we use @xmath157 , a general two point correlation function on the @xmath143sheeted riemann surface in @xmath5-dimensional space for later use , instead of the specific @xmath158 defined above .
accordingly , @xmath159 is then given as [ eq : entropy ] s_a = - _ n
1 e^- dm^2 d^d+1 x ( g^(n ) ( , ) - n g^(1 ) ( , ) ) . here and in the following ,
we will leave it understood that the first term in the integrand is integrated over the @xmath114-sheeted geometry , whereas the second is integrated over a one - sheeted geometry . there should be no confusion as the superscript of @xmath160 generally indicates the geometry .
the benefit of the above approach is that the two point correlation function or green s function , defined in terms of certain differential equation obtained from the equation of motion , can be solved for on the @xmath143sheeted riemann surface thus enabling us to compute the entanglement entropy .
although cc s work only considers massive @xmath116 boson fields , it is also applicable to our case .
the price one has to pay is to to introduce a mass term for regularization . at the end of the calculation the inverse mass , which is the correlation length of the system , shall be considered to be on the same scale as @xmath2 : @xmath161 , where @xmath2 is the characteristic length scale of the subsystem .
the validity of such consideration is well - established in other cases,@xcite where the correlation length is either set by finite temperature or mass .
the only modification necessary to apply the above to a bosonized fermi surface in higher dimensions is to introduce a sum over the patch index . through the remainder of this work , instead of the general geometry considered before , we work with a special half - cylinder geometry as shown in fig .
[ subfig : geometry ] : the system is infinite in the @xmath162 direction while obeying periodic boundary condition along the @xmath163 direction with length @xmath2 . the system is cut along the @xmath163 axis so that we are computing the entanglement entropy between the two half planes .
we require @xmath2 to be large so that it can be considered @xmath164 unless otherwise noted .
we choose such this simple geometry for the following reasons .
cutting the system straight along the @xmath163 direction , yielding a two half - plane geometry , is a straightforward @xmath118 generalization of the semi - infinite chain geometry considered in cc .
it makes any straight line intersect the boundary only once , dividing it into two semi - infinite segments , for all patch directions as in the 1d case , except for lines parallel to the @xmath165 patch direction .
the degrees of freedom associated with this special patch do not contribute to the entanglement entropy , since they are not coupled ( have no dynamics ) along @xmath162 , and are of measure zero in the large patch number limit anyway . for this simple geometry ,
the @xmath118 @xmath114-sheeted geometry is constructed from @xmath114 identical copies @xmath166 sliced along `` branch cuts '' @xmath167 and then appropriately glued together along these cuts .
this happens exactly as in 1d , and the @xmath168 coordinate is so far a mere spectator .
this defines an @xmath143sheeted , or in this case more appropriately the @xmath143layered , replica manifold which is a simple enough generalization of the @xmath116 case . the @xmath114-sheeted riemann surface , as discussed in sec .
[ sec : ee_free_fermion ] and shown in fig .
[ fig : riemann_surface ] , now acquires an extra direction @xmath163 perpendicular to the @xmath169 plane .
it can still be implemented by imposing the same @xmath170 periodicity boundary conditions on @xmath46 , the angular variable of the polar coordinates @xmath171 in the @xmath169 plane .
therefore , we can safely make use of the cc result , i.e. the solution to the green s function on a @xmath143sheeted riemann surface , to the free fermion theory , and can further use it as a starting point for treating the interacting theory .
this is obviously true for the patch with @xmath172 , but it also holds for general @xmath42 as we shall validate as the following . for a general patch direction @xmath42 , the noninteracting green s function associated with this patch embodies correlations in the affine @xmath173 `` planes '' .
the geometry of each such `` plane '' is that of the @xmath114-sheeted riemann surface of the @xmath116 problem , as we will now argue . with each patch direction
we thus associate a different foliation of the @xmath114-sheeted @xmath118 geometry into @xmath116 counterparts .
to be more precise , for given patch @xmath174 , instead of the cartesian coordinates @xmath175 , we consider a parallel / perpendicular decomposition @xmath176 for each sheet via @xmath177 where the @xmath178 are the perpendicular unit vectors aligned with the patch @xmath174 .
the natural choice of coordinates for a given patch is to choose polar coordinates within the @xmath173 plane : @xmath179 because these are the coordinates in which the @xmath173 planes restricted to each sheet are naturally glued together by extending the range of @xmath180 to @xmath181 , as we will now show .
the shift @xmath182 of @xmath183 is necessary as to ensure that @xmath144 , the location of the onset of the branch cut , corresponds to @xmath184 which is what makes these coordinates convenient .
the @xmath173 planes are now defined by fixed @xmath185 . if we can establish that the @xmath181 periodicity of @xmath46 is equivalent to a @xmath181 periodicity of @xmath180 , then cc s solution would be justified in the above set - up so that the non - interacting green s function @xmath186 can be expressed through cc s result .
this can be achieved by establishing a one - to - one correspondence ( mapping ) between @xmath46 and @xmath180 .
the mapping is intuitively constructed , as shown in fig .
[ subfig : bc ] , as the vertical projection from the @xmath173 plane ( the blue plane in fig .
[ subfig : bc ] ) onto the @xmath169 plane along the @xmath187 direction .
consider moving the projection of @xmath188 in the @xmath189 plane around the origin @xmath114 times ( the red circle ) .
it is clear that @xmath188 ( on the blue ellipse ) follows its projection while also moving around the origin @xmath114 times , always being on the same sheet . in particular ,
the branch cut is always traversed simultaneously for @xmath190 .
the @xmath173 planes , the leaves of our foliation , thus have the familiar 1 + 1@xmath5 @xmath114-sheeted geometry , and @xmath180 obeys the same @xmath170 periodicity as @xmath46 . finally , the periodicity condition of the @xmath163 direction is necessary for the total entanglement entropy to be finite ; it also provides the only length scale for the subsystem which is needed for extracting the scaling behavior of entanglement entropy .
however , if we are only concerned with the integral form of the entanglement entropy as in eq .
( [ eq : gk ] ) , not requiring it to be finite as a whole , but rather requiring only the entanglement entropy per unit length to be finite , we may take the @xmath168 direction to be infinite .
this point of view will be taken here and in the following in order to simplify our calculation . in order to treat the interacting theory , in this section we re -
derive the free fermion result for the half cylinder geometry via the replica trick .
later we shall generalize the method to include interactions . rewriting the hamiltonian eq .
( [ eq : bh_free ] ) in terms of the non - chiral fields , and adding the mass term by hand , we have & h[( ; ) ] = _ d^2x ( ( _ x_ ( ; ) ) ^2 + ( _ x_ ( ; ) ) ^2 ) + ( ; ) ^2 . for convenience
, we use the lagrangian formalism and work with the @xmath191 representation through the rest of this work .
switching to imaginary time @xmath192 , and rescaling the coordinates in the following manner : [ eq : rescale ] , , we obtain the following lagrangian density in the @xmath89 representation : = - [ ( _ ( ; ) ) ^2 + ( _ ( ; ) ) ^2 + ( ( ; ) ) ^2 ]
. then we can work out the euler - lagrangian ( e - l ) equation of motion .
making use of the e - l equation of motion , we find the green s function @xmath193 satisfies the following differential equation : [ eq : diff_free_g ] & - ( _ ^2 + _ x_^2 - 1 ) g_0^(n)( , ; , ) + & = c _ , ( -_y ) ( x_ - y_ ) ^d-1(_ - _ ) , where @xmath194 .
the rescaling makes the green s function dimensionless , thus easier to handle when it comes to computing @xmath195 .
the extra factor @xmath196 generated on the right hand side ( @xmath197 ) will be canceled by the jacobian of the integral over the green s function , leaving only a factor of @xmath198 .
all that needs to be computed is then an integral over the dimensionless @xmath160 .
therefore , it is legitimate to ignore this factor from now on .
the @xmath199-functions originate from the commutator eq .
( [ eq : bcom ] ) , and are _ coarse - grained_. after we include the patch index , perform the integral over @xmath200 , and take the @xmath114 derivative , eq . ( [ eq : entropy ] ) becomes [ eq : entropy_reduced ] & s_a = ( m^2 a_0 ^ 2 ) _ n 1 _ ( c_g( ; n ) - n c_g(;1 ) ) , where @xmath201 is an ultraviolet cutoff , and c_g( ; n ) = d^d+1 x g_0^(n)(, ; , ) .
the exponential factor in eq .
( [ eq : entropy ] ) becomes one after the @xmath202 limit is applied .
note that @xmath203 .
our major task is now computing @xmath204 . observing that there is no @xmath205 dependence on the left hand side ( @xmath206 ) of eq .
( [ eq : diff_free_g ] ) , we can write @xmath207 and we obtain a @xmath116 equation & -(_^2 + _ x_^2 + 1 ) g_0,b^(n)(;_,,_ , ) + & = ( -_y ) ( x_ - y_ ) in which @xmath208 is as defined in eq .
. the same equation appears in cc .
we shall also suppress the subscript @xmath209 unless necessary , as it is normally already specified in the notation for @xmath210 .
the transverse part of the integral in @xmath204 can be factored out as @xmath211 recalling our discussion about eq .
( [ eq : bcom ] ) , this is a coarse - grained @xmath199-function . at short distances , instead of a divergence , we should use @xmath212 therefore , the transverse direction integral becomes @xmath213 identifying @xmath214 as the surface element @xmath215 , for a given patch the integration can be rewritten as @xmath216 .
this leaves us with only an integral over @xmath217 .
the solution for the @xmath116 green s function on the @xmath143sheeted replica manifold is given in cc : [ eq:(1 + 1)d_g ] g_0,b^(n)(;_x,_y ) & = _
k=0^ d_k _ k / n(_x-_y ) + g_k / n(r_x , r_y ) , where @xmath218 for @xmath219 , @xmath220 , @xmath221 , and @xmath222 and @xmath223 are the modified bessel functions of the first and second kind respectively . @xmath224 and @xmath46 are again the polar coordinates of the @xmath225 plane , and we have suppress the index @xmath30 of @xmath224 as only one patch direction is involved .
the integral over @xmath158 is & d^2r_x g_0,b^(n)(;_x,_x ) = dr_x r_x _ k d_k g_k / n(r_x , r_x ) .
the integral is divergent since the integrand @xmath226@xmath227 , a consequence of the fact that we are calculating the partition function of an infinite system . but
this divergence should be canceled in @xmath228 . to regularize the divergence
, we use the euler - maclaurin ( e - m ) summation formula following cc , and sum over @xmath134 first : [ eq : eml_sum ] _ k=0^d_k
f(k ) & = _
0^f(k ) dk - f(0 ) + & - _ j=2 ^ f^(2j-1)(0 ) , where @xmath229 are the bernoulli numbers , @xmath230 .
note that the first term , the integral over @xmath134 , is always canceled by rescaling @xmath231 in @xmath232 . for the remaining terms , which contain derivatives with respect to @xmath134 , we may add a constant , in this case @xmath233 , under the derivative , which allows us to pull the derivative outside the integral .
the integrand now is well - behaved at infinity . to be more precise , according to eq .
( [ eq : eml_sum ] ) , we need to compute @xmath234 so we have c_g(;n ) - n c_g(;1 ) = . combining the above results into eq .
( [ eq : entropy_reduced ] ) and converting the sum over @xmath30 into an integral around the fermi surface , we obtain eq .
( [ eq : gk ] ) for this geometry . in this part
, we derive the differential equations of the green s functions for the quadratic boson theory with inter - patch coupling , and provide an iterative solution . including the fermi liquid interaction @xmath235 , the hamiltonian becomes & h[(;)]= d^2x where @xmath236 is order @xmath237 .
this hamiltonian can be written in terms of the non - chiral fields as @xmath238 where we have made use of the fact that @xmath239 , which is required by time - reversal symmetry . here
the summation over @xmath30 is restricted to a semicircle .
this hamiltonian contains generalized type kinetic terms ( inter - patch coupling due to interaction ) which are not diagonal . to obtain the corresponding lagrangian
, one needs to invoke the general legendre transformation@xcite , and obtains the following lagrangian densities , respectively , in terms of @xmath89 or @xmath88 : _ & = + _ & = , where & f_1( , ) = g_ , + g_-,- - g_,- - g_- , , + & f_2( , ) = g_ , + g_- , - + g_- , + g_,- , and @xmath240 is defined through @xmath241 \}^{-1 } = i + [ h_{1(2)}(\s,\t)].\ ] ] here , @xmath242 is the identity matrix , and @xmath243 $ ] is the matrix formed by @xmath244 , @xmath245 . applying this result and making use equations of motion obtained from the hamiltonian
, we obtain the lagrangians @xmath246 or @xmath247 . here
we arbitrarily choose to work with @xmath248 .
then , by making use of the e - l equation of motion , applying the same rescaling eq .
( [ eq : rescale ] ) , and letting @xmath249 , we obtain the differential equations that the interacting green s function @xmath250 s satisfies : [ eq : g_resc ] & - ( _ ^2 + _ ^2 - 1 ) g^(n)(,;, ) + & + _ l ( h_2(l , ) _ ^2 + f_1(l , ) _
l _ ) g^(n)(l,;, ) + & = c _ , ( - ) ( x_-x_ ) ) ( x_ - x_ ) .
here the jacobian due to change of variables is the same as in the free fermion case .
the entanglement entropy is still given by eq .
( [ eq : entropy_reduced ] ) , but replacing @xmath251 with @xmath252 in @xmath204 . in the following ,
we omit the replica index @xmath114 in the green s function unless different values of @xmath114 are involved in a single equation .
as is well known , differential equations such as the above can be converted to an integral form@xcite relating the full green s function to the noninteracting one .
this leads to an iterative ( perturbative ) definition of the former in terms of the latter . in the present case ,
this integral equation reads [ eq : iteration_sol ] & g( , ; , ) + & = g_0( , ; , ) + d^3 z g_0(, ; , ) + & ( _ l ( h_2(l , ) _
^2 + f_1(l , ) _
l _ ) g(l , ; , ) ) + & = g_0( , ; , ) + g( , ; , ) . given this equation
, we can now compute the green s function and thus the entanglement entropy perturbatively in powers of @xmath98 . in eq .
, the @xmath253 term is the same as that of the free fermions , thus yields the same contribution to entanglement entropy . to study how the correction term @xmath254 affects the entanglement entropy
, we need to study @xmath255 where @xmath256 denotes the @xmath257th order correction .
there are two distinctive types of terms in the perturbative expansion of @xmath258 . in general , at order @xmath257 , we have in total @xmath259 integrals .
let us examine one of the many terms contributing to the m - th order correction , to be summed over patch indices : [ eq : in - patch_o(m ) ] & d^3x ^(m ) g(, ; , ) + & ~d^3x _ i=0^m-1(d^3z_i ) g_0(, ; , _ 0 ) + & _ _ 0 ^ 2 g_0(l_0,l_0 ; _ 0 , _ 1 ) _ _ i^2 g_0(l_i , l_i ; _ i , _
i+1 ) + & _ _ m-1 ^ 2 g_0(, ; _ m-1 , ) . here
we only include the @xmath260derivatives .
in general we would also have spatial ( @xmath261 ) derivative terms , as well as terms with mixed derivatives .
but @xmath262 and @xmath42 directions are equivalent . using rotational symmetry , and the fact that the two different derivatives in each term are with respect to independent variable that are each integrated over
, one can see that all terms are identical except for @xmath174-dependent pre - factors .
the two categories of terms are defined by the set @xmath263 : 1 ) @xmath264 @xmath265 , i.e. with intra - patch coupling only ; and 2 ) @xmath266 containing inter - patch coupling .
we shall label the two categories as @xmath267 setting @xmath264 for all @xmath99 s in eq .
, first we consider the transverse direction @xmath268 we can immediately integrate out the transverse component of all @xmath269 s and obtain @xmath270 in the last line we use again the fact that the transverse @xmath199-function is a coarse - grained one ( eq . ) .
the rest of @xmath271 is obtained by substituting @xmath272 with the @xmath116 green s function @xmath273 .
although a direct computation is possible , we first give a general argument that for any @xmath257 the contribution to entanglement entropy from @xmath274 vanishes .
we do so by making a comparison with the 1d case where a rigorous solution is available . for the 1d luttinger liquid with only forward scattering
, the entanglement entropy can be calculated directly via bosonization and the result remains at @xmath275 in the presence of interactions .
the calculation is possible because , in our language , there are only two patches , so the transformation which diagonalizes the hamiltonian is not plagued by the nonlocality issue we encounter in the 2d theory .
however , we can also treat the 1d case with our perturbative approach .
the resulting series of integrals turns out to be _ identical _ to the one obtained from the intra - patch contributions in the higher dimensional case except for the transverse @xmath276function .
therefore , we argue that at all orders , the intra - patch coupling terms have vanishing contribution to the entanglement entropy
. we shall demonstrate such behavior explicitly up to second order in @xmath98 later on . for terms with inter - patch coupling
, we find that they are of order @xmath111 comparing to the leading term according a scaling argument .
the crucial observation here is that , as long as @xmath266 , we do not encounter the factor @xmath277 , eq . because for @xmath278 ( z_1 ^()- z_^( ) ) ( z _ ^(l)- z_1^(l ) ) = , where @xmath180 ( @xmath279 ) is the angle between @xmath42 ( @xmath280 ) and the @xmath162-axis in the @xmath281 plane .
therefore , when we integrate out the @xmath282 transverse @xmath199-functions , the factor @xmath277 would be suppressed by even a single @xmath283 . to examine the remaining integral
, we can ignore the angular part as it can not affect the scaling behavior .
the asymptotic expansion of @xmath223 and @xmath222 for real @xmath224 at large value is@xcite & k_(r ) e^-r , + & i_(r ) , where @xmath284 . by using the above asymptotic expansion of bessel functions ,
the leading term for @xmath285 behaves as @xmath286 .
all of these terms peak around @xmath287 and are otherwise exponentially suppressed .
we may therefore again estimate this integral by letting @xmath288 and removing @xmath289 of the integrals .
the remaining integrals yield , at the leading order , @xmath290
. however , at the leading order , there is no @xmath291 dependence . according to the formalism in sec .
[ subsec : ee_of_free_fermion_rev ] , such terms have no contribution to the entanglement entropy .
therefore , the term that contributes to the entanglement entropy is the next order which behaves as @xmath292 and is of order @xmath293 , leading only to a correction @xmath294 to the entanglement entropy .
next , we shall demonstrate in detail our above analysis , for both inter - patch and intra - patch coupling terms by explicit calculation up to the second order .
the first order term correction to @xmath295 is [ eq : cg1 ] & ^(1 ) c_g( ; n ) = d^3x d^3z g_0(, ; , ) + & ( h_2(, ) _
_ z^2 + f_1(,)_z_^2 ) g_0(, ; , ) . as we have pointed , it is sufficient to calculate either piece of the two terms due to the equivalence of the imaginary time direction and the real space direction .
the other piece should be just the same except for the coefficient .
here we choose to compute the first term .
the transverse degrees of freedom provide an overall factor counting the total degrees of freedom as discussed in the general case .
then we can also integrate out the angular degrees of freedom in the @xmath296 plane , both @xmath297 and @xmath298 as defined in eq .
( [ eq:(1 + 1)d_g ] ) , after which one obtains [ eq : dg1 ] & ^(1)c_g( ; n ) ~_k ^(1 ) g_k / n _ a _ xd_k , where & ^(1 ) g_k / n = dr_x dr_z r_x r_z g_k / n ( r_x , r_z ) + & ( _ r_z^2 - ) g_k / n ( r_x , r_z ) .
the two @xmath134 summation is reduced to one due to orthogonality of the angular function @xmath299 . by employing the e - m formula and properties of the bessel functions ,
we show in appendix [ apdx : a ] that sum over @xmath134-values in eq . can be converted into an integral , which cancels in eq .
for the same scaling reasons discussed above , following eq . .
therefore , we find that the contribution of eq . to the entanglement entropy vanishes .
the second order correction is & ^(2)c_g(;n ) = d^3x d^3z d^3z_1 g_0(, ; , ) + & _ l ( h_2(l, ) _
_ z^2 + f_1(l, ) _
z_l_z_ ) g_0(l , l;,_1 ) + & ( h_2(, ) _ _ z_1 ^ 2 + f_1(, ) _
z_1^2 ) g_0(,;_1 , ) .
@xmath300 for @xmath301 : & d^3x d^3z d^3z_1 g_0(, ; , ) + & ( h_2(, ) _ _ z^2 + f_1(, ) _ z_^2)g_0( , ; , _ 1 ) + & ( h_2(, ) _ _ z_1 ^ 2 + f_1(, ) _
z_1^2 ) g_0(,;_1 , ) .
according to our general discussion , we only need to consider the following piece : [ eq : ksum2 ] & d^3x d^3z d^3z_1 g_0(, ; , ) _ _ z^2 g_0( , ; , _ 1 ) + & _ _
z_1 ^ 2 g_0(,;_1 , ) + & = ( 2)^-1 _ a _ xd_k _ k ^(2)g_k / n , where [ eq : dg2 ] & ^(2)g_k / n = dr_x dr_z dr_1 r_x r_z r_1
g_k / n ( r_x , r_z ) + & ( _ r_z^2 - k^2/(r_z n)^2 ) g_k / n ( r_z , r_1 ) + & ( _ r_1 ^ 2 - k^2/(r_1 n)^2 ) g_k / n ( r_1 , r_x ) . in the above , we have proceeded as in the first order calculation , integrating out the angular part first to obtain the expression for @xmath302 .
after a lengthy but similar calculation as for the first order ( see appendix b ) , we find , using the e - m formula : [ eq : em2alex ] & d_k ^(2)g_k / n= dr_x dr_z dr_1 r_x r_z r_1 + & _ 0^dk p_k / n(r_x , r_z , r_1 ) + & + , where @xmath303 is the product of @xmath232 dependent terms in eq . .
the usual scaling argument for the integral shows that the entire expression is proportional to @xmath114 , and thus cancels the second ( @xmath304 ) term in eq .
: & s ( ) ~- d^2x ( g_n - n g_1 ) ) _ n=1 .
therefore , at the second order level for the @xmath301 piece we still have no correction to the scaling law of entanglement entropy .
@xmath300 for @xmath305 : the integrand we need to consider is [ eq:2nd_order ] & g_0(, ; , ) + & ( h_2(l, ) _ _ z^2 + f_1(l, ) _
z_l_z_ ) g_0(l , l;,_1 ) + & ( h_2(, ) _ _ z_1 ^ 2 + f_1(, ) _
z_1^2 ) g_0(,;_1 , ) . + the first thing to notice in eq . ( [ eq:2nd_order ] )
is that we have derivatives along directions different from the patch normal direction @xmath42 acting on the non - interacting green s function .
we expand this term as @xmath306 for the first term , we can decompose the derivative @xmath307 into terms that act along @xmath308 and along its transverse direction , respectively .
the non - interacting green s function only depends on the transverse coordinates via @xmath309 , which indicates that those derivative terms vanish .
thus it is @xmath310 . for the second term , we integrate by parts with respect to @xmath311 , which leads to ( including now the first @xmath253 factor , which depends on @xmath312 ) @xmath313 therefore
, the overall integrand is proportional to @xmath314 .
note that the @xmath315 dependence only appears in these @xmath199-functions , we can integrate it out , leaving only @xmath316 . secondly , it is sufficient to focus on the following terms in the integrand @xmath317 to ease the presentation . for other combinations , the rest of this section is equally applicable with minor modifications that only leads to different coefficients and do not affects the scaling analysis .
we first perform the intra - patch integration d^3x g_0,b( ; , ) g_0,b(;_1,)= h( ; , _ 1 ) , where & h( ; , _ 1 ) = _ k _ k / n(_z,_z_1 ) ( ( r_z - r_1 ) + & ( r_z k_z+ i_1 - r_1 i_1- k_z ) + ( r_1 -r_z ) + & ( r_1k_1 + i_z - r_z i_z- k_1 ) ) .
so for a given @xmath30 the contribution to entanglement entropy due to coupling with patch @xmath318 can be written as [ eq:2nd_order_scaling ] & | | d^3z d^3z_1 ^2(_1 - ) ( _ _ z_1 ^ 2 h( ; , _ 1 ) + & _ ^2 g_0,b ( l;_1 , ) + _ z_ _ _
z_1 ^ 2 h( ; , _ 1 ) _ z_l g_0,b ( l;_1 , ) ) + & = | | d^3z d_1 ( _ _ z_1 ^ 2 h( ; , _ 1 ) _ ^2 g_0,b ( l;_1 , ) + & + _ z_ _ _
z_1 ^ 2 h( ; , _ 1 ) _ z_l g_0,b ( l;_1 , ) ) | _ , where @xmath319 indicate the two spatial components of @xmath312 . as we argued in previous section , for extracting the order of magnitude of the result
it is sufficient to set @xmath320 in the final line of eq .
( [ eq:2nd_order_scaling ] ) and remove the integral over @xmath321 .
we also note that the derivatives do not alter the leading power of @xmath224 , owing to the presence of the exponential function .
therefore , it is sufficient to examine d^3z ( h( ; , _ 1 ) g_0,b ( l;_1,))|__1 = . at the lowest order in @xmath322 , we have g_0,b(;r , r ) ~i_(r ) k_(r ) ~1/r , & h(;r , r ) ~r i_(r ) k_+(r ) - r k_(r ) i_-(r ) + = & ( 1++ ) ( 1 - + ) + & - ( 1-+ ) ( 1 + + ) + = & + ( ) .
since the @xmath323 derivative does not alter the leading powers , we extract the leading term to be ( h( ; , _ 1 ) _
^2 g_0,b ( ; _ 1 , ) ) |__1 = z ~. for a triple integral over @xmath324 , one would get a linear divergence , i.e. the result would be @xmath325 .
this is indeed the case as we have already seen in previous calculation .
however , at the lowest order , everything is independent on @xmath326
. actually what finally appears in the the entanglement entropy are the @xmath327derivatives of these terms appearing in the e - l summation formula .
this means the leading term has vanishing contribution to the entanglement entropy .
the first term contributing to entanglement entropy is then @xmath328 the upper limit of which is order @xmath293 and only leads a correction up to @xmath294 to the free fermion entanglement entropy .
in this paper , we developed an intuitive understanding of the logarithmic correction to the area law for the entanglement entropy of free fermions in one and higher dimensions on equal footing
the criticality associated with the fermi surface ( or points ) .
then we used the tool of high dimensional bosonization to compute the entanglement entropy , and generalized this procedure to include fermi liquid interactions . in the presence of such interactions we calculated the entanglement entropy for a special geometry perturbatively in powers of the interaction strength up to the second order , and
find no correction to the leading scaling behavior .
we also point out that the situation is the same at higher orders .
our results thus strongly suggest that the leading scaling behavior of the block entanglement entropy of a fermi liquid is the _ same _ as that of a free fermi gas with the same fermi surface , not only for the special block geometry studied in this paper , but for arbitrary geometries .
explicit demonstration of the latter is an obvious direction for future work . in the special geometry in which we performed explicit calculations using the replica trick ,
a mass - like term is introduced to regularize the theory at long distance , as is done in closely related contexts@xcite . for a fermi liquid ( which is quantum - critical ) the corresponding length scale @xmath329
must be identified with the block size @xmath2 , and is thus _ not _ an independent length scale . on the other hand , such a mass - like term
can also describe a superconducting gap due to pairing . in particular , for a weak - coupling superconductor , @xmath330 , the superconducting coherence length ,
is much longer than all microscopic length scales , but finite nevertheless . in this case
it is _ independent _ of @xmath2 , and the interplay between the two is interesting . for @xmath331 ,
the fermi liquid result ( [ eq : toy_model ] ) still holds .
but for @xmath332 , the logarithmic factor in the entanglement entropy _ saturates _ at @xmath333 , and we expect : @xmath334 which agrees with the conjecture made in ref.@xcite .
more generally , fermi liquids are ( perhaps the best understood ) examples of quantum critical phases ( or points ) in high dimensions . unlike in 1d where conformal symmetry powerfully constrains the behavior of entanglement entropy , our understanding of entanglement properties of such high - dimensional quantum critical phases or points ( many of them have fermi surfaces but are _ not _ fermi liquids )
is very limited .
our work can be viewed as a step in that general direction .
furthermore , the formalism developed in this work has potential applicability to systems with composite or emergent fermions with fermi surfaces as well , or more generally , _ non - fermi liquid _ phases with fermi surfaces .
the system studied in ref.@xcite , where there is an emergent spinon fermi surface , is a potential example . this work is supported support by national science foundation under nsf grant no .
dmr-0704133 and dmr-1004545 ( w.x.d . and k.y . ) .
as is supported by the national science foundation under nsf grant no .
throughout the appendix , we shall denote the modified bessel functions @xmath336 , @xmath337 as @xmath338 , @xmath339 for simplicity with @xmath340 .
we also have @xmath341 which shall be shortened as @xmath342 . &
^(1 ) g_k / n = dr_x dr_z r_x r_z g_k / n ( r_x , r_z ) + & ( _ r_z^2 - ) g_k / n ( r_z , r_x ) . expanding @xmath343 , and noting the identities @xmath344 , @xmath345 where @xmath346 , we get & ( _ r_z^2 - ) g_k / n ( r_z , r_x ) + = & - + ( r_x - r_z ) ( 1 - _ r_z)i_zk_x + & + ( r_z - r_x ) ( 1 - _ r_z)i_xk_z . then integrating over @xmath347 first , and making use of the following formula dx x x_^2(x ) = x^2 ( x_^2(x ) - x_- 1(x ) x_+1 ( x ) ) , where @xmath348 can be the first or second kind of modified bessel function , @xmath349 or @xmath350 , and the identities @xmath351 in addition to those given above , @xmath335 is reduced to [ eq : simplified_dg ] & ^(1 ) g_k / n = dr r(- i k + r^2 ( i^2-i_+i_- ) k^2 ) .
we apply the same strategy as in sec .
[ subsec : ee_of_free_fermion_rev ] , making use of the e - m formula to do the sum over the @xmath134-index in eq . .
this converts the sum into a divergent integral over @xmath134 which cancels in eq . as before , and
a sum over terms of the form @xmath352 that turn out to vanish , as we will now show .
again , we can include proper constants under the derivative into the integrand .
these derivatives then act on well defined integrals .
the first term has been discussed in sec .
[ subsec : ee_of_free_fermion_rev ] : @xmath353 the second term can be shown to be @xmath354 summing the two terms together , we find ( _ k^j ^(1 ) g_k / n)_k0 = _ k^j ( - ) = 0 , for all @xmath355 .
let us first compute the integral : & h(r , r_1 ) = - i_1k + ( f_1(r , r_1 ) + ki_1 ) , + & f_1(r , r_1 ) = k k_1 ( r_1 ^ 2 ( i_1 ^ 2-i_1,+ i_1,- ) -i_1 ^ 2 ) - i i_1 + & ( r^2 ( k^2 - k_+ k_- ) - k^2 ) + k i_1 ( f(r ) - f(r_1 ) + & - i k + i_1 k_1 ) , + & f(r ) = 2 dr r i k = r^2 i k + r^2 i_+ k_- . the @xmath357 term in @xmath358 results from @xmath199-functions ( derivatives of @xmath199-functions ) coming from the derivative applied on the step function ( @xmath359 ) . the remaining part comes from terms involving a product of two @xmath46-functions . here
one needs to distinguish between @xmath360 and @xmath361 , which gives rise to the terms in @xmath362 and @xmath363 , respectively .
the remaining integral can be carried out by applying the identities of bessel functions @xmath242 and @xmath364 used in appx .
a. in applying the e - m formula to the sum over @xmath134 in , we again arrive at a divergent k - integral that can be rescaled and subsequently canceled ( see and below ) , and a sum over derivative terms that are well - behaved . in the latter terms , we always add proper constants under the derivatives to regularize the integrand at infinity , as before .
we divide into two terms .
the first is the one containing the @xmath199-function . after integrating out @xmath365 ,
this term becomes the second term is expanded to @xmath366 due to the complexity of @xmath367 , we examine each of the three terms of @xmath367 separately .
the first term is simple . applying those identities of @xmath364 s and @xmath242 s and including the proper constant , we get [ eq : h1 ] h. li and f. d. m. haldane , _ entanglement spectrum as a generalization of entanglement entropy : identification of topological order in non - abelian fractional quantum hall effect states _ ,
* 101 * , 010504 ( 2008 ) . a preliminary discussion of this idea , as well as its connection with multi - dimensional bosonization was presented earlier by one of us@xcite .
a similar idea has been independently discussed by b. swingle@xcite , based on a gapless mode - counting argument which is equivalent to our toy model . in this work ,
we construct the local degrees of freedom explicitly via high - dimensional bosonization , and use the constructed bosonic theory to compute the entanglement entropy directly .
we also generalize the calculation to include fermi liquid interaction in a special geometry , which is non - trvial and shows subleading contribution @xmath368 that can not be accounted for by mode - counting .
f. d. m. haldane , _
luttinger s theorem and bosonization of the fermi surface _
, in _ proceedings of the international school of physics `` enrico fermi '' , course cxxi `` perspectives in many - particle physics '' , r. a. broglia and j. r. schrieffer eds . , north - holland , amsterdam ( 1994 ) _ , pp 5 - 29 ( 1993 ) ; arxiv : cond - mat/0505529 .
there is a recently announced proof of widom s conjecture by a. v. sobolev , _ quasi - classical asymptotics for pseudodifferential operators with discontinuous symbols : widom s conjecture _ , , * 44*:4 ( 2010 ) , 86 - 90 . | the logarithmic violations of the area law , i.e. an area law " with logarithmic correction of the form @xmath0 , for entanglement entropy are found in both 1d gapless fermionic systems with fermi points and for high dimensional free fermions .
the purpose of this work is to show that both violations are of the same origin , and in the presence of fermi liquid interactions such behavior persists for 2d fermion systems . in this paper
we first consider the entanglement entropy of a toy model , namely a set of decoupled 1d chains of free spinless fermions , to relate both violations in an intuitive way .
we then use multi - dimensional bosonization to re - derive the formula by gioev and klich [ phys .
rev .
lett .
96 , 100503 ( 2006 ) ] for free fermions through a low - energy effective hamiltonian , and explicitly show the logarithmic corrections to the area law in both cases share the same origin : the discontinuity at the fermi surface ( points ) . in the presence of fermi liquid ( forward scattering )
interactions , the bosonized theory remains quadratic in terms of the original local degrees of freedom , and after regularizing the theory with a mass term we are able to calculate the entanglement entropy perturbatively up to second order in powers of the coupling parameter for a special geometry via the replica trick . we show that these interactions do not change the leading scaling behavior for the entanglement entropy of a fermi liquid . at higher orders , we argue that this should remain true through a scaling analysis . |
the flexibility and bistability are the two key points in the next generation display technologies .
the flexibility implies the display would be thin , lightweight , and ultimately , paper like , meaning it would be cheap enough to be disposable .
the bistability implies the technology would be ecologically friendly . in the bistable display
, the image does not need to be refreshed until rewritten and , therefore , a low level of power consumption is expected for still images .
unfortunately , for motion pictures , the bistability has no advantages in power savings over nonbistable technologies such as liquid crystal displays ( lcds ) , plasma display panels ( pdps ) , and displays based on organic light emitting diodes commonly referred to as oleds .
the display technologies based on particles are the most prominent candidates for a flexible and bistable displays .
the earliest display based on particles dates back to 1970 s when ota@xcite filed for a patent .
since then various particle - displays based on electrophoresis and electrowetting principles have emerged to form what is now referred to as `` e - paper technologies '' in the industry.@xcite in electrophoresis , particles are usually suspended in a fluid . because the speed at which particles move inversely vary with fluid density , particle - displays based on electrophoresis have slow response time , typically on the order of @xmath0@xcite this makes motion pictures unsuitable for electrophoretic particle - displays .
the issue of slow response time in electrophoretic particle - displays , however , has been resolved with the unveil of quick response liquid powder display ( qr - lpd ) by hattori et al.@xcite the qr - lpd is distinguished from the rest of particle based displays in that it uses air as the particle carrying medium rather than fluid . because the particles in qr - lpd move in air , its response time is at @xmath1 which is even faster than lcds .
the submillisecond response time makes qr - lpd the only candidate based on particle - display capable of handling motion pictures , and researches are being conducted for particle - displays with air as the particle carrying medium.@xcite common to all particle - displays , regardless of whether air or fluid is used as the particle carrying medium , is the lack of continuous grayscale . here ,
the terminology , `` continuous grayscale , '' is referred to as the number of grayscale levels required to produce desired number of colors . in principle , a display device with continuous grayscale can generate an infinite range of colors .
that being clarified , a high grayscale range is essential to quality displays . without it , the images displayed on monitors
would be dull.@xcite recently , it has been reported particle - display based on qr - lpd can generate up to 16 grayscale levels ( i.e. , 4 bits ) , which corresponds to the capability of generating 4096 colors.@xcite to the proponents of other competing flat panel display technologies , such as flexible lcds , pdps , or oleds , 4 bits of gray scale range could hardly be considered a technological milestone . however , considering only 4 grayscale levels were possible just a few years ago for particle - displays , 16 grayscale levels is a significant technological advancement for the e - paper technology.@xcite the most well known e - paper technology , e - ink , obtains different gray states through modulation of voltages supplied to the control electrodes .
the same mechanism is employed by particle - displays based on qr - lpd for achieving different gray states.@xcite because the voltage is modulated at a value lower than the saturation voltage , which is the voltage required to display either all black or all white for a simple black - and - white display , the displayed image would have intensity somewhere between that of completely black and completely white .
this approach to achieve different gray states , however , is done at the cost of losing bistability . since the pixel is being constantly modulated to sustain a gray state , the situation is equivalent to motion pictures and the power savings from bistability no longer applies to gray states . in spite of the lost bistability advantages over the other competing flexible display technologies ,
the speed at which particle can be modulated is limited by its finite inertia ( mass ) and this places a practical limit on the extent to which the grayscale levels of particle - display can be enhanced by the aforementioned method .
most recently , chim@xcite demonstrated a 64 grayscale levels ( or 6bits ) for particle - display based on qr - lpd for his masters thesis project at delft university of technology .
however , to generate 64 grayscale levels , the data driver , which is a serial to parallel shift register that shifts data words of 6 bits per clock cycle , is required .
( color online ) transflective display based on charged particles . ] in this work , a display architecture based on charged particles is presented . unlike the previous particle - display technologies , the proposed device has potential to generate continuous grayscale levels without sacrificing the benefit of bistability .
also , the proposed particle - display technology , in principle , does not require complicated voltage modulation schemes to generate different gray states.@xcite
the cross - sectional schematic of an optical shutter based on charged particles of same polarity is illustrated in fig .
[ fig : transflective_0 ] . in the figure ,
the two electrodes , where each is labeled top and bottom electrodes , constitute the control electrodes . because the light must be transmitted through the control electrodes , the electrodes are chosen from optically transparent conductors .
the chamber , wherein the particles reside , can be a vacuum , filled with noble gas , or filled with air .
the metallic reflectors , which forms the lateral surface of the chamber , are electrically connected to one of the control electrodes .
this make metallic reflectors not only to reflect light , but also function as to keep particles from aggregating to the lateral surface of chamber . the optically transparent passivation layer , which is treated on the inner surface of the chamber , functions to prevent charge transfer between particles and conductors . for positively charged particles inside the chamber , the hydrophobic treatment on the surface of the passivation layer ,
which is not explicitly shown in the figure , for example , prevents particles from sticking to the surface .
( color online ) transmission of light through a medium filled with total of @xmath2 suspended spherical particles . both of the compressed and
the uncompressed cases are shown . ]
the intensity of light transmitted through a medium filled with suspended spherical particles , illustrated in fig .
[ fig : cylinder_0 ] , is given by @xmath3,\label{eq : transmitted_intensity_bh}\ ] ] where @xmath4 and @xmath5 are , respectively , the real and the imaginary part of the complex refractive index for the medium and the particles suspended in it ; @xmath6 is the initial intensity , @xmath7 is the wavelength of incidence light , @xmath8 is the height ( or thickness ) of the volume containing charged particles , and @xmath9 is the volume fraction of particles.@xcite in explicit form , the volume fraction of particles is @xmath10 where @xmath2 is the total number of particles in the optical chamber , @xmath11 is the volume of a single particle , @xmath12 is the height of compressed cylinder containing charged particles , and @xmath13 is the cross - sectional area of the cylindrical chamber illustrated in fig .
[ fig : cylinder_0 ] . throughout this work
, i shall refer to terms such as `` compressed state '' or `` compressed particle volume '' to denote the compression of the volume containing particles .
insertion of @xmath9 into eq .
( [ eq : transmitted_intensity_bh ] ) gives @xmath14 where @xmath15 is just a constant .
the complex dielectric constant , @xmath16 and the complex refractive index , @xmath17 are expressed in form as @xmath18 where @xmath19 and @xmath4 are the real parts , and @xmath20 and @xmath5 are the imaginary parts . for the complex refractive index , @xmath4 represents the real refractive index and @xmath5 is the extinction coefficient ( or the absorption coefficient ) .
the two , @xmath21 and @xmath17 are related by the expression @xmath22 with the real and imaginary parts inserted for @xmath21 and @xmath23 from eq .
( [ eq : complex - optical - constants ] ) , i have @xmath24 the resulting expression can be rearranged to yield @xmath25 equation ( [ eq : complex - optical - constants - algebra ] ) can only be true if and only if the real and the imaginary parts are equal to zero independently .
this requirement yields the two expressions connecting @xmath26 to the optical constants @xmath27 and , the two expressions are @xmath28 stoller et al.@xcite measured the complex dielectric constant , @xmath29 for the single gold nanoparticle . for the gold nanoparticles of diameters @xmath30 and @xmath31 assuming a spherical morphology , they observed a reasonably good correspondence existing between complex dielectric constants of the bulk gold and the gold nanoparticles for the wavelength range of roughly @xmath32 to @xmath33 their finding is of significant importance , as it allows the bulk gold dielectric constants , which is readily available , to be used for the gold nanoparticles , which is not so readily available . since the real and the imaginary parts of the complex dielectric constant are related to the optical constants @xmath34 thru eq .
( [ eq : e1-e2-complex - dielectric - constant ] ) , the @xmath4 and @xmath5 for the gold nanoparticles are readily available once @xmath19 and @xmath20 are known from the bulk counterpart .
the vice versa is also true , of course .
fortunately , johnson and christy measured the optical constants for the bulk copper , silver , and the bulk gold.@xcite justified by the findings of stoller et al.,@xcite for @xmath35 assuming the gold nanoparticle of radius @xmath36 the @xmath4 and @xmath5 measurements from the work of johnson and christy,@xcite @xmath37 can be utilized for the gold nanoparticles to plot eq .
( [ eq : transmitted_intensity ] ) for the transmitted intensity . for a spherical gold nanoparticle , the particle volume ,
@xmath38 is given by @xmath39 where @xmath40 is the particle radius .
assuming the diameter of @xmath41 for the gold nanoparticle , @xmath11 becomes @xmath42 to plot eq .
( [ eq : transmitted_intensity ] ) , i shall assume the following values for the chamber parameters , ( the height @xmath8 and the cross - sectional area @xmath13 ) , and the particle number , @xmath43 @xmath44 schematic of curves illustrating the intensity , @xmath45 of transmitted light as a function of compression height , @xmath46 ] before going ahead with plotting eq .
( [ eq : transmitted_intensity ] ) , it is worthwhile to double check if the total number of particles , @xmath43 specified in eq . ( [ eq : han ] ) is reasonable . with chamber parameters as defined in eq .
( [ eq : han ] ) , the volume of the cylindrical chamber is @xmath47 suppose the charged spherical particles can be compressed and enclosed in a confined volume in such a way that neighboring particles actually touch each other .
how many particles , under such restrictions , can be fitted inside the chamber defined by eq .
( [ eq : han ] ) ?
the answer is @xmath48 and its expression is @xmath49 for the spherical particle of radius @xmath36 the @xmath48 is roughly @xmath50 for the charged particles , the condition of neighboring particles actually touching each other is not possible due to coulomb repulsion , unless the external compression force is infinite .
nonetheless , the expression for @xmath51 defined in eq .
( [ eq : nmax_particle_num ] ) , provides the upper limit for @xmath2 inside the chamber .
the physical optical shutter based on charged particles must be compressible in order to allow variations in transmission intensity , which imposes the condition , @xmath52 equivalently , the variability of transmission intensity thru compression requires @xmath53 and this further modifies the condition for @xmath2 as @xmath54 the inequality condition for @xmath43 defined in eq .
( [ eq : zero_n_nmax ] ) , can be understood from the illustration shown in fig .
[ fig : transmission_intensity_schematic ] , where some of the possible curves for the transmission intensity , @xmath45 as a function of the compression height , @xmath55 are shown .
the cases where @xmath56 and @xmath57 are represented by the curves corresponding to @xmath58 at @xmath59 and @xmath60 respectively .
similarly , the the curve corresponding to @xmath58 at @xmath61 in the figure represents the case where @xmath62 for @xmath63 the particles in the chamber are already maximally compressed and , therefore , no light gets transmitted . on the other hand , for @xmath64
all light gets transmitted through the optical shutter , as the inside of the optical shutter is a void .
array of metallic nanoparticles in grid formation .
the parameter @xmath65 is the grid ( or particle - particle ) spacing and @xmath7 is the wavelength of the incidence light . for the charged particles inside the chamber ,
the parameter @xmath65 can only be thought of as the time averaged mean particle - particle distance between the nearest neighboring particles .
nonetheless , the grid representation illustrated here serves the purpose of initiating the crude connection between the wavelength of incidence light , @xmath66 and the total particle number , @xmath67 ] for the case of any finite @xmath2 satisfying the condition defined in eq .
( [ eq : zero_n_nmax ] ) , the curve for the transmission intensity must necessarily lie in the region between the two extreme cases , @xmath56 and @xmath63 as schematically demonstrated in fig . [
fig : transmission_intensity_schematic ] .
three such curves , corresponding to finite @xmath43 are illustrated in the figure : ( 1 ) the upper curve represented by @xmath68 ( 2 ) the curve represented by @xmath69 and ( 3 ) the lower curve represented by @xmath68 where @xmath70 is a constant and @xmath71 is a function of @xmath46 the two curves corresponding to @xmath72 only appreciably varies in transmission intensity with compression within the window of @xmath73 where @xmath74 such curves are associated with @xmath43 in which the @xmath2 is finite but close to either @xmath56 or @xmath62 because the transmission intensity only appreciably varies within @xmath75 for such choices of @xmath43 the gray states are difficult to achieve as it requires very precise control of the compression height . for example , assuming the particles can be compressed at the increment in @xmath76 the fact that @xmath77 makes it difficult to achieve gray states
. to achieve gray states , the @xmath76 which defines the sensitivity of compression , must be much smaller than @xmath78 consequently , the transmission intensity curves for the aforementioned choices of @xmath2 are only good for displaying either completely bright or completely dark transmission states .
the nanoparticles of radius @xmath40 are separated by @xmath79 which is the nearest surface to surface separation between two neighboring particles .
the imaginary particles of radius @xmath80 are , however , in closed packed formation for a given nearest surface to surface separation of @xmath81 ] the optimal design for the presented optical shutter is achieved by choosing @xmath2 for the total particle number in the chamber in such way that the curve for the transmission intensity goes like @xmath69 where @xmath70 is a constant , in fig . [
fig : transmission_intensity_schematic ] . because the curve is linear for the transmission intensity
, the window of range in which compression can be done is maximized , @xmath82 for this particular choice of @xmath43 the @xmath76 which is the increment at which particles can be compressed , is much less than @xmath83 i.e. , @xmath84 and the number of different gray states that can be achieved is given by @xmath85 in principle , the @xmath86 can be made finer and finer as desired . in reality , the fineness of @xmath86 is limited by the system design . nevertheless , with the right choice of @xmath43 the grayscale levels in number of @xmath87 as defined in eq .
( [ eq : ngray ] ) , is possible with the presented optical shutter based on charged particles .
the question remains to be answered is this : how do we go about obtaining the right @xmath2 ? to answer this , i shall refer to the illustration shown in fig .
[ fig : grid ] . due to the wave nature of light , the total particle number and the transmission intensity depend on the wavelength , @xmath66 of the incidence light . as a rudimentary assumption , the transmission loss of an electromagnetic wave passing through an array of metallic particles decreases for @xmath88 and increases for @xmath89 where @xmath90 which is schematically illustrated in fig .
[ fig : grid ] .
the case where @xmath88 represents the situation in which @xmath2 is very small , whereas the case where @xmath89 represents the situation in which @xmath91 provided the @xmath92 is not too large .
it is , therefore , not too bad to impose the condition , @xmath93 for @xmath65 in estimating for the total number of particles in the chamber . for @xmath35
this condition for @xmath65 yields the value of @xmath94 to figure out exactly how man particles can be fitted inside the chamber under restriction @xmath93 the fig .
[ fig : particle_num ] is referred to .
since @xmath65 is the closest distance between surfaces of two nearest neighbor particles , i shall visualize an imaginary particle of radius @xmath95 @xmath96 assuming the imaginary particles are close - packed inside the chamber , the problem becomes identical to the previous case , which resulted in eq .
( [ eq : nmax_particle_num ] ) . with @xmath97 of eq .
( [ eq : r_imaginary ] ) inserted for @xmath40 in eq .
( [ eq : nmax_particle_num ] ) , the expression becomes @xmath98 where @xmath48 has been replaced by @xmath99 with @xmath100 defined in eq .
( [ eq : nmim_particle_num ] ) , the number of particles inside the chamber may be chosen according to the inequality , @xmath101 it can be easily verified that @xmath102 for @xmath103 which corresponds to the half wavelength of @xmath35 @xmath100 is roughly @xmath104 this value for @xmath100 is much larger than zero , but it is much smaller than @xmath51 which has the value @xmath105 from eq . ( [ eq : nmax_number ] ) . for the optical shutter involving charged particles of spherical morphology and the cylindrical chamber of specifications defined in eq .
( [ eq : han ] ) , the inequality condition for @xmath43 eq . ( [ eq : n_condition ] ) , becomes @xmath106 equation ( [ eq : transmitted_intensity ] ) has been computed using gold nanoparticles as the charged particles ( the gold nanoparticle was chosen only because its optical constant data , @xmath4 and @xmath107 were readily available ) . for the charged particles and the incidence light , the parameters defined in eq .
( [ eq : bulk - gold - n - k ] ) were used .
the parameters defined in eq .
( [ eq : han ] ) were used for the chamber specification . using eq .
( [ eq : transmitted_intensity ] ) , the transmission intensity for different values of @xmath2 were considered and the results are shown in fig .
[ fig : transmission_intensity ] .
the @xmath108 curve is the case where particle number is relatively low in the chamber .
as it can be observed , for low particle numbers in the chamber , the intensity does not vary well with compression except for small @xmath55 which is consistent with the curve represented by @xmath72 in fig .
[ fig : transmission_intensity_schematic ] .
contrarily , the @xmath109 curve corresponds to the case where too many particles are inside the chamber .
although the transmission intensity varies linearly with compression , which is a good characteristic of an optical shutter , the output intensity is far too low even for the brightest state . for @xmath110@xmath111 the brightest state only transmits @xmath112 of the initial input intensity , i.e. , @xmath113 as expected , the @xmath114@xmath115 curve in fig .
[ fig : transmission_intensity ] , which corresponds to the case where @xmath116 for @xmath117 most resembles the curve represented by @xmath118 in fig .
[ fig : transmission_intensity_schematic ] .
however , the brightest output intensity is only @xmath119 of the input intensity of the incidence light . compared to lcds , where only @xmath120 of the intensity of the incidence light from back light unit gets transmitted , the output intensity of @xmath119 is already @xmath30 times more efficient than lcds .
( color online ) transmission intensity as a function of particle number , @xmath43 and the compression height , @xmath46 the abbreviation `` mil '' in the figure denotes a million , i.e. , @xmath121 ] the same principle , which is inherent in eq .
( [ eq : transmitted_intensity ] ) , applies to the proposed optical shutter , fig .
[ fig : transflective_0 ] . in the presented device
, the chamber is filled with charged particles of same polarity and this can be identified with the medium filled with suspended particle in fig .
[ fig : cylinder_0 ] . the on set of electric field inside the chamber , which is done by controlling the voltage over one of the electrodes , causes particles to be compressed in volume , as illustrated in fig .
[ fig : transflective_1 ] .
since the particles are assumed to be positively charged , they are compressed in the direction of electric field .
eventually , the compression comes to a stop when particle - particle coulomb repulsion counterbalances the compression induced by the control electrodes . in principle
, the level of compression for the particle volume can be varied continuously . because the intensity of transmitted light varies with compression , i.e. , eq .
( [ eq : transmitted_intensity ] ) , the display based on charged particles has the potential to generate continuous gray levels . also , since the control electrodes form a capacitor , provided there is no leakage ( or negligible ) current across the capacitor , the electric field inside the chamber can be sustained even when the device is disconnected from power .
this opens up the possibility of a bistable mode for gray states as well .
the presented device works as an optical shutter , provided the charged particles can be effectively prevented from piling up at the surface of dielectric walls .
when a charged particle is brought close to the dielectric surface , the bound charges within the dielectric get redistributed in order to reduce the field originating from the charged particle placed near vicinity of dielectric surface .
such is illustrated in fig .
[ fig : bc1 ] . for the case where a positively charged particle is placed near the surface of a dielectric ,
the surface bound charges of opposite polarity get induced and distributed near the inner surface of dielectric . as a result ,
the positively charged external particle gets pulled to the dielectric surface and , eventually , sticking to the surface of dielectric , which process is illustrated in fig . [ fig : bc2 ] .
if there are more than one positively charged particles placed at the vicinity of dielectric surface , the aforementioned process continues and , eventually , layers get formed , for example , the layers @xmath13 and @xmath122 in the figure .
this process does not continue indefinitely , however , as the bound charges of opposite polarity within the dielectric get eventually shielded by the positively charged particles forming layers over the dielectric surface . because the particles in each layer are charged with same polarity
, the coulomb repulsion keeps the two particles from touching each other . assuming the layer @xmath122 is sufficient to shield completely the negatively charged surface bound charges within the dielectric
, the remaining positively charged particles in the region between two dielectric walls would have no other places to go , except to continuously bounce back and forth within the region , which region has been indicated by @xmath123 in fig .
[ fig : bc2 ] .
( color online ) the on set of electric field compresses the filled volume for charged particles . ] the aforementioned description for an optical shutter , however , has a serious problem which is associated with the particle layers forming on the surface of dielectric walls . assuming the path of light propagation is along the horizontal axis
, the light enters the device from the left and exits at the right or vice versa . in principle , the compression of charged particles in region , @xmath124 controls the intensity of transmitted light .
the problem arises because the charged particles forming layers @xmath13 and @xmath122 on the surface of dielectric walls may no longer be optically transparent . as previously discussed , using the illustration in fig .
[ fig : grid ] as an example , metallic particles in an array of grid formation severely reduces the intensity of transmitted light for grid spacing , @xmath79 much less than the wavelength , @xmath66 of the incidence light .
i now discuss the ways to resolve this complication .
( color online ) charged particles near dielectric wall . ]
( color online ) charged particles in region @xmath123 are shielded from the negatively charged surface bound charges within the dielectric wall . ]
( color online ) transparent charged layer formed by isam or electron beam irradiation method effectively plays the role of positively charged particle layers at @xmath13 and @xmath125 ] it is well known that the surface of a typical @xmath126 glass becomes hydrophilic in an open air as a result of the silane group at the surface of glass combining with oxygen.@xcite this has the effect of making the surface of glass to be slightly negatively charged and any positively charged particles in vicinity would be attracted to the glass surface forming layers such as those , e.g. , @xmath13 and @xmath127 illustrated in fig .
[ fig : bc2 ] . unless the attracted , positively charged , particles are optically transparent
, the device would not be able to function as an optical shutter for no light would to pass through the device .
necessarily , the role of layers @xmath13 and @xmath122 in fig .
[ fig : bc2 ] must be played out by particles of same polarity as those residing in region @xmath124 but distinguished from those residing in region @xmath123 in that they are optically transparent .
one way to achieve this is to chemically treat the surface of glass so as to make it hydrophobic .
the hydrophobic treatment of glass electrically neutralizes the glass surface , thereby significantly reducing the oppositely charged particles from sticking to the glass surface . for large and weakly charged positive particles , simple electrical neutralization of glass surface by hydrophobic treatment is sufficient to keep particles from permanently sticking to the surface . in such system , charged particles
are continuously bounced off of the wall and do not stick to the surface . however , for smaller and well charged positive particles , simple hydrophobic treatment of glass surface is not sufficient to prevent oppositely charged particles from sticking to the glass surface . for such cases , it is necessary to make the surface of glass net positively charged .
this can be easily done by utilizing the technique known as `` ionically self assembled monolayer '' ( isam ) , in which technique charged particles , polymers , or monomers of preferred polarity are physically attached to the surface.@xcite alternatively , and more directly , the glass may be irradiated with electron beam to physically embed charged particles inside the glass medium.@xcite in this latter method , for example , net negative charges may be made to accumulate inside the glass .
this has the effect of inducing positive charges on the surface of glass , which repels positive charged particles in vicinity of the glass surface , thereby preventing particles from sticking to the surface . in summary ,
the role of positively charged particle layers , @xmath13 and @xmath122 in fig .
[ fig : bc2 ] , may be effectively get taken cared by the aforementioned isam or the electron beam irradiation methods ; and , using these alternative techniques , the glass surface repels charged particles inside the chamber and it is optically transparent , as illustrated in fig .
[ fig : bc3 ] . to utilize charged particles in displays , a quantitative understanding of how design parameters , such as @xmath128 for the sub - pixel dimensions and @xmath129 for the charged particle , enter into the compression mechanism is required . here
, @xmath130 is the particle radius ( assuming a spherical particle ) , @xmath131 is the particle mass density , and @xmath132 is the net charge which the particle holds . describing the compression mechanism in terms of the aforementioned design parameters is the task for the next section . the optical shutter presented in this proposal relies on the density of particles in chamber to control the intensity of transmitted light .
an analogy can be made to the driving under misty weather .
when the density of water vapor suspended in the atmosphere is heavy , one is obscured in his or her viewing distance , as less light reaches the eye .
contrarily , the amount of light reaching the eye increases with a reduction in density of water vapor suspended in the atmosphere , thereby enabling the driver to see far distances . the particles in chamber of the proposed optical shutter ranges in diameter anywhere from a few nanometers to several microns , assuming a spherical morphology .
this range for the particle size , although small macroscopically , is much too large to be considered for a treatment within quantum domain , where the quantum theory must be used for a description .
therefore , the classical theory suffices for the description here . since the charged particles in the system , as a whole ,
behave like a classical gas , the description is carried out in the realm of statistical physics . to keep the topic presented here self - contained
, i shall briefly summarize the kind of manipulations and approximations assumed in obtaining expressions which are considered crucial to the initial development of the analysis .
the charged particles in chamber can be treated as classical particles obeying the maxwell - boltzmann statistics.@xcite the @xmath133 charged particle under influence of external forces , for example , gravitational and electric forces , assumes the energy @xmath134 where @xmath135 is the center of mass momentum for the @xmath133 charged particle and the term associated with it is the kinetic energy , @xmath136 is the interaction energy with external influences , and @xmath137 is the energy contribution arising only if the particle is not monatomic . in explicit form , @xmath136 can be expressed as @xmath138^{1/2}}+m_{i}gz_{i}\\ & + q_{i}ez_{i},\end{aligned}\ ] ] where @xmath2 is the number of particles in volume , the @xmath139 and @xmath140 denote respectively the net charges for the @xmath133 and @xmath141 particles , @xmath142 is the electric field magnitude , and the constants @xmath143 and @xmath144 in mks system of units . for a monatomic particle , the energy contributions from the internal rotation and vibration with respect to its center of mass vanishes
, @xmath145 therefore , the @xmath133 monatomic charged particle under the influence of external forces assumes the energy @xmath146^{1/2}},\label{eq : ui}\end{aligned}\ ] ] where , for convenience , all particles in the system are assumed to be identically charged with same polarity , i.e. , @xmath147 the gravity and electric field have directions , and this information must be taken into account in eq .
( [ eq : ui ] ) . to do this ,
the parameter @xmath148 is first restricted to a domain @xmath149 with @xmath148 restricted to a domain defined by @xmath150 the gravitational potential energy of a particle , @xmath151 increases with positive @xmath152 and decreases with negative @xmath152 as @xmath148 increase .
therefore , the direction of gravity in eq .
( [ eq : ui ] ) can be taken into account by @xmath153 the electric potential energy of a particle , @xmath154 increases with positive @xmath142 and decreases with negative @xmath142 as @xmath148 increase . for a positively charged particle , its electric potential energy increases as it moves against the direction of electric field and decreases as it moves in the direction of electric field .
the direction of electric field inside the chamber can hence be taken into account in eq .
( [ eq : ui ] ) by @xmath155 the probability of finding the particle with its center of mass position in the ranges @xmath156 and @xmath157 can be expressed as @xmath158 where @xmath159 is the temperature in units of degree kelvin @xmath160 and @xmath161 is the boltzmann constant , @xmath162 with eq .
( [ eq : ui ] ) , @xmath163 becomes @xmath164^{1/2}}\right)\nonumber \\ & \times\exp\left[-\left(\frac{m_{i}g+qe}{k_{\textup{b}}t}\right)z_{i}\right]d^{3}\mathbf{r}_{i}.\label{eq : ps1}\end{aligned}\ ] ] the presence of repulsive coulomb interaction , @xmath165^{1/2}},\ ] ] makes eq .
( [ eq : ps1 ] ) difficult and this term must be approximated .
the configuration depicted in fig .
[ fig : cylinder ] is referred for the analysis .
the @xmath133 particle is the top most particle and all other particles are ahead of it in the direction of compression .
[ fig : cylinder ] ] as the @xmath133 particle gets compressed in the direction of electric field , it experiences net electric field given by @xmath166 where @xmath167 is the electric field inside the chamber generated by external electrodes and @xmath168 is the electric field produced by all other particles inside the chamber . because @xmath167 and @xmath168 are oppositely directed , the magnitude @xmath169 is given by @xmath170 to estimate @xmath171 fig .
[ fig : cylinder ] is considered .
the vectors @xmath172 @xmath173 and @xmath174 satisfy @xmath175 in cylindrical coordinates , @xmath176 @xmath177 and @xmath174 become @xmath178 where @xmath179 represents the cylindrical coordinates for particle labeled as @xmath180 in fig .
[ fig : cylinder ] . with @xmath177 and @xmath174
thus defined , eq .
( [ eq : b_c_a ] ) becomes @xmath181 at @xmath182 the electric field contributed from particle labeled as @xmath180 is given by @xmath183}{\left[\rho_{2}^{2}+\left(\eta+\varepsilon - z_{2}\right)^{2}\right]^{3/2}}.\end{aligned}\ ] ] because both gravitational and electrical forces are assumed to depend on @xmath184 coordinate only , the @xmath185 and @xmath186 components of @xmath187 average to zero to become @xmath188^{3/2}}\mathbf{e}_{3}.\ ] ] all particles in the cylinder , not just the particle labeled as @xmath180 in fig .
[ fig : cylinder ] , contributes to form @xmath189 hence , @xmath190^{3/2}}\mathbf{e}_{3},\label{eq : e_total_0}\ ] ] where @xmath2 is the number of charged particles in the chamber . for @xmath2
sufficiently large , the coordinates @xmath191 and @xmath192 can be replaced by @xmath193 and @xmath194 in the continuum limit , the summation symbol gets replaced by @xmath195 and the charge @xmath132 is replaced by @xmath196 where volume is that of cylinder illustrated in fig .
[ fig : cylinder ] and @xmath197 is the total charge inside it . in the continuum limit then , where @xmath2 is assumed to be sufficiently large , eq .
( [ eq : e_total_0 ] ) can be approximated by @xmath198^{3/2}}\mathbf{e}_{3},\end{aligned}\ ] ] which result , integrating over the @xmath199 becomes @xmath200^{3/2}}\mathbf{e}_{3}.\label{eq : e_total_1}\end{aligned}\ ] ] with the change of variable , @xmath201 the @xmath202 integral in eq . ( [ eq : e_total_1 ] ) becomes @xmath203^{3/2 } } & \rightarrow\frac{\left(\eta+\varepsilon - z\right)}{2}\int\frac{dx}{x\sqrt{x}}\\ & = -\frac{\left(\eta+\varepsilon - z\right)}{\sqrt{x}}.\end{aligned}\ ] ] with @xmath204 reverted back to the original variable , the @xmath202 integral becomes @xmath205^{3/2 } } & = \left.-\frac{\left(\eta+\varepsilon - z\right)}{\sqrt{\rho^{2}+\left(\eta+\varepsilon - z\right)^{2}}}\right|_{0}^{\xi}\\ & = 1-\frac{\left(\eta+\varepsilon - z\right)}{\sqrt{\xi^{2}+\left(\eta+\varepsilon - z\right)^{2}}}.\end{aligned}\ ] ] insertion of the result into eq .
( [ eq : e_total_1 ] ) yields @xmath206 with the change of variable , @xmath207 the @xmath184 integral in eq . ( [ eq : e_total_2 ] ) becomes @xmath208 with @xmath209 reverted back to the original variable , the @xmath184 integral becomes @xmath210 insertion of the result into eq .
( [ eq : e_total_2 ] ) gives the @xmath211 @xmath212\mathbf{e}_{3}.\end{aligned}\ ] ] since the @xmath184 coordinate of the @xmath133 particle is given by @xmath213 the expression for @xmath168 may be rewritten in terms of @xmath148 as @xmath214\mathbf{e}_{3}.\end{aligned}\ ] ] the parameter @xmath215 has been introduced for a mathematical convenience to assure that the @xmath133 particle is the upper most particle residing at the top surface of the compressed volume . taking the limit @xmath216
the previous expression for @xmath168 becomes @xmath217 since @xmath148 is the @xmath184 coordinate for the @xmath133 particle , which is the particle residing at the top surface of the compressed volume in fig . [
fig : cylinder ] , the @xmath168 defined in eq .
( [ eq : e_repulsion ] ) represents the coulomb repulsion acting on the particle residing at the top surface of the compressed volume from all other ones within the compressed volume .
insertion of eq .
( [ eq : e_repulsion ] ) into eq .
( [ eq : e_net ] ) gives @xmath218 the @xmath219 in current form is only an approximation because the expression for @xmath211 eq .
( [ eq : e_repulsion ] ) , is an approximation .
the equality can be made by replacing @xmath220 where @xmath221 is the effective charge to be determined experimentally . with @xmath222
the expression for @xmath219 becomes @xmath223 where the subscript @xmath224 of @xmath225 has been dropped for convenience .
what is the implication of @xmath219 ? the @xmath226 of eq .
( [ eq : ui ] ) , which is the energy term assumed by the @xmath133 charged particle under the influence of external forces , can be rearranged in form as @xmath227^{1/2}}\vphantom{\frac{\frac{\frac{1}{1}}{\frac{1}{1}}}{\frac{\frac{1}{1}}{\frac{1}{1}}}}\right\ } .\end{aligned}\ ] ] one notices that the term in the summation is the electric field contribution from the @xmath141 particle acting on the @xmath133 particle , @xmath228^{1/2}}.\end{aligned}\ ] ] furthermore , one finds @xmath229 and @xmath226 can be equivalently expressed as @xmath230 since @xmath231 eq .
( [ eq : e_net ] ) , the @xmath226 becomes @xmath232 insertion of eq .
( [ eq : ui_approximation ] ) into eq .
( [ eq : ps0 ] ) gives an alternate expression for probability density for finding particle with its center of mass position in the ranges @xmath156 and @xmath233 @xmath234d^{3}\mathbf{r}_{i}\nonumber \\ & \times\exp\left(-\frac{\mathbf{p}_{i}^{2}}{2m_{i}k_{\textup{b}}t}\right)d^{3}\mathbf{p}_{i},\label{eq : ps20}\end{aligned}\ ] ] which is different from the previous expression , eq .
( [ eq : ps1 ] ) , but now manageable . with eq .
( [ eq : e_net_final ] ) inserted for @xmath235 eq .
( [ eq : ps20 ] ) becomes @xmath236\exp\left(-\frac{\mathbf{p}_{i}^{2}}{2m_{i}k_{\textup{b}}t}\right)d^{3}\mathbf{r}_{i}d^{3}\mathbf{p}_{i},\label{eq : ps21}\end{aligned}\ ] ] equation ( [ eq : ps21 ] ) may be integrated over all possible @xmath204 and @xmath209 values lying within in the container and each components of the momentum may be integrated from @xmath237 to @xmath238 @xmath239dz_{i}\int_{x}\int_{y}dx_{i}dy_{i}.\label{eq : ps22}\end{aligned}\ ] ] the double integral over @xmath204 and @xmath209 gives slice area of the chamber at @xmath240 @xmath241 @xmath242 where @xmath243 is the radius of cylindrical chamber depicted in fig .
[ fig : cylinder ] .
the momentum integrals are obtained utilizing the well known integral formula@xcite @xmath244 where solutions are given by @xmath245 the momentum integrals become @xmath246 ] with eqs .
( [ eq : ps22a ] ) and ( [ eq : ps22b ] ) , eq .
( [ eq : ps22 ] ) becomes @xmath247dz_{i}.\end{aligned}\ ] ] with the following definitions , @xmath248 the previous expression for @xmath249 simplifies to @xmath250dz_{i},\label{eq : ps23}\end{aligned}\ ] ] where @xmath15 is the constant of proportionality to be determined from the normalization condition , @xmath251 it can be shown @xmath252dx_{i}\right\ } ^{-1},\label{eq : ccccc}\end{aligned}\ ] ] where @xmath253 is a dummy integration variable and it should not be confused with the coordinate @xmath204 of the cylinder .
what can be said about @xmath221 defined in @xmath70 ?
the effective coulomb repulsion from the remaining @xmath254 charged particles inside the chamber acting on the @xmath133 charged particle , see fig .
[ fig : cylinder ] , is proportional to @xmath255 where @xmath221 must be determined empirically from measurements . in principle , @xmath221 takes into account the spatial configuration of the @xmath254 charged particles in the system because it effectively describes the system , which is illustrated in fig .
[ fig : cylinder ] , in terms of the two body problem ( see fig . [
fig : cylinder_qeff ] ) . because the total charge in the imaginary cylinder must be conserved
, it must be true that @xmath256 and , this implies the condition @xmath257 for describing the trend of volume compression involving charged particles , eq . ( [ eq : qeff_condition ] ) provides the way to estimate @xmath258 once @xmath221 is defined , eq . ( [ eq : ps23 ] ) may be plotted for the most probable height of the compressed volume , which volume contains the @xmath2 charged particles in the system .
that being said , combining eqs .
( [ eq : ps23 ] ) and ( [ eq : ccccc ] ) , the probability density for the most probable height of the compressed volume containing @xmath2 charged particles becomes @xmath259dx_{i}\right\ } ^{-1}\exp\left[\alpha\left(\frac{1}{z_{i}}+\frac{1}{\xi}-\sqrt{\frac{1}{z_{i}^{2}}+\frac{1}{\xi^{2}}}\right)\right.\nonumber \\ & \left.-\beta z_{i}\vphantom{\sqrt{\frac{1}{z_{i}^{2}}}}\right]dz_{i},\label{eq : ps24}\end{aligned}\ ] ] where @xmath70 and @xmath260 are defined in eq .
( [ eq : alpha_beta ] ) . before plotting @xmath261
i shall explicitly define the charge @xmath262 particle mass @xmath263 and the electric field magnitude @xmath264 in nature , the charge is quantized and , therefore , it is convenient to express @xmath132 in terms of the charge number @xmath265 @xmath266 where @xmath267 is the fundamental charge unit and @xmath268 is the number of electrons removed from the particle . for the sake of simple analysis ,
the particles in the system are assumed to be spheres of identical radius .
the mass of each particle would then be given by @xmath269 where @xmath130 is the radius of sphere and @xmath131 is the mass density .
finally , the electric field generated in the chamber by control electrodes is @xmath270 where @xmath271 is the voltage applied to the top electrode ( the other electrode has been grounded ) .
the approximation @xmath272 in electric field comes about because the effects of passivation layers inside the chamber have been neglected for simplicity .
defined in eq .
( [ eq : ps24 ] ) .
[ fig : cylinder2 ] ] having defined @xmath262 @xmath263 and @xmath273 the configuration illustrated in fig .
[ fig : cylinder2 ] is referred to plot @xmath261 eq .
( [ eq : ps24 ] ) .
the parameters for the configuration are given the following values : @xmath274 for the purpose of plotting @xmath275 defined in eq .
( [ eq : ps24 ] ) , i shall assume , see eq .
( [ eq : qeff_condition ] ) , @xmath276 where @xmath132 is defined in eq .
( [ eq : q - def ] ) .
the voltages of @xmath277 @xmath278 @xmath279 @xmath280 and @xmath281 are considered for the top electrode ( the bottom electrode is grounded ) . with @xmath271
thus defined , the electric field generated inside chamber is given by eq .
( [ eq : e ] ) .
the gravity of @xmath282 was assumed inside the chamber .
the directions for the electric field @xmath283 and the gravitational force @xmath284 are determined from eqs .
( [ eq : g_direction ] ) and ( [ eq : e_direction ] ) .
since both @xmath142 and @xmath152 are positive , according to the convention defined in eqs .
( [ eq : g_direction ] ) and ( [ eq : e_direction ] ) , the electric field @xmath283 and the gravitational force @xmath284 are both directed in @xmath285 which is the negative @xmath184 axis .
the @xmath275 of eq .
( [ eq : ps24 ] ) is computed numerically utilizing simpson method for the integral.@xcite the simpson method routine was coded in fortran 90 .
that being said , the results are summarized in figs .
[ fig : probability ] and [ fig : probability_zoom ] , where the three smaller peaks of fig .
[ fig : probability ] are magnified and replotted in fig .
[ fig : probability_zoom ] . at @xmath277 that is , when there is no electric field inside the chamber other than the static fields from particles , the particles are distributed to occupy the entire volume of the chamber .
this is indicated by the peak occurring at the physical height of the chamber , @xmath286 at @xmath287 an electric field of roughly @xmath288 is generated inside the chamber . because particles are positively charged , they are compressed in the direction of electric field , i.e. , the @xmath289 direction .
this force , which induced particle volume compression , eventually gets counter balanced by the coulomb repulsion and the compression ceases . for the case where control electrode is held at @xmath287
the compression ceases at roughly @xmath290 ( see fig . [
fig : probability_zoom ] ) and this marks the most probable height of the compressed volume for the case . finally ,
with @xmath291 applied to the control electrode , the compressed volume state is reached where all particles are cluttered near the floor of the chamber , thereby resulting in very high particle density , as illustrated in fig .
[ fig : probability ] .
if charged particles are to be useful for any display applications , the charged particle system must be insensitive to gravitational effects , if not negligible .
the effect of gravity on the most probable height for the compressed volume has been investigated by reversing the direction of gravity ( but , keeping all other conditions unchanged ) in fig .
[ fig : probability_zoom ] .
the case where @xmath292 was selected for comparison .
the result is shown in fig .
[ fig : probability_g_effect ] , where it shows that the most probable height for the compressed volume is only negligibly affected by the gravity .
the influence of cylinder radius @xmath243 on the most probable height for the compressed volume has also been investigated .
again , the case of @xmath292 was selected from fig .
[ fig : probability_zoom ] for comparison by considering @xmath293 @xmath294 and @xmath295 all other conditions were kept unmodified .
the result , fig .
[ fig : probability_eta_effect ] , reveals a decrease in height for the most probable compressed volume with increasing @xmath243 as expected .
( color online ) most probable height of the compressed volume containing charged particles . ]
( color online ) the three small peaks in fig . [
fig : probability ] are magnified for detail .
each peak represents the most probable height for the compressed volume , where there are @xmath2 charged nanoparticles inside the volume . ]
( color online ) gravity has negligible effect on most probable height of the compressed volume . ]
( color online ) the height of the compressed volume decreases with the increased radius of the cylindrical chamber . ]
the intensity of light transmitted through a medium filled with charged particles goes like , eq .
( [ eq : transmitted_intensity ] ) , @xmath296 where both @xmath15 and @xmath12 have unit of meter , and @xmath12 is the height of the compressed volume containing charged particles . referring to fig .
[ fig : probability_zoom ] , the compressed height @xmath12 corresponds to the @xmath184 axis where the probability curve is maximum . in principle
, the compression height @xmath12 can be varied continuously by controlling the voltages applied to the control electrode .
this implies the intensity of light output from the proposed optical shutter can be varied continuously , thereby generating continuous grayscale levels for the device . in reality , the number of grayscale levels that can be achieved in the presented optical shutter is given by eq .
( [ eq : ngray ] ) , @xmath297 where the fineness of @xmath86 is limited by the system design .
this work is not the first kind to address potential applications with charged particles .
szirmai@xcite experimented with alumina powders to study electrosuspension as early as 1990 s . in szirmais@xcite
experiment , alumina powder of @xmath298 in diameter was placed inside an electrically insulating cylindrical vessel , which is similar in configuration with fig .
[ fig : cylinder2 ] .
the initial charging of alumina powder was done by a process of field emission , which can be achieved by applying high voltage to the control electrodes(field emission is the phenomenon in which electrons get emitted from the surface of host material , such as nanoparticles , due to the presence of high electric fields ) .
szirmai,@xcite however , does not quantitatively address the compression states of volume containing charged particles in terms of the design parameters , as his motive was not in discussing possible applications of charged particles for displays . in his experiment , the control electrodes , in principle ,
could be supplied with whatever high voltages required by it to do the job ; therefore , the quantitative understanding of how design parameters , such as @xmath299 @xmath300 @xmath301 @xmath302 @xmath303 @xmath304 @xmath43 @xmath265 @xmath305 and @xmath306 enter into the picture of particle volume compression never was an issue .
the competition has always been fierce and it will always remain so among different display manufacturers . in the near future , when paperlike displays become dominant , the most important deciding factor to who stays in and goes out of business would be determined by the power efficiency of their products .
that being said , a low operation voltage for the control electrodes is crucial for all e - paper technologies and the proposed device based on charged particles is no exception .
the light intensity out of each sub - pixel based on proposed charged particle display technology varies as illustrated in eq .
( [ eq : i_hc ] ) , where @xmath12 is the most probable compression height corresponding to the voltage difference of @xmath271 applied to the control electrodes , see fig .
[ fig : probability ] . with
the @xmath271 restricted to certain range , say @xmath307 one can not arbitrarily choose the other parameters which constitute the design parameters , i.e. , @xmath299 @xmath300 @xmath301 @xmath302 @xmath303 @xmath304 @xmath43 @xmath265 and @xmath308 for example , if too many electrons are removed from each of the aluminum particles , i.e. , @xmath265 the voltage of @xmath309 applied to one of the control electrodes ( the other grounded ) may not be sufficient enough to overcome the coulomb repulsion between particles and compress the particle volume to a level where dark state is reached , assuming @xmath310 defines the brightest state . on the other end , if too many charged particles are present in a chamber , i.e. , the particle number @xmath43 the brightest state achieved by setting @xmath310 for the control electrode may be too dark . the quantitative description of the height @xmath12 of the compressed particle volume in terms of the so called `` design parameters '' thru the expression @xmath261 eq .
( [ eq : ps24 ] ) , ables the design of particle based display with potential to generate continuous grayscale .
the presented optical shutter based on charged particles portrays bistability at all states , including the gray states .
this is possible because the two optically transparent electrodes act as a capacitor , which has the property of sustaining electric fields even when the device is removed of the power supply .
to illustrated how the bistability is achieved for all states , including the gray states , the illustration shown in fig .
[ fig : capacitor ] is considered .
i shall begin with an isolated capacitor , in which the two electrodes of the capacitor are electrically neutral , resulting in zero electric field inside the region between the two electrodes . with the switch closed ,
the top electrode is quickly accumulated with a net positive charge , @xmath311 and the bottom electrode gets accumulated with a net negative charge , @xmath312 the potential difference between the two electrodes results in the creation of electric field inside the capacitor , as illustrated in stage 2 of fig .
[ fig : capacitor ] . assuming the charged particles reside in the region between the two electrodes , the electric field generated inside the capacitor is responsible for the compression of volume containing charged particles .
now , when the switch is opened , the net charge of @xmath313 remains in the top electrode and the net charge of @xmath314 remains in the bottom electrode , provided the capacitor is ideal , i.e. , free from the leakage of electrical current .
therefore , for an ideal capacitor , the electric field is maintained forever inside the region between the two electrodes , thereby sustaining the gray states even when the device is removed of the power supply .
( color online ) the bistability is maintained by the electric field stored in the capacitor . ] in the real system , the role of switch is played by a semiconductor transistor , which is very far from being an ideal switch and it has finite leakage of current .
this deficiency in semiconductor transistor makes the proposed device only semi - bistable , meaning the device must be refreshed regularly .
this , however , is about to change with the current developments in micro electromechanical systems ( mems ) based switches , which literally has zero leakage current for the open state.@xcite initially , the mems based switch has been developed in an attempt to replace the dynamic random access memory ( dram ) architecture for the memory sector of business .
however , as it lacks in switching speed , it will be a while before mems based switches can permanently replace the drams . as for its use as a switch in display technology ,
the mems based switches already show plenty of speed . combined with mems based switches ,
which has zero leakage current for the open switch mode , the proposed optical shutter based on charged particles opens up the possibility of realizing the bistability mode for all states , including the grayscale states .
the pioneering work by szirmai,@xcite hattori et al.,@xcite and others have exposed the potential applications with charged particles . utilizing charged particles in display technologies ,
however , requires a quantitative understanding of how design parameters , such as @xmath299 @xmath300 @xmath301 @xmath302 @xmath303 @xmath304 @xmath43 @xmath265 @xmath305 and @xmath306 enter into particle volume compression . in this work ,
an expression for the compressed state , which incorporates the design parameters , has been presented .
the result should find its role in the development of displays based on charged particles .
the author acknowledges the support for this work provided by samsung electronics , ltd .
w. chim , `` a flexible electronic paper with integrated display driver using single grain tft technology , '' msc .
thesis , computer engineering , dept . of electrical engineering , delft university of technology ( 2009 ) . | an optical shutter based on charged particles is presented .
the output light intensity of the proposed device has an intrinsic dependence on the interparticle spacing between charged particles , which can be controlled by varying voltages applied to the control electrodes .
the interparticle spacing between charged particles can be varied continuously and this opens up the possibility of particle based displays with continuous grayscale . |
recently , bouncing models of the universe were revived in the framework of string cosmology @xcite , since it is possible to generate a scale invariant spectrum of density fluctuations in the pre - bounce phase .
specific realizations are e.g. cyclic / ekpyrotic models of the universe @xcite or the pre - big bang scenario @xcite .
the main problem for such models to succeed lies in the fact that the bounce itself is often singular in the models studied so far @xcite .
this makes matching conditions a necessity @xcite which are , unfortunately , rather ambiguous .
in addition , fluctuations often become non perturbative near a singularity @xcite . in the few toy models of regular bouncing cosmologies known in the literature
, the actual bounce has a strong impact on the evolution of perturbations @xcite .
what is more , the models are often technically challenging at the perturbative level and unfortunately inconsistent methods were proposed in the literature .
consequently , bouncing models have not been taken seriously as alternatives to the inflationary paradigm @xcite . in this article
we will construct a simple regular toy model , for which one can compute analytically how perturbations evolve through the bounce .
we find a red spectral index for scalar perturbations that is ruled out by observations @xcite .
furthermore , the detailed mechanism providing the bounce has a strong impact on the spectrum . in order to generate a bounce within pure four dimensional general relativity
, one has to violate various energy conditions .
the usual mechanism in the literature involves the addition of an extra exotic matter field that becomes dominant during the bounce @xcite ( see also @xcite for an @xmath2 regularized bounce , @xcite for a non - singular cosmology achieved by higher derivative modification of general relativity , or @xcite for a bounce induced by a general - covariant , @xmath3-duality - invariant , non - local dilaton potential ) .
the presence of a second matter field , and perturbations in it , is one physical origin for the sensitivity to the type of bounce one is considering .
we will not add an additional matter field , but use an ingredient motivated by string theory : extra dimensions .
if we assume one additional dimension , the effective four dimensional friedmann equation will show corrections proportional to @xmath0 at high energy densities .
if the extra dimension is space - like , we have the usual randall sundrum setup @xcite . in that case ,
@xmath4 and @xmath0 have the same sign , and no bounce will occur
. however , if we assume the additional dimension to be time - like , this sign will flip and thus it will cause a bounce when @xmath0 becomes dominant @xcite .
we are aware of complexities associated with an extra time - like dimension ; for instance , existence of tachyonic modes and possible violation of causality , or possible appearance of negative norm states ( see @xcite for more discussion regarding these issues ) .
we will not touch on those issues , but take the modification of the friedmann equations as a simple toy model that could also arise by other means .
the outline of the article is as follows : first , we work out the background solution based on the above idea in section [ secbgr ] .
we find a whole class of simple , analytically known , bouncing cosmologies , either big bang avoiding ones , for usual matter content , or big rip avoiding ones , for phantom matter
. we will then focus on one specific big bang avoiding bounce by considering radiation as the matter content this seems to be the most conservative choice to us .
in section [ secpert ] we discuss scalar perturbations and compute analytically the spectrum of fluctuations in the post bounce era , for bunch - davis vacuum initial condition in the pre - bounce era .
we find a red spectrum with a spectral index of @xmath1 , thus ruling out this model as a realistic one .
we also find that the background cosmology around the bounce dictates the deformation of the spectrum through the transition . in section
[ interpret ] , we conclude that it is challenging for the bounce in cyclic / ekpyrotic models of the universe to leave the spectrum unaffected . the only way to check if this is the case is by finding a regularized version of the proposed scenarios and following the perturbations explicitly through the bounce . in section [ secvector ] , we discuss vector perturbations and find that they remain perturbative during the bounce . finally in section [ sectensor ] , we derive analytically the evolution of gravitational waves as they pass through the bounce .
we find a blue spectrum of gravitational waves in the post bounce era on large scales , and an amplitude that depends on the details of the bounce .
we use a metric with negative signature , scale factor @xmath5 , cosmic time @xmath6 and , for simplicity , we work with a flat universe , so that @xmath7 considering a modified version of the randall sundrum ( rs ) scenario @xcite , originating e.g. by having an extra time - like dimension @xcite we have the modified einstein equations @xmath8 where @xmath9 , @xmath10 and @xmath11 is the fundamental 5-dimensional planck mass . for simplicity and
since we are primarily interested in the high energy modifications due to @xmath12 , we fine tuned the model such that the effective four dimensional cosmological constant vanishes and we neglected the projected weyl tensor .
@xmath12 is quadratic in the energy momentum tensor @xmath13 and is given by @xcite @xmath14 \ , .
\label{smunu}\end{aligned}\ ] ] we have a different sign in front of @xmath12 in ( [ einstein ] ) compared to the usual rs setup @xcite .
this modification will yield a class of non singular bounces we shall examine below . if we consider an ideal fluid with @xmath15 the quadratic term becomes @xmath16 we may rephrase the quadratic corrections in terms of a second ideal fluid with @xmath17 where @xmath18 .
the `` two fluids '' are of course related and represent only one degree of freedom . if we have an equation of state between @xmath19 and @xmath20 , that is @xmath21 , we get @xmath22 with @xmath23 .
it will become useful to write @xmath24 so that @xmath25 . as a consequence of the equations of state we get @xmath26 the friedmann equation reads now @xmath27 and has the solution @xmath28^{1/n}\,,\end{aligned}\ ] ] for @xmath29 .
thus for any positive @xmath30 we get a regular bounce at @xmath31 with minimal scale factor @xmath32 e.g. for dust @xmath33 , radiation @xmath34 , stiff ( holographic ) matter ( @xmath35 ) etc .
negative @xmath30 corresponds to matter that violates the null energy condition ( nec ) .
however , there is no big rip , but a bounce at the maximal scale factor @xmath36 .
such a phantom bounce was recently proposed in @xcite , but no analytic solution was given .
if we also introduce the characteristic time scale of the bounce @xmath37 we may write @xmath38 the relation between cosmic time @xmath6 and conformal time @xmath39 is given by @xmath40 where @xmath41 is a generalized hypergeometric function , but we will not make use of this relation in the following analysis .
let us turn our attention to perturbations and examine how they evolve through a big bang avoiding bounce with minimal scale factor .
this is the most interesting case , because it could be seen as a simple toy model for a regularized version of the bounce occurring in the cyclic scenario .
it is a toy model , since we remain at the four dimensional effective description throughout .
we will comment on this issue in section [ interpret ] . in the following
, we have to be careful , since the so called sound speed " @xmath42 diverges near the bounce . to be specific , for our background solutions one
gets @xmath43 so that @xmath44 diverges at @xmath45 , that is at @xmath46 . as a consequence ,
the usual equations governing the evolution of perturbations that involve @xmath44 can not be used near the bounce .
therefore , we will derive the relevant equations of motion from first principles in the next section .
for the time being we will work with conformal time @xmath39 . the most general perturbed metric involving only scalar metric perturbations in the longitudinal gauge
is given by @xcite @xmath47\,.\end{aligned}\ ] ] note that the two bardeen potentials agree with the gauge invariant scalar metric perturbations there is no residual gauge freedom .
the perturbed energy momentum tensor is given by @xcite @xmath48 where @xmath49 and @xmath50 denote background quantities given by ( [ rho+ ] ) and @xmath51 is the velocity potential so that @xmath52 considering no anisotropic stress , that is @xmath53 , the off - diagonal einstein equations yield @xmath54 so that only one scalar metric degree of freedom is left . considering @xmath55 , where @xmath56 is the entropy perturbation , and defining @xmath57 as well as @xmath58 we have @xmath59 and from perturbing ( [ smunu ] ) we get @xmath60 \delta ^i_j \end{array } \right)\,.\end{aligned}\ ] ] with @xmath61 the perturbed einstein equations read @xmath62\frac{\rho _ 0}{\lambda}\right]\,,\\ & & \left[\phi\mathcal{h}+\phi ^{\prime}\right]_{,i}=-\frac{a^2}{2}\kappa^2\rho_0v_{,i}(1+w ) \left(1-\frac{\rho _ 0}{\lambda}\right)\,.\label{einstein3}\end{aligned}\ ] ] note that the right hand side of ( [ einstein1 ] ) will become zero at @xmath63 .
henceforth , one might expect difficulties in deriving a single second order differential equation for @xmath64 .
a method to deal with this problem , in the case of adiabatic perturbation or @xmath65 , was advocated in @xcite and subsequently used e.g. in @xcite and the previous versions of the article : one splits @xmath64 in two components , each of which satisfies a regular second order differential equation . however ,
this method is inconsistent with other constraint equations , like the fluid conservation equations in the case of true two fluid models , as was explicitly shown in @xcite , or in our model the fact that the density of the second fluid is related to the first fluid .
there , an alternative method was introduced , which we shall employ in the following .
since the universe was radiation dominated before the period of recombination , let us restrict ourselves to @xmath66 such that @xmath67 where , @xmath68 is the rescaled time .
now , let us combine ( [ einstein1 ] ) and ( [ einstein2 ] ) to @xmath69\\ & & -\left(1-\frac{r}{\lambda a^4}\right)\left [ \phi^{\prime\prime}+3\mathcal{h}\phi^{\prime}+(2\mathcal{h}^{\prime}+\mathcal{h}^2)\phi\right]\,,\end{aligned}\ ] ] where we used @xmath70 , since @xmath66 and @xmath71 for radiation .
converting this equation to rescaled cosmic time @xmath72 , making use of the background solution ( [ specifica ] ) for @xmath73 and introducing the rescaled momentum @xmath74 yields @xmath75\frac{\tilde{k}^2}{\sqrt{1+x^2}}\right)\phi_k&=&0\,,\end{aligned}\ ] ] where @xmath76 is a fourier - mode of @xmath64 and _ dot _ denotes a derivative with respect to @xmath72 . before we go ahead and solve this equation in different regimes ,
let us have a look at the problematic region around @xmath77 , the boundaries of the region where the _ null energy condition _ is violated .
the nontrivial question is whether @xmath76 remains continuous and smooth near these points or not .
this question has been the object of study in @xcite , which provides analytic solutions and a detailed discussion in the case of adiabatic perturbations .
consequently , we will expand @xmath76 near these points to find the two independent analytic solutions , as proposed in @xcite .
our final goal is to compute the spectrum @xmath78 long after the bounce ( @xmath79 ) , while specifying the initial conditions for each fourier mode @xmath76 well before the bounce ( @xmath80 ) . to achieve that
, we will find approximate analytic solutions within different regions and match them together . in this way
, we will be able to compute analytically the spectral index @xmath81 .
let us start in the limit @xmath82 . in that case one can introduce the gauge invariant variable @xmath83 in terms
of which the action takes the simple form of a scalar field ( see @xcite for a detailed discussion ) so that the equation of motion reads @xmath84 with @xmath85 and @xmath86 since the universe is radiation dominated , @xmath87 vanishes and we can solve ( [ eomv ] ) @xmath88 where we imposed quantum mechanical initial conditions for @xmath83 , that is @xmath89 and the normalization condition @xmath90 . the variable @xmath83 is related to @xmath76 via @xmath91 in that regime .
if we use the background solution in the limit @xmath80 we get @xmath92 as an approximate analytical solution for @xmath64 .
the overall scale is given by @xmath93 which is a free parameter in our model , subject to mild constraints , e.g. @xmath94 should be larger than the planck time , but smaller than the time of nucleosynthesis . however , we will not need to specify its value if we scale the spectrum appropriately . similarly , for @xmath79 the general solution of @xmath95
can be written as @xmath96 where the @xmath97 are constants .
for @xmath98 we can neglect the @xmath99-dependence of ( [ maineom ] ) , that is we can use @xmath100 which is solved by @xmath101 where @xmath102\,,\\ d_1&=&1+\frac{\sqrt{41}}{4}\,,\\ n_2&=&\left[\frac{15-\sqrt{41}}{8},\frac{9-\sqrt{41}}{8}\right]\,,\\ d_2&=&1-\frac{\sqrt{41}}{4}\,,\end{aligned}\ ] ] and @xmath103 is the generalized hypergeometric function . close to @xmath104 we can taylor - expand ( [ maineom ] ) so that we get @xmath105 where we introduced @xmath106 we will now follow @xcite to solve this equation near @xmath104 : if @xmath107 remains bounded , we get one independent solution from which we can deduce the other one by means of the wronskian method . in this way , we arrive at @xmath108 note that it would be justified to neglect the @xmath99 dependence in this solution . for the detailed reasoning as to why this solution is appropriate we refer the reader to @xcite .
last but not least , we have to find a solution for the actual bounce , that is for @xmath109 and @xmath110 . the relevant equation of motion in this regime is @xmath111 which is solved by @xmath112 with @xmath113 , @xmath114 and @xmath115 whittakers functions .
note that the solution is well behaved throughout the actual bounce .
all that is left to do is to match all the solutions smoothly . to be specific one has to use ( [ initialphi ] ) for @xmath116 , ( [ solbefore ] ) for @xmath117 , ( [ soltaylor ] ) for @xmath118 , ( [ solbounce ] ) for @xmath119 , ( [ soltaylor ] ) for @xmath120 , ( [ solbefore ] ) for @xmath121 and finally ( [ solaway ] ) for @xmath122 with some small @xmath123
this straightforward but tedious calculation results in a spectral index of @xmath124 which is independent of the choice of @xmath123 .. the reason the index is not effected by the choice of @xmath123 lies in the fact that ( [ soltaylor ] ) becomes independent of @xmath99 if @xmath110 . ]
this index is of course in conflict with the observed nearly scale invariant spectrum .
a few words regarding the validity of the matching procedure might be in order .
lets first consider the matching at @xmath125 : to either consider the term @xmath126 or the term stemming from @xmath127 in ( [ maineom ] ) and match at the point where both are equal is indeed a standard procedure that was employed and tested ( e.g. numerically ) at various instances .
for example , in the framework of a bouncing universe it was used in @xcite and verified numerically in the limit @xmath110 , .
one can simply check the validity of this matching for our case since the solutions of ( [ afternec ] ) as well as equations ( [ initialphi ] ) and ( [ postbouncesol ] ) , around @xmath128 behave as @xmath129 where @xmath130 s and @xmath131 s are constant .
this ensures a smooth matching of these solutions at these points .
also , one could successfully use the same matching procedure in the study of fluctuations in an inflationary universes , even though a more elegant procedure is available ( see e.g. chapter 13 of @xcite for a detailed discussion ) .
the remaining matchings occur close to @xmath132 and do not influence the spectral index as long as we are confident that the solutions are regular ( as is the case for our scenario ) , since all equations dictating the matching conditions are independent of @xmath99 in the limit of @xmath110 . as a consequence
, we will be able to give a simplified analytical argument in the next section yielding the same spectral index .
thereafter , we compare our result with related studies in the literature like @xcite .
we would like to provide some simple insight into our result of @xmath1 from the previous section .
as we mentioned earlier , it can be shown that the solutions of the corresponding equations around @xmath133 have the following asymptotic @xmath72 dependence @xmath134 this makes an analytic tracking of the calculation very easy : on the one hand , the smooth matching conditions are satisfied by requiring @xmath135 and on the other hand , the proper @xmath99 dependence of @xmath136 and @xmath137 can be enforced through the the appropriate initial conditions leading to ( [ initialphi ] ) .
taylor expanding ( [ initialphi ] ) in the limit @xmath138 results in @xmath139 so that we have @xmath140 next , note that the transfer functions relating @xmath141 and @xmath142 to the post - bounce coefficients @xmath143 and @xmath144 have to be independent of @xmath99 , since the equations governing the regime in between @xmath145 and @xmath146 are independent of @xmath99 .
in fact , our equation have regular solutions throughout this region so that it is reasonable to deduce @xmath147 where @xmath148 and @xmath149 have to be some constant numbers .
consequently the total transfer functions will be @xmath150 finally , making use of the above equations and @xmath151 for small @xmath99 according to ( [ a1b1 ] ) , we can calculate the power spectrum for @xmath152 , the non - decaying mode of ( [ finalmode ] ) @xmath153 henceforth , we conclude a spectral index of @xmath154 in order to provide a quick check of our analytic arguments , we integrated equation ( [ maineom ] ) with a 4/5th - order runge - kutta method , as implemented in maple v9 . starting the numerical treatment at @xmath145 with initial conditions given by ( [ initialphi ] ) and evaluating the final spectrum at @xmath155 yields a spectral index of @xmath1 , fig.[fig1 ] .
the computations in maple were done with 120 digits accuracy ; nevertheless , it should be noted that the amplitude can not be recovered properly by the simple numerical method used , due to the non - trivial nature of the equation of motion near the bounce .
however , this does not effect the spectral index since the actual bounce is @xmath156-independent , as can be seen in ( [ eomforsolbounce ] ) .
the numerical integration of ( [ maineom ] ) yields the spectrum ( [ defspectrum ] ) evaluated at @xmath155 ( dots ) ; the solid line corresponds to a spectral index of @xmath1 , in agreement with the analytic result ( [ analyticresult ] ) .
higher order effects in @xmath99 are accounted for by letting @xmath99 cover a broad range from @xmath157 down to @xmath158 . ]
our result of the previous section clearly shows that for the specific bouncing cosmology considered , both modes ( the initial constant one and the growing one ) will pass through the bounce , but it is the mode with the redder spectrum , i.e. the growing mode for the bunch - davis vacuum , that dictates the final shape of the post - bounce spectrum . henceforth , we observe mode mixing in a model with only one degree of freedom and no spatial curvature .
one has to be more careful to draw such conclusion for more general cases involving two different fluids ( constrained by an adiabatic condition in other backgrounds ) , since the matching procedure at @xmath159 is not that simple in general . to be more specific
, the feature of the bounce which is crucial for determining the approximate behavior of the out coming spectrum , besides knowing the existence and regularity of the solutions throughout the bounce region , is the correction to the equation of motion ( [ afternec ] ) due to the second fluid .
this is the region where modes just turned non oscillatory and corrections due to the bounce have to be considered , because they are crucial for the solution .
recently , a wide class of bouncing cosmologies with two independent fluids was discussed in @xcite by bozza and veneziano .
one might think that , in their notation , our class of bouncing universes corresponds to @xmath160 , @xmath161 , @xmath162 and @xmath66 .
this would put it into region c of @xcite and a spectral index of @xmath163 would result according to @xcite .
however the model of @xcite differs from our bouncing universe , since @xmath160 only enforces the adiabatic condition on each fluid separately and while no intrinsic entropy modes for each fluid are considered , two independent initial conditions for each of these fluids are assumed . in our model
the `` two fluids '' are indeed related , so that the total pressure becomes a function of the total energy ( as a consequence only one initial condition can be specified ) . in other words ,
we have enforced the adiabatic condition in its strong interpretation .
it should also be noted , the approximation scheme employed there , differs considerably from ours , involving an expansion in @xmath156 and a dimensional argument for the contribution of the bounce ( no explicit solution for the bounce was used ) .
henceforth , the results of @xcite can not be applied to our model .
a few words regarding ekpyrotic / cyclic scenarios are needed : in @xcite there is a mechanism present that generates a scale invariant spectrum before the bounce @xcite .
it was shown in @xcite , that this spectrum will not get transferred to the post - bounce era , if perturbations evolve independently of the details of the bouncing phase . furthermore , the main conclusion of @xcite was , that a smooth bounce in four dimensions is not able to generate a scale invariant spectrum via the mode mixing technique . however , all these arguments were specific to bouncing cosmologies in four dimensional gravity .
so far , the effective four dimensional description breaks down for all proposed cyclic models @xcite near the bounce , and the bounce itself is singular @xcite .
thus , it is of prime interest to search for a fully regularized bounce in cyclic / ekpyrotic models of the universe , so that one can compute the spectrum of perturbations explicitly in a full five dimensional setting - see e.g. @xcite for a recent proposal and @xcite for an argument that mode mixing can occur in the full five dimensional setup .
our results confirm that it is challenging for such a model to succeed , but not impossible .
vector perturbations ( vp ) have recently caught attention in the context of bouncing cosmologies , due to their growing nature in a contracting universe @xcite . in the context of a regularized bounce that we are discussing ,
vp remain finite and well behaved , as we shall see now .
the most general perturbed metric including only vp is given by @xcite @xmath164 where the vectors @xmath165 and @xmath166 are divergenceless , that is @xmath167 and @xmath168 .
a gauge invariant vp can be defined as @xcite @xmath169 the most general perturbation of the energy momentum tensor including only vp is given by @xcite @xmath170 where @xmath171 and @xmath172 are divergenceless .
furthermore the perturbation in the 4-velocity is related to @xmath172 via @xmath173 gauge invariant quantities are given by @xmath174 and @xmath171 . from now on we work in newtonian gauge , where @xmath175 so that @xmath176 coincides with @xmath165 and @xmath177 with @xmath51 .
note that there is no residual gauge freedom after going to newtonian gauge . for simplicity
we assume the absence of anisotropic stress , @xmath178 , so that @xmath179 the resulting equations of motion read @xmath180 equation ( [ eoms ] ) is easily intergrated for each fourier mode , yielding @xmath181 where @xmath182 is time independent . since @xmath183
, there is no divergent mode present .
equation ( [ eomv ] ) is an algebraic equation for each fourier mode of the velocity perturbation @xmath172 .
since @xmath184 becomes zero shortly before and after the bounce , it seems like @xmath172 has to diverge before the bounce
however , all that this is telling us is that @xmath172 is not the right variable to focus on : we should focus on the combination of @xmath172 and @xmath185 that appears in the perturbation of the effective total energy momentum tensor , that is @xmath186 .
this combination is well behaved , and in fact one can check that it is always smaller than either @xmath187 or @xmath188 if we impose appropriately small initial conditions for @xmath182 .
thus the perturbative treatment of vp is consistent .
in this section we calculate the amplitude and spectral index of gravity waves , corresponding to tensor perturbations of the metric , in a radiation dominated universe .
because the complete five dimensional calculation is very complicated we will continue using the theory of linear perturbations in a four dimensional effective field theory .
we include only the corrections @xmath189 , due to fifth dimension , entering through the evolution of the background . in the linear theory of cosmological perturbations
, tensor perturbations can be added to the background metric via @xcite @xmath190\,,\ ] ] where @xmath191 has to satisfy @xmath192 thus @xmath191 has only two degrees of freedom which correspond to the two polarizations of gravity waves .
one can show ( see @xcite ) that the amplitude @xmath193 for each of these polarizations has to obey @xmath194 where @xmath195 .
note that gravity waves do not couple to energy or pressure and that they are gauge invariant from the start .
if the second term dominates over @xmath196 in the above equation @xmath197 will exhibit oscillatory solutions , and if the gradient term is negligible we will have @xmath198 , or , in other words , @xmath193 is almost constant . using the standard fourier decomposition we get for each fourier mode @xmath199
therefore , @xmath200 sets a scale for the behavior of each mode .
this is very similar to the role of the hubble radius for inflationary scenarios , but rather than being almost constant , @xmath200 is more like a potential barrier in quantum mechanics : @xmath201 as fig .
[ v ] shows , @xmath202 rises up to @xmath203 around the bounce but falls off very quickly as @xmath204 .
@xmath202 sets a scale for the behavior of each mode , it rises up to @xmath203 around the bounce but falls off very quickly as @xmath205 . ]
thus , for @xmath206 , the solutions of equation ( [ eqgrav ] ) are oscillatory .
this implies that we can set our initial conditions according to the bunch - davis vacuum at the initial time @xmath207 .
we get @xmath208 long before the bounce and it will remain basically unchanged throughout the bounce .
we easily compute the spectrum of perturbations and the spectral index for these modes to be @xmath209 implying that , for @xmath206 , we are simply left with vacuum fluctuations with an ever decaying amplitude . but first , we need to estimate the solutions for @xmath225 ( i.e. @xmath226 ) that is for @xmath227 . unfortunately , unlike the case of most potential barriers in quantum mechanics , we can not use the wkb approximation to calculate the outgoing spectrum , since the condition @xmath228 is not satisfied in this regime .
fortunately we can still approximate the solution of equation ( [ eqgrav ] ) perturbatevely in orders of @xmath229 ) for @xmath230 .
this approach is very helpful for analyzing most models of bouncing universes , since the equations governing the evolution of perturbations are similar during the bounce , see for example @xcite . ]
@xmath231\\ \nonumber & & + b_2\int^\eta { d\tau\over a^2}\left[1-k^2\int^\tau d\varsigma ~a^2\int^\varsigma { d\rho \over a^2}\right ] + \cdot\cdot\cdot\,.\end{aligned}\ ] ] for @xmath232 , since @xmath233 and also@xmath234 ) to @xmath235\,.\ ] ] notice that @xmath236 for the above solution , since @xmath237\sim\rm{const.}\ ] ] at @xmath238 .
therefore , @xmath239 can be smoothly matched to our solutions from equations ( [ solutioni ] ) and ( [ solutioniii ] ) at the transition points @xmath214 by equating @xmath239 and @xmath240 .
we obtain the transfer function by relating the coefficients @xmath136 and @xmath137 to @xmath143 and @xmath144 .
these satisfy the relations @xmath241
\label{transmatrix1}\\ & & + e^{-ik\eta_i}[0.65~ i \tilde{k}^{-1}]b_1\ , , \nonumber\\ b_3&\simeq&e^{ik\eta_i}[-0.65~i\tilde{k}^{-1}]a_1 \label{transmatix2}\\ & & + e^{-ik(2\eta_{_{tr}}+\eta_i)}\bigl [ -0.65~ i \tilde{k}^{-1}+2.32 \tilde{k}^{-1/3}\bigr ] b_1\,.\nonumber\end{aligned}\ ] ] the above result is valid for general initial conditions .
we first note that , to lowest order in @xmath242 , the constant mode in the pre - collapse phase ( growing mode for h ) , matches onto the growing mode in the post - collapse phase ( constant mode for h ) , without any change in the spectral shape . as we mentioned before , for a bunch - davis vacuum one simply takes @xmath223 and @xmath224 .
the resulting power - spectrum for such a case is @xmath243 the above spectrum is oscillatory and , for @xmath244 , it is oscillating so fast that one could conclude an almost flat ( @xmath245 ) effective spectrum , with a decaying amplitude .
however , if @xmath246 or in other words @xmath247 , which is the case for super - hubble wavelengths , the above equation simplifies to @xmath248 thus , we get a non - decaying mode which has a blue spectrum ( @xmath249 ) .
notice , the amplitude of the power spectrum was dictated directly by the scales and details of the bounce .
this result is in agreement with the conclusions of @xcite , where a scale factor somewhat similar to our background scale factor was used .
we constructed a wide class of regular bouncing universes , with a smooth transition from contraction to expansion .
this was motivated by open questions regarding the evolution of perturbations during singular bouncing cosmologies in string cosmology , occurring e.g. in the cyclic scenario .
the bounce is caused by the presence of an extra time - like dimension , which introduces corrections to the effective four dimensional einstein equations at large energy densities .
if matter respecting the null energy condition is present , a big bang avoiding bounce results , and for phantom matter a big rip avoiding bounce emerges .
we find analytic expressions for the scale factor in all cases . for scalar perturbations
we compute analytically a spectral index of @xmath1 , thus ruling out such a model as a realistic candidate .
we also find the spectrum to be sensitive to the bounce region , in agreement to the majority of regular toy models discussed in the literature .
we conclude that it is challenging , but not impossible , for cyclic / ekpyrotic models to succeed , if one can find a regularized version .
next , we discussed vector perturbations which are known to be problematic in bouncing cosmologies , due to their growing nature .
we checked explicitly that they remain perturbative and small if compared to the background energy density and pressure .
we concluded with a discussion of gravitational waves .
the transfer function was computed analytically and we showed that the shape of the spectrum is unchanged in general . in the special case of vacuum initial conditions ,
the spectral index in post bounce era for super - hubble wavelengths is blue .
however , the amplitude turned out to be sensitive to the details of our model . to summarize , having an explicit realization of e.g. the cyclic model of the universe that features a regular bounce it might be possible to produce a scale invariant spectrum .
henceforth , the challenge for the cyclic model is twofold : 1 . to find a convincing mechanism that regularizes the bounce .
2 . to deal with the perturbations in a full five dimensional setup and to take into account the presence of an inhomogeneous bulk as inevitable by the presence of branes .
we would like to thank robert brandenberger for helpful comments and discussions and scott watson as well as niayesh afshordi , diana battefeld , zahra fakhraai , andrei linde and v. bozza for comments on the draft .
we would also like to thank the referee of prd for drawing our attention to a possible problem associated with the splitting method used in previous versions of this article . t.b .
would like to thank the physics department of mcgill university and g.g would like to thank the physics department at harvard university for hospitality and support .
t.b . has been supported in part by the us department of energy under contract de - fg02 - 91er40688 , task a. j. e. lidsey , d. wands and e. j. copeland , phys .
rept . * 337 * , 343 ( 2000 ) [ arxiv : hep - th/9909061 ] .
j. khoury , b. a. ovrut , p. j. steinhardt and n. turok , phys .
d * 64 * , 123522 ( 2001 ) [ arxiv : hep - th/0103239 ] . j. khoury , b. a. ovrut , n. seiberg , p. j. steinhardt and n. turok , phys .
d * 65 * , 086007 ( 2002 ) [ arxiv : hep - th/0108187 ] .
p. j. steinhardt and n. turok , phys .
d * 65 * , 126003 ( 2002 ) [ arxiv : hep - th/0111098 ]
. s. gratton , j. khoury , p. j. steinhardt and n. turok , phys .
d * 69 * , 103505 ( 2004 ) [ arxiv : astro - ph/0301395 ] .
a. j. tolley , n. turok and p. j. steinhardt , phys .
d * 69 * , 106005 ( 2004 ) [ arxiv : hep - th/0306109 ] .
m. gasperini and g. veneziano , astropart .
phys . * 1 * , 317 ( 1993 ) [ arxiv : hep - th/9211021 ] .
m. gasperini and g. veneziano , phys .
rept . * 373 * , 1 ( 2003 ) [ arxiv : hep - th/0207130 ] .
t. j. battefeld , s. p. patil and r. brandenberger , phys .
d * 70 * , 066006 ( 2004 ) [ arxiv : hep - th/0401010 ] .
n. turok , m. perry and p. j. steinhardt , phys .
d * 70 * , 106004 ( 2004 ) [ erratum - ibid .
d * 71 * , 029901 ( 2005 ) ] [ arxiv : hep - th/0408083 ] .
j. khoury , b. a. ovrut , p. j. steinhardt and n. turok , phys .
d * 66 * , 046005 ( 2002 ) [ arxiv : hep - th/0109050 ] .
r. durrer and f. vernizzi , phys .
d * 66 * , 083503 ( 2002 ) [ arxiv : hep - ph/0203275 ] .
j. martin , p. peter , n. pinto - neto and d. j. schwarz , phys .
d * 65 * , 123513 ( 2002 ) [ arxiv : hep - th/0112128 ] .
r. brandenberger and f. finelli , jhep * 0111 * , 056 ( 2001 ) [ arxiv : hep - th/0109004 ] . n. goheer , p. k. s. dunsby , a. coley and m. bruni , arxiv : hep - th/0408092 .
p. peter and j. martin , arxiv : hep - th/0402081 .
j. martin and p. peter , phys .
d * 68 * , 103517 ( 2003 ) [ arxiv : hep - th/0307077 ] .
l. e. allen and d. wands , phys .
d * 70 * , 063515 ( 2004 ) [ arxiv : astro - ph/0404441 ] .
p. peter and n. pinto - neto , phys .
d * 66 * , 063509 ( 2002 ) [ arxiv : hep - th/0203013 ] .
v. bozza and g. veneziano , phys .
b * 625 * , 177 ( 2005 ) [ arxiv : hep - th/0502047 ] .
v. bozza and g. veneziano , jcap * 0509 * , 007 ( 2005 ) [ arxiv : gr - qc/0506040 ] .
v. bozza , arxiv : hep - th/0512066 .
f. finelli , jcap * 0310 * , 011 ( 2003 ) [ arxiv : hep - th/0307068 ] . c. cartier , arxiv : hep - th/0401036 . t. j. battefeld and g. geshnizjani , arxiv : hep - th/0506139 , to appear in prd , comments . m. gasperini , m. giovannini and g. veneziano , phys .
b * 569 * , 113 ( 2003 ) [ arxiv : hep - th/0306113 ] .
m. gasperini , m. giovannini and g. veneziano , nucl .
b * 694 * , 206 ( 2004 ) [ arxiv : hep - th/0401112 ] .
a. linde , arxiv : hep - th/0205259 .
d. n. spergel _ et al . _
[ wmap collaboration ] , astrophys .
j. suppl .
* 148 * , 175 ( 2003 ) [ arxiv : astro - ph/0302209 ] .
patrick peter , nelson pinto - neto , diego a. gonzalez , jcap 0312:003,2003 .
[ hep - th 0306005 ] r. brandenberger , v. mukhanov and a. sornborger , phys .
d * 48 * , 1629 ( 1993 ) [ arxiv : gr - qc/9303001 ] .
s. tsujikawa , r. brandenberger and f. finelli , phys .
d * 66 * , 083513 ( 2002 ) [ arxiv : hep - th/0207228 ] .
l. randall and r. sundrum , phys .
lett . *
83 * , 3370 ( 1999 ) [ arxiv : hep - ph/9905221 ] .
p. brax , c. van de bruck and a. c. davis , arxiv : hep - th/0404011 . y. shtanov and v. sahni , phys .
b * 557 * , 1 ( 2003 ) [ arxiv : gr - qc/0208047 ] .
m. g. brown , k. freese and w. h. kinney , arxiv : astro - ph/0405353 . y. s. piao and y. z. zhang , arxiv : gr - qc/0407027 .
i. bars , arxiv : hep - th/0106021 . t. shiromizu , k. i. maeda and m. sasaki , phys .
d * 62 * , 024012 ( 2000 ) [ arxiv : gr - qc/9910076 ] .
r. maartens , phys .
d * 62 * , 084023 ( 2000 ) [ arxiv : hep - th/0004166 ] . s. nojiri and s. d. odintsov , phys .
d * 70 * , 103522 ( 2004 ) [ arxiv : hep - th/0408170 ] .
j. m. bardeen , phys .
d * 22 * , 1882 ( 1980 ) .
v. f. mukhanov , h. a. feldman and r. h. brandenberger , phys .
rept . * 215 * , 203 ( 1992 ) .
r. h. brandenberger , arxiv : hep - th/0501033 .
m. abramowitz and i. stegun , _ handbook of mathematical functions _ , ( dover publ .
, new york , 1985 ) .
p. creminelli , a. nicolis and m. zaldarriaga , arxiv : hep - th/0411270 .
m. giovannini , arxiv : hep - th/0502068 .
t. j. battefeld , s. p. patil and r. h. brandenberger , arxiv : hep - th/0509043 , to appear in prd .
p. mcfadden , n. turok and p. j. steinhardt , arxiv : hep - th/0512123 .
t. j. battefeld and r. brandenberger , phys .
d * 70 * , 121302(r ) ( 2004 ) [ arxiv : hep - th/0406180 ] .
m. giovannini , phys .
d * 70 * , 103509 ( 2004 ) [ arxiv : hep - th/0407124 ] .
a. a. starobinsky , jetp lett . * 30 * , 682 ( 1979 ) [ pisma zh .
30 * , 719 ( 1979 ) ] .
r. h. brandenberger , arxiv : hep - th/0509099 .
r. h. brandenberger , arxiv : hep - th/0509159 .
t. battefeld and s. watson , arxiv : hep - th/0510022 , to appear in rev . mod . | we consider a simple toy model of a regular bouncing universe .
the bounce is caused by an extra time - like dimension , which leads to a sign flip of the @xmath0 term in the effective four dimensional randall sundrum - like description .
we find a wide class of possible bounces : big bang avoiding ones for regular matter content , and big rip avoiding ones for phantom matter .
focusing on radiation as the matter content , we discuss the evolution of scalar , vector and tensor perturbations .
we compute a spectral index of @xmath1 for scalar perturbations and a deep blue index for tensor perturbations after invoking vacuum initial conditions , ruling out such a model as a realistic one .
we also find that the spectrum ( evaluated at hubble crossing ) is sensitive to the bounce .
we conclude that it is challenging , but not impossible , for cyclic / ekpyrotic models to succeed , if one can find a regularized version . |
classically , a _ chord diagram _ is a collection of chords of a circle ; the intersection graph for these chords is called a _ circle
graph_. circle graphs have been intensively studied in graph theory , with classifications given by even and itai @xcite and bouchet @xcite , among others . in recent years ,
interest in chord diagrams has spread to topologists as part of the theory of finite type knot and link invariants @xcite , as well as the theory of virtual knots @xcite .
a natural extension of knot theory is to look at spatial embeddings of more complex graphs , and it is natural to ask whether some analogue of chord diagrams could be equally useful in this context .
the goal of this paper is to define a reasonable notion of chord diagram for general graphs , and to construct some tools to study them . in particular
, we will look at a variety of intersection graphs for these chord diagrams .
we will use these intersection graphs to determine when an embedding of a planar graph can be extended to an embedding of a chord diagram on that graph . as an application , we will define _ gauss codes _ for immersions of arbitrary graphs and give algorithms for determining whether a crossing sequence is realizable as a gauss code .
acknowledgement : the authors would like to acknowledge the hospitality of waseda university , tokyo , and professor kouki taniyama during the international workshop on knots and links in a spatial graph in july , 2004 , where the idea for this project was conceived .
the second author was supported by an lmu faculty research grant .
a chord diagram for a circle may be characterized as simply a circle with a set of labeled points indicating the endpoints of the chords .
this is the idea we use to define a chord diagram for a general graph .
[ d : chord ] let g be a graph with edge set e and vertex set v. a * chord diagram on g of degree n * is a collection of 2n points in @xmath0 ( so each point is in the interior of an edge ) , each labeled from a set @xmath1 such that each label is used exactly twice .
the two points labeled @xmath2 are called the * endpoints * of chord @xmath2 .
often , we will want to look at _ oriented _ chord diagrams of directed graphs .
this allows us to record whether the chord is on the same or opposite sides of the edges at each endpoint .
[ d : orientchord ] let g be a * directed * graph with edge set e and vertex set v. an * oriented chord diagram on g of degree n * is a collection of 2n points in @xmath0 , each labeled from a set @xmath3 such that exactly two points are labeled from each set @xmath4 .
these two points are called the * endpoints * of chord @xmath2 ; if they have the same label , the endpoints are said to have the same orientation .
if one is labeled @xmath2 and the other @xmath5 , they are said to have opposite orientations .
generally , we will represent a chord diagram visually by drawing an arc between the endpoints . in an oriented chord diagram
, the arc will be on the _ same _ side of the edges at the endpoints when the endpoints have _ opposite _ orientation , and on opposite sides when the endpoints have the same orientation .
this terminology is geometrically motivated - an oriented line segment in the plane naturally has left and right sides , and when the arc is on the same side of the edges at each endpoint , the bases for @xmath6 given by the ordered pair of the tangent vectors to the chord and the edge have opposite orientations at the two endpoints , as shown in figure [ f : orient ] .
[ h ] @xmath7 two examples of ( oriented ) chord diagrams are shown in figure [ f : chordexamples ] .
[ h ] @xmath8 much of the work done on chord diagrams for circles has been in terms of their intersection graphs , also known as _
circle graphs_. these graphs have also been useful in the recent applications of chord diagrams to knot and link invariants @xcite .
so it is natural to look at some kind of intersection graph for chord diagrams on more general graphs . for chord diagrams on the circle
, two chords are said to intersect if their endpoints alternate around the circle . in a more general graph , there may not be a way to move around " the graph , or there may be many , so we need a different criterion .
another point of view of circle graphs is as proper circular - arc graphs " - in other words , two chords intersect if the arcs on the circle between their endpoints overlap , but neither is a proper subset of the other @xcite . by replacing arcs on the circle " with paths in the graph , " we can generalize this notion to other graphs . [
d : igchord ] given a chord diagram c on a ( possibly directed ) graph g , the * intersection graph for c * is the graph @xmath9 whose vertices are the chords of c , and vertices @xmath2 and @xmath10 ( corresponding to chords @xmath2 and @xmath10 ) are adjacent if given any ( undirected ) path @xmath11 in g between the endpoints of @xmath2 and any ( undirected ) path @xmath12 in g between the endpoints of @xmath10 , @xmath11 and @xmath12 properly intersect ( meaning that they overlap - perhaps only in one vertex - but neither is a subset of the other ) .
so for two chords to intersect , _ all _ paths between their endpoints must intersect ; for them to be disjoint , there need only be one pair of paths which are disjoint ( or in which one properly contains the other ) .
this is clearly a fairly restrictive notion of intersection , but it seems to be the one which best matches our visual intuition .
figure [ f : igchord ] shows the intersection graphs for the chord diagrams in figure [ f : chordexamples ] .
note that , for chord diagrams on circles , this definition agrees with the usual definition of the intersection graph .
[ h ] @xmath13
a common situation is to study chord diagrams when the chords are real " chords - in other words , they are chords of regions in some planar embedding of the graph . in this section , we will explore chord diagrams and their intersection graphs when we are given a planar graph together with a particular embedding in the plane .
our main result is to give conditions when the embedding of the graph can be extended to an embedding of the entire chord diagram .
[ d : respect ] let @xmath14 be a directed planar graph , @xmath15 be an oriented chord diagram on @xmath14 , and @xmath16 be a particular embedding of @xmath14 . consider a chord @xmath2 of @xmath15 , with endpoints on edges @xmath17 and @xmath18 in @xmath14 .
we say @xmath2 * respects * @xmath19 if : 1 .
both endpoints of @xmath2 lie on the boundary of a single region r of @xmath20 ( i.e. a single connected component of @xmath21 ) .
if r is on the same side ( left or right ) of @xmath17 and @xmath18 , then the endpoints of @xmath2 have opposite orientations ; if r is on the opposite sides of @xmath17 and @xmath18 , then the endpoints of @xmath2 have the same orientation ( assuming @xmath17 and @xmath18 are not cut edges ) .
3 . if either @xmath17 or @xmath18 is a cut edge , then the endpoints of @xmath2 may have either orientation .
if @xmath14 is not directed , then only the first condition is required .
we say that @xmath15 * respects * @xmath19 if every chord of @xmath15 respects @xmath19
. we should observe that if @xmath17 and @xmath18 are both on the boundaries of two regions @xmath22 and @xmath23 , then the definition does not depend on which region is used ; if @xmath22 is on the same side of both edges , then so is @xmath23 , and vice - versa .
we define the intersection graph of @xmath15 with respect to @xmath19 by restricting our attention to paths in @xmath14 which lie on the boundary of a single region of @xmath20 .
[ d : igplanar ] given a chord diagram c on a ( possibly directed ) planar graph g , and an embedding @xmath16 , the * intersection graph for c with respect to f * is the graph @xmath24 whose vertices are the chords of c , and vertices @xmath2 and @xmath10 ( corresponding to chords @xmath2 and @xmath10 ) are adjacent if given any ( undirected ) path @xmath11 on the boundary of a single region of @xmath20 between the endpoints of @xmath2 and any ( undirected ) path @xmath12 on the boundary of a single region of @xmath20 between the endpoints of @xmath10 , @xmath11 and @xmath12 properly intersect .
remarks : 1 .
if @xmath2 does not have both endpoints on the boundary of a single region of @xmath20 , then there is no path between the endpoints which lies on the boundary of a single region . in this case
, the condition for adjacency is vacuously true , so @xmath2 is adjacent to every other chord in the chord diagram .
2 . the intersection graphs shown in figure [ f : igchord ] are also the intersection graphs with respect to the embeddings shown .
since the number of paths in a graph may be large , it is not clear whether @xmath9 can be constructed in polynomial time .
however , @xmath24 certainly can , since an endpoint for a chord can lie on the boundary of at most two regions , there are at most 4 possible paths between the endpoints for each chord .
so the number of pairs of paths which need to be compared is at most @xmath25 , where @xmath26 is the number of chords .
if two chords intersect by definition [ d : igchord ] , then they certainly intersect by definition [ d : igplanar ] , but the converse is not true .
so @xmath27 for every embedding @xmath19 .
the last remark gives rise to the following open question : given a planar graph g and a chord diagram c on g , is @xmath28 ?
our main focus in this section is to study when a planar embedding of a graph @xmath14 can be extended to give a planar embedding of a chord diagram @xmath15 on @xmath14 . to be precise , by an embedding of the chord diagram "
we mean an embedding of the graph @xmath29 constructed from @xmath14 by adding vertices at the endpoints of all the chords in @xmath15 , and edges between the endpoints of each chord in @xmath15 .
we will also look at embeddings of _ oriented _ chord diagrams - in this case , we will insist that the embedded chords be on the same or opposite sides of the edges at each endpoint , as dictated by the orientations ( see section [ ss : defchord ] ) .
our first result is well - known ( an equivalent form can be extracted from @xcite , for example ) , but is given here for completeness and as a warm - up for our later results .
[ p : circle ] let g be a directed graph consisting of a single loop , and let @xmath16 be an embedding of g ( so @xmath20 is a circle ) .
let c be an oriented chord diagram on g. then f extends to an embedding of c if and only the following conditions are met : 1 .
@xmath9 is bipartite .
the endpoints of every chord in c have opposite orientation . if c is not oriented , then the first condition is sufficient .
proof : we first note that , since every embedding of the circle in the plane is isotopic , the particular embedding @xmath19 is irrelevant . and , since @xmath20 has exactly two regions , both bounded by the circle , every path on the circle is also a path in the boundary of a region , so @xmath30 .
we first show the sufficiency of the conditions of the proposition .
since the endpoints of the chords have opposite orientations , the chord is on the same side of the circle at each endpoint , so each chord can be drawn entirely within one region of @xmath20 ( if @xmath15 is not oriented , we can simply choose to draw the chords on the same side of the circle at each endpoint ) . since @xmath9 is bipartite
, we can divide the chords into two sets , denoted the blue " chords and the red " chords , so that no two blue chords intersect , and no two red chords intersect .
we will draw the blue chords in one region of @xmath20 ( inside the circle ) and the red chords in the other region ( outside the circle ) .
choose one of the blue chords , @xmath31 , and connect the endpoints by an arc inside the circle .
this gives an embedding of @xmath32 , which divides the inside of the circle into two regions .
since no other blue chord intersects @xmath31 , any other chord @xmath33 must have both endpoints on the boundary of one of these regions , and so there is an arc connecting these endpoints in @xmath34 . continuing in this way , we can embed all of the blue arcs inside the circle ; by the same argument , we can embed all the red arcs outside the circle .
this gives us an extension of the embedding to @xmath15 .
conversely , if we have an embedding of @xmath15 , then each chord must lie in a single region of @xmath20 , and so must be on the same side of the circle at each endpoint , so the endpoints must have opposite orientations .
in addition , the chords inside the circle must all be disjoint , as are the chords outside the circle .
so the vertices of the intersection graph can be divided into two independent sets , and so @xmath9 is bipartite .
@xmath35 + + remarks : 1 .
since there are polynomial - time algorithms to determine whether a graph is bipartite , proposition [ p : circle ] shows that we can determine whether the embedding extends to the chord diagram in polynomial time .
proposition [ p : circle ] easily extends to general cycles ( in which the edges may not all be oriented the same way ) by replacing the condition that the endpoints of each chord have opposite orientation with the condition that every chord must respect the embedding , as in definition [ d : respect ] .
now we want to consider more complicated graphs .
topologically , the simplest graphs after the circle are the @xmath36-graphs , which are the graphs consisting of two vertices and @xmath26 edges between these vertices .
figure [ f : theta ] shows several examples .
[ h ] @xmath37 the spatial embeddings of these graphs have been studied extensively @xcite , so they are a natural set of graphs for us to consider .
[ p : theta ] let g be a directed @xmath36-graph with vertices v and w , and with all edges directed from v to w. let @xmath16 be a planar embedding of g , and let c be an oriented chord diagram on g. label the edges of g by @xmath38 according to their ( counterclockwise ) cyclic order around v in f(g ) , and label the region of f(g ) enclosed by @xmath39 and @xmath40 by @xmath41 ( @xmath42 is the region enclosed by @xmath43 and @xmath17 ) .
then f extends to an embedding of c if and only if the following conditions are met : 1 .
the vertices of @xmath24 can be labeled by the regions @xmath41 so that : * if a vertex is labeled @xmath41 , then the corresponding chord in c has both endpoints on the boundary of region @xmath41 in f(g ) . * for each @xmath41 , the set of vertices labeled @xmath41 is independent .
if both endpoints of a chord are on the same edge of g , then they have opposite orientations ; if the endpoints of a chord are on different edges of g , they have the same orientation ( i.e. @xmath15 respects @xmath19 , as in definition [ d : respect ] ) .
if c is not oriented , then we can ignore the second condition .
proof : we will first show that the conditions are sufficient .
the idea is simply to draw the chord @xmath31 corresponding to a vertex @xmath44 labeled @xmath41 in the region @xmath41 .
since both endpoints of @xmath31 lie on the boundary of @xmath41 , they can be connected by an arc in @xmath41 .
if both endpoints of @xmath31 are on one edge ( @xmath39 or @xmath40 ) , then they have opposite orientations ( since @xmath15 respects @xmath19 ) , and the chord is on the same side of the edge at both endpoints ; therefore , the arc is an embedding of the oriented chord .
on the other hand , if @xmath31 has one endpoint on @xmath39 and the other on @xmath40 , then its endpoints have the same orientation , and @xmath31 should be on opposite sides of the edges .
but the region @xmath41 is on the left side of @xmath39 and the right side of @xmath40 , so again the arc in @xmath41 is an embedding of the oriented chord .
so the embedding of @xmath14 can be individually extended to each chord in @xmath15 . we still need to show it can be extended to all the chords at once .
the obstruction to this extension is when chords cross each other - i.e. when their endpoints alternate around the boundary of a region .
we will first consider what happens if there _ is _ such an obstruction .
[ l : thetaalternate ] if two chords @xmath31 and @xmath33 both have their endpoints on the boundary of region @xmath41 , and their endpoints alternate along the ( undirected ) cycle around the boundary of @xmath41 , then the corresponding vertices @xmath44 and @xmath45 are adjacent in @xmath24 .
proof of lemma : if @xmath31 has both endpoints on the same edge , say @xmath39 , then @xmath33 has exactly one endpoint on the segment @xmath39 between the endpoints of @xmath31 , and any path between the endpoints of @xmath33 must contain one of the endpoints of @xmath31 , but not both .
so any path between the endpoints of @xmath33 will intersect , but not contain , any path between the endpoints of @xmath31 . on the other hand ,
if neither @xmath31 nor @xmath33 have both endpoints on the same edge , then they each have one endpoint on @xmath39 and one on @xmath40 .
then any path between the endpoints of the chords with lies on the boundary of a region of @xmath20 must run along the boundary of @xmath41 , since it is the only region whose boundary contains both endpoints .
there are two such paths between the endpoints of @xmath31 , each of which contains one endpoint of @xmath33 .
so we conclude that if the endpoints of @xmath31 and @xmath33 alternate around the boundary of @xmath41 , then @xmath44 and @xmath45 are adjacent .
this proves our lemma .
equivalently , if @xmath44 and @xmath45 are _ not _ adjacent , then the endpoints of @xmath31 and @xmath33 must _ not _ alternate around the boundary of @xmath41 .
since the set of vertices labeled @xmath41 is independent , none of the corresponding chords have alternating endpoints around the boundary of @xmath41 .
so these chords can be embedded in @xmath41 as disjoint arcs , as in proposition [ p : circle ] .
this can be done for each label @xmath41 , resulting in an embedding of @xmath15 which extends @xmath19 .
it remains to show that the conditions are necessary .
assume that @xmath19 can be extended to @xmath15
. then each chord is embedded as an arc in some region @xmath41 .
if both endpoints of the chord are on the same edge , the chord is on the same side at both endpoints , and the endpoints must have opposite orientations . on the other hand ,
if they are on distinct edges @xmath39 and @xmath40 , then the chord is on opposite sides of the edges at the two endpoints ( the left side of @xmath39 and the right side of @xmath40 , so the endpoints must have the same orientation .
so the second condition is necessary .
now label the vertex @xmath44 in @xmath24 corresponding to chord @xmath31 by the region in which @xmath31 is embedded in @xmath46 .
obviously , both endpoints of @xmath31 must be on the boundary of this region . since we have embedded our chord diagram , chords which are in the same region @xmath41 do not cross , so their endpoints do not alternate around the boundary of @xmath41 .
then there are clearly non - overlapping paths between the endpoints of the chords , so the corresponding vertices are not adjacent in @xmath24 .
so vertices with the same label will not be adjacent .
this shows that the first condition is necessary , and completes the proof .
@xmath35 + + remark : it is easy to check that these conditions can be checked in polynomial time .
the coloring of most chords is forced by the location of their endpoints .
those that are left have both endpoints on the same edge , and it is enough to show that the subgraph in @xmath24 induced by the remaining chords on each edge is bipartite . the arguments of section [ ss : theta ] can easily be generalized to any planar graph without _ cut edges _ ( so that each edge in a planar embedding of the graph bounds a different region on either side ) .
recall that @xmath47 is a _ cut edge _ of a connected graph @xmath14 if @xmath48 is not connected .
[ p : nocut ] let g be a directed planar graph with no cut edges .
let @xmath16 be a planar embedding of g , and let c be an oriented chord diagram on g. label the regions of f(g ) by @xmath41 ( @xmath49 number of regions ) .
then f extends to an embedding of c if and only if the following conditions are met : 1 .
c respects f , as in definition [ d : respect ] ( so every chord in c has both endpoints on the boundary of some region @xmath41 ) .
the vertices of @xmath24 can be labeled by the regions @xmath41 so that : * if a vertex is labeled @xmath41 , then the corresponding chord in c has both endpoints on the boundary of region @xmath41 in f(g ) . * for each @xmath41 , the set of vertices labeled @xmath41 is independent . before we begin the proof of proposition [ p : nocut ] , we need a technical lemma , which generalizes lemma [ l : thetaalternate ] in the last section . [
l : alternate ] consider g , f and c as in the statement of proposition [ p : nocut ] .
if c and d are chords in c whose endpoints alternate around the boundary of region @xmath50 in f(g ) , then the corresponding vertices @xmath44 and @xmath45 are adjacent in @xmath24 . proof : since @xmath14 does not have cut edges , the boundary of @xmath50 is a _ cycle _ in @xmath14 .
assume that @xmath44 and @xmath45 are _ not _ adjacent .
then there are paths @xmath51 between the endpoints of @xmath31 and @xmath52 between the endpoints of @xmath33 , each along the boundary of a region in @xmath20 , such that either @xmath53 or one path is contained in the other - without loss of generality , @xmath54 .
so neither of the endpoints of chord @xmath33 lie on @xmath51 . since the endpoints of @xmath31 and @xmath33 alternate around the boundary of @xmath50 , @xmath51
can not be contained in the boundary of @xmath50 , or it would contain one of the endpoints of @xmath33 .
so @xmath51 is in the boundary of some other region @xmath55 , meaning that the endpoints of @xmath31 lie on the boundary between regions @xmath50 and @xmath55 .
first consider the case when @xmath51 and @xmath52 are disjoint .
then the union of @xmath51 with one of the paths between the endpoints of @xmath31 along the boundary of region @xmath50 gives a cycle @xmath56 ( or at least a circuit which contains a cycle @xmath56 ) which contains one endpoint of @xmath33 .
so @xmath52 must intersect @xmath56 .
if @xmath52 lies along region @xmath50 , then it will contain one of the endpoints of @xmath31 , and so wo nt be disjoint from @xmath51 .
but since @xmath52 ca nt pass _ through _ region @xmath50 ( since this is a region of the planar embedding of @xmath14 ) , it must ( by the jordan curve theorem ) pass from one side of @xmath56 to the other , and so must intersect @xmath51 , as shown in figure [ f : alternate ] .
[ h ] @xmath57 this contradicts the assumption that @xmath51 and @xmath52 are disjoint . therefore , @xmath54 , so the endpoints of @xmath31 are in @xmath52 .
since the endpoints of @xmath31 are on the boundary between regions @xmath50 and @xmath55 , and we know that @xmath52 can not be on the boundary of region @xmath50 , it must lie on the boundary of region @xmath55 .
so the endpoints of @xmath33 are also on the boundary between regions @xmath50 and @xmath55 . since @xmath54 ,
the endpoints of @xmath31 and @xmath33 do _ not _ alternate around the boundary of region @xmath55 , which is a cycle .
so there are paths @xmath58 and @xmath59 between the endpoints of chords @xmath31 and @xmath33 on the boundary of region @xmath55 , with @xmath60 . but this is impossible , by the argument in the last paragraph . therefore , @xmath51 and @xmath52 must properly intersect , so @xmath44 and @xmath45 are adjacent .
@xmath35 + + proof of proposition [ p : nocut ] : our proof is very similar to the proof of proposition [ p : theta ] .
we will first show that the conditions are sufficient .
once again , the idea is simply to draw the chord @xmath31 corresponding to a vertex @xmath44 labeled @xmath50 in the region @xmath50 .
since both endpoints of @xmath31 lie on the boundary of @xmath50 , they can be connected by an arc in @xmath50 .
the first condition ensures that this arc will be an embedding of the oriented chord @xmath31 .
so the embedding of @xmath14 can be individually extended to each chord in @xmath15 .
we still need to show it can be extended to all the chords at once .
let @xmath33 be another chord on region @xmath50 , with corresponding vertex @xmath45 .
by lemma [ l : alternate ] , if @xmath44 and @xmath45 are _ not _ adjacent , then the endpoints of @xmath31 and @xmath33 must _ not _ alternate around the boundary of @xmath50 . since the set of vertices labeled @xmath50 is independent , none of the corresponding chords have alternating endpoints around the boundary of @xmath50 .
so these chords can be embedded in @xmath50 as disjoint arcs , as in proposition [ p : circle ] .
this can be done for each label @xmath50 , resulting in an embedding of @xmath15 which extends @xmath19 .
it remains to show that the conditions are necessary .
assume that @xmath19 can be extended to @xmath15 .
then each chord is embedded as an arc in some region @xmath50 , with endpoints on edges @xmath17 and @xmath18 . if @xmath50 is on the same side of both edges , then
so is the embedded chord , and the endpoints must have opposite orientations ; if it is on opposite sides , the endpoints must have the same orientation .
so the first condition is necessary .
now label the vertex @xmath44 in @xmath24 corresponding to chord @xmath31 by the region in which @xmath31 is embedded in @xmath46 .
obviously , both endpoints of @xmath31 must be on the boundary of this region .
chords which are in the same region @xmath50 do not cross , so their endpoints do not alternate around the boundary of @xmath50 ( since the boundary of @xmath50 is a cycle ) . then there are clearly non - overlapping paths between the endpoints of the chords , so the corresponding vertices are not adjacent in @xmath24 .
so vertices with the same label will not be adjacent .
this shows that the second condition is necessary , and completes the proof .
[ c : nocut ] given a planar graph @xmath14 with no cut edges , a chord diagram @xmath15 on @xmath14 , and a planar embedding @xmath16 , there is a polynomial time algorithm to determine whether @xmath19 extends to an embedding of @xmath15 .
proof : the conditions of proposition [ p : nocut ] can be checked in polynomial time . the coloring of most chords is forced by the location of their endpoints .
those that are left have both endpoints on boundaries of the same two regions , and it is enough to show that the subgraph in @xmath24 induced by the remaining chords lying on each pair of regions is bipartite .
@xmath35 now we want to consider planar graphs which _ do _ have cut edges . as an introduction , we will consider the simplest such graph , the complete graph on two vertices @xmath61 ( i.e. the line segment consisting of two endpoints and the edge between them ) .
why do cut edges pose difficulties ?
one difficulty is that the boundaries of the regions in a planar embedding of the graph are no longer cycles in the graph .
for example , we might expect a line segment to be no more complicated than a circle . but consider the example in figure [ f : lineexample ] .
[ h ] @xmath62 the sequence of endpoints along the graph is @xmath63 . according to definition [ d : igplanar ] , all three of these chords intersect .
so the intersection graph has a cycle of length 3 , and so is not bipartite .
and yet we _ can _ extend the embedding of the line segment to an embedding of the chord diagram , so proposition [ p : circle ] does not extend to @xmath61 .
the fact that chords can go around the end " of the graph increases the difficulty of the problem .
however , we would still like to exploit our results from section [ ss : nocut ] .
so our plan is to transform a graph _ with _ cut edges into a graph _ without _ cut edges , and hope solving the problem in this new graph will give us the solution for the original graph . to this end
, we introduce an operation called blowing up " an edge in a planar embedding of a graph or chord diagram .
[ d : blowup ] let g be a planar directed graph with an edge e ( which is not a loop ) , and let @xmath16 be a planar embedding of g. say that the endpoints of e are vertices v and w , with e oriented from v to w. then the * blowup of @xmath47 in @xmath14 with respect to @xmath19 * is the graph @xmath64 formed by removing edge @xmath47 and replacing it with two edges e and @xmath65 , both directed from @xmath66 to @xmath67 .
figure [ f : blowup ] illustrates this operation .
[ h ] @xmath68 remark : figure [ f : blowup ] also shows that the embedding @xmath19 of @xmath14 is easily modified to give an embedding @xmath69 of @xmath64 . in particular , this means that @xmath64 is also a planar graph . what is the effect of blowing up an edge @xmath47 on a chord diagram @xmath15 on @xmath14 ?
if @xmath70 is a chord of @xmath15 , label its endpoints by @xmath71 and @xmath72 .
if @xmath73 is on edge @xmath47 in @xmath14 , it will be doubled in @xmath64 , appearing on both @xmath47 and @xmath65 .
in other words , the sequence of endpoints along @xmath47 and @xmath65 will both be the same as the original sequence along @xmath47 .
the chord " @xmath70 now consists of all pairings of a point labeled @xmath71 and a point labeled @xmath72 ; there are @xmath74 such pairings , where @xmath75 is the number of endpoints of the original chord @xmath70 which were on edge @xmath47 .
we give each pairing its own name , @xmath76 , yielding a new chord diagram @xmath77 on @xmath64 .
if the chord diagram @xmath15 is oriented , then the orientations of @xmath71 and @xmath72 are also replicated in @xmath64 .
if the chord @xmath70 in @xmath15 respected the embedding @xmath19 , and @xmath70 had an endpoint on @xmath47 , then only half of the set of chords labeled @xmath70 in @xmath64 will respect the embedding @xmath69 .
so we obtain the _ reduced _ oriented chord diagram @xmath78 on @xmath64 by discarding those chords @xmath70 which do not respect @xmath69 . the intersection graph @xmath79 is defined as before .
we will add to this graph edges between all the chords which came from the chord @xmath70 , to get a new intersection graph @xmath80 .
figure [ f : chordblowup ] illustrates this process for an example with @xmath81 .
[ h ] @xmath82 [ p : line ] let @xmath83 be a directed line segment , and let @xmath16 be an embedding .
let @xmath15 be an oriented chord diagram on @xmath14 .
let @xmath64 be the result of blowing up the single edge @xmath47 of @xmath14 , @xmath69 be the resulting embedding of @xmath64 , and @xmath78 be the reduced oriented chord diagram on @xmath64 obtained from @xmath15 . then @xmath19 can be extended to an embedding of @xmath15 if and only if @xmath80 is bipartite .
proof : in this case , every chord in @xmath15 has both endpoints on @xmath47 , so each chord of @xmath15 corresponds to 4 chords in @xmath77 , and hence 2 chords in @xmath78 .
we first show the sufficiency . if @xmath80 is bipartite , then so is its subgraph @xmath79 . since every chord in @xmath15 respects @xmath19 ( since the only edge is a cut edge ) , @xmath15 respects @xmath19 ; so by definition @xmath78 respects @xmath69 and contains a chord corresponding to every chord in @xmath15 .
so by proposition [ p : circle ] , @xmath69 extends to an embedding of @xmath78 .
moreover , by using the partition of vertices from @xmath80 , we can extend @xmath69 so that exactly one of each pair of chords @xmath84 is inside the cycle @xmath64 and one is outside .
the chords which are outside give an extension of @xmath19 to @xmath15 .
conversely , if @xmath19 extends to @xmath15 , then we can draw the chords of @xmath15 as disjoint arcs in the single region of @xmath20 .
these arcs can then be drawn as arcs outside the cycle @xmath64 in @xmath85 .
but there is another extension of @xmath19 to @xmath15 obtained by taking the reflection of the first embedding across the axis defined by the image of @xmath14 .
these arcs can be drawn _ inside _ the circle @xmath85 .
together , these give an extension of @xmath69 to an embedding of @xmath78 , so @xmath79 is bipartite . since the arcs for the chords @xmath70 and
@xmath86 in this extension are on opposite sides of the circle , adding edges between them in the intersection graph will not affect the partition into two independent sets , so @xmath80 is also bipartite . @xmath35 notice that , in figure [ f : chordblowup ] , the intersection graph @xmath80 is bipartite , and the chord diagram @xmath78 can be embedded as shown .
we can extend the ideas of the last section to prove the analogue of proposition [ p : nocut ] for chord diagrams on graphs with cut edges .
our plan is simply to blow up every cut edge , as in section [ ss : chordline ] , and consider the resulting graph , which no longer has cut edges . consider a graph @xmath14 , and let @xmath87 be the set of cut edges .
let @xmath88 be the result of blowing up every edge in @xmath87 .
if @xmath15 is an oriented chord diagram on @xmath14 , then @xmath89 is the corresponding chord diagram on @xmath88 .
if @xmath16 is an embedding of @xmath14 in the plane , then we have a corresponding embedding @xmath90 of @xmath88 .
the regions of @xmath91 are essentially the same as the regions of @xmath20 , together with the new regions created by the blown up edges ( one new region for each cut edge in @xmath14 ) .
if we label the regions of @xmath20 by @xmath92 , then we can give the corresponding regions of @xmath91 the same labels ; in addition , we will give _ all _ the regions created by blowing up cut edges on the boundary of @xmath41 the same label @xmath93 .
the region " @xmath93 is the union of all these regions , and its boundary is the union of the images @xmath94 of the cut edges of @xmath41 in @xmath15 , in the same order that they are found in the boundary of @xmath41 in @xmath89 ( so the boundary of @xmath93 is a subset of the boundary of @xmath41 ) .
this is illustrated in figure [ f : boundaries ] .
[ h ] @xmath95 as usual , a chord of @xmath89 _ respects _ @xmath90 if it satisfies definition [ d : respect ] , treating @xmath93 as a single region .
so we can consider the reduced chord diagram @xmath96 which results from removing all the chords in @xmath89 which do not respect @xmath90 .
the intersection graph @xmath97 is defined as usual , with the exception that a path " on the boundary of a region @xmath93 is taken to be a sequence of consecutive edges in the boundary of @xmath93 ( as defined in the last paragraph ) , even though these edges may not be adjacent in @xmath88 . as in section [ ss : chordline ] , @xmath98 is defined by adding to @xmath97 edges between vertices corresponding to chords in @xmath96 which came from the same chord in @xmath15 .
[ p : cut ] let g be a directed planar graph with a set of cut edges e , and let @xmath88 be the result of blowing up the edges in e , as described above .
let @xmath16 be a planar embedding of g , and let c be an oriented chord diagram on g. label the regions of f(g ) by @xmath41 ( @xmath49 number of regions ) .
we define @xmath90 and @xmath89 as described above .
then f extends to an embedding of c if and only if the following conditions are met : 1 .
c respects f. 2 .
the vertices of @xmath98 can be labeled by the regions @xmath41 and @xmath93 so that : * if a vertex is labeled @xmath50 , then the corresponding chord in @xmath96 has both endpoints on the boundary of region @xmath50 in @xmath91 . * for each region
@xmath50 ( @xmath41 or @xmath93 ) , the set of vertices labeled @xmath50 is independent . *
if @xmath70 was a chord of @xmath15 with both endpoints on a cut edge in the boundary of @xmath41 , then exactly one of the chords @xmath84 in @xmath96 is labeled @xmath41 , and the other is labeled @xmath93 .
proof : we first show the sufficiency . from proposition
[ p : nocut ] , the conditions imply that @xmath90 extends to an embedding of @xmath96 . moreover , for every chord in @xmath15 , there is exactly one corresponding chord of @xmath96 whose image lies in the union of the regions @xmath41 ( as opposed to the regions @xmath93 ) .
so we can then collapse " the regions @xmath93 ( reversing the blowup operations ) to obtain an embedding of @xmath15 which extends @xmath19
. for the necessity , assume that @xmath19 does extend to @xmath15 .
then , for any chord @xmath70 of @xmath15 such that @xmath99 is an arc in the region @xmath41 , there is a corresponding arc in the region @xmath41 in the complement of @xmath91 .
these arcs are all still disjoint .
this provides an embedding of almost all chords of @xmath96 ; label the vertices in @xmath98 corresponding these chords by the region in which each is embedded .
the vertices labeled @xmath41 will still be independent , as they were in @xmath24 .
the only remaining chords are doubles " of chords which have already been embedded in some @xmath41 , and which correspond to chords in @xmath15 which had both endpoints on cut edges on the boundary of @xmath41 .
so these chords have both endpoints on the boundary of @xmath93 , and the corresponding vertices can be labeled @xmath93 .
it only remains to check that the vertices labeled @xmath93 will be independent .
however , the order of the endpoints of the chords labeled @xmath93 around the boundary " of @xmath93 is exactly the reverse of the order of the endpoints of their doubles " around the boundary of @xmath41 .
so the endpoints of two chords alternate around the boundary of @xmath93 only if the endpoints of their doubles alternate around the boundary of @xmath41 . since the vertices labeled @xmath41 are all independent , so are the vertices labeled @xmath93 . so the conditions of the proposition are satisfied , completing the proof .
[ c : cut ] given a planar graph @xmath14 , a chord diagram @xmath15 on @xmath14 , and a planar embedding @xmath16 , there is a polynomial time algorithm to determine whether @xmath19 extends to an embedding of @xmath15 .
proof : from corollary [ c : nocut ] , we can check whether @xmath90 extends to @xmath96 in polynomial time .
the only condition that is left to check is the last one , that one of each pair @xmath84 can be labeled @xmath41 and the other @xmath93 . since , for a connected graph
, there is at most one division into two independent sets ( vertices with paths which are odd or even length to a given vertex ) , this can be done very quickly after determining that the set of chords with endpoints on @xmath41 and @xmath93 is bipartite . @xmath35
as an application , we will use chord diagrams and intersection graphs for graphs to study gauss codes for graphs .
the gauss code was first developed by c. f. gauss to study closed curves in the plane which intersect themselves only in transverse double points ( called crossings " ) .
label the crossings by some set of symbols ( such as integers or the letters of the alphabet ) , and give the curve an orientation .
beginning at an arbitrary crossing , we obtain a gauss code for the curve by writing down the sequence of labels of the crossings passed as we traverse the curve , following its orientation .
the gauss code provides a convenient combinatorial representation of the curve .
however , not all sequences which look like " gauss codes actually represent closed curves in the plane . the problem is to determine ( in polynomial time ) which sequences of symbols can be obtained from a closed curve in this way - i.e. , which sequences are _ realizable _ as the gauss code for a closed curve in the plane .
there are several different solutions to this problem @xcite , we will describe one in section [ ss : circle ] which uses proposition [ p : circle ] .
kauffman s recent development of the theory of virtual knots @xcite was motivated in part by a desire to realize the unrealizable " gauss codes .
this paper was largely motivated by a desire to extend kauffman s work to spatial graphs , and study virtual " spatial graphs . as a preliminary ,
we need to extend the notion of a gauss code to general graphs , and consider when an abstract code is realizable by an immersion of a graph in a plane . in future work , this will lead to the introduction of virtual spatial graphs @xcite . our goal is to generalize to the situation where we have an immersion of an arbitrary graph in the plane , rather than a circle . in this case , as described in section [ ss : graph ] , the gauss code " is a set of sequences associated with the edges of the graph .
we will give a algorithm , based on proposition [ p : cut ] , for determining whether an abstract gauss code can be realized as the gauss code for an immersed graph and , if so , constructing the desired immersion .
we will first describe a method for determining whether a classical gauss code is realizable by an immersed closed curve in the plane .
this problem was first solved by dehn @xcite , and a clear presentation of his solution is given by read and rosenstiehl @xcite .
our method builds on dehn s , and is equivalent to that given by de fraysseix and ossona de mendez @xcite , but is presented in the language of chord diagrams developed in section [ s : chord ] .
an oriented closed curve in the plane gives rise to a crossing sequence , the _ gauss code _ for the curve , by labeling the @xmath26 self - crossings of the curve , and then writing down the sequence of @xmath100 labels encountered as we trace the curve following its orientation ( obviously , the code is only unique modulo cyclic permutations ) .
an example is shown in figure [ f : gaussloop ] .
the problem , initially posed by gauss , is to determine whether an arbitrary sequence of length @xmath100 , containing two occurrences each of @xmath26 symbols , is realizable as the gauss code of a closed plane curve .
we will call such a sequence a _ crossing sequence _ , and refer to the symbols as _ crossings_. we follow read and rosenstiehl by defining a _ splitting _ of a crossing sequence at each crossing , and then studying the resulting _ split sequence _ or _ split
code_. the notion of splitting a crossing is motivated by the idea of smoothing a self - intersection of a curve in the plane into two non - intersecting arcs .
we consider a small neighborhood of a self - crossing @xmath102 . inside this neighborhood , we replace the two intersecting arcs with two disjoint arcs with the same endpoints , as shown in figure [ f : split ] .
this replacement is done so as to preserve the number of components of the curve , which requires reversing the orientation of one of the arcs from @xmath102 to itself , so that the resulting curve is still oriented . aside from the change of orientation , the operation is entirely local , so if we label points @xmath104 and @xmath105 on the two new arcs , we can draw a chord @xmath106 between them which does not intersect the rest of the curve and is contained within the original neighborhood of the crossing , as shown in figure [ f : split ] .
notice that , if the chord @xmath106 is oriented , its endpoints at @xmath104 and @xmath105 will have opposite orientations . [
d : splitting ] given a sequence @xmath107 , * splitting s at p * means to replace s by @xmath108 , where we say that @xmath104 and @xmath105 have opposite orientations , and @xmath109 is the result of writing the subsequence @xmath110 in reverse order and reversing the orientation of any previously split symbols in @xmath110 . given a crossing sequence @xmath55 , the result of splitting @xmath55 at
_ every _ crossing is called a _ split sequence _ or _ split code _ for @xmath55 , and is denoted @xmath111 .
the split code is _ not _ unique - it depends upon the order in which the crossings are split .
figure [ f : splitcode ] illustrates this procedure with an example of smoothing all the crossings in a closed plane curve , and the corresponding operations on the gauss code in order to obtain the split code .
the result of performing the splitting process on a closed plane curve is a _ simple _ closed curve ( i.e. an embedded circle ) labeled by the split code .
we will call this the _
split curve_. clearly , the split curve is an oriented chord diagram on the circle , in the sense of definition [ d : orientchord ] . to reconstruct the original curve from the split curve , we simply need to connect each pair of points @xmath104 and @xmath105 by the chord @xmath106 shown in figure [ f : split ] , and contract the chords to bring the points back together , reversing the splitting process . so reconstructing the original curve
is done by extending the embedding of the underlying circle of the split curve to an embedding of the oriented chord diagram represented by the split curve .
this leads directly to the following result .
[ t : closedcurve ] @xcite a crossing sequence s is realizable as the gauss code of a closed curve in the plane if and only if the the oriented chord diagram on the circle represented by the split sequence s * can be embedded in the plane .
proof : from the discussion above , if the crossing sequence @xmath55 is realizable , then there is an embedding of the chord diagram @xmath111 .
conversely , an embedding of @xmath111 can be turned into a collection of intersecting closed loops by expanding each chord into a crossing , reversing the splitting operation in figure [ f : split ] .
it only remains to check that this process will yield a single closed curve with gauss code @xmath55 ( this is the _ complete traceability _ of @xcite ) .
say that @xmath111 was produced from @xmath55 via a series of sequences @xmath113 , where @xmath114 is the result of splitting crossing @xmath11 in @xmath93 .
we will expand the chords between pairs of symbols in @xmath111 in the reverse order from how they were split - i.e. the last symbol to be split is the first chord added .
so we begin with the chord @xmath115 between @xmath116 and @xmath117 in @xmath118 . expanding this chord into a crossing , and reversing
the orientation on the arc of @xmath119 originally directed from @xmath116 to @xmath117 , exactly reverses the splitting process of definition [ d : splitting ] .
a key observation is that , since @xmath116 and @xmath117 have opposite orientations , the result of expanding the chord is a closed curve with one component .
since we have exactly reversed the final splitting operation , the result is an oriented closed curve @xmath120 whose gauss code is @xmath121
. we can continue this process , always choosing to expand the chord corresponding to the last crossing which was split , so that its endpoints have opposite orientation and the resulting curve still has a single component . in this way
, we can successively construct an oriented closed curve @xmath122 for each sequence @xmath93 .
ultimately , we will have a curve @xmath15 whose gauss code is the original crossing sequence @xmath55 , as desired .
@xmath35 + + remarks : there are several remarks that should be made at this point . 1 .
to check whether a particular split code is embeddable , we apply the criteria of proposition [ p : circle ] .
so we check that the two occurrences of each symbol have opposite orientation and that the intersection graph is bipartite .
the importance of theorem [ t : closedcurve ] is that this can be done in polynomial time . 2 .
comparing with the _ d - switch _ operation of de fraysseix and ossona de mendez @xcite , the condition that the two occurrences of each symbol have opposite orientation is equivalent to saying that a symbol and its twin are not adjacent in the interlacement graph .
so , the result of theorem [ t : closedcurve ] then also follows from the proof of theorem 6 in @xcite .
kauffman @xcite provides another characterization , in which the condition that each pair of symbols have opposite orientations is replaced by the condition that the crossing sequence @xmath55 is _ evenly intersticed _
, meaning that between two occurrences of any symbol is a word of even length .
however , kauffman does not prove that this requirement is sufficient .
it is not hard to show that , if @xmath55 is evenly intersticed , then the number of unsplit symbols between a pair of split symbols ( at any stage in the process of splitting @xmath55 ) is even exactly when the split symbols have opposite orientations . since , in @xmath111 , the number of unsplit symbols is 0 , and hence even , all pairs of split symbols have opposite orientations , so by proposition [ p : circle ] and theorem [ t : closedcurve ] the crossing sequence is realizable .
this shows that kauffman s characterization is also valid .
not every sequence of symbols @xmath111 which satisfies the conditions of theorem [ t : closedcurve ] is the split code of a crossing sequence . for example
, if the reconstruction algorithm is applied to the sequence @xmath123 , the result will be a curve with two components ; however , this is not a contradiction because this sequence _ can not _ be obtained as the split code of a crossing sequence . to see this ,
observe that if we unsplit " symbol @xmath124 we are left with @xmath125 - since the two occurrences of @xmath126 have the same orientation , this could not be the result of a splitting operation ; attempting to unsplit @xmath126 leads to a similar situation .
we would like to extend the ideas of the last section to general directed graphs . in this section
, our graphs will be _ vertex - oriented _ , meaning that we have a cyclic order on the edges incident to each vertex . given a graph @xmath14 , a _ gauss code for g _ consists of a set of symbols @xmath126 and a set of sequences in these symbols , one for each edge of the graph .
each sequence will begin with the vertex at the start of the edge , and the position of the edge in the cyclic ordering around that vertex , and end with the vertex at the end of the edge , with the position of the edge in the cyclic ordering around the end vertex . in between are symbols from @xmath126 .
notice that the abstract graph @xmath14 can be reconstructed from the set of gauss codes . while an abstract graph has many possible gauss codes , we can assign a unique code ( up to cyclic permutation of the vertex orderings ) to any immersion of the graph in the plane , just as we do for immersions of circles . in this case ,
the sequence of symbols just records the order of the crossings along each edge .
an example is shown in figure [ f : example ] .
our first observation is that every symbol in the set of sequences ( other than the vertices ) must appear exactly twice .
this allows us to create a _ split code _ from the original gauss code . as in the previous sections ,
the motivation for the split code is to split the crossings of an immersion of a graph in the plane .
there are two ways to split a crossing ( one is shown in figure [ f : split ] ) - we want to perform the split so that the graph remains connected . if both splits preserve the connectedness of the graph , we can choose one arbitrarily .
as before , we also want to keep track of the relative local orientations of the smoothed arcs .
these operations can be defined abstractly in terms of the gauss code .
if @xmath102 is a symbol in the alphabet @xmath126 , then splitting the code at @xmath102 means one of the following operations : 1 .
if there is a sequence @xmath128 , then the new sequence is @xmath129 .
so the subsequence @xmath110 has been reversed .
if there are two sequences @xmath130 and @xmath131 , then the new sequences are either @xmath132 and @xmath133 or @xmath134 and @xmath135 .
these are illustrated in figure [ f : split2 ] .
notice that , in the second case , the underlying graph is changed in either choice , but at least one of these graphs is connected .
we will always make a choice which keeps the underlying graph connected . also notice that the first choice preserves orientations along each edge segment , while the second choice reverses the orientations along segments coming from one of the two edges .
the result of performing this splitting operation on every symbol in the set of sequences gives a split code for the original gauss code .
the underlying abstract graph for the split code is called the _
split graph_. the splitting process is not unique - the final set of sequences after all crossings have been split will depend on the order in which they were split , and how they were split .
any of these possibilities is called a split code for the original gauss code .
a split code can be viewed as an oriented chord diagram on its split graph .
[ l : planar ] if a gauss code is derived from an immersion of a connected graph , then the underlying split graph for any split code is planar and connected ; moreover , the split code determines a unique ( up to isotopy ) spherical embedding of the split graph .
proof : if we begin with an immersion of a graph and split every crossing , we are left with a graph embedded in the plane ; this implies that the graph underlying the split code is planar , and the split code is realized by some planar embedding .
the embedding is connected since we choose our splittings to keep the graph connected .
it remains to show that this embedding is unique ( up to isotopy on the sphere ) .
we will describe an algorithm for constructing a planar embedding from the split code .
we will show that ( up to isotopy on the sphere ) the algorithm produces at most one embedding , and that it fails to produce an embedding respecting the split code only if one does not exist .
in fact , we will not use all the information of the split code , only the underlying graph and the order of the edges around each vertex .
the idea of the algorithm is to trace out faces of the planar embedding .
begin with an arbitrary vertex @xmath66 , and an edge @xmath47 incident to @xmath66 .
embed @xmath66 and @xmath47 .
say that @xmath67 is the other endpoint of @xmath47 .
let @xmath19 be the next edge adjacent to @xmath67 after @xmath66 ( using the cyclic ordering of the edges adjacent to @xmath67 ) , and add @xmath19 to the embedded graph .
continue this process , moving one edge around each vertex .
the first time we return to a vertex we have visited before , we will have two choices of how to draw the edge - clockwise or counterclockwise around the previously embedded parts of the graph .
but these choices are isotopic on the sphere .
once this first choice is made , all future edges will connect two points on the boundary of a ( topological ) disk , so all possible ways to draw them will be isotopic . as a result , there are no more choices to be made , and the resulting embedding is unique . when we come to an edge we have reached before , skip it and move to the next unused edge on that vertex . if there are no more unused edges on the vertex , then backtrack to the last vertex with an unused edge .
continue in this way until all the edges have been drawn ( the procedure will terminate , since the graph is connected ) .
if we are able to do this without edges crossing , we will have a planar embedding of the split graph , uniquely determined by the underlying graph and the vertex orientations .
@xmath35 lemma [ l : planar ] allows us to embed the split graph in the plane . as in earlier sections , if the original gauss code is realizable then we can recover an immersion of a graph from this embedding code by connecting the two occurrences of each symbol in the split graph by disjoint chords which respect the orientations of the symbols , and then expanding the chords as in figure [ f : expand2 ] . proof : if @xmath55 is realizable , then the split code @xmath111 can be embedded in the plane as an oriented chord diagram .
so the underlying split graph can be embedded ( and is therefore planar ) , and this embedding extended to the chord diagram .
conversely , if the chord diagram can be embedded , we can expand each chord into a crossing as in figure [ f : expand2 ] .
it only remains to check that the result is an immersion of @xmath14 with gauss code @xmath55 . as in theorem [ t
: closedcurve ] , we expand the chords between symbols in the reverse order that the symbols were split .
say that @xmath111 was produced from @xmath55 via a series of sequences @xmath113 , where @xmath114 is the result of splitting crossing @xmath11 in @xmath93 . since each splitting was chosen to keep the graph connected , expanding the splits in the reverse order
will also keep the graph connected .
our goal is for these expansions to exactly reverse the effects of the splitting process on the original gauss code .
there is one possible ambiguity : the expansion process shown in figure [ f : expand2 ] assumes that crossings between different edges were always split so as to preserve the orientations of the edges ; but , in order to keep the graphs connected , it may have been necessary to reverse the orientations of one of the two edges , as shown in figure [ f : split2 ] .
so the expansion may leave one of the edges with the
wrong " orientation .
however , since we have recorded all the sequences @xmath93 , this is easily fixed - simply compare the code for our current graph with the corresponding @xmath93 .
they will differ in at most the orientation of one edge , so we can simply reverse the orientation of this edge in our graph , and continue the process .
the final result will be an immersion of @xmath14 with gauss code @xmath55 .
@xmath35 proof : producing the split code can be done in linear time , and there are linear - time algorithms for determining whether the split graph is planar @xcite . finally , by corollary [ c : cut ] , we can determine whether the embedding of @xmath138 extends to @xmath111 in polynomial time as well . @xmath35
+ remark : in the theory of spatial graphs , a spatial embedding of a graph can be described by the _
gauss code _ for a particular diagram of the knot ( a projection of the knot to the plane ) . in this context
, the gauss code comes with additional information at each crossing , indicating which edge crosses over the other and the sign of the crossing .
corollary [ c : gauss ] implies that there is a polynomial time algorithm to determine whether a gauss code is realizable in this context as well .
once we have determined whether the code can be realized by an immersion of the graph in the plane , and constructed such an immersion , using theorem [ t : graph ] , we simply need to check whether the sign and over / under information at each crossing are compatible .
10 d. bar - natan : on the vassiliev knot invariants , _ topology _ 34 , no .
423 - 472 , 1995 d. bar - natan and s. garoufalidis : on the melvin - morton - rozansky conjecture , _ invent .
_ 125 , pp .
103 - 133 , 1996 j. birman and x. s. lin : knot polynomials and vassiliev s invariants , _ invent .
_ 111 , pp . 225 - 270 , 1993 a. bouchet : circle graph obstructions , _ j. comb .
theory , series b _ 60 , p.107 - 144 , 1994 j. boyer and w. myrvold : on the cutting edge : simplified @xmath139 planarity by edge addition , _ j. graph algorithms and applications _ , vol . 8 , no .
241 - 273 , 2004 s. v. chmutov , s. v. duzhin , and s. k. lando : vassiliev knot invariants i , ii , iii , _ advances in soviet mathematics _ 21 , pp .
117 - 145 , 1994 m. dehn : ber kombinatorische topologie , _ acta math .
_ 67 , 1936 , pp .
123 - 168 s. even and a. itai : queues , stacks and graphs , in _ theory of machines and computations _
( z. kohavi and a. paz , eds ) , academic press , new york , 1971 , pp .
71 - 86 t. fleming and b. mellor : virtual spatial graphs , preprint , 2005 , arxiv : math.gt/0510158 h. de fraysseix and p.
ossona de mendez : on a characterization of gauss codes , _ discrete comput .
_ , v. 22 , 1999 , pp .
287 - 295 h. goda : bridge index for theta curves in the 3-sphere .
_ topology appl .
_ , v. 79 , 1997 , pp .
177 - 196 .
l. kauffman : virtual knot theory , _ europ .
j. combinatorics _ , v. 20 , 1999 , pp .
663 - 691 s. kinoshita : on @xmath140-curves in @xmath141 and their constituent knots , _ topology and computer science ( atami , 1986 ) _ , kinokuniya , tokyo , 1987 , pp .
211 - 216 s. lando : on a hopf algebra in graph theory ,
_ j. comb .
theory , series b _ 80 , p.104 - 121 , 2000 t. a. mckee and f. r. mcmorris : _ topics in intersection graph theory _ , siam , philadelphia , 1999 b. mellor : the intersection graph conjecture for loop diagrams , _ j. knot theory ramif .
_ 9 , no . 2 , 2000 b. mellor : a few weight systems arising from intersection graphs , _ michigan math
. j. _ , vol .
509 - 536 , 2003 r. nikkuni : delta edge - homotopy on theta curves , _ math .
cambridge philos .
_ , v. 138 , 2005 , pp .
401 - 420 .
r. c. read and p. rosenstiehl : on the gauss crossing problem , _ colloq .
janos bolyai _ , v. 18 , _ combinatorics _ , keszthely , hungary , 1976 , pp
. 843 - 876 | chord diagrams on circles and their intersection graphs ( also known as _ circle graphs _ ) have been intensively studied , and have many applications to the study of knots and knot invariants , among others . however , chord diagrams on more general graphs have not been studied , and are potentially equally valuable in the study of spatial graphs .
we will define chord diagrams for planar embeddings of planar graphs and their intersection graphs , and prove some basic results . then , as an application , we will introduce gauss codes for immersions of graphs in the plane and give algorithms to determine whether a particular crossing sequence is realizable as the gauss code of an immersed graph . |
heavy quark symmetry @xcite implies that in the @xmath8 limit matrix elements of the weak currents between a @xmath0 meson and an excited charmed meson vanish at zero recoil ( where in the rest frame of the @xmath0 the final state charmed meson is also at rest ) . however , in some cases at order @xmath1 these matrix elements are not zero @xcite
. since most of the phase space for semileptonic @xmath0 decay to excited charmed mesons is near zero recoil , @xmath9 corrections can be very important .
this paper is concerned with rates for @xmath0 semileptonic decay to excited charmed mesons , including the effects of @xmath1 corrections .
the use of heavy quark symmetry resulted in a dramatic improvement in our understanding of the spectroscopy and weak decays of hadrons containing a single heavy quark , @xmath10 . in the limit where the heavy quark mass goes to infinity , @xmath8 ,
such hadrons are classified not only by their total spin @xmath11 , but also by the spin of their light degrees of freedom ( i.e. , light quarks and gluons ) , @xmath12 @xcite . in this limit hadrons containing a single heavy quark
come in degenerate doublets with total spin , @xmath13 , coming from combining the spin of the light degrees of freedom with the spin of the heavy quark , @xmath14 .
( an exception occurs for baryons with @xmath15 , where there is only a single state with @xmath16 . )
the ground state mesons with @xmath17 flavor quantum numbers contain light degrees of freedom with spin - parity @xmath18 , giving a doublet containing a spin zero and spin one meson . for @xmath19
these mesons are the @xmath2 and @xmath20 , while @xmath21 gives the @xmath0 and @xmath22 mesons .
excited charmed mesons with @xmath23 have been observed .
these are the @xmath3 and @xmath4 mesons with spin one and two , respectively .
( there is also evidence for the analogous @xmath21 heavy meson doublet . ) for @xmath24 , the @xmath3 and @xmath4 mesons have been observed to decay to @xmath25 and are narrow with widths around @xmath26mev .
( the @xmath27 and @xmath28 strange mesons decay to @xmath29 . ) in the nonrelativistic constituent quark model these states correspond to @xmath30 orbital excitations . combining the unit of orbital angular momentum with the spin of the light antiquark leads to states with @xmath31 and @xmath32 .
the @xmath33 doublet , @xmath34 , has not been observed .
presumably this is because these states are much broader than those with @xmath35 . a vast discrepancy in widths is expected since the members of the @xmath33 doublet of charmed mesons decay to @xmath25 in an @xmath36-wave while the members of the @xmath32 doublet of charmed mesons decay to @xmath25 in a @xmath2-wave .
( an @xmath36-wave @xmath37 amplitude is allowed by total angular momentum conservation , but forbidden in the @xmath8 limit by heavy quark spin symmetry @xcite . )
the heavy quark effective theory ( hqet ) is the limit of qcd where the heavy quark mass goes to infinity with its four velocity , @xmath38 , fixed .
the heavy quark field in qcd , @xmath10 , is related to its counterpart in hqet , @xmath39 , by @xmath40 h_v^{(q ) } \,,\ ] ] where @xmath41 and the ellipses denote terms suppressed by further powers of @xmath1 .
putting eq .
( [ hqetfield ] ) into the part of the qcd lagrangian involving the heavy quark field , @xmath42 , gives @xmath43 the hqet lagrangian @xcite @xmath44 is independent of the mass of the heavy quark and its spin , and so for @xmath45 heavy quarks with the same four velocity @xmath38 there is a @xmath46 spin - flavor symmetry .
this symmetry is broken by the order @xmath1 terms @xcite in @xmath47 , @xmath48 \,,\ ] ] where @xmath49 the first term in eq .
( [ lag ] ) is the heavy quark kinetic energy .
it breaks the flavor symmetry but leaves the spin symmetry intact .
the second is the chromomagnetic term , which breaks both the spin and flavor symmetries .
( in the rest frame , it is of the form @xmath50 , where @xmath51 is the heavy quark color magnetic moment . )
the hadron masses give important information on some hqet matrix elements .
the mass formula for a spin symmetry doublet of hadrons @xmath52 with total spin @xmath53 is @xmath54 where the ellipsis denote terms suppressed by more powers of @xmath1 and @xmath55 is the number of spin states in the hadron @xmath52 .
the parameter @xmath56 is the energy of the light degrees of freedom in the @xmath8 limit , @xmath57 determines the heavy quark kinetic energy , satisfy the standard covariant normalization @xmath58 . ] @xmath59 and @xmath60 determines the chromomagnetic energy @xmath61 @xmath56 and @xmath57 are independent of the heavy quark mass , while @xmath60 has a weak logarithmic dependence on @xmath62 . of course they depend on the particular spin symmetry doublet to which @xmath52 belong . in this paper , we consider heavy mesons in the ground state @xmath63 doublet and the excited @xmath23 and @xmath33 doublets .
we reserve the notation @xmath56 , @xmath57 , @xmath60 for the ground state multiplet and use @xmath64 , @xmath65 , @xmath66 and @xmath67 , @xmath68 , @xmath69 for the excited @xmath23 and @xmath33 doublets , respectively .
the average mass @xmath70 , weighted by the number of helicity states @xmath71 is independent of @xmath60 .
the spin average masses for the lowest lying charmed mesons is given in table [ tab : charm ] . identifying the @xmath72 resonances observed at lep with the bottom @xmath35 meson doublet we can use their average mass , @xmath73gev @xcite , to determine the differences @xmath74 and @xmath75 : @xmath76\over m_b - m_c } \simeq -0.23\ , { \rm gev}^2 \ , .
\end{aligned}\ ] ] the numerical values in eq .
( [ hqet_diff ] ) follow from the choices @xmath77gev and @xmath78gev . to the order we are working , @xmath79 and @xmath80 in eq .
( [ hqet_diff ] ) can be replaced by @xmath81 and @xmath82 .
this changes the value of @xmath74 only slightly , but has a significant impact on the value of @xmath75 . the value of @xmath74 given in eq .
( [ hqet_diff ] ) has considerable uncertainty because the experimental error on @xmath83 is large , and because it is not clear that the peak of the @xmath72 mass distribution corresponds to the narrow @xmath32 doublet . and
@xmath84 masses could also be used to determine @xmath74 from the relation @xmath85 with the analog of eq .
( [ hqet_diff ] ) used to fix @xmath86 , and @xmath87mev
. the @xmath88 has not been observed , but its mass can be determined from @xmath89 $ ] .
however , because of uncertainties in the @xmath90 and @xmath84 masses and the unknown order @xmath91 term , this relation does not give a more reliable determination of @xmath74 than eq .
( [ hqet_diff ] ) . ] .charmed meson spin multiplets ( @xmath24 ) . [ cols="^,^,^,^,^,^",options="header " , ] we considered the zero recoil matrix elements of the weak currents between a @xmath0 meson and other excited charmed mesons at order @xmath1 .
only the corrections to the states contribute and these were expressed in terms of matrix elements of local operators .
our results have implications for @xmath0 decay sum rules , where including the contributions of the excited charmed meson states strengthens the bounds on @xmath92 ( the slope of the isgur - wise function for @xmath93 ) , on @xmath57 , and on the zero recoil matrix element of the axial - vector current between @xmath0 and @xmath20 mesons .
the latter bound has implications for the extraction of @xmath94 from exclusive @xmath95 decay .
we thank a. le yaouanc , m. neubert , a. vainshtein for useful discussions , and a. falk for keeping us from being disingenuous .
this work was supported in part by the department of energy under grant no .
de - fg03 - 92-er 40701 .
in this appendix we compute order @xmath7 and order @xmath97 corrections to the @xmath98 form factors . at this order both the current in eq .
( [ 1/mcurrent ] ) and the order @xmath1 corrections to the lagrangian in eq .
( [ lag ] ) receive corrections .
matrix elements of the kinetic energy operator , @xmath99 , enter proportional to @xmath100 to all orders in @xmath7 due to reparameterization invariance @xcite .
the matrix elements involving the chromomagnetic operator are probably very small and have been neglected .
order @xmath7 corrections to the @xmath101 flavor changing current in the effective theory introduce a set of new operators at each order in @xmath1 , with the appropriate dimensions and quantum numbers .
the wilson coefficients for these operators are known @xmath102-dependent functions @xcite , which we take from @xcite .
the vector and axial - vector currents can be written at order @xmath7 as @xmath103\ , h^{(b)}_v + { \alpha_s\over\pi}\ , \big [ v^{\mu\,(1 ) } + v^{\mu\,(2 ) } \big ] + \ldots \ , , \nonumber\\ * a^\mu & = & \bar h^{(c)}_{v'}\ , \bigg [ \gamma^\mu\gamma_5-{i\overleftarrow d\!\!\!\!\!\slash\ , \gamma^\mu\gamma_5 \over 2\,m_c } + { i\gamma^\mu\gamma_5 \overrightarrow d\!\!\!\!\!\slash \over 2\,m_b}\ , \bigg]\ , h^{(b)}_v + { \alpha_s\over\pi}\ , \big [ a^{\mu\,(1 ) } + a^{\mu\,(2 ) } \big ] + \ldots \,,\end{aligned}\ ] ] where the ellipses denote terms higher order in @xmath7 and @xmath1 .
superscripts @xmath104 denote corrections proportional to @xmath7 , @xmath105\ , h^{(b)}_v \ , , { \nonumber}\\ * a^{\mu\,(1 ) } & = & \bar h^{(c)}_{v'}\ , \big [ c_{a1}\gamma^\mu+c_{a2}v^\mu+c_{a3}v'^\mu \big]\,\gamma_5\ , h^{(b)}_v \,.\end{aligned}\ ] ] the terms with superscript @xmath106 in eq .
( [ a1 ] ) denote corrections proportional to @xmath107 , @xmath108 \\ * & - & { i\overleftarrow d_{\!\lambda}\over 2\,m_c}\ , \bigg [ 2 c_{v3 } g^{\mu\,\lambda } + \bigg(\gamma^\lambda + 2v^\lambda\ , { \frac{\overrightarrow\partial}{\partial w}}\bigg ) \big(c_{v1}\gamma^\mu+c_{v2}v^\mu+c_{v3}v'^\mu\big ) \bigg ] \bigg\}\ , h^{(b)}_v , { \nonumber}\\ * a^{\mu\,(2 ) } = \bar h^{(c)}_{v'}\ , & \bigg\ { & { i\overrightarrow d_{\!\lambda}\over 2\,m_b}\ , \bigg [ \big(c_{a1}\gamma^\mu+c_{a2}v^\mu + c_{a3}v'^\mu\big)\,\gamma_5\ , \bigg(\gamma^\lambda + 2v'^\lambda\ , { \frac{\overleftarrow\partial}{\partial w}}\bigg ) + 2 c_{a2 } g^{\mu\,\lambda}\,\gamma_5\ , \bigg ] { \nonumber}\\ * & - & { i\overleftarrow d_{\!\lambda}\over 2\,m_c}\ , \bigg [ 2 c_{a3 } g^{\mu\,\lambda } + \bigg(\gamma^\lambda + 2v^\lambda\ , { \frac{\overrightarrow\partial } { \partial w}}\bigg ) \big(c_{a1}\gamma^\mu+c_{a2}v^\mu+c_{a3}v'^\mu\big ) \bigg]\ , \gamma_5\ , \bigg\}\ , h^{(b)}_v \,.{\nonumber}\end{aligned}\ ] ] in these expressions the covariant derivatives , @xmath109 , act on the fields @xmath110 or @xmath111 , and partial derivatives with respect to @xmath102 , @xmath112 , act on the coefficient functions @xmath113 and @xmath114 .
using eqs .
( [ a2 ] ) and ( [ a3 ] ) it is straightforward to include the order @xmath7 and @xmath107 corrections using trace formalism presented in sec .
ii . the corrections with superscript ( 1 ) simply change the form of @xmath115 in eq .
( [ lo ] ) , while those with superscript ( 2 ) change @xmath115 in eq .
( [ 1/mc1 ] ) .
the @xmath116 form factors were defined in eq .
( [ formf1 ] ) , and their expansions in terms of isgur - wise functions at leading order in @xmath7 were given in eq .
( [ expf ] ) .
the order @xmath7 and order @xmath107 corrections modify the results for @xmath117 in eq .
( [ expf ] ) to @xmath118 .
the functions @xmath119 are given by @xmath120 \tau { \nonumber}\\ * & + & \varepsilon_c\,(w-1)\ , \big\ { [ 3c_{a1}-2(w-1)c_{a3 } ] \tau_1- ( 3c_{a1}+4c_{a3})\tau_2 \big\ } { \nonumber}\\ * & - & \varepsilon_b\,\big[(\bar\lambda'+\bar\lambda)(w-1)c_{a1}- 2(\bar\lambda'-w\bar\lambda)(w+1)c_{a1}'+2\,(w\bar\lambda'-\bar\lambda ) c_{a2}\big ] \tau { \nonumber}\\ * & + & \varepsilon_b\,(w-1)\ , \big\ { [ ( 2w+1)c_{a1}-2(w-1)c_{a2 } ] \tau_1 + ( c_{a1}-4c_{a2})\tau_2 \big\ } , \end{aligned}\ ] ] @xmath121 \tau { \nonumber}\\ * & + & \varepsilon_c\ , ( w^2 - 1 ) \big\ { [ 3c_{v1}+2(w+2)c_{v3 } ] \tau_1 - ( 3c_{v1}+2c_{v3})\tau_2 \big\ } { \nonumber}\\ * & - & \varepsilon_b\ , ( w+1 ) \big[(\bar\lambda'+\bar\lambda)(w-1)c_{v1 } - 2(\bar\lambda'-w\bar\lambda)(w-1)c_{v1}'+4(w\bar\lambda'-\bar\lambda)c_{v2 } \big ] \tau { \nonumber}\\ * & + & \varepsilon_b\,(w^2 - 1 ) \big\ { [ ( 2w+1)c_{v1}+2(w+2)c_{v2 } ] \tau_1 + ( c_{v1}-2c_{v2})\tau_2 \big\ } , \end{aligned}\ ] ] @xmath122 \tau - 2\varepsilon_c\ , ( w\bar\lambda'-\bar\lambda ) \big [ 3c_{v1}'+2c_{v2}+2(w+1)c_{v2 } ' \big ] \tau { \nonumber}\\ * & - & \varepsilon_c \big\ { [ ( 4w-1)c_{v1}-2(2w+1)(w-1)c_{v2}-2(w+2)c_{v3 } ] \tau_1 { \nonumber}\\ * & & \phantom{\varepsilon_c } + [ 5c_{v1}+2(1-w)c_{v2}+2c_{v3}]\tau_2 \big\ } { \nonumber}\\ * & - & \varepsilon_b\ , \big\ { 3 ( \bar\lambda'+\bar\lambda)c_{v1}- 6(\bar\lambda'-w\bar\lambda)c_{v1}'+2[(w-1)\bar\lambda'+(3w+1)\bar\lambda ] c_{v2}{\nonumber}\\ * & & \phantom{\varepsilon_b } - 4(\bar\lambda'-w\bar\lambda)(w+1)c_{v2 } ' \big\}\tau { \nonumber}\\ * & + & \varepsilon_b \big\ { [ 3(2w+1)c_{v1}+2(2w^2 + 1)c_{v2 } ] \tau_1 + [ 3c_{v1}+ 2(w-2)c_{v2 } ] \tau_2 \big\ } , \end{aligned}\ ] ] @xmath123 \tau { \nonumber}\\ * & + & 2\varepsilon_c\ , ( w\bar\lambda'-\bar\lambda ) \big\ { 2c_{v1 } + ( w-2)c_{v1}'- 2 [ c_{v3}+(w+1)c_{v3 } ' ] \big\ } \tau { \nonumber}\\ * & + & \varepsilon_c \big\ { [ ( 2+w)c_{v1}+2(w^2 - 3w-1)c_{v3 } ] \tau_1 + [ ( 3w+2)c_{v1}+(4w-2)c_{v3 } ] \tau_2 \big\ } { \nonumber}\\ * & + & \varepsilon_b\ , \big [ ( \bar\lambda'+\bar\lambda)(w+2)c_{v1}+ 2(\bar\lambda'-w\bar\lambda ) ( 2-w)c_{v1}'+4\bar\lambda'(w+1)c_{v2}{\nonumber}\\ * & & \phantom{\varepsilon_b } - 2(\bar\lambda'+\bar\lambda)(w-1)c_{v3 } + 4(\bar\lambda'-w\bar\lambda)(w+1)c_{v3 } ' \big ] \tau { \nonumber}\\ * & - & \varepsilon_b\ , \big\ { [ ( 2w^2 + 5w+2)c_{v1}+2w(2+w)c_{v2}+2(1+w-2w^2)c_{v3 } ] \tau_1 { \nonumber}\\ * & & \phantom{\varepsilon_b } + [ ( 2+w)c_{v1}-2wc_{v2}-2(w-1)c_{v3}]\tau_2 \big\ } .\end{aligned}\ ] ] here @xmath124 and @xmath125 are functions of @xmath102 , and prime denotes a derivative with respect to @xmath102
. note that at zero recoil @xmath126 is known in terms of @xmath74 and @xmath127 , as expected from our results in sec .
ii . for @xmath128 decay , the @xmath7 and order @xmath107 corrections modify the leading order form factors in eq .
( [ expk ] ) to @xmath129 .
the functions @xmath130 are @xmath131 \\ * & - & \varepsilon_b\,\big\ { [ ( \bar\lambda'+\bar\lambda)c_{v1}-2 ( \bar\lambda'-w\bar\lambda ) c_{v1}']\tau - [ ( 2w+1)c_{v1}+2wc_{v2 } ] \tau_1 - ( c_{v1}+2c_{v2})\tau_2 \big\ } , { \nonumber}\end{aligned}\ ] ] @xmath132 - \varepsilon_b\ , \big\ { [ ( \bar\lambda'+\bar\lambda)(w-1)c_{a1}- 2(\bar\lambda'-w\bar\lambda)(w+1)c_{a1}'{\nonumber}\\ * & & -2(w\bar\lambda'-\bar\lambda)c_{a2}]\tau - ( w-1)[c_{a1}(\tau_1+\tau_2 ) + 2(wc_{a1}-wc_{a2 } -c_{a2})\tau_1 ] \big\},\end{aligned}\ ] ] @xmath133 \tau_1 + c_{a2}\tau_2 \big\ } { \nonumber}\\ * & + & \varepsilon_b\,\big\ { [ ( \bar\lambda'+3\bar\lambda)c_{a2 } - 2(\bar\lambda'-w\bar\lambda)c_{a2 } ' ] \tau - ( 2w+3)c_{a2}\tau_1 - c_{a2}\tau_2 \big\ } , \end{aligned}\ ] ] @xmath134 { \nonumber}\\ * & + & \varepsilon_b\,\big\ { [ ( \bar\lambda'+\bar\lambda)(c_{a1}+c_{a3 } ) -2 \bar\lambda'c_{a2 } - 2 ( \bar\lambda'-w\bar\lambda ) ( c_{a1}'+c_{a3}')]\tau { \nonumber}\\ * & & \phantom{\varepsilon_b } -(c_{a1}+c_{a3})(\tau_1+\tau_2 ) - 2w(c_{a1}-c_{a2}+c_{a3})\tau_1 \big\ } .\end{aligned}\ ] ] to compute the corrections to the results obtained in sec .
ii , it is sufficient to expand the wilson coefficients @xmath124 and @xmath125 to linear order in @xmath102 .
we take @xmath124 and @xmath125 and their first derivatives at zero recoil from ref .
@xcite . to evaluate these
, we choose to integrate out the @xmath135 and @xmath136 quarks at a common scale @xmath137 , giving for @xmath124 and @xmath125 @xmath138 \over 3(1-z)^3 } \simeq -1.20 \ , , { \nonumber}\\ * c_{a3}(1 ) & = & { 2z\ , [ 1 + 2z-3z^2+(5z-1)\ln z ] \over 3 ( 1-z)^3 } \simeq 0.42 \,.\end{aligned}\ ] ] the derivatives @xmath139 and @xmath140 at zero recoil are @xmath141 \over 27\,(1-z)^3 } \simeq 0.20 \ , , { \nonumber}\\ * c_{v2}'(1 ) & = & { 2 [ 2 + 3z-6z^2+z^3 + 6z\,\ln z ] \over 9\,(1-z)^4 } \simeq 0.21 \ , , { \nonumber}\\ * c_{v3}'(1 ) & = & { 2z [ 1 - 6z+3z^2 + 2z^3 - 6z^2\,\ln z ] \over 9\,(1-z)^4 } \simeq 0.05 \ , , { \nonumber}\\ c_{a1}'(1 ) & = & - { 2 [ 7 + 9z-9z^2 - 7z^3 + 3(2 + 3z+3z^2 + 2z^3)\,\ln z ] \over 27\,(1-z)^3 } \simeq 0.64 \ , , { \nonumber}\\ * c_{a2}'(1 ) & = & { 2 [ 2 - 33z+9z^2 + 25z^3 - 3z^4 - 6z(1 + 7z)\,\ln z]\over 9\,(1-z)^5 } \simeq 0.37 \ , , { \nonumber}\\ * c_{a3}'(1 ) & = & - { 2z [ 3 - 25z-9z^2 + 33z^3 - 2z^4 - 6z^2(7+z)\,\ln z]\over 9\,(1-z)^5 } \simeq -0.12 \ , .
\end{aligned}\ ] ] here @xmath142 , and the numbers quoted are for @xmath143 .
m. feindt , talk given at the second international conference on @xmath0 physics and @xmath144 violation , honolulu , march 1997 ; + g. eigen , talk given at the seventh international symposium on heavy flavor physics , santa barbara , july 1997 .
s. godfrey and n. isgur , phys .
d32 ( 1985 ) 189 ; + s. godfrey and kokoski , phys .
d43 ( 1991 ) 1679 ; + a.b .
kaidalov and a.v .
nogteva , sov .
47 ( 1988 ) 321 ; + j. rosner , comments nucl .
( 1986 ) 109 . | exclusive semileptonic @xmath0 decays into excited charmed mesons are investigated at order @xmath1 in the heavy quark effective theory .
differential decay rates for each helicity state of the four lightest excited @xmath2 mesons ( @xmath3 , @xmath4 , @xmath5 , and @xmath6 ) are examined . at zero recoil ,
@xmath1 corrections to the matrix elements of the weak currents can be written in terms of the leading isgur - wise functions for the corresponding transition and meson mass splittings .
a model independent prediction is found for the slope parameter of the decay rate into helicity zero @xmath3 at zero recoil .
the differential decay rates are predicted , including @xmath1 corrections with some model dependence away from zero recoil and including order @xmath7 corrections . ratios of various exclusive branching ratios are computed .
matrix elements of the weak currents between @xmath0 mesons and other excited charmed mesons are discussed at zero recoil to order @xmath1 .
these amplitudes vanish at leading order , and can be written at order @xmath1 in terms of local matrix elements . applications to @xmath0 decay sum rules and factorization are presented . |
a series of precision measurements @xcite of the polarized structure function @xmath0 off proton and deuteron targets has considerably improved our knowledge on the spin structure of the nucleon over the last two years . in combination with several older measurements @xcite
, these experiments now cover an @xmath1-range of @xmath2 , although the @xmath3 range for fixed @xmath1 is still rather restricted . in the ` naive ' parton model @xmath4
can , like the unpolarized structure function @xmath5 , be expressed in terms of the probability distributions for finding quarks with spin parallel or antiparallel to the longitudinally polarized parent proton : @xmath6 \\ g_1(x , q^2 ) & = & \frac{1}{2 } \sum_q\ ; e_q^2\ ; [ \delta q(x ) + \delta\bar q(x ) ] \ ; , \label{naiveg1}\end{aligned}\ ] ] where @xmath7 furthermore , the naive parton model predicts exact scaling behaviour for the above distributions , i.e. independence of the @xmath3 scale of the measurement .
these distributions are intrinsic , nonperturbative features of the nucleon , and can therefore at present only be determined from a fit to the structure function data .
some insight can however be gained from thermodynamical models @xcite or from the light - cone wavefunctions of partons in the nucleon @xcite .
perturbative qcd yields corrections @xcite to the simple parton model picture , which are manifest in a scale - dependence of the parton distributions .
the quantitative features of these corrections will be discussed in section 2 .
various groups have used the recent data on @xmath0 to determine the polarized parton distributions , taking into account the leading - order qcd corrections .
the concepts of the various approaches and their results are compared in section 3 .
finally , section 4 contains a brief summary .
in the qcd corrected parton model @xmath4 is expressed in terms of parton densities for the polarization of quarks and gluons , @xmath8\ ; \nonumber \\ & & \times \left\{\delta(1-y ) + { \alpha_s(q^2 ) \over 2\pi } \delta c_q(y ) + \ldots \right\ } \nonumber \\ & & + \ ; \frac{1}{9 } \int_x^1 \ ; { dy\over y}\ ; \delta g(x / y , q^2 ) \ ; \left\ { n_f\ ; { \alpha_s(q^2 ) \over
2\pi}\;\delta c_g(y ) + \ldots \right\}. \label{qcdg1}\end{aligned}\ ] ] the emission of collinear partons gives rise to an evolution of the parton densities @xcite , @xmath9 this equation only determines how the distributions change with @xmath3 , not the distributions themselves .
the boundary conditions for the solution enter as initial distributions @xmath10 and @xmath11 .
even if the precise form of the polarized parton distributions is still not yet known , several qualitative features of the @xmath3 dependence can be determined from the splitting functions ( fig . 1 ) .
the evolution of the polarized valence quark density is identical to the unpolarized one : the distribution decreases in the large @xmath1 region and increases for smaller @xmath1 .
the polarized singlet distribution is changed by two splitting processes : @xmath12 lowers the distribution at large @xmath1 and increases at small @xmath1 while @xmath13 slightly increases it at large @xmath1 and lowers it at small @xmath1 .
the overall change of @xmath14 is therefore sensitive to the relative magnitude of the quark and gluon polari- zations .
the evolution of @xmath15 is dominated by the @xmath16 splitting , which strongly increases the distribution at small and medium @xmath1 .
only if the gluon polarization is initially smaller than the total quark polarization , will effects from @xmath17 contribute visibly to the increase of @xmath15 .
figure 1 : qualitative description of the evolution of polarized valence quark ( a ) , polarized singlet quark ( b ) and polarized gluon ( c ) densities . solid line : @xmath18 at @xmath19 , dashed line : @xmath20 at @xmath21 .
although the change of the polarized parton distributions with increasing @xmath3 is determined by the above evolution equations , the distributions themselves are incalculable in perturbative qcd .
the starting distributions at some low scale @xmath19 reflect the _ nonperturbative _ spin - structure of the nucleon and can only be fitted to the experimental data .
various groups have performed these fits using the leading - order evolution equations . in the following
we will restrict our discussion to the most recent parametrizations from glck - reya - vogelsang ( grv @xcite ) , the la plata group ( lp @xcite ) and to our own results ( gs @xcite ) . these parametrizations differ in various aspects : * * data selection : * + + ' '' '' + expt . &
@xmath1 & @xmath22 $ ] & grv & lp & gs94 & gs95 + + ' '' '' + e130-p & 0.180 0.7 & 3.5 10 & & @xmath23 & @xmath23 & @xmath23 + + ' '' '' + emc - p & 0.010 0.7 & 1.5 70 & @xmath23 & @xmath23 & @xmath23 & @xmath23 + + ' '' '' + smc - p & 0.003 - 0.7 & 1.0 60 & @xmath23 & @xmath23 & @xmath23 & @xmath23 + + ' '' '' + e143-p & 0.029 0.8 & 1.3 10 & @xmath23 & & & @xmath23 + + ' '' '' + smc - d93 & 0.006 0.6 & 1.0 30 & @xmath23 & @xmath23 & & @xmath23 + + ' '' '' + smc - d95 & 0.003 0.7 & 1.0 60 & & & & @xmath23 + + ' '' '' + e143-d & 0.029 0.8 & 1.0 30 & @xmath23 & & & @xmath23 + + ' '' '' + e142-n & 0.030 0.6 & 1.0 10 & @xmath23 & & @xmath23 & @xmath23 + * * construction of @xmath0 : * the above experiments have measured the cross section asymmetry @xmath24 therefore , one has to use @xmath25 for the extraction of @xmath0 .
the various parametrizations use either the unpolarized parton distributions of refs .
@xcite ( grv ) and @xcite ( lp ) or the slac / nmc parametrizations @xcite ( gs ) . *
* positivity of the distributions : * the most fundamental constraint on the polarized parton distribution is the positivity of the individual helicity distributions .
this forces the polarized distributions to be less in magnitude than the unpolarized ones , @xmath26 this constraint is either incorporated by defining polarizations ( grv , lp ) @xmath27 or by constraining combinations of parameters in the initial distributions ( gs ) . *
* constraints from the ellis - jaffe sum rule : * the ellis - jaffe sum rule @xcite @xmath28 relates the first moment of @xmath0 to the expectation values @xmath29 of the conserved nonsinglet axial vector currents . the nonconserved singlet axial vector current @xmath30 is then matched to the experimental data .
note that the ellis - jaffe sum rule is independent of @xmath3 in a leading - order model .
therefore , only an average value of all experimental measurements can be reproduced .
furthermore , the above procedure is not unique , as the different parametrizations use different assumptions for the flavour decomposition of the sea quark polarization . *
* constraints from asymptotic estimates : * the experimental data on @xmath0 are presently insufficient for a global determination of all polarized parton distributions . therefore some properties such as the behaviour at small and large @xmath1 have to be estimated from theoretical considerations .
the estimates used differ from parametrization to parametrization , the most common being : color coherence and regge arguments at small @xmath1 , and counting rules at large @xmath1 ( see ref .
@xcite for a recent review ) .
as the data on @xmath31 and @xmath32 have almost the same statistical quality , the non - singlet distributions @xmath33 and @xmath34 can be determined from these measurements .
figure 2 illustrates the agreement between the parametrizations .
the differences arise mainly from the data included in the fit and from different assumptions on the magnitude of the sea quark polarization entering the ellis - jaffe sum rule .
figure 2 : parametrizations of the polarized valence quark distributions .
the sea quark and gluon distributions are much more sensitive to the various constraints used to determine the distributions .
as these differ for all parametrizations , the corresponding distributions span a wide range of possibilities , as can be seen in figs . 3 and 4 .
figure 3 : parametrizations of the polarized sea ( @xmath35 ) and polarized strange ( @xmath36 ) quark distributions .
figure 4 : parametrizations of the polarized gluon distribution . without experimental data on processes other than the structure function @xmath4 , it will be very difficult to distinguish between these different possiblities .
future experiments at rhic , cern and slac , together with final state measurements at a polarized hera collider @xcite and at hermes , could provide this information on @xmath37 and @xmath38 from various different processes .
in this talk , we have given a brief review of the quantitative features of the @xmath3 evolution of polarized parton densities .
comparing various recent parametrizations of these densities , we have found that the present data on @xmath39 determine the valence quark polarization in the nucleon to some accuracy .
in contrast , the polarization of the quark sea and the gluon are strongly dependent on additional theoretical constraints imposed on the distributions .
this situation is not expected to improve significantly with more precise data on @xmath4 , which clearly shows the need for complementary measurements on polarized nucleons .
financial support from the uk pparc ( wjs ) , and from the gottlieb daimler- und karl benz - stiftung and the studienstiftung des deutschen volkes ( tg ) is gratefully acknowledged .
this work was supported in part by the eu programme `` human capital and mobility '' , network `` physics at high energy colliders '' , contract chrx - ct93 - 0357 ( dg 12 coma ) .
999 smc collaboration : b. adeva et al .
, phys . lett . *
b302 * ( 1993 ) 553 , d. adams et al .
, phys . lett . *
b329 * ( 1994 ) 399 .
+ slac - e143 collaboration : k. abe et al . , phys .
* 74 * ( 1995 ) 346 ; * 75 * ( 1995 ) 25 .
slac - yale collaboration : m.j .
alguard et al .
* 37 * ( 1976 ) 1261 ; g. baum et al . ,
* 45 * ( 1980 ) 2000 ; * 51 * ( 1983 ) 1135 .
+ emc collaboration : j. ashman et al .
, nucl . phys . *
b328 * ( 1989 ) 1 . + slac - e142 collaboration : d.l .
anthony et al .
* 71 * ( 1993 ) 959 .
c. bourrely and j. soffer , nucl .
b445 * ( 1995 ) 341 .
brodsky , m. burkhardt and i. schmidt , nucl .
b441 * ( 1995 ) 197 . m.a .
ahmed and g.g .
ross , phys .
lett . * 56b * ( 1975 ) ; nucl .
* b111 * ( 1976 ) 441 .
+ k. sasaki , progr .
* 54 * ( 1975 ) 1816 .
+ g. altarelli and g. parisi , nucl
b126 * ( 1977 ) 298 .
r. mertig and w.l .
van neerven , leiden preprint inlo - pub-6 - 95 .
m. glck , e. reya , m. stratmann and w. vogelsang , dortmund preprint do - th 95/14 .
m. glck , e. reya and w. vogelsang , dortmund preprint do - th 95/13 . d. de florian et al .
, phys . rev .
* d51 * ( 1995 ) 37 .
t. gehrmann and w.j .
stirling , z. phys .
* c65 * ( 1995 ) 461 .
t. gehrmann and w.j .
stirling , durham preprint dtp/95/82 , in preparation .
m. glck , e. reya and a. vogt , z. phys . * c67 * ( 1995 ) 433 .
martin , r.g .
roberts and w.j .
stirling , phys .
b306 * ( 1993 ) 145 .
whitlow , phd thesis , stanford university ( 1990 ) , slac - report 357 .
+ p. amaudruz et al . , phys .
b295 * ( 1992 ) 159 . j. ellis and r.l .
jaffe , phys . rev .
* d9 * ( 1974 ) 1444 , erratum * d10 * ( 1974 ) 1669 . g. ladinsky , d. von harrach , s.e .
rock , v. breton , these proceedings . | a review of the present knowledge on polarized parton distributions is given .
the effects of perturbative evolution on these distributions are discussed qualitatively and a comparison of various recent parametrizations is made .
dtp/95/78 + september 1995 + 1.cm * present status of polarized * + + t. gehrmann and w.j .
stirling + + _ talk given at the workshop on the prospects of spin physics at hera , desy - zeuthen , august 28 - 31 , 1995 _ |
the density - morphology relation @xcite , indicating an increasing elliptical and s0 population fraction with increasing density , has been known to exist over six orders of magnitude in galaxy density @xcite .
it is only recently that large surveys have shown just how smoothly the galaxy properties change with galaxy density .
for example @xcite find that gas deficiency gradually decreases out to two abell radii from the cluster center and @xcite and @xcite find that the star formation rate depends on local galaxy density .
these results have led to the concept that galaxies may already be affected by their environment well before they fall into the dense cluster core ( _ pre - processing _ ) . while many processes have been suggested to affect galaxies in the cluster environment such as ram - pressure stripping @xcite , starvation @xcite , harassment @xcite and tidal distortions due to the cluster potential and galaxy - galaxy interactions @xcite , it is still an open question what would affect the galaxies in the lower density outskirts of the clusters .
recently @xcite studied one such case , ngc 4522 , which is at a large projected distance from the center of virgo , yet appears to be undergoing ram - pressure stripping due to bulk motions of the icm , as the m49 subcluster is merging with virgo proper . in our new vla hi survey of the virgo cluster , we find candidates that appear to be feeling the cluster influence for the first time in the outskirts of the cluster . a number of galaxies located at intermediate distances from m87 reveal long gas tails , all pointing roughly away from the cluster center . here
we test our assumption that these tail galaxies are recent arrivals in the cluster and review possible causes of their gas tails , i.e. how galaxies can be modified before they enter the dense cluster core .
viva , vlaimaging of virgo galaxies in atomic gas , is a new vla hi survey of spiral galaxies in the virgo cluster .
the goal of the survey is to study details of different environmental effects .
the virgo cluster is ideal for this study not only due to its nearness but also because it is dynamically young @xcite and very likely to contain various mechanisms at work .
we have sampled @xmath150 galaxies throughout the cluster from near the dense core to the outskirts .
the selected galaxies are located at projected distances of 0.3 - 3.3 mpc from the cluster center assuming a distance of 16 mpc to virgo @xcite , which corresponds to @xmath2 - 4.3 virial radii @xcite .
the new survey was done with the vla in the c - short array and additional 10 fields ( 13 galaxies ) previously observed in c array were also included in the database .
the typical spatial and spectral resolutions are 15@xmath3 and 10 km s@xmath4 , respectively , which are a factor of 3 and 4 better than the previous vla hi survey of virgo spirals @xcite .
the rms of the data is typically 0.4 mjy , yielding an hi column density sensitivity of @xmath5 @xmath6 in 3@xmath7 .
the data have been calibrated and continuum subtracted with the astronomical imaging processing system ( aips ) .
the data from the nrao archive have been processed in the same way as the new survey data . to maximize the sensitivity
, we applied a weighting scheme intermediate between uniform and natural but somewhat closer to a natural weighting scheme ( robust=1 , * ? ? ? * ) .
we also made tapered images with twice the beamsize to bring out faint extended features .
we will present further details of the observations and the data reduction in chung et al .
( 2007 , in preparation ) .
[ tbl - tail ] llrrrrrcrrrrrrrrlrrc & & & & + & & & & & & & & & & & & & & & & & & & + & & & & & & & & & & & & & & & & & & & + 4294 & sbc & 12.3 & 359 & 14 & 17 & 15.9 & 172 & @xmath8 & 27 & 8.2 & 2.0 & 11 - 26 & 0.7 & 1.04 & 0.97 & n4299 & 127 & 27 & 0.2 + 4299 & sabd & 12.5 & 237 & 7 & 12 & 11.8 & 169 & @xmath9 & 14 & 9.0 & 2.6 & 11 - 18 & 0.7 & 1.09 & 0.93 & n4294 & 127 & 27 & 0.2 + 4302 & sc & 12.5 & 1150 & 24 & 26 & 14.9 & 199 & @xmath10 & 16 & 8.9 & 0.9 & 10 - 20 & 0.9 & 0.74 & 1.24 & n4298 & 14 & 11 & 0.5 + 4330 & sc & 13.1 & 1564 & 25 & 13 & 4.1 & 180 & @xmath11 & 13 & 5.0 & 0.4 & 6 - 13 & 0.6 & 1.19 & 0.86 & v0706 & 75 & 77 & 4.3 + 4396 & scd & 13.1 & -125 & 15 & 14 & 9.9 & 170 & @xmath12 & 13 & 3.6 & 0.7 & 19 - 39 & 1.0 & 0.64 & 1.37 & n4419 & 133 & 186 & 1.1 + 4424 & sba & 12.4 & 439 & 16 & 10 & 1.7 & 158 & @xmath13 & 22 & 10.7 & 0.8 & 3 - 6 & 0.9 & 0.77 & 1.21 & n4445 & 79 & 74 & 1.2 + 4654 & sc(r ) & 11.0 & 1038 & 23 & 28 & 34.3 & 300 & @xmath14 & 32 & 12.9 & 4.1 & 20 - 47 & 0.9 & 0.68 & 1.32 & n4639 & 27 & 81 & 1.2 + + + +
the survey collected hi data on 53 galaxies in total , 48 spirals and 5 systems of other types ( i m , de ) . among them , 7 spiral galaxies revealed a long hi extension ( see plate 1 ) .
these tail galaxies have the following properties in common in the hi morphology : 1 ) the hi is extended well beyond the optical disk only on one side ; 2 ) the tails differ from tidal bridges , i.e. there is no optical counterpart at the tip of the tail down to @xmath15 mag arcsec@xmath16 in the sloan digital sky survey ( sdss ) images for extended features like the tails ; 3 ) the projected length of the tail is larger than the half of the symmetric part in hi ( @xmath17 ; see table 1 ) . besides the similarities in gas morphology , the tails are also pointing roughly away from m87 as shown with arrows in plate 1 .
these hi tail galaxies are located at intermediate distances of @xmath0 mpc from m87 in projection .
all but ngc 4654 are found to the west of m87 , where the x - ray emission appears to be more elongated .
they show a wide range of hi gas deficiency ( see * ? ? ?
* for the definition ) , from @xmath9 to @xmath18 .
the hi mass in the tail varies from 10 to 40% of the total .
the optical and hi properties are summarized in table 1 and detailed descriptions of individual tails are given in the following section .
* ngc 4299/4294 * : these two galaxies are separated by 5.7@xmath19 ( @xmath20 kpc ) in projection and @xmath21 km s@xmath4 in velocity . the tail of ngc 4294 is almost perpendicular to the stellar ( and the hi ) disk and pointing to the sw .
ngc 4299 has one tail pointing to the sw , parallel to ngc 4294 s tail and another tail pointing to se .
the second tail of ngc 4299 is much broader and lower in hi surface density than the other one .
the combined hi mass of ngc 4299 s tails is @xmath22 , @xmath23 of the total emission .
the hi position - velocity diagram ( pvd , figure [ fig - tailxv ] ) shows a 2@xmath7 connection between the se end of ngc 4294 s disk , where the stellar disk is more extended .
we see no optical counterpart along the hi tails or between the two galaxies down to the limit mentioned .
* ngc 4302 * : the hi is mildly truncated within the stellar disk in the south ( @xmath24 , but a faint stellar envelope is seen beyond the hi disk ) and the gas tail is extended to the north with no optical counterpart .
a nearby galaxy , ngc 4298 is only 2.4@xmath19 ( @xmath25 kpc ) to the west at a similar velocity ( @xmath26 km s@xmath4 ) .
the hi of ngc 4298 is more extended to the nw while its stellar disk is more extended to the se .
the two galaxies overlap in velocity but there is no hi in between down to 2@xmath7 .
while ngc 4298 looks disturbed both optically and in hi , ngc 4302 does not show any obvious signatures of a tidal disturbance .
* ngc 4330 * : its hi gas is truncated within the stellar disk to the ne ( @xmath27 ) and the hi tail extends to the sw .
a deeper optical image shows that also to the sw the undisturbed stellar disk extends well beyond the hi tail ( abramson et al .
, in preparation ) .
we do not see any obvious optical counterpart along the hi tail but its galex ( galaxy evolution explorer ) image shows a @xmath28 tail from the sw end of the optical disk along the hi to the nw , which is displaced from the optical disk . * ngc 4396
* : the hi morphology resembles that of ngc 4330 , except that at the opposite side of the hi tail ( the se ) the hi extends as far as the stellar disk . at that side
the hi is compressed and h@xmath29 emission from star formation is enhanced . along the tail
no stellar light or @xmath28 is seen down to the sdss and galex limits .
* ngc 4424 * : the hi is severely truncated within the optical disk ( @xmath30 ) and the tail extends from the south to sw .
the tail shows a weak velocity gradient ( fig .
[ fig - tailxv ] ) toward m49 as well as m87 .
the tail is clearly detected out to @xmath31 ( @xmath32 kpc ) , although there is a hint of faint gas even further to the sw in the direction of m49 , which itself is the center of a subcluster of galaxies ( plate 1 ) .
the stellar morphology and kinematics appear to be strongly disturbed by a gravitational interaction @xcite .
* ngc 4654 * : this is the only galaxy in the tail sample that also has a strong asymmetry in the stellar disk .
the hi shows a compressed edge on one side and a long tenuous tail on the other side .
sdss images show that starlight extends beyond the ridge of compressed hi in the west , implicating an icm - ism interaction as @xcite suggested .
the tail galaxies are located at intermediate distances from the cluster center at a projected distance of 0.6 - 1 mpc from m87 .
the length and the width of the tails vary , but the tails are all pointing roughly away from the cluster center ( m87 ) , suggesting that a single global mechanism such as the cluster potential or ram - pressure due to the icm might be responsible for these hi tails .
the cluster potential however , is not enough to affect galaxies significantly at those distances .
the tidal acceleration due to the cluster potential ( @xmath33 , @xmath34 where @xmath35 is the gravitional constant , @xmath36 is the cluster mass within a radius @xmath37 , and @xmath38 is the size of the galaxy ) is typically @xmath12 ( km s@xmath4)@xmath39 kpc@xmath4 while the gravitational acceleration of the galaxy ( @xmath40 where @xmath41 is the galaxy s rotational velocity ) is always larger than the cluster potential by 3 orders of magnitude , with a typical value of @xmath42 ( km s@xmath4)@xmath39 kpc@xmath4 . alternatively galaxies can be gas stripped by ram - pressure when @xmath43 , where @xmath44 is the icm density , @xmath45 is the galaxy s velocity @xmath46 the icm , and @xmath47 is the ism surface density ( * ? ? ?
* see also voller 2001 ) . in figure 2
we compare the estimated ram - pressure strength with the restoring force per unit area in the disk of each galaxy . rather than using the observed radial velocities , we use the simulations of @xcite to estimate the range of relative velocities a galaxy can have at a given projected distance from the cluster center ( 1250 to 1900 km s@xmath4 ) .
the error bar in the ram - pressure indicates this range .
the main uncertainty in the restoring force is the surface density of the gas at the time of stripping .
we considered two extremes , the mean hi surface density in the tail , and the peak value measured in the tail .
figure 2 shows that ram - pressure could exceed the restoring force for 5 out of 7 galaxies . as seen in plate 1 ,
ngc 4330 , ngc 4424 and ngc 4302 ( and its neigbour ngc 4298 ) clearly show that their hi disk is truncated to well within the stellar disk indicating that these three galaxies must have been affected by a gas - gas interaction .
the systemic velocity of ngc 4330 and ngc 4302 is close to the cluster mean , indicating that they move mostly in the plane of the sky .
this combined with the fact that their hi tails point away from the cluster center , suggests these galaxies are falling into the cluster along highly radial orbits .
ngc 4424 , whose systemic velocity is highly blueshifted with respect to the cluster mean , could be falling in from the back .
another possibility is that it has passed close to m49 and has been stripped by the icm of that subcluster .
its tail points directly toward the subcluster . in this scenario
the ngc 4424 tail would be more comparable to another tail in virgo recently found by @xcite , which appears to be gas stripped from ngc 4388 by the hot icm gas in the subgroup of m86 .
similarly a passage close to m49 could have caused the gravitational disturbance of ngc 4424 , as m86 might have gravitationally disturbed ngc 4388 . at the other extreme , the two galaxies for which our simple estimate of ram - pressure appears too low to have an effect
, still show signs of gas - gas interaction . both galaxies ,
ngc 4396 and ngc 4654 , have hi compressed on one side , while the hi tail at the opposite side does not seem to have an optical counterpart . in deeper optical images ,
ngc 4654 reveals more stellar light beyond the hi disk to the nw ( 3.2 ) .
in fact @xcite concludes that the hi morphology and kinematics of ngc 4654 can best be explained by the combined effect of ram - pressure stripping and a tidal disturbance by its companion , ngc 4639 .
on the other hand , ngc 4396 looks optically undisturbed .
several explanations are possible , either the galaxy is affected by viscous turbulent stripping instead of the simple momentum transfer , or the icm pressure is enhanced due to bulk motions or local clumping of the gas , as it is likely in ngc 4522 ( kenney et al .
2004 ) .
additionally , the kinematics of the tails also support an icm - ism interaction for most cases . for all galaxies except ngc 4294 and possibly ngc 4330
, the velocity gradient in the hi tails is toward the cluster mean ( fig . [ fig - tailxv ] ) as it is expected if the hi gas has been stripped by the cluster gas . we have seen for ngc 4654 that the tidal interaction with a nearby companion could have made the galaxy more vulnerable to ram - pressure stripping .
we may well ask whether tidal interactions with neighbors could help explain the existence of the other tails .
@xcite shows that if galaxies fall in as groups , the effect of the cluster potential gets greatly enhanced . in table 1
we list the nearest companion for each of the galaxies .
almost all have neighbors that are close in projection ( @xmath48100 kpc ) and in velocity ( @xmath48100 km s@xmath4 ) . in a dense cluster environment being close in projection
does not necessarily imply physical association , on the other hand the cumulative effect of many distant encounters can also loosen the outer parts ( harassment ) .
it is thus not unlikely that gravity plays some role , yet morphologically all the systems show some evidence for gas - gas interactions as well .
the only exception is the ngc 4299/4294 pair , although a parallel orientation of two tidal tails is rarely seen in interacting systems , and the tails have no stellar counterparts down to the sdss limit
. in analogy with the @xcite result tidal interactions may make it more easy to ram - pressure strip galaxies by bringing material to the outer disk @xcite . to conclude , while previous hi imaging showed that highly hi deficient galaxies in the center of virgo have very small hi disks ( e.g. * ? ? ?
* ) , we are now providing hi images of galaxies that are beyond the strongly hi deficient zone ( @xmath49mpc ; * ? ? ?
* see also fig . 3 in solanes et al .
our images show how galaxies just outside the virial radius may begin to lose their gas . in this region ( 0.6@xmath501 mpc from m87 ) ,
seven of the 16 galaxies imaged by us ( out of 27 spiral galaxies in the vcc catalog with [email protected] , binggeli , sandage & tammann 1985 ) show hi tails .
thus at least 26% of the large spiral galaxies in this region of virgo seem to be recent arrivals being stripped of gas .
interestingly , since it only affects the loosely bound gas in the outer parts of the galaxies , the simple momentum transfer ram - pressure picture still works for most of these galaxies , though viscous turbulent stripping , which can remove the gas more efficiently from edge - on encounters @xmath46 the icm @xcite , could affect some galaxies ( e.g. ngc 4302 ) .
it seems then that galaxy pre - processing occurs through mild gas - gas and/or tidal interactions before the galaxies enter the dense part of the cluster .
once they are within a virial radius gas gets removed within a gigayear through momentum transfer ram - pressure stripping of the more higly radially infalling galaxies and possibly viscous turbulent stripping of galaxies on more circular orbits .
bekki , k. 1999 , , 510 , l15 binggeli , b. , sandage , a. , & tammann , g. a. 1985 , , 90 , 1681 binggeli , b. 1999 , in the radio galaxy messier 87 , ed .
r@xmath52ser & k. meisenheimer ( berlin : springer ) , 9 b@xmath52hringer , h. , briel , u. g. , schwarz , r. a. , voges , w. , hartner , g. , & tr@xmath53mper , j. 1994 , , 368 , 828 briggs , d. s. 1995 , phd thesis , new mexico institute of mining and technology , socorro , nm cayatte , v. , van gorkom , j. h. , balkowski , c. , & kotanyi , c. 1990 , , 100 , 604 cort@xmath54s , j. r. , kenney , j. d. p. , & hardy , e. 2006 , , 131 , 747 dressler , a. 1980 , , 236 , 351 g@xmath55mez , p. l. , et al .
2003 , , 584 , 210 gunn , j. e. , & gott , j. r. 1972 , , 176 , 1 haynes , m. p. , & giovanelli , r. 1984 , , 89 , 758 kenney , j. d. p. , koopmann , r. a. , rubin , v. c. , & young , j. s. 1996 , , 111 , 152 kenney , j. d. p. , van gorkom , j. h. , vollmer , b. 2004 , , 127 , 3361 koopmann , r. a. , & kenney , j. d. p. 2004
, , 613 , 866 larson , r. b. , tinsley , b. m. , & caldwell , c. n. 1980 , , 237 , 692 lewis , i. , et al .
2002 , , 334 , 673 mihos , j. c. 2001 , asp conf .
ser . 230 : galaxy disks and disk galaxies , 230 , 491 mihos , j. c. 2004 , clusters of galaxies : probes of cosmological structure and galaxy evolution , 277 moore , b. , katz , n. , lake , g. , dressler , a. , & oemler , a. 1996 , , 379 , 613 nulsen , p. e. j. 1982 , , 198 , 1007 oosterloo , t. , & van gorkom , j. 2005 , , 437 , l19 phookun , b. , & mundy , l. g. 1995 , , 453 , 154 postman , m. & geller , m. j. 1984 , , 281 , 95 schindler , s. , binggeli , b. , & b@xmath52hringer , h. 1999 , , 343 , 420 solanes , j. m. , manrique , a. , garc@xmath56a - g@xmath55mez , c. , gonz@xmath57lez - casado , g. , 2001 , , 548 , 97 tully , r. b. , & shaya , e. j. 1984 , , 281 , 31 vollmer , b. , cayatte , v. , balkowski , c. , & duschl , w.j .
2001 , , 561 , 708 vollmer , b. 2003 , , 525 , 539 vollmer , b. , & huchtmeier 2006 , , in press ( astro - ph/0611162 ) yasuda , n. , fukugita , m. , & okamura , s. 1997 , , 108 , 417 | in a new hi imaging survey of virgo galaxies ( viva : vla imaging of virgo galaxies in atomic gas ) , we find 7 spiral galaxies with long hi tails .
the morphology varies but all the tails are extended well beyond the optical radii on one side .
these galaxies are found in intermediate - low density regions ( @xmath0 mpc in projection from m87 ) .
the tails are all pointing roughly away from m87 , suggesting that these tails may have been created by a global cluster mechanism .
while the tidal effects of the cluster potential are too small , a rough estimate suggests that simple ram - pressure stripping indeed could have formed the tails in all but two cases .
at least three systems show hi truncation to within the stellar disk , providing evidence for a gas - gas interaction .
although most of these galaxies do not appear disturbed optically , some have close neighbors , suggesting that tidal interactions may have moved gas outwards making it more susceptible to the icm ram - pressure or viscosity .
indeed , a simulation study of one of the tail galaxies , ngc 4654 , suggests that the galaxy is most likely affected by the combined effect of a gravitational interaction and ram - pressure stripping .
we conclude that these one - sided hi tail galaxies have recently arrived in the cluster , falling in on highly radial orbits .
it appears that galaxies begin to lose their gas already at intermediate distances from the cluster center through ram - pressure or turbulent viscous stripping and tidal interactions with neighbours , or a combination of both . |
prominences / filaments are fascinating coronal magnetic structures , whose dynamics and properties are not well - understood yet .
the long life of the so - called quiescent prominences ( several weeks ) suggests that the cool and dense prominence material is maintained against gravity and thermally shielded from the much hotter and much rarer solar corona by means of some not well - known processes .
however , it is believed that the magnetic field must play a crucial role in both the support and isolation of prominences .
high - resolution observations of solar filaments reveal that they are formed by a myriad of horizontal structures called threads ( e.g. , * ? ? ?
* ) , which have been observed in the spines and barbs of both active region and quiescent filaments @xcite .
the width of these fine - structures is typically in the range 0.2 arcsec
0.6 arcsec , which is close to the resolution of present - day telescopes , whereas their lengths are between 5 arcsec and 20 arcsec @xcite .
threads are assumed to be the basic substructures of filaments and to be aligned along magnetic field lines . from the point of view of theoretical modeling ,
prominence threads are interpreted as large coronal magnetic flux tubes , with the denser and cooler ( prominence ) region located at magnetic field dips that correspond to the observed threads . although some theoretical works have attempted to model such structures ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , there are some concerns about their formation and stability that have not been resolved yet .
small amplitude oscillations , propagating waves and mass flows are some phenomena usually observed in prominences and prominence threads ( see some recent reviews by * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
periods of small - amplitude prominence oscillations cover a wide range from less than a minute to several hours , and they are usually attenuated in a few periods @xcite . focusing on prominence threads ,
some works have detected oscillations and waves in such fine - structures ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
in particular , @xcite and @xcite suggested the presence of groups of near threads that moved in phase , which may be a signature of collective oscillations . on the other hand , mass flows along magnetic field lines
have been also detected @xcite , with typical flow velocities of less than 30 km s@xmath1 in quiescent prominences , although larger values have been detected in active region prominences @xcite .
regarding the presence of flows , a phenomenon which deserves special attention is the existence of the so - called counter - streaming flows , i.e. , opposite flows within adjacent threads @xcite . motivated by the observational evidence
, some authors have broached the theoretical investigation of prominence thread oscillations by means of the magnetohydrodynamic ( mhd ) theory in the @xmath2 approximation .
first , some works @xcite focused on the study of the ideal mhd oscillatory modes supported by individual nonuniform threads in cartesian geometry .
later , @xcite considered a more representative cylindrical thread and obtained more realistic results with respect to the spatial structure of perturbations and the behavior of trapped modes .
subsequently , the attention of authors turned to the study of collective oscillations of groups of threads , and the cartesian geometry was adopted again for simplicity .
hence , @xcite investigated the collective fast modes of systems of nonidentical threads and found that the only nonleaky mode corresponds to that in which all threads oscillate in spatial phase .
later , @xcite considered the limit of a periodic array of threads and obtained a similar conclusion .
therefore , these results seem to indicate that all threads within the prominence should oscillate coherently , even if they have different physical properties . however , one must bear in mind that the cartesian geometry provides quite an unrealistic confinement of perturbations , and so systems of more realistic cylindrical threads might not show such a clear collective behavior .
the next obvious step is therefore the investigation of oscillatory modes of systems of cylindrical threads .
the first approach to a similar problem was done by @xcite , who considered a system of two identical , homogeneous cylinders embedded in an unlimited corona .
although @xcite applied their results to coronal loops , they are also applicable to prominence threads .
these authors numerically found that the system supports four trapped kink - like collective modes .
these results have been analytically re - obtained by @xcite , by considering the thin tube approximation and bicylindrical coordinates .
subsequently , @xcite made use of the @xmath0-matrix theory of acoustic scattering to study the collective oscillations of arbitrary systems of non - identical cylinders .
although the scattering theory has been previously applied in the solar context ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , the first application to the study of normal modes of magnetic coronal structures has been performed by @xcite .
they concluded that , contrary to the cartesian case of @xcite , the collective behavior of the oscillations diminishes when cylinders with nonidentical densities are considered , the oscillatory modes behaving in practice like individual modes if cylinders with mildly different densities are assumed .
the present study is based on @xcite and applies their technique to the investigation of mhd waves in systems of cylindrical prominence threads .
moreover , we extend their model by considering some effects neglected by them . here , the more general @xmath3 case is considered , allowing us to describe both slow and fast magnetoacoustic modes .
in addition , the adiabatic assumption is removed and , following previous papers @xcite , the effect of radiative losses , thermal conduction and plasma heating is taken into account .
the detection of mass flows in prominences has motivated us to include this effect in our study , and so the presence of flows along magnetic field lines is also considered here .
therefore , the present work extends our recent investigation ( * ? ? ?
* hereafter paper i ) , which was focused on individual thread oscillations , to the study of collective mhd modes in prominence multi - thread configurations with mass flows . on the other hand , the longitudinal structure of threads ( e.g. , * ? ? ?
* ) is neglected in the present investigation .
for this reason , the effect of including a longitudinal variation of the plasma physical conditions within threads should be investigated in a future work .
finally , the prominence multi - thread model developed here could be an useful tool for future seismological applications ( similar to that of * ? ? ?
this paper is organized as follows .
the description of the model configuration and the mathematical method are given in
[ sec : math ] .
then , the results are presented in [ sec : results ] .
first , the case of two identical prominence threads is investigated in [ sec:2treads ] .
later , this study is extended to a configuration of two different threads in
[ sec : ntreads ] .
finally , our conclusion is given in
[ sec : conclusions ] .
our equilibrium system is made of an arbitrary configuration of @xmath4 homogeneous and unlimited parallel cylinders , representing prominence threads , embedded in an also homogeneous and unbounded coronal medium .
each thread has its own radius , @xmath5 , temperature , @xmath6 , and density , @xmath7 , where the subscript @xmath8 refers to a particular thread .
on the other hand , the coronal temperature and density are @xmath9 and @xmath10 , respectively .
cylinders are orientated along the @xmath11-direction , the @xmath12-plane being perpendicular to their axis .
the magnetic field is uniform and also orientated along the @xmath11-direction , @xmath13 being the magnetic field in the @xmath8-th thread , and @xmath14 in the coronal medium .
in addition , steady mass flows are assumed along magnetic field lines , with flow velocities and directions that can be different within threads and in the corona .
thus , @xmath15 represents the mass flow in the @xmath8-th thread , whereas @xmath16 corresponds to the coronal flow . for simplicity , in all the following expressions a
subscript 0 indicates local equilibrium values , while subscripts @xmath8 or @xmath17 denote quantities explicitly computed in the @xmath8-th thread or in the corona , respectively . such as shown in paper i , linear
nonadiabatic magnetoacoustic perturbations are governed by the next equation for the divergence of the velocity perturbation , @xmath18 , @xmath19 \delta + \tilde{\lambda}_0 ^ 2 { v_{\mathrm{a 0}}}^2 \frac{{\partial}^2}{{\partial}z^2 } \nabla^2 \delta = 0 , \label{eq : basic}\ ] ] where @xmath20 is the following operator , @xmath21 while @xmath22 is the alfvn speed squared and @xmath23 is the nonadiabatic sound speed squared , @xmath24 , \label{eq : lambda}\ ] ] @xmath25 and @xmath26 being the adiabatic sound speed squared and the adiabatic ratio , respectively .
terms with @xmath27 , @xmath28 , and @xmath29 are related to nonadiabatic mechanisms , i.e. radiative losses , thermal conduction , and heating ( see paper i for details ) . finally , @xmath30 is the doppler - shifted frequency @xcite , @xmath31 where @xmath32 is the oscillatory frequency and @xmath33 is the longitudinal wavenumber .
considering cylindrical coordinates , namely @xmath34 , @xmath35 , and @xmath11 for the radial , azimuthal , and longitudinal coordinates , respectively , we can write @xmath36 in the following form , @xmath37 where the function @xmath38 contains the full radial and azimuthal dependence . by inserting this last expression into equation ( [ eq : basic ] ) ,
the following helmholtz equation is obtained , @xmath39 where @xmath40 is the laplacian operator for the @xmath34 and @xmath35 coordinates , and @xmath41 @xmath42 are the radial wave number and the nonadiabatic tube speed squared , respectively .
moreover , due to the presence of nonideal terms @xmath43 is a complex quantity . since nonadiabatic mechanisms produce a small correction to the adiabatic wave modes , @xmath44 and the dominant wave character depends on the sign of @xmath45 . here
, we investigate nonleaky modes , which are given by @xmath46 .
we impose no restriction on the wave character within threads . in order to solve equation ( [ eq : hem ] )
, we consider the technique developed by @xcite based on the study of wave modes of an arbitrary configuration of cylinders by means of the @xmath0-matrix theory of acoustic scattering .
the novelty with respect to the work of @xcite is that the method is applied here to solve a helmholtz equation for the divergence of the velocity perturbation ( our eq .
[ [ eq : basic ] ] ) whereas @xcite considered an equation for the total pressure perturbation in the @xmath2 approximation ( their eq .
the present approach allows us to study the more general @xmath3 case , therefore slow modes are also described .
in addition , nonadiabatic effects and mass flows are easily included in our formalism .
however , the rest of the technique is absolutely equivalent to that of @xcite , and therefore the reader is refered to their work for an extensive explanation of the mathematical technique ( see also an equivalent formalism in * ? ? ?
we next give a brief summary of the method . the main difference between our application and that of @xcite is in the definition of the @xmath0-matrix elements .
these elements are obtained by imposing appropriate boundary conditions at the edge of threads , i.e. , at @xmath47 , where @xmath48 is the radial vector corresponding to the position of the @xmath8-th thread center with respect to the origin of coordinates . in our case , these boundary conditions are the continuity of the total pressure perturbation , @xmath49 , and the lagrangian radial displacement , @xmath50 .
expressions for these quantities as functions of @xmath36 and its derivative are , @xmath51 @xmath52 expressions for the rest of perturbations are given in appendix a of paper i. thus , in our case the @xmath0-matrix elements have the following form , @xmath53 where @xmath54 and @xmath55 are the hankel function of the first kind and the bessel function of order @xmath56 , respectively , while the prime denotes the derivative taken with respect to @xmath34 .
note that the denominator of @xmath57 vanishes at the normal mode frequencies of an individual thread .
this can be easily checked by comparing it to the dispersion relation of a single thread , see equation ( 19 ) of paper i , in which bessel @xmath58 functions are used instead of hankel functions .
the equivalence between both kinds of functions is given in @xcite .
thus , following @xcite , the internal @xmath59 field of the @xmath8-th thread is , @xmath60 whereas the external net field is , @xmath61 where @xmath62 is the azimuthal angle corresponding to the position of the @xmath8-th thread center with respect to the origin of coordinates , and @xmath63 and @xmath64 are constants
. equations ( [ eq : int ] ) and ( [ eq : ext ] ) allow us to construct the spatial distribution of @xmath36 .
subsequently , the rest of perturbations can be obtained . finally , the constants @xmath63 form a homogeneous system of linear algebraic equations , @xmath65 for @xmath66 .
once both integers @xmath56 and @xmath67 are truncated to a finite number of terms , the non - trivial ( i.e. , non - zero ) solution of system ( [ eq : system ] ) gives us a dispersion relation for the oscillatory frequency , @xmath32 , which is enclosed in the definitions of @xmath68 and @xmath69 . in the next sections ,
we apply the method to obtain the oscillatory frequency and the spatial distribution of perturbations of the wave modes supported by prominence thread configurations .
we assume that @xmath33 is real , so a complex frequency is obtained , namely @xmath70 .
the imaginary part of the frequency appears due to the presence of nonadiabatic mechanisms . the oscillatory period , @xmath71 , and the damping time , @xmath72 , are related to the frequency as follows , @xmath73
now , we consider a configuration of two identical threads ( see fig . [
fig : model ] ) .
their physical conditions are typical of prominences ( @xmath74 k , @xmath75 kg m@xmath76 ) while the coronal temperature and density are @xmath77 k and @xmath78 kg m@xmath76 , respectively .
their radii are @xmath79 km , and the distance between centers is @xmath80 km .
the magnetic field strength is 5 g everywhere .
the flow velocity inside the cylinders is denoted by @xmath81 and @xmath82 , respectively , whereas the flow velocity in the coronal medium is @xmath83 .
unless otherwise stated , these physical conditions are used in all calculations .
first , we consider no flow in the equilibrium , i.e. @xmath84 .
we fix the longitudinal wavenumber to @xmath85 , which corresponds to a wavelength within the typically observed range .
in addition to the four kink modes described by @xcite , i.e. the @xmath86 , @xmath87 , @xmath88 , and @xmath89 modes , where @xmath90 or @xmath91 denote symmetry or antisymmetry of the total pressure perturbation with respect to the @xmath92-plane , and the subscripts refer to the main direction of polarization of motions , we also find two more fundamental collective wave modes ( one symmetric and one antisymmetric ) mainly polarized along the @xmath11-direction , which we call @xmath93 and @xmath94 modes following the notation of @xcite .
these new solutions correspond to slow modes which are absent in the investigation of @xcite due to their @xmath2 approximation . as was stated by @xcite ,
a collective wave mode is the result of a coupling between individual modes . so the reader must be aware that in the present work we indistinctly use both expressions , i.e. , collective modes and coupled modes , to refer to wave solutions whose perturbations have significant amplitudes in both threads .
the total pressure perturbation field , @xmath49 , and the transverse lagrangian displacement vector - field , @xmath95 , corresponding to the six fundamental modes are displayed in figure [ fig : eigen ] .
on the other hand , figure [ fig : cut ] displays a cut of the cartesian components of the lagrangian displacement ( @xmath96 , @xmath97 , and @xmath98 ) at @xmath99 , again for these six solutions . for simplicity ,
only the real part of these quantities are plotted in both figures , since their imaginary parts are equivalent .
one can see in figure [ fig : cut ] that the amplitude of the longitudinal ( magnetic field aligned ) lagrangian displacement , @xmath98 , of the @xmath93 and @xmath94 modes is much larger than the amplitude of transverse displacements , @xmath96 and @xmath97 , such as corresponds to slow modes in @xmath100 homogeneous media , while the contrary occurs for the @xmath86 , @xmath87 , @xmath88 , and @xmath89 fast kink solutions .
next , figure [ fig : distkink]a displays the ratio of the real part of the frequency of the four kink solutions to the frequency of the individual kink mode , @xmath101 ( from paper i ) , as a function of the distance between the center of cylinders , @xmath102 .
this figure is equivalent to figure 3 of @xcite and , in agreement with them , one can see that the smaller the distance between centers , the larger the interaction between threads and so the larger the separation between frequencies . on the other hand ,
figure [ fig : distkink]b shows the ratio of the damping time to the period of the four kink modes as a function of @xmath102 .
we see that the damping times are between 4 and 7 orders of magnitude larger than their corresponding periods
. therefore , dissipation by non - adiabatic mechanisms can not be responsible for the observed damping times of transverse thread oscillations , as was pointed out in paper i. recently , @xcite found that the mechanism of resonant absorption can provide kink mode damping times compatible with those observed .
regarding slow modes , figure [ fig : distslow]a displays the ratio of the real part of the frequency of the @xmath93 and @xmath94 solutions to the frequency of the individual slow mode , @xmath103 ( from paper i ) .
one can see that the frequencies of the @xmath93 and @xmath94 modes are almost identical to the individual slow mode frequency , and so the strength of the interaction is almost independent of the distance between cylinders .
this is consistent with the fact that transverse motions ( responsible for the interaction between threads ) are not significant for slow - like modes in comparison with their longitudinal motions .
therefore , the @xmath93 and @xmath94 modes essentially behave as individual slow modes , contrary to kink modes , which display a more significant collective behavior .
finally , figure [ fig : distslow]b shows @xmath104 corresponding to the @xmath93 and @xmath94 solutions versus @xmath102 .
one sees that both slow modes are efficiently attenuated by non - adiabatic mechanisms , with @xmath105 , which is in agreement with previous studies @xcite and consistent with observations .
the aim of the present section is to assess the effect of flows on the behavior of collective modes . with no loss of generality , we assume no flow in the corona , i.e. @xmath106 . on the other hand , the flow velocities in both cylinders , namely @xmath107 and @xmath108 , are free parameters .
we vary these flow velocities between -30 km s@xmath1 and 30 km s@xmath1 , which correspond to the range of typically observed flow velocities in filament threads ( e.g. , * ? ? ?
these flow velocities are below the critical value that determines the apparition of the kelvin - helmholtz instability ( see details in * ? ? ?
* ) . in our configuration
, a positive flow velocity means that the mass is flowing towards the positive @xmath11-direction , whereas the contrary is for negative flow velocities . from paper
i ( see also * ? ? ?
* ) we know that the symmetry between waves whose propagation is parallel ( @xmath109 ) or anti - parallel ( @xmath110 ) with respect to magnetic field lines is broken by the presence of flows . hence , we must take into account the direction of wave propagation in order to perform a correct description of the wave behavior .
following paper i , we call parallel waves those solutions with @xmath109 , while anti - parallel waves are solutions with @xmath110 .
we begin this investigation with transverse modes .
first , we assume @xmath111 20 km s@xmath1 and study the behavior of the oscillatory frequency when @xmath108 varies ( see fig . [
fig : phase ] ) . since frequencies are almost degenerate and , therefore , almost indiscernible
if they are plotted together , we use the notation of @xcite and call low - frequency modes the @xmath86 and @xmath89 solutions , while high - frequency modes refer to @xmath87 and @xmath88 solutions .
in addition , we restrict ourselves to parallel propagation because the argumentation can be easily extended to anti - parallel waves . to understand the asymptotic behavior of frequencies in figure [ fig : phase ] ,
we define the following doppler - shifted individual kink frequencies : @xmath112 @xmath113 since @xmath107 is fixed , @xmath114 is a horizontal line in figure [ fig : phase ] , whereas @xmath115 is linear with @xmath108 .
three interesting situations have been pointed by means of small letters from @xmath116 to @xmath17 in figure [ fig : phase ] .
each of these letters also corresponds to a panel of figure [ fig : eigenampl ] in which the total pressure perturbation field of the @xmath86 mode is plotted .
the three different situations are commented in detail next ( remember that in all cases @xmath117 km s@xmath1 ) : * @xmath118 @xmath119 km s@xmath1 ( @xmath120 ) .
this corresponds to a situation of counter - streaming flows . from figure
[ fig : phase ] we see that the frequency of low - frequency modes is close to @xmath115 , whereas that of high - frequency solutions is near @xmath114 .
thus , these solutions do not interact with each other and low - frequency ( high - frequency ) solutions are related to individual oscillations of the second ( first ) thread .
this is verified by looking at the total pressure perturbation field in figure [ fig : eigenampl]a , that shows that only the second thread is significantly perturbed .
therefore , for an external observer this situation corresponds in practice to an individual thread oscillation .
* @xmath121 @xmath122 km s@xmath1 ( @xmath123 ) .
the flow velocities and their directions are equal in both threads . in such a situation , low- and high - frequency modes couple . at the coupling ,
an avoided crossing of the solid and dashed lines is seen in figure [ fig : phase ] .
because of this coupling solutions are related no more to oscillations of an individual thread but they are now collective and , for this reason , figure [ fig : eigenampl]b shows a significant pressure perturbation in both threads .
* @xmath124 @xmath125 km s@xmath1 ( @xmath126 ) .
this case is the opposite to situation @xmath118 .
therefore , in practice the present situation corresponds again to an individual thread oscillation .
our argumentation is supported by figure [ fig : ampli ] , which displays the amplitude of the transverse lagrangian displacement , @xmath127 , at the center of the second thread as a function of @xmath108 , for parallel and anti - parallel kinklike waves .
the displacement amplitude at the center of the first thread is always set equal to unity .
the three previously commented situations have been pointed again in figure [ fig : ampli ] .
we clearly see that the displacement amplitude is only comparable in both threads , and so their dynamics is collective , when their flow velocities are similar .
next we turn our attention to slow modes .
the behavior of the @xmath93 and @xmath94 modes is like that of low- and high - frequency kinklike solutions , so we comment them in short for the sake of simplicity . @xmath93 and @xmath94 solutions
can only be considered collective when the flow velocity is the same in both threads because , in such a case , the @xmath93 and @xmath94 modes couple .
if different flows within the threads are considered , the @xmath93 and @xmath94 slow modes lose their collective aspect and their frequencies are very close to the doppler - shifted individual slow frequencies , @xmath128 @xmath129 then , the @xmath93 and @xmath94 solutions behave like individual slow modes .
it is worth to mention that the coupling between slow modes is much more sensible to the flow velocities in comparison with fast modes , and the @xmath93 and @xmath94 solutions quickly decouple if @xmath107 and @xmath108 slightly differ .
an example of this behavior is seen in figure [ fig : eigenamplslow ] , which displays the total pressure perturbation field of the @xmath94 mode for @xmath130 km s@xmath1 .
although the difference of the flow velocities is insignificant , one can see that in this situation the @xmath94 mode essentially behaves as the individual slow mode of the second thread .
the main idea behind these results is that fast or slow wave modes with a collective appearance ( i.e. , modes with a similar displacement amplitude within all threads ) are only possible when the doppler - shifted individual kink ( eqs .
[ [ eq : wkleft ] ] and [ [ eq : wkright ] ] ) or slow ( eqs . [ [ eq : wsleft ] ] and [ [ eq : wsright ] ] ) frequencies are similar in both threads . in a system of identical threads ,
this can only be achieved by considering the same flow velocities within all threads , since all of them have the same individual kink and slow frequencies . however ,
if threads with different physical properties are considered ( i.e. , different individual frequencies ) , it is possible that the coupling may occur for different flow velocities .
this is explored in the next section .
now we consider a system of two nonidentical threads and focus first on kink modes . from
[ sec:2treads ] we expect that collective kink motions occur when the doppler - shifted individual kink frequencies of both threads coincide .
the relation between flow velocities @xmath107 and @xmath108 for which the coupling takes place can be easily estimated from the equation , @xmath131 note that the individual kink frequency of each thread is different , i.e. , @xmath132 .
then , the relation between flow velocities at the coupling is , @xmath133 this last expression can be simplified by considering the approximate expression for the kink frequency in the long - wavelength limit , @xmath134 for @xmath135 , where the @xmath136 sign is for parallel waves and the @xmath137 sign is for anti - parallel propagation .
then , one finally obtains , @xmath138 where the meaning of the @xmath136 and @xmath137 signs is the same as before . in the case of identical threads , @xmath139 and so @xmath140 .
thus the flow velocity must be the same in both threads to obtain collective motions , as we concluded in [ sec:2treads ] .
an equivalent analysis can be performed for slow modes and one obtains , @xmath141 in general , the coupling between slow modes occur at different flow velocities than the coupling between kink modes .
this makes difficult the simultaneous existence of collective slow and kink solutions in systems of non - identical threads .
next , we assume a particular configuration of two non - identical threads in order to verify the later argumentation .
thread radii are @xmath142 km and @xmath143 km , whereas their physical properties are @xmath144 k , @xmath145 kg m@xmath76 , and @xmath146 k , @xmath147 kg m@xmath76 .
coronal conditions are those taken in [ sec:2treads ] ( i.e. , @xmath77 k , @xmath78 kg m@xmath76 ) .
the magnetic field strength is 5 g everywhere and the distance between the thread centers is @xmath148 km .
we assume @xmath149 km s@xmath1 . in the present case ,
also four kinklike solutions are present , which are grouped in two almost degenerate couples . for this reason
, we again refer to them as low- and high - frequency kink solutions .
their frequency as a function of @xmath108 is displayed in figure [ fig : phasediff ] . at first sight
, we see that solutions couple for a particular value of @xmath108 , as expected .
applying equation ( [ eq : velrelation ] ) , and considering that @xmath150 km s@xmath1 and @xmath151 km s@xmath1 , we obtain @xmath152 , and since @xmath153 km s@xmath1 , we get @xmath154 km s@xmath1 for parallel propagation .
we see that the approximate value of @xmath108 obtained from equation ( [ eq : velrelation ] ) is in good agreement with figure [ fig : phasediff ] .
in addition , we obtain that for parallel propagation collective dynamics appear in a situation of counter - streaming ( opposite ) flows .
this result is of special relevance because counter - streaming flows have been detected in prominences @xcite and might play a crucial role in the collective behavior of oscillations . on the contrary , in the anti - parallel propagation case we obtain @xmath155 km s@xmath1 from equation ( [ eq : velrelation ] ) , and
so both flows are in the same direction and large a quite value of @xmath108 is obtained in comparison with the parallel propagation case . regarding slow modes , considering that @xmath156 km s@xmath1 and @xmath157 km s@xmath1 , equation ( [ eq : velrelationslow ] ) gives @xmath158 km s@xmath1 for parallel propagation and @xmath159 km s@xmath1 for anti - parallel waves .
note that in our particular example the flow velocities needed for the coupling situation are realistic and within the range of typically observed velocities .
however , if threads with very different physical properties and , therefore , with very different alfvn and sound speeds are considered , the coupling flow velocities could be larger than the observed values .
this means that the conditions necessary for collective oscillations of systems of threads with very different temperatures and/or densities may not be realistic in the context of solar prominences .
in this work , we have assessed the effect of mass flows on the collective behavior of slow and fast kink magnetosonic wave modes in systems of prominence threads .
we have seen that the relation between the individual alfvn ( sound ) speed of threads is the relevant parameter which determines whether the behavior of kink ( slow ) modes is collective or individual . in the absence of flows and
when the alfvn speeds of threads are similar , kink modes are of collective type . on the contrary ,
perturbations are confined within an individual thread if the alfvn speeds differ . in the case of slow modes ,
the conclusion is equivalent but replacing the alfvn speeds by the sound speeds of threads . on the other hand ,
when flows are present in the equilibrium , one can find again collective motions even in systems of non - identical threads by considering appropriate flow velocities .
these velocities are within the observed values if threads with not too different temperatures and densities are assumed . however , since the flow velocities required for collective oscillations must take very particular values , such a special situation may rarely occur in real prominences .
therefore , if coherent oscillations of groups of threads are observed in prominences ( e.g. , * ? ? ?
* ) , we conclude that either the physical properties and flow velocities of all oscillating threads are quite similar or , if they have different properties , the flow velocities within threads are the appropriate ones to allow collective motions . from our point of view
, the first option is the most probable one since the flow velocities required in the second case correspond to a very peculiar situation .
this conclusion has important repercussions for future prominence seismological applications , in the sense that if collective oscillations are observed in large areas of a prominence , threads in such regions should possess very similar temperatures , densities , and magnetic field strengths here , we have only considered two - thread systems , but the method can be applied to an arbitrary multi - thread configuration .
so , the model developed here could be used to perform seismological studies of large ensembles of prominence threads if future observations provide with positions and physical parameters of such systems .
ballester , j. l. , & priest , e. r. 1989 , , 225 , 213 ballester , j. l. 2006 , phil .
a , 364 , 405 banerjee , d. , erdlyi , r. , oliver r. , & oshea , e. 2007 , , 246 , 3 bogdan , t. j. , & zweibel , e. g. 1987 , , 312 , 444 daz , a j. , oliver r. , erdlyi , r. , & ballester , j. l. 2001 , , 379 , 1083 daz , a j. , oliver r. , & ballester , j. l. 2002 , , 580 , 550 daz , a j. , oliver r. , & ballester , j. l. 2003 , , 402 , 781 daz , a j. , oliver r. , & ballester , j. l. 2005 , , 440 , 1167 daz , a. j. , & roberts , b. 2006 , , 236 , 111 lin , y. , engvold , o. , & wiik , j. e. 2003 , , 216 , 109 lin , y. 2004 , p.h .
d. thesis , university of oslo lin , y. , engvold , o. , rouppe van der voort , l. h. m. , wiik , j. e. , & berger , t. e. 2005 , , 226 , 239 lin , y. , engvold , o. , rouppe van der voort , l. h. m. , & van noort , m. 2007 , , 246 , 65 lin , y. , martin , s. f. , & engvold , o. 2008 , in asp conf .
383 , subsurface and atmospheric influences on solar activity , ed .
r. howe , r. w. komm , k. s. balasubramaniam , & g. j. d. petrie ( san francisco : asp ) , 235 luna , m. , terradas , j. , oliver , r. , & ballester , j. l. 2008b , , in preparation molowny - horas , r. , wiehr , e. , balthasar , h. , oliver , r. , & ballester , j. l. 1999 , joso annual report 1998 , astronomical institute tatranska lomnica , 126 okamoto , t. j , et al .
2007 , science , 318 , 1557 oliver , r. , & ballester , j. l. 2002 , , 206 , 45 rempel , m. , schmitt , d. , & glatzel , w. 1999 , , 343 , 615 schmitt , d. , & degenhardt , u. 1995 , rev .
astron . , 8 , 61 soler , r. , oliver , r. , & ballester , j. l. 2007 , , 471 , 1023 soler , r. , oliver , r. , & ballester , j. l. 2008 , , 684 , 725 ( paper i ) terra - homem , m. , erdlyi , r. , & ballai , i. 2003 , , 217 , 199 van doorsselaere , t. , ruderman , m. s. , and robertson , d. 2008 , , 485 , 849 yi , z. , engvold , o. , & kiel , s. l. 1991 , , 132 , 63 yi , z. , & engvold , o. 1991 , , 134 , 275 zirker , j. b. , engvold , o. , & yi , z. 1994 , , 150 , 81 | high resolution observations of solar filaments suggest the presence of groups of prominence threads , i.e. the fine - structures of prominences , which oscillate coherently ( in phase ) .
in addition , mass flows along threads have been often observed . here
, we investigate the effect of mass flows on the collective fast and slow nonadiabatic magnetoacoustic wave modes supported by systems of prominence threads .
prominence fine - structures are modeled as parallel , homogeneous and infinite cylinders embedded in a coronal environment .
the magnetic field is uniform and parallel to the axis of threads .
configurations of identical and nonidentical threads are both explored .
we apply the @xmath0-matrix theory of acoustic scattering to obtain the oscillatory frequency and the eigenfunctions of linear magnetosonic disturbances .
we find that the existence of wave modes with a collective dynamics , i.e. those that produce significant perturbations in all threads , is only possible when the doppler - shifted individual frequencies of threads are very similar .
this can be only achieved for very particular values of the plasma physical conditions and flow velocities within threads . |
in the recent report on the detailed analysis of solar neutrino oscillations @xcite done by the super - kamiokande collaboration @xcite , it has been suggested that solutions with large mixing angles are favored while solutions with small mixing angles are disfavored at the 95@xmath18 confidence level .
neutrino oscillations controlled by such large mixing angle have also been observed and confirmed for atmospheric neutrinos @xcite .
both observed oscillations are characterized by the same property of neutrino mixings , namely , large neutrino mixings , which are theoretically consistent with bimaximal mixing scheme @xcite .
the difference arises in their oscillation scales denoted by @xmath19 for atmospheric neutrinos and @xmath20 for solar neutrinos , which are specified by @xmath21 ev@xmath22 and @xmath23 ev@xmath22 , thereby , indicating a hierarchy of @xmath24 .
neutrino oscillations arise for massive neutrinos @xcite .
since neutrino masses of @xmath25 ev are implied by @xmath19 , models for neutrino oscillations must be equipped with mechanisms generating such tiny masses of neutrinos @xcite , which include breaking of the lepton number conservation for majorana neutrino masses .
it has been long known @xcite that such lepton number breaking interactions are inevitably contained in the minimal @xmath26 model for electroweak interactions ( called the 331 model ) @xcite because a charged lepton ( @xmath27 ) and its antiparticle ( @xmath28 ) together with a neutrino ( @xmath29 ) are placed in a triplet of @xmath30 as @xmath31=(@xmath32 , @xmath33 , @xmath34)@xmath35 . as a result , the model provides a charged - lepton mass term included in @xmath36 , which simultaneously contains @xmath37 as a neutrino mass term .
namely , the need for majorana mass term is linked to the existence of the massive charged leptons .
therefore , the 331 model is suitable for incorporating majorana neutrino masses @xcite . with this nice feature of the 331 model in mind ,
we introduce the bimaximal mixing scheme into the model to accommodate observed neutrino oscillations .
one of the possible scenario for bimaximal mixing scheme is to employ a new @xmath38 symmetry based on @xmath1 ( @xmath39 @xmath10 ) @xcite .
neutrino mixings are predicted to be maximal for solar neutrinos with @xmath20=0 but arbitrary for atmospheric neutrinos with @xmath40 .
the hierarchy of @xmath24 is trivially satisfied since @xmath20=0 .
one has to next explain how a nonvanishing @xmath20 is induced while preserving @xmath24 .
an elaborated mechanism is to generate @xmath20 as radiative effects @xcite , whose generic smallness explains the relative smallness of @xmath20 over @xmath19 @xcite . in this article
, we construct our interactions that induce phenomenologically consistent neutrino and charged - lepton mass matrix , @xmath41 and @xmath42 , with appropriate radiative effects for @xmath20 .
since the model describes both matrices in a unified form , masses for charged leptons and neutrinos are correlated to each other .
the mass term of @xmath36 transforms as @xmath43 and @xmath44 , which requires @xmath45 and/or @xmath46 as higgs scalars .
the coupling of the triplet higgs scalar to @xmath36 is antisymmetrized with respect to the @xmath30-index .
the spin and statistics , then , requires antisymmetrized flavor indices carried by @xmath31 , in turn , generating antisymmetrized @xmath42 with eigenvalues corresponding to @xmath47 , which contradicts with the observed hierarchy of @xmath48 .
the triplet higgs scalar alone is not consistent with the observed lepton mass spectrum . therefore , the sextet scalar , which provides symmetrized mass terms , is the key ingredient of the 331 model @xcite .- singlet vector - like heavy @xmath49electron " without introducing a sextet scalar @xcite has been advocated to describe consistent mass spectra of neutrinos and charged leptons @xcite . ] because the sextet higgs scalar contains both an @xmath50-doublet scalar for charged leptons and an @xmath50-triplet scalar for neutrinos , we naturally expect @xmath49degeneracy `` in the structure of their mass matrices .
the charged leptons should contain a diagonal mass term such as the @xmath15-mass , which forces @xmath41 to have a diagonal term .
however , @xmath41 should not contain diagonal terms to be consistent with bimaximal mixing scheme based on @xmath10 . to avoid this @xmath49degeneracy '' , we use the conservation of the @xmath10 quantum number supplemented by an appropriate @xmath3 parity as well as a specific breaking pattern of @xmath10 yielding a residual discrete symmetry .
the collaboration of these symmetries allows neutrinos to develop no diagonal mass terms . as a result
, atmospheric neutrino oscillations are controlled by those tree - level off - diagonal masses while solar neutrino oscillations are to be induced by one - loop effects . and
a discrete symmetry have been done in ref.@xcite . ] in the following section , we first introduce particles and their interactions and discuss how the required vacuum alignment is realized . section iii deals with the diagonalization of @xmath41 , which reflects rotation effects due to the diagonalization of @xmath42 .
the bimaximal condition on the neutrino masses determines a pattern of the matrix elements for the electron mass .
one - loop radiative mass of @xmath51 is also calculated .
the discussions in sec.iv include to examine @xmath20 by utilizing the correlation between @xmath41 and @xmath42 and to confirm that the partial @xmath49degeneracy " between @xmath41 and @xmath42 suppresses the rotation effects to yield the lma solution based on the radiative effects .
the final section is devoted to summary and discussions .
the model contains neutrinos and charged leptons in triplets of @xmath52 ( @xmath53=1,2,3 ) and the greek letters of @xmath54 , respectively , stand for the three families and three @xmath30-indices . ] with @xmath55 for @xmath56 and @xmath57 for @xmath58 , where the superscript @xmath59 of @xmath60 stands for the charge conjugation , and two ( anti-)sextet higgs scalars @xmath61 with @xmath10=(0 , 2 ) for ( @xmath9 , @xmath11 ) .
since @xmath9 can couple to @xmath62 ( @xmath8=2,3 ) , @xmath63 mixing terms are generated by @xmath64 and are to be responsible for the bimaximal structure .
on the other hand , @xmath65 ( and also @xmath66 ) are generated by @xmath67 via its coupling to @xmath68 ( @xmath69=2,3 ) .
since @xmath68 also supply couplings to @xmath70 and @xmath71 , if @xmath72 is also induced , then it disturbs the nice feature of the bimaximal mixing scheme based on @xmath10 .
therefore , in order for the present approach to be acceptable , the simultaneous alignment of @xmath73 and @xmath67 should be dynamically allowed .
it is demonstrated that this vacuum alignment of @xmath74 and @xmath75 is indeed possible to occur as a result of the emergence of a residual discrete symmetry specific to @xmath38 .
the 331 model is specified by the @xmath76-charge .
let @xmath77 be the @xmath76 quantum number , then the hypercharge , @xmath78 , is given by @xmath79 and the electric charge @xmath80 is given by @xmath81 , where @xmath82 is the @xmath83 generator with tr@xmath84 @xmath85 .
let us first summarize the whole particle content in our model , where the quantum numbers are specified by ( @xmath30 , @xmath76 ; @xmath38 ) : @xmath86 for leptons , @xmath87 for quarks @xmath88 for triplet higgs scalars , and @xmath89 for ( anti-)sextet higgs scalars .
these higgs scalars have the following vacuum expectation values ( vev s ) : @xmath90 and quarks and leptons will acquire masses via these vev s , where the orthogonal choice of the vev s in eq.([eq : vev_triplet ] ) and the vanishing vev of @xmath91 in eq.([eq : vev_sextet ] ) are to be ensured by appropriate higgs interactions .
the pure @xmath30-anomaly is cancelled in a vectorial manner .
the anomalies from triplets of the three families of leptons and of the three colors of the third family of quarks are cancelled by those from antitriplets of three colors of the first and second families of quarks @xcite .
other anomalies including @xmath76 are also cancelled .
the scale of the 331 model is set by @xmath92 , yielding @xmath93 and we expect the magnitude of @xmath92 is of order tev . to realize phenomenologically viable interactions
, we further impose a @xmath3 parity on the model , where @xmath3=+ for @xmath56 , @xmath94 and @xmath11;=@xmath95 for @xmath58 and @xmath9;=@xmath8 for @xmath96 and @xmath97 and similarly for quarks .
hereafter , we restrict ourselves to the lepton sector .
the yukawa interactions for leptons are , then , controlled by the following lagrangian : @xmath98 where @xmath99 s denote the yukawa couplings with @xmath100 .
the possible term of the @xmath101-type is forbidden by @xmath3 .
the higgs interactions are described by self - hermitian terms composed of @xmath102 ( @xmath103=@xmath94 , @xmath96 , @xmath97 ) and @xmath104 , which include the potential term of @xmath105 : @xmath106 where @xmath107 s represent coupling constants and @xmath108 and by the non - self - hermitian terms in @xmath109 where @xmath110 denotes a mass scale and @xmath111 stand for coupling constants .
there is a leptonic number associated with @xmath31 , which can be taken to be @xmath112 for @xmath113 and 0 for all others . since the @xmath114 conservation is broken both spontaneously by @xmath113 and explicitly by @xmath115
, there is no harmful nambu - goldstone boson .
although the lepton number can not be globally assigned to @xmath30-multipulets , to argue majorana masses for neutrinos , it is useful to define the lepton number @xmath116 to classify interactions @xcite in the similar way used for the electric charge defined in the standard @xmath117 model . following the classification lately done in ref.@xcite
, the @xmath116-number can be introduced in the form of @xmath118 where @xmath119 = 2 for @xmath120 ; = 4/3 for @xmath97 ; = 2/3 for @xmath121 ; = 1/3 for @xmath122 ; = 0 for @xmath123 and @xmath124 ; = @xmath952/3 for @xmath125 , @xmath94 , @xmath96 and @xmath113 ; = @xmath952 for @xmath126 , which yields @xmath116 = ( 1 , 1 , @xmath127 ) for @xmath122 as usual . the assignment for leptons and higgs scalars is shown in table [ tab : lnumber0 ] . among our interactions constrained by @xmath10 ,
only @xmath128 breaks the @xmath116 conservation ; therefore , it is this interaction that gives the nonvanishing @xmath129 .
a few comments are in order to support our choice of vev s of @xmath130 and @xmath131 : 1 .
the nonvanishing vev of @xmath132 is triggered by the @xmath116-violating @xmath128 for @xmath133 since @xmath134 , @xmath135 and @xmath136 ; 2 .
the nonvanishing vev of @xmath137 is triggered by the @xmath116-conserving @xmath138 for @xmath139 since @xmath135 and @xmath136 and by the @xmath116-conserving @xmath140 for @xmath141 since @xmath134 ; 3 . the orthogonal choice of vev s of @xmath94 , @xmath96 and @xmath97 as in eq.([eq : vev_triplet ] )
is supported by @xmath105 if all @xmath107 s are negative .
it is because @xmath105 gets lowered if @xmath94 , @xmath96 and @xmath97 develop vev s .
so , one can choose vev s such that @xmath142 @xmath143 0 , @xmath144 @xmath143 0 and @xmath145 @xmath143 0 .
thus , these vev s are dynamically aligned by our higgs potential . on the other hand
, we can not argue whether @xmath146 acquire vev s or not since there are no triggers such as an @xmath116-violating @xmath147 and an @xmath116-conserving @xmath148 , which can , respectively , align @xmath91 and @xmath149 to be nonvanishing .
however , so far these interactions are forbidden by the @xmath10 conservation . to invoke such dynamical alignment , thus ,
requires @xmath10-breaking interactions . to introduce @xmath10-breaking effects calls for careful consideration because of the specific alignment of vev s of @xmath11 , _
i.e. _ @xmath73 and @xmath67 .
its breaking readily induces dangerous interactions that creates @xmath91 such as tadpole interactions for @xmath74 . owing to the vacuum alignment of @xmath73
, @xmath10-breaking interactions should conserve @xmath116 .
there is one candidate that can keep this specific alignment , which is an @xmath116-conserving @xmath150 : @xmath151 where @xmath152 represents a breaking of the @xmath10 conservation . since
@xmath38 is broken by the unit of @xmath153=4 , there is still a residual conservation due to a @xmath3 symmetry of @xmath154 to be denoted by @xmath155 , which is shown in table [ tab : lnumber0 ] together with @xmath10 and @xmath3 .
this @xmath155 symmetry can be used to constrain possible forms of @xmath10-breaking interactions .
the mass terms for @xmath14 and @xmath15 are created by @xmath67 , which is now supported by @xmath156 for @xmath157 .
therefore , @xmath156 serves as a trigger for @xmath67 .
the vev of @xmath91 would be created by the following @xmath116-violating interactions of * tr(@xmath158 ) since @xmath64 , * @xmath147 since @xmath159 , * @xmath160 since @xmath134 , @xmath135 and @xmath136 , * @xmath161 and @xmath162 since @xmath135 , @xmath136 and @xmath163 and * tr(@xmath164 ) , tr(@xmath165 ) and tr(@xmath166 ) since @xmath163 .
fortunately , all of these interactions are forbidden by the interplay between @xmath3 and @xmath155 .
tabulated in table [ tab : lnumber ] are @xmath3 and @xmath155 for non - self - hermitian higgs interactions allowed by @xmath0 , where the reader can recognize how these dangerous interactions are excluded . at this moment ,
we reach the plausible situation , where the alignment of @xmath167 is dynamically supported .
the resulting mass matrices for charged leptons and neutrinos take the form of @xmath168 where @xmath169 are expressed in terms of vev s : @xmath170 leading to the proportionality of @xmath171 and @xmath172 stand for one - loop radiative masses to be estimated in the next subsection .
we simply assume @xmath173 ( @xmath69=2,3 ) so as to reproduce the hierarchy of @xmath174 by the mechanism discussed in ref.@xcite .
since the magnitude of neutrino masses @xmath175 is controlled by @xmath129 , let us explain how naturally small @xmath129 is obtained .
an estimate of @xmath176 can be found in the analyses of ref.@xcite applied to our higgs potential .
it calls for @xmath133 as has been expected in the vacuum alignment .
the smallness of @xmath129 is , then , ascribed to either @xmath177 @xcite or @xmath178 @xcite , the latter of which is known as the type ii seesaw mechanism @xcite , where we use the mechanism due to @xmath177 by simply expecting that @xmath92 is of order tev .
one will find that @xmath179a few@xmath180 provides neutrino masses about @xmath181 ev .
this is in fact a fine - tuning of @xmath182 to reproduce neutrino masses so as to meet the value of @xmath181 .
however , we have already known this tiny magnitude , which is give by @xmath183 .
the relative smallness of @xmath184 is linked to the underlying chiral symmetry recovered at the limit of @xmath184 @xmath185 0 .
this feature dictates the so - called @xmath49naturalness " @xcite .
similarly , since the @xmath116 conservation is respected in the limit of @xmath186 , the process of taking @xmath177 is also regarded to be natural although both cases require other new physics to explain why such tiny values are realized .
our program to discuss patterns of neutrino oscillations is based on the use of the proportionality of @xmath175 to @xmath187 indicated by eq.([eq : nupropto ] ) .
since to generate the electron mass and to reproduce @xmath19 will , respectively , determine @xmath187 and @xmath175 , which yield the estimation of @xmath188 .
the magnitude of @xmath129 derived in eq.([eq : vev_nu ] ) now determines that of @xmath182 once appropriate values of vev s are given and , then , @xmath189 ( @xmath69=2,3 ) to be defined in eqs.([eq : neutrino_ee ] ) and ( [ eq : neutrino_mutau ] ) are calculable for this estimated @xmath182 . in the end ,
solar neutrino oscillations controlled by @xmath189 are found to be consistent with the lma solution .
we first discuss the diagonalization of @xmath42 .
the following unitary matrix @xmath190 is obtained under the approximation of @xmath173 ( @xmath69=2,3 ) and transforms @xmath42 into @xmath191=@xmath192=diag.(@xmath193 , @xmath66 , @xmath65 ) : @xmath194 up to the second order of @xmath195 with @xmath196 where @xmath69=@xmath197 for @xmath198 , @xmath199 and @xmath200 .
the masses of @xmath201 , @xmath14 and @xmath15 are given by @xmath202 where @xmath203 to be consistent with the observed mass pattern of the charged lepton mass matrix , we should present a mechanism generating plausible mass textures such as the hierarchical pattern @xcite and the democratic pattern @xcite .
however , in the present context of the model , we are not aimming at clarifying all the physical aspect of the lepton sector but , instead , at clarifying the correlation of neutrino mass matrix with charged lepton mass matrix . here
, we simply parameterize @xmath204 by adopting two distinct ways to realize @xmath48 : hierarchical and democratic mass textures and discuss their influences in neutrino mixings .
the resulting parameterization is given by 1 .
the hierarchical pattern of @xmath205 and @xmath206 leading to @xmath207 with @xmath208 ( @xmath209 ) and 2 . the democratic pattern of @xmath210 and @xmath211 where @xmath212 and @xmath213 specify @xmath214 and @xmath215 ( @xmath216 ) , leading to @xmath217 .
therefore , we have found the estimate of @xmath218 that reproduces @xmath184 .
neutrinos acquire radiative masses of @xmath189 ( @xmath69=2,3 ) generated by one - loop interactions corresponding to diagrams depicted in fig.[fig : loopdiagrams_ee ] and fig.[fig : loopdiagrams_mutau ] .
it is convenient to shift the basis for charged - leptons from the original one for @xmath42 to the intermediate one , where @xmath14 and @xmath15 have diagonal masses of @xmath219 and @xmath220 , respectively .
the corrections due to @xmath221 are neglected because of @xmath222 . in this basis , the couplings of @xmath9 to the lepton triplets are described by @xmath223 ( @xmath8=2,3 ) with ( @xmath224 , @xmath225)=(@xmath226 , @xmath227 ) and @xmath228 are the rotated @xmath229-couplings similarly defined as ( @xmath230 , @xmath231)=@xmath232 , @xmath233 ) , which appear in the figures . the result of the calculation of @xmath189 ( @xmath69=2,3 ) is given by @xmath234v_\eta \nonumber \\ & & + 2\lambda_2\lambda_1 v^{(0)}_\ell f^2_{1i } m_{\ell^i } \left [ i(m^2_{\ell^i } , m^2_{s^{(0)+ } } , m^2_{s^{(0)- } } , m^2_{\eta^+ } ) + i(m^2_{\ell^i } , m^2_{s^{(0)- } } , m^2_{s^{(0)+ } } , m^2_{\eta^- } ) \right ] v_\eta v_\rho v_\chi , \label{eq : neutrino_ee } \\
\delta m^{rad}_{ij}&= & 2\lambda_2\lambda_b v^{(0)}_\ell f_{1i}f_{1j } m_{\ell^i } \left [ i(m^2_{\ell^i } , m^2_{s^{(+)+ } } , m^2_{s^{(0)- } } , m^2_{\eta^+ } ) + i(m^2_{\ell^i } , m^2_{s^{(+)- } } , m^2_{s^{(0)+ } } , m^2_{\eta^- } ) \right ] v_\eta v_\rho v_\chi , \label{eq : neutrino_mutau}\end{aligned}\ ] ] for @xmath69=2,3 and @xmath184=0 in this basis , with @xmath235 and @xmath236 as in eq.([eq : unitaryentries ] ) , where @xmath237 the majorana mass @xmath238 for @xmath51 is induced by effective couplings corresponding to fig.[fig : loopdiagrams_ee ] : @xmath239 for @xmath103=@xmath96 or @xmath97 , and @xmath240 with similar terms scrambled by possible permutations of the fields .
both couplings conserve @xmath10 . on the other hand , the same effective coupling as eq.([eq : inducedmajorana_ee2 ] ) but with the replacement of @xmath241 , thus of the @xmath10-breaking type , yields @xmath172 ( @xmath69=2,3 ) for @xmath242 , which correspond to fig.[fig : loopdiagrams_mutau ] .
now , let us turn to examining the neutrino sector .
the neutrino mass matrix is also transformed into @xmath243 by @xmath190 of eq.([eq : unitarymatrix ] ) to maintain diagonal weak currents and is calculated , up to the second order of @xmath244 , to be : @xmath245 where @xmath246 with @xmath247 it is obvious to see that @xmath243 satisfies tr(@xmath243)=tr(@xmath41)=@xmath248 and tr(@xmath249)= tr(@xmath250)=2(@xmath251 . since @xmath252 ,
the diagonal neutrino masses given by ( @xmath253 , @xmath254 , @xmath255 ) are calculated to be : @xmath256 up to the second order of @xmath257 equivalent to the second order of @xmath244 , where @xmath258 with @xmath259 and @xmath260 as the mixing angle for atmospheric neutrinos and @xmath261 the masses of @xmath262 and @xmath263 can be further converted into : @xmath264 , \label{eq : eachmassterms3}\end{aligned}\ ] ] where other irrelevant terms involving @xmath172 are omitted .
the mass of @xmath265 is obtained by noticing the relation of tr(@xmath243)=@xmath248 that yields @xmath266 from which we find that @xmath267 to enhance the bimaximal structure in @xmath243 is achieved by setting @xmath268 , where @xmath269 can be approximated to be : @xmath270 because @xmath271 .
the condition of @xmath268 thus , requires , for the hierarchical case with @xmath272 and @xmath273 , @xmath274 and , for the democratic case with @xmath275 , @xmath276 the essence of the present research lies in the fact that the bimaximal mixing scheme for neutrinos , in turn , determines how the electron mass is derived .
the bimaximal condition of eq.([eq : bimaximalcond1 ] ) combined with eq.([eq : masses23_1 ] ) yields @xmath277 while eq.([eq : bimaximalcond2 ] ) combined with eq.([eq : masses23_2 ] ) yields @xmath278 and @xmath279 where @xmath280 .
the neutrino oscillation parameters of @xmath19 and @xmath281 are expressed in terms of our neutrino masses as @xmath282 leading to @xmath283 with @xmath284 the ratio @xmath285 arises from the radiative effects and @xmath286 receives the first order contributions from the rotation due to the diagonalization of @xmath287 .
we are now in a position to estimating @xmath288 to meet the lma solution . before going to its explicit evaluation , we list our assumptions on numerical values of the relevant mass parameters : * @xmath289 (= @xmath290=174 gev ) for @xmath291 , * @xmath292 as the simplest case to keep @xmath293 for weak boson masses , * @xmath294 and @xmath295 , where @xmath92 , presumably of order tev , is left unspecified and is to be measured in the unit of 10@xmath296 .
let us first reproduce @xmath21 ev@xmath22 , which requires that @xmath297 ev@xmath22 from eq.([eq : atm_m2 ] ) .
the relation of @xmath298 ( @xmath8=2,3 ) together with eqs.([eq : bimaximalchargedcond1]-[eq : bimaximalchargedcond3 ] ) for @xmath187 , then , leads to @xmath299 where three terms to be placed in the parenthesis , respectively , denote the cases with ( hierarchical masses with @xmath300 , democratic masses with @xmath301 , democratic masses with @xmath302 ) . in the following , the same notation is used . since @xmath129 is expressed in terms of @xmath182 and vev s as estimated in eq.([eq : vev_nu ] ) , we find the constraint on @xmath182 given by @xmath303 therefore , the magnitude of @xmath182 is not arbitrary but fixed to be @xmath304 by vev s in order to obtain @xmath21 ev@xmath22 .
next , we calculate @xmath20 , which depends on the sum of @xmath238 and @xmath305 given by eqs.([eq : neutrino_ee ] ) and ( [ eq : neutrino_mutau ] ) with @xmath306 . to find numerical values of @xmath189 , we use the following property of the integral @xmath307 of eq.([eq : integral ] ) : @xmath308 by using the assumptions on mass parameters stated above , we know that the second term in eq.([eq : neutrino_ee ] ) is proportional to @xmath309 and all other terms including @xmath172 of eq.([eq : neutrino_mutau ] ) are proportional to @xmath310 . in eq.([eq : neutrino_ee ] ) , the first term thus dominates the second term because of @xmath311 for @xmath312 and @xmath313 .
the other terms proportional to @xmath310 , the third and fourth terms together with @xmath305 , are more suppressed by the factor of @xmath314 compared with the first term since @xmath315 are at most of order unity .
therefore , the first term in @xmath238 gives the most dominant contributions to @xmath20 and those from all other terms can be safely neglected .
the radiative mass of @xmath238 saturated by the mediating @xmath15 lepton contribution turns out to be : @xmath316 where , in the democratic cases , the factor of @xmath317 is due to @xmath318 in eq.([eq : bimaximalchargedcond2 ] ) and the factor 1/2 arises from @xmath319 .
the rotated couplings of @xmath228 used in these calculations are approximated to be ( @xmath229 , @xmath320 , @xmath321 ) ( @xmath322 ) for @xmath323=@xmath127 for @xmath8=2 and @xmath323=1 for @xmath8=3 . by inserting the above result of eq.([eq : numerics_rad1 ] ) into @xmath285 shown in eq.([eq : ratiosolarrad ] ) with @xmath324 ev , we reach the following estimate : @xmath325 where @xmath326 is used in eq.([eq : numerics_rad1 ] ) .
this estimate together with eq.([eq : lambda2_estimate ] ) for @xmath182 yields @xmath327 where @xmath328 is bounded from below by @xmath280 . by taking @xmath329{\cos\phi } , 2.0 - 3.6 \right),\end{aligned}\ ] ] we obtain @xmath281=(@xmath330 ) ev@xmath22 relevant to the lma solution to the solar neutrino problem and @xmath331 since @xmath296=174 gev , @xmath92 roughly lies in the range of @xmath332 tev , which is the right order of @xmath92 (= @xmath333(1 tev ) ) as anticipated for the scale of @xmath0 @xmath185 @xmath117 . for @xmath334 tev as a typical value
, namely , @xmath335=2.3 , we find that @xmath336 leading to @xmath337 ev@xmath22 for @xmath338 ev@xmath22 .
finally , we consider effects from @xmath286 , which may disturb the behavior specific to the lma solution indicated by @xmath285 .
since @xmath286 is controlled by @xmath265 as in eq.([eq : ratiosolarrad ] ) , it can be found that the leading terms of @xmath339 contained in @xmath265 turns out to vanish .
the cancellation can be seen from the insertion of the leading terms of @xmath340 into the equation of @xmath341 appearing in the numerator of @xmath265 in eq.([eq : thirdmass ] ) that turns out to vanish .
the next nonvanishing contributions to @xmath265 are given by the second order terms in @xmath341 that result in the third order contributions to @xmath265 , which is beyond our approximation because possible third order terms for @xmath257 ( @xmath69=2,3 ) , which have not been included , directly give the third order contributions as can be seen from the definition of @xmath265 in eq.([eq : neutrinomassparam ] ) .
therefore , we expect that @xmath286 becomes at most @xmath342 , which is about @xmath343 ) as shown in eqs.([eq : bimaximalchargedcond1])-([eq : bimaximalchargedcond3 ] ) , leading to @xmath344 .
this estimate of @xmath286 should be compared with eq.([delta_sol_lma2 ] ) for the radiative case and in fact gives more suppressed contributions .
it is , then , concluded that we can safely neglect @xmath286 and that the model can provide the lma solution to the solar neutrino problem as indicated by @xmath285 .
similarly , the third neutrino @xmath345 will also acquire masses at most of order @xmath346 , which is about @xmath347 ev . the radiative correction to @xmath255 arising from @xmath348
is estimated to be about @xmath349 since @xmath350 and @xmath351 , we find that @xmath352 ev for @xmath353 tev .
thus , altogether we observe that the bound on @xmath255 is given by @xmath354 ev .
the cancellation of the diagonalization effects on the neutrino masses of @xmath355 is a reasonable consequence because the ( 1,2 ) and ( 1,3 ) entries of @xmath41 are exactly proportional to those of @xmath42 ( see eqs.([eq : leptontextureh])-([eq : leptontextureb2 ] ) ) so that the rotation effects due to @xmath244 , which induce @xmath356 , @xmath357 , @xmath358 , are partially cancelled .
however , the diagonalization effects manifest themselves in the matrix element @xmath359 that connects @xmath51 to @xmath345 .
the chooz and paloverde experiments imply that @xmath360 @xcite . for the mass matrix @xmath243 of eq.([eq : numassesweak ] ) , this element is calculated to be @xmath361 \nonumber \\ & = & -\frac{1}{m_\nu}\left [ \left ( c_\alpha \delta m^\nu_{13 } + s_\alpha \delta m^\nu_{12 } \right)\delta u_\mu - \left ( c_\alpha \delta m^\nu_{12 } - s_\alpha \delta m^\nu_{13 } \right)\delta u_\tau \right ] , \label{eq : ue3}\end{aligned}\ ] ] leading to @xmath362 with @xmath363 for eq.([eq : bimaximalchargedcond2 ] ) allowing @xmath364 , which includes the typical value of @xmath365 .
our model predicts that @xmath366 , which are consistent with the current experimental bound .
we have successfully demonstrated that the minimal 331 model with the global @xmath1 and @xmath3 symmetry , which involves two ( anti-)sextet higgs scalars , @xmath113 , indeed explains solar neutrino oscillations based on the lma solution . the lma solution is explicitly obtained by the use of the plausible parameter values for masses and couplings .
the model has the following properties , which do not depend on the parameterization , that 1 .
the global @xmath1 symmetry is explicitly broken by @xmath150 down to a discrete @xmath155 symmetry defined by @xmath154 that forbids the @xmath116-breaking interactions for @xmath11 , the absence of which is the essence of our discussions to be consistent with the bimaximal neutrino mixing scheme based on @xmath1 ; 2 . the @xmath116 conservation is explicitly broken by @xmath367 , which generates the vev of @xmath129 for neutrinos calculated to be @xmath368 , where @xmath133 is the trigger for @xmath369 ; 3 .
the mechanism generating the diagonal electron mass via @xmath187 ( @xmath8=2,3 ) is linked to the requirement realizing the bimaximal neutrino mixing via @xmath175 owing to the proportionality of @xmath370 ; 1 .
the smallness of @xmath371 ev for @xmath19 is realized by @xmath372 estimated to be @xmath373 ; 2 .
solar neutrino oscillations are induced by * the one - loop radiative effects , which yield @xmath374 corresponding to the lma solution provided @xmath0 @xmath185 @xmath117 occurs at @xmath332 tev , * the effects of the simultaneous diagonalization of the neutrino and charged - lepton mass matrices , which are expected to yield @xmath375 and which certainly yield the @xmath359 matrix element estimated to be @xmath376 , + where , in the present model , the dominance of one - loop radiative effects yields the lma solution .
it should be stressed that the choice of @xmath92=@xmath333(1 tev ) automatically leads to the lma solution .
the present radiative mechanism generates the majorana mass of @xmath51 , @xmath238 , given by the effective coupling of @xmath377 .
this coupling is induced by @xmath128 and @xmath138 as in fig.[fig : loopdiagrams_ee ] and is of the @xmath10-conserving type .
the less - dominant radiative masses are induced by @xmath378 for @xmath379 with @xmath69=1 contributing to @xmath238 and for @xmath380 with @xmath69=2,3 contributing to @xmath172 .
the @xmath380-terms involve the @xmath10-breaking coupling of @xmath150 as in fig.[fig : loopdiagrams_mutau ] .
the splitting of @xmath381 and @xmath382 is provided by @xmath238 , which is about @xmath383 ev , leading to the lma solution .
the mass of @xmath345 is provided by @xmath348 at most of order @xmath384 ev but the diagonalization due to @xmath287 may give @xmath255 @xmath385 @xmath347 ev . 1 . for the hierarchical texture with @xmath386 , @xmath387 2 .
for the democratic texture with @xmath388 , @xmath389 with @xmath390 and @xmath391 constrained by @xmath280 , 3 . for the democratic texture with @xmath392 , @xmath393 with @xmath394 , where @xmath395=@xmath188 , @xmath396 and the asterisks stand for the suppressed entries to generate @xmath20
. the partial @xmath49degeneracy " between @xmath41 to @xmath42 suppresses the rotation effects on @xmath281 due to the diagonalization of @xmath41 and @xmath42 . as a result ,
the solar neutrino oscillations are , respectively , characterized by @xmath337 ev@xmath22 for @xmath353 tev .
of course , if our mechanism explaining neutrino oscillations works further , one should next consider some reasons for the emergence of the approximate equality of @xmath397 in the hierarchical case and of the hierarchy of either @xmath301 or @xmath302 in the democratic case .
the masses of higgs scalars are simply assumed to be about @xmath296 for those related to the weak boson masses for the order - of - magnitude estimate for radiative neutrino masses , namely , @xmath398
@xmath399 @xmath400 @xmath399 @xmath401 @xmath399 @xmath402 @xmath399 @xmath296 . on the other hand , the @xmath97 scalar that generates masses for the exotic quarks and gauge bosons
is assumed to have @xmath403 @xmath399 @xmath92(@xmath404 @xmath296 ) because @xmath405 is related to the spontaneous breakdown of @xmath0 to @xmath117 .
we have estimated @xmath92 to lie in @xmath332 tev in order to reproduce the observed neutrino mass pattern .
it is further known that the 331 model contains peculiar particles , _
i.e. _ doubly - charged bilepton gauge bosons @xcite with masses of the order @xmath406 , that couple to @xmath407 .
since @xmath92 can be @xmath332 tev , the bilepton gauge bosons can be as light as and even lighter than 1 tev .
it is , therefore , quite conceivable that we will observe the bilepton gauge bosons with masses around 1 tev , which provide striking phenomena with clean @xmath408-@xmath408 jets @xcite .
now equipped with the plausible mechanism that generates phenomenologically consistent masses for neutrinos and charged leptons , the 331 model may be a physically interested model of leptons .
it not only explains three families of quarks and leptons because of the specific anomaly - cancellation mechanism used together with the asymptotic free condition on @xmath409 but also involves majorana mass terms demanded by charged - lepton mass terms .
there remain dynamical questions about the smallness of the @xmath116-violating coupling of @xmath182 and the fine tuning of the lepton mass parameters to reproduce @xmath410 , which require some new physics beyond @xmath411 .
it is also required to have extensive analyses on physics of the quark and gauge boson sectors as well to fully clarify potential power of the 331 model .
d. v. ahluwalia , ; v. barger , p. pakvasa , t.j .
weiler and k. whisnant , ; a. baltz , a.s .
goldhaber and m. goldhaber , ; m. jezabek and y. sumino , ; r.n .
mohapatra and s. nussinov , ; y. nomura and t. yanagida , ; i. starcu and d.v.ahluwalia , ; c. jarlskog , m. matsuda , s. skadhauge and m. tanimoto , ; q. shafi and z. tavartkiladze , ; ; p.h .
frampton and s.l .
glashow , ; c.h . albright and s.m .
barr , ; h. georgi and s.l .
glashow , ; r.n .
mohapatra , a. prez - lorenzana and c. a. de s. pires , .
t. yanagida , in _ proceedings of the workshop on unified theories and baryon number in the universe _ edited by a. sawada and a. sugamoto ( kek report no.79 - 18 , tsukuba , 1979 ) , p.95 ; ; m. gell - mann , p. ramond and r. slansky , in _ supergravity _ edited by p. van nieuwenhuizen and d.z .
freedmann ( north - holland , amsterdam 1979 ) , p.315 ; r.n .
mohapatra and g. senjanovi , .
valle and m. singer , ; r. barbieri and r.n .
mohapatra , ; j. liu , ; r. foot , o. f. hernndez , f. pisano and v. pleitez , ; v. pleitez and m.d .
tonasse , ; ; ; j. liu and d. ng , ; p.h .
frampton , p.i .
krastev and j.t .
liu , ; f. pisano , v. pleitez and m.d .
tonasse , hep - ph/9310230 v2 ( feb , 1994 ) ; m. @xmath412zer , ; f. pisano , j.a .
silva - sobrinho and m.d .
tonasse , f. -z .
chen , ; m.b .
tully and g.c .
joshi , hep - ph/9810282.v2 ( jan,1999 ) ; y. okamoto and m. yasu@xmath413 , ; n.a .
long and d.v .
soa , .
t. kitabayashi and m. yasu , ; ; hep - ph/0102228 v3 ( feb . ,
2001 ) to be published in phys . lett .
b ( 2001 ) ; t. kitabayashi , hep - ph/0010341 ( oct . ,
2000 ) ; hep - ph/0103195 ( mar . , 2001 ) to appear in phys .
d ( 2001 ) .
tully and g.c .
joshi , hep - ph/0011172 to appear in phys .
rev d ( 2001 ) ; j.c .
montero , c.a .
de s. pires and v. pleitez , .
see for example , g.c .
w. grimus and l. lavoura , ; w. grimus and h. neufeld , ; l. bento and j.w.f .
valle , ; s.y .
tsai , ; k.s .
babu , r.n .
mohapatra and l. rothstein , ; j.t .
peltoniemi , a.yu . smirnov and j.w.f .
valle , in ref .
@xcite ; j.t .
peltoniemi , d. tommasini and j.w.f .
valle , ; j.t .
peltoniemi and j.w.f .
e. ma , ; ; ; ; ; for recent discussions , see for example , a.s .
joshipura and s.d .
rindani , ; r.n .
mohapatra , a. prez - lorenzana and c.a .
de s. pires , in ref.@xcite .
d. chang and a. zee , ; r. adhikari , e. ma and g. rajasekaran , . w. grimus and h. neufeld , ; w. grimus and l. lavoura , ; t. kitabayashi and m. yasu , in ref.@xcite ; t. kitabayashi , in ref.@xcite . for supersymmetric models ,
see for example , e. ma , m. raidal and u. sarkar , ; j.c .
romo , m.a .
daz , m. hirsch , w. porod and j.w.f .
valle , ; f. takayama and m. yamaguchi , ; o.c.w .
kong , ; ; e.j .
chun , ; s. davidson and m. losada , ; m. hirsch , m.a .
daz , w. porod , j.c .
romo and j.w.f .
valle , .
g. t hooft , in _ recent development in gauge theories _
, proceedings of the cargese summer institute , cargese , france , 1979 , edited by g. t hooft _ et al .
_ , nato advanced study institute series b : physics vol .
59 ( plenum press , new york , 1980 ) . for recent studies ,
see for example , p.h .
frampton and x .- hu guan , ; f. pisano , j.a .
silva - sobrinho and m.d .
tonasse , ; b. dion , t. gregoire , d. london , l. marleau and h. nadeau , ; m.b .
tully and g.c .
joshi , ; p.h .
frampton , ; p.h .
frampton and a. ra , .
.[tab : lnumber0 ] @xmath10 , @xmath119 , @xmath3 and @xmath155 for leptons and higgs scalars , where the lepton number @xmath116 is related to @xmath119 as @xmath116 = @xmath415 + @xmath119 . [ cols="<,^,^,^,^,^,^,^",options="header " , ] | in the minimal @xmath0 gauge model with a global @xmath1 ( @xmath2 ) symmetry and a discrete @xmath3 symmetry , it is found that the interplay between neutrinos and charged leptons contained in triplets of @xmath4=(@xmath5 , @xmath6 , @xmath7 ) ( @xmath8=1,2,3 ) naturally leads to the large mixing angle ( lma ) msw solution .
the model includes two ( anti)sextet higgs scalars , @xmath9 with @xmath10=0 and @xmath11 with @xmath10=2 , which , respectively , couple to @xmath12 for the electron mass and masses of atmospheric neutrinos and to @xmath13 for the @xmath14- and @xmath15-masses and one - loop radiative neutrino masses relevant to solar neutrinos .
this mechanism is realized by utilizing an additional residual discrete symmetry supplied by explicitly broken @xmath10 , which guarantees the absence of tree - level neutrino mass terms of the @xmath13-type .
pure rotation effects due to the diagonalization of neutrino and charged - lepton mass matrices are estimated to yield @xmath16 but the radiative effects supersede the rotation effects to yield @xmath17 as the lma solution .
epsf # 1#2#3#4#1 * # 2 * ( # 4 ) # 3 # = `` 0023 $ = ' ' 0024 % = `` 0025 = ' ' 705c = `` 336e = ' ' 336f = `` 331a = ' ' 331b = `` 3328 = ' ' 3329 = " 0350 |
the first indication that stellar composition plays a role in , and may be affected by , the formation of planets was the enhanced metallicity ( parameterized as [ fe / h]=log(n@xmath1/n@xmath2 ) - log ( n@xmath1/n@xmath2)@xmath3 ) of solar - type exoplanet host stars ( gonzalez 1997 ) . while initially suggested as a sign of accretion of hydrogen - depleted material onto the star , the planet - metallicity correlation is now established as a mostly primordial effect ( e.g. , santos et al.2004 ; fischer & valenti 2005 ) .
recently , the correlation was shown to be dependent on planet mass ( e.g. , sousa et al .
2008 ; ghezzi et al .
2010 ) , being the weakest for terrestrial - sized planets ( e.g. , buchhave et al .
if host star metallicity is considered a proxy for heavy element concentration in the protoplanetary disk , this planet mass dependence seems to support the core accretion process of planet formation , as a larger concentration of heavy elements is needed to form giant planet cores ( @xmath410 m@xmath5 ; ida & lin 2004 ) rapidly enough to allow time for gaseous envelope accretion ( @xmath6100 m@xmath5 ; pollack et al .
1996 ) prior to gas disk dissipation after a few million years ( e.g. , wyatt et al .
is it still possible that planet formation may alter host star composition ?
melndez et al . (
2009 ; m09 ) suggest the sun is deficient by @xmath420% in refractory elements , with condensation temperatures @xmath7 900 k , relative to volatile elements when compared to `` solar twins '' without detected planets .
m09 , as well as studies finding similar results ( e.g. , ramrez et al .
2010 ; gonzalez et al .
2010 ) , interpret the negative [ x / h ] or [ x / fe ] vs t@xmath8 slopes of refractory elements as a star `` missing '' rock - forming material that has been sequestered in terrestrial planets ( chambers 2010 ) .
however , this interpretation has been questioned .
other studies have found no increase of negative slopes in hotter or more metal - rich stars ( ramrez et al.2014 ) ; no difference in slopes between hosts / non - hosts after correcting for galactic chemical evolution ( schuler et al .
2011b ) , correlations between the @xmath9 slope and stellar age and log @xmath10 ( adibekyan et al .
2014 ) ; and even the opposite of expected trends positive / negative slopes in small / giant planet hosts ( gonzlez hernndez et al.2011 ) . to complicate the picture , stars hosting close - in giant planets may instead accrete refractory - rich material that would have formed small planets ( e.g. , raymond et al . 2011 ; schuler et al 2011b ) .
thus , if / how planet formation influences stellar composition remains uncertain , and is challenging to disentangle from the local or global composition of the disk and the star s position in / motion through the galaxy ( e.g. , nehag et al .
2014 ; schuler et al .
binary host stars provide an opportunity for constraining how planet formation affects host stars , as they likely experience similar environments throughout their evolution ( e.g. , kratter 2011 ) .
desidera et al .
( 2004 ; 2006 ) found that the majority of @xmath450 wide binaries ( not known to host planets ) had @xmath0[fe / h]@xmath110.03 , suggesting larger metallicity differences are exceptional .
gratton et al .
( 2001 ) examined six wide binaries across a larger range of elements , finding that four had identical abundances .
however , two stars were significantly different , with @xmath0[fe / h][email protected] and [email protected] and similar differences for other refractory elements ; the authors suggested these differences were due to pollution by dusty protoplanetary disk or rocky planetary material . radial velocity ( rv ) monitoring of both `` polluted '' systems to search for planets has been inconclusive due to changing stellar activity cycles . within binary systems
known to host planets , the results are also ambiguous : there is not yet consensus on the significance of differences between 16 cyg a and b ( the latter hosts a @xmath42.4 m@xmath13 planet ; schuler et al .
2011a ; ramrez et al.2011 ; tucci maia et al . 2014 ) , and similar studies indicate no abundance differences between stars in a single - host ( hat - p-1 , liu et al .
2014 ) and dual - host ( hd 20782/1 , mack et al .
2014 ) systems . recently ,
rv monitoring of the southern component ( xo-2s ; [email protected] m@xmath14 ; desidera et al .
2014 ) of the @xmath44600 au binary xo-2 revealed two giant ( 0.26 m@xmath13 , 1.37 m@xmath13 ) , slightly eccentric ( @xmath150.18 , 0.15 ) exoplanets at 0.13 and 0.48 au ( desidera et al .
xo-2n ( [email protected] m@xmath14 ; burke et al .
2007 ) was already known to host a transiting hot jupiter ( 0.62 m@xmath13 , @xmath16=0 , @xmath17= 0.04 au ) .
this system is one of only four known dual - planet - hosting binaries ( xo-2 , hd 20782/1 , kepler-132 , and wasp-94 ) .
here we expand the analysis of teske et al .
( 2013 ) to investigate whether there are abundance differences between the stars , how such differences depend on derived stellar parameters , and how they may relate to the formation of different types of planets .
the observations and data reduction for xo-2n and -2s are detailed in teske et al .
the @xmath1860,000 , s / n@xmath4170 - 230 data were acquired in february 2012 with the high dispersion spectrograph at the 8.2 m subaru telescope , and reduced using standard techniques with the iraf software package . the stellar parameter analysis and uncertainty calculations
are also explained in teske et al .
( 2013 ) and listed in table [ tab : stellar_params ] .
the abundances were determined using the local thermodynamic equilibrium ( lte ) spectral analysis code moog ( sneden 1973 ) , with model atmospheres interpolated from the kurucz atlas9 nover grids .
equivalent widths were measured with ` splot ' in iraf , and abundances were normalized to solar , derived from a reflected - moonlight spectrum taken with the same configuration during the same run , on a line - by - line basis . in xo-2n , 49 fe i and 8 fe ii lines were measured ; in xo-2s , 49 fe i and 10 fe ii lines were measured ; the line properties and equivalent widths are provided in table [ tab : lines ] . as shown in recent studies of gaia benchmark stars ( jofre et al .
2014 ) and transiting exoplanet host stars ( e.g. , torres et al . 2012 ) , different analysis methods lead to different stellar parameters and thus different abundances .
to investigate this effect thoroughly , particularly the different t@xmath19s , we explored several alternative methods , used widely in other exoplanet host star studies , for deriving stellar parameters for xo-2n and -2s .
* alternative parameter set 1 * first , we repeated our analysis relative to solar , using by - hand iraf measurements of the lines in ghezzi et al .
this list includes a smaller number of fe i and ii lines ( in our case 23 fe i for n / s , and 9/10 fe ii for n / s ) , but all have lab - measured and thus more reliable gf - values .
kurucz atlas9 odfnew grids were used as input models , though the difference between nover and odfnew are insignificant .
the resulting parameters are listed as `` alt param 1 '' in table [ tab : stellar_params ] .
* alternative parameter set 2 * second , we adopted the full analysis approach of ghezzi et al .
( 2010 ) and derived parameters based on absolute abundances , not relative to solar .
the ghezzi et al .
( 2010 ) line list , with lab - measured gf - values , was compiled specifically for an absolute analysis .
again , the measurements were done by - hand in iraf and the stellar models were odfnew , using the same ews as `` alt param 1 '' .
the resulting parameters , are listed as `` alt param 2 '' in table [ tab : stellar_params ] .
* alternative parameter set 3 * third , we used odfnew models with a different and longer line list of 72/67 fe i and 9/8 fe ii lines for n / s , compiled from sousa et al .
the equivalent widths were measured with ares ( sousa et al .
2007 ) ; lines @xmath200.30 dex from the average in the sun were removed and an interative 2@xmath21 clipping was applied to the stars during the parameter determination .
the resulting parameters are listed as `` alt param 3 '' in table [ tab : stellar_params ] . in almost every case , within 1@xmath21 total errors
all derived parameter pairs for xo-2n and -2s overlap , indicating the stars are similar , potentially identical , but could also differ by as much as @xmath4200 k in t@xmath19 , @xmath40.25 dex in log @xmath10 , @xmath40.10 in [ fe / h ] , and @xmath40.07 km / s in @xmath22 . in the first three parameter sets ,
t@xmath19 of xo-2n @xmath23 xo-2s , which agrees with the original xo-2nb discovery paper ( burke et al .
2007 ) , t@xmath19s calculated from the irfm photometric calibration of casagrande et al .
( 2010 ) using colors from burke et al .
( 2007 ) , and the recent analysis of harps - n spectra by damasso et al .
the original and alternative parameter set 3 both find log @xmath24 log @xmath25 , while alternative parameters 1 and 2 find the opposite ; however , given the log @xmath10 errors , the magnitude of @xmath0log @xmath10 between the two stars is uncertain ( and could be zero ) .
interestingly , a higher [ fe / h ] in xo-2n is consistent across all derived parameters ; this also concurs with damasso et al .
( 2015 ) . despite the possibility of the stars having the same [ fe / h ] within errors , the fact that all four parameter sets found using different lines measurements and analyses techniques find [ fe / h]@xmath26[fe / h]@xmath27 points towards a real metallicity difference .
we return to this point in 3 .
all abundances were derived directly from equivalent width ( ew ) measurements performed with the ` splot ' task in iraf , except [ o / h ] , which was derived through spectral synthesis in moog .
abundance determinations of fe , ni , c , and o are detailed in teske et al .
( 2013 ) . for elements new to this work , line lists ( table [ tab : lines ] ) were compiled from melndez et al .
( 2014 ) and schuler et al .
( 2011b ) ; atomic parameters are from numerous sources and compiled in vald ( piskunov et al . 1995 ) .
all @xmath0(n - s ) abundance differences are listed in table [ tab : diffs ] .
the only element besides fe for which we measured two ionization states was titanium .
we removed 1@xmath21 outliers based on the mean and standard deviation of the line - by - line solar - normalized abundances .
this procedure changed the [ ti i / h ] and [ ti ii / h ] values by @xmath110.025 , but reduced the dispersion in the abundances by roughly a factor of two .
bergemann ( 2011 ) found the discrepancy between solar ti i and ii is solved by including non - lte line formation , and suggested that ti ii may be a better [ ti / h ] indicator in solar - type stars .
we tested whether non - lte corrections to ti i lines in xo-2n and -2s , calculated using statistical equilibrium and line formation codes described in bergemann ( 2011 ) and mafags - os models with the same stellar parameters derived here , change the [ ti i / h ] values .
we find these corrections do not significantly alter [ ti i / h ] , but as these are the first non - lte corrections calculated for cool , metal - rich stars , we provide the a(ti i)@xmath28-a(ti i)@xmath29 corrections in table [ tab : lines ] .
we report a total [ ti / h ] based on the lines of both species , with 1@xmath21 outliers removed within each species as described above .
the results of our measurements are in table [ tab : lines ] , including wavelength , @xmath30 , log @xmath31 , ews for each element for the sun , xo-2n , and xo-2s .
the final derived stellar parameters and their 1@xmath21 uncertainties are in table [ tab : stellar_params ] , and the @xmath32(x / h)@xmath33 and @xmath21 values calculated on a line - by - line basis for xo-2s
xo-2s are in table [ tab : diffs ] .
next we discuss the resulting abundances from the four different parameter sets derived above .
the advantage of studying binary stars is the likelihood that their formation environments were similar , so their compositions
assuming like xo-2n and -2s they are of similar mass and temperature should also be similar .
small differences in binary star abundances were previously suggested to be related to planet formation / migration ( e.g. , tucci - maia et al .
2014 ) . to investigate the xo-2 system for such abundance differences , we compare one star directly to the other , @xmath0(xo-2n - xo-2s ) . in figure [ fig2 ]
we show the line - by - line ( n - s ) differences and standard deviations of the differences ( error bars ) for our four parameter sets .
the black asterisks in figure [ fig2 ] represent the largest possible differences between ( n - s )
-204 k @xmath0t@xmath19 , 0.27 dex @xmath0log @xmath10 , and 0.11 dex [ fe / h ] . based on these differences , xo-2n may have less refractories than xo-2s , save a few elements si , fe , and ni .
these `` original '' parameter results also imply xo-2n is enhanced in volatile elements ; the slope for t@xmath34500 k elements is negative ( @xmath351.66e-4 dex / k ) .
however , the abundance differences , and thus slopes , change with different stellar parameters , as suggested in 2 . using a smaller line list , relative - to - solar / absolute results in the blue closed / open triangles in figure [ fig2 ] .
these parameters correspond to ( n - s ) differences of -83/-72 k @xmath0t@xmath19 , -0.09/-0.11 dex @xmath0log @xmath10 , and 0.09/0.11 dex [ fe / h ] .
interestingly the relative - to - solar and absolute analyses produce a very similar pattern , indicating that xo-2n is enhanced in most refractory elements and depleted in volatiles .
the red points in figure [ fig2 ] represent abundances represent parameters derived from a longer line list than the `` original '' parameter set , with measurements made with ares , and ( n - s ) differences of 3 k @xmath0t@xmath19 , 0.06 dex @xmath0log @xmath10 , and 0.08 dex [ fe / h ] .
these abundance differences show an even greater enhancement / depletion in xo-2n of refractory / volatile elements versus the abundances marked with blue triangles .
the different @xmath0(n - s ) abundances , and corresponding t@xmath8-slopes , resulting from different stellar parameters make it challenging to conclude whether any of the enhancements / depletions between xo-2n and -2s are real , within the bounds of measurement errors . to investigate how dependent the abundance differences are on @xmath0t@xmath19 and @xmath0log @xmath10 , which change the most between our different analyses
, we used the same ews and three test parameter sets that varied only in @xmath0log @xmath10 .
for these tests , we adopted @xmath0t@xmath19 -130 k ( n - s ) , based on the spread of the points in figure [ fig2 ] ; this @xmath0t@xmath19 is between the @xmath0t@xmath19s of the black asterisks ( -204 k ) and red circles ( 3 k ) .
specifically , the parameters for xo-2n / s were 5390/5520 k , 0.43/0.33 dex [ fe / h ] , and 1.20 km / s for @xmath22 , with log @xmath10 varying between 4.30/4.40 , 4.35/4.35 , and 4.40/4.30 .
the results of changing @xmath0log @xmath10 between xo-2n and -2s are shown in figure [ fig3 ] as blue ( -0.1 dex ) , green ( 0 dex ) , and orange ( 0.1 dex ) stars.changing @xmath0log @xmath10 has a minimal effect on the abundance differences , except for @xmath0(o / h ) , which is known to depend strongly on log @xmath10 . for most elements , the log @xmath10 tests result in @xmath0(n - s ) abundances of @xmath40 , falling somewhere in between the black , blue , and red points from figure [ fig2 ] .
however , in all three tests in figure [ fig3 ] , the abundances of si , fe , and ni are still above the zero difference line .
this is remarkably consistent with all four parameter sets derived in 2 and plotted in figure [ fig2 ] , which all show si , fe , and possibly ni enhancements in xo-2n .
varying many different aspects of the analysis and stellar parameters does not alter this enhancement pattern , providing strong evidence for a real abundance difference in si , fe , and ni between xo-2n and -2s .
robinson et al . ( 2006 ) and brugamyer et al .
( 2011 ) found differences in [ si / fe ] between known host and non - host stars ; robinson et al .
( 2006 ) also found that host stars are enhanced in nickel .
while these studies compare known hosts and non - hosts , they point specifically at si and ni as important to giant planet formation .
this is interesting given these are two elements we find enhanced in xo-2n , the hot jupiter host . indeed , plotting brugamyer et al .
( 2011 ) s [ si / h ] values against planet semi - major axis shows a dearth of silicon - poor hosts to close - in ( @xmath36 au ) planets , suggesting that silicon may play a role in close - in giant planet formation .
the enrichment of certain elements in xo-2n could be due to accretion of material preferentially enhanced in these elements .
simulations have shown that a significant fraction of giant planets migrating in disks as a result of planet - planet scattering cause much , if not all , of rocky disk material to be accreted onto the star ( raymond et al .
2011 ) . the rv evidence for a second object orbiting xo-2n ( damasso et al .
2015 ; knutson et al . 2014
; narita et al .
2011 ) adds weight to the planet - planet scattering migration scenario ( rasio & ford 1996 ; nagasawa et al .
2008 ) , by which xo-2nb could have migrated inward and subsequently had its eccentricity and spin - orbit alignment angle damped ( narita et al .
2011 ) .
xo-2s s lesser amount of fe , si , and ni may instead indicate that these elements are important to _ small _ planet formation ; xo-2s may have formed small planets that were ejected .
kaib et al .
( 2013 ) specifically studied wide binaries , finding that @xmath470% of 1 m@xmath14 - 1 m@xmath14 systems with the separation of xo-2n and -2s were unstable and ejected a planet within 10 gyr .
based on dynamical simulations of wide binary planet systems , morbidelli ( 2014 ) found that 73% of planets are ejected , and that 60% of wide binaries , versus 41% of isolated systems , lost a planet . in all parameter cases , a few refractory elements ( si , fe , ni ) stand out as enhanced in the cooler xo-2n .
this pattern is different from other abundance studies of binary exoplanet hosts .
liu et al . (
2014 ) found @xmath0[x / h ] values consistent with zero in the hat - p-1 g0+f8 binary .
though there is not consensus regarding differences between 16 cyg a ( g1.5 ) and b ( g3 ) , the proposed differences ( ramrez et al . 2011 ; tucci maia et al . 2014 ) either show @xmath40.04 dex uniform depletion with t@xmath8 , or a range of depletions increasing with t@xmath8 at the -0.03 to -0.05 dex level .
( we switched a - b values to b - a , to stay consistent with cooler - hotter star . ) mack et al .
( 2014 ) , who measured the double - host binary hd20781/hd20782 ( g9.5/g1.5 ) , show in their figure 3 non - zero differences for o , al , v , cr , fe , and co. are we seeing an effect in @xmath0[x / h ] trends due to t@xmath19 , moving from hotter stars and no differences to cooler stars and more significant differences ?
right now this question is hard to answer due to different analysis techniques , line lists , and the host star nature of these systems ( whether or not both stars host planets ) . in gratton et al .
( 2001 ) s figure 1 , plotting @xmath0[fe / h ] versus @xmath0t@xmath19 for stars without confirmed planets , there is no clear pattern
@xmath0[fe / h ] is near zero for stars with @xmath0t@xmath19 @xmath410 , 90 , 275 , and 330 k , and @xmath0[fe / h ] = 0.05 and 0.09 for stars with @xmath0t@xmath19 @xmath4120 and 275 k. desidera et al .
( 2006 ) s data hint at larger @xmath0[fe / h ] differences at larger @xmath0t@xmath19s , but desidera et al .
( 2004 ) s data show no real dependence of @xmath0[fe / h ] on @xmath0t@xmath19 .
more uniform abundances studies , and a larger sample , are needed to understand the temperature dependence of binary host star abundance differences and whether they are related to planet formation .
we performed a detailed stellar abundance analysis on xo-2n and xo-2s , binary stars of similar temperature and mass that host planets of different types , to look for effects of planet formation .
though binary stars are expected to have approximately identical abundances , some studies report differences between host stars that may be due to planet formation .
when comparing the stars relative to one another , there are hints of abundance differences .
we test how dependent these differences , and the @xmath0(x / h ) slopes with t@xmath8 , are on stellar parameters using two different line lists , solar - normalized and absolute abundance analyses , and measurements made by hand in iraf and automatically with ares .
most of the resulting @xmath0(x / h ) values vary widely , but a few refractory elements
si , fe , and possibly ni
consistently appear to be enhanced in xo-2n .
this pattern is independent of analysis method and adopted t@xmath19 and log @xmath10 value , giving confidence to the conclusion that xo-2n and -2s have real abundance differences .
this work adds to the few binary host star analyses , which may prove to be the most useful for constraining both the effects of planet formation on stellar abundances , and the bulk compositions of giant exoplanets .
however , it also illustrates how sensitive binary host star abundances and t@xmath8 slopes are to derived stellar parameters ; such studies approach the limit of measurement errors and should be considered carefully .
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 .
we thank the referee for their comments and edits that improved the paper .
_ facilities : _ lccccccccccc orig params , xo-2s & 5547 & 59 & 4.22 & 0.24 & 1.24 & 0.07 & 0.28 & 0.14 & 49 fe ii , 10 fe ii & yes & iraf + orig params , xo-2n & 5343 & 78 & 4.49 & 0.25 & 1.22 & 0.09 & 0.39 & 0.14 & 49 fe ii , 8 fe ii & yes & iraf + alt params 1 , xo-2s & 5523 & 49 & 4.44 & 0.11 & 1.35 & 0.08 & 0.36 & 0.04 & 23 fe i , 10 fe ii & yes & iraf + alt params 1 , xo-2n & 5440 & 69 & 4.35 & 0.19 & 1.29 & 0.09 & 0.45 & 0.06 & 23 fe i ,
9 fe ii & yes & iraf + alt params 2 , xo-2s & 5447 & 47 & 4.36 & 0.11 & 1.08 & 0.06 & 0.28 & 0.04 & 23 fe i , 10 fe ii & no & iraf + alt params 2 , xo-2n & 5375 & 78 & 4.25 & 0.20 & 1.02 & 0.11 & 0.39 & 0.07 & 23 fe i , 9 fe ii & no & iraf + alt params 3 , xo-2s & 5403 & 39 & 4.40 & 0.13 & 1.06 & 0.07 & 0.39 & 0.06 & 67 fe i , 8 fe ii & yes & ares + alt params 3 , xo-2n & 5406 & 32 & 4.46 & 0.08 & 1.11 & 0.05 & 0.47 & 0.05 & 72 fe i , 9 fe ii & yes & ares + lcccccccccc c i & 5052.17 & 7.86 & -1.304 & 33.9 & 36.4 & 33.0 + & 5380.34 & 7.69 & -1.615 & 19.4 & 22.4 & 20.5 + o i & 6300.30 & 0.00 & -9.717 & & & +
na i & 4751.82 & 2.10 & -2.078 & 13.1 & & 48.0 + & 5148.84 & 2.10 & -2.044 & 12.4 & 42.0 & 45.8 + & 6154.23 & 2.10 & -1.547 & 36.2 & 82.9 & 89.3 + & 6160.75 & 2.10 & -1.246 & 55.4 & 103.7 & 113.0 + lcccccccc @xmath0(c / h ) & 0.165 & 0.007 & -0.013 & 0.009 & -0.029 & 0.009 & -0.066 & 0.013 + @xmath37 & 0.16 & & -0.075 & & -0.078 & & -0.075 & + @xmath38 & -0.073 & 0.006 & 0.054 & 0.038 & 0.069 & 0.043 & 0.075&0.019 + @xmath39 & -0.013 & 0.025 & 0.066 & 0.009 & 0.079 & 0.008 & 0.075 & 0.020 + @xmath40 & -0.042 & 0.066 & 0.069 & 0.010 & 0.084 & 0.009 & 0.087 & 0.021 + @xmath41 & 0.104 & 0.032 & 0.068 & 0.030 & 0.075 & 0.032 & 0.050 & 0.029 + @xmath42 & 0.105 & 0.061 & -0.057 & 0.069 & -0.071 & 0.071 & -0.107 & 0.079 + @xmath43 & -0.109 & 0.050 & 0.096 & 0.042 & 0.114 & 0.041 & 0.088 & 0.041 + @xmath44 & -0.042 & 0.115 & 0.022 & 0.070 & 0.039 & 0.084 & 0.063 & 0.099 + @xmath45 & -0.086 & 0.023 & 0.075 & 0.027 & 0.112 & 0.032 & 0.163 & 0.035 + @xmath46 & -0.113 & 0.071 & 0.043 & 0.043 & 0.062 & 0.046 & -0.006 & 0.051 + @xmath47 & -0.105 & 0.033 & 0.072 & 0.042 & 0.090 & 0.041 & 0.073 & 0.028 + @xmath48 & 0.083 & 0.067 & 0.091 & 0.066 & 0.109 & 0.071 & 0.074 & 0.077 + @xmath49 & 0.061 & 0.054 & 0.061 & 0.056 & 0.077 & 0.061 & 0.075 & 0.060 + @xmath50 & 0.018 & 0.049 & 0.038 & 0.101 & 0.051 & 0.112 & 0.039 & 0.073 + @xmath51 & 0.019 & 0.002 & -0.003 & 0.007 & 0.010 & 0.008 & -0.051 & 0.009 + ( xo-2n - xo-2s ) relative abundances versus t@xmath8 for our test cases , keeping @xmath0t@xmath19 for ( n - s ) at -130 k , @xmath0[fe / h ] for ( n - s ) at 0.1 dex , and varying @xmath0 log @xmath10 ( n - s ) from -0.1 ( blue stars ) , to 0 ( green stars ) , to + 0.1 ( orange stars ) dex .
a dotted line shows zero difference .
, scaledwidth=100.0% ] | the chemical composition of exoplanet host stars is an important factor in understanding the formation and characteristics of their orbiting planets .
the best example of this to date is the planet - metallicity correlation .
other proposed correlations are thus far less robust , in part due to uncertainty in the chemical history of stars pre- and post - planet formation . binary host stars of similar type present an opportunity to isolate the effects of planets on host star abundances . here
we present a differential elemental abundance analysis of the xo-2 stellar binary , in which both g9 stars host giant planets , one of which is transiting .
building on our previous work , we report 16 elemental abundances and compare the @xmath0(xo-2n xo - s ) values to elemental condensation temperatures .
the @xmath0(n - s ) values and slopes with condensation temperature resulting from four different pairs of stellar parameters are compared to explore the effects of changing the relative temperature and gravity of the stars .
we find that most of the abundance differences between the stars depend on the chosen stellar parameters , but that fe , si , and potentially ni are consistently enhanced in xo-2n regardless of the chosen stellar parameters .
this study emphasizes the power of binary host star abundance analysis for probing the effects of giant planet formation , but also illustrates the potentially large uncertainties in abundance differences and slopes induced by changes in stellar temperature and gravity . |
in the last decade the study of the evolution of the qso luminosity function and of the clustering properties of qsos has been based on statistical analysis of even larger qso catalogues . for redshifts
lower than 2.2 the color techniques provide good selection methods .
now days the qso luminosity function evolution at @xmath3 is well established ( e.g. la franca & cristiani 1997 , goldschmidt & miller 1998 ) .
this kind of studies are mainly based on the most statistically significant qso samples such us the durham / aat qso sample ( boyle et al . 1990 ) at faint magnitudes ( @xmath4 ) , and the bright qso survey ( schmidt & green 1983 ) , the large bright qso survey ( hewett et al .
1995 ) , the edinburgh uvx qso survey ( goldschmidt et al . 1992 , miller et al . in prep ) , the homogeneous bright qso survey ( hbqs , cristiani et al .
1995 ) at bright magnitudes ( @xmath5 ) .
in total these samples collect about a thousand of qsos with @xmath3 .
but these samples are still not ideally suited to study the qso clustering and its evolution ( see andreani & cristiani 1992 ) .
the best qso samples for clustering analysis should have the highest possible surface densities and cover a contiguous areas , such as the planned qso 2df redshift survey ( croom et al .
1998 ) . in order to improve the signal - to - noise ratio in the estimate of the clustering of qsos
, we have built a new sample of 429 qsos down to @xmath6 over a contiguous area of 24.6 deg@xmath1 .
this sample has allowed the measure of a difference on the qso clustering evolution in comparison with that found for galaxies ( la franca et al .
1998 ) .
the survey is located in the south galactic pole ( sgp ) region , where some of the durham / aat sample areas are included ( boyle et al .
1990 ) , and part of the high - redshift qso survey of warren et al .
( 1991b ) was carried out .
the region has been also studied by campusano ( 1991 ) .
the central area of 1.7 deg@xmath1 in our region has been covered by the eso imaging survey in the b , v and i band ( eis , da costa 1998 , nonino et al .
1999 , prandoni et al . 1999 ,
zaggia et al . 1999 ) . the acquisition of the corresponding u data is planned for 1999 .
further deep multicolor imaging over an area of about half a sq.deg .
is planned with the wfi at the eso 2.2 m telescope .
moreover about 10 deg@xmath1 of the southern area of the region is included in the 2df qso redshift survey ( croom et al .
our survey is based on a set of ukst plates in the @xmath7 , @xmath8 and @xmath9 bandpass ( see blair & gilmore ( 1982 ) for a definition of the various systems ) .
ukst plates subtend @xmath10 .
details of the photographic plate material are listed in table 1 .
the plates have been scanned in mapping and imaging mode on the cosmos microdensitometer ( @xcite ) .
the mapping mode used a spot size of 16 @xmath11 m ( fwhm ) , corresponding to 1 arcsec . for each band - pass the digitized mapping mode data have been added together in order to increase the signal to noise ratio .
the reliability of the co - adding technique has been verified by many tests on ukst plates ( hawkins 1991 ) .
the resulting coadded digitized data has been analyzed by the cosmos crowded - field analysis software ( beard et al .
the resulting tables contained the instrumental magnitudes , the area above the threshold , the intensity weighted @xmath12 and @xmath13 positions , and other useful parameters .
we discriminated between point - like and extended sources using the cosmos image parameters , that is , on morphological grounds .
we defined a plane in which stellar locus was well determined , and drew the separation line near the locus .
for typically @xmath14 , we used the log(isophotal area ) versus magnitude plane . at fainter magnitudes , where stellar peak surface brightness ( @xmath15 ) are not saturated , the log@xmath16 versus magnitude plane provided a better separation .
the two separation lines produced the same star - galaxy ratio at the overlap magnitude .
relative photometry , using instrumental magnitudes , was produced by cosmos .
calibration was based on published sequences ( hawkins & bessell 1988 , @xcite ) .
lllcr + * emulsion & * filter & * number & * date & * exp .
+ & & & & * time + & & & & _ ( min . ) + iiao & ug1 & u 2639 & 1976 - 09 - 28 & 180 iiiaj & ug1 & u 6326 & 1980 - 09 - 03 & 180 iiiaj & ug1 & u 6380 & 1980 - 09 - 15 & 180 iiiaj & gg395 & j 9764 & 1984 - 11 - 23 & 70iiiaj & gg395 & j 9765 & 1984 - 11 - 23 & 70iiiaj & gg395 & j 9766 & 1984 - 11 - 23 & 70iiiaj & gg395 & j 9770 & 1984 - 11 - 24 & 70iiiaj & gg395 & j 9771 & 1984 - 11 - 24 & 70127 - 04 & rg630 & r 3498 & 1977 - 08 - 15 & 90iiia - f & rg630 & r 4676 & 1978 - 12 - 02 & 90 iiia - f & rg630 & r 9594 & 1984 - 09 - 22 & 90 iiia - f & rg630 & r 9672 & 1984 - 10 - 17 & 90 iiia - f & rg630 & r11331 & 1986 - 09 - 02 & 90 _ * * * * * * the candidates were extracted from a rectangle in the sky centered at @xmath17(1950 ) : @xmath18 @xmath19 @xmath20 and @xmath21(1950 ) : @xmath22 @xmath23 @xmath24 with limits @xmath25 , @xmath26 , @xmath27 , @xmath28 ( i.e. the limits in right ascension are the two lines distant @xmath29 from the meridian at @xmath17(1950 ) : @xmath18 @xmath19 @xmath20 ; see cristiani et al .
1995 , their table 1 ) .
all the objects inside a radius of 0.15 degrees centered at @xmath17(1950 ) : @xmath30 @xmath31 @xmath32 , and @xmath21(1950 ) : @xmath33 @xmath34 @xmath35 have been excluded from the catalogue as this region includes a crowded globular cluster .
it results a total area of 24.55 deg@xmath1 .
we have selected as candidates all the uvx `` not extremely extended '' objects with @xmath36 , satisfying a type of modified braccesi less - restricted two color criterion ( la franca et al .
1992 , cristiani et al .
the completeness of the selection has been tested against the qsos already known in the field with redshift in the range @xmath2 .
we have selected @xmath37 of the 176 blue qsos already known in the field with @xmath36 .
the completeness becomes @xmath38 if the `` red '' qsos from warren et al .
( 1991a ) are included in the comparison .
we have used the 3.6 m eso telescope , equipped with the meudon - eso fiber optic system mefos ( bellenger et al .
1991 ) , to obtain low - resolution spectra of the qso candidates in four observational campaigns between 1993 and 1995 .
mefos is a multifibre positioner which enables the conventional use of the eso boller & chivens spectrograph to be extended to multiobject spectroscopy .
the spectra of up to 29 objects inside the @xmath39 field of the prime focus of the 3.6 m telescope can be simultaneously recorded .
it consists of 29 remotely controlled arms , each carrying two _ spectroscopic fibers _ of 2.5 arcsec of diameter , one for the object , and one for the sky sampling .
we used the eso grating n. 13 ( 150 grooves / mm ) with a resolution of 35 in the wavelength range @xmath40 .
the ccd was a 512@xmath41512 tektronics with 27 @xmath11 m pixels ( eso n. 32 ) .
wavelength calibration was carried out by comparison with exposures of he and ar lamps .
the campaign dates with the number of pointings and the exposure time are listed in table 2 .
the data reduction was performed under midas and followed the optimized spectra extraction from fiber spectrographs of lissandrini et al .
no absolute flux calibration has been applied .
the spectra have been flux calibrated in relative fluxes just for the sake of facilitating the identification .
only the most evident cosmic rays hits have been removed . | spectra , position , magnitudes and colors are presented for 485 faint ( @xmath0 ) emission line objects selected with the ultraviolet - excess ( uvx ) criterion on a area of 24.6 deg@xmath1 in the south galactic pole .
the objects were selected from the analysis of pixel - to - pixel stacking of cosmos scans of ukst u j and r plates . the candidates were observed with the meudon - eso fiber optics system ( mefos ) at the eso 3.6 m telescope .
429 type 1 agns have been identified ( 373 in the redshift range @xmath2 ) .
this sample has allowed the measure of a difference on the qso clustering evolution in comparison with that found for galaxies ( la franca et al .
1998 ) .
the region is part of the eso imaging survey ( eis ) and of the 2df qso redshift survey . |
since the successful fabricating of graphene in 2004 , @xcite its excellent transport properties , @xcite such as the exceptional electron and thermal transport properties at room temperature , have attracted a lot of research interest to apply it in making future electronic devices .
@xcite many ideas and attempts have been going on .
one of the most fundamental and revolutionary attempts is to control the valley degree of freedom to realize the signal s storage and transport.@xcite supposed an achievement of this idea in future , graphene will probably replace all semiconductors in conventional electronics and bring a new era in many fields . before the coming of the new era , there is a long way to improve the control over transport in graphene . inspired by the crucial role of the doping in semiconductor ,
many researchers have studied the effects related to disorder in graphene,@xcite which consists of adsorbed atoms ( or molecules ) , charged impurities , vacancies , or other topological defects .
it is anticipated to modify the transport property of pristine graphene to fulfill the requirement .
recently , a peculiar topological line - defect in graphene was reported experimentally by lahiri et al .
@xcite this topological line - defect is created by alternating the stone - thrower - wales defect and divacancies , leading to a pattern of repeating paired pentagons and octagons , as shown in fig .
1 . in particular
, it is found in this experiment that this line - defect has metallic characteristics .
great attention has been attracted to the line - defect and further investigations manifest a variety of its promising applications.@xcite although a lot of theoretic proposals are presented in graphene to realize the valley filter before , all of them are not realized till now due to the extreme difficulty in the experimental implementation . after the discover of line - defect , gunlycke and white @xcite reported that a valley polarization near @xmath0 can be achieved by scattering off a line - defect.@xcite this original idea subsequently arouse physics community s intensive interests.@xcite as a new discovery
, the extended line - defect is very promising to realize the valley polarization in graphene .
however , because the @xmath0 valley polarization only maintains at a large incident angle in these reported results.@xcite a serious problem will be encountered in experiment that the electron must always follow the direction of the line - defect to maintain a high valley polarization .
this transport characteristic is almost fatal to utilize efficiently the high valley polarization .
the purpose of this paper is to address a feasible approach to realize a @xmath0 valley polarization at a small incident angle by using a resonator , which consists of several paralleling line - defects . given a high valley polarization by scattering off one line - defect , a question is putted forward naturally : what will happen in the presence of two or more line - defects ? motivated by finding an efficient realization of the high valley polarization , we therefore study the electron transmission coefficient in the presence of several line - defects .
intriguingly , it is found that an obvious change happens to the transmission coefficients with varying the distance between two line - defects .
in particular , for the electron with a definite momentum and incident angle , the transmission coefficients can reach to one at a definite distance between two line - defects .
simple numerical analysis argues that the large transmission coefficients should be attributed to the standing wave structure due to the electron s quantum interference , which is also named as fabry - p@xmath1rot interference in optics .
based on a further numerical simulation , an efficient device to realize a @xmath0 valley polarization is further proposed for six line - defect configuration at the end .
this fascinating finding opens a way to realize the valley polarization by line - defects .
the rest of this paper is organized as follows : in sec .
[ sec : models ] , we introduce the hamiltonian in the tight - binding model and present the formula of the transmission coefficient . the numerical results are discussed in sec .
[ sec : discussions ] .
finally , a brief summary is given in sec .
[ sec : conclusions ] .
in the tight - binding approximation , a single layer of graphene with a line - defect , sketched in fig .
1(a ) , can be described by the following hamiltonian : @xmath2 where @xmath3 and @xmath4 represent the electron annihilation operators on the graphene site @xmath5 and line - defect sites , respectively , and @xmath6 refers to the nearest neighboring sites .
supposed a variation less than @xmath7 in the hopping terms @xmath8 and @xmath9,@xcite here we adopt @xmath10 .
is from the fourier transformation along y axis . ] because of the existence of line - defect , the translation symmetry along @xmath11 axis breaks down for a graphene system , as sketched in fig .
however , the translation symmetry along @xmath12 axis still retains and therefore @xmath13 is a good quantum number , namely a conserved quantity . for the sake of the convenience in computing , we do the fourier transform only along @xmath12 axis as follows : @xmath14 @xmath15 represents the lattice constant in pristine graphene . after the fourier transformation , the hamiltonian in eq .
( [ m1 ] ) is decoupled into @xmath16 .
note that here @xmath17 can be effectively represented by a quasi - one - dimensional lattice model , as shown in fig .
readily , we can describe the hamiltonian @xmath17 as follows : @xmath18 where @xmath19 , @xmath20 $ ] , and @xmath21 represents the unit length between the neighboring supercells at the graphene part . here
, @xmath22 represent the position of a supercell , @xmath23 takes the integer number from 1 to 4 to denote the different columns in a supercell , and @xmath24 in @xmath25 corresponds to the up / down site in the same column in fig .
1(c ) . or @xmath26 dirac cone .
( b ) the relationship between the electron s incident angle and momentums .
( c ) and ( d ) manifest the transmission coefficient in @xmath27 and @xmath26 valleys . in this case , there is only one line - defect . here
, the dotted line represents the approximate result in the low energy limit : @xmath28 according to the landauer - bttiker formula and the hamiltonian in eq .
( [ m3 ] ) , the transmission coefficient of the line - defect system at zero temperature and low bias voltage can be represented as : @xcite @xmath29},\end{aligned}\ ] ] where @xmath30 is the retarded / advanced green s function related to the line - defect hamiltonian and @xmath31 with the retarded / advanced self - energy @xmath32 .
note that here the left / right lead is represented by a semi - infinite quasi - one - dimensional graphene lattice sketched in fig .
1(c ) and the sample hamiltonian refers to the line - defect hamiltonian , namely @xmath33 .
( b ) and @xmath26 ( c ) are mapped as a function of the incident angle @xmath23 in the presence of two line - defects . here
, electron s energy is set as @xmath34 . ]
it is worth noting that we can readily compute the transmission coefficients of the electron with a fixed incident angle @xmath23 in two dirac points , @xmath27 and @xmath26 , by using the above formula . as shown in fig .
1(b ) , two dirac points @xmath27 and @xmath26 which are located at @xmath35 $ ] for a pristine graphene are now cast at @xmath36 and @xmath37 for a graphene with a line - defect , @xmath38 $ ] .
therefore , an electron in two valleys satisfies the following relations : @xmath39 and @xmath40 , where @xmath41 represents electron s group velocity along the @xmath42 direction .
combined with the dirac electron s linear dispersion relation @xmath43 , the transmission coefficients can be readily mapped as a function of the electron incident angle @xmath23,@xcite where @xmath44 . for simplicity in discussing
, we will never distinguish the difference of the dirac points in two brillouin zones in the following and uniformly adopt @xmath27 and @xmath26 to represent the two nonequivalent dirac points .
besides , some previous papers reported that in the vicinity of the dirac point , the transmission coefficient of low energy electron will obey the law : @xmath45 . in order to make this work essentially self - contained
, we also confirm this conclusion and present the related derivations in appendix .
even though the transmission coefficient in the vicinity of the dirac point was given by an approximated method in many references,@xcite to the best of our knowledge , the exact numerical calculation without any analytical approximations is not unambiguously shown before . given the realization of the valley polarization using the line - defect in future , what the transmission coefficient behaviors at a relatively large energy is a fundamental question . in fig .
2 , the transmission coefficient is mapped as a function of the electron s incident angle for different energies .
it can be observed the line of @xmath46 ( red ) overlaps exactly with the dotted line , @xmath47 , which represents the transmission coefficient of low energy electron in the vicinity of the dirac point by an approximate method and is also given in appendix .
one subtle difference between the numerical and approximate results is zero transmission probability nearby @xmath48 for exact numerical result .
however , this physical observation is not manifested in the approximate result ( dotted line ) . a severe deviation from the dotted line
is clearly seen for the lines with @xmath49 , which indicates that a excellent valley transmission can only be maintained at not very large energy .
( left ) and @xmath26 ( right ) valleys . in this case
, two line - defects are considered and the fixed distance between them is @xmath50 . ]
the search of the valley polarization material or device attracts physics community greatly .
although a proposal was given recently by scattering off a line - defect,@xcite it looks much difficult for graphene with only one line - defect to make electron leave away from the line - defect and finally reach the key device , and meanwhile maintain at a high valley polarization in this process .
inspired by the valley polarization in the line - defect graphene , we expect this problem can be resolved by scattering off two or several line - defects . in fig .
3 , the transmission behavior is mapped as a function of the incident angle in the presence of two line - defects .
3(a ) shows a lattice model of graphene with two line - defects and meanwhile presents a schematic interpretation of @xmath51 which represents the distance between two nearest neighbouring line - defects . using the formula ( 4 ) ,
the transmission coefficients in the @xmath27 and @xmath26 valleys are mapped as a function of the incident angle in fig 3(b ) and 3(c ) .
a mirror symmetry with respect to @xmath52 can be seen clearly between them . comparing to the case of one line - defect
, there is almost no distinct change for the lines of @xmath53 both in fig .
3(b ) and 3(c ) .
however , the transmission coefficient starts varying for @xmath54 and in particular some transmission peaks can be observed obviously at small incident angles for the lines of @xmath55 and @xmath56 .
this exciting finding makes us fascinated to explore what happens to the two - line - defect graphene and why there is a high transmission probability at some incident angles . which is exactly the green circle path in fig .
4 . two lines ( blue and red ) correspond to that of @xmath27 and @xmath26 valleys , respectively . in this case , two line - defects are considered and the fixed distance between them is @xmath50 . ] in order to demonstrate a whole physical picture , we map the transmission coefficient as the function of @xmath57 in fig .
the left and right maps , which correspond to the @xmath27 and @xmath26 valleys , respectively , manifest a clear symmetry .
intriguingly , two large transmission regions ( dark red ) can be observed both in the @xmath27 and @xmath26 valleys .
note that the positive ( negative ) @xmath58 represents an incident electron along positive x direction ( negative x direction ) .
these peaks remind us some quantum interferences occur in the presence of two line - defects , e.g. electron s resonant tunneling phenomena .
to confirm the origin of the high transmission above , we plot two lines selectively in fig .
these two lines which belong to @xmath27 and @xmath26 valleys respectively correspond to half circle of the green route in fig .
4 , representing @xmath59 . obviously , the transmission peaks at @xmath60 and @xmath61 are from the intersection of the green circle route and the high transmission region ( dark red ) . when substituting @xmath62 and @xmath63 into the formula : @xmath64 and @xmath65 , we find two quantities are equivalent within the physical error range .
this is to say , the @xmath0 transmission does originate from the resonant tunneling , which is often observed in the system of quantum dot.@xcite valley is given in the presence of several paralleling line - defects : ( a ) @xmath66 , ( b ) @xmath67 , ( c ) @xmath68 , and ( d ) @xmath69 .
note that here @xmath70 represents the number of paralleling line - defects . in each case , four lines with differen @xmath71 , the distance of two nearest neighbouring line - defects , are manifested . here
, electron s energy is set as @xmath34 ] according to the quantum theory of wave - particle duality , the low energy electron in the vicinity of the dirac point is inclined to exhibit undulation characteristics . when scattering off two line - defects , a standing wave state between two line - defects helps electron resonantly tunnel through the barriers of the line - defects .
consequently , we can observe this perfect transmission phenomenon in above figures .
using the valley polarization formula @xmath73 , the @xmath0 valley polarization can not be realized at very small incident angle in the presence of two line - defects , e.g. at the positions of the @xmath60 and @xmath61 peaks .
nevertheless , an exciting thing is that @xmath0 valley polarization can almost be reached when @xmath23 takes the value of @xmath74 $ ] , a relatively small incident angle . at least , the results argue that the high valley transmission at a small incident angle is feasible by scattering off two line - defects .
( left ) and @xmath26 ( right ) valleys . in this case
, six line - defects are considered and the fixed distance between arbitrary two neighboring line - defects is @xmath75 . ] inspired by the above results , we expect an excellent case could occur to the graphene with several line - defects . in fig . 6 ,
the transmission coefficient only in @xmath27 valley is shown in the presence of several paralleling line - defects : ( a ) @xmath66 , ( b ) @xmath67 , ( c ) @xmath68 , and ( d ) @xmath69 , where @xmath70 represents the number of paralleling line - defects . in fig .
6(a ) with @xmath66 , almost no obvious change can be observed with comparison to the case of @xmath76 in fig .
nevertheless , one extra transmission peak still appears in the line of @xmath55 ( red ) . without question
, it should be attributed to the effect of multiple interference between line - defects . when setting @xmath67 in fig .
6(b ) , except @xmath77 , the transmission coefficients of all other lines become active and manifest some peaks at some fixed incident angles . for the cases of @xmath68 in fig .
6(c ) and @xmath69 in fig .
6(d ) , more transmission peaks appears and overlap each other .
however , the electron transmission coefficient in case of @xmath77 decreases gradually with increasing the number of paralleling line - defects .
it is because at @xmath34 electron s wave length is considerably larger than the distance of two nearest neighbouring line - defects , @xmath77 . for sake of concreteness
, here we restrict to the case of six line - defects and make great efforts to manifest clearly a picture of a @xmath0 valley polarization . in fig .
7 , the transmission coefficients in two valleys are mapped in the panel of @xmath78 .
it is clearly shown that comparing to the case of two line - defects , more large transmission regions ( dark red ) exist in fig .
7 , the case of six line - defects . as explained in fig . 4 and 5
, these high transmission regions should likewise be from the resonant tunneling due to the quantum interference , which is also named as fabry - p@xmath1rot interference in optics . in fig .
8 , the transmission coefficient is plotted as a function of the incident angle @xmath23 , which follows half of the green path in fig .
note that here the electron momentum @xmath79 is taken as @xmath80 .
obviously , the marking a , b , and c transmission peaks should be attributed to the three intersections between the green path and the large transmission regions ( dark red ) in fig . 7 . here ,
@xmath81 , @xmath82 , and @xmath83 illustrate that the a transmission peak is attributed to the intrinsic scattering property of a line - defect and the peaks at b and c are from the resonant tunneling of the positive and negative first standing waves , respectively . much importantly ,
not only a perfectly resonant transmission but also a @xmath0 polarization can be obtained at the @xmath84 resonant peak by the polarization formula : @xmath73 . in a conclusion , we predict that the perfect @xmath0 polarization can be achieved by scattering off many line - defects , which origins in the quantum resonant tunneling phenomenon . even though only exhibiting the cases of two and six line - defects here , in principle this high transmission and thus the @xmath0 polarization can be realized in graphene with arbitrary number of line - defects .
this feature shed a light on constructing a feasible valley polarization resonator by using multiple line - defects . which is exactly the green circle path in fig .
two lines ( blue and red ) correspond to that of @xmath27 and @xmath27 valleys , respectively . in this case , six line - defects are considered and the fixed distance between arbitrary two neighboring line - defects is @xmath75 . ]
in summary , we study the valley polarization in the presence of several paralleling line - defects .
importantly , the valley polarization changes with varying the distance between the line - defects and the number of the line - defects .
in particular , the valley polarization can reach to the maximum value of one , namely a 100@xmath85 polarization , at a small incident angle .
further numerical calculations manifest that @xmath0 polarization should be attributed to the behavior of electron quantum interference , more precisely the phenomenon of the standing wave between the line - defects .
the findings make us get rid of helplessness in the choose of the electron s incident angle and provides a reliable way to realize a controllable valley polarization in experiment . with the great advance in graphene related technologies , a valley polarization resonator and also a detector of a valley polarization , being the key components in valleytronics , are promising to be realized in the system of the line - defect graphene . in the preparation of this paper
, we notice a reference on two paralleling line - defect in recent@xcite and intriguingly , nearly a perfect perfect valley polarization is also observed .
this report also illustrates that the valley filter of several paralleling line - defects should feasible in experiment .
yang liu and juntao song want to thank liwei jiang and yisong zheng for their help in the preparation of this paper .
we gratefully acknowledge the financial support from nsf - china under grant nos . 11204065(j.t.s . ) , 11274089(y.l . ) , 10974043(y.x.l . ) , and 11274364(q.f.s . ) .
is also supported by rfdphe - china under grant no .
in this appendix , we present a detailed derivation for the transmission coefficient @xmath86 of low energy electron in the vicinity of the dirac point . note that similar derivations can also reference to some previous papers.@xcite in the landauer - bttiker formula , eq .
( [ m4 ] ) , the green s function @xmath87 can be readily expressed as : @xmath88 . here
, @xmath89 is just a two - by - two matrix related to @xmath33 and therefore the key point is to obtain the retarded self - energy function , @xmath90 .
that can be solved by the surface green s function . in the following ,
we firstly present an analytic method of obtaining the surface green s function @xmath91 and then derive the transmission coefficient of low energy electron in the vicinity of the dirac point . based on the schematic representation in fig .
1(c ) , a hamiltonian of a semi - infinite right lead can be written as : @xmath92 where @xmath93 means a hopping term from the @xmath94 sub - layer in the @xmath95 supercell to the @xmath96 sub - layer in the @xmath97 .
obviously , the relations hold true : @xmath98 , @xmath99 , @xmath100 , and @xmath101 .
thus , the eigen equation @xmath102[\mathcal{c}]=e[\mathcal{c}]$ ] can be written as : @xmath103 where @xmath104 and @xmath105 $ ] represents the eigenvalue and eigenstate , respectively .
for simplicity , we cast the above eigen equation into a subspace and consequently obtain the following expression , @xmath106 here , @xmath107 where @xmath108 .
note that other quantities , such as @xmath109 , @xmath110 and @xmath111 , can be also obtained by using dyson equation of green s function .
for simplicity in the following derivation , we define some symbols as : @xmath112 where @xmath113 takes a form of @xmath114 following the similar derivation processes of the eigen equation above , the green s function can be written readily as : @xmath115 after a gauge transformation by a unitary matrix , which is denoted as : @xmath116 the @xmath113 matrix , all @xmath117 matrixes , and green s function in eq .
( [ a6 ] ) can be transformed into : @xmath118 where @xmath119 and @xmath120 .
note that the green s function must be diagonalized if all coefficients in eq .
( [ a6 ] ) , namely all @xmath121 , can be diagonalized by this unitary transformation .
there are many known mathematic method to calculate an inverse matrix of a tridiagonal matrix .
however , we refrain from introducing these methods here .
we use @xmath126 to denote the determinant of the tridiagonal matrix on the left side of eq .
( [ a9 ] ) and @xmath127 to denote the determinant of the tridiagonal submatrix except the former @xmath22 columns and rows of the tridiagonal matrix.@xcite then , we can obtain the recursive equations : @xmath128 and @xmath129 . in the limit of infinite long lead , corresponding to an infinite large tridiagonal matrix in eq .
( [ a9 ] ) , the relation @xmath130 maintains . from the formula @xmath131
, we can readily obtain @xmath132 where @xmath133 is adopted . performing a inverse unitary transformation with respect to eq .
( [ a8 ] ) , the surface green s function of a semi - infinite long lead can be expressed as : @xmath134 where @xmath135 and @xmath136 .
now , we get an exact expression about the surface green s function of a semi - infinite long lead . combining the self energy function and the green s function of the line - defect @xmath137 substituting the above formula into the landauer - bttiker formula , eq .
( [ m4 ] ) , we can express the transmission coefficient exactly as : @xmath138,\end{aligned}\ ] ] where @xmath139 , @xmath140 , and @xmath141 . in the low energy limit , it is not difficult to find @xmath142 , @xmath143 , and @xmath144 .
thus , the transmission coefficient is approximately written as:@xcite @xmath145 note that we still reserve the term of @xmath146 in @xmath147 and @xmath148 because some terms in @xmath147 and @xmath148 are in the same order as @xmath146 .
assuming incident electron s energy is much smaller than the hopping terms @xmath149 , @xmath8 and @xmath9 , an approximate expression of the transmission coefficient can be written as : @xmath150/2}{2\tau^2_1\tau_2\cos{\theta}(1-m_1)+e^2 + 4\tau^4_1\cos^2{\theta}e^2}\end{aligned}\ ] ] following the relations : @xmath39 , @xmath151 , @xmath152 , @xmath153 , and @xmath154 , we adopt the approximation in the low energy limit : @xmath155 , @xmath156 , and @xmath157 in eq .
( [ a14 ] ) .
we can readily obtain @xmath158 that is namely the transmission coefficient of low energy electron in the vicinity of two dirac valleys.@xcite + + k. s. novoselov , a. k. geim , s. v. morozov , d. jiang , y. zhang , s. v. dubonos , i. v. grigorieva , and a. a. firsov , science * 306 * , 666 ( 2004 ) . k. s. novoselov , a. k. geim , s. v. morozov , d. jiang , m. i. katsnelson , i. v. grigorieva , s. v. dubonos , and a. a. firsov , nature ( london ) * 438 * , 197 ( 2005 ) . k. i. bolotin , k. j. sikes , z. jiang , g. fudenberg , j. hone , p. kim , and h. l. stormer , solid state commun .
* 146 * , 351 ( 2008 ) .
a. a. balandin , s. ghosh , w. bao , i. calizo , d. teweldebrhan , f. miao , and c. n. lau , nano lett .
* 8 * , 902 ( 2008 ) .
a. rycerz , j. tworzydo , and c.w .
j. beenakker , nat . phys . * 3 * , 172 ( 2007 ) .
z. z. zhang , k. chang , and k. s. chan , appl .
. lett . * 93 * , 062106 ( 2008 ) .
h. zeng , j. dai , w. yao , d. xiao , and x. cui , nat . nanotech . * 7 * , 490 ( 2012 ) .
y. jiang , t. low , k. chang , m. i. katsnelson , and f. guinea , phys .
* 110 * , 046601 ( 2013 ) .
f. zhai , y. ma , and y .- t .
zhang , j. phys . : condens .
matter * 23 * , 385302 ( 2011 ) .
w. long , q .- f .
sun , and j. wang , phys .
* 101 * , 166806 ( 2008 ) .
h. jiang , z. qiao , h. liu , j. shi , and q. niu , phys .
lett . * 109 * , 116803 ( 2012 ) .
zhang , j. hu , b. a. bernevig , x. r. wang , x.c .
xie , and w. m. liu , phys .
lett . * 102 * , 106401 ( 2009 ) .
j. song , h. liu , h. jiang , q .- f .
sun , and x.c .
xie , phys .
b * 86 * , 085437 ( 2012 ) .
s. s. alexandre , a. d. lcio , a. h. castro neto , and r. w. nunes , nano lett .
, 2012 , * 12 * , 5097 ( 2012 ) .
d. a. bahamon , a. l. c. pereira , and p. a. schulz , phys .
b * 83 * , 155436 ( 2011 ) .
s. okada , t. kawai , and k. nakada , j. phys .
* 80 * , 013709 ( 2011 ) .
l. kou , c. tang , w. guo , and c. chen , acs nano * 5 * , 1012 ( 2011 ) .
d. gunlycke and c. t. white , phys .
lett . * 106 * , 136806 ( 2011 ) .
l , z. liu , h .- b .
yao , l .- w .
jiang , w .- z .
gao , and y .- s .
zheng , phys .
b * 86 * , 045410 ( 2012 ) .
l. jiang , x. lv , and y. zheng , phys .
a * 376 * , 136 ( 2011 ) .
h. jiang , l. wang , q .- f .
sun , and x. c xie , phys .
b * 80 * , 165316 ( 2009 ) ; j. song , h. liu , h. jiang , q .- f . sun , and x. c. xie , phys .
b * 85 * , 195125(2012 ) .
t. p. pareek , phys .
lett . * 92 * , 076601 ( 2004 ) ; y. xing , q .- f .
sun , and j. wang , phys .
b * 73 * , 205339 ( 2006);phys .
b * 80 * , 235411 ( 2009 ) . _ electronic transport in mesoscopic systems _ , edited by s. datta ( cambridge university press 1995 )
. j. song , q .- f .
sun , j. gao , and x. c. xie , phys .
b * 75 * , 195320 ( 2007 ) .
j. liu , j. song , q .- f .
sun , and x.c .
xie , phys .
b * 79 * , 161309 ( 2009 ) .
l. jiang , y. zheng , c. yi , h. li , and t. l , phys .
b * 80 * , 155454 ( 2009 ) ; | the electron transport of different conical valleys is investigated in graphene with extended line - defects .
intriguingly , the electron with a definite incident angle can be completely modulated into one conical valley by a resonator which consists of several paralleling line - defects .
the related incident angle can be controlled easily by tuning the parameters of the resonator .
therefore , a controllable @xmath0 valley polarization , as well as the detection of the valley polarization , can be realized conveniently by tuning the number of line - defects and the distance between two nearest neighbouring line - defects .
this fascinating finding opens a way to realize the valley polarization by line - defects . with the advancement of experimental technologies , this resonator is promising to be realized and thus plays a key role in graphene valleytronics . |
since topological concepts were first introduced to understand the behaviour of electrons in the quantum hall effect @xcite , such ideas have become a powerful theme in condensed matter physics , applicable to a wide - range of different systems @xcite . in topological materials
, the eigenstates making up an electronic energy band can be characterised by nontrivial topological invariants , which are robust against any perturbations that do not close the energy band gaps .
these invariants can be directly related to the existence of remarkable topological edge states localised near the system boundary , which lead , for example , to the robust and precise quantisation of conductance in the quantum hall effect @xcite .
while the study of real materials is usually complicated , the essential physics of various topological phenomena can be captured through simple lattice models .
perhaps the most famous of these is the harper - hofstadter ( hh ) model of a charged electron hopping on a 2d tight - binding square lattice in a uniform magnetic field @xcite .
not only does this model exhibit the quantum hall effect thanks to its topological energy bands and unidirectionally - propagating edge states , but it has a rich fractal energy spectrum known as the hofstadter butterfly .
most importantly , this model has opened up the way for the study of quantum hall physics beyond traditional solid - state materials , beginning with its realisation in solid - state superlattice devices @xcite , ultracold atomic gases @xcite and silicon photonics @xcite .
recently , topological concepts have entered the realm of classical physics with specially - designed mechanical meta - materials @xcite and arrays of coupled classical harmonic oscillators such as pendula @xcite , gyroscopes @xcite and electrical circuits @xcite . for many of these systems ,
topological properties are encoded in the normal - mode dynamical matrix , in analogy with the electronic hamiltonian of a topological lattice model .
the demonstration of topological edge states in these classical systems has also ignited the hope that such set - ups could find practical applications , for instance , as loss - free propagating acoustic waveguides @xcite .
in neutral systems , such as atoms , photons and classical oscillators , the simulation of topological phenomena has been made possible by important interdisciplinary developments in the creation of so - called synthetic gauge fields @xcite . in the case of the hh model , the effect of the magnetic field on a charged electron is to modify the tight - binding tunnelling amplitudes by complex spatially - dependent phases called the peierls phases @xcite .
the route to simulating this model with a neutral system is therefore to find an alternative mechanism to generate suitable complex hopping phases . in ultracold gases ,
this can be , for example , through the control of internal atomic states , by laser - assisted tunneling or lattice shaking @xcite , while in photonics , this has been realised through the insertion of link resonators in ring - resonator arrays @xcite or by twisting waveguides helically along the direction of light propagation to simulate a time - dependent modulation @xcite . finally , in classical systems
, analogous effects arise naturally for coupled gyroscopes @xcite , with angular - momentum biasing @xcite , or can be hard - wired into the coupling - connections in lattices of pendula @xcite or electrical elements @xcite . in this paper
, we propose a different conceptual approach for how to introduce non - trivial coupling phases , and hence how to simulate topological phenomena , in classical harmonic oscillator lattices through the application of temporally - periodic modulations to system parameters . in this approach
, we build on advances and methods from ultracold gases @xcite and photonics @xcite , where time - periodic modulations were previously introduced to realise topological lattice models . for classical harmonic oscillators ,
the first steps in this direction were presented in @xcite for the case of two coupled oscillators ; here , we develop these ideas further to propose a modulation scheme that simulates the analogue of the hh model in a large array of coupled oscillators . even though our discussion is focused on mechanical pendulum systems ,
the same ideas straightforwardly extend to other types of oscillators such as lumped - element electric circuits of @xcite . as in the gyroscopic lattices of @xcite ,
our proposal breaks time - reversal symmetry and can be associated with energy bands describing the collective excitation modes of the coupled pendula system , which are characterised by a nontrivial chern number : the topological invariant of the quantum hall effect .
we go beyond these previous works to show how , not only the edge states , but the hallmarks of the chern number itself could be seen experimentally in a system of classical harmonic oscillators . while these systems focus on the haldane model @xcite , our scheme focuses on the hh model , where , thanks to the easy tunability of our synthetic flux , we show that the full hofstadter butterfly of states could also be observed .
finally , we note that our proposal stands in contrast to the recent experiments on pendula @xcite and electrical circuits @xcite that did not break time - reversal symmetry , but instead realised two copies of the hh model with opposite synthetic flux : one copy for each of the two pseudo - spin " degrees of freedom defined in these systems . in the quantum analogy
, this leads to the quantum spin hall effect in which the edge states are not topologically robust as they are unprotected against perturbations which flip the pseudo - spin " .
the paper is organized as follows . in sec .
[ sec : nem ] we introduce the system of coupled classical harmonic oscillators and highlight the limit in which newton s equations of motion are analogous to the heisenberg equations for a tight - binding model . in sec .
[ sec : hhm ] , we introduce the specific frequency modulation scheme from which we recover the hh model in the high frequency expansion of floquet theory . the efficiency of our idea is quantitatively assessed in sec .
[ sec : results ] where numerical simulations of the full classical equations of motion , are compared to the predictions of the hh model .
we find that the oscillation amplitude spectra exhibit wide peaks corresponding to the bulk energy bands ; we show that these peaks build up to give the hofstadter butterfly as a function of the synthetic magnetic flux .
we excite topologically - protected edge states showing that these are chiral and immune from back - scattering .
we then discuss a classical counterpart of the integer quantum hall effect from which we can extract the chern number . in sec .
[ sec : instability ] , we investigate the dynamical stability of the system and the role of the counter - rotating - wave terms , together with the consequent limitations of the mapping onto the hh model .
we comment in sec .
[ sec : experimental ] on possible experimental implementations of our proposal using arrays of either mechanical pendula or electric circuits , before drawing conclusions in sec .
[ sec : conclusions ] .
and @xmath0 .
pendula on the edges are coupled to a rigid wall.,scaledwidth=40.0% ] we consider a square lattice of @xmath1 sites .
each lattice site , labelled by two indexes @xmath2 , hosts a pendulum of mass @xmath3 and frequency @xmath4 , whose oscillation plane is rigidly fixed , for example , along the unit - cell diagonal , as shown in fig .
[ fig : system ] .
we assume that the natural frequencies of the pendula are periodically modulated in time around their mean value @xmath5 .
the displacement of each pendulum from its equilibrium point along the fixed direction is denoted by @xmath6 .
all the pendula are coupled to their nearest neighbours , while the pendula on the edges of the lattice are coupled to a fixed wall .
the coupling is provided , for example , by springs of rest length equal to the lattice spacing and of constant @xmath7 and @xmath0 along the two @xmath8 and @xmath9 directions respectively . in the small oscillation regime
the elongation of each spring is proportional to the difference of the spatial displacement of the pendula that it connects @xmath10 and , in absence of any pre - tension , the force is determined by the standard hooke s law .
including loss mechanisms with a friction coefficient @xmath11 , newton s equations of motion for a pendulum driven by the external force @xmath12 read : @xmath13 the goal of this section is to show that in a suitable regime newton s equations of motion - can be rewritten in the same form as the heisenberg equation for a tight - binding hamiltonian , where the bosonic annihilation and creation operators are replaced by @xmath14-numbers . starting from the displacements @xmath6 and the corresponding momenta @xmath15 , classical complex variables corresponding to the annihilation operators of quantum harmonic oscillator are defined as : @xmath16 it is straightforward to see that the square modulus @xmath17 is proportional to the oscillation energy of the @xmath2-th pendulum .
assuming for simplicity that the product @xmath18 is constant for every site @xmath19 and defining @xmath20 and @xmath21 , we can combine eq .
and eq . into equations for the complex amplitudes @xmath22 of the form : @xmath23 \alpha_{i , j } \\ & -{\mathrm{i}}\left [ \frac{\tilde{\omega}_{i , j}}{2}- \frac{\omega_{i , j}^2(t)}{2\tilde{\omega}_{i , j}}+{\mathrm{i}}\gamma_{i , j } + 2\omega_x+2\omega_y\right ] \alpha^*_{i , j}\\ & + { \mathrm{i}}\sum_{\pm1}\omega_x\left(\alpha_{i\pm1,j}+\alpha^*_{i\pm1 , j}\right ) \\ & + { \mathrm{i}}\sum_{\pm1}\omega_y\left(\alpha_{i , j\pm1}+\alpha^*_{i , j\pm1}\right)+ { \mathrm{i}}f_{i , j}^\text{ex}(t ) , \end{split } \label{alphadrivendot}\ ] ] where the damping rate @xmath24 .
the external driving force is taken to be monochromatic at frequency @xmath25 with amplitude proportional to @xmath26 , @xmath27 equations can be simplified under the assumption that the natural frequencies @xmath28 of the pendula are much larger than all other frequencies in the problem and that the driving frequency @xmath25 is comparable to the @xmath28 s . in this case ,
the fastest time dependence of @xmath22 is proportional to @xmath29 , while that of the conjugate variables @xmath30 is proportional to @xmath31 .
it can be seen that in the limit @xmath32 the contribution of the @xmath33 s to the motion equation for @xmath34 is negligible as the fast oscillations quickly average to zero , in which the @xmath34 terms only have a ( relatively ) slow time - evolution while the @xmath33 terms oscillate at @xmath35 and therefore quickly average to zero . ] . as a result
, it is legitimate to neglect in eq .
the terms involving the complex conjugate variables @xmath33 , an approximation that in quantum optics is well - known under the name of the rotating wave approximation ( rwa ) . applying the rwa
, we interpret eq . without driving and dissipation , as the classical analogue of the heisenberg equation of motion obtained from the following hamiltonian : @xmath36
this is a bosonic tight - binding hamiltonian , where the terms on the first line account for the on - site energy and those on the second line describe the hopping . in the hamiltonian ,
the rwa consists of neglecting all non - energy - conserving terms proportional to @xmath37 and @xmath38 and only keep the energy - conserving ones @xmath39 .
in the previous section , we have shown how newton s equations of motion can be rewritten in the rwa in the same form as the heisenberg equation of a tight - binding model . in this section
we proceed to identify a suitable modulation scheme for the @xmath4 that reduces to an effective hh model in the floquet picture . inspired by the realization of the hh model with ultra - cold atoms in optical lattices @xcite and by related proposals in the photonic context @xcite , we take the natural frequency of the pendula , _
i.e. _ a part of the single - particle on - site energy in the tight - binding analogy , to be temporally and spatially modulated according to : @xmath40 with a spatial dependence @xmath41 direction , sketching the spatial modulation of the natural frequencies of the pendula with period @xmath42.,scaledwidth=45.0% ] the temporal modulation @xmath43 is taken as periodic with frequency @xmath44 , and the static spatial modulation : @xmath45 is a saw - tooth function along the @xmath8-direction , of period @xmath46 , as schematically shown in fig .
[ fig : system_x ] . while hopping along @xmath9 is unaffected by both the spatial and temporal modulations , bare hopping along @xmath8 would be strongly suppressed by the spatial modulation eq . .
hopping in this direction is however restored by the temporal modulation : when the frequency of the modulation is chosen to be on resonance with the detuning of neighbouring sites , a uniform coupling is restored over the whole lattice @xcite . given the specific sawtooth form of eq .
chosen for @xmath47 , a bi - harmonic temporal modulation is needed : @xmath48 \end{split } \label{timemodulation}\ ] ] where the amplitude @xmath49 is position independent .
the first component at frequency @xmath44 restores a uniform hopping between the pairs of pendula with natural frequency difference @xmath44 , while the second component at frequency @xmath50 addresses the pairs with a frequency difference @xmath50 . as we shall see in the next subsection ,
the position - dependence of the phase @xmath51 offers us a way to control the phase of the hopping along @xmath8 @xcite . in order to obtain the desired hh model ,
a possible choice is the following : @xmath52 with the modulations introduced in the previous paragraph , we now show how the equations of motion in eq .
take the form of the heisenberg equations of motion for the hh model by applying the floquet formalism of time - periodic systems at high - frequency @xcite . substituting the chosen forms of the temporal and static modulation eq .
and eq . into eq . and performing the rwa
, we straightforwardly obtain : @xmath53 \alpha_{i , j}\\ & + { \mathrm{i}}\sum_{\pm 1 } \left[\omega_x \alpha_{i\pm1 , j}+ \omega_y \alpha_{i , j\pm1}\right ] + { \mathrm{i}}f_{i , j}^\text{ex}(t ) .
\end{split } \label{alpharwadot}\ ] ] in the following we focus on the case where a single @xmath54 site is driven .
we also assume that the fastest frequency of the temporal modulation @xmath55 is fast compared to all other frequencies except the bare natural oscillation frequency : @xmath56 the first inequality is required by the rwa , while the latter justifies our calculating the effective dynamics after a time - average over the period @xmath57 of the temporal modulation . in this regime ,
a simple expression for the time - evolution of the whole system at stroboscopic times @xmath58 can be obtained using floquet theory @xcite . to this purpose
we introduce new variables : @xmath59 we then look for the steady - state solution , where @xmath60 oscillates with the same frequency as the external driving force , which is tuned in the vicinity of the natural frequency of the driven pendulum @xmath61 . an effective equation of motion can be obtained for the @xmath60 variables by time - averaging eq . over one period of the fast temporal modulation ,
that is by applying the magnus expansion to the lowest order @xcite .
after some calculations , time - averaged equations for @xmath60 are found .
the full forms of the coefficients of the effective equations are expressed as sums of bessel functions and can be found in the appendix [ app : fulleff ] .
a much simpler expression for the effective equations of motion is obtained by considering only the terms that give the largest contribution in these sums : @xmath62 where , with the definition of the modulation phase in eq .
, we have introduced the following quantities : @xmath63 as detailed in appendix [ app : fulleff ] , the condition of validity for eq .
is that the next order of the bessel functions can be neglected , @xmath64 from this expression , it is apparent that a very large period of the spatial modulation @xmath46 would be optimal , with a linear ramp in the bare frequencies of the harmonic oscillators in eq . .
while this configuration is possible for ultracold atomic systems @xcite , it is not possible for these classical harmonic oscillators as the condition @xmath65 for the rwa does not allow us to use an arbitrarily large value of @xmath46 .
more details of these issues can be found in sec .
[ sec : experimental ] . in the following
, we shall see that a good compromise can be found already for @xmath46 as small as @xmath66 . in eq .
, the bare coupling frequencies @xmath67 and @xmath68 are renormalised by the bessel functions , and the effective couplings along the two directions are different . in order to recover the usual hh model with equal hopping in the two directions ,
it is useful to start from suitably chosen @xmath69 that will give the same effective @xmath70 in the two directions : @xmath71 one must of course be careful in choosing the amplitude of the modulation @xmath49 , such that the frequencies in eq . satisfy the inequality of eq . . as with the hopping terms , so is the amplitude of the driving term in eq .
renormalised by the bessel functions .
we choose the bare driving force on the pumped site as : @xmath72 in order to keep the effective driving intensity @xmath73 constant as we vary @xmath49 or @xmath44 . without the last two terms describing driving and dissipation ,
eq . is the classical version of the heisenberg equations derived from the hh hamiltonian : @xmath74 \end{split } \label{hhh}\ ] ] where @xmath75 is the energy at which the hh model is coherently probed .
the non - trivial hopping phase in eq . is determined by the modulation phase in eq . and controlled by the parameter @xmath76 .
as usual , the magnetic flux enclosed within a plaquette is calculated by summing the phases accumulated on each link when hopping around a plaquette of the lattice . with the definition in eq . , the sum of the phases gained by the complex hopping elements in eq . , is uniform for the whole lattice and equal to : @xmath77 as a result ,
our model of coupled pendula is a classical simulator of the hh model and allows us to generate the hofstadter butterfly by simply tuning the phase of the temporal modulations according to eq . .
the tunability of this artificial magnetic field is a main advantage of our scheme compared to other systems , where the magnetic flux is fixed @xcite .
we note that any other choice of the temporal modulation phase @xmath51 leading to the same flux per plaquette would give an equivalent hh model that only differs by a gauge transformation . before proceeding , it is important to note that the analysis we have done so far is strictly valid only in the deep rwa limit @xmath78 where the counter - rotating - wave terms @xmath79 in eq . can be safely neglected and effective equations such as eq . can be derived . for lower values of @xmath80 , the first effect beyond the rwa is a global shift in the detuning frequency : @xmath81 , with : @xmath82 this shift
is predicted by including the counter - rotating - wave part of the @xmath79 in eq . , that in the frame rotating at @xmath80 corresponds to a non - rotating term and
so includes the leading order of the counter - rotating - wave effects . as discussed in @xcite , this has the form of a classical analogue of the bloch - siegert shift . for even smaller @xmath80
, the counter - rotating - wave terms can also lead to dynamical instabilities and will be discussed in sec .
[ sec : instability ] .
in order to assess the validity of the effective hh model , obtained via the floquet approach , we have performed numerical simulations of the full equations of motion including the counter - rotating - wave terms as well as the full spatial and temporal modulations of the oscillation frequencies .
we have quantitatively compared these predictions to those from the heisenberg equation of motion for the hh model in eq . .
we have numerically integrated the set of @xmath83 time - dependent equations for a square lattice of @xmath1 sites with a standard fourth - order runge - kutta until a steady state is achieved at long evolution times @xmath84 . to match with the floquet picture ,
we have performed a stroboscopic sampling at times @xmath85 . at these times @xmath86 , under a monochromatic drive at @xmath25 the steady - state amplitudes have the form @xmath87 , where @xmath88 are the oscillation amplitudes .
the total intensity is obtained as the sum of the oscillation amplitudes @xmath89 of all the pendula @xmath90 .
the response spectra describe the total intensity as a function of the detuning @xmath91 .
depending on the value of @xmath92 and the position @xmath54 of the driven site , different behaviours are expected . for @xmath92 within a band of the bulk hh model and @xmath54 located far from the edges , the bulk of the system is excited and so the response is dominated by delocalized band states .
when @xmath92 belongs to a gap between two energy bands , the total oscillation is suppressed unless @xmath54 is located close to an edge of the system , so that edge states can be excited .
a detailed study of these regimes will be the subject of the next subsection . to facilitate the reader in following the discussion , it is useful to first review some fundamental facts about the hh model . in the energy - flux plane , the states organize themselves in a recursive and self - similar structure , known as the hofstadter butterfly .
the structure of the bands and the gaps is the simplest when the magnetic flux is a rational number @xmath93 with @xmath94 being co - prime positive integers @xcite . in this special case
, one has exactly @xmath95 sub - bands separated by @xmath96 energy gaps .
when @xmath95 is even , the two middle bands touch at dirac points .
the eigenstates making up the hofstadter bands are associated with a non - trivial topological invariant , called the chern number @xmath97 .
this invariant is related to the transverse conductivity of the integer quantum hall effect @xcite . in a finite system with non - zero @xmath97 ,
unidirectionally propagating states exist at the edges of the system at energies located in the gaps between bands .
since there are no states available for backscattering , these states are topologically protected from disorder @xcite . in most optical @xcite and mechanical @xcite experiments
so far , the presence of such chiral edge states represents the smoking gun of the non - trivial topology .
we now show that such edge states can be identified also in our scheme
. lattice with a flux per plaquette @xmath98 and losses @xmath99 where the central lattice site is pumped .
the blue circles and orange triangles correspond to the spectra calculated from the steady state of eq . as a function of the detuning from the external driving frequency @xmath100 , including the shift in eq . .
parameters are : @xmath101 , @xmath102 , @xmath103 for the blue circles , and @xmath104 , @xmath105 , @xmath103 for the orange triangles .
black curve shows the spectra as obtained from the driven - dissipative hh model @xcite .
green areas show the position of the hh bands .
frequencies indicated by arrows are used in fig .
[ fig : topologicalstate].,scaledwidth=48.0% ] the response spectra presented in fig .
[ fig : spectra ] are calculated by driving the central site of a lattice of @xmath106 pendula with a flux per plaquette of @xmath98 in units of the flux quantum .
losses are chosen to be @xmath99 , which is large enough to ensure that the excitation does not reach the edges of the system and that the response is only determined by the bulk properties .
the position of the hh bands is highlighted by the green shaded area .
the accuracy of the floquet effective dynamics is assessed by comparing the numerical result with the prediction of the driven - dissipative hh model studied in @xcite .
the blue circles show the response spectrum of a system of bare frequency @xmath101 , temporal modulation frequency @xmath102 and amplitude such that @xmath103 from eq . .
this corresponds to @xmath107 , @xmath108 and @xmath109 .
the orange triangles are instead calculated for @xmath104 , @xmath105 , and again @xmath103 , which corresponds to @xmath110 , and the same @xmath67 and @xmath68 as before . for the latter set of parameters , the inequality in eq .
is better satisfied and indeed the peak amplitudes of the pendula - spectra agree very well with the driven - dissipative hh model depicted with a solid black line . with the former set of parameters , the two central peaks are difficult to resolve but the external regions of high intensity still coincide well with the position of the hh bands . for even smaller values of @xmath80
the deviation from the hh model would be much more dramatic , as the rwa is no longer valid and the contribution of the counter - rotating - wave terms is not negligible any more . pendula with @xmath98 , @xmath111 , @xmath103 , @xmath102 and @xmath101 , obtained by pumping a single site on the lower edge for different values of the detuning of the pumping frequency @xmath112 .
panel * a * and * c * are for @xmath113 located in a gap and show the excitation of an edge state with clockwise chirality . in panel
* c * , the blue square indicates the position of a defect composed of @xmath114 missing pendula on the left edge .
panel * b * is for @xmath115 located again in a gap and shows the excitation of an edge state with counter - clockwise chirality .
panel * d * is for @xmath116 located in an energy band and shows the excitation of bulk states .
the arrow indicates the position of the pumped site on the lower edge.,title="fig:",scaledwidth=24.0% ] pendula with @xmath98 , @xmath111 , @xmath103 , @xmath102 and @xmath101 , obtained by pumping a single site on the lower edge for different values of the detuning of the pumping frequency @xmath112 .
panel * a * and * c * are for @xmath113 located in a gap and show the excitation of an edge state with clockwise chirality . in panel * c * , the blue square indicates the position of a defect composed of @xmath114 missing pendula on the left edge .
panel * b * is for @xmath115 located again in a gap and shows the excitation of an edge state with counter - clockwise chirality .
panel * d * is for @xmath116 located in an energy band and shows the excitation of bulk states .
the arrow indicates the position of the pumped site on the lower edge.,title="fig:",scaledwidth=24.0% ] pendula with @xmath98 , @xmath111 , @xmath103 , @xmath102 and @xmath101 , obtained by pumping a single site on the lower edge for different values of the detuning of the pumping frequency @xmath112 .
panel * a * and * c * are for @xmath113 located in a gap and show the excitation of an edge state with clockwise chirality . in panel * c * , the blue square indicates the position of a defect composed of @xmath114 missing pendula on the left edge .
panel * b * is for @xmath115 located again in a gap and shows the excitation of an edge state with counter - clockwise chirality .
panel * d * is for @xmath116 located in an energy band and shows the excitation of bulk states .
the arrow indicates the position of the pumped site on the lower edge.,title="fig:",scaledwidth=24.0% ] pendula with @xmath98 , @xmath111 , @xmath103 , @xmath102 and @xmath101 , obtained by pumping a single site on the lower edge for different values of the detuning of the pumping frequency @xmath112 .
panel * a * and * c * are for @xmath113 located in a gap and show the excitation of an edge state with clockwise chirality . in panel * c * , the blue square indicates the position of a defect composed of @xmath114 missing pendula on the left edge .
panel * b * is for @xmath115 located again in a gap and shows the excitation of an edge state with counter - clockwise chirality .
panel * d * is for @xmath116 located in an energy band and shows the excitation of bulk states .
the arrow indicates the position of the pumped site on the lower edge.,title="fig:",scaledwidth=24.0% ] in fig .
[ fig : topologicalstate ] we show the steady state intensity distribution obtained by pumping one site in the middle of the bottom edge with different values of the detuning @xmath112 corresponding to the arrows in fig . [ fig : spectra ] .
system parameters are @xmath101 , @xmath102 and @xmath103 as for the blue spectra of fig .
[ fig : spectra ] , while losses @xmath111 are smaller in order to allow the steady state amplitude oscillations to propagate over an adequate distance .
the intensity distributions shown in panels * a * , * b * and *
c * are obtained by driving the system in the band gap where the chiral edge - states are expected .
the pendula with the biggest oscillation amplitudes are indeed localized on the edge of the system .
as one can see by comparing panels * a * and * b * , the direction of unidirectional propagation changes when the sign of @xmath92 is changed , as expected from the hh model .
the distance covered along the edge before decaying is set by the ratio of the group velocity of the edge state over the loss rate , as usual in such driven - dissipative systems @xcite .
the topological robustness of such edge states is highlighted in panel * c * , where a defect modelled as @xmath117 missing pendula was introduced on the left edge of the lattice .
the edge state is found to propagate around the defect , without being scattered into the bulk .
finally , panel * d * shows the steady - state intensity distribution when the detuning corresponds to a hh band : in this case , the excitation penetrates into the bulk and shows no preferred chirality .
pendula with @xmath99 , @xmath103 , @xmath105 and @xmath104 , obtained by pumping the central site .
panels * a * and * c * are obtained from full numerical integration of newton s equation of motion , while panels * b * and * d * are obtained from a driven - dissipative hh model as in @xcite , and are shown for comparison .
the first row is for @xmath98 and @xmath118 corresponding to the center of the highest band .
the second row is for @xmath119 and @xmath120 corresponding to the middle of the central band.,title="fig:",scaledwidth=24.0% ] pendula with @xmath99 , @xmath103 , @xmath105 and @xmath104 , obtained by pumping the central site .
panels * a * and * c * are obtained from full numerical integration of newton s equation of motion , while panels * b * and * d * are obtained from a driven - dissipative hh model as in @xcite , and are shown for comparison .
the first row is for @xmath98 and @xmath118 corresponding to the center of the highest band .
the second row is for @xmath119 and @xmath120 corresponding to the middle of the central band.,title="fig:",scaledwidth=24.0% ] + pendula with @xmath99 , @xmath103 , @xmath105 and @xmath104 , obtained by pumping the central site .
panels * a * and * c * are obtained from full numerical integration of newton s equation of motion , while panels * b * and * d * are obtained from a driven - dissipative hh model as in @xcite , and are shown for comparison .
the first row is for @xmath98 and @xmath118 corresponding to the center of the highest band .
the second row is for @xmath119 and @xmath120 corresponding to the middle of the central band.,title="fig:",scaledwidth=24.0% ] pendula with @xmath99 , @xmath103 , @xmath105 and @xmath104 , obtained by pumping the central site .
panels * a * and * c * are obtained from full numerical integration of newton s equation of motion , while panels * b * and * d * are obtained from a driven - dissipative hh model as in @xcite , and are shown for comparison .
the first row is for @xmath98 and @xmath118 corresponding to the center of the highest band .
the second row is for @xmath119 and @xmath120 corresponding to the middle of the central band.,title="fig:",scaledwidth=24.0% ] while the existence of the chiral edge states is one hallmark of a quantum hall system , we now demonstrate that characteristic features of the driven - dissipative hh model can also be observed in the response of the bulk of the system . in fig .
[ fig : steadystate ] , we present the intensity distribution of the system when the central site of the lattice is driven with a detuning frequency resonant with an energy band . for the panels in the upper row ,
the magnetic flux is @xmath98 , while @xmath119 for the panels in the lower row ; losses are @xmath99 in all panels .
panels * a * and * c * show the distributions of the full numerical integration with @xmath104 , @xmath105 , @xmath103 and can be compared with panels * b * and * d * that are obtained from the driven - dissipative hh model as in @xcite .
we notice that the two pattern distributions are in excellent qualitative agreement , in particular where the periodicity of the pattern is concerned . as expected , when the natural frequency @xmath80 is reduced away from the rwa limit , the spatially symmetric structure of the oscillation amplitude pattern is broken and the agreement with the driven - dissipative hh model is less good ( not shown ) . and frequency detuning @xmath121
the total intensity of the oscillation is depicted with the color scale , where lower intensities are in darker colors .
black regions indicating negligible intensities correspond to the energy gaps between energy bands .
the peculiar structure of the hofstadter butterfly pattern is clearly visible .
system parameters are : @xmath104 , @xmath105 , @xmath99 , @xmath103 , for a lattice of @xmath106 sites where only the central one is pumped.,scaledwidth=48.0% ] the scheme that we have proposed in the previous sections allows us to change the magnetic flux @xmath76 per plaquette by simply adjusting the phase of the modulation of each pendula in eq . . in this way ,
our scheme is sufficiently flexible to explore a wide portion of the flux - frequency plane . collecting a large number of numerically calculated spectra of the total intensity for different values of @xmath76 in a single color - plot
, we obtain the pattern shown in fig .
[ fig : butterfly ] .
although the presence of dissipation does not allow us to resolve details of the fractal structure that are smaller than the loss rate , the emerging picture shows a close resemblance to the hofstadter butterfly @xcite .
such a result would constitute the first experimental visualization of the hofstadter butterfly in classical physics . in order to speed up the numerical calculations needed to plot fig .
[ fig : butterfly ] , the equations of motion in eq . have been solved by expanding the solution in the fourier basis : @xmath122 and then truncating the series to a value @xmath123 that is sufficiently large to ensure the convergence of the solution .
solving the linear system for the fourier components coupled by the time - dependent modulation is equivalent to , but much faster than , solving the full differential equations until reaching the steady state .
more details on this decomposition method are found in appendix [ app : fourier ] .
it is interesting to note that the agreement with the well - known hofstadter butterfly gets worse in the top and bottom regions of the spectra and eventually breaks down .
those areas correspond to very low values of the magnetic flux @xmath76 for which the number of bands is large and the smoothing has most dramatic consequences .
to understand the break - down from the butterfly structure , we checked the solutions with the full numerical runge - kutta integration method and we observed that in these regions the system is dynamically unstable .
this means that a stationary steady state can not exist and therefore the hofstadter butterfly pattern completely breaks down .
these deviations stem from the strength of the counter - rotating - wave - terms as controlled by the amplitude of the temporal modulation @xmath49 . in our simulation , we take @xmath124 to be fixed which , from eq . , results in an increase of @xmath49 as the flux is decreased .
larger @xmath49 implies a larger contribution from counter - rotating - wave terms , that results in the emergence of a parametric instability @xcite .
we note that the spectra obtained from the fourier decomposition method appear to be still stable despite the exponential growth of the runge - kutta solution because such a method begins from the ansatz of a steady - state solution and so can not predict the instability .
more insights into the instability will be given later in the paper .
in the previous sections , we have demonstrated that the system of frequency - modulated coupled harmonic oscillators can be a classical simulator of the hh model . as a final point
, we now show that an analogue of the integer quantum hall effect ( iqhe ) can be observed even in a classical mechanical context .
a main feature of the iqhe is that a current appears in the direction orthogonal to the applied electric field where the associated conductivity is an integer multiple of the conductance quantum @xmath125 .
this integer is related to the chern number of the populated bands of the hh model . for the bosonic model considered here ,
these bands can be populated by a suitable excitation scheme .
our proposal is inspired by a recent work on photonic systems @xcite .
pendula with @xmath98 , @xmath99 , @xmath103 , @xmath105 and @xmath104 , obtained by pumping the central site with a detuning corresponding to the middle of the lowest band @xmath126 . for panel * a * there is no external force @xmath127 , and the centre of the intensity distribution indeed remains at the centre of the system . for panel * b * , the force @xmath128 is directed along the @xmath9-direction as indicated by the arrow and induces a sizeable rightward shift of the centre of the distribution towards the positive @xmath8 direction as expected from the iqhe.,title="fig:",scaledwidth=24.0% ] pendula with @xmath98 , @xmath99 , @xmath103 , @xmath105 and @xmath104 , obtained by pumping the central site with a detuning corresponding to the middle of the lowest band @xmath126 . for panel
* a * there is no external force @xmath127 , and the centre of the intensity distribution indeed remains at the centre of the system .
for panel * b * , the force @xmath128 is directed along the @xmath9-direction as indicated by the arrow and induces a sizeable rightward shift of the centre of the distribution towards the positive @xmath8 direction as expected from the iqhe.,title="fig:",scaledwidth=24.0% ] in order to introduce an analogue of the electric field in our model , we need to generate a linear potential acting on the excitation field @xmath22 .
applying such a potential along the positive @xmath9-direction would correspond to a term of the form @xmath129 in eq . .
equivalently , and perhaps more straightforwardly , the analogue electric field could be introduced through an additional linear gradient of the bare oscillation frequency along the @xmath9-direction in eq . , with @xmath130 . in fig .
[ fig : halleffect]-*a * we show the intensity distribution of a system of @xmath131 pendula with @xmath98 and @xmath99 in the absence of the external force and for a detuning @xmath126 such that only the lowest band is excited with a resulting population that is approximately uniform . as discussed in @xcite , these conditions can be fulfilled provided the energy width of the band @xmath132 is much smaller than the band gap @xmath133 and the loss rate falls in between these two energy scales .
the intensity pattern is symmetric both in @xmath8 and in @xmath9 and is centred around the middle point .
when a non - zero force @xmath128 along the vertical direction is applied as in panel * b * , the oscillation pattern clearly shifts to the right in the direction orthogonal to the applied force .
this is a manifestation of the lorentz force and a clear signature of a hall - like effect . in order to assess the quantized nature of this hall effect
, we can quantify the lateral shift as @xmath134 and compare it to the prediction for the integer quantum hall effect in driven - dissipative systems @xcite , @xmath135 where @xmath95 is the denominator of the rational flux @xmath93 and @xmath97 is the chern number of the excited band .
the real number @xmath136 quantifies the ( spurious ) contribution of the bandwidth , that is responsible for a non - uniform population of the band , as well of the other neighbouring bands .
the quantized nature of the iqhe is most apparent in the small @xmath137 limit in the coefficient of the leading term . &
@xmath138 & @xmath139 + & & @xmath140 & & @xmath141 + & & @xmath142 & @xmath143 & @xmath142 + & & @xmath144 & & + & & @xmath142 & @xmath145 & @xmath145 + & & @xmath146 & & + & & @xmath142 & @xmath142 & @xmath147 + & & & & @xmath148 + & & @xmath143 & @xmath149 & @xmath143 + in table [ table : chern ] we summarize the estimated chern numbers @xmath150 of different bands for four different @xmath76 as calculated from eq . for a system of @xmath106 lattice sites , @xmath104 , @xmath105 , @xmath151 and @xmath152 .
the constant @xmath136 , that does not depend on the losses @xmath137 , is eliminated by calculating the linear coefficient in the shift given in eq . for two values of @xmath153
choosing any value in this frequency range only affects the calculation of the chern number to the least - significant digit .
when the method is not applicable because @xmath154 , the corresponding chern number of these bands was not calculated . in all other cases , the agreement with the chern number @xmath97 of the hh model is good and confirms the expected integer quantum hall effect in a classical mechanics context .
in this section we discuss in more detail the instability observed in the small flux region of the hofstadter butterfly of fig.[fig : butterfly ] and the corrections to the hh model due to the counter - rotating - wave terms in the equations of motion eq . .
as the amplitude of these terms is proportional to @xmath155 , their importance for a fixed @xmath156 increases as the magnetic flux @xmath76 is lowered according to eq . , explaining why the instability was only observed close to the upper and lower edges of the butterfly pattern . in order to more easily
access this physics , we have introduced a cleaner strategy to reinforce the counter - rotating - wave terms by decreasing the bare frequency of the pendula @xmath80 while keeping other parameters fixed .
this approach comes with the great advantage that we can compare energy spectra with the same value of the magnetic flux as the strength of the counter - rotating - wave terms increases . in the following we allow
the bare frequency to be @xmath157 , by relaxing the constraint in eq . .
recently , there have been several works discussing the role of the counter - rotating - wave terms in bosonic hamiltonians on topology @xcite .
however , exception made for a work on edge state instabilities @xcite , only a little attention has been so far devoted to the interplay between topology and parametric instabilities .
we now investigate how counter - rotating - wave terms lead to the appearance of complex energy modes in the spectrum and to a dramatic distortion of the hh bands which is accompanied by a disappearance of topological edge states . to this end , it is convenient to recast the equations in eq . , together with the corresponding equations for @xmath33 , without driving and dissipation , in the matrix form : @xmath158 where @xmath159 is a column vector .
the eigenvalues of the @xmath160 time periodic matrix @xmath161 represent the oscillation frequencies of the system . due to the specific form of the coupling between the variables @xmath162 and @xmath163
, we notice that this matrix is not hermitian with respect to the standard positive - definite inner product .
its block - structure : @xmath164 where @xmath165 and @xmath166 are @xmath167 real symmetric matrices , periodic with the modulation period @xmath57 , guarantees that it is hermitian with respect to the indefinite - inner product defined by the matrix : @xmath168 and of common use in the bogoliubov theory of dilute bose gases @xcite . the non - hermiticity of @xmath161 with respect to the standard inner product is crucial for the appearance of the complex energy - modes and the emergence of the instability . to treat the time - dependent problem
, we can apply the floquet theory of time - periodic systems . the evolution within one period @xmath169 of the modulation
is then expressed as that with a time - independent stroboscopic floquet matrix @xmath170 , defined as the time - ordered exponential @xcite : @xmath171 if @xmath161 were a hermitian matrix , then @xmath172 would be unitary and the eigenvalues of the stroboscopic floquet matrix @xmath170 would always be real @xcite . in our system , instead , the matrix @xmath173 describes , in terms of the complex variables in eq . , the one - period evolution of the system which is symplectic in the original @xmath6 and @xmath15 variables . as a result ,
this matrix is not hermitian with respect to the standard inner product , which implies that , in general , the eigenvalues of the floquet matrix can also be complex , so the system can be dynamically unstable @xcite .
the matrix @xmath172 is the _ mapping at a period _ of the system with periodically varying parameters @xcite , and it is the classical analogue of the stroboscopic time - evolution operator : @xmath174 .
it is known from classical mechanics textbooks that if all the eigenvalues @xmath175 of the mapping at a period @xmath172 are distinct and lie on the unit circle in the complex plane , the system is strongly stable @xcite . since the floquet matrix is related to @xmath172 by eq .
, the stability of the system can be checked just by looking at the eigenvalues of @xmath170 .
notice that the condition on the eigenvalues of @xmath172 to lie on the unit circle implies that the floquet spectra has to be real .
in fact , if @xmath176 is an eigenvalue of the floquet matrix @xmath170 , then it is related to the eigenvalue of @xmath172 as : @xmath177 from eq . .
it is straightforward to check that if @xmath178 , then @xmath179 is on the unit circle . in analogy with the above - mentioned bogoliubov theory @xcite
, it directly follows from the specific form of the matrix @xmath161 that if @xmath180 is an eigenvector of @xmath170 with eigenvalue @xmath176 , then @xmath181 is also an eigenvector but with opposite eigenvalue @xmath182 .
the @xmath183 eigenvalues of @xmath170 can be then distinguished according to their _ krein signature _
@xcite : @xmath184 that is the sign of the norm corresponding to the @xmath136 scalar product of the above - mentioned bogoliubov theory @xcite . by definition ,
the krein signature is either @xmath185 .
all stable eigenvalues of @xmath170 have either positive or negative krein signature , while all unstable eigenvalues have zero krein signature ; this also holds for the eigenvalues of @xmath172 . in the complex plane .
red points corresponds to eigenvalues with krein signature @xmath186 , points in blue are the eigenvalues with @xmath187 and green points are unstable eigenvalues with @xmath188 .
panel * a * shows a stable configuration for @xmath189 , where all eigenvalues lies on the unit circles .
panel * b * shows the unstable configuration @xmath190 , where some eigenvalues are clearly not lying on the unit circle .
parameters are : @xmath191 , @xmath42 , @xmath105.,title="fig:",scaledwidth=24.0% ] in the complex plane .
red points corresponds to eigenvalues with krein signature @xmath186 , points in blue are the eigenvalues with @xmath187 and green points are unstable eigenvalues with @xmath188 .
panel * a * shows a stable configuration for @xmath189 , where all eigenvalues lies on the unit circles .
panel * b * shows the unstable configuration @xmath190 , where some eigenvalues are clearly not lying on the unit circle .
parameters are : @xmath191 , @xmath42 , @xmath105.,title="fig:",scaledwidth=24.0% ] in fig .
[ fig : krein ] we show the representation on the unit circle of the eigenvalues of the matrix @xmath172 , coloured according to their krein signature .
eigenvalues with positive @xmath192 are shown in red , eigenvalues with negative @xmath192 are displayed in blue and unstable eigenvalues with zero krein signature are displayed in green .
as the ratio @xmath193 is changed , the eigenvalues @xmath175 of @xmath172 rotate on the unit circle .
whenever eigenvalues with opposite krein signature meet , they can leave the unit circle , meaning that the system becomes unstable .
this instability happens every half - round - trip : since the eigenvalues @xmath175 perform a full rotation on the unit circle when @xmath193 is increased by @xmath194 , the instability occurs each time @xmath195 , with @xmath196 . , modulo @xmath44 , associated with either a positive krein - signature @xmath192 or with @xmath188 and @xmath197 .
the system has a magnetic flux @xmath191 , with periodic boundary conditions ; driving and dissipation are not included .
the energy spectra are presented as a function of @xmath193 , for a fixed value of the amplitude modulation @xmath198 .
when all the eigenvalues at a particular frequency are stable , we denote them in black , otherwise we denote them either in red or blue .
red points are purely real modes , while blue points are the unstable modes , plotted according to the real part of their complex energy .
panel * a * shows the eigenvalues in units of @xmath44 , while in panel * b * the same eigenvalues are shifted by @xmath199 as defined in eq . .
parameters are : @xmath191 , @xmath42 , @xmath105.,title="fig:",scaledwidth=50.0% ] , modulo @xmath44 , associated with either a positive krein - signature @xmath192 or with @xmath188 and @xmath197 .
the system has a magnetic flux @xmath191 , with periodic boundary conditions ; driving and dissipation are not included .
the energy spectra are presented as a function of @xmath193 , for a fixed value of the amplitude modulation @xmath198 .
when all the eigenvalues at a particular frequency are stable , we denote them in black , otherwise we denote them either in red or blue .
red points are purely real modes , while blue points are the unstable modes , plotted according to the real part of their complex energy .
panel * a * shows the eigenvalues in units of @xmath44 , while in panel * b * the same eigenvalues are shifted by @xmath199 as defined in eq . .
parameters are : @xmath191 , @xmath42 , @xmath105.,title="fig:",scaledwidth=50.0% ] in fig .
[ fig : stability ] the floquet energy spectra @xmath176 of the system are shown as the bare frequency @xmath80 is changed for a fixed value of @xmath44 .
the magnetic flux is @xmath191 and periodic boundary conditions are used , so that the edge states do not appear in the spectra . with no loss of information , in the figure we only present stable eigenvalues associated with a positive krein - signature and unstable eigenvalues with @xmath197 . in panel
* a * , the eigenvalues are displayed modulo @xmath44 , while in panel * b * the same eigenvalues @xmath176 are shifted by @xmath199 as defined in eq . .
when all the eigenvalues at a particular frequency are stable , we denote them in black , otherwise we denote them either in red or blue .
the points in blue correspond to the modes which have become unstable , plotted here according to the real part of the energy .
the points in red are eigenmodes with purely real energies .
we refer to frequencies for which there are blue points as instability regions .
as anticipated above , we observe that the instability regions occur every integer and half - integer times of @xmath193 .
in particular , purely imaginary modes start to develop around @xmath200 and @xmath201 .
as @xmath193 is increased , the size of the instability regions along @xmath80 gets thinner and the deviation from @xmath179 correspondingly weaker .
still we find that the instabilities persist even for a very high ratio @xmath193 in the conservative case considered here .
however , once losses are included in the model , stability improves @xcite .
this explains why in the large @xmath80 calculations shown in the previous sections instabilities were only observed at very small values of the magnetic flux .
finally , in panel * b * , we show how the detailed structure of the floquet quasi - energy spectra changes as the ratio @xmath193 is varied . to allow a direct comparison of the spectra
, we have shifted each of these by the quantity @xmath199 as defined in eq . .
for large values of @xmath193 the spectra consists of @xmath66 bands in agreement with the hh model .
as @xmath193 is decreased , the band structure of the effective floquet hamiltonian remains initially qualitatively similar to the one of the hh hamiltonian with minor deviations due to the counter - rotating - wave terms .
as @xmath193 is reduced further , the deviations become more significant and the band - gaps eventually close .
we have numerically checked that the topological edge states are no longer present once the bands - gaps have closed around @xmath202 .
we conclude the work by commenting on possible experimental realizations of our proposal .
as already mentioned , even though our attention was focussed on a classical mechanical implementation based on coupled pendula , all our results hold for harmonic oscillators of any type . in all cases ,
the main challenge is to have a sufficiently high quality factor @xmath203 .
in fact , a high q is crucial to have enough freedom to choose the frequencies of the system to satisfy well the following inequality chain : @xmath204 these inequalities are required to be in the rotating wave limit , to find simple forms for the effective coupling amplitudes between neighbouring pendula , and to have clear access to the different spectroscopic features . as discussed around fig .
[ fig : system_x ] , the ratio @xmath193 also controls the number of sites @xmath46 that can be accommodated in a period of the spatial modulation . in particular , the sawtooth profile is needed for the scalability of the system of classical oscillators with a fixed quality factor @xmath205 .
this can be seen by instead considering a simple linear ramp of the bare frequency of the pendula for a system of length @xmath206 . in this case
, the frequency difference between pendula would range up to @xmath207 which , for a sufficiently large system , could exceed @xmath80 and thus violate eq . .
the sawtooth modulation with finite @xmath46 is instead scalable to any arbitrarily large system while keeping the bare frequency @xmath80 to satisfy eq . .
we note that the atomic systems of @xcite can use a linear ramp of an external potential because they do not suffer from these difficulties , since counter - rotating - wave terms are not present . for optical cavities , instead ,
the bare cavity frequency and the @xmath205 factor are typically large enough for this constraint to be practically irrelevant for realistic system sizes @xcite . while the spatial modulation of the oscillation frequency is rather straightforward to implement in pendula ,
a practical implementation of the temporal modulation needed for the floquet scheme may be obtained , as sketched in @xcite , by mounting a magnet on each pendulum and letting it interact with an externally - controlled time - dependent magnetic field .
quantitatively we need a minimum @xmath205-factor of around @xmath208 to observe chiral edge states in the scheme that we propose , before encountering the band - closing of fig .
[ fig : stability]-*b*. while reaching such a value of the q factor in coupled pendula systems requires a significant , though still not dramatic , improvement of recent experimental set - ups @xcite , it should be much more achievable using conventional rlc circuits . to go even further , the spectra of the hh model in fig .
[ fig : spectra ] and the hofstadter butterfly in fig .
[ fig : butterfly ] were obtained using higher q factors of @xmath209 and @xmath210 , respectively .
such high quality factors can be achieved in mechanical systems using micro - resonators as done in optomechanical oscillators @xcite .
in fact , recently there has been a related proposal of realizing topological models using such optomechanical systems in the limit of the rotating - wave - approximation @xcite .
in this work , we have theoretically proposed a scheme to realize non - trivial topological models in classical systems of coupled pendula or coupled lumped - element rlc circuits through a periodic time - modulation of system parameters . within a floquet framework
, we have shown that suitably - designed temporal modulations of the natural oscillation frequencies effectively produce the complex coupling amplitudes of the hh model , where the synthetic magnetic flux is easily tunable .
we have assessed the accuracy of the floquet picture by numerically solving the full equations of motion and we have shown that hallmarks of the hh model , such as the topologically - robust chiral edge states and the hofstadter butterfly pattern , can be observed in the excitations of the system .
furthermore , we have illustrated a protocol to extract the topological chern number of the bands in a classical counterpart of the integer quantum hall effect .
finally , we extended our investigation to the regime where the analogy with the quantum lattice model breaks down , investigating the crucial role played by the counter - rotating - wave terms in the stability of the system and the persistence of topological effects .
the exploration of topological effect in classical - mechanical systems is opening up many interesting new avenues of research .
the use of floquet modulation schemes to engineer topological models brings with it particular advantages .
firstly , time - modulation can naturally be used to break time - reversal symmetry and so to create edge states that are fully topologically protected .
secondly , the scheme discussed here is only one of many possible floquet modulation schemes that could be applied to arrays of classical oscillators ; other schemes may open up the way to build classical simulators of topological models with other interesting properties and with even more exotic topological invariants @xcite .
looking to the future , our proposal can be naturally extended to include a confining external potential , leading to novel oscillation features associated with analogue momentum - space magnetism @xcite . beyond the small oscillation regime
considered so far , systems of pendula also show nonlinear effects such as amplitude - dependent oscillation frequencies .
the interplay of such nonlinearities with topological features may lead to a variety of new and unexpected effects .
in this appendix , we give more details on the derivation of the hofstadter - like effective equation of motion in eq . .
from the magnus expansion applied to time - periodic hamiltonians at first order @xcite , one gets the effective hamiltonian as : @xmath211 for high modulation frequency .
the effective heisenberg equations that are derived from this effective hamiltonian are @xmath212=-{\mathrm{i}}\frac{1}{t}\int_0^t [ \beta_{i , j},{\hat{\mathcal{h}}}(t)]$ ] .
therefore , from the equations of motion eq . , substituting the transformed variables of eq . and integrating over a period of the temporal modulation ,
the following effective equations of motion are found : @xmath213 the couplings along the @xmath8 and @xmath9 direction get renormalised , as does the external driving force .
the expression for the effective coupling along @xmath9 is : @xmath214 where we have used the anger - jacobi expansion to obtain the bessel functions @xmath215 of @xmath216-th order with argument @xmath8 .
@xmath156 and @xmath217 are defined as in eq .
and eq . . the effective coupling along @xmath8 depends on the position along the static spatial modulation @xmath47
we first focus on the couplings between pendula that have a natural - frequency difference of @xmath218 , where the @xmath219 indicates that the hopping is calculated going towards the left ( @xmath145 ) or towards the right ( @xmath142 ) .
we have that : @xmath220 this equation describes the hopping along the small steps of the static spatial modulation .
the coupling between pendula that have a natural - frequency difference of @xmath221 is : @xmath222 where this time @xmath223 indicates the hopping calculated going towards the left ( @xmath142 ) or towards the right ( @xmath145 ) .
this is because the frequency - difference along the `` big step '' of the static modulation has an opposite sign to the ones along the `` small steps '' .
finally , the effective driving force : @xmath224 we now assume that @xmath225 , with @xmath226 and take only the largest term in the sums . in eq . , , and we have that @xmath227 , while in eq .
we must take @xmath228 .
remarkably , in this approximation , the hopping along @xmath8 is uniform and does not depend on the position along the static - spatial modulation , thus allowing for clear definitions of @xmath229 , @xmath230 and @xmath231 : @xmath232 by combining eq . with the effective equations of motion in eq .
to lowest order in @xmath156 and @xmath49 , we obtain exactly the equations in eq . . from eq
. , and we notice that the second largest term in the sums is proportional to @xmath233 , that is very small if @xmath234 or when considering a large period of the static modulation @xmath235 .
we found that an optimal combination of these two requirements that fulfils the inequality eq . with the renormalisation of eq .
is to have @xmath103 with @xmath42 .
we now comment on the fourier decomposition that was used as an alternative method to the full - time integration with runge - kutta for solving the set of differential equations in eq . to produce the butterfly spectra in fig .
[ fig : butterfly ] . from the properties of periodically driven systems @xcite , we know that , during a floquet evolution between two stroboscopic times @xmath236 , the system exhibits micro - motion , with a period set by the frequency of the temporal modulation @xmath44 .
we are interested in the steady state , therefore , we search for a solution of eq .
that oscillates at the frequency of the external driving @xmath25 . to this end , we expand the @xmath237 in a fourier series , using both @xmath25 and @xmath44 as harmonics : @xmath238 where @xmath239 are the time - independent fourier amplitudes , assuming that all the time - dependencies are in the exponential term . the sum over @xmath240
is truncated to a finite number @xmath123 , that is large enough to ensure a convergent solution .
the sum over @xmath241 takes only two values @xmath242 , since we have assumed that the external driving force is a cosine with a defined frequency @xmath25 . by substituting eq . in eq . and
isolating the component proportional to @xmath243 , we have a set of linear algebraic equations that can be inverted to find the coefficients @xmath239 . in order to simulate the real experimental situation
, we performed all the calculations using the full numerical integral of eq . , except for hofstadter butterfly where the full numerical integration would have been computationally demanding .
we have verified the good agreement between the two methods finding a mean error that is less than @xmath244 .
we note that the two methods are in fact equivalent when the system is dynamically stable .
however , when there is an instability , the system does not reach the steady state .
this is clearly seen in the full numerical integration method , where the solution shows the typical exponential growth of an unstable system . in the fourier method the steady state
is imposed by the decomposition eq .
itself and so the emergence of an instability can not be predicted , as discussed in the main text .
g. l. yu , r. v. gorbachev , j. s. tu , a. v. kretinin , y. cao , r. jalil , f. withers , l. a. ponomarenko , b. a. piot , m. potemski , d. c. elias , x. chen , k. watanabe , t. taniguchi , i. v. grigorieva , k. s. novoselov , v. i. falko , a. k. geim and a. mishchenko , _ hierarchy of hofstadter states and replica quantum hall ferromagnetism in graphene superlattices _ , nat
. phys . * 10 * , 525 ( 2014 ) . c. r. dean ,
l. wang , p. maher , c. forsythe , f. ghahari , y. gao , j. katoch , m. ishigami , p. moon , m. koshino , t. taniguchi , k. watanabe , k. l. shepard , j. hone and p. kim , _
hofstadter s butterfly and the fractal quantum hall effect in moir superlattices _ , nature * 497 * , 598 ( 2013 ) .
m. aidelsburger , m. atala , m. lohse , j. t. barreiro , b. paredes and i. bloch , _ realization of the hofstadter hamiltonian with ultracold atoms in optical lattices _ , phys .
lett . * 111 * , 185301 ( 2013 ) .
m. mancini , g. pagano , g. cappellini , l. livi , m. rider , j. catani , c. sias , p. zoller , m. inguscio , m. dalmonte and l. fallani , _ observation of chiral edge states with neutral fermions in synthetic hall ribbons _ , science * 349 * , 1510 ( 2015 ) .
j. struck , c. lschlger , m. weinberg , p. hauke , j. simonet , a. eckardt , m. lewenstein , k. sengstock and p. windpassinger , _ tunable gauge potential for neutral and spinless particles in driven optical lattices _ , phys .
* 108 * , 225304 ( 2012 ) .
m. aidelsburger , m. atala , s. nascimbne , s. trotzky , y .- a . chen and i. bloch ,
_ experimental realization of strong effective magnetic fields in an optical lattice _ , phys .
* 107 * , 255301 ( 2011 ) .
p. jain , a. s. bradley and c. w. gardiner , _ quantum de laval nozzle : stability and quantum dynamics of sonic horizons in a toroidally trapped bose gas containing a superflow _ , phys .
a * 76 * , 023617 ( 2007 ) . | we theoretically propose how to observe topological effects in a generic classical system of coupled harmonic oscillators , such as classical pendula or lumped - element electric circuits , whose oscillation frequency is modulated fast in time . making use of floquet theory in the high frequency limit , we identify a regime in which the system is accurately described by a harper - hofstadter model where the synthetic magnetic field can be externally tuned via the phase of the frequency - modulation of the different oscillators .
we illustrate how the topologically - protected chiral edge states , as well as the hofstadter butterfly of bulk bands , can be observed in the driven - dissipative steady state under a monochromatic drive . in analogy with the integer quantum hall effect , we show how the topological chern numbers of the bands can be extracted from the mean transverse shift of the steady - state oscillation amplitude distribution .
finally we discuss the regime where the analogy with the harper - hofstadter model breaks down . |
the physical mechanisms which govern the evolution of proto - planetary nebulae ( ppne ) are poorly documented because relatively few objects are known to be in this rapid transition between the agb and planetary nebula phase ( kwok 1993 ) .
the role played by high - velocity winds which interact with the slowly expanding agb envelope has been recognized as essential in the shaping of the planetary nebulae .
however , the details of the interaction and the precise evolution from the symmetric agb envelope to the asymmetries which characterize planetary nebulae are not well understood .
afgl 2688 ( the ` egg nebula ' ) is one of the prime examples of a ppn which has evolved from the agb phase about a hundred years ago ( jura & kroto 1990 ) .
the nearly circular slowly expanding agb envelope ( with a diameter of about @xmath1 ) is shocked by a warm , optically thin , fast wind ( e.g. , young et al .
studies at high spatial resolution ( e.g. , bieging & nguyen - quang - rieu 1996 ) have shown that the bulk of the molecular gas is concentrated in the central 4@xmath2 ( or @xmath3 cm for an adopted distance of 1 kpc ) , coincident with the dark lane seen in optical images , with weaker extensions along the north - south axis .
observations at near - infrared wavelengths have revealed that high - velocity h@xmath4 gas is present both in the north - south and east - west direction suggesting a quadrupolar outflow ( e.g. , cox et al .
1997 ) .
this paper presents high angular resolution observations ( @xmath5 ) made in the co j=21 line using the iram interferometer which provide a detailed picture of the morphology and the kinematics of the gas recently ejected by the central star in afgl 2688 .
the results of the observations are summarized in fig . 1 .
in the velocity integrated co map , the distribution of the co emission consists of a central core @xmath6 in diameter , with extensions in both the north - south and east - west directions . at negative ( approaching ) velocities ,
the co is detected in the northern and eastern extensions of the nebula .
along the north extension , velocities increase away from the centre ; the tip of the eastern arm is seen over a range of velocities from @xmath7 to @xmath8 ; and the highest velocity gas is found near the centre of the nebula up to @xmath9 . at positive ( receding ) velocities ,
the co gas is extended to the south and west , with a similar velocity structure to the blue - shifted gas , reversed about the systemic velocity .
the channel maps provide definite evidence of two distinct high - velocity outflow directions in afgl 2688 , one along a north - south axis at a p.a . of @xmath10 , and the other in a roughly orthogonal direction east - west .
detailed examination of the co data reveals that the two main outflows are resolved into a series of more collimated , bipolar outflows which are symmetric in direction and velocity about the center ( cox et al . 1999 ) and which are identified in fig . 1 : four collimated outflows in the east - west direction , and three in the north - south direction .
it is striking that , for most of these outflows , the tips correspond precisely with the h@xmath4 peaks seen in the hst image ( fig . 1 ) .
the only exceptions are in the central regions and to the west side of the equatorial plane , where the high - velocity red - shifted co gas ( located behind the dense central gas ) has no clear h@xmath4 counterpart in the hst image , most likely as a result of extinction in the near - infrared .
along all the outflow axes ( a to g ) , the co velocity increases with distance from the centre .
this implies that the observed co gas is entrained by high - velocity jets which could be atomic or ionized .
the velocity of the co gas in the outflows is also higher by a factor of @xmath112.5 than the velocity of the h@xmath4 which is close to the expansion velocity of the agb envelope ( kastner et al .
the high - velocity jets entrain the co gas and shock the molecular hydrogen of the agb envelope . assuming the model - dependent value of @xmath12 for the inclination of the north - south axis to the plane of the sky with the northern lobe towards the observer ( yusef - zadeh et al .
1984 ) , the deprojected flow velocity of f1 ( at @xmath13 from the centre ) is @xmath14 , corresponding to a kinematic age of @xmath15 years . in comparison ,
the projected velocities of the jets d and e are @xmath16 , implying ages of @xmath11250 and 125 years , if the same inclination is assumed .
finally , we note that the position of the central exciting star of afgl 2688 recently derived from polarisation measurements ( j. kastner , private communication ) lies within the positional errors at the intersection of the outflows .
in addition to the collimated outflows , the plateau de bure observations reveal in the central @xmath17 a shell - like structure expanding with a velocity of @xmath18(cox et al .
this co structure is not aligned with any of the outflow axes and lies at a p.a . of @xmath19 , which is comparable to that of the 1.3 mm dust continuum emission .
the size of this structure ( @xmath20 or @xmath21 ) implies that it was ejected @xmath11500 years ago and that it could trace the last episode of mass - loss on the agb ; the expansion velocity of this structure is much slower than that of the envelope ( @xmath22- see , e.g. , young et al .
neither the distribution of the dense gas at the center nor its kinematics indicate the presence of an equatorial disk in afgl 2688 .
the present data thus do not support the interpretation of a rotating equatorial disk to explain the east - west kinematics in afgl 2688 ( cf . bieging & nguyen - quang - rieu 1996 , kastner et al .
1999 ) . instead ,
the co kinematics reveal the presence of a central expanding shell - like structure ( not seen in h@xmath4 ) and high - velocity gas with a morphology similar to the shocked h@xmath4 gas tracing multiple , collimated outflows .
the observations discussed here reveal the detailed structure and kinematics of the molecular outflows in afgl 2688 .
a series of young ( a few 100 years ) , high velocity , collimated jets originate from the central star , and are directed along the north - south optical axis and in the east - west direction .
the detailed correlation of the co outflows with h@xmath4 emission seen in afgl 2688 provides direct evidence for their interaction with the nearly spherical envelope ejected on the agb .
the presence of multiple , collimated outflows in two , roughly orthogonal directions can not be explained by the standard two - winds model ( e.g. , balick 1987 ) .
proto - planetary nebulae with morphologies similar to afgl 2688 have recently been found ( e.g. , kwok et al . 1998 ) and multipolar jets appear to be a common phenomenon in young planetary nebulae ( forveille et al .
1998 , sahai & trauger 1998 ) .
whatever detailed processes are involved , the high - velocity winds must be intimately linked to the abrupt transition in the evolution of the star after the agb phase . one important consequence of the jets is their shaping effect on the molecular envelope ejected on the agb .
they generate complex point symmetries in the envelope which later emerge in the ionized gas during the pn phase .
balick b. 1987 , aj 94 , 671 bieging j.h . , nguyen - quang - rieu 1996 , aj 112 , 706 cox p. , maillard j .-
, huggins p.j .
1997 , aa 321 , 907 cox p. , et al .
1999 , aa ( submitted ) forveille t. , huggins p.j .
, bachiller r. , cox p. 1998 apj 495 , l111 jura m. , kroto h. 1990 apj 351 , 222 kastner j.h .
, henn l.a .
, weintraub d.a . , gatley i. 1999 , in _ agb stars _ , iau symp .
191 , t. le bertre , a. lebre , c. waelkens ( eds ) , asp conf .
series , p. 431
kwok s. 1993 , araa 31 , 63 kwok s. , su y.l.s . , hrivnak b.j .
1998 , apj 501 , l117 sahai r. , hines , d.c . ,
kastner , j.h .
et al . 1998 , apj 492 , l163 sahai r. , trauger , j.t .
1998 , aj 116 , 1357 young k. , serabyn g. , phillips t.g . ,
knapp g.r .
, gsten r. , schultz a. 1992 , apj 385 , 265 yusef - zadeh f. , morris m. , white r.l . 1984 , apj 278 , 186 | we present high resolution ( @xmath0 ) imaging of the proto - planetary nebula afgl 2688 in the co ( j=21 ) line using the iram interferometer .
the observations reveal with unprecedented detail the structure and the kinematics of the gas ejected by the star over the past few hundred years and exemplify the mechanism by which point symmetries are imprinted on the structure of planetary nebulae at early stages of their formation .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in |
motion estimation is a fundamental task for the analysis of spatiotemporal data , the prototypical example of which are sequences of images taken by a camera . the term _ optical flow _ has been coined to designate the apparent motion in such data . its accurate and efficient estimation has been a major topic in the fields of computer vision and image processing for more than 30 years . however , the applicability of optical flow algorithms is by no means limited to flat two - dimensional projections of real world scenes .
the advance of microscopy techniques has led to a particularly promising application of optical flow : cell motion analysis .
reliable optical flow algorithms supplied with microscopy images of sufficiently high spatial and temporal resolution can obviously help understanding cellular dynamics in transparent organisms , see for example @xcite .
the particular dataset we are working with in this article depicts a living zebrafish embryo during early embryogenesis .
main feature of this dataset are the embryo s endodermal cells , which have been labelled with a fluorescent protein and are known to develop on the surface of the zebrafish s spherical yolk , see fig .
[ fig : data3d ] and sec . [
sec : data ] .
the distribution of these cells can be modelled by a nonnegative function @xmath2 depending on time @xmath3 and position @xmath4 on the @xmath0-sphere , such that the number @xmath5 is directly proportional to the fluorescence response of a point @xmath4 at time @xmath3 .
the models we propose below are motivated , but surely not restricted , to this specific type of data .
we argue that the problem of extracting and analysing motion from spherical data is sufficiently general so as to be of potential interest to a wider audience .
degrees around the @xmath6-axis .
see sec .
[ sec : experiments ] for more details on the data and preprocessing.,title="fig:",scaledwidth=49.0% ] degrees around the @xmath6-axis .
see sec .
[ sec : experiments ] for more details on the data and preprocessing.,title="fig:",scaledwidth=49.0% ] + degrees around the @xmath6-axis .
see sec .
[ sec : experiments ] for more details on the data and preprocessing.,title="fig:",scaledwidth=49.0% ] degrees around the @xmath6-axis .
see sec .
[ sec : experiments ] for more details on the data and preprocessing.,title="fig:",scaledwidth=49.0% ] + our motion models are based on the optical flow equation @xmath7 which every vector field @xmath8 describing the temporal evolution of @xmath2 should ( approximately ) satisfy for all @xmath4 and @xmath3 . here
@xmath9 denotes the surface gradient on the sphere .
we derive this equation in more detail in sec . [
sec : of ] . directly solving the optical flow equation for @xmath8
is infeasible .
we therefore use tikhonov regularisation to compute an approximate solution to .
tikhonov regularisation consists in minimising a functional which is a weighted sum of two terms .
the first one , usually called data or similarity term , is the squared @xmath10 norm of the left hand side of .
the second term is a regularising functional @xmath11 , which in this article will always be a sobolev @xmath12 norm ( sec .
[ sec : reg ] ) .
these norms are introduced in sec .
[ sec : sobolev ] .
next , we extend the motion estimation method outlined above to two types of decomposition models that are also variational in nature ( sec . [
sec : decomposition ] ) .
while the input of those is again @xmath2 , their outputs are now two or more vector fields capturing different structural parts of the total motion .
both models are adaptations of recently proposed image decomposition techniques to the optical flow setting .
the first one is a @xmath1 decomposition .
its idea is to replace @xmath8 in the data term with a sum @xmath1 , and then to add two different regularising functionals , one for @xmath8 and one for @xmath13 .
the second one is a hierarchical model . roughly speaking
, one repeatedly minimises an optical flow functional , while the amount of regularisation is constantly decreased . in every iteration ,
the @xmath8 in the data term is replaced by a sum @xmath14 , where only @xmath8 is optimised and the @xmath15 are the results from previous iterations .
finally , all optimisation problems are solved by projecting them onto finite dimensional spaces spanned by vector spherical harmonics ( sec .
[ sec : numerics ] ) .
one advantage of this method is that it automatically yields helmholtz decompositions of all computed vector fields . mathematical background on ( vector )
spherical harmonics is presented in secs .
[ sec : sh ] and [ sec : vsh ] . in sec .
[ sec : experiments ] we provide details of our implementation , give a more detailed account of the used microscopy data , and show experimental results with these data .
to summarise , the main novelties presented in this article are decomposition models for optical flow on the sphere together with their application to microscopy data . the first variational optical flow method is usually attributed to @xcite , where they used an @xmath16 seminorm for regularisation . in @xcite
it was shown that this particular choice leads to a well - posed problem .
we refer to @xcite for a gentle introduction to optical flow , to @xcite for an overview of different optical flow functionals and to @xcite for a recent survey and benchmark .
optical flow algorithms have only recently been extended to data defined on non - euclidean domains . in @xcite images defined on the sphere were treated , whereas in @xcite the original functional by horn and schunck was generalised to @xmath0-riemannian manifolds and well - posedness was verified .
most recently , optical flow on evolving manifolds has been considered in @xcite .
horn and schunck @xcite numerically solved the variational problem by applying a finite difference scheme to the euler - lagrange equations .
a similar approach was adopted in @xcite after parametrisation of the surface . in @xcite , however , the problem was solved by finite elements on a surface triangulation .
finally , we mention the work by schuster and weickert @xcite , where they used projection methods to solve the optical flow equation in the plane . instead of tikhonov - regularising their solution , they solely relied on regularisation by discretisation .
the main reason for choosing a numerical method based on vector spherical harmonics in this article is that @xmath12-type regularisers , for arbitrary real @xmath17 , are handled very easily in contrast to most other methods .
the aim of @xmath1 image decomposition models , as pioneered in @xcite , is to separate the cartoon and texture parts of images . while the cartoon component should capture large - scale structural components and
should therefore be piecewise smooth , the texture component is supposed to consist of high - frequency oscillating patterns . since the original model was promising but hard to implement , a large number of modifications and approximations have been proposed . in some of them
the problematic @xmath18-norm was approximated by an @xmath19 norm @xcite .
recently , @xmath1 models have been extended to the @xmath20 optical flow setting @xcite .
hierarchical models , originally introduced in @xcite for image analysis , have not yet been tried in combination with optical flow .
they have , however , the preferable property of producing arbitrarily fine multiscale descriptions of input data . as a concluding remark about vector field decompositions ,
let us remark that helmholtz - hodge decompositions of motion fields have enjoyed a certain degree of attention in recent years , not only in the plane @xcite , but also on surfaces @xcite .
applying optical flow algorithms to cell microscopy data has become increasingly popular lately .
see for example @xcite and the references therein .
we highlight the article @xcite , where also endodermal cells of zebrafish embryos have been analysed .
there the authors point out that , although of immense importance for developmental biology , only little is known about the motion behaviour of this type of cells .
let @xmath21 be the two - sphere embedded in @xmath22 . for functions @xmath23 we define the laplace - beltrami operator by @xmath24 where @xmath25 is the usual laplacian of @xmath22 and @xmath26 is the radially constant extension of @xmath2 to @xmath27 .
the eigenvalues of @xmath28 are @xmath29 the corresponding eigenspaces @xmath30 have dimension @xmath31 and are mutually orthogonal in @xmath32 .
their direct sum equals @xmath32 .
every eigenfunction @xmath33 lies in @xmath34 and is called _ ( scalar ) spherical harmonic of degree _ @xmath35 .
their name derives from the equivalent characterisation of @xmath30 as the restriction to @xmath36 of the space of harmonic polynomials @xmath37 that are homogeneous of degree @xmath35 . from now on @xmath38 always refers to a particular orthonormal basis of @xmath32 consisting of real - valued spherical harmonics . for numerical experiments we use the so - called fully normalised spherical harmonics ,
see ( * ? ? ?
5.2 ) for a detailed construction .
denote by @xmath39 the outward unit normal to @xmath36 and let @xmath40 be the surface gradient of @xmath23 .
[ def : vsh ] let @xmath41 and @xmath33 . whenever a function @xmath42 that does not vanish identically admits one of the following three representations @xmath43
then @xmath44 is called a _ vector spherical harmonic of degree @xmath35 and type @xmath45_. for obvious reasons we refer to types @xmath0 and @xmath46 as _ tangential vector spherical harmonics_. note that there is no tangential spherical harmonic of degree @xmath47 . .
scalar spherical harmonics are depicted using a blue to red colour bar .
type @xmath0 vector spherical harmonics are visualised with red arrows , type @xmath46 ones with blue arrows .
note that the length of the vectors has been scaled for better illustration.,title="fig:",scaledwidth=49.0% ] .
scalar spherical harmonics are depicted using a blue to red colour bar .
type @xmath0 vector spherical harmonics are visualised with red arrows , type @xmath46 ones with blue arrows .
note that the length of the vectors has been scaled for better illustration.,title="fig:",scaledwidth=49.0% ] + .
scalar spherical harmonics are depicted using a blue to red colour bar .
type @xmath0 vector spherical harmonics are visualised with red arrows , type @xmath46 ones with blue arrows .
note that the length of the vectors has been scaled for better illustration.,title="fig:",scaledwidth=49.0% ] .
scalar spherical harmonics are depicted using a blue to red colour bar .
type @xmath0 vector spherical harmonics are visualised with red arrows , type @xmath46 ones with blue arrows .
note that the length of the vectors has been scaled for better illustration.,title="fig:",scaledwidth=49.0% ] + .
scalar spherical harmonics are depicted using a blue to red colour bar .
type @xmath0 vector spherical harmonics are visualised with red arrows , type @xmath46 ones with blue arrows .
note that the length of the vectors has been scaled for better illustration.,title="fig:",scaledwidth=65.0% ] we are mainly interested in the space @xmath48 of square integrable tangent vector fields on @xmath36 endowed with the inner product @xmath49 where @xmath50 is the usual surface measure on the sphere .
an orthonormal basis of this space is obtained from by setting @xmath51 for all @xmath52 and @xmath53 . in fig .
[ fig : vspharm ] a handful of the elements of both bases and are depicted .
every @xmath54 has the following fourier series representation @xmath55 in particular , we have parseval s identity @xmath56 for a comprehensive and unified treatment of both scalar and vector spherical harmonics we refer to @xcite . for an arbitrary real number @xmath17 , the space @xmath57 is commonly defined as the domain of @xmath58 .
see @xcite or ( * ? ? ?
6.2 ) for example . in this section
we introduce the spaces @xmath59 by means of the vectorial counterpart of @xmath60 . for tangent vector fields
@xmath13 we define the laplace - beltrami operator by @xmath61 where application of @xmath60 to @xmath13 is understood componentwise and @xmath62 is the orthogonal projector onto the tangent plane @xmath63 , compare ( * ? ? ?
* def . 5.26 ) .
the tangential vector spherical harmonics introduced in def . [ def : vsh ] are eigenfunctions of this operator to the same eigenvalues as their scalar counterparts : if we let @xmath64 then @xmath65 for every @xmath66 .
the @xmath67 are as defined in , the only difference being that now @xmath68 and therefore the spectrum of @xmath69 is strictly positive . applying functional calculus , we formally define the @xmath17-th power of @xmath69 by @xmath70 finally , for every @xmath71 , set @xmath72 note that , in contrast to the scalar setting , this functional is an actual norm , rather than only a seminorm .
we therefore define , for every real @xmath17 , @xmath59 as the space of all distributions @xmath73 for which the series in is finite . clearly , if @xmath74 is any sequence satisfying @xmath75 for two positive constants @xmath76 , @xmath77 and for all @xmath35 , then replacing @xmath78 with @xmath79 in leads to an equivalent norm and thus to the same space . for every sequence of positive numbers
@xmath79 we denote the resulting norm simply by @xmath80 .
let @xmath81 \subset \mathbb{r}$ ] be a time interval . we assume to be given a scalar time - varying ( brightness ) function @xmath82 the problem of estimating optical flow
consists in tracking the temporal evolution of the data @xmath2 by means of a time - dependent vector field .
our optical flow model is based on the so - called _ brightness constancy assumption _ : we assume existence of a function @xmath83 satisfying @xmath84 for all @xmath4 and @xmath3 . intuitively this means that for every starting point @xmath4 on the sphere , the function @xmath2 remains constant along the trajectory @xmath85 .
in addition we require that @xmath86 is a diffeomorphism of @xmath36 for every @xmath3 .
the first equation in implies that @xmath87 this equation is typically written in terms of the vector field @xmath88 whose integral curves are the trajectories @xmath85 , which is defined by the equation @xmath89 . the resulting _ optical flow equation
_ reads @xmath90 solving the optical flow equation directly is problematic . in general , a solution to need not exist , and if it exists , it can not be unique .
the typical remedy is tikhonov regularisation , where one minimises a functional of the form @xmath91 with @xmath92 being a regularising functional that incorporates a - priori knowledge about desirable solutions .
the parameter @xmath93 controls the amount of regularisation . in the context of optical flow one
usually tries to enforce spatial ( and temporal ) smoothness on the solution .
a natural candidate for @xmath92 would then be the squared sobolev @xmath94 ( semi-)norm , which penalises first derivatives in space and time equally , compare @xcite another , and in fact more popular , possibility is to drop time regularisation , in which case minimisation of is equivalent to minimising @xmath95 for each instant @xmath3 separately .
this corresponds to the original approach of horn and schunck @xcite .
from now on we denote the above data term by @xmath96 . instead of we consider the more general class of optical flow functionals
@xmath97 the regularisation parameter is omitted , as it can be considered a constant factor in the sequence @xmath74 .
functional forms the basic optical flow setting of this article .
all variational models considered here are extensions of .
the following two decomposition models are inspired by techniques that have recently been developed in the context of image analysis .
the fact that motion estimation based on can be viewed as denoising of vector - valued images suggests the translation of said image decomposition models to the optical flow setting @xcite .
the aim is now not to extract one , but two vector fields @xmath8 and @xmath13 in such a way that they capture different structural parts of the total motion @xmath98 of @xmath2 .
the idea is to solve the following variational problem @xmath99 choosing , for instance , the two regularisers to be @xmath100 and @xmath101 norms , respectively , would lead to a model which , in spirit , comes closest to the image decomposition model considered in @xcite . generalising to a decomposition into @xmath102 , instead of two , constituents is possible , but will not be considered here @xcite .
hierarchical image decomposition models have been introduced in @xcite . there
, an original image is decomposed by repeatedly applying denoising steps .
the input of one such step is the residual of the previous one . in every step the degree of regularisation
is decreased .
in contrast to @xmath1 models , hierarchical decomposition models provide multiscale descriptions of the input data . this iterative procedure can be transferred to the optical flow setting as follows .
let @xmath103 , @xmath104 , be a sequence of norms as defined in section [ sec : sobolev ] , such that @xmath105 for all @xmath35 and @xmath106 . for every such sequence of sequences
we propose the following iterative scheme , @xmath107 the resulting sequence of accumulated solutions @xmath108 provides a multiscale representation which , with an appropriate choice of sequences @xmath109 , can be made arbitrarily fine .
the hierarchical model as formulated above is a slight generalisation of the originally proposed one , in the sense that the sequences of regularising functionals considered in @xcite are always of the form @xmath110 .
that is , the regularising functional of step @xmath106 is the same as the one of previous steps save for a smaller regularisation parameter .
requirement allows a more general setup .
the helmholtz decomposition theorem states that every continuously differentiable tangent vector field @xmath111 on the sphere can be uniquely represented as the sum of a consoidal ( curl - free ) and a toroidal ( divergence - free ) vector field . more precisely , there exist two uniquely determined tangent vector fields @xmath112 and @xmath113 satisfying @xmath114 see ( * ? ? ?
5.3 ) , for example .
the projection method we use to numerically solve the variational problems presented above leads to solutions @xmath111 that are finite fourier sums @xmath115 where @xmath116 .
now , from the definition of basis , and the fact that @xmath117 for all sufficiently smooth functions @xmath118 , the helmholtz decomposition of @xmath111 is obtained immediately @xmath119
in the first subsection below we describe the numerical optimisation of the optical flow functional based on the optical flow equation and explain the modifications necessary for the hierarchical decomposition .
subsection [ sec : u+v_num ] is dedicated to the @xmath1 decomposition model .
finally , we explain how the resulting spherical integrals are approximated ( sec .
[ sec : discretisation ] ) . for convenience
we relabel the orthonormal basis using a single index @xmath120 and write , for example , @xmath121 from now on .
let @xmath122 and let @xmath74 be a sequence comparable to @xmath123 in the sense of .
then , a minimiser of @xmath124 , if it exists , has to be in @xmath59 .
we solve the problem of finding @xmath125 by a projection method .
that is , we let @xmath8 range only over a finite - dimensional subspace @xmath126 of @xmath59 , where @xmath127 and @xmath128 is a finite index set .
the unknown vector field can now be written as @xmath129 and the problem of finding an optimal @xmath130 simplifies to a minimisation problem over @xmath131 . plugging into the optical flow functional gives
@xmath132 which is minimal , if the optimality conditions @xmath133 , for all @xmath134 , are satisfied .
they read @xmath135 or in matrix - vector form @xmath136 where @xmath137 is the vector of unknown coefficients , the elements of matrix @xmath138 read @xmath139 @xmath140 is a diagonal matrix that corresponds to the regularisation term and the right hand side @xmath141 is given by @xmath142 with a slight abuse of notation we identified the set @xmath143 with @xmath144 in the definitions of @xmath145 .
we continue to do so below .
the hierarchical model only needs a minor modification for the case @xmath146 .
we can rewrite the data term from as @xmath147 where @xmath148 .
therefore , only the right hand side of the optimality system has to be updated in every step .
for simplicity we can assume that the approximation space @xmath126 is the same in every step , so that @xmath149 has the representation @xmath150 , where the coefficients @xmath151 are already known from previous steps .
letting @xmath152 denote the right hand side of the optimality system for step @xmath106 , we calculate @xmath153 or simply @xmath154 the projection approach explained above is easily adapted to the @xmath1 decomposition problem . again , let @xmath155 be real numbers and choose two sequences @xmath74 , @xmath156 so that @xmath157 is a norm for @xmath158 and @xmath159 is one for @xmath59 .
now , we solve @xmath160 where @xmath161 are finite dimensional spaces . proceeding as in the previous section ,
we obtain the following optimality conditions @xmath162 where the coefficients @xmath163 and @xmath164 are as defined in and , respectively .
concatenating the two coefficient vectors @xmath165 and @xmath166 into a single vector @xmath167 , so that the @xmath168 occupy the first @xmath169 entries while the @xmath170 occupy the last @xmath171 entries , the linear system reads @xmath172 the matrix @xmath173 is given by @xmath174 where @xmath175 and @xmath176 is concatenated from two versions of @xmath177 in the same way as @xmath111 .
it remains to discuss the numerical evaluation of the integrals , .
first , we approximate the @xmath0-sphere @xmath36 with a polyhedron @xmath178 defined by a set @xmath179 of vertices and a set @xmath180 of triangular faces .
each triangle @xmath181 is associated with a tuple @xmath182 identifying the corresponding vertices @xmath183 . how the triangulated sphere is obtained in practice ,
is explained in sec .
[ sec : acquisition ] .
second , in every experiment data @xmath2 are given only at the vertices and for two time steps @xmath184 and @xmath185 .
we set @xmath186 and @xmath187 and extend both functions to all of @xmath188 by linear interpolation on every triangle .
thus we obtain two continuous piecewise linear functions @xmath189 , @xmath190 .
the time derivative of @xmath2 is approximated by a simple forward difference @xmath191 which is again piecewise linear on @xmath188 .
the surface gradient @xmath192 is replaced by a vector field @xmath193 that is constant on every triangle .
it is given by @xmath194 where @xmath195 is the height vector of the triangle @xmath196 pointing from vertex @xmath197 to the opposite side , compare ( * ? ? ?
3.3.3 ) .
finally , we also replace the fully normalised scalar spherical harmonics @xmath198 by their piecewise linear approximations @xmath199 defined on @xmath188 . as before we obtain piecewise constant approximations @xmath200 of @xmath201 .
the resulting approximated integrals read @xmath202 where @xmath203 denotes the area of @xmath204 , and @xmath205
the data which motivated the study of the proposed decomposition models are time - lapse volumetric ( 4d ) images .
the obtained sequence depicts a live zebrafish embryo during the gastrula period , approximately five to ten hours after fertilisation . with the help of confocal laser - scanning microscopy ,
endoderm cells expressing a green fluorescence protein were recorded separately from the background . for details on the imaging techniques and the fluorescence marker we refer to @xcite and @xcite , respectively .
the sequence obtained by the microscope captures a cuboid region of approximately @xmath206 .
the spatial resolution is @xmath207 voxels and the intensity is in the range @xmath208 .
a total number of 77 images were taken , one every @xmath209 . in the following
, the microscopic data will be denoted by @xmath210 during this early stage , endodermal cell proliferation is known to take place on a so - called monolayer @xcite . in other words , cells move and divide without stacking and , as the yolk is ball - shaped , admit for the extraction of a spherical image sequence . for further explanations and numerous illustrations of the developmental process of zebrafish embryos
we refer to @xcite .
we extracted spherical images from the dataset by first fitting a sphere to the approximate cell centres in each pair of consecutive frames .
for simplicity we restrict our attention to one such pair of frames which we denote by @xmath211 .
cell centres are typically characterised by local maxima in intensity and can be found by applying a gaussian filter and simple thresholding . without loss of generality the radius of the fitted sphere
is assumed to be @xmath212 . in a second step ,
we created a point grid @xmath213 starting from an icosahedron inscribed in the sphere . in each iteration
every triangular face is split into four sub - triangles by connecting the edge midpoints with each other and projecting them onto the sphere .
thus , the total number of faces is @xmath214 , where @xmath106 is the number of refinements . in our experiments we found that @xmath215 iterations suffice . in order to project the volumetric time - lapse data @xmath216 , onto the grid @xmath217
, we define @xmath218 } \bar{f}_j^{\delta}(c v_{i}),\ ] ] for the said pair of consecutive frames @xmath219 . here
@xmath220 is a piecewise linear extension to @xmath22 of @xmath216 and @xmath221 is sufficiently large .
deviations of the monolayer from a perfect sphere are thereby corrected .
the obtained data are subsequently scaled to the range @xmath222 $ ] .
note that , in contrast to our previous work @xcite , here we consider the unfiltered microscopy data for optical flow estimation .
the support of the obtained data is contained in the northern hemisphere .
thus , it suffices to consider only half of the triangulation leading to a total number of around 164000 faces .
figure [ fig : data3d ] depicts a sample of two frames @xmath223 . in fig .
[ fig : data2d ] , a top view of the two frames is shown . in order to visualise our results
we will apply the standard flow colour - coding @xcite using a colour disk .
figure [ fig : colourcoding ] ( rightmost image ) depicts this colour space .
each vector is assigned a colour determined by its angle and length . however , this colour - coding is defined for planar vector fields only . as a possible remedy
we suggest to first project tangential velocities to the plane and then correct the length . to this end , let @xmath224 be the orthogonal projector of @xmath22 onto the @xmath225-@xmath226-plane . accordingly , given a tangent vector field @xmath13 , the planar vector field which we visualise is @xmath227 this construction is chosen so that it preserves the length of @xmath13 .
this additional rescaling is different to @xcite .
the resulting colour image is finally mapped back onto the hemisphere .
[ fig : colourcoding ] shows a tangent vector field visualised with the proposed approach . from now on we will visualise velocity fields only in top view , as in the right hand side of fig .
[ fig : colourcoding ] .
in addition , for every figure the colour disk s radius @xmath228 was chosen to be equal to the length of the longest vector under consideration .
specific values of @xmath228 are given in table [ tab : radii ] . .radii
@xmath228 of the colour disks used in the different experiments below . [ cols="<,<,<,<,<,<,<,<,<,<,<",options="header " , ] as a second way of illustrating steady velocity fields , we employ streamlines , see e.g. @xcite . in all our experiments we consider time as fixed and compute the optical flow @xmath13 for one pair of frames , cf .
[ sec : reg ] . given a tangential velocity field @xmath13 and a starting point @xmath229 ,
a streamline @xmath230 on @xmath36 is the solution to the ordinary differential equation @xmath231 numerically , we approximated by solving @xmath232 where @xmath233 is a step size , for a number of approximately 1300 initial points @xmath234 and @xmath235 iterations .
the step size was chosen as @xmath236 .
this use of integral curves is different from @xcite , where we computed approximate cell trajectories in a nonsteady velocity field .
the visualisation by means of the colour coding is rich in detail and is even capable of indicating individual cell motion .
nevertheless , it fails to deliver intuition about the helmholtz decomposition .
streamlines provide the anticipated effect .
we performed numerous experiments and minimised functionals , , and as outlined in sec .
[ sec : numerics ] for the two frames shown in fig .
[ fig : data3d ] and fig .
[ fig : data2d ] . in all experiments the finite - dimensional spaces introduced in sec .
[ sec : numerics ] were chosen as @xmath237 all resulting linear systems were solved using the generalized minimal residual method ( gmres ) on an intel xeon e5 - 1620 @xmath238 workstation with @xmath239 ram .
solutions converged to a relative residual of 0.02 within 100 iterations .
the runtime was governed by the evaluation of the integrals , cf .
[ sec : discretisation ] , and amounts to approximately five hours for the chosen bases and the chosen triangulation .
nevertheless , once the integrals are computed they can be used in all of the proposed models and the linear systems can be solved in a few seconds for different parameters and different norms .
our matlab implementation and the data are available on our website .
( top ) , with @xmath240 , @xmath241 , and increasing values @xmath242 , @xmath243 , and @xmath244 from left to right .
the middle row depicts the curl - free component whereas the bottom row depicts the divergence - free component .
the embryo s body axis roughly runs from bottom left to top right in all images.,title="fig:",scaledwidth=32.0% ] ( top ) , with @xmath240 , @xmath241 , and increasing values @xmath242 , @xmath243 , and @xmath244 from left to right .
the middle row depicts the curl - free component whereas the bottom row depicts the divergence - free component .
the embryo s body axis roughly runs from bottom left to top right in all images.,title="fig:",scaledwidth=32.0% ] ( top ) , with @xmath240 , @xmath241 , and increasing values @xmath242 , @xmath243 , and @xmath244 from left to right .
the middle row depicts the curl - free component whereas the bottom row depicts the divergence - free component .
the embryo s body axis roughly runs from bottom left to top right in all images.,title="fig:",scaledwidth=32.0% ] + ( top ) , with @xmath240 , @xmath241 , and increasing values @xmath242 , @xmath243 , and @xmath244 from left to right .
the middle row depicts the curl - free component whereas the bottom row depicts the divergence - free component .
the embryo s body axis roughly runs from bottom left to top right in all images.,title="fig:",scaledwidth=32.0% ] ( top ) , with @xmath240 , @xmath241 , and increasing values @xmath242 , @xmath243 , and @xmath244 from left to right .
the middle row depicts the curl - free component whereas the bottom row depicts the divergence - free component .
the embryo s body axis roughly runs from bottom left to top right in all images.,title="fig:",scaledwidth=32.0% ] ( top ) , with @xmath240 , @xmath241 , and increasing values @xmath242 , @xmath243 , and @xmath244 from left to right .
the middle row depicts the curl - free component whereas the bottom row depicts the divergence - free component .
the embryo s body axis roughly runs from bottom left to top right in all images.,title="fig:",scaledwidth=32.0% ] + ( top ) , with @xmath240 , @xmath241 , and increasing values @xmath242 , @xmath243 , and @xmath244 from left to right .
the middle row depicts the curl - free component whereas the bottom row depicts the divergence - free component .
the embryo s body axis roughly runs from bottom left to top right in all images.,title="fig:",scaledwidth=32.0% ] ( top ) , with @xmath240 , @xmath241 , and increasing values @xmath242 , @xmath243 , and @xmath244 from left to right .
the middle row depicts the curl - free component whereas the bottom row depicts the divergence - free component .
the embryo s body axis roughly runs from bottom left to top right in all images.,title="fig:",scaledwidth=32.0% ] ( top ) , with @xmath240 , @xmath241 , and increasing values @xmath242 , @xmath243 , and @xmath244 from left to right .
the middle row depicts the curl - free component whereas the bottom row depicts the divergence - free component .
the embryo s body axis roughly runs from bottom left to top right in all images.,title="fig:",scaledwidth=32.0% ] . with increasing parameter @xmath245 the streamlines change colours from yellow ( bright ) to green ( dark ) .
the top row shows the total motion whereas middle and bottom rows depict the curl - free and divergence - free parts of the helmholtz decomposition , respectively .
images are arranged in accordance with fig .
[ fig : of1].,title="fig:",scaledwidth=32.0% ] . with increasing parameter @xmath245 the streamlines change colours from yellow ( bright ) to green ( dark ) .
the top row shows the total motion whereas middle and bottom rows depict the curl - free and divergence - free parts of the helmholtz decomposition , respectively .
images are arranged in accordance with fig .
[ fig : of1].,title="fig:",scaledwidth=32.0% ] . with increasing parameter @xmath245 the streamlines change colours from yellow ( bright ) to green ( dark ) .
the top row shows the total motion whereas middle and bottom rows depict the curl - free and divergence - free parts of the helmholtz decomposition , respectively .
images are arranged in accordance with fig .
[ fig : of1].,title="fig:",scaledwidth=32.0% ] + . with increasing parameter @xmath245 the streamlines change colours from yellow ( bright ) to green ( dark ) .
the top row shows the total motion whereas middle and bottom rows depict the curl - free and divergence - free parts of the helmholtz decomposition , respectively .
images are arranged in accordance with fig .
[ fig : of1].,title="fig:",scaledwidth=32.0% ] . with increasing parameter @xmath245 the streamlines change colours from yellow ( bright ) to green ( dark ) .
the top row shows the total motion whereas middle and bottom rows depict the curl - free and divergence - free parts of the helmholtz decomposition , respectively .
images are arranged in accordance with fig .
[ fig : of1].,title="fig:",scaledwidth=32.0% ] . with increasing parameter @xmath245 the streamlines change colours from yellow ( bright ) to green ( dark ) .
the top row shows the total motion whereas middle and bottom rows depict the curl - free and divergence - free parts of the helmholtz decomposition , respectively .
images are arranged in accordance with fig .
[ fig : of1].,title="fig:",scaledwidth=32.0% ] + . with increasing parameter @xmath245 the streamlines change colours from yellow ( bright ) to green ( dark ) .
the top row shows the total motion whereas middle and bottom rows depict the curl - free and divergence - free parts of the helmholtz decomposition , respectively .
images are arranged in accordance with fig .
[ fig : of1].,title="fig:",scaledwidth=32.0% ] . with increasing parameter @xmath245 the streamlines change colours from yellow ( bright ) to green ( dark ) .
the top row shows the total motion whereas middle and bottom rows depict the curl - free and divergence - free parts of the helmholtz decomposition , respectively .
images are arranged in accordance with fig .
[ fig : of1].,title="fig:",scaledwidth=32.0% ] . with increasing parameter @xmath245 the streamlines change colours from yellow ( bright ) to green ( dark ) .
the top row shows the total motion whereas middle and bottom rows depict the curl - free and divergence - free parts of the helmholtz decomposition , respectively .
images are arranged in accordance with fig .
[ fig : of1].,title="fig:",scaledwidth=32.0% ] in the first experiment , we minimised functional @xmath124 as defined in for @xmath240 and different values of @xmath17 and @xmath246 . figure [ fig : of1 ] , top row , depicts the optical flow field for @xmath241 and values @xmath242 , @xmath243 , and @xmath244 .
the presented results are in accordance with our findings in @xcite . as explained in sec .
[ sec : decomposition ] , a helmholtz decomposition @xmath247 is obtained immediately .
figure [ fig : of1 ] , middle row , shows @xmath112 whereas fig .
[ fig : of1 ] , bottom row , shows @xmath113 . furthermore , in fig .
[ fig : of1stream ] , streamlines for the same velocity fields are portrayed and the individual plots are arranged accordingly . in addition , fig . [
fig : of2 ] shows the velocity fields for parameters @xmath248 , @xmath243 , @xmath249 , and @xmath250 .
are shown in the top row .
parameters were chosen as @xmath248 and @xmath243 , @xmath249 , and @xmath250 and are arranged from left to right .
the bottom row depicts the corresponding streamlines.,title="fig:",scaledwidth=32.0% ] are shown in the top row .
parameters were chosen as @xmath248 and @xmath243 , @xmath249 , and @xmath250 and are arranged from left to right .
the bottom row depicts the corresponding streamlines.,title="fig:",scaledwidth=32.0% ] are shown in the top row .
parameters were chosen as @xmath248 and @xmath243 , @xmath249 , and @xmath250 and are arranged from left to right .
the bottom row depicts the corresponding streamlines.,title="fig:",scaledwidth=32.0% ] + are shown in the top row .
parameters were chosen as @xmath248 and @xmath243 , @xmath249 , and @xmath250 and are arranged from left to right .
the bottom row depicts the corresponding streamlines.,title="fig:",scaledwidth=32.0% ] are shown in the top row .
parameters were chosen as @xmath248 and @xmath243 , @xmath249 , and @xmath250 and are arranged from left to right .
the bottom row depicts the corresponding streamlines.,title="fig:",scaledwidth=32.0% ] are shown in the top row .
parameters were chosen as @xmath248 and @xmath243 , @xmath249 , and @xmath250 and are arranged from left to right .
the bottom row depicts the corresponding streamlines.,title="fig:",scaledwidth=32.0% ] in a next experiment , we computed a minimiser for functional in order to obtain a @xmath98 decomposition of the optical flow .
the sequences @xmath251 and @xmath252 were chosen as @xmath253 and @xmath254 , respectively . in figs .
[ fig : u+v ] and [ fig : u+v2 ] , the resulting decomposition is shown for two different parameter settings .
the motion field in fig .
[ fig : u+v ] was obtained with parameters @xmath255 , @xmath256 , @xmath257 , and @xmath258 . as anticipated ,
@xmath8 and @xmath13 capture different structural parts of the motion .
@xmath8 is sufficiently smooth whereas @xmath13 contains spatial oscillations .
the result in fig .
[ fig : u+v2 ] was computed by setting @xmath259 , @xmath256 , @xmath260 , and @xmath261 .
expectedly , the velocity field @xmath8 is , by choice of @xmath262 , smoother than in the previous setting .
in addition , fig .
[ fig : detail ] illustrates the characteristics of the velocity fields during a cell division in more detail .
while @xmath8 is smooth , @xmath13 clearly indicates the cell division .
( left ) into structural parts @xmath8 ( middle ) and @xmath13 ( right ) .
sequences were chosen as @xmath253 and @xmath254 and parameters were set to @xmath255 , @xmath256 , @xmath257 , and @xmath258.,title="fig:",scaledwidth=32.0% ] ( left ) into structural parts @xmath8 ( middle ) and @xmath13 ( right ) .
sequences were chosen as @xmath253 and @xmath254 and parameters were set to @xmath255 , @xmath256 , @xmath257 , and @xmath258.,title="fig:",scaledwidth=32.0% ] ( left ) into structural parts @xmath8 ( middle ) and @xmath13 ( right ) .
sequences were chosen as @xmath253 and @xmath254 and parameters were set to @xmath255 , @xmath256 , @xmath257 , and @xmath258.,title="fig:",scaledwidth=32.0% ] ( left ) into structural parts @xmath8 ( middle ) and @xmath13 ( right ) .
sequences were chosen as @xmath253 and @xmath254 and parameters were set to @xmath259 , @xmath256 , @xmath260 , and @xmath261.,title="fig:",scaledwidth=32.0% ] ( left ) into structural parts @xmath8 ( middle ) and @xmath13 ( right ) .
sequences were chosen as @xmath253 and @xmath254 and parameters were set to @xmath259 , @xmath256 , @xmath260 , and @xmath261.,title="fig:",scaledwidth=32.0% ] ( left ) into structural parts @xmath8 ( middle ) and @xmath13 ( right ) .
sequences were chosen as @xmath253 and @xmath254 and parameters were set to @xmath259 , @xmath256 , @xmath260 , and @xmath261.,title="fig:",scaledwidth=32.0% ] + of velocity fields obtained by the hierarchical decomposition .
images are aligned from left to right and top to bottom . at iteration @xmath106 the sequence
was set to @xmath263 with parameters @xmath241 and @xmath264.,title="fig:",scaledwidth=32.0% ] of velocity fields obtained by the hierarchical decomposition .
images are aligned from left to right and top to bottom .
at iteration @xmath106 the sequence was set to @xmath263 with parameters @xmath241 and @xmath264.,title="fig:",scaledwidth=32.0% ] of velocity fields obtained by the hierarchical decomposition .
images are aligned from left to right and top to bottom . at iteration
@xmath106 the sequence was set to @xmath263 with parameters @xmath241 and @xmath264.,title="fig:",scaledwidth=32.0% ] + of velocity fields obtained by the hierarchical decomposition .
images are aligned from left to right and top to bottom . at iteration @xmath106 the sequence
was set to @xmath263 with parameters @xmath241 and @xmath264.,title="fig:",scaledwidth=32.0% ] of velocity fields obtained by the hierarchical decomposition .
images are aligned from left to right and top to bottom .
at iteration @xmath106 the sequence was set to @xmath263 with parameters @xmath241 and @xmath264.,title="fig:",scaledwidth=32.0% ] of velocity fields obtained by the hierarchical decomposition .
images are aligned from left to right and top to bottom . at iteration @xmath106 the sequence
was set to @xmath263 with parameters @xmath241 and @xmath264.,title="fig:",scaledwidth=32.0% ] + of velocity fields obtained by the hierarchical decomposition .
images are aligned from left to right and top to bottom . at iteration @xmath106 the sequence
was set to @xmath263 with parameters @xmath241 and @xmath264.,title="fig:",scaledwidth=32.0% ] of velocity fields obtained by the hierarchical decomposition .
images are aligned from left to right and top to bottom . at iteration @xmath106 the sequence
was set to @xmath263 with parameters @xmath241 and @xmath264.,title="fig:",scaledwidth=32.0% ] of velocity fields obtained by the hierarchical decomposition .
images are aligned from left to right and top to bottom . at iteration @xmath106 the sequence
was set to @xmath263 with parameters @xmath241 and @xmath264.,title="fig:",scaledwidth=32.0% ] of velocity fields obtained by the hierarchical decomposition .
images are aligned from left to right and top to bottom . at iteration @xmath106 the sequence
was set to @xmath265 with parameters @xmath266 and @xmath242.,title="fig:",scaledwidth=32.0% ] of velocity fields obtained by the hierarchical decomposition .
images are aligned from left to right and top to bottom .
at iteration @xmath106 the sequence was set to @xmath265 with parameters @xmath266 and @xmath242.,title="fig:",scaledwidth=32.0% ] of velocity fields obtained by the hierarchical decomposition .
images are aligned from left to right and top to bottom . at iteration
@xmath106 the sequence was set to @xmath265 with parameters @xmath266 and @xmath242.,title="fig:",scaledwidth=32.0% ] + of velocity fields obtained by the hierarchical decomposition .
images are aligned from left to right and top to bottom . at iteration @xmath106 the sequence
was set to @xmath265 with parameters @xmath266 and @xmath242.,title="fig:",scaledwidth=32.0% ] of velocity fields obtained by the hierarchical decomposition .
images are aligned from left to right and top to bottom .
at iteration @xmath106 the sequence was set to @xmath265 with parameters @xmath266 and @xmath242.,title="fig:",scaledwidth=32.0% ] of velocity fields obtained by the hierarchical decomposition .
images are aligned from left to right and top to bottom . at iteration @xmath106 the sequence
was set to @xmath265 with parameters @xmath266 and @xmath242.,title="fig:",scaledwidth=32.0% ] as a final experiment , we computed two types of hierarchical decompositions as proposed in sec .
[ sec : decomposition ] .
first we chose @xmath263 such that @xmath246 is halved in every iteration .
solutions @xmath267 were obtained using parameters @xmath241 and @xmath264 . in fig .
[ fig : hierarchical ] , the subsequence @xmath268 is shown .
as @xmath106 increases , the motion field expands on the details . in a second run , @xmath269
was set to @xmath270 , decreasing the exponent of @xmath271 by @xmath272 in every step .
we iteratively computed solutions @xmath273 with parameters @xmath266 and @xmath242 , which were kept constant this time .
figure [ fig : hierarchical2 ] depicts the subsequence @xmath274 .
note that in figs .
[ fig : hierarchical ] and [ fig : hierarchical2 ] the colour - coded visualisation is relative to the chosen subsequence . as an exception , here we allowed a maximum number of @xmath275 iterations instead of @xmath276 for the linear system solve to ensure a relative residual of 0.025 in the first step of the hierarchical decomposition .
we provided a set of variational methods for the analysis of motion fields .
while their applicability is limited to data given on the sphere , the proposed models have great flexibility in terms of possible regularising functionals .
in fact , the chosen numerical method based on tangential vector spherical harmonics allows for a straightforward usage of sobolev @xmath12 norms for any real @xmath17 . combined with both @xmath1 and hierarchical decomposition models , which we adapted to the spherical optical flow setting ,
this flexibility makes it possible to capture different motion characteristics with ease .
feasibility of the proposed models was verified on a microscopy dataset depicting endodermal cells of a zebrafish embryo .
we thank pia aanstad from the university of innsbruck for sharing her biological insight and for kindly providing the microscopy data .
this work has been supported by the vienna graduate school in computational science ( ik i059-n ) funded by the university of vienna .
in addition , we acknowledge the support by the austrian science fund ( fwf ) within the national research networks photoacoustic imaging in biology and medicine " ( project s10505-n20 , reconstruction algorithms for pai ) and geometry + simulation " ( project s11704 , variational methods for imaging on manifolds ) .
j. abhau , z. belhachmi , and o. scherzer . on a decomposition model for optical flow . in _
energy minimization methods in computer vision and pattern recognition _ ,
volume 5681 of _ lecture notes in computer science _ , pages 126139 .
springer - verlag , berlin , heidelberg , 2009 .
g. aubert and p. kornprobst .
, volume 147 of _ applied mathematical sciences_. springer , new york , second edition , 2006 . partial differential equations and the calculus of variations , with a foreword by olivier faugeras .
f. frhauf , c. pontow , and o. scherzer .
texture enhancing based on variational image decomposition . in matine bergounioux , editor , _ mathematical image processing _
, volume 5 of _ springer proceedings in mathematics _ ,
pages 127140 , berlin heidelberg , 2011 .
springer .
a. imiya , h. sugaya , a. torii , and y. mochizuki .
variational analysis of spherical images . in a.
gagalowicz and w. philips , editors , _ computer analysis of images and patterns _ , volume 3691 of _ lecture notes in computer science _ , pages 104111 .
springer berlin , heidelberg , 2005 . c. kirisits , l. f. lang , and o. scherzer .
optical flow on evolving surfaces with an application to the analysis of 4d microscopy data .
in a. kuijper , k. bredies , t. pock , and h. bischof , editors , _ ssvm13 : proceedings of the fourth international conference on scale space and variational methods in computer vision _ ,
volume 7893 of _ lecture notes in computer science _ , pages 246257 , berlin , heidelberg , 2013 .
springer - verlag .
t. kohlberger , e. memin , and c. schnrr .
variational dense motion estimation using the helmholtz decomposition . in l. d. griffin and m. lillholm , editors , _ scale space methods in computer vision _ , volume 2695 of _ lecture notes in computer science _ , pages 432448
springer , berlin , 2003 . c. melani , m. campana , b. lombardot , b. rizzi , f. veronesi , c. zanella , p. bourgine , k. mikula , n. peyriras , and a. sarti .
cells tracking in a live zebrafish embryo . in _ proceedings of the 29th annual international conference of the ieee engineering in medicine and biology society ( embs 2007 ) _ , pages 16311634 , 2007 .
p. quelhas , a. m. mendona , and a. campilho .
optical flow based arabidopsis thaliana root meristem cell division detection . in a.
campilho and m. kamel , editors , _ image analysis and recognition _ , volume 6112 of _ lecture notes in computer science _ , pages 217226 .
springer berlin heidelberg , 2010 .
b. schmid , g. shah , n. scherf , m. weber , k. thierbach , c. campos prez , i. roeder , p. aanstad , and j. huisken .
high - speed panoramic light - sheet microscopy reveals global endodermal cell dynamics .
, 4:2207 , 2013 .
a. torii , a. imiya , h. sugaya , and y. mochizuki . .
in m. de gregorio , v. di maio , m. frucci , and c. musio , editors , _ brain , vision , and artificial intelligence _ ,
volume 3704 of _ lecture notes in computer science _ , pages 527536 .
springer berlin , heidelberg , 2005 .
l. vese and s. osher . modeling textures with total variation minimization and oscillating patterns in image processing .
, 19(13):553572 , 2003 .
special issue in honor of the sixtieth birthday of stanley osher .
j. weickert , a. bruhn , t. brox , and n. papenberg .
a survey on variational optic flow methods for small displacements . in o.
scherzer , editor , _ mathematical models for registration and applications to medical imaging _ , volume 10 of _ mathematics in industry _ , pages 103136 .
springer , berlin heidelberg , 2006 . | we propose a number of variational regularisation methods for the estimation and decomposition of motion fields on the @xmath0-sphere .
while motion estimation is based on the optical flow equation , the presented decomposition models are motivated by recent trends in image analysis .
in particular we treat @xmath1 decomposition as well as hierarchical decomposition .
helmholtz decomposition of motion fields is obtained as a natural by - product of the chosen numerical method based on vector spherical harmonics .
all models are tested on time - lapse microscopy data depicting fluorescently labelled endodermal cells of a zebrafish embryo . |
multi - wavelength high spatial resolution observations of ngc 4151 ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) have provided unique information on the continuum and emission line morphology in the central few 100 pc ( @xmath4 mpc , @xmath5 pc , * ? ? ?
* ) , contributing to our understanding of the nature of the nuclear emission and its interaction with the host galaxy ( see @xcite for a review ) .
in this paper we investigate further these topics , using high quality deep _
acis observations of the nearby prototypical seyfert 1 galaxy ngc 4151 , which allow the morphological investigation of spectral line emission in the inner circum - nuclear region ( @xmath6 pc in projection ) .
in particular , we will address the radio jet
host galaxy interaction and the nature of the fe k@xmath2 line emission .
high resolution radio surveys show broad existence of collimated radio outflows or jets in seyfert galaxies , albeit less prominent than those of radio galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
ever since the _ hubble space telescope _
emission line imaging resolved the nlr morphology in a number of seyfert galaxies with linear radio structures @xcite , it has been actively debated whether the radio jet plays a competing role against nuclear photoionization in the narrow - line region ( nlr ) ionization structure @xcite .
_ chandra _ x - ray observations of the nuclear regions of nearby seyfert galaxies appear to offer powerful diagnostics for discriminating between photoionization by the nuclear radiation or collisional ionization by a radio jet @xcite , although the importance of radio jets in shaping nlr environments seems to differ case - by - case @xcite .
the nuclear region of ngc 4151 is known to host a two - sided , @xmath0300 pc - long radio outflow along position angle ( p.a . ) @xmath077@xmath7 ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . within the central 100 pc of the galaxy
, @xcite identified a faint , highly collimated jet underlying discrete components that are shock - like features associated with the jet
gas clouds interactions .
although the distribution of the ionized gas and the kinematics in the nlr imply that shock ionization originated from jet cloud interactions is unlikely to be the dominant source of ionizing photons compared to the nucleus @xcite , @xcite find evidence that some nlr clouds are responsible for bending the jet .
near - ir emission - line mapping of the nuclear region by @xcite also reveals enhancement of [ feii ] emission , suggesting the presence of shock heating in addition to nuclear photoionization @xcite . in our previous work using the _ chandra _ hrc image of ngc 4151
, @xcite found an excess of x - ray emission at these jet
cloud interaction locations .
however , no spectral information could be obtained because of the very limited energy resolution of the hrc .
acis spectral imaging data of the ngc 4151 nuclear region @xcite were limited by both heavy pile - up and sensitivity , and did not allow detailed high spatial resolution comparison to the radio jet and emission line gas , nor spectral studies of x - ray emission from spatially resolved features .
our 200 ks acis - s observation of ngc 4151 ( pi fabbiano ; see * ? ? ?
* and wang et al .
2011 for previous work on this data set ) allows us to pursue these investigations , by comparing the morphology of the neix and ovii with that of the radio jet .
moreover , with these data we can address the outstanding controversy on the spatial properties of the 6.4 kev fe k@xmath2 emission of ngc 4151 .
in seyfert galaxies , this emission line is expected to originate from cold matter near the nucleus ( @xmath8 pc ) , either the obscuring torus ( e.g. , * ? ? ? * ) or an accretion disk ( e.g. , * ? ? ? * ; * ? ? ? * ) .
these model predictions , however , conflicted with the finding of @xcite , who reported spatially resolved narrow iron k@xmath2 line emission in the _ chandra _ hetg observation , concluding that @xmath9 of this emission originates in the enlr at distance up to 6 from the nucleus ( @xmath0400 pc across ) .
this conclusion has been more recently contested by the _
xmm_-newton based work of @xcite .
the details of our acis observations and data reduction are described in ( * ? ? ?
* hereafter paper i ) .
briefly , ngc 4151 was observed by _
chandra _ for a total of 180 ks ( after screening for high background intervals ) with the spectroscopic array of the advanced ccd imaging spectrometer ( acis - s ; * ? ? ?
* ) in 1/8 sub - array mode during march 27 - 29 , 2008 .
the data were reprocessed following standard procedures , using ciao ( version 4.2 ) with the caldb 4.2.1 provided by the _
chandra _ x - ray center ( cxc ) .
subpixel event repositioning ( `` static '' method in li et al .
2004 , using the corner split and 2-pixel split events ) and subpixel binning techniques were applied to the acis images to improve the spatial resolution . during our previous work
@xcite and paper i , we became well aware of the complexity in data analysis caused by the bright nuclear emission of ngc 4151 .
first , photon pile - up is present in the nuclear region even with the reduced frame - time of our observation .
we established that for the soft x - ray emission , pile - up is moderate ( @xmath10 ) at @xmath11 .
second , although the _ chandra _ s psf is highly centrally peaked , the contamination from the scattered nuclear emission to the extended emission is not negligible due to the brightness of the nucleus and the broader psf in the higher energy range .
we performed _
psf simulations that provide an estimate of the expected contamination from the nuclear emission in an extended feature .
detailed explanation of the analysis leading to these conclusions is presented in paper i. we have taken into account this information in the following analysis . the 27 kev emission is dominated by the unresolved nucleus @xcite . since we are interested in the extended x - ray emission , we used the energy range below 2 kev where this emission is prominent ( see paper i ) . to investigate the general spectral dependency of the morphology of this soft x - ray emission ,
images in three spectral bands below 2 kev were extracted from the merged data : 0.30.7 kev ( `` soft band '' ) , 0.71.0 kev ( `` medium band '' ) , and 12 kev ( `` hard band '' ) . following @xcite , exposure maps were created for the individual bands to obtain exposure - corrected flux images .
figure [ 3color ] presents the resulting false color composite image of the central @xmath12 ( 450 pc on a side ) region of ngc 4151 , where the soft , medium , and hard band images are shown in red , green , and blue , respectively .
the images in the three band have been smoothed with a @xmath13 gaussian kernel .
it clearly shows bright structured soft x - ray emission along the northeast ( ne ) southwest ( sw ) direction , which is also the direction of the radio jet .
in particular , the medium band image shows a jet - like @xmath02 linear extension ( contours in figure [ 3color ] ) .
this medium band is dominated by the neix emission ( see below ) . in the following we concentrate on the analysis of this inner 2-radius circum - nuclear region .
high spectral resolution grating spectra of ngc 4151 @xcite have shown that the emission in these soft spectral bands is almost entirely dominated by lines .
although the spectral resolution of acis ccd can not provide unique identifications of the strongest soft x - ray emission lines ( @xmath142 kev ) seen in the nuclear spectrum @xcite , we can identify the dominant transitions guided by the hetg observations @xcite .
most notably the blended lines appear as three strong lines in the acis spectrum , approximately centered at 0.57 kev , 0.68 kev , 0.91 kev ( c.f . figure 4 in yang et al .
2001 ; figure 3 in @xcite ) .
we then extracted the x - ray emission in three narrow energy intervals ( 0.530.63 kev , blended ovii f , i , r ; 0.630.73 kev , blended oviii ly@xmath2 and ovii rrc ; 0.850.95 kev , blended neix f , i , r ) to create line strength images , highlighting regions of these prominent emission lines .
this is a reasonable approach since the line emission dominates over the weak underlying continuum @xcite in these narrow bands and reveal substructures that are not obvious in the broadband images ( e.g. , * ? ? ?
the resulting images are shown in figures [ lines ] and [ o3 ] , showing the x - ray emission - line structure close to the nucleus ( the central @xmath15 , @xmath0130 pc ) in the context of the radio outflow and the optical emission line clouds .
the alignment between these images is done using the peak of x - ray emission , [ oiii ] emission , and the radio core . since the astrometry of our _ chandra _ image is accurate to @xmath16 , our interpretation of these features is not affected by significant alignment uncertainties .
the position of the nucleus is indicated with a cross .
the ovii , oviii , and neix emission line images all show extended morphology and some structures , closely following the p.a . of the large scale extended nlr traced by the [ oiii ] emission ( e.g. , * ? ? ?
* ; * ? ? ?
* ) . in particular
, the linear feature seen in the medium - band is clearly present in the neix image .
there appears to be two x - ray enhancements bracketed by the optical clouds ( indicated by arrows in figure [ o3 ] ) that are close to the radio knot features c2 and c5 @xcite .
these `` hot spots '' are better visualized in the neix / ovii ratio image , shown in figure [ jet ] together with the vlba jet @xcite and the nir [ feii ] line emission @xcite .
the locations where the jet appears to be intercepted by the optical clouds and where prominent [ feii ] emission arise , are characterised by a neix / ovii ratio of @xmath17 , which is significantly ( @xmath18 ) higher than the ratio in the surrounding regions ( @xmath19 ) .
we also examined the optical clouds with high velocity dispersion , possibly associated with jet - cloud impact ( marked as crosses in figure [ jet ] ; see figure 7 in mundell et al . 2003 ) , and find an elevated neix / ovii ratio of @xmath20 ( with marginal significance ) .
the eastern neix / ovii hot spot appears to be associated with clear interactions between the jet and the gas cloud , where shock heating becomes important .
we note that the observed line flux of ovii is more suppressed than neix in the presence of the galactic absorption column ( @xmath21 @xmath22 ) , and xspec @xcite simulations show the absorption corrected ratio may systematically decrease by 10% .
thus the presence of neix / ovii hot spots is not affected , unless there is significant differential absorption in the region .
we extracted x - ray spectra from both the high neix / ovii emission line ratio regions and the surrounding low ratio region ( figure [ jet]b ) , and fit them jointly ( with the same model components but normalizations set free ) for comparison using xspec ( version 12.5 ; arnaud 1996 ) .
spectra and instrument responses were generated using ciao tool _ specextract_. the background spectrum
is taken from a nearby source - free region on the same ccd node .
spectra were grouped to have a minimum of 20 counts per energy bin to allow for @xmath23 fitting . the contribution from the bright nuclear emission can not be neglected here ; to have a self - consistent model , we included a fraction of the scattered emission predicted from the psf simulation of the point - like ngc 4151 nucleus .
we have made use of the cloudy photoionization modeling code , last described by @xcite , to model the soft x - ray emission . using cloudy version c08.00 , which enables a cloudy / xspec interface @xcite , we attempted to produce the soft part of the x - ray spectrum assuming an open plane - parallel geometry ( `` slab '' ) .
the dimensionless ionization parameter @xcite is defined as @xmath24 , where @xmath25 is the hydrogen number density , @xmath26 is the distance to the inner face of the model slab , @xmath27 is the speed of light , and @xmath28 is the emitting rate of hydrogen ionizing photons ( s@xmath29 ) by the ionizing source . it was previously noted that the spectral fit to the soft x - ray continuum of the nucleus is quite uncertain because of heavy absorption @xcite , therefore we adopted the broken power - law form in @xcite for the agn continuum ( @xmath30 for the energy range between 13.6 ev and 0.5 kev , and @xmath31 for @xmath32 kev ) .
normalizing to the observed x - ray luminosity of the nucleus @xcite , we obtain @xmath33 ( photons s@xmath29 ) .
we varied @xmath34 and the column density @xmath35 of the model slab to create spectral model grids , which were fed to xspec .
we quickly find that , using a single photoionized component alone can not reproduce the observed soft x - ray emission in ngc 4151 . with a high photoionization component ( @xmath36 )
, we were able to produce the hydrogen - like neon and oxygen lines emission ( e.g. , oviii , nex ) , which are unambiguously present in all the grating spectra ( e.g. , * ? ? ? * ; * ? ? ?
* ; * ? ? ?
however , the helium - like transitions , in particular the ovii and neix emission line features , are not fitted acceptably . adding a second lower ionization photoionized component ( @xmath37 ) ,
as detected in @xcite , we were able to obtain a significantly improved fit ( @xmath38 ) , except for residuals in the 0.71.1 kev range that are particularly strong for the hot spot spectrum .
adding additional photoionized components to the model does not further improve the fit when an f - test is performed . allowing different absorption columns for
the models also does not give a better fit ( reduced @xmath39 ) , indicating that the high neix / ovii ratio is not due to higher obscuration .
the absorption column required by the fits is consistent with the low galactic extinction towards the nlr region found in @xcite . instead , these residuals disappear when a thermal emission component ( @xmath40 ; * ? ? ?
* ) with a temperature of @xmath41 kev is added to the model , suggesting the presence of collisionally ionized fe l emission .
the spectral fitting results are summarized in table 1 and shown in figure [ 2reg ] .
the absorption column is consistent with the low galactic extinction towards the nlr region found in @xcite .
lastly , we note that fitting the data with only combinations of absorbed optically - thin thermal emission ( @xmath40 ) gave strong residuals of emission line features ( reduced @xmath42 ) , even when the abundance @xmath43 is allowed to vary .
the subpixel resolution acis image also allows a similar comparison of the x - ray emission with the optical nlr clouds resolved in the @xmath44 image . taking advantage of the spectral capability of acis
, we measured the 0.32 kev counts for the same [ oiii ] clouds identified in @xcite , and derived their 0.52 kev x - ray fluxes using the spectral model above . to subtract the nuclear contribution
, we used a local region at the same radii but with different azimuthal angles than the clouds .
the results are listed in table 2 .
figure [ ratio ] shows these [ oiii ] to soft x - ray ratios , comparing the measurements with hrc results in @xcite and photoionization models in @xcite .
the [ oiii]/x - ray ratios at the two x - ray hot spot locations ( corresponding to radio knots c2 and c5 in @xcite ) were also measured , and found to be uniformly low , @xmath03 .
this was previously noted in @xcite , and implies enhanced x - ray emission compared to other clouds under nuclear photoionization , which is likely associated with the jet cloud interaction and consistent with their association with high neix / ovii ratio regions .
the 6.4 kev fluorescent iron k@xmath2 emission of neutral or mildly ionized cold material is readily visible in the acis spectra of the nucleus @xcite and the extended region ( figure [ fek1]a ) . another fluorescent feature , sii k@xmath2 emission at 1.74 kev , is also clearly present in the extended emission ( figure [ fek1]b ) .
both spectra of extended emission were extracted from a @xmath45 annular region centered on the nucleus .
is this `` extended '' fluorescent line emission truly originating from outside the nuclear region , or is it the result of the psf wing spreading out point - like nuclear features ? to determine the spatial distribution of the fe k@xmath2 line emitting material , we followed the procedure in @xcite : a continuum image , taken to be the average of the counts in the 5.96.2 kev and 6.56.8 kev bands , was subtracted from the image extracted in the 6.26.5 kev band , resulting in a `` pure '' fek@xmath2 line image , shown in figure [ fek2]a .
the fe k@xmath2 emission appears mostly circular in the @xmath46 , but shows faint extended emission towards the north and west .
the si k@xmath2 line is blended with the mgxii ly@xmath47 and sixiii @xmath48 emission @xcite , thus the above continuum subtraction is not applied .
we extracted emission in the 1.71.8 kev range to obtain the si k@xmath2 emission map ( figure [ fek2]b ) .
it is mostly concentrated within a @xmath49 circle , with a slight elongation along the optical bi - cone direction ( northeast southwest ) .
however , at such energies the wings of the energy - dependent telescope psf spread a small but non - negligible amount of the nuclear photons to large radii .
thus the presence of `` extended '' fe k@xmath2 emission or si k@xmath2 emission in the @xmath50 region may in fact come from the psf - scattered unresolved nuclear emission . to investigate this point
, we further compared the observed radial profile of the fe k@xmath2 emission band ( 6.26.5 kev ) and the radial profile in the same band for a point - like nucleus simulated with marx ( see * ? ? ?
* ) , shown in figure [ fek]a . a similar comparison for the si k@xmath2 ( 1.71.8 kev ) is shown in figure [ fek]b .
the extracted observed profiles shown in figure [ fek ] agree well with the psf profiles , except for the inner @xmath51 where the simulation overpredicted the observed emission due to heavy pile up in the psf core .
the simulated psfs indicate that the @xmath46 region contains 89% of the point source emission in the 6.26.5 kev band and 94% of the point source emission in the 1.71.8 kev band .
approximately 6% of the nuclear fe k@xmath2 emission is expected to be spread within the @xmath52 extended regions , the extent of fe k@xmath2 seen in figure [ fek1]a .
taking into account the contribution from the psf wings , the remaining extended fe k@xmath2 emission is @xmath53 of the total fe k@xmath2 emission .
this should be considered an upper limit for the truly extended emission , because the nuclear fe k@xmath2 emission is likely to be underestimated due to pile up .
similarly , we estimated that the extended si k@xmath2 emission is @xmath54 of the total si k@xmath2 emission . as pointed out by the referee , the measured extended emission could easily disappear with a slightly broader simulated psf given that the ditherblur parameter in the marx simulation is not known a priori .
therefore , we conclude that although there is faint extended emission seen in the fe and si fluorescent line images , at most 5% of the observed fe k@xmath2 emission and 4% of the si k@xmath2 emission can be truly spatially extended ; most of these emissions are consistent with the psf scattering of a strong nuclear component .
both our previous work using hrc image of ngc 4151 nuclear region @xcite and this work using acis find that most of the nlr clouds in the central 150 pc radius region of ngc 4151 have a relatively constant [ oiii ] to soft x - ray ratio ( @xmath010 ; figure [ ratio ] and table 2 ) , no matter the distance of the clouds from the nucleus .
this ratio indicates a uniform ionization parameter and a density decreasing as @xmath55
. this may be consistent with a photoionized wind scenario ( e.g. , bianchi et al .
2006 ) as the nlr clouds are outflowing ( e.g. , kaiser et al .
the [ oiii]/x - ray ratios at the locations of jet - cloud collision are lower ( @xmath03 ) , leading to the suggestion of enhanced x - ray emission in these regions due to shock heating in addition to photoionization .
this conclusion is supported by the results of multiwavelength imaging studies .
the high spatial resolution emission line images of ovii , oviii , and neix emission show extended structures , in particular two `` hot spots '' in the neix image , which are close to the radio knot features ( figures [ lines ] and [ o3 ] ) . at these locations ( 1 from the nucleus )
, @xcite noted that the morphology of the highly collimated radio outflows appears more disturbed , suggesting interaction with some dense clouds here .
there are also clearly optical clouds with high velocity dispersion associated with the jet
cloud impact @xcite .
the radio knots are near to the location of strong [ feii ] emission @xcite , falling between the peaks of the [ feii ] emission . altogether , the morphology suggests a scenario where the radio ejecta runs into a denser cloud in the inhomogeneous ism and results in locally enhanced shock heating ( e.g. , * ? ? ?
cloud interaction has been suggested to explain the enhanced [ feii ] emission , which could be due to the iron unlocked from grains by the shocks @xcite
. the neix emission , which traces the higher ionization gas , becomes more prominent with the extra collisional ionization from the jet cloud interaction .
@xcite also noted that the poor fit to the ovii line profile in the grating spectra could be due to presence of a non - photoionization component .
this possibility is further supported by our spectral fits , which suggest that the neix / ovii enhancements have a thermal origin , requiring the presence of a collisionally ionized plasma ( the @xmath41 kev component ) in addition to a photoionized nuclear outflow @xcite .
the emission measure of the @xmath40 component allows us to estimate the electron density @xmath56 ( @xmath57 ) and the thermal pressure @xmath58 ( @xmath59 ) . the emitting volume ( @xmath60 ) of the hot gas , assuming that the depth along the line of sight is comparable to the other dimensions , is @xmath61 @xmath62 for the high neix / ovii ratio hot spots .
the filling factor @xmath63 is assumed to be 100% and both of @xmath56 and @xmath58 have a weak dependence on @xmath63 ( @xmath64 ) .
there is @xmath65 uncertainty in @xmath66 , which is propagated to the estimated pressure and energy . for the @xmath67 kev emission , we derived an average density @xmath68 @xmath69 , a hot gas pressure @xmath70 dyne @xmath22 , a thermal energy content of @xmath71 ergs , and a cooling time of @xmath72 yr . the estimated internal pressure of the radio jet based on the synchrotron minimum energy assumption ranges between @xmath73 and @xmath74 dyne @xmath22 ( pedlar et al . 1993 ) , to be confined by the hot gas pressure of @xmath74 dyne @xmath22 , which further supports our collisional ionization scenario .
assuming @xmath75 k for the photoionized clouds and @xmath76 @xmath69 @xcite , the thermal pressure from the photoionized gas is @xmath77 dyne @xmath22 , which is comparable to the thermal pressure of the hot ism , implying a possible pressure equilibrium between the collisionally ionized hot gas and the photoionized line - emitting cool clouds .
we further estimate the age of the interaction from the approximate crossing time of the @xmath78 region ( dimension of the radio knot ; * ? ? ?
* ) with a characteristic velocity of @xmath79 km s@xmath29 , the local sound speed .
this is approximately the thermal velocity for the 0.58 kev gas ( @xmath80 km s@xmath29 ) , and also at the order of velocity dispersion of emission line gas seen in @xcite . assuming the strong shock jump conditions , a crude estimate of the shock velocity @xmath81 can be obtained using @xmath82 k @xmath83 km s@xmath84 @xcite . for the @xmath85
k x - ray - emitting gas , a @xmath81 of @xmath0700 km s@xmath29 relative to the downstream material is required .
this appears consistent with a supersonic but relatively slow radio jet , which is constrained to be sub - relativistic from radio proper motion estimates ( e.g. , @xmath86 km s@xmath29 ; * ? ? ?
thus we obtain a characteristic timescale of @xmath87 yr . if the jet cloud interaction converts kinematic energy into heating of the hot gas , a lower limit can be placed on the kinematic luminosity of the jet , @xmath88 erg s@xmath29 .
it is interesting to compare this observed @xmath89 to the jet power directly inferred from the synchrotron emission . using the @xmath90@xmath91 scaling relation ( equation 1 ) in @xcite and a total radio luminosity at 1.4 ghz ( @xmath92 erg s@xmath29 ;
* ) , we find @xmath93 erg s@xmath29 , suggesting that @xmath1 of the jet power is deposited in the ism .
however , recalling that the @xmath90@xmath91 relation was derived from a sample of radio loud galaxies with x - ray cavities ( see @xcite for review on the @xmath90@xmath91 scaling relation ) while here ngc 4151 is of low radio power , we also estimate directly the jet energy flux using the pressure in the knots @xcite based on the minimal energy assumption , giving a similar @xmath94 erg s@xmath29 .
we further note that , the crossing time @xmath95 is comparable to the cooling time @xmath96 , implying that the hot gas is prominent only locally ( close to the radio knots ) . indeed , along the path of the radio jet ,
the clouds that are spatially close to the jet impact spots ( e.g. , clouds # 3,4 bracketing knot c2 ; # 9,10 for knot c5 ) show [ oiii]/x - ray ratios indistinguishable from others ( figure [ ratio ] ) .
if these clouds are physically related to the ism interacting with the jet , instead of being projected close to the knots in our line - of - sight , their x - ray emission shows little evidence for excess shock heating in addition to nuclear photoionization .
this localized heating is perhaps related to the highly collimated radio ejecta in ngc 4151 @xcite , in contrast to the more expanded lobe - like radio outflow in mrk 3 @xcite , where the nlr gas appears broadly impacted .
we plan to investigate in future work whether the [ oiii]/x - ray ratio , such as @xmath974 measured here and seen in mrk 3 and ngc 1386 @xcite , together with other line ratios ( e.g. , [ feii]/[pii ] ; @xcite ) could be useful diagnostics for strong jet
ism interaction in seyfert galaxies . in terms of energetics
, we summarize our findings in ngc 4151 : ( a ) the jet power estimated from the radio power is @xmath98 erg s@xmath29 , a few percent of the current agn energy output ( @xmath99 erg s@xmath29 ; * ? ? ?
* ) , which is close to the canonical @xmath05% of the radiant energy adopted in theoretical models for efficient agn feedback ( e.g. , * ? ? ?
* ; * ? ? ?
* ) ; ( b ) @xmath1 of the jet power is observed to have been deposited into the host ism in the nuclear region through the interaction between the radio jet and the dense medium ; ( c ) the shock heating due to the jet
cloud interaction appears localized to the impact spots , and most clouds are consistent with being photoionized by the nucleus .
@xcite claimed that the narrow iron line emission was spatially resolved in the _ chandra _ hetg observation and @xmath100 of the fei k@xmath2 emission comes from the enlr at distance up to 6extent ( @xmath0400 pc across ) .
this is intriguing , since the putative pc - scale torus or more compact broad - line region ( * ? ? ?
* and references therein ) are expected to be the primary location of narrow fei k@xmath2 emission .
@xcite cautioned that due to the sensitivity of the short grating observation in @xcite , the 13 arcsec region off the peak of cross - dispersion profile contains few photons .
therefore the detection of spatially extended iron line emission can not be significant .
a high signal - to - noise xmm - newton spectrum of the neutral fe k@xmath2 suggests that all line flux originates in a nearly compton - thick torus that is not resolvable at current resolution @xcite .
our results show that the extended fe k@xmath2 emission , if present , is only at @xmath101 level of the observed fe k@xmath2 emission , when the contribution from the psf wings is taken into account .
this is consistent with our constraint on the extent of the si fluorescent emission and supports the conclusions of @xcite , while it is in strong disagreement with the @xmath102 reported in @xcite .
we note that in @xcite the psf is represented as a narrow gaussian with @xmath103 , when the _ chandra _ mission was new .
the marx model is now much more advanced so that a more reliable estimate becomes possible . at 6.4 kev ,
the encircled energy fraction of a point source is only @xmath104 ( see @xmath105 proposers observatory guide chapter 4 table 4.2 ) in such a psf core . hence a significant fraction of the nuclear emission could have been measured in ogle et al . as spatially extended fe k@xmath2 emission .
in this paper we present spectral analysis and emission line images from deep _ chandra _ observation of ngc 4151 , aiming to resolve and characterise the x - ray emission in the inner @xmath0130 pc - radius nuclear region .
the findings are summarized as follows : 1 .
we have obtained high spatial resolution x - ray narrow - band images of ovii , oviii , and neix line emission , which are blended at the acis spectral resolution .
the images show extended structures that are spatially correlated with the radio outflow and optical [ oiii ] emission .
2 . we find strong evidence for jet
ism interaction , including morphological correspondences with regions of x - ray enhancement , peaks of nir [ feii ] emission , and optical clouds .
this is further strengthened by the presence of a @xmath67 kev collisionally ionized component in the spectral fitting of the hot spots , and by the excess of x - ray emission in addition to nuclear photoionization as indicated by a low [ oiii]/x - ray ratio .
we find a possible pressure equilibrium between the collisionally ionized hot gas and the photoionized cool clouds .
the estimated velocity of the shocks from the jet
cloud impact is @xmath0700 km s@xmath29 .
we estimate that the jet power in ngc 4151 is close to a few percent of the current agn energy output ( @xmath106 erg s@xmath29 ) .
the derived thermal energy in the hot gas suggests that @xmath1 of the jet power is deposited into the host ism in the nuclear region through the interaction between the radio - jet and the dense medium .
the [ oiii]/x - ray ratios of nlr clouds bracketing the radio knots show little deviation from other photoionized clouds , indicating a localized impact on the ism by the highly collimated jet .
we investigate the spatial extent of the fluorescent features , including the fe k@xmath2 emission and the si k@xmath2 emission .
our results show that both are dominated by point - like emission , consistent with an origin of unresolved inner structure such as a torus .
the extended fe k@xmath2 emission is @xmath3 of the observed fe k@xmath2 emission , which is in strong disagreement with the @xmath102 reported in @xcite .
we thank the anonymous referee for helpful suggestions .
this work is supported by nasa grant go8 - 9101x ( pi : fabbiano ) and grant go1 - 12009x ( pi : wang ) .
we acknowledge support from the cxc , which is operated by the smithsonian astrophysical observatory ( sao ) for and on behalf of nasa under contract nas8 - 03060 .
cgm acknowledges financial support from the royal society and research councils u.k .
j. w. thanks g. ferland , t. kallman , s. bianchi , a. marinucci , and s. chakravorty for extensive advices on photoionization modeling , p. nulsen for jet power discussion , t. storchi - bergmann and r. riffel for providing the gemini nifs maps .
this research has made use of data obtained from the _ chandra _ data archive , and software provided by the cxc in the application packages ciao and sherpa .
, g. p. , bautz , m. w. , ford , p. g. , nousek , j. a. , & ricker , jr .
, g. r. 2003 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol .
4851 , society of photo - optical instrumentation engineers ( spie ) conference series , ed .
j. e. truemper & h. d. tananbaum , 2844 250 pc of ngc 4151 ( a ) ovii ; ( b ) oviii+ovii rrc ; ( c ) neix .
the nucleus position is indicated with a cross .
the contours outline the radio outflow in the 1.4 ghz merlin map ( mundell et al .
the acis images have been rebinned to @xmath107 per pixel , and smoothed with a @xmath13 gaussian kernel . [ lines ] ] 250 pc of ngc 4151 as in figure [ lines ] , but with contours showing the @xmath44/foc [ oiii]@xmath108 emission line clouds ( winge et al .
( a ) ovii ; ( b ) oviii+ovii rrc ; ( c ) neix .
the acis images have been rebinned to @xmath107 per pixel , and smoothed with a @xmath13 gaussian kernel .
[ o3 ] ] high - ratio & 4585 & @xmath109 & 20.5 & @xmath110 & @xmath111 & 22.5 & @xmath112 & @xmath113 & @xmath114 & @xmath115 & @xmath116 + low - ratio & 1317 & @xmath109 & 20.5 & @xmath117 & @xmath111 & 22.5 & @xmath118 & @xmath113 & @xmath119 & @xmath120 & @xmath121 + cccccccc 1 & 1.76 & 114 & 1.1 & 369@xmath12224 & 1.5 & 7.4 & 12 + 2 & 1.44 & 93.6 & 1.9 & 688@xmath12229 & 2.7 & 6.9 & 12 + 3 & 1.01 & 65.6 & 2.2 & 581@xmath12238 & 2.3 & 9.4 & 12 + 4 & 0.78 & 50.7 & 1.3 & 470@xmath12239 & 1.9 & 6.9 & 7 + 5 & 1.0 & 65.0 & 1.5 & 168@xmath12229 & 0.7 & 22.2 & 75 + 6 & 0.6 & 39 . & 1.1 & 99@xmath12244 & 0.4 & 27.8 & 110 + 7 & 0.4 & 26 . & 3.4 & 1320@xmath12296 & 5.3 & 6.4 & 5 + 8 & 0.4 & 26 .
& 3.3 & 560@xmath122100 & 2.2 & 14.7 & 13 + 9 & 0.85 & 55.2 & 2.4 & 694@xmath12247 & 2.8 & 8.6 & 3 + 10 & 0.89 & 57.8 & 1.6 & 266@xmath12235 & 1.1 & 15.0 & 8 + 11 & 2.15 & 139.7 & 0.6 & 264@xmath12217 & 1.1 & 5.7 & 15 + c2 & 0.9 & 58.5 & 0.8 & 1180@xmath12244 & 4.7 & 1.7 & 2 + c5 & 0.93 & 60.4 & 1.4 & 1052@xmath12243 & 4.2 & 3.3 & 3 + | we have studied the x - ray emission within the inner @xmath0150 pc radius of ngc 4151 by constructing high spatial resolution emission line images of blended ovii , oviii , and neix .
these maps show extended structures that are spatially correlated with the radio outflow and optical [ oiii ] emission .
we find strong evidence for jet gas cloud interaction , including morphological correspondences with regions of x - ray enhancement , peaks of near - infrared [ feii ] emission , and optical clouds . in these regions , moreover , we find evidence of elevated neix / ovii ratios ; the x - ray emission of these regions also exceeds that expected from nuclear photoionization .
spectral fitting reveals the presence of a collisionally ionized component .
the thermal energy of the hot gas suggests that @xmath1 of the estimated jet power is deposited into the host interstellar medium through interaction between the radio jet and the dense medium of the circum - nuclear region .
we find possible pressure equilibrium between the collisionally ionized hot gas and the photoionized line - emitting cool clouds .
we also obtain constraints on the extended iron and silicon fluorescent emission .
both lines are spatially unresolved .
the upper limit on the contribution of an extended emission region to the fe k@xmath2 emission is @xmath3 of the total , in disagreement with a previous claim that 65% of the fe k@xmath2 emission originates in the extended narrow line region . |
special relativity ( sr ) and general relativity ( gr ) are at the basis of our understanding of space and time and thus are fundamental for the formulation of physical theories . without sr
we can not explain the phenomena in high energy physics and in particle astrophysics , without gr there is no understanding of the gravitational phenomena in our solar system , of the dynamics of galaxies and of our universe , and , finally , of the physics of black holes .
both theories are linked by the fact that the validity of st is necessary for gr . due to the overall importance of these theories , a persistent effort to improve the experimental tests of their foundations
is mandatory .
modern tests of sr and gr using ultrastable oscillators have been performed on earth @xcite , and an experiment is planned on the international space station @xcite .
furthermore , new results from quantum gravity theories predict small deviations from sr and gr giving additional motivation to improve tests on sr and gr . for example ,
loop gravity and string theory predict modifications of the maxwell equations @xcite .
these modifications lead to an anisotropic speed of light and to a dispersion in vacuum , thus violating the postulates of sr .
quantum gravity also predicts , besides a violation of the weak equivalence principle @xcite , a violation of the universality of the gravitational red shift @xcite . although the amount of sr violations predicted by quantum gravity are too small to be in the range of experimental capabilities in the near future , these predictions open up the window for violations for such basic principles . since we do not know the true
quantum gravity theory we also do not know the true
parameters of the theory and therefore the predicted range of violations of sr and gr is no final prediction , but merely a hint .
( -5,-3)(7,5)(-4.6,-3.5)(7,5 ) ( -3,0)0.47 ( -3,0),title="fig:",width=37 ] ( -0.5,1.5)(0,0)(4.5,1.7 ) ( -3,-1.5)perigee @xmath0 10000 km ( 3.7,4.4)apogee @xmath0 40000 km ( -0.5,3.4)(4.8,2.2 ) ( -0.5,3.3)(1.5,-1.7 ) ( -1.7,3.3)(0,0.3)(0.6,0.2 ) ( -0.3,0.3)(0.3,1 ) ( 0.4,0.42)(0.3,0.45 ) ( -0.5,3.3 ) ( -1.5,0.5)(-0.5,1.5)(-3.5,1.5)(-4.5,0.5)(-1.5,0.5 ) ( 0,0)(1,0.5 ) ( -1,0)(1,2 ) ( -1,0)(-1,2 ) ( 1,0)(1,2 ) ( 0,2)(1,0.5 ) ( 0.5,0.5)(1.5,1.5)(4.5,1.5)(3.5,0.5)(0.5,0.5 ) ( 0,2)(0,9 ) ( -1.4,3.5)@xmath1 ( -2,4.3)(-4,4.3 ) ( -3,4.6)to sun ( 4,-0.45)2.8 ( 4.75,-0.15 ) ( 3.2,4.2)(3.2,-1.2)(1.2,-3.2)(-4.2,-3.2)(-6.2,-1.2)(-6.2,4.2)(-4.2,6.2)(1.2,6.2)(3.2,4.2 ) ( 0,-0.8)(0.8,0 ) ( 1.2,-0.8)(2,0 ) ( 0,-2)(0.8,-1.2 ) ( 1.2,-2)(2,-1.2 ) ( 0,-0.8)(0,0)(0.8,0 ) ( 1.2,0)(2,0)(2,-0.8 ) ( 0,-1.2)(0,-2)(0.8,-2 ) ( 1.2,-2)(2,-2)(2,-1.2 ) ( 2,-1.2)(2,-0.8 ) ( 0.8,-2)(1.2,-2 ) ( 0,-1.2)(0,-0.8 ) ( 0.8,0)(1.2,0 ) ( -5.5,-1.3)(-4,-0.7 ) ( -4,-0.9)(-4,-0.7)(-5.5,-0.7)(-5.5,-1.3)(-4,-1.3)(-4,-1.1 ) ( 0.7,3)(1.3,4.5 ) ( 0.9,3)(0.7,3)(0.7,4.5)(1.3,4.5)(1.3,3)(1.1,3 ) ( -6,0.7)(-4.5,1.3 ) ( -6,2.7)(-4.5,3.3 ) ( -5.25,1.3)(-5.25,2.7 ) ( -1,4.2 ) ( 0,0)0.30180 ( 0,0)0.30180 ( 0,0.3)(0,1.2 ) ( -1.9,1 ) ( 0,0)0.30180 ( 0,0)0.30180 ( 0.5,0.5)(1.5,1.5 ) ( -1.5,-1.5)(-0.5,-0.5 ) ( -0.5,0.5)(-1.5,1.5 ) ( -1.5,2.5)(-0.5,3.5 ) ( -4,-1)(-2.5,-1 ) ( -2.5,-1)(1.6,-1 ) ( 2,-1)(0.4,-1 ) ( -1,-1)(-1,0 ) ( -1,0)(-1,2 ) ( -1,2)(-1,4.2 ) ( -4.5,3)(-2.75,3 ) ( -2.75,3)(-1,3 ) ( 1,3)(1,1.25 ) ( 1,2.5)(1,-1.6 ) ( 1,-2)(1,-0.4 ) ( 1,1)(0,1 ) ( 0,1)(-1,1 ) ( -1,1)(-1.9,1 ) ( -2.2,1)(-3,1 ) ( -4.5,1)(-3.64,1 ) ( -3.34,1.3)(-3.34,2.9 ) ( -3.34,3.1)(-3.34,5.4 ) ( -3.34,1)@xmath2 ( -4.5,-1.6)laser 1 ( 1.6,3.75)laser 2 ( 0,-2.3 ) ( 0,-2.7 ) ( -2.2,5.7)to data recording ( 2.7,1.5)thermal isolation ( -5.1,0.4)atomic ( -5.2,-0.1)clock ( -5,4.1)comb ( -4.9,3.6)generator by means of the proposed mission optis an improvement of three tests of sr and gr by up to three orders of magnitude is projected : llc experiment & present accuracy & projected accuracy + [ 1ex ] ' '' '' michelson morley experiment ( mm ) & @xmath3 @xcite & @xmath4 + kennedy thorndike experiment ( kt ) & @xmath5 @xcite & @xmath6 + universality of gravitat . red shift ( lpi ) & @xmath7 @xcite & @xmath8 + [ 1ex ] the main features of the optis mission can be seen in fig.[optisprinzip ] : a spinning drag free satellite orbits the earth .
the satellite payload consists of two lasers , two orthogonal optical cavities , a femtosecond laser comb generator , and an atomic clock .
the two cavities are used for the mm experiment which searches for differences in the velocity of light in different directions .
the atomic clock represents an independent clock of different physical nature
. a comparison between the atomic clock and an optical cavity can be performed by means of the comb generator . a search for a dependence of their frequency ratio with respect to a change of the velocity of the satellite or with respect to a change of the gravitational potential amounts to a kt test and to a test of the universality of the gravitational red shift , respectively .
the most precise experiments testing the constancy of the speed of light use cavities .
the wave vector magnitude @xmath9 of an electromagnetic wave in a cavity of length @xmath10 is given by @xmath11 and the frequency @xmath12 of an outcoupled wave by @xmath13 .
if the velocity of light depends on the orientation of the cavity and on the velocity @xmath14 of the laboratory , @xmath15 , so will the frequency @xmath16 .
fig.[fig : schemetest ] shows a schematic setup for a search of an orientation and velocity dependence of the frequency . since sr is based on an orientation and velocity _
in_dependent speed of light , the search for an orientiation and velocity dependence of the frequency amounts to a test of sr .
( 1,-2)(9,7)(5,1.4)(0,0.02)10(0,0)(4,1 ) ( 4.4,1.6)(5.2,1.4)(5.7,2.3)(4.9,2.5)(4.4,1.6 ) ( 5.2,1.4)(5.7,2.3)(5.7,1.7 ) ( 2.7,1.3)(5,2 ) ( 5,2)(5,4 ) ( 5,4)(5,5 ) ( 3.7,1.6)(-0.05,-0.015)20(0,0)(0.25,0.4 ) ( 5,-1)(5,0.36 ) ( 5,5)(0,0)0.30180 ( 0,0)0.30180 ( 0,0.3)(0,1.8)(0.7,-0.2)(0.7,1.3 ) ( 7.3,5.1 ) ( 6.5,1.8)mirror ( 3,2.8)cavity ( -0.3,1.2)turn table ( 5,0.8)(0.5,-0.992157)(1,-0.968246)(1.5,-0.927025)(2,-0.866025)(2.5,-0.780625)(3,-0.661438)(3.5,-0.484123)(3.7,-0.379967)(3.9,-0.222205 ) ( 8.3,-0.2)@xmath17 ( 4,-2)(7,-1 ) ( 7,-1.4)@xmath18 according to common test theories @xcite , the orientation and velocity dependence of the velocity of light is parametrized according to @xmath19
this means that all anomalous terms vanish for vanishing velocity @xmath14 . here
@xmath17 is the angle between the velocity with respect to the cosmic preferred frame @xmath18 and the cavity axis .
in addition , an expansion with respect to @xmath20 has been used , for simplicity .
if special relativity is valid , then @xmath21 .
the parameters @xmath22 and @xmath23 in the mansouri sexl and robertson test theories are given by [ cols="<,<,<",options="header " , ] in order to test the isotropy of space , or the isotropy of the velocity of light , one has to mount the cavity on a turn table and look for a variation of the frequency as the turn table is rotated .
this setup has been used in experiments since 1955 , see fig.[fig : historymm ] . in terms of the relative variation of the velocity of light , of the robertson parameter , and of the mansouri
sexl parameter , the most accurate result is @xcite @xmath24 for more interpretation see @xcite .
a hypothetical dependence of the velocity of light on the velocity @xmath14 of the laboratory can be tested by changing the velocity of the cavity and looking for a variation of the frequency . in earth
based experiments , the rotation of the earth around its axis ( @xmath25 ) or around the sun ( @xmath26 ) can be used , where @xmath27 is the velocity with respect to the cosmic microwave background , the cosmologically preferred frame .
because of technical reasons only the rotation of the earth around its own axis has been used so far . in terms of the same parameters as above ,
the most accurate result is @xcite @xmath28 in the framework of einstein s gr , the comparison of two identical clocks of frequency @xmath29 located in different gravitational potentials @xmath30 and @xmath31 yields @xmath32 . the gravitational red shift does not depend on the type of clock .
this is the universality of the gravitational red shift , an aspect of local position invariance .
if einstein s theory is not correct , then the red shift may depend on the clock @xmath33 with @xmath34 . in einstein
s gr @xmath35 .
if two different clocks are displaced together in a gravitational potential , a relative frequency shift @xmath36 may occur , which is proportional to the difference of the gravitational potential relative to the uinitial position .
( 0,-1)(16,11)(1,-1.5)(16,-1.5 ) ( 1,-1.5)(1,10 ) ( 0.1,10)@xmath37 ( 1,-1.5)(1,0)15(0,0)(0,0.2 ) ( 2,-2)1880 ( 3,-2)1890 ( 4,-2)1900 ( 5,-2)1910 ( 6,-2)1920 ( 7,-2)1930 ( 8,-2)1940 ( 9,-2)1950 ( 10,-2)1960 ( 11,-2)1970 ( 12,-2)1980 ( 13,-2)1990 ( 14,-2)2000 ( 0.3,-1)@xmath38 ( 1,-1)(15,-1 ) ( 0.3,0)@xmath39 ( 1,0)(15,0 ) ( 0.3,1)@xmath40 ( 1,1)(15,1 ) ( 0.3,2)@xmath41 ( 1,2)(15,2 ) ( 0.3,3)@xmath42 ( 1,3)(15,3 ) ( 0.3,4)@xmath43 ( 1,4)(15,4 ) ( 0.3,5)@xmath44 ( 1,5)(15,5 ) ( 0.3,6)@xmath45 ( 1,6)(15,6 ) ( 0.3,7)@xmath46 ( 1,7)(15,7 ) ( 0.3,8)@xmath47 ( 1,8)(15,8 ) ( 0.3,9)@xmath48 ( 1,9)(15,9 ) ( 2,8)(0,0 ) ( 4.4,7)(0,0)@xmath49 morley & miller 1904 ( 6.4,7.2)(0,0)@xmath49 tomaschek 1924 ( 6.6,5.5)(0,0)@xmath49 kennedy 1926 ( 6.7,6.8)(0,0)@xmath49 illingworth 1927 ( 7,6.5)(0,0)@xmath49 joos 1930 ( 9.5,4)(0,0)@xmath49 essen 1955 ( 9.8,3.2)(0,0)@xmath49 cedarholm et al 1958 ( 10.4,3.2)(0,0)@xmath49 jaseja et al 1964 ( 11.9,2.5)(0,0)@xmath49 brillet & hall 1979 ( 14.3,-1)(0,0)@xmath49 optis ( proposed ) for the hydrogen maser @xmath50 ( gp - a experiment , @xcite ) , verifying the gravitational red shift .
for a test of the universality of the gravitational red shift by means of a comparison between a microwave cavity and an atomic cesium clock , the best result is @xmath51 @xcite .
the advantages of a sattelite mission for doing experiments for testing sr and gr are the following : * the high orbital velocity ( see eq.([genansatz ] ) .
for the proposed ( elliptic ) orbit , @xmath14 varies between @xmath52 and @xmath53 over half the orbit period @xmath54 .
this value is 20 times larger than the change of the velocity of the earth s surface over a period of 12 h. * the shorter period @xmath55 of the velocity modulation as compared to 24 h on earth .
it relaxes the requirements on the optical resonators and is directly relevant for the kt and lpi tests .
it is indirectly useful also for the elimination of systematic effects in the mm test . *
the difference of the gravitational potential in the highly eccentric orbit is @xmath56 .
this is about three orders of magnitude larger than the difference of the potential of the sun which an earth bound observer experiences .
this is relevant for the lpi test . *
the microgravity environment minimizes distortions of the optical resonators .
this is relevant for the mm test . * a variable spin frequency of the satellite permits elimination of sytematic effects .
* long integration times ( longer than 6 months ) .
in this and the following section we discuss the science payload of the satellite , the relevant performance specifications and the resulting requirements for the orbit and the satellite bus structure . in order to perform the mm test , the satellite has to spin around its axis .
a typical rotation period is @xmath57 . for the elimination of systematic errors
, the rotation frequency @xmath1 can be varied . for the kt test
the timescale is the orbit time @xmath58 .
the main subsystems of the experimental payload are optical resonators , ultra stable lasers , an optical frequency comb generator and an ultra stable microwave oscillator .
these components are interconnected and supplemented by the locking and stabilization electronics , the optical bench , the drag free control system ( discussed in the next section ) , and an advanced thermal control system . the lisa pre phase a report @xcite showed that a 3stage passive thermal isolation is sufficient to achieve a level of thermal fluctuations below @xmath59 at @xmath60 .
this performance ( assuming no shadow phases ) would be sufficient for the optis random noise requirements . to supress the fluctuations correlated with the rotation of the satellite , an improvement of the thermal stability by one order of magnitude would be required .
this could be achieved either by an additional stage of passive isolation , or by adding an active temperature stabilization .
the optical resonators are the central part of the experimental setup . in the baseline configuration ,
two crossed standing wave resonators are implemented by optically contacting four highly reflective mirrors to a monolithic spacer block with two orthogonal holes made from a low expansion glass ceramic ( ule , zerodur ) . with a technically feasible finesse of @xmath61 and a length of @xmath62 each
, these resonators should exhibit linewidths of @xmath63 .
low expansion glass ceramics are designed for a minimal thermal expansion coefficient ( @xmath64 ) at room temperature and can be manufactured in many different shapes and dimensions .
resonators made from these materials are well suited for laser stabilization , although aging effects cause a continuous shrinking of the material and thereby frequency drifts of typically @xmath65 .
the monolithic construction of the resonator block serves to strongly reduce the effect of shrinking and of temperature fluctuations on the mm experiment : a high degree of common mode rejection ( two orders of magnitude ) in the differential frequency measurement of the lasers locked to the resonators is expected . for the kt and the redshift experiment ,
on the other hand , the aging related frequency drift is critical .
it has to be modelled well enough to keep the unpredictable residual part below @xmath66 ( i.e. @xmath67 ) over the signal half period @xmath68 .
this is feasible @xcite . for the projected measurement sensitivity the temperature stability requirements are as follows :
a level @xmath69 for random fluctuations over the spin period time scale and a level @xmath70 for a temperature modulation correlated with the rotation of the satellite . for the kt as well as for the redshift experiment
the requirements are @xmath71 for random temperature fluctuations on a time scale of the orbit period and @xmath72 for temperature modulation correlated with the orbital motion .
the length of the reference resonators is affected by accelerations , with a typical sensitivity of @xmath73 for a resonator length of @xmath74 @xcite .
this leads to a requirement of @xmath75 for random residual accelerations and @xmath76 for residual accelerations correlated with the rotation of the satellite .
the drag - free control system ( see next section ) is designed to meet these requirements .
the lasers used for the optis mission should have high intrinsic frequency stability , narrow linewidth and high intensity stability .
these requirements are best fulfilled by diode pumped monolithic nd : yag lasers , which are also used in gravity wave detectors ( geo600 , ligo , virgo , lisa ) and many other high precision experiments @xcite .
such lasers are already available in space qualified versions . locking the lasers to the reference resonators using the pound drever
hall frequency modulation method requires fast , low noise photodetectors and an optimized electronic control system .
the intrinsic noise of the photodetectors has to be sufficiently low to allow shot noise limited detection .
the residual amplitude fluctuations of the lasers have to be actively suppressed . to prevent thermally induced length changes of the resonators by absorbed laser radiation
, the intensity of the laser beams has to be actively stabilized to a relative level of @xmath77 in the frequency range @xmath78 to @xmath79 .
the whole optical setup should be stable and isolated from vibrations to prevent frequency fluctuations caused by vibration induced doppler shifts .
even very small displacements ( less then @xmath80 ) of the laser beams relative to the reference resonators are known to cause substantial frequency shifts @xcite .
these requirements can be fulfilled by using a well designed monolithic optical bench on which all optical components are stably mounted as close together as possible .
the kt and the universality of red shift experiments require an independent frequency reference .
this reference should be based on the difference of two energy levels in an atomic or molecular system .
atomic clocks based on hyperfine transitions in cesium or rubidium atoms are suited for this task .
they are available in space qualified versions with relative instabilities of better than @xmath81 for the relevant time scale @xmath82 of several hours .
the atomic clock can also serve as the reference for the microwave synthesizer required to mix down the beat signal between the two stabilized lasers in the mm experiment from typically 1 ghz to a lower frequency for data acquisition and analysis .
the required relative instability @xmath83 on the timescale @xmath84 of @xmath85 is thereby satisfied . for the frequency comparison between the atomic clock and the lasers stabilized to the reference
resonators it is necessary to multiply the microwave output of the atomic clock ( @xmath86 ) into the optical range ( @xmath87 ) .
thanks to recent important progress in the field of frequency metrology , this can now be done reliably amd simply by using femtosecond optical comb generators @xcite . here
the repetition rate of a mode locked femtosecond laser of @xmath88 is locked to an atomic clock .
its optical spectrum ( a comb of frequencies spaced at exactly the repetition rate ) is broadened to more than one octave by passing the pulses through a special optical fiber . by measuring and stabilizing the beatnote between the high frequency part of the comb and the frequency doubled low frequency part , it is then possible to determine the absolute frequency of each component of the comb relative to the atomic clock . in a final step the frequency of the cavity stabilized lasers
is then compared to the closest frequency component of the comb by measuring their beatnote with a fast photodetector .
compact diode pumped comb generators with very low power consumption , as required for space applications , are already under development .
the experimental requirements define an optimum mission profile . in particular , the requirements of attitude control , maximum residual acceleration , and temperature stability result in the following specifications : * the satellite needs a drag - free attitude and orbit control system for all 6 degrees of freedom .
thrust control must be possible down to @xmath89n by means of ion thrusters ( field emission electical propulsion , feep ) . * for drag - free control
the satellite needs an appropriate reference sensor . *
feeps are not effective for orbit heights less than 1000 km , because of the high gas density in lower orbits . * to avoid charging of the capacitive reference sensor by interactions with high energy protons , highly eccentric orbits , where the satellite passes the van
belt , are not appropriate .
* the kt experiment requires a low orbit , because the experimental resolution depends directly on the satellite orbit velocity . *
the precise attitude control requires a star sensor with a resolution of 10 arcsec . * mechanical components for attitude control ( e.g. fly wheel or mechanical gyros )
can not be used , because of the sensitivity of the experiment to vibrations .
* temperature stability requirements during integration times of more than 100 s can only be realized on orbits without or with rare eclipse intervals .
* the thermal control of the satellie structure must achieve a stabilioty of @xmath90 .
* the mission time is 6 months minimum .
optis is designed to be launched on a micro
satellite with limited technological performance . considering all experimental requirements , technological feasibility , launch capability , and design philosophy
the most feasible solution is to launch the satellite in a high elliptical orbit , attainable via a geo transfer orbit ( gto ) by lifting the perigee . in this scenario ,
the satellite is first launched as an auxiliary payload ( asap5 ) by an ariane 5 rocket into the gto with an apogee of 35800 km and a perigee of only 280 km .
an additional kick motor on the satellite will lift the satellite in its final orbit with a perigee of about 10000 km , corresponding to @xmath91 .
a minimum height of 10000 km enables one to use ion thrusters ( feeps ) and avoids flying through the van
allen belt . also , the orbit eccentricity of @xmath92 is high enough to attain sufficient velocity differences for the kt experiment .
although the relatively high orbit reduces the in orbit velocity of the satellite by a factor of 2.8 compared to a low earth circular orbit , it is still 20 times higher than for an earth based experiment .
the satellite has cylindrical shape , with a height and a diameter of about 1.5 m. it spins around its cylindrical axis which is always directed to the sun .
the front facing the sun is covered by solar panels . behind the front plate serving as thermal shield a cylindrical box for on
board electronics and thruster control is located .
the experimental box whose temperature is actively controlled is placed below the electronics .
a stringer structure below the experimental box carries the kick motor and the fuel tank .
4 clusters of ion thrusters for fine ( drag free ) attitude control as well as 3 clusters of cold gas thrusters for coarse attitude control and first acquisition operation are mounted circumferentially .
the structure elements of the entire bus have to satisfy the extreme requirement for passive thermal control .
their thermal expansion coefficient has to be less than @xmath93 .
the total satellite mass is about 250 kg including 90 kg of experimental payload .
the total power budget is estimated to be less than 250 w. during experimental and safe mode coarse attitude and orbit control are based on a sun sensor ( 1 arcsec resolution ) and a star sensor ( 10 arcsec resolution ) .
fiber gyros are used for spin and de spin maneuvers and serve to control the cold gas thrusters .
the fine attitude control , also called the drag free control , must be carried out with an accuracy of @xmath94 within the signal bandwidth of @xmath95 to @xmath96 hz , depending on the satellite s spin rate .
the general principle of drag free control is to make the satellite s trajectory as close as possible a geodesic .
therefore , a capacitive reference sensor @xcite is used .
the sensor unit consists of a test mass whose movements are measured capacitively with respect to all 6 degrees of freedom . apart from the sensing electrodes , electrodes for servo
control surround the test mass and compensate its movement relative to the satellite structure .
the signal is also used to control the satellite s movement via the ion thrusters .
thus , the test mass falls quasi freely and is shielded against all disturbances by the satellite , in particular against solar pressure and drag . because the servo - control is influenced by back
action effects , the test mass and the servo
control form a spring mass system whose spring constant and eigenfrequnecy must be adapted to the signal bandwidth .
therefore , electrode surfaces , charging by external sources as well as the precision of the test mass and the electrode alignments influence the measurement of @xmath97 directly and make necessary repeated in
orbit calibration maneuvers @xcite .
the chosen orbit avoids charging effects as much as possible .
feep ion thrusters ( field emission electric propulsion ) must be used to overcome ( 1 ) solar radiation pressure acting on the satellite and disturbing its free fall behaviour , and ( 2 ) to control the residual acceleration down to @xmath98 in the signal bandwidth .
the first requirement sets an upper limit for the thrust : linear forces acting on the satellite are less than @xmath99 in all 3 axes and maximum torques are about @xmath100 .
the second condition determines the resolution of thrust control which have to be done with an accuracy of about @xmath101 .
a feep able to satisfy these requirements is the indium liquid metal ion source ( lmis ) of the austrian research centers seibersdorf ( arcs ) @xcite .
thrust is produced by accelerating indium ions in a strong electrical field .
a sharpened tungsten needle is mounted in the centre of a cylindrical indium reservoir bonded to a ceramic tube which houses a heater element for melting the indium .
ion emission is started by applying a high positive potential between the tip covered with a thin indium film and an accelerator electrode . to avoid charging a neutralizer emitting electones
is installed .
the feeps have a length of 2 cm , a diameter of 4 mm , and weigh only several grams sufficient for the entire mission .
a continuous thrust of @xmath102 per thruster is available . for a satellite diameter of about 1.5 m ( the maximum asap 5 size ) , the solar radiation pressure of ca . @xmath103 and the radiation pressure of the earth albedo of @xmath104 sum up to a total drag of about @xmath105 .
considering thruster noise , misalignments and other disturbing effects , a continuous thrust of @xmath106 would be sufficient .
to control all 6 degrees of freedom a minimum of 3 clusters of 4 thrusters is necessary . to guarantee a continuous thrust with some redundancy 4 clusters are desirable .
the power consumption is less than 3 w on average with peaks up to 25 w.
the proposed optis mission is capable to make considerable improvements , up to three orders of magnitude , in the tests of sr and gr .
it is designed to be a low cost mission .
it is based on using ( i ) recent laboratory developments in optical technology and ( ii ) the advantages of space conditions : quit environment , long integration time , large velocities and large potential differences .
the optical technology includes an optical comb generator , stabilized lasers and highly stable cavities .
we remark that an alternative optical cavity system consisting of a monolithic silicon block operated at @xmath107 , the temperature where the thermal expansion coefficient vanishes , should be studied @xcite .
the advantage of this approach could be a significantly reduced level of creep and therefore a corresponding improvement of the kt test . | a new satellite based test of special and general relativity is proposed . for the michelson
morley experiment we expect an improvement of at least three orders of magnitude , and for the kennedy thorndike experiment an improvement of more than one order of magnitude .
furthermore , an improvement by two orders of the test of the universality of the gravitational red shift by comparison of an atomic clock with an optical clock is projected .
the tests are based on ultrastable optical cavities , an atomic clock and a comb generator . |
after the recent discovery of pentaquark states @xcite the central question to be addressed now concerns the structure of these new baryons .
they could be : a ) uncorrelated quarks inside a bag @xcite ; b ) a @xmath2 molecule bound by a van - der waals force @xcite ; c ) a `` @xmath2 '' bound state in which @xmath3 and @xmath4 are not separately in color singlet states @xcite ; d ) a diquark - triquark @xmath5 bound state @xcite and e ) a diquark - diquark - antiquark ( dda ) state .
this last one was advanced by jaffe and wilczek ( jw ) @xcite and is quite appealing because it can explain two unusual features of pentaquarks : their small mass and decay width .
instantons generate strong attractive quark - quark interactions with the formation of low mass diquarks , which , in turn , lead to relatively low mass pentaquarks .
this was verified in @xcite , @xcite and @xcite . moreover , in @xcite , it was shown that the dda configuration may lead to strongly suppressed transition amplitudes ( to meson - baryon states ) for a reasonable choice of its spatial wave function , namely , two separated extended diquark balls overlapping only partially and the antiquark at the center of the system .
pentaquark configurations can be implemented in qcd sum rules ( qcdsr ) @xcite and in lattice qcd @xcite by a proper choice of the interpolating current . a current for configuration c )
has been proposed by zhu @xcite and for configuration e ) three different currents were proposed in @xcite , @xcite and @xcite . given the impulse of this field and the prospects of new measurements
, we may expect that the `` pentaquark wave '' will last still for a long time . a great effort will be devoted to understand the structure of these objects . for the qcdsr community ( and for lattice
qcd studies as well ) this means that more attention will have to be given to the properties of the interpolating currents . at this point , we can already compare the three calculations presented in @xcite , @xcite and @xcite for the mass of the @xmath6 .
although they present different implementations of the dda scheme , they produce very similar results for the pentaquark mass .
this indicates that a more careful analysis has to be performed .
the purpose of the present work is twofold .
we will use qcdsr to calculate the mass of the recently measured @xmath0 and we will also take the opportunity to perform a careful comparison of the results obtained with two different currents , which we will call i and ii .
current i was introduced by us in @xcite and current ii is the one proposed in @xcite . as it will be seen , both currents give approximately the same mass for the @xmath0 but have different sensitivity to the continuum threshold parameter , different convergence of the operator product expansion ( ope ) and also different strength of the pole , with respect to the continuum .
this comparative study , which will later be extended to the currents suggested in @xcite and also in @xcite , will give us a better understanding of these currents and may lead to the choice of the best current .
in the qcdsr approach @xcite , the short range perturbative qcd is extended by an ope expansion of the correlators , which results in a series in powers of the squared momentum with wilson coefficients .
the convergence at low momentum is improved by using a borel transform .
the expansion involves universal quark and gluon condensates .
the quark - based calculation of a given correlator is equated to the same correlator , calculated using hadronic degrees of freedom via a dispersion relation , providing sum rules from which a hadronic quantity can be estimated .
the qcdsr calculation of hadronic masses centers around the two - point correlation function given by ( q)id^4 x e^iqx t(x)(0 ) , [ cor ] where @xmath7 is an interpolating field ( a current ) with the quantum numbers of the hadron we want to study . following the diquark - diquark - antiquark scheme ,
we can write two independent interpolating fields with the quantum numbers of @xmath0 : _ 1(x)=1_abc ( d_a^t(x ) c _ 5 s_b(x ) ) [ d_c^t(x ) c _ 5 s_e(x ) + s_e^t(x ) c _ 5 d_c(x ) ] c|u^t_e(x ) [ eta1 ] _
2(x)=1_abc ( d_a^t(x ) c s_b(x ) ) [ d_c^t(x ) c s_e(x ) + s_e^t(x ) c d_c(x ) ] c|u^t_e(x ) [ eta2 ] where @xmath8 and @xmath9 are color index and @xmath10 is the charge conjugation operator . as in the nucleon case ,
where one also has two independent currents with the nucleon quantum numbers @xcite , the most general current for @xmath0 is a linear combination of the currents given above : ( x)= , [ cur ] with @xmath11 being an arbitrary parameter . in the case of the nucleon , the interpolating field with @xmath12 is known as ioffe s current @xcite . with this choice for @xmath11 ,
this current maximizes the overlap with the nucleon as compared with the excited states , and minimizes the contribution of higher dimension condensates . in the present case
it is not clear a priori which is the best choice for @xmath11 .
this will be studied in what follows .
we also employ the following interpolating field operator for the pentaquark @xmath0 @xcite : @xmath13 it is easy to confirm that this operator produces a baryon with @xmath14 , @xmath15 and strangeness @xmath16 .
the parts , @xmath17 and @xmath18 , give the scalar @xmath19 ( @xmath20 ) and the pseudoscalar @xmath21 ( @xmath22 ) @xmath23 diquarks , respectively .
they both belong to the anti - triplet representation of the color su(3 ) .
the scalar diquark corresponds to the the @xmath24 @xmath25 quark with zero angular momentum .
it is known that a gluon exchange force as well as the instanton mediated force commonly used in the quark model spectroscopy give significant attraction between the quarks in this channel .
inserting eq .
( [ cur ] ) into the integrand of eq .
( [ cor ] ) we obtain t(x)(0)= t^2_11(x ) + t(_12(x)+_21(x))+_22(x ) calling @xmath26 and @xmath27 we get _ ij(x)= t_i(x)_j(0)= _
abc_abccs_ee^t(-x)c\ { - . + + + + - + - +
+ } , [ expa ] where @xmath28 and @xmath29 are the light and strange quark propagators respectively . in order to evaluate the correlation function @xmath30 at the quark level ,
we first need to determine the quark propagator in the presence of quark and gluon condensates . keeping track of the terms linear in the quark mass and taking into account quark and gluon condensates ,
we get @xcite s_ab(x)&= & t[q_a(x)_b(0)]=i_ab2 ^2x^4-m_q_ab4 ^ 2x^2-i32 ^ 2x^2t^a_ab g^a _ ( ^+^ ) + & -&_ab12 -m_q32 ^ 2t^a_abg^a_^(-x^2 ) + i_ab48m_q - x^2_ab 2 ^ 63 + & + & ix^2_ab2 ^ 73 ^ 2m_q - x^4_ab2 ^ 103 ^ 3 , [ prop ] where we have used the factorization approximation for the multi - quark condensates , and we have used the fixed - point gauge @xcite . inserting ( [ prop ] ) into ( [ expa ] ) we obtain a set of diagrams which contribute to the ope side of the correlation function . in the case of the current i , fig .
1 shows the classes of diagrams which we are considering . in each diagram , except the perturbative one , all possible permutations of the lines are implicit .
lorentz covariance , parity and time reversal imply that the two - point correlation function in eq .
( [ cor ] ) has the form ( q)= _ 1(q^2)+ _ q ( ) .
[ stru ] a sum rule for each scalar invariant function @xmath31 and @xmath32 , can be obtained . as in ref .
@xcite , in this work we shall use the chirality even structure @xmath33 to obtain the final results but we shall also discuss the chirality odd structure @xmath31 , showing its limitations for our present purposes .
the phenomenological side is described , as usual , as a sum of pole and continuum , the latter being approximated by the ope spectral density . in order to suppress the condensates of higher dimension and at the same time reduce the influence of higher resonances we perform a standard borel
transform @xcite : ( m^2 ) _ n , q^2 ( q^2)^n+1 ( - ) ^n ( q^2 ) [ borel ] ( @xmath34 ) with the squared borel mass scale @xmath35 kept fixed in the limit .
for current ii we repeat the steps mentioned above , substituting ( [ eq : if ] ) into ( [ cor ] ) , making use of the expansion ( [ prop ] ) , picking up the terms multiplying the structure @xmath36 and finally performing a borel transform ( [ borel ] ) . in this case the diagrams that have to be considered are shown in fig .
after borel transforming each side of @xmath37 and transferring the continuum contribution to the ope side we obtain the following sum rule at order @xmath38 : @xmath39 where @xmath40 refers to the current employed and the spectral densities , up to order 6 are given by : @xmath41 with @xmath42 , @xmath43 , @xmath44 , @xmath45 , @xmath46 , @xmath47 mev and @xmath48 to extract the @xmath0 mass , @xmath49 , we take the derivative of eq .
( [ soma ] ) with respect to @xmath50 and divide it by eq .
( [ soma ] ) . repeating the same steps leading to ( [ soma ] ) , ( [ rho1 ] ) and ( [ rho2 ] ) for the chirality odd structure @xmath51 we arrive at @xmath52 where
@xmath53 and @xmath54 in the complete theory , the mass extracted from the sum rule should be independent of the borel mass @xmath55 .
however , in a truncated treatment there will always be some dependence left .
therefore , one has to work in a region where the approximations made are supposedly acceptable and where the result depends only moderately on the borel variables .
before proceeding with the numerical analysis , a remark is in order .
a comparison between results obtained with different currents is more meaningful when they describe the same physical state , i.e. , those with the same quantum numbers . concerning spin ,
all currents considered in our work have the same spin (= 1/2 ) . concerning the parity , the situation is more complicated . in qcd sum rules ,
when we construct the current , it has a definite parity .
current i has parity @xmath56 and current ii has parity @xmath57 .
however , currents can couple to physical states of different parities . as well discussed in @xcite , in order to know the parity of the state in qcdsr , we have to analyze the chiral - odd sum rule .
if the r.h.s of this sum rule ( containing the spectral density coming from qcd ) is positive , then the parity of the corresponding physical state is the same as the parity of the current .
if it is negative the parity is the opposite of the parity of the current . performing this analysis
we might determine in both cases the parity of the state .
however it turns out that for both currents , in the chiral - odd sum rule the ope does not have good convergence , i.e. , terms containing higher order operators are not suppressed with respect to the lowest order ones .
this sum rule is thus ill defined and nothing can be said about the parity of the state .
the comparative study of the currents is still valid because we can compare other properties of these currents
. a possible outcome of this study might be that one current has defects , which are so severe that we are forced to abandon it .
if this turns out to be case , the determination of the parity of the associated states becomes irrelevant .
having presented the main formulas in the last section , the next step will be to introduce numerical values for the masses and condensates , choose reasonable values for the free ( or partially constrained ) parameters , which are the continuum threshold , the borel mass at which the mass sum rule is evaluated and , in the case of current i , the value of @xmath11 . as a result
we obtain values for @xmath49 . in doing these calculations we must remember that _ it is not enough to obtain a pentaquark mass consistent with the experimental number_. there is a list of requirements that must be fulfilled : + i ) the physical observables , such as masses and coupling constants , must be approximately independent of the borel mass ( this is the so called borel stability ) .
+ ii ) the right hand side ( rhs ) of the sum rule ( [ soma ] ) must be positive , since the left hand side ( lhs ) is manifestly positive .
+ iii ) the operator product expansion ( ope ) must be convergent , i.e. , the terms appearing in ( [ rho1 ] ) and ( [ rho2 ] ) must decrease with the increasing order of the operator .
+ iv ) the pole contribution must be dominant , i.e. , the integral in ( [ soma ] ) must be at least 50 % of the integral over the complete domain of invariant masses ( @xmath58 ) .
+ v ) the threshold parameter @xmath59 must be compatible with the energy corresponding to the first excitation of an usual baryon .
+ vi ) the current - state overlap @xmath60 must be sizeable .
the larger is @xmath60 th better is the current .
as it will be seen the above conditions are not easy to be simultanously satisfied and may be useful in discriminating between different currents . in the numerical analysis of the sum rules ,
the values used for the condensates are : @xmath61 , @xmath62 , @xmath63 with @xmath64 and @xmath65 .
we define the continuum threshold as : s_0 = ( 1.86+)^2 ^2 [ s0 ] we evaluate our sum rules in the range @xmath66 .
the results are shown in table i. the first column in table i gives the values of @xmath11 considered here .
we notice that some values were excluded , as for example @xmath67 , because they would lead to a violation of condition ii ) above .
the simbol s stands for a current composed by scalar diquarks only , i.e. , @xmath68 . the motivation for studying this current is to verify if it gives a smaller mass for the pentaquark than those obtained with other currents . according to the instanton description of diquark dynamics
, this should be the case . in the second and third columns of table i we list the values adopted for the strange quark mass and continuum threshold respectively .
the fourth column shows @xmath49 and the borel mass squared at which the particle mass was extracted .
the fifth and sixth columns show the ratios of terms containing dimension 4 ( quark and gluon condensates ) and 6 ( mixed condensates ) operators and the perturbative term respectively .
ideally these numbers should be smaller than one , the second being smaller than the first .
finally , the last column shows the strength of the pole . in practice
this quantity is obtained dividing the integral in the right hand side of ( [ soma ] ) by the same integral with the upper limit @xmath69 .
this percentage must be as large as possible , but usually we accept values around 50 % .
as it can be seen in table i , it is hard to satisfy conditions i ) - vi ) simultaneously . in particular
, when we have a very good ope convergence , the strength of the pole is very weak and vice versa .
we have to look for a compromise and we believe that the choices indicated in bold face are the best . for same choices for @xmath38 and @xmath70 we observe that the masses found with scalar diquark currents , s , are only slightly smaller than the others .
this means either that instanton dynamics was not captured by our choice of currents and diagrams or that the interaction between the pseudoscalar diquarks ( included in the mixed currents ) is more attractive than expected .
the connection between our approach and instantons deserves further investigation . in order to further illustrate these results
, we consider the parameters of the first line in bold face , @xmath71 gev and @xmath72 gev , and contruct , figures 3 , 4 and 5 , showing the borel mass dependence of @xmath49 , of the ope terms ( in absolute value ) and of the percentage of the pole contribution respectively .
for these choices the value of the current - state overlap is : ^q _ i 5.4 10 ^ -9 [ lambnos ] in view of the variety of results presented in table i and in the figures , it would be somewhat artificial to invent a procedure to determine the average value of the @xmath73 mass and its corresponding error .
if we simply take the average over all values presented we arrive at : m _ = 1.87 0.22 [ mia ] restricting ourselves to the parameter combinations which satisfy the requirements i ) - vi ) and taking the average we obtain : m _ = 1.85 0.05 [ mib ] these numbers are very close to the experimental value .
using the same numerical inputs quoted in the last subsection we evaluate now the sum rules obtained with current ii .
the results are shown in table ii , which has the same format as table i except for the fact that , now , there is no @xmath11 parameter .
the same comments made for the results of table i apply here . choosing the parameters of the first line in bold face , @xmath71 gev and @xmath74 gev , we present in figures 6 , 7 and 8 the borel mass dependence of @xmath49 , of the ope terms and of the percentage of the pole respectively .
as it can be seen , with current ii we tend to overestimate @xmath49 , unless very low threshold parmeters or quark masses are used .
besides , the pole contribution is always smaller than 40 % . on the other hand , the borel stability seen in fig
6 is remarkable .
this suggests that we could choose a lower value for the borel mass , thereby increasing the pole contribution without significantly changing @xmath49 . in order to extract from table ii
some value for the pentaquark mass , we repeat the procedure applied in the last subsection .
the simple average gives : m _ = 2.00 0.16 [ miia ] and the average over the best parameter choices leads to : m _ = 1.88 0.04 [ miib ] these numbers are somewhat high but still compatible with data .
finally , for these parameters the current - state overlap is : ^q _ i 1.3 10 ^ -9 [ lamboka ] comparing ( [ lambnos ] ) and ( [ lamboka ] ) we observe that the coupling of current i to the the @xmath73 state is four times larger than the coupling of current ii to this state .
this speaks in favor of current i.
in conclusion , we have presented a qcd sum rule study of the @xmath0 pentaquark mass using a diquark - diquark - antiquark scheme .
we have employed two possible currents ( i and ii ) which implement this picture .
several parameter choices were considered for the currents .
we could conclude that for both currents it was possible to find a set of parameters and a borel window so that conditions i ) - vi ) were fulfilled and a reasonable value for the mass was found .
the ratio pole / total , presented in the tables is still small and probably is the most important drawback of the studied currents . at the present stage
we may say that the pentaquark @xmath0 can be reasonably well described by qcd sum rules and the value of its mass is given by ( [ mib ] ) or by ( [ miib ] ) .
these numbers can be considered the same within the uncertainties .
concerning the currents , there are differences in the borel stability , in the sensitivity to the strange quark mass , in the continuum threshold and in the their couplings to the @xmath73 state . | in a diquark - diquark - antiquark picture of pentaquarks , we use two interpolating currents to calculate the mass of the recently measured @xmath0 state in the framework of qcd sum rules .
we show that , even though yielding similar values for @xmath1 ( and close to the experimental value ) , these currents differ from each other in what concerns the strength of the pole , convergence of the ope and sensitivity to the continuum threshold parameter . |
fusion of two atomic nuclei is possible if they approach each other to a short distance ( @xmath1 cm ) . to come that close , they need to go through a coulomb barrier of the height of few mev .
the penetration probability for such a barrier at room temperature ( energy of relative motion @xmath2mev ) is practically zero ( @xmath3 @xcite ) , but this probability rapidly grows when the kinetic energy of the nuclei increases .
for example , for the @xmath4 system at the energy of 30kev the penetration probability becomes @xmath5 @xcite .
therefore , the obvious way to fuse the nuclei is to raise the temperature of their mixture . in this way
the so called thermo - nuclear reactions happen in the stellar bodies , in nuclear weapons , and in the tokamak@xcite .
+ alternatively , the nuclei ( which we want to fuse ) can be put in a bound system such as a molecule , where they sit next to the barrier relatively long . in such a case , even at zero relative energy , the nuclei are separated by a thinner barrier . for example , the two deuterons sitting at rest near each other at a distance of @xmath6 , have to overcome the same barrier as two free moving deuterons with relative energy of @xmath7ev ( which is equivalent to the temperature of @xmath8 ) .
+ this , however , does not help much . for a deuterium molecule @xmath9 in its ground state
the penetration probability is still too low , namely , @xmath10 @xcite and the fusion rate is @xmath11 @xcite .
these numbers can be significantly increased if we make the size of the molecule smaller .
this can be achieved if one electron in the molecule is replaced with the muon , which is approximately 200 times heavier . as a result ,
the nuclei find themselves at a distance that is 200 times smaller @xcite . for the deuterons ,
the thickness of the barrier becomes the same as at the collision energy almost equal to 3kev ( this corresponds to the temperature of @xmath12 ) .
+ this is already a significant gain .
the penetration probability for the muonic molecule @xmath13 is @xmath14 @xcite .
with such a probability , one muon can help to fuse , i.e. can catalyze the fusion of many nuclear pairs before it decays ( muon lifetime is @xmath15s ) .
the muon - catalyzed fusion has been observed and well studied both experimentally and theoretically ( see , for example , refs.@xcite ) , but turned out to be inefficient as a new source for the energy production .
+ in the present paper , we suggest and experimentally explore yet another possible approach to fusion of light nuclei . the idea is to make a crystal out of the atoms
whose nuclei we want to fuse . in this crystal ,
the nuclei sit next to each other at an atomic distance and oscillate around the equilibrium positions .
a crystal is just a huge molecule and of course the probability of spontaneous fusion of neighbouring nuclei is negligible , the same as in the ordinary molecules .
an experiment aimed to observe the spontaneous fusion in the lithium deuteride crystal put an upper bound on the fusion rate per nuclear pair as @xmath16 @xcite .
+ however , we can try to shake the crystalline lattice by an external force ( electromagnetic wave , for example ) .
the nuclei swayed around their equilibrium positions , acquire kinetic energy relative to each other . as a result , the original boltzmann distribution of the nuclei over the oscillation levels is changing and the higher levels are populated . from these higher energy levels
the nuclei can tunnel through the coulomb barrier and fuse .
this is what we observed when irradiated the lid crystal with the x - rays .
the rate of such a fusion turned out to be very low ( one event every half an hour ) , but still easily measurable .
the nuclei we want to fuse , are the isotopes of hydrogen and lithium , namely , @xmath17 and @xmath18 .
their compound nucleus , @xmath19 , has no stable states .
the threshold energies for its spontaneous disintegration in various pairs of fragments are shown in fig .
[ fig.8be_spectrum ] ( the data are taken from ref .
@xcite ) .
as is seen , the @xmath20-@xmath18 threshold lies @xmath21 above the ground state ( which is also unstable ) . and
this is above the thresholds for all the other two - body channels , namely , @xmath22 , @xmath23 , @xmath24 , @xmath25 , and @xmath26 .
+ when sitting at the nodes of a crystalline lattice , the nuclei oscillate around their equilibrium positions , but the oscillation energy is negligible on the nuclear scale .
they therefore can be considered as being at rest relative to each other .
in other words , the nuclear pair @xmath27 in a crystal is practically at the threshold energy .
+ if @xmath20 and @xmath18 in the crystal , overcome huge coulomb barrier and fuse , then there is no way back for them .
indeed , for the resulting excited ( resonant ) state @xmath28 to decay back into the channel @xmath29 , the deuteron and lithium nuclei must overcome the same huge coulomb barrier , while they have practically zero relative kinetic energy .
it is much more easier to decay into one of the channels whose thresholds are below and where the kinetic energy is above the coulomb barrier .
for example , in the channel @xmath30 ( which lies @xmath31 below the ground state of the compound nucleus ) the relative kinetic energy is greater than in the @xmath27 channel by the amount of @xmath32 , i.e. is always well above the barrier .
+ if we restrict our model consideration to only the two - body channels , then our 8-body problem involves six - channels , namely , @xmath33 t+{}^5\mathrm{li}+0.5928\,\mathrm{mev}\\[3 mm ] { } ^3\mathrm{he}+{}^5\mathrm{he}+0.9008\,\mathrm{mev}\\[3 mm ] n+{}^7\mathrm{be}+3.3811\,\mathrm{mev}\\[3 mm ] p+{}^7\mathrm{li}+5.0257\,\mathrm{mev}\\[3 mm ] \alpha+\alpha+22.3726\,\mathrm{mev } \end{array } \right.\ ] ] and all of these channels are open .
in other words , if @xmath20 and @xmath18 fuse , the final outcome would be one of the pairs on the right hand side of eq .
( [ channels ] ) with the energy release shown for each channel .
+ around the @xmath34-threshold , the compound nucleus @xmath19 has several very wide and overlapping meta - stabele states ( resonances ) .
the one that is located most closely to the threshold , has the quantum numbers : ( total spin and parity ) @xmath35 , ( isospin ) @xmath36 , @xmath37 , and @xmath38 @xcite .
this wide state is shown in fig .
[ fig.8be_spectrum ] as a strip of verdical bars .
+ recent analysis of experimental data , based on the @xmath39-matrix parametrization @xcite , showed that at the near - threshold energies the inelastic collision of @xmath20 and @xmath18 mainly leads to formation of that resonance @xmath40 .
the authors of ref .
@xcite also showed that this resonance predominantly decays into the @xmath41-channel .
they found that the partial width for such a decay is @xmath42 , which gives @xmath43 .
in other words , after its formation , this resonance decays in the @xmath41-channel with the probability of 96% .
+ ignoring the remaining 4% , we can assume that , if the fusion of our @xmath20 and @xmath18 happens , almost the only outcome is the @xmath41 pair , @xmath44 the resulting @xmath45-particles equally share the energy and momentum .
therefore the fusion event could be identified by detecting at least one of the two @xmath45-particles moving in the opposite directions with the energies of @xmath46 .
nuclear reactions at extremely low energies ( @xmath47 ) are significantly suppressed by the repelling coulomb forces . the probability @xmath48 that the colliding nuclei tunnel through the coulomb barrier is equal to the ratio @xmath49 , where @xmath50 is the wave function of their relative motion at the distance @xmath51 , and @xmath52 is the classical turning point .
it can be shown ( see eq .
( 134.10 ) of ref .
@xcite ) that for a pure coulomb potential , the ratio of @xmath53 to absolute square of the plane wave is given by @xmath54 where @xmath55 is the sommerfeld parameter that involves the nuclear charges , @xmath56 and @xmath57 , and the reduced mass @xmath58 of the nuclear pair . for the purpose of estimating
, we can assume that for distances @xmath59 the relative motion of the nuclei in the crystalline cell is described by a plane wave , and thus eq .
( [ psipsiratio ] ) gives us the penetration probability for the coulomb barrier .
+ in our problem , the nuclei are confined to finite volumes of space in the crystal .
within its cell , a nucleus moves to and fro , periodically colliding with the barriers .
each of these collisions is an attempt to tunnel through . if the size of the cell is @xmath60 and velocity of the nucleus is @xmath61 , then the attempts are repeated with the period @xmath62 , i.e. with the frequency @xmath63 .
+ therefore , the number of transitions through the barrier per second ( i.e. the transition rate ) is @xmath64 . as we mentioned in sec .
[ sec.fusionchannels ] , if the deuteron ( or @xmath18 ) manages to pass to the other side of the barrier , there is no way back ( for that , it needs to tunnel once more through the same barrier , while the penetration probability is very small ) .
this means that if the penetration takes place , the nuclear system ends up in one of the inelastic channels given by eq .
( [ channels ] ) . among these channels ,
the reaction ( [ dli6to2alpha ] ) has the highest probability , @xmath65 . therefore , if a deuteron in the crystal oscillates with the energy @xmath66 , the reaction ( [ dli6to2alpha ] ) happens with the rate @xmath67 where @xmath68 is the mass of deuteron .
apparently , the same is valid for the tunneling of a lithium nucleus , and the corresponding reaction rate @xmath69 can be obtained in the same way .
lithium hydride is an ionic crystal with simple cubic structure . in each pair of @xmath70 and @xmath71 atoms ,
one electron is transferred from the lithium to the deuterium . as a result ,
the crystal consists of positive ions of lithium and negative ions of deuterium .
the extra electron is loosely bound to the deuterium , which makes the radius of the ion @xmath72 approximately twice as much as the radius of @xmath73 ( see , for example , ref .
schematically , the structure of lithium deuteride crystal is depicted in figures [ fig.crystal.plane ] and [ fig.crystal.3d ] .
if we consider just bare nuclei sitting at the nodes of the lattice , they repel each other with the coulomb forces and at short distances attract each other with the strong forces .
the surrounding electrons make the configuration stable and partly screen the coulomb repulsion . for a deuteron nucleus ,
the neighbouring @xmath18 nuclei create the potential profile schematically shown in fig .
[ fig.crystal.pot ] .
there are three orthogonal axes along which the deuteron moves in such a potential . in order to fuse with @xmath18 ,
the deuteron have to tunnel through one of the six potential barriers surrounding it .
+ the height @xmath74 of the barrier can be estimated as the coulomb repulsion energy of the charges @xmath75 and @xmath76 at the distance @xmath77 , where @xmath78fm @xcite and @xmath79fm @xcite are the nuclear radii of the deuteron and @xmath18 , respectively .
this gives @xmath80 apparently , the central minimum of the potential shown in fig .
[ fig.crystal.pot ] , is not the same as the zero potential energy for an isolated @xmath27 pair .
indeed , the deuteron is pushed away from both sides .
this effectively lifts the deuteron up against the barrier .
+ if @xmath81 is the distance between the nuclei @xmath20 and @xmath18 in the crystal @xcite , and @xmath82 is the shift of the deuteron from its equilibrium position , then the coulomb forces acting on it from the neighbouring lithium nuclei , generate the potential energy : @xmath83 at the equilibrium point ( @xmath84 ) the potential is @xmath85 this is how much the deuteron in the crystal is lifted up against the free - space coulomb barrier . actually , the deuteron energy is even a bit higher .
this additional energy , however , is associated with the thermal oscillations and therefore is small , namely , of the order of @xmath86 , i.e. @xmath87 at room temperature .
+ the deuteron sitting in the central well of the potential shown in fig .
[ fig.crystal.pot ] , oscillates around its equilibrium point . in order
to estimate the fusion rate , we need to know the energy levels ( spectrum ) of its oscillations , as well as the distribution of the statistical ensemble of the deuterons over these levels . for the estimation purpose
, this problem can be simplified if we approximate the central well by another potential for which both the spectrum and wave functions are known analytically .
+ looking at fig . [ fig.crystal.pot ] , one might guess that the best approximation would be a parabola , i.e. a harmonic oscillator potential .
this however is wrong because the curve shown in that figure is schematic .
actually , if we accurately plot the function ( [ vx2coulombs ] ) on the interval @xmath88 , it is practically flat everywhere except for the left and right ends of this interval , where it quickly raises to @xmath74 .
the reason is that @xmath89 is too large as compared to the range of distances where the coulomb potential is comparable with @xmath74 .
actual potential - well looks as is shown in fig .
[ fig.crystal.pot.actual ] , where we removed the central part , which is long and almost flat .
+ of course , in addition to the coulomb fields generated by the nuclei , there are also electric fields due to the electron shells of the ions . however , these fields are weak as compared to the height @xmath90 of the coulomb barriers . indeed , the binding energy of an electron to the nucleus ( and therefore of the nucleus to the shell ) in a hydrogen atom is @xmath91 , which is five orders of magnitude smaller than the height of the barrier .
the reason of relative weakness of the electron field in the crystal is that , in contrast to the nuclei , the electrons act not as point - like charges but as a space charge of the electron cloud distributed within a volume of few angstrom size .
+ for the purpose of estimating , we can assume that the electron cloud is a uniformly charged sphere .
for a charge @xmath92 , the interaction potential with the charge @xmath93 uniformly distributed within a sphere of radius @xmath39 , is ( see eq .
( 2 - 104 ) of ref .
@xcite ) @xmath94 where @xmath51 is the distance of charge @xmath92 from the centre of the sphere .
if we assume ( for the sake of estimating ) that the electron shell of the ion @xmath72 is a uniformly charged sphere ( @xmath95 ) of the radius @xmath96 ( @xmath97 : see fig . [ fig.crystal.plane ] )
, then the strongest attraction ( at @xmath98 ) for the deuteron from its own electron shell is @xmath99 . in other words , the bottom of the potential shown in fig .
[ fig.crystal.pot.actual ] , is not exactly flat . at its centre , it has an additional shallow `` dent '' ( potential well ) of the depth @xmath100 , which is spherically spread to a distance of @xmath101 .
+ similarly , within the electron cloud of the @xmath73 ion , there is another `` dent '' on the potential curve of the depth @xmath102 ( because the radius of this cloud is half of the deuterium radius ) .
therefore the potential energy of the deuteron nucleus in the crystal ( along the line connecting two neighbouring lithium nuclei ) looks like is schematically shown in fig .
[ fig.crystal.vde ] .
+ it should be noted that despite its depth , the bottom of the lithium `` dent '' is a bit higher relative to the deuterium one because it is superimposed on the center of the coulomb repulsion .
it is clear that such a superposition slightly reduces both the height and the thickness of the coulomb barrier . of course the curve shown in fig .
[ fig.crystal.vde ] , is an exaggeration .
actual depths of the `` dents '' are five orders of magnitude smaller than the height @xmath74 of the barrier .
this means that , if plotted correctly , actual curve looks practically as a square - well potential of the depth @xmath74 and width @xmath103 .
it should be emphasized that we can only ignore the electron `` dents '' when consider the motion of the deuteron along any of the three orthogonal lines , i.e. directly towards one of the neighbouring lithium nuclei , and when the excitations are @xmath104 . in all other directions and at lower oscillation energies the central `` dent '' is important because it holds the deuteron in its place .
+ therefore , a good approximation for the potential @xmath105 is a square well .
moreover , we can replace it with an infinitely deep square well , because the excitations that we are going to consider ( @xmath106 ) are very small as compared to @xmath107 .
the advantage of using the infinite square well is that we know both the spectrum and wave functions for such a potential analytically : @xmath108 where @xmath68 is the deuteron mass .
apparently , everything that was said about the potential well for the deuteron , is valid for the lithium nucleus as well .
the only parameter that is different in such a case , is the mass of the nucleus .
the ground - state energies for both nuclei in the corresponding potential wells are very low , namely , @xmath109 and @xmath110 .
the excitation levels are very dense . on the interval @xmath111\,\mathrm{kev}$ ] there are 12,746 and 22,030 levels for the deuteron and @xmath18 , respectively ( see fig .
[ fig.levels ] ) . as we have seen , in the lithium deuteride crystal the deuteron is effectively lifted up against the coulomb barrier to the energy of @xmath112 .
this gives for the exponential ( gamow ) factor in eq .
( [ psipsiratio ] ) a negligible value , @xmath113 such a small value means that the spontaneous fusion in the crystal ( although is possible in principle ) is far beyond our capacity to detect it .
+ the main idea of the experiment that we describe in the present paper , is to irradiate the crystal with x - rays , which may excite the oscillations of the nuclei ( near their equilibrium positions ) to such a level where they could tunnel through the coulomb barrier as is shematically shown in fig .
[ fig.excitation ] .
apparently , the higher the excitation energy is , the greater the penetration probability would be . under the influence of an electromagnetic wave , the stimulated excitation ( or de - exitation ) rate @xmath114 , i.e. the probability of transition from a state @xmath115 to another state @xmath116 per second ,
can be found as ( taking into account only the dominant dipole transitions ) @xcite : @xmath117 where @xmath118 ( for deuteron ) or @xmath119 ( for lithium ) , @xmath120 is the angle between the @xmath82-axis and the polarization of the photon , @xmath121 , and @xmath122 is the volume density of electromagnetic energy per unit frequency interval . since we use lithium deuteride in the form of powder , the angle @xmath120 is random . relative to the photon polarization
, the @xmath82-axis can have any orientation uniformly distributed within full @xmath123 solid angle .
the averaging of @xmath124 over this solid angle gives a factor of @xmath125 .
+ the dipole matrix element in eq .
( [ pmn_cos ] ) can be found using the wave functions ( [ sqwell ] ) , @xmath126 \left[\frac{1}{(m - n)^2}-\frac{1}{(m+n)^2}\right]\ .\ ] ] as it should be , this matrix element is non - zero only if @xmath127 and @xmath128 have different parities . in such a case @xmath129 and @xmath130
using the first of eqs .
( [ sqwell ] ) , we can replace @xmath127 and @xmath128 with the corresponding energies
@xmath131 and @xmath132 , @xmath133 which gives the following stimulated transition rate ( averaged over the photon polarizations ) : @xmath134 as is seen , the rate of high - energy excitations ( @xmath135 ) diminishes as @xmath136 .
the statistical ensemble of deuterons in the crystal is in the thermodynamical equilibrium , and thus the population @xmath137 of each level ( i.e. the probability that a particular deuteron occupies the level @xmath128 ) can be found using the boltzmann distribution , @xmath138 at the room temperature , @xmath139 , this distribution gives the following average energy for the oscillations of the deuteron in the square well potential : @xmath140 the corresponding energy for the @xmath18-nucleus is almost the same , @xmath141 .
these small values are obtained because only a couple of dozens of the lowest levels are populated with appreciable probabilities .
+ when the crystal is exposed to x - rays , the statistical ensemble of deuterons ( or similarly @xmath18 nuclei ) is not in the thermodynamical equilibrium anymore
. its distribution over the energy levels is changing .
the deuterons jump up and down due to absorption and emission of photons .
+ if the flux of external photons is steady , a new dynamical equilibrium is formed with constant population @xmath142 for each level @xmath143 .
the time - evolution of the populations can be described by the so called master equation ( see , for example , ref.@xcite ) , @xmath144 where @xmath114 is the rate of @xmath145-transition stimulated by the external radiation , and @xmath146 is the rate of spontaneous transition from a higher level @xmath128 to a lower level @xmath127 .
the stimulated transition rate is given by eq .
( [ pmnemen ] ) , and the spontaneous rate ( in the same dipole approximation ) can be found as@xcite @xmath147 involving the same matrix ( [ mxn_ee ] ) . as we said before ,
the elements of this matrix are non - zero only if @xmath127 and @xmath128 have different parities , i.e. @xmath148 .
+ using boltzmann distribution ( [ boltzmannpop ] ) as the initial conditions at @xmath149 , we can , in principle , numerically solve the system of differential equations ( [ mastereq ] ) up to such time @xmath150 when all the populations reach constant values . in practice , however , this is difficult to do .
the reason is that the system ( [ mastereq ] ) consists of too many equations ( see fig .
[ fig.levels ] ) and therefore becomes numerically unstable .
+ if we manage to solve the differential equations ( [ mastereq ] ) , we obtain much more information than we actually need .
indeed , such a solution would give us the full history of how the initial boltzmann distribution evolves with time .
but we only need the final stationary distribution , when all @xmath151 reach their constant values and their derivatives on the left hand sides of these equations become zero .
this gives a homogeneous system of linear equations , @xmath152 with the normalisation condition @xmath153 where the matrix @xmath154 is composed of the transition rates @xmath114 and @xmath146 as is given in eq .
( [ mastereq ] ) .
+ it is easy to proof ( using the method of mathematical induction ) that the symmetry of the transition rates with respect to permutations of their subscripts implies that for each column of matrix @xmath155 , the sum of its elements is zero .
these means that the lines of this matrix are linearly dependent and thus @xmath156 . in other words , the linear system ( [ homogeneous ] )
always has a unique non - trivial solution , no matter what the initial distribution is .
+ a homogeneous linear system with the additional condition ( [ normalization ] ) and the symmetry properties described above , can be transformed to an equivalent non - homogeneous system , which can be numerically solved using the gauss - seidel iterative procedure .
an algorithm for such a solution can be found in ref .
@xcite . + in our experiment , we irradiated the sample with the x - rays whose spectrum covered the energy interval of @xmath157\,\mathrm{kev}$ ] .
the sample consisted of many layers of lithium deuteride alternated with polymer detectors ( for details , see sec .
[ sec.sample_preparation ] ) .
the total depth of the sample along the direction of the radiation was @xmath158 .
the radiation intensity attenuated in the sample due to absorption and the divergence of the rays . in our calculations , we used the intensity averaged over the depth of the sample . the averaged spectrum of x - rays is shown in fig .
[ fig.xspectrum ] .
it is given in terms of the number @xmath159 of photons in the unit energy interval falling on the area of @xmath160 per second .
+ using this spectrum , @xmath161 , we calculated the corresponding volume density @xmath122 of electromagnetic energy per unit frequency interval as follows .
in one cubic centimeter , there are those photons that fall on @xmath160 of its side during the time @xmath162 needed for the first bunch of photons to reach the opposite side of the 1 cm cube .
the total electromagnetic energy in the cube is the number of photons multiplied by @xmath163 .
dividing this by @xmath164 , we obtain the energy density per unit frequency , @xmath165 .
+ this density is needed in finding the stimulated transition rates ( [ pmnemen ] ) . with these rates and the rates ( [ pmnsp ] ) of the spontaneous transitions
, we solved the linear system ( [ homogeneous ] ) and thus found the stationary probability distribution for occupying various energy levels in the potential square - well .
this was done for both the deuteron and @xmath18 nuclei ( the only difference is the mass ) .
the resulting distributions are shown in fig .
[ fig.population ] .
when the crystal is exposed to the x - rays , the nuclei oscillating within the square - well potentials , acquire some kinetic energy . as a result , the probability for them to penetrate through the coulomb barrier increases . the corresponding fusion rate for a deuteron ( and similarly for lithium )
is given by eq .
( [ ratew ] ) . + in that equation
, the energy @xmath66 does not have a definite value , but can be any @xmath132 of the square - well spectrum for the deuteron ( or lithium , which is different ) , with the probability distributions @xmath166 and @xmath167 shown in fig .
[ fig.population ] .
this is similar to the reactions in stellar plasma @xcite , where velocities of the colliding nuclei are distributed according to maxwell s probability density .
therefore the observable fusion rate is the following average @xmath168 where @xmath169 stands for either @xmath20 or @xmath70 .
+ each deuteron has six neighbouring lithium nuclei and similarly each lithium nucleus has six hydrogen isotopes surrounding it .
however , the crystal is not pure @xmath170 compound . in some nodes of the lattice
it can be a proton ( instead of @xmath20 ) or @xmath171 ( instead of @xmath18 ) . in our experiment
, it was used a sample containing @xmath172 of lithium hydride powder with natural isotope composition of lithium ( mass fractions of @xmath18 and @xmath171 being @xmath173 and @xmath174 , respectively ) and enriched with the hydrogen isotope @xmath17 ( @xmath175 and @xmath176 ) .
+ therefore , if you find a hydrogen atom in the crystal , it is a deuterium with the probability of @xmath177 .
and a lithium atom has in its centre the @xmath18 isotope with the probability of @xmath178 .
therefore , a deuteron can find around itself a @xmath18 nucleus with the probability of @xmath179 , and for a @xmath18 isotope the probability to find a deuteron nearby is @xmath180 .
+ since there are two possibilities for the same fusion event to happen : either the deuteron or the lithium gets through the barrier , the total ( `` bulk '' ) fusion rate for the whole crystal can be found as a sum of the corresponding contributions : @xmath181 where @xmath182 and @xmath183 are the numbers of available deuterons and @xmath18 nuclei .
+ the effective molar mass @xmath127 of our crystal - powder is @xmath184 the number of hydrogen and lithium atoms ( any isotopes ) is the same , namely , @xmath185 , where @xmath186 is the avogadro number .
this gives @xmath187 and therefore @xmath188 numerical calculations with the x - ray spectrum shown in fig .
[ fig.xspectrum ] , give the following results for the single - nucleus rates : @xmath189 the corresponding `` bulk '' rate for the whole sample is @xmath190 for the total exposure time in our experiment , @xmath191 hours , we therefore should have expected to register @xmath192 fusion events .
as it follows from the theoretical estimate ( see previous section ) , the fusion events we are looking for , are rare ( roughly , one event every half an hour ) .
this fact determines the sample design and the choice of the way of registering the events .
the low rate of the reaction implies that we need to expose the sample to the x - rays for a relatively long time ( at least a hundred hours ) to get a statistically meaningful result
. the detector should be able to efficiently distinguish the fusion events from all possible background signals and to accumulate the information during whole exposure period .
+ based on this , we used the polymer track detectors cr-39 in direct contact with the active material ( lithium hydride ) .
each fusion event results in a pair of @xmath45-particles moving in the opposite directions with the same energy of @xmath46 .
these particles leave tracks in the polymer material , which can be identified after the experiment is completed . since only few hundred events are expected , the probability that different tracks overlap each other is practically zero .
the background tracks left by other charged particles , can be easily excluded using specific properties of the tracks belonging to the @xmath45-particles from the fusion events ( see sec .
[ sec.detectors ] ) .
lithium hydride ( @xmath193 ) is a solid crystalline compound , usually available in the form of chunks or powder .
it is extremely hygroscopic , absorbing water from the air , via the chemical reaction @xmath194 for our experiment , this is a destructive process and therefore we avoided the contacts of @xmath193 with atmospheric air by all available means .
all the manipulations with the crystals were done in an anaerobic chamber ( glove - box ) filled with dry @xmath195 gas . besides that , some desiccants were present in the chamber .
+ the crystalline chunks were ground into powder form .
this powder was placed between square plates ( @xmath196 each ) of the plastic detectors ( cr-39 ) , as is schematically shown in fig .
[ fig.slojka ] .
we put in parallel several units like the one shown in this figure , and the whole assembly was enclosed in a hermetic pvc container .
the total number of the detector plates was 85 .
+ the lithium hydride we used , was not pure @xmath197-compound .
the isotope composition of lithium atoms was practically natural , namely , 92.41% of @xmath198 and 7.59% of @xmath199 .
as far as the hydrogen atoms are concerned , the substance was enriched with deuterium up to 98% . to generate the x - rays , we used a xylon y.tu 225-d02 tube .
the tube was operated with a potential difference of 100kv and the current of 30ma . in order to find out what was the spectrum of the rays
, we used available data for the spectrum of this source , measured at the same voltage , but with a much lower current , namely , @xmath200 , and at a different distance .
we simply scaled these data to actual current and distance , used in our experiment .
+ the length of our sample was 13.2 cm .
passing through the sample , the x - rays attenuate both because of interaction with the material and because of spreading ( the rays are not collinear ) . as a result
the intensity of the radiation is not uniform through the sample . in our theoretical estimate
, we used average intensity of the x - rays in the sample . the corresponding spectrum ( obtained by scaling the original data and averaging them )
is shown in fig .
[ fig.xspectrum ] .
we identify the fusion events by registering the @xmath45-particles that emerge from the decay of the resulting @xmath201 resonance . however , there are other ( background ) sources of @xmath45-particles that might be distorting our counting . among them
the most common is radon ( and its decay products ) that is present practically everywhere .
in addition to that , our measurements were done at the premises of necsa ( nuclear energy corporation south africa ) near pretoria , where a nuclear reactor was operating .
although an appropriate radiation shielding was used , we can not exclude that some thermal neutrons from the reactor could reach our sample and generate @xmath45-particles in collisions with @xmath18 via the reaction @xmath202 our detectors were kept isolated from the atmospheric air all the time . and
since @xmath45-particles can not penetrate through a plastic container , we can safely ignore the possibility of counting false events due to radon and its progeny .
the only source of possible false events is the reaction ( [ neutron ] ) .
however , the @xmath45-particles from the fusion events have a much higher energy ( 11.1mev ) than those from this reaction .
this fact provides us with a reliable way of distinguishing the fusion and false events .
when analysing the detectors , we only counted the tracks left by particles with the energy higher than 6mev ( for the details see sec .
[ sec.detectors ] ) .
the fusion events were registered by identifying the appropriate tracks left by @xmath45-particles in the solid state nuclear track detectors cr-39 .
original tracks are very narrow channels where the structure of the plastic material is damaged by the particle .
these channels become much wider and visible under a microscope magnification , after etching . for etching
, we used a 6.25 mol/@xmath203 solution of naoh at 70 @xmath204c . +
this solution removes the plastic material not only from the channels , but from all the surfaces .
however the rate of etching at the damages is higher .
there are certain empirical formulae that enable one to calculate the etching rates at various conditions ( see ref .
@xcite and the references therein ) .
it is therefore possible to numerically model the shapes of the tracks for different etching times .
for such a modelling , we used the standard code track_test developed by the authors of ref .
+ depending on its kinetic energy , the @xmath45-particle penetrates into the plastic material and damages it to a certain depth ( typically a dozen of microns ) . when etched , such a track forms a pit in the shape of a sharp cone , as is shown in fig .
[ fig.tracks](a ) .
the longer we etch it , the deeper and wider this cone becomes .
this goes on until the depth of the pit becomes equal to the penetration length .
after that the pit only becomes wider and retains practically the same depth . as a result , the bottom of the pit transforms from a sharp cone to a smooth spherically concave surface , as is schematically shown in fig .
[ fig.tracks](b ) .
+ the sharp cone and concave shapes have different optical properties .
a smooth concave surface reflects the light and even focus it to a small dot , which is not possible for a sharp cone .
therefore just looking into the etched tracks , we can distinguish `` finished '' ( track end has been reached ) and `` unfinished '' tracks . using the track_test software , we can calculate the length of the track for any given energy @xmath205 of the @xmath45-particle as well as calculate the depth of the pit for a given etching time @xmath150 .
therefore we can choose @xmath150 such that all the tracks for @xmath205 higher than certain threshold @xmath206 , form sharp cone pits , while for @xmath207 the pits reflect the light .
+ since we need to exclude the tracks resulted from the reaction ( [ neutron ] ) , where the maximal energy is 4.78mev , we could choose the threshold energy to be 5mev .
however , to be absolutely sure , we used @xmath208 .
this corresponds to etching time @xmath209hours .
after the etching , we looked for completely dark pits ( sharp cones ) with certain diameter of the opening .
these tracks could not be done by anything else but the @xmath45-particles from the fusion reaction .
it should be noted that we can not register all the fusion events that occur in the sample .
first of all , the @xmath45-particles from the reactions that take place far away from a detector plate , can not reach the detector because the distance they can go through in any solid media is very short .
moreover , even if they reach a detector surface from afar , they loose a significant part of the kinetic energy and thus can not be registered if the remaining energy is less than @xmath208 .
+ our calculations show that in the lithium - hydride the original energy of 11.1mev of the @xmath45-particle is reduced to the minimally acceptable energy of 6mev on the distance of @xmath210 mm .
this means that , when we count the events , we can only register the events happening within a thin layer of the lithium - hydride of 0.095 mm thickness , which is in contact with the detector .
since in total we have 85 detectors and for each of them the surface area in contact with lih is [email protected]@xmath212 ( most of them have lih on both sides ) , the total volume of lih , from which we register the @xmath45-particles , is 0.77@xmath213 and the corresponding mass is 0.61 g .
this effective mass of the active material was used in our theoretical estimate of the fusion rate .
+ this is not the end of the story . the efficiency of the detection is reduced further by the fact that the @xmath45-particles are emitted isotropically in all directions .
for example , if the fusion event happens at the maximal acceptable distance from the detector and the @xmath45-particle moves at a non - zero angle relative to the normal to the detector surface , the actual distance it has to pass in lih is greater than the @xmath214 and thus it looses too much energy .
similar losses take place at any distance within the thin layer , when the angle is too large .
moreover , the tracks with the angle to the normal greater that @xmath215 can not be identified using our simple method . + estimating all possible losses , we conclude that we can only register of about 40% of the events happening in 0.61 g of the material .
this means that from the number of detected events we can only derive a lower bound for the reaction rate .
this is sufficient for a first and simple experiment .
actually , the main goal of our experiment was to establish if the fusion in the crystal could be induced by x - rays or not .
the sample composed of alternating layers of lithium - hydride and the plastic detector - plates ( as is described in sec . [ sec.sample_preparation ] ) , was exposed to the x - ray radiation for the duration of 111.466 hours .
then , after the etching of the detectors , we identified ( in total on all 85 plates ) 88 tracks that belonged to the @xmath45-particles from the fusion reaction ( [ dli6to2alpha ] ) . since we could not register all the fusion events ( some of them were missing as is explained in sec .
[ sec.efficiency ] ) , actual number @xmath216 of the events ( happened in 0.61 g of lih ) is greater , @xmath217 . taking into account that the substance we used was not pure lithium - deuteride ( see sec .
[ sec.fusion_induced_x ] for explanations ) , we conclude that under the used electromagnetic radiation ( the spectrum is given in fig .
[ fig.xspectrum ] ) the fusion rate for a single @xmath218 pair is @xmath219 this result is of the same order of magnitude as was predicted by our theoretical estimate ( [ 2body_rates ] ) .
the total number of events @xmath220 , theoretically expected for our experimental conditions and duration , is in accordance with the actual observation , @xmath217 .
+ apparently , the fusion rate turned out to be too low for any possible applications of this process in the energy production .
however , the electromagnetically induced nuclear reactions in crystals can be used in a different way . in principle , this offers a new way of measuring the cross sections ( or the astrophysical s - factors ) of such reactions ( not only fusion ) at extremely low energies , which are not accessible in the direct collision experiments .
+ of course , such an approach would require a more rigorous theoretical description of all the processes involved .
in particular , the electron screening has to be taken into account in a proper way
. this could be done by appropriately modifying the barrier penetration factor and using an accurate model ( instead of the square well ) for the potential schematically shown in fig .
[ fig.crystal.vde ] .
there is no doubt that taking into account the electron clouds is not an easy task .
indeed , the configuration of their space charge is dynamically changing when the nuclei move . it is therefore needed some theoretical effort before the nuclear reactions in crystals could be used for measuring the corresponding cross sections at low energies .
+ at present , the only alternative to the direct collision experiments at low energies is the so called `` trojan horse '' method , where the projectile is carried to the close proximity of the target as a cluster inside another nucleus ( the `` horse '' ) .
this `` horse '' may have the kinetic energy well above the coulomb barrier and thus neither its penetrability nor the screening effects play any role .
when near the target , the projectile is detached from the `` horse '' and interacts with the target at a low energy while practically all the original kinetic energy is taken away by the `` empty horse '' ( see @xcite , and the references therein ) .
+ although in the `` trojan horse '' method the complicated screening effects are avoided , it still relies on an involved theoretical analysis .
therefore , an additional experimental method for determining the s - factors would be of help .
the `` trojan horse '' and the crystal methods would complement each other .
+ experimentally , the main advantage of using crystals is the fact that here the effective flux of the nuclei is extremely dense . indeed ,
roughly the avogadro number of projectiles collide with the avogadro number of target nuclei .
with such a density ( as we have seen in our experiment ) even the reactions with very low probabilities can be observed . in order to narrow the interval of the collision energies
, a synchrotron radiation can be used instead of the x - rays .
99 r. balian , j .- p .
blaizot , p. bonche , _ `` cold fusion in a dense electron gas '' _ , j.phys . france , * 50 * , 2307 - 2311 ( 1989 ) a.c
. sips , g. giruzzi , s. ide , c. kessel , t.c .
luce , j.a . snipes , j.k .
stober , _ `` progress in preparing scenarios for operation of the international thermonuclear experimental reactor '' _ , physics of plasmas , 2015 , * 22(2 ) * , pp
. 1 - 24 ( 2015 ) . h. s. picker , _
`` on the fusion of hydrogen isotopes in ordinary molecules '' _ , nucleonica , * 25 * , 1491 - 1494 ( 1980 ) j. d. jackson , _ `` catalysis of nuclear reactions between hydrogen isotopes by @xmath221 mesons '' _ , phys .
, * 106 * , 330 - 339 ( 1957 ) h.e .
rafelski , d. harley , g.r .
shin , j. rafelski , _
`` cold fusion : muon - catalysed fusion '' _ , j.phys .
b : at.mol.opt.phys . ,
* 24 * , pp . 1469 - 1516 ( 1991 ) .
balin et al . , _ `` high precision study of muon catalyzed fusion in d2 and hd gases '' _ , physics of particles and nuclei ( jinr , dubna ) , v. 42(2 ) , pp .
361 - 414 ( 2011 ) .
s. e. jones , _
`` muon - catalysed fusion revisited '' _ , nature , * 321 * , pp .
127 - 133 ( 1986 ) .
v. b. belyaev , m. b. miller , yu .
g. sobolev , a. v. sermyagin , i. v. kuznetzov , and e. bialkovski , _ `` molecular - nuclear transition @xmath222 : search with a paired @xmath223 telescope '' _ , few - body systems , * 38 * , pp .
103 - 107 ( 2006 ) .
d. r. tilley , j. h. kelley , j. l. godwin , d. c. millener , j. e. purcell , c. g. sheu , h. r. weller , _
`` energy levels of light nuclei @xmath224 '' _ , nucl . phys .
, * a745 * , pp .
155 - 362 ( 2004 ) .
j. grineviciute , l. lamia , a.m. mukhamedzhanov , c. spitaleri , m. la cognata , _ `` low - energy r - matrix fits for the @xmath225 @xmath226 factor '' _ , phys .
, * c91 * , 014601 ( 2015 ) .
l. d. landau , e. m. lifshitz , _
`` quantum mechanics '' _ , pergamon press , ( 1965 ) .
d. laplaze , _ `` lattice dynamics of lithium hydride and lithium deuteride : effect of long - range tree - body forces '' _ , j.phys.c : solid state phys . , * c10 * , pp . 3499 - 3520 ( 1977 ) .
table xli of : p. j. mohr , b. n. taylor , d. b. newell , _ `` codata recommended values of the fundamental physical constants : 2010 '' _
, rev.mod.phys . , * 84(4 ) * , pp . 1527 - 1605 ( 2012 ) .
e. g. nadjakov , k. p. marinova , yu .
p. gangrsky , _
`` systematics of nuclear charge radii '' _ , atomic data and nuclear data tables , * 56 * , pp .
133 - 157 ( 1994 ) .
a. bohr , b. r. mottelson , _
`` nuclear structure '' _ , volume i , p.173 , w. a. benjamin , inc . , new york , amsterdam ( 1969 ) .
d. i. blokhintsev , _
`` quantum mechanics '' _ , d. reidel publishing company , dordrecht , holland ( 1964 ) .
n. g. van kampen , _
`` stochastic processes in physics and chemistry '' _ , third edition , north - holland ( 2007 ) .
http://www.mathpages.com/home/kmath175/kmath175.htm c. e. rolfs , w. s. rodney , _
`` cauldrons in the cosmos '' _ , the university of chicago press , chicago , ( 1988 ) .
a. a. azooz , s. h. al - niaemi , m. a. al - jubbori , _
`` a parameterization of nuclear track profiles in cr-39 detector '' _ , computer physics communications , vol . 183 , pp . 2470 - 2479
d. nikezic , k. n. yu , _
`` computer program track_test for calculating parameters and plotting profiles for etch pits in nuclear track materials '' _ , computer physics communications , vol . 174 , pp . 160 - 165 ( 2006 ) . c. spitaleri et al . , _
`` measurement of the 10 kev resonance in the @xmath227 reaction via the trojan horse method '' _ , phys .
, * c90 * , 035801 ( 2014 ) . | the nuclei that constitute a crystalline lattice , oscillate relative to each other with a very low energy that is not sufficient to penetrate through the coulomb barriers separating them .
an additional energy , which is needed to tunnel through the barrier and fuse , can be supplied by external electromagnetic waves ( x - rays or the synchrotron radiation ) .
exposing to the x - rays the solid compound lid ( lithium - deuteride ) for the duration of 111 hours , we have detected 88 events of the nuclear fusion @xmath0 .
our theoretical estimate agrees with what we observed .
one of possible applications of the phenomenon we found , could be the measurements of the rates of various nuclear reactions ( not necessarily fusion ) at extremely low energies inaccessible in accelerator experiments . |
2d imaging methods , such as optical microscopy , are still preferable to 3d imaging methods due to their high level of specificity and high resolution properties .
histological sections ( slices ) obtained through 2d imaging methods provide useful information for the diagnosis or the study of pathology .
3d volume reconstruction from these 2d slices is required in order to fully appreciate anatomical structures @xcite .
typically , a 3d volume is reconstructed by registering ( aligning ) the 2d sections with respect to a chosen reference and stacking successive aligned sections @xcite .
as the acquisition processes of different 2d histological images are performed independently , slice misalignment and deformation is often unavoidable . figure [ img : badslices ] shows examples of histological slices with non - cohorent distortions , tears , hole and missing parts .
the deformation varies from section to section and non - cohorent distortions may exist in consecutive sections . choosing an arbitrary slice as a reference slice leads to errors in 3d volume reconstruction , hence , the reference slice should be chosen properly not to contain distortions in order to achive high quality volume reconstruction @xcite . [
cols="^ " , ]
in this study , we present a novel method for reconstructing 3d rat brain volumes from 2d histological images .
the framework bases on three fundamental premises .
( 1 ) all histological images must be standardized for accurate registration leading to 3d volume reconstruction .
( 2 ) for accurate and succesful registrations in consecutive slices , a reliable feature space must be taken into account .
( 3 ) for automatic 3d volume reconstruction , the reference slice must be chosen properly by avoiding slices with high noise , distortions and other factors . to validate the reconstructed volume ,
the smoothness of the volume is considered .
experimental results indicate that the reconstructed volume is highly accurate .
this research is funded by the european commission fp6 marie curie action programme ( mest - ct-2005 - 021170 ) under the cmiag ( collaborative medical image analysis on grid ) project .
t. ju , j. warren , j. carson , m. bello , i. kakadiaris , w. chiu , c. thaller , and g. eichele , `` 3d volume reconstruction of a mouse brain from histological sections using warp filtering , '' _ j. neurosci . methods _ ( 156 ) , pp .
84110 , 2006 .
s. wirtz , b. fischer , j. modersitzki , and o. schmitt , `` superfast elastic registration of histologic images of a whole rat brain for 3d reconstruction , '' in _ medical imaging 2004 : image processing .
edited by fitzpatrick , j. michael ; sonka , milan .
proceedings of the spie , volume 5370 , pp .
328 - 334 ( 2004 ) .
_ , * 5370 * , pp . 328334 , may 2004 .
u. baci and l. bai , `` multiresolution elastic medical image registration in standard intenstiy scale , '' in _ 20th brazilian symposium on computer graphics and image processing ( sibgrapi ) _ , october 2007 .
y. ge , j. udupa , l. nyul , l. wei , and r. grossman , `` numerical tissue characterization in ms via standardization of the mr image intensity scale , '' _ journal of magnetic resonance imaging _ * 12 * , pp . 715721 ,
november 2000 . a. mackenzie - graham , e. lee , i. dinov , m. bota , d. shattuck , s. ruffins , h. yuan , f. konstantinidis , a. pitiot , y. ding , g. hu , r. jacobs , and a. toga , `` a multimodal , multidimensional atlas of the c57bl/ 6j mouse brain , '' _
j.anat _ ( 204 ) , pp .
93102 , 2004 . | in this paper , we propose three novel and important methods for the registration of histological images for 3d reconstruction .
first , possible intensity variations and nonstandardness in images are corrected by an intensity standardization process which maps the image scale into a standard scale where the similar intensities correspond to similar tissues meaning .
second , 2d histological images are mapped into a feature space where continuous variables are used as high confidence image features for accurate registration .
third , we propose an automatic best reference slice selection algorithm that improves reconstruction quality based on both image entropy and mean square error of the registration process .
we demonstrate that the choice of reference slice has a significant impact on registration error , standardization , feature space and entropy information .
after 2d histological slices are registered through an affine transformation with respect to an automatically chosen reference , the 3d volume is reconstructed by co - registering 2d slices elastically . |
within the standard model the extraction of the cabbibo - kobayashi - maskawa ( ckm ) matrix elements @xmath0 ( and @xmath6 ) is an outstanding topic of @xmath7-meson physics .
several ways have been utilized that are summarized , e.g. , by the particle data group @xcite .
presently , the value of @xmath0 extracted from inclusive decays is somewhat larger than from exclusive decays , e.g. , in @xmath8 .
a fruitful method to extract @xmath0 from exclusive decays is to reparameterize the decay data in such a way that they may be fitted by a smooth ( e.g. , linear ) function @xmath9 of @xmath10 , where @xmath11 , and @xmath12 is the 4-momentum transfer . doing so it is possible to extrapolate to the point of zero recoil of the @xmath13-meson , i.e. , @xmath2 , that is not directly measurable .
this procedure is particularly favored in the context of heavy quark expansion ( hqet ) @xcite but also useful to compare to other approaches since in this context the notion of the whole decay spectrum is not needed to extract @xmath0 . in hqet the value of @xmath14 at zero recoil is normalized up to corrections of order @xmath15 ( where @xmath16 denotes the mass of the @xmath17- or @xmath18-quark ) .
however the required fitting and extrapolation procedure leads to some errors , where the statistical error is under control and presently in the order of @xmath19 @xcite .
alternatively , quark models have been proven very useful as they provide not only predictions for @xmath20 for all @xmath10 and @xmath0 , but also numerous testable results for quite different processes @xcite .
a general overview on bound state models for heavy hadron decay form factors has been given by ref .
other approaches to the physics of heavy quark has profited from are qcd sum rules @xcite and lattice qcd @xcite .
since @xmath20 is known for a quark model one may ask for the implications on the empirical value for @xmath0 , if the zero recoil result is contrasted to the one obtained by using the whole decay spectrum .
both methods are frequently used but not yet compared to each other directly . in addition , relativistic quark models also allow us to describe heavy - to - light transitions , in particular @xmath21 important to determine @xmath6 , see , e.g. , ref .
@xcite . in this sense
the heavy - to - heavy transitions provide an important test case and bear an impression of possible model uncertainties . in this context the merit of semileptonic @xmath22 transitions
may be considered twofold : as already mentioned they provide an very good source to extract the cabbibo - kobayashi - maskawa ( ckm ) matrix element @xmath0 that has to be contrasted to inclusive and nonleptonic decays . on the other hand ,
weak decays ( in general ) provide important complementary information for qcd - motivated modeling of the underlying quark structure of mesons ( in general hadrons ) .
in addition , they may be considered useful to discuss the different relativistic approximations used in this context .
we choose an approach utilizing the instantaneous bethe - salpeter equation to treat the @xmath23-system within a relativistically covariant formalism @xcite .
the model is able to describe the meson mass spectrum for low radial excitations .
it has been applied to the calculation of leptonic decays , viz .
decay constants , @xmath24-decays @xcite , and to elastic form factors of mesons @xcite as well as to charmonium and bottomonium @xcite .
relativistic quark models have been investigated , e.g. , in refs .
the bethe - salpeter approach provides a consistent treatment of two - body bound states as well as the coupling of an external field via the mandelstam formalism @xcite . in order to actually solve the bound state problem several reasonable approximations are necessary or practical : i ) the quark propagators are assumed to be free propagators irrespective of confinement that is introduced via a confining kernel , ii ) quark masses are assumed to be constant ( i.e. constituent quark mass ) which is reasonable for heavy quarks , since current quark masses and constituent quark masses needed for reproducing the mesonic mass spectrum are rather close to each other , iii ) we utilize ladder approximation for the interaction kernel , and iv ) using an instantaneous interaction in addition leads to computational advantages , as it provides rpa - type equations @xcite that can be solved by introducing an effective hamiltonian @xcite in a formally covariant way .
the specific model used here , has been solved for the @xmath23-system in @xcite and applied to a wide range of phenomena @xcite including the heavy quark sector @xcite .
details of the model may therefore be found in the references given in the introduction .
here , i give a short survey and summarize some results . within the approximations given above the @xmath25-integration in the bethe - salpeter ( bs )
equation may be performed .
the resulting salpeter amplitude in the rest frame of the bound state with mass @xmath26 is the given by @xmath27 where @xmath28 is the full bethe - salpeter amplitude .
note , that the relative momentum @xmath29 appearing in eq .
( [ eqn : amplitude ] ) may be written in a covariant fashion @xcite . the resulting salpeter equation is then given by @xmath30 \,\gamma^0\,\lambda^+_2(-{\bf p } ) } { m+\omega_1+\omega_2 } \nonumber \\ & & - \int \!\!\frac{d^3p'}{(2\pi)^3}\ , \frac{\lambda^+_1({\bf p})\,\gamma^0\ , [ v({\bf p},{\bf p}\,')\,\phi({\bf p}\ , ' ) ] \,\gamma^0\,\lambda^-_2(-{\bf p } ) } { m-\omega_1-\omega_2}. \label{salpeter}\end{aligned}\ ] ] here @xmath31 , and we introduce energy projection operators @xmath32 in obvious notation , where @xmath33 is the standard dirac hamiltonian ( for details see , e.g. , refs.@xcite )
. the dynamical input of the model is defined by a confinement plus one gluon exchange ( oge ) kernel , @xmath34 .
confinement is introduced as a mixture of a scalar and a vector type kernel in the following way , @xmath35 = { \cal v}^s_c(({\bf p}-{\bf p}\,')^2)\,[\phi({\bf p}\ , ' ) - \gamma^0\,\phi({\bf p}\,')\,\gamma^0 ] .
\label{cc}\ ] ] due to the instantaneous approximation it is possible to introduce the same spatial dependence ( in the rest system of the meson ) as used in the nonrelativistic case , viz . in co -
ordinate space , @xmath36 the mixture of a scalar and a vector spin structure has been introduced in order to give an improved description of the spin orbit splitting .
other mixtures have been advocated in the literature , and also anomalous tensor - type confinement has been discussed , see , e.g. , ref . @xcite .
however , the consequences concerning , e.g. , the mass spectrum or regge behavior have not been studied yet . for the oge kernel
, we chose the coloumb gauge for the gluon propagator .
this way it is possible to retain a covariant formulation within an instantaneous treatment of the bethe - salpeter equation , and it allows to substitute @xmath12 by @xmath37 .
the oge kernel then reads @xcite @xmath38 = { \cal v}_g(({\bf p}-{\bf p}\,')^2 ) } \nonumber \\ & & \times \left [ \gamma^0 \phi({\bf p}\ , ' ) \,\gamma^0 -\frac{1}{2}\,\left ( { \mbox{\boldmath$\gamma$ } } \phi({\bf p}\,')\ , { \mbox{\boldmath$\gamma$ } } + ( { \mbox{\boldmath$\gamma$}}\hat{x } ) \phi({\bf p}\,')\ , ( { \mbox{\boldmath$\gamma$}}\hat{x})\ , \right)\,\right ] , \label{couleq}\end{aligned}\ ] ] with the operator @xmath39 , and @xmath40 where @xmath41 is introduced as `` running '' coupling as discussed in refs .
@xcite . to solve the salpeter equation numerically , eq .
( [ salpeter ] ) is rewritten as an eigenvalue problem ( rpa - equations ) , see , e.g. , ref .
this way it is possible to utilize the variational principle to find the respective bound states . to this end the salpeter amplitude @xmath42 is expanded into a reasonable large number of basis states used as a test function . as a suitable choice of basis states
we have taken laguerre polynomials and found that about ten basis states lead to sufficient accuracy , see also refs .
@xcite . for completeness , the parameters of the model given in ref .
@xcite are shown in table [ parameters ] .
these are the quark masses , the offset @xmath43 and slope @xmath44 of the confinement interaction eq .
( [ vconf ] ) and the saturation value @xmath45 .
they are determined to give a good overall description of the meson mass spectrum ( heavy and light mesons as well as charmonium and bottomonium ) @xcite .
semileptonic decays are treated in current - current approximation . for a transition @xmath46 the lagrangian
is given by @xmath47 with the ckm matrix element @xmath48 and the fermi constant @xmath49 .
the leptonic @xmath50 and hadronic currents @xmath51 are defined by @xmath52 the relevant transition amplitudes @xmath53 for @xmath54 and @xmath55 of the hadronic current can be decomposed due to the lorentz structure of the current , thus introducing form factors .
a standard representation of the form factors is given in terms of @xmath56 , @xmath57 ( or @xmath58 , @xmath59 ) for @xmath60 transitions , and @xmath61 , @xmath62 , @xmath63 , @xmath64 for @xmath65 transitions @xcite .
the exact definitions , and further references have been given , e.g. , in ref .
note that @xmath66 due to kinematical reasons . helicity amplitudes @xmath67 and @xmath68 in terms of the above form factors have been given by krner and schuler in a series of papers @xcite and are compiled by the particle data group @xcite .
using the helicity amplitudes the formulas for the decay spectrum used here is given in ref .
@xcite and will no be repeated here .
the respective decay rates into specific helicity states @xmath69 , @xmath70 are also given in the literature , see , e.g. , ref .
@xcite . to determine the form factors from the model , we follow the general prescription by mandelstam @xcite , see , e.g. , ref .
@xcite for a textbook treatment .
the lowest order contribution ( relativistic impulse approximation ) to the current ( sometimes referred to as triangle graph ) is given in fig .
[ kernel ] , and written as ( consider , e.g. , the anti - quark current , flavor indices suppressed ) @xmath71 where @xmath72 and @xmath73 denote the relative momenta of the incoming and outgoing @xmath74 pair , @xmath75 is the momentum transfer .
the quark feynman propagator is denoted by @xmath76 .
the dirac coupling to point - like particles is consistent with the use of free quark propagators . in eq .
( [ transitioncurrent ] ) the amputated bethe - salpeter amplitude or vertex function @xmath77 is given by @xmath78^{-1 } \,\chi_p(p)\;[s^f_{{\overline{q}}}(-p_{{\overline{q}}})]^{-1}.\ ] ] it may be computed in the rest frame from the equal time amplitude @xmath79 using the bethe - salpeter equation @xmath80 .
\label{vert}\end{aligned}\ ] ] finally , using the lorentz transformation properties of the field operators that define the bethe - salpeter amplitude @xcite , we can calculate the bethe - salpeter amplitude in any reference frame via @xmath81 where @xmath82 is the pure lorentz boost , and @xmath83 the corresponding transformation matrix for dirac spinors .
due to the reconstruction of the full bethe - salpeter amplitude sketched above , the transition matrix element eq .
( [ transitioncurrent ] ) is manifestly covariant .
the analysis of experimental data on heavy - to - heavy transitions now widely uses the notion of heavy quark expansion @xcite .
following ref .
@xcite we for one introduce the ratios @xmath84 and @xmath85 , @xmath86 \frac{v(q^2)}{a_1(q^2)},\\ r_2(\omega)&\equiv&\left[1-\frac{q^2}{(m_b+m_{d^*})^2}\right ] \frac{a_2(q^2)}{a_1(q^2)},\end{aligned}\ ] ] where @xmath87
. for @xmath88 decays the standard form factors may then be related to the ones used for the heavy quark expansion by @xcite @xmath89 h_{a_1}(\omega),\\ a_2(q^2)&=&\kappa_{bd^*}\,r_2(\omega)h_{a_1}(\omega),\\ v(q^2)&=&\kappa_{bd^*}\,r_1(\omega)h_{a_1}(\omega).\end{aligned}\ ] ] where @xmath90 . in the heavy quark mass limit ( @xmath91 )
@xmath92 where @xmath93 is a universal function known as isgur - wise function @xcite . for the case
@xmath94 the standard form factors are related to the heavy quark form factors via @xcite , @xmath95 where @xmath96 . in the heavy quark mass limit @xmath97 and @xmath98
. in the model approach used here the heavy quark mass limit has been performed numerically by multiplying @xmath16 with a large factor and keeping all other parameters as given in table [ parameters ] . to evaluate the transition matrix elements eq .
( [ transitioncurrent ] ) the meson amplitudes are then calculated by diagonalizing the eigenvalue problem with the large quark masses .
due to numerical reasons the heavy quark masses can not be chosen too large .
the function resulting from this numerical limiting procedure is then defined to be the isgur wise function @xmath99 of the bethe - salpeter - model , where @xmath100 within 0.1% .
this function is shown as a solid line in fig .
[ fig : isgurwise ] .
note that at this stage @xmath101 does not include radiative corrections that will be given below . in the same fashion the ratios @xmath102 and @xmath103 tend to unity within less than 0.1% when numerically increasing the heavy quark masses . for finite masses the experimental ratios @xmath104 assuming constant values
have recently been extracted by cleo @xcite .
the latest values are @xcite , @xmath105 that have to be contrasted to the long - dashed ( @xmath102 ) and dashed - dotted ( @xmath103 ) curve shown in fig .
[ fig : isgurwise ] .
the model ratios vary slowly by roughly 10% over the whole @xmath10 range , which is smaller than the experimental error .
utilizing the bethe - salpeter model to describe mesons as @xmath23 states the exclusive decay spectra for @xmath8 and @xmath94 have been calculated and compared to the experimental data .
earlier the ckm matrix element @xmath0 has been determined by a least squared fit to the whole spectrum of @xmath8 @xcite .
we have now redone this analysis on the basis of the improved data and also included the recently measured @xmath94 decay spectrum .
in addition , we present a new analysis for this model approach at zero recoil of @xmath88 that is commonly used for heavy - to - heavy transitions to extract @xmath0 .
this enables us to compare the difference between the energy dependent and the zero - recoil analysis quantitatively , at least for the bethe - salpeter model discussed here and the nonrelativistic approach given earlier @xcite .
we now turn to the extraction the ckm matrix element @xmath0 .
the differential decay rate for @xmath8 is given by @xcite , @xmath106.\end{aligned}\ ] ] the formula is written in a way that @xmath107 reduces to @xmath108 in the heavy quark mass limit , where @xmath109 denotes the radiative corrections @xcite .
for finite masses the function @xmath110 contains all the symmetry breaking effects .
experiments are given in a way that all well known factors are divided out in the decay rate and only @xmath111 is left over .
the corresponding data points of a recent cleo measurement are shown in fig .
[ fig : cleo ] .
the result is particularly smooth and may be fitted by a linear curve .
the fit to the data done by the cleo collaboration is also shown in fig .
[ fig : cleo ] as dashed - dotted line .
other lines reflect the model results utilizing different assumptions .
the solid line is calculated using the exact formula for the decay rate as given , e.g. , in @xcite ( i.e. , with @xmath104 @xmath10-dependent ) divided by the same factor as the experiments are .
the ckm matrix element @xmath0 is then determined by a least squared fit to all data points .
radiative corrections have been included in the dominant form factor @xmath112 , expressed as an overall factor @xmath109 @xcite . unlike earlier estimates that imply a correction of approx .
1% @xcite ( i.e. , smaller than the model uncertainty and therefore neglected in the earlier analysis
@xcite ) a recent two loop calculation leads to substantial value of @xmath113@xcite , which has to be included into the analysis .
the result is @xmath114 . to see the model dependence of the different analyses used in this context we now take @xmath115 , @xmath102 and @xmath103 constant , i.e. , @xmath116 , @xmath117 , and this ckm matrix element that leads to the long - dashed line shown in fig .
[ fig : cleo ] .
it is obvious that the curve slightly deviates from the solid one that includes the @xmath10-dependence of @xmath102 and @xmath103 .
for the same form factor however using the value of @xmath118 from the cleo fit is given by the short - dashed curve .
the resulting ckm matrix element is obviously larger by @xmath119 .
the function @xmath107 that leads to fig .
[ fig : cleo ] can be approximated by a quadratic fit to @xmath120 with the parameters @xmath121 and @xmath17 .
the slope of @xmath110 extracted by cleo @xcite assuming a linear dependence on @xmath10 ( @xmath122 ) is @xmath123 , and show as dashed - dotted line in fig .
[ fig : cleo ] .
the respective parameters for the quadratic fit of the curve discussed are shown in tab .
[ tab : slope ] . within the notion of the heavy quark
effective theory the form factor @xmath124 can be expanded into orders of @xmath125 . since symmetry breaking effects in semileptonic decays are of second order only @xcite , lowest order terms are usually written as @xmath126 where @xmath127 has to be determined .
corrections vary from @xmath128 @xcite to @xmath129 @xcite . a recent discussion and appreciation of the different approaches is given by martinelli @xcite .
the relativistic model discussed here leads to value of @xmath130 .
the values for @xmath131 that have been extracted by different experiments are quite consistent and lead to an overall fit of @xcite @xmath132 for the empirical slope parameter given in the last line of tab . [
tab : slope ] . using this value and the radiative corrections given above we extract the ckm matrix element @xmath0 for the relativistic bethe salpeter model to be @xmath133 that is larger by 7% compared to the spectrum dependent analysis .
the differential decay rate for the @xmath94 is given by @xcite , @xmath134 where again @xmath135 reduces to the isgur wise function in the heavy quark mass limit .
recent experimental data for the relevant part of the spectrum @xmath136 are shown in fig .
[ fig : cleobd ] . the linear fit to the data given by the cleo collaboration is also shown ( as a dashed - dotted line ) .
the model results utilizing the full spectrum dependent analysis leads to the solid line in fig .
[ fig : cleobd ] . for comparison the result of the zero recoil method utilized in the previous paragraph
is also shown .
the radiative corrections have been assumed to be in the same order as in the @xmath8 transition .
obviously the model is capable to provide a good description of the experimental data .
we have analyzed the exclusive decay rates of @xmath8 and @xmath94 within a relativistic constituent quark model .
the interaction kernel has been taken instantaneous . this way the bethe salpeter equation reduces to a salpeter equation as given in eq .
( [ salpeter ] ) .
the interaction consists of a one gluon exchange evaluated in the coulomb gauge and a linear confinement given in co - ordinate space .
the model parameters have been fixed to describe the mass spectrum of all observed mesons ( not only heavy mesons ) in a satisfactory manner @xcite .
the interaction current to describe the weak decay process has been introduced via the mandelstam formalism . to this end
the ( instantaneous ) amplitude has been reconstructed using the lorentz transformation properties of the field operators .
the only parameter left to describe the exclusive decay spectra is the ckm matrix element @xmath0 . for @xmath8 two methods have been compared .
one uses the complete spectrum , viz .
the functional dependence of @xmath137 , emerging from the quark model .
the ckm matrix element is then fixed by a least squared fit .
the other analysis utilizes the `` zero recoil '' method used in the context of heavy quark expansion . here only the empirical value of @xmath138 is used that is gained from an extrapolation of the experimental data ( e.g. by a linear fit ) to the zero recoil point @xmath2 and @xmath102 , @xmath103 assumed constant .
the ckm matrix element @xmath0 is then extracted using a singular value of the model at @xmath139 .
clearly , this method is not consistent with the underlying model , but widely used to extract @xmath0 from the zero recoil point . comparing the full model result to
the zero - recoil result leads to different values for the ckm matrix elements @xmath0 by approximately @xmath140 . in view of this result
it seems obvious that this kind of model uncertainty steaming from the different treatment of the @xmath10-dependence of the spectrum may show up in the determination of @xmath0 besides the statistical error .
this result may also be relevant for other more `` model independent '' analyses . for comparison
the @xmath94 decay has been calculated for the different assumptions and also leads to a good overall description of the experimental spectrum .
i would like to thank g. zller for his previous contributions to this work .
i am grateful to d. melikhov for valuable comments on the manuscript and to t. mannel for his interest .
also i would like to thank r. faustov for discussions on some general issues .
et al . _ ,
phys . rev . *
d 54 * , 1 ( 1996 ) , and 1997 off - year partial update for the 1998 edition available on the pdg www pages ( url : http://pdg.lbl.gov/ ) .
n. isgur , m. b. wise , phys .
* b 232 * , ( 1989 ) 113 ; * b 237 * , 527 ( 1990 ) .
cleo collaboration , s. sanghera _
et al . _ ,
phys . rev .
* d 47 * , 791 ( 1993 ) ; b. barish _
et al . _ ,
phys . rev . *
d 51 * , 1014 ( 1995 ) ; j.e .
. lett . * 76 * 3898 ( 1996 ) .
m. wirbel , b. stech , m. bauer , z. phys .
* c 29 * , 637 ( 1985 ) .
n. isgur , d. scora , b. grinstein , m.b .
wise , phys .
d 39 * , 799 ( 1989 ) .
w.jaus , phys . rev .
* d 41 * 3394 ( 1990 ) , ebd .
* d 53 * , 1349 ( 1996 ) .
galkin , a.yu .
mishurov , r.n .
faustov , sov . j. nucl
* 55 * ( 1992 ) 1207 , r.n .
faustov , v.o .
galkin , a.yu .
mishurov , j. nucl . phys .
* 55 * ( 1992 ) 1080 ( english : jinr print dubna , russia e2 - 91 - 451 ) , r.n .
faustov , v.o .
galkin , a.yu .
mishurov , phys .
b 356 * , 516 ( 1995 ) ; _ ibid .
_ * b367 * e391 ( 1996 ) ; r.n .
faustov , v.o .
galkin , z. phys .
* c 66 * , 119 ( 1995 ) .
d. scora , nathan isgur , phys . rev . * d 52 * , 2783 ( 1995 ) .
g. zller , s. hainzl , c.r .
mnz , m. beyer , z. phys . * c 68 * , 103 ( 1995 ) ; g. zller , _ diploma thesis _
( in german , unpublished , bonn 1994 ) .
grach , i.m .
narodetskii , s. simula , phys . lett .
* b 385 * , 317 ( 1996 ) .
d. melikov , phys . lett .
* b 394 * , 385 ( 1997 ) , _ ibid . _
* b 380 * ( 1996 ) , phys . rev . * d 53 * , 2460 ( 1996 ) .
cheng , c .- y .
cheung , c .- w .
hwang , phys .
d 55 * , 1559 ( 1997 ) .
a. le yaouanc invited plenary talk on _ ivth international workshop on progess in heavy quark physics _
held in rostock 1997 , to be published in the proceedings ( eds .
m. beyer , t. mannel , h. schrder , rostock university press , 1998 ) .
v. braun invited plenary talk on _ ivth international workshop on progess in heavy quark physics _ held in rostock 1997 , to be published in the proceedings ( eds .
m. beyer , t. mannel , h. schrder , rostock university press , 1998 ) .
c. sachrajda invited plenary talk on _ ivth international workshop on progess in heavy quark physics _ held in rostock 1997 , to be published in the proceedings ( eds
. m. beyer , t. mannel , h. schrder , rostock university press , 1998 ) .
b. stech , z. phys .
* c 75 * , 245 ( 1997 ) .
j.resag , c.r.mnz , b.c.metsch , h.r.petry , nucl . phys . * a 578 * 397 ( 1994 ) .
c.r.mnz , j.resag , b.c.metsch , h.r.petry , nucl . phys . * a 578 * , 418 ( 1994 ) .
c.r.mnz , j.resag , b.c.metsch , h.r.petry , phys . rev .
* c52 * , 2110 ( 1995 ) .
j.resag , c.r.mnz , nucl . phys . * a 590 * , 735 ( 1995 ) .
laga , phys . rev .
* d 45 * 305 , 317 ( 1992 ) . h. hersbach , phys . rev . * a 46 * 3657 ( 1992 ) , * d 47 * 3027 ( 1993 ) .
e.hummel , j.tjon , phys . rev . *
c 42 * , 423 ( 1990 ) ; g.rupp , j.a.tjon , phys . rev . * c 41 * , 472 ( 1990 ) ; p.c.tiemeijer , j.a.tjon : phys .
* b 277 * , 38 ( 1992 ) ; phys .
rev . * c49 * 494 ( 1993 ) , * c 48 * 896 ( 1994 ) .
t.murota , progr .
. phys . * 69 * , 181 ( 1983 ) , ibid . 1498 ( 1983 ) .
kalinovsky , c. weiss : z. phys . * c 63 * , 275 ( 1994 ) .
salpeter , h.a .
bethe , phys . rev . * 84 * , 1232 ( 1951 ) .
s.mandelstam , proc .
soc . * 233 * , 248 ( 1955 ) a. bilal , p. schuck , phys .
* d 31 * , 2045 ( 1985 ) .
e. klempt , b.c .
metsch , c.r .
mnz , h.r .
petry , phys .
b361 * , 160 ( 1995 ) , w.i .
giersche , c.r .
mnz , phys . rev . * c 53 * , 2554 ( 1996 ) , c.r .
mnz , nucl . phys . * a 609 * , 364 ( 1996 ) .
krner , g.a .
schuler , z. phys .
* c 38 * , 511 ( 1988 ) 511 [ e * c 41 * , 690 ( 1989 ) ] ; * c 46 * , 93 ( 1990 ) .
m. neubert , phys . rep . * 245 * , 259 ( 1994 ) .
cleo collaboration , g. gollin invited plenary talk on _ ivth international workshop on progess in heavy quark physics _ held in rostock 1997 , to be published in the proceedings ( eds .
m. beyer , t. mannel , h. schrder , rostock university press , 1998 ) .
s. resag , m. beyer , z. phys .
* c 63 * 121 ( 1994 ) .
d.lurie , _ particles and fields _ , ( interscience publishers , 1968 ) .
m. neubert , phys . lett . * b 264 * , 455 ( 1991 ) .
a. czarnecki , phys .
lett . * 76 * 4124 ( 1996 ) .
luke , phys .
* b 252 * , 447 ( 1990 ) .
a.f . falk , m. neubert , phys . rev . * d47 * 2965 , 2982 ( 1993 ) .
t. mannel , phys .
, * d 50 * 428 ( 1994 ) .
shifman , n.g .
uraltsev , a. vainshtein , phys . rev . *
d51 * , 2217 ( 1995 ) .
g. martinelli , nucl .
. meth . * a 384 * , 241 ( 1996 ) .
@xmath141 } & \mbox{[gev ] } & \mbox{[gev ] } & \mbox{[gev ] } & \mbox{[gev ] } & \mbox{[gev / fm ] } & \mbox{[fm ] } & \\[1ex ] \hline \vspace{1ex } 0.200 & 0.440 & 1.738 & 5.110 & -1.027 & 1.700 & 0.1 & 0.391\\ \hline\hline \end{array}\ ] ] | * abstract : * in the framework of a relativistic covariant bethe - salpeter model for the quark - antiquark system we present a renewed determination of the cabbibo - kobayashi - maskawa matrix element @xmath0 . complementing an earlier analysis applied to the whole decay spectrum for @xmath1 we now also employ the `` zero - recoil method '' that uses the end point of the decay spectrum ( @xmath2 ) and is suited for heavy - to - heavy transitions .
the averaged experimental value extracted from the data at zero recoil , @xmath3 , then leads to @xmath4 .
this value is somewhat larger than the one that uses the whole decay spectrum for the model analysis .
we also contrast this result to a nonrelativistic model and to recent experiments on the @xmath5 semileptonic decay . 13.30.ce , 12.39.ki , 12.15.hh , 13.20.he |
quantum states of light are efficient carriers of quantum information with applications in quantum computing , quantum cryptography , and quantum networks .
their most serious drawback is the absence of suitable non - linear interactions which can produce and transform the desired states .
measurement back action presents one means to accommodate the required non - linearities and paves the way for universal quantum computing with the tools of linear optics , but certain resource states such as single - photon states @xcite or superpositions of coherent states @xcite are required , while even more involved states , such as cluster states @xcite , are needed to perform successfully with high probability .
an efficient method to prepare general quantum states of light deterministically would constitute a significant step forward for optical quantum information processing .
there are a number of different strategies to produce non - classical states of light .
squeezed states and multi - mode states with quantum correlated amplitudes can be produced directly by optical parametric oscillators and beam splitters , while these states can be further transformed into states with non - gaussian amplitude distributions by conditional measurements @xcite .
the probabilistic nature of the schemes , however , limits the variety of states that can be produced by this method .
another successful approach uses laser excitation of an atomic ensemble with all atoms in the same single atom state and detection of a spontaneously emitted photon to conditionally prepare a superposition state with different atoms populating the final state of the spontaneous emission process .
the atomic state is generated at a random instant and the ensemble is then ready to deterministically emit a single photon at the moment when the experimentalist couples the singly occupied state to an optically excited state @xcite .
to @xmath0 , and during the conversion process , occupation by one atom translates into a single photon in a corresponding spatio - temporal light mode . for the case
illustrated , the states @xmath1 , @xmath2 , @xmath3 , and @xmath0 are occupied initially .
( b ) preparation of the atomic state .
the excitation from @xmath4 to a rydberg state @xmath5 is off - resonant if more than a single atom is excited , and a collectively distributed single excitation can be subsequently transferred to a unit atomic population of the single - atom ground state level @xmath6 .
( c ) the feeding of state @xmath6 by the attempted transfer @xmath7 can also be made conditional on the population of another state @xmath8 if that state is excited to a rydberg state @xmath9 interacting with @xmath5 . ] in the present paper , we propose a source , which is capable of producing a large variety of quantum states of light .
we focus on coding of quantum bits for quantum computing and communication , and we use multiple spatial and temporal modes to provide multi - bit registers with different entangled states .
the basic idea is to prepare the equivalent state in an ensemble of cold atoms and , in a step - wise process , convert the atomic state to a quantum state of light , fig .
[ convert](a ) , as proposed for two - mode states in @xcite . to avoid the requirement of individual addressing of the atoms and to facilitate the efficient conversion to light
, we encode the quantum state in the collective populations of different internal states of the atoms as proposed in @xcite .
the transfer of precisely one atom in the ensemble to an initially unoccupied internal state is carried out via intermediate excitation of a rydberg excited state and makes use of the so - called rydberg blockade mechanism @xcite .
it is thus possible to initialize the ensemble with all atoms populating a definite internal `` reservoir '' state , denoted @xmath4 in the following , and to subsequently encode , in principle , any quantum superposition states of an @xmath10-bit register based on @xmath10 different internal states , labeled @xmath11 , with each qubit value set to zero if the state is unpopulated , and to unity , if the corresponding state is populated by precisely one atom .
this is ensured by the universality of the collective quantum computing proposal with rydberg blockade @xcite . here
, however , we also have the freedom to process the state of the atomic ensemble between light emission events .
we can thus reuse the atomic levels and generate entangled light states with more than @xmath10 qubits .
our current task with the rydberg blocked ensemble is a more dedicated one than quantum computing , and rather than appeal to universality , we will describe explicit schemes which in few , efficient steps use the rydberg interaction , the physical properties of the internal atomic level structure , and the collective emission properties of the ensemble to provide the optical states of interest .
the rydberg blockade mechanism @xcite utilizes the large dipole - dipole interaction between atoms in certain rydberg states that are not separated by more than about @xmath12 m . due to this interaction , states with two or more atoms in rydberg excited states have energies that are significantly different from the sum of the single atom excitation energies , and therefore resonant excitation of a few @xmath13 sized ensemble from the reservoir ground atomic state will coherently drive the system towards the state with a single coherently distributed rydberg excitation in the ensemble . the rabi frequency for this transition is collectively enhanced by the factor @xmath14 , where @xmath15 is the number of atoms .
this enhancement and the blocking of further excitation has been seen for a pair of atoms @xmath16 @xcite , and blocking and enhanced coupling has also been observed in larger samples of atoms @xcite . after a @xmath17 pulse on the coherent transition , there is a single ensemble rydberg excitation and one can subsequently drive this collectively shared population into a low lying state , fig .
[ convert](b ) . to provide the coupling in a fully selective manner between the desired internal states , we suggest to use polarization selection rules and the optical frequency resolution of different zeeman ground states of the atoms in a homogeneous magnetic field switched on during state selective processes . in the following ,
the ket @xmath18 denotes a state with @xmath19 atoms in the single - atom state @xmath6 , @xmath20 , and the remaining atoms in the reservoir state @xmath4 . in writing this state ,
phase factors arising due to the different positions of the atoms within the laser traveling wave fields are taken into account .
the state @xmath21 , for instance , with one atom in @xmath1 and zero atoms in @xmath2 to @xmath22 is @xmath23 is the number of atoms in the ensemble , @xmath24 means that the @xmath25th atom is in the internal state @xmath6 , @xmath26 is the position of the @xmath27th atom , and @xmath28 is the resulting wave vector of the light pulses used to bring the ensemble to the above state , i.e. , @xmath29 , where @xmath30 is the wave vector of the @xmath31th laser pulse and @xmath32 is @xmath33 if a photon is absorbed and @xmath34 if a photon is emitted during the @xmath31th pulse .
if we apply a @xmath17-pulse with wave vector @xmath35 to transfer the atom in the state @xmath1 to the optically excited state @xmath36 , the amplitude for emission of a photon with wave vector @xmath37 and polarization @xmath38 when the atom decays to the state @xmath4 is proportional to @xmath39 , where @xmath40 is the dipole operator and @xmath41 .
the transition probability @xmath42 is thus proportional to @xmath43 the factor @xmath44 suppresses emission in all directions except @xmath45 , for which the sum is approximately @xmath15 and @xmath46 .
@xmath47 must also equal the atomic excitation frequency .
the factor @xmath44 is replaced by @xmath48 if the atom decays to a different hyperfine level , and such transitions are thus also suppressed . with an ensemble of about @xmath49 atoms it is possible both to satisfy the requirement of all atoms being at most @xmath50 apart , while the ensemble is still large enough to ensure directional emission of the photon @xcite .
a detailed analysis of light emission from various arrangements of atoms can be found in @xcite . to map a general superposition of the states @xmath18 to @xmath10 temporally distinct light pulses , we simply transfer one internal state population after the other to the optically excited state as illustrated schematically for @xmath51 in fig .
[ convert](a ) , and we can even renew the populations of the atomic states during the conversion process if desired .
we now turn to a description of explicit schemes to generate particular quantum states of light .
we start with the @xmath52-mode ghz state @xmath53 , where @xmath54 denotes a single / zero - photon state in the @xmath55th mode . to prepare this state
, we first apply a @xmath56-pulse between @xmath4 and a rydberg level @xmath9 followed by a @xmath17-pulse between @xmath9 and the ground state level @xmath8 , which leads to a superposition of having zero and one atom in @xmath8 .
a train of entangled pulses , all containing either zero or one photon , can then be emitted by conditional feeding of the state @xmath6 , as shown in fig .
[ convert](c ) .
after @xmath57 emissions , we apply a @xmath17-pulse between @xmath8 and @xmath9 , a @xmath17-pulse between @xmath9 and @xmath4 , and a @xmath17-pulse between @xmath8 and @xmath9 , which coherently changes the population in @xmath8 between zero and one . the last mode
is then released by transferring the population in @xmath8 to an excited state . from the collective atomic state @xmath58
( the first term is indicated with circles , and the second with crosses ) .
( a ) a light pulse transfers the populations in @xmath2 and @xmath59 to the excited states @xmath60 and @xmath61 , respectively , and the subsequent atomic decay to the state @xmath4 leads to emission of the first photon , which is @xmath62-polarized if @xmath60 is populated and @xmath63-polarized if @xmath61 is populated .
( b ) the second photon is similarly obtained via spontaneous emission after transfer of the populations in the states @xmath1 and @xmath3 to the excited states . ]
the maximally entangled two - mode bell state @xmath64 , where @xmath63 and @xmath62 denote right and left polarized photons , respectively , can be prepared , if we first prepare the atoms in the collective state @xmath65 , making use of the five hyperfine states in fig .
[ englert ] .
this is done as follows : _ i _ ) transfer one atom from @xmath4 to @xmath2 via a rydberg level .
_ ii _ ) use a raman transition between @xmath2 and @xmath59 to obtain the transformation @xmath66 .
_ iii _ ) move an atom from @xmath4 to @xmath3 if and only if @xmath59 is unoccupied , and an atom from @xmath4 to @xmath1 if and only if @xmath2 is unoccupied using the conditional method in fig .
[ convert](c ) . the optical release of the state @xmath67 is then achieved as explained in fig . [ englert ] .
if the light pulses are released by spontaneous raman transitions in the atoms , the temporal shape of the emitted photon pulse can be controlled by the coupling laser amplitude .
conversely , incident light pulses with a known temporal shape can be transferred via raman processes to collective atomic population in different states . by subsequent application of rydberg quantum gates ,
the state can be processed and analyzed , and the ensemble can thus , for example , be used to distinguish between all four polarization or photon number encoded bell states . as a specific application of the two - mode optical bell state
, we implement the double trine scheme for quantum key distribution , proposed recently by tabia and englert @xcite .
this scheme has the appealing features that the users of the communication channel , alice and bob , do not need to share a common reference frame and it is possible to extract @xmath68 key bits per trine state which is close the the theoretical maximum .
the scheme applies a sequence of groups of three photons with orthogonal polarizations @xmath63 and @xmath62 .
the groups are chosen at random among three mixed quantum states @xmath69 , @xmath70 , each being a direct product of a completely mixed state @xmath71 of the @xmath31th photon and a pure bell state @xmath72 of the other two photons .
to produce such a state , we prepare the atomic state @xmath67 and interweave its conversion to light with the emission of a mixed polarization state of a single photon .
the mixed state @xmath71 can be generated by using a classical random decision to determine whether an atom is moved from @xmath4 to @xmath60 or from @xmath4 to @xmath61 via a rydberg level .
the subsequent decay back to @xmath4 then produces the photon with the apparent random polarization .
note that the emission of the randomly polarized photon may occur independently of the state of the levels @xmath1 , @xmath2 , @xmath59 , and @xmath3 , and the three photons can thus be emitted in any order , which allows alice to prepare any of the states @xmath69 .
the detection by bob consists simply in measuring the polarization of the individual photons in any basis and does not require any further complicated processing of the states . .
( a ) in each iteration step , a @xmath56-pulse is applied to the ( raman ) transition between @xmath2 and @xmath59 .
an atom is then moved from @xmath4 to @xmath1 if and only if @xmath59 is unoccupied , and an atom is moved from @xmath4 to @xmath3 if and only if @xmath2 is unoccupied .
( note that the sum of the populations in @xmath2 and @xmath59 is always one atom . )
( b ) the populations in @xmath1 and @xmath3 are transferred to @xmath60 and @xmath61 , respectively , which leads to emission of a photon , and the process is repeated .
the bonds between the emitted photons illustrate the cluster state entanglement , and the graph representing the state is shown below the photons . ]
recently , lindner and rudolph proposed to use a quantum dot to prepare one - dimensional cluster states of photons @xcite . following their approach ,
such cluster states can be produced as shown in fig .
[ rudolph ] . starting with all atoms in @xmath4 ,
the initialization consists in moving one atom to the state @xmath59 via a rydberg level .
we then use a raman transition to apply a @xmath56 pulse between @xmath2 and @xmath59 . using the controlled feeding of atomic states shown in fig .
[ convert](c ) , we can apply the transformations @xmath73 and @xmath74 and subsequently transfer the populations in @xmath1 and @xmath3 to @xmath60 and @xmath61 , respectively .
after emission of a photon thus entangled with the atomic population remaining in states @xmath2 and @xmath59 , one applies again a @xmath56-pulse between @xmath2 and @xmath59 followed by a new rydberg controlled feeding of the populations in @xmath1 and @xmath3 and emission of the next photon .
this procedure is iterated and one gradually builds up a one - dimensional cluster state of light pulses .
after the desired number of iterations , the cluster is decoupled from the atomic state by measuring the state of the last emitted photon in the basis @xmath75 . , and one atom is moved to @xmath76 .
( a ) in each iteration step , a @xmath56-pulse is applied to the transition between @xmath2 and @xmath59 ( pulse 1 ) and to the transition between @xmath77 and @xmath76 ( pulse 2 ) .
the @xmath17-pulses 3 - 6 then change the phase of the state by @xmath17 unless @xmath59 and @xmath76 are both unoccupied .
atoms are moved from @xmath4 to @xmath1 , @xmath3 , @xmath0 , and @xmath78 conditioned on @xmath59 , @xmath2 , @xmath76 , and @xmath77 being unoccupied , respectively , and photons are emitted by transferring first the populations in @xmath1 and @xmath3 and subsequently the populations in @xmath0 and @xmath78 to @xmath60 and @xmath61 .
( b ) the first 8 emitted photons and ( c ) the corresponding graph . ] the behavior of one - dimensional cluster states can be efficiently simulated on a classical computer , and they are hence not useful resource states for cluster state quantum computing @xcite .
our method can , however , be generalized to higher - dimensional cluster states by involving more internal levels of the atoms . to generate , for instance , the two - dimensional cluster state shown in fig . [ rudolph2](c )
, we can use eight hyperfine states to prepare two one - dimensional cluster states in parallel and then add a controlled phase gate in each iteration step to obtain the bonds between them .
the details of the processing of the state of the ensemble between each emission of photons are shown in fig .
[ rudolph2](a ) , and the emission follows the same procedure as for the one - dimensional cluster state . the cluster state is decoupled from the atomic ensemble by measuring the state of the last two emitted photons .
larger two - dimensional cluster states can be obtained by weaving more one - dimensional cluster states together , and one can also form tubes and three - dimensional structures . in reality , the finite number of internal states of the atoms restricts the possibilities , and to comply with this , we suggest to emit the light pulses from a one- or two - dimensional array of individually addressable ensembles with separations that are small enough to allow long distance rydberg phase gates @xcite between nearest neighbor ensembles .
in conclusion , we have proposed a multi - purpose , multi - mode photonic source and constructed explicit schemes to prepare ghz states , double trine states , and cluster states .
we have only considered states with at most one photon per mode , but we anticipate that the same mechanism allows emission of more photons into the same mode @xcite .
the collective encoding generalizes naturally to this case by allowing atomic state populations exceeding one .
experiments on rydberg mediated one- and two - bit operations are still in their infancy , but according to theoretical estimates , their fidelities may be high enough that entangled photonic states with tens of modes may be prepared .
a rydberg state mediated quantum gate may typically be accomplished in about @xmath79 s , and recent experimental results with pulsed laser fields even demonstrate rabioscillations to rydberg states taking place on a time scale of about @xmath33 ns @xcite .
the time required to emit a single photon can be as short as @xmath80 s @xcite .
these time scales are much faster than rydberg state lifetimes on the order of milliseconds .
the shape of the laser pulses used to excite the atoms leaves some freedom to tailor the shape and duration of the emitted pulses , and the emission direction is controllable and may vary from mode to mode .
the number of photons that can be emitted by a single ensemble is limited in practice , because the interactions with the ensemble are not perfectly symmetric .
the collective state of the ensemble may , however , be brought back to the symmetric state by optical pumping of all atoms into the reservoir state or by a transfer of the atomic internal state populations via the rydberg gate mechanism to a new ensemble from which the photon emission can proceed .
99 e. knill , r. laflamme , and g. milburn , nature ( london ) * 409 * , 46 ( 2001 ) .
t. c. ralph , a. gilchrist , g. j. milburn , w. j. munro , and s. glancy , phys .
a * 68 * , 042319 ( 2003 ) .
r. raussendorf and h. j. briegel , phys .
* 86 * , 5188 ( 2001 ) .
k. wakui , h. takahashi , a. furusawa , and m. sasaki , opt .
express * 15 * , 3568 ( 2007 ) .
j. s. neergaard - nielsen , b. m. nielsen , h. takahashi , a. i. vistnes , and e. s. polzik , opt .
express * 15 * , 7940 ( 2007 ) .
a. ourjoumtsev , h. jeong , r. tualle - brouri , and p. grangier , nature ( london ) * 448 * , 784 ( 2007 ) .
duan , m. d. lukin , j. i. cirac , and p. zoller , nature ( london ) * 414 * , 413 ( 2001 ) .
a. kuzmich , w. p. bowen , a. d. boozer , a. boca , c. w. chou , l .- m .
duan , and h. j. kimble , nature ( london ) * 423 * , 731 ( 2003 ) .
s. chen , y .- a .
chen , t. strassel , z .- s .
yuan , b. zhao , j. schmiedmayer , and j .- w .
pan , phys .
lett . * 97 * , 173004 ( 2006 ) .
d. porras and j. i. cirac , arxiv:0704.0641 .
e. brion , k. mlmer , and m. saffman , phys .
lett . * 99 * , 260501 ( 2007 ) .
d. jaksch , j. i. cirac , p. zoller , s. l. rolston , r. ct , and m. d. lukin , phys .
lett . * 85 * , 2208 ( 2000 ) .
m. d. lukin , m. fleischhauer , r. cote , l. m. duan , d. jaksch , j. i. cirac , and p. zoller , phys .
lett . * 87 * , 037901 ( 2001 ) .
a. gatan , y. miroshnychenko , t. wilk , a. chotia , m. viteau , d. comparat , p. pillet , a. browaeys , and p. grangier , nature physics * 5 * , 115 ( 2009 ) .
e. urban , t. a. johnson , t. henage , l. isenhower , d. d. yavuz , t. g. walker , m. saffman , nature physics * 5 * , 110 ( 2009 ) .
r. heidemann , u. raitzsch , v. bendkowsky , b. butscher , r. lw , l. santos , and t. pfau , phys .
lett . * 99 * , 163601 ( 2007 ) .
m. saffman and t. g. walker , phys .
a * 66 * , 065403 ( 2002 ) .
l. h. pedersen and k. mlmer , phys .
a * 79 * , 012320 ( 2009 ) .
d. porras and j. i. cirac , phys .
a * 78 * , 053816 ( 2008 ) .
g. tabia and b .-
englert , arxiv:0910.5375 .
n. h. lindner and t. rudolph , phys .
lett . * 103 * , 113602 ( 2009 ) .
after submission of the present paper , a related quantum dot proposal to produce photonic 2d cluster states has been described in s. e. economou , n. lindner , and t. rudolph , arxiv:1003.2410 .
m. a. nielsen , reports on mathematical physics * 57 * , 147 ( 2006 ) .
m. saffman and k. mlmer , phys . rev .
a * 78 * , 012336 ( 2008 ) .
t. pfau , private communication . | we propose the implementation of a light source , which can deterministically generate a rich variety of multi - mode quantum states .
the desired states are encoded in the collective population of different ground hyperfine states of an atomic ensemble and converted to multi - mode photonic states by excitation to optically excited levels followed by cooperative spontaneous emission . among our examples of applications ,
we demonstrate how two - photon entangled states can be prepared and implemented in a protocol for reference frame free quantum key distribution and how one - dimensional as well as higher - dimensional cluster states can be produced . |
momentum space is assumed to be a manifold @xmath2 with a metric @xmath5 and a connection @xmath6 .
the geodesic distance @xmath7 from the origin to a point @xmath8 is interpreted as the rest mass of a particle with momentum @xmath9 .
the non - linear addition rule of momenta @xmath10 , which should be found experimentally , determines the connection : @xmath11 introduce an inverse operation @xmath12 satisfying @xmath13 which turns incoming momenta into outgoing momenta .
now we can write the conservation law associated with any interaction vertex as a non - linear equation @xmath14 .
for example , a three vertex can be written as @xmath15 .
the order of addition is important , since it corresponds to micro - causal structure of interactions@xcite .
the torsion of momentum space is a measure of the non - commutativity and the curvature is a measure of non - associativity of addition rule of momenta.@xcite dynamics of point particles is defined by the action : @xmath16 where @xmath17 is an affine ( time ) parameter along the trajectory of the particle and an interaction labeled by @xmath18 happens at @xmath19 for each particle ; @xmath20 are _ hamiltonian spacetime coordinates _ which are defined as being canonically conjugate to @xmath21 : @xmath22 and @xmath23 .
the mass shell condition @xmath24 is imposed by the lagrange multiplier @xmath25 .
the interaction part of the action is a lagrange multiplier times the conservation of momenta @xmath26 . by varying the action
, we get the equations of motion , the two of which we will be concerned with : @xmath27 @xmath28 indicates an incoming / outgoing particle respectively .
equation ( [ four velocity ] ) tells us how free particles propagate on one cotangent space ( hamiltonian spacetime " @xmath29 ) of their phase space ; and equation ( [ end points ] ) describes how cotangent spaces @xmath29 of different particles are connected by interaction events @xmath30 .
physics should be invariant under momentum space diffeomorphisms , i.e. redefinition of coordinates on the momentum manifold .
such a transformation is given by : @xmath31 s.t .
the geodesic distance between the origin and @xmath9 is preserved . under this transformation , @xmath32
, @xmath33 transforms like a covector : @xmath34 . by applying these transformations to our action
we see that it is unaltered , as desired .
in relativity the causal relationships between events are expressed as geometrical relationships between points on the spacetime manifold , while in rl we do not have a universal spacetime .
we thus have to come back to the most fundamental notion of causality between events .
define event b to be in the _ causal past _ of event a if in the process that is being considered , there exists a sequence of events @xmath35 s.t . from each event there
exists an outgoing free - propagating particle coming to the next event . we write it as @xmath36 .
vice versa we can define _ causal future _ @xmath37 .
the above definition of causal relationship is actually a strict partial order . in analogy with the notion in general relativity , one type of causality violation is if @xmath38 events @xmath39 s.t.@xmath40 i.e. causal loop .
let us check the existence of the simplest causal loop .
assume two bi - particle collision events @xmath41 and @xmath42 , defined by conservation laws @xmath43 and @xmath44 .
particle 0 with momentum @xmath45 is created from event @xmath41 and then collides with another particle at event @xmath42 .
the twist is now to consider particle 1 with momentum @xmath46 created at event @xmath47 and then colliding with another particle at event @xmath41 , which creates the particle 0 .
we use @xmath48 and @xmath49 to label the starting point and ending point of particle @xmath45 s worldline which lives on @xmath50 ; and similarly @xmath51 and @xmath52 for particle @xmath46 , as in fig[[phasespace ] ] .
particle 0 freely proporgates from event a to b for proper time @xmath53 on @xmath54 ; the end of its worldline is related with the starting point of particle 1 s worldline by interaction b : @xmath55 similarly , the other part of the loop is described by @xmath56 where @xmath57 are given by equation of motion ( [ four velocity ] ) .
thus the existence of a causal loop process is equivalent to the following equation having a physical solution for @xmath58 and proper times @xmath59 : @xmath60 where the matrix @xmath61 and @xmath62 .
+ the above is a general condition .
once we specify the geometry of momentum space and write down the conservation laws of two vertices , the above condition ( [ general condition ] ) will then give a system of four linear equations with six free unknowns : @xmath58 , @xmath53 and @xmath63 .
based on the cramer s rule , if the matrix @xmath64 has rank four , i.e. the determinant is non - zero , any physical choice of @xmath53 and @xmath63 yields a unique solution for @xmath58 .
if the matrix @xmath65 has rank less than four , we can fix a ratio @xmath66 first and then check the number of remaining unknowns of the system of the homogeneous equations . assuming the number of unknowns is @xmath67 ( @xmath68 now ) , if @xmath69 s.t .
the new matrix of coefficients of the system of equations has rank @xmath70 , then we can always get at least one physical solution for @xmath71 . when for @xmath72 the new coefficients matrix has a rank @xmath73 , only then the equation does not have physical solutions , which means the causal loop ( corresponding to these specific interaction vertices ) does not occur
. the solution will be invariant under the momentum space diffeomorphsim , because the equation ( [ general condition ] ) is just based on the equations of motion , which are invariant . in the special relativity limit of rl , beacuse of the linearity of momentum space , which also implies trivial isomorphism between cotangent spaces , the transport operators @xmath74
, @xmath75 are all identity matrices .
this makes the above equation ( [ general condition ] ) immediately degenerate to @xmath76 , which does nt have physical solution .
it means that this kind of causal loop will not occur in special relativity .
the causal loop formed by two interaction events is just the simplest case .
similar processes can be constructed by other forms of vertices and more events : @xmath77 , @xmath78 , @xmath79 , which conditions enjoy the similar form with the simplest case : @xmath80 these compose a system of four linear equations with @xmath81 unknowns , where @xmath82 is the number of interactions that form the causal loop .
compared with eq .
( [ general condition ] ) , the above conditions are even easier to be satisfied , since there are more unknowns in the same number of equations .
more general causal loops that contain branching out can be always decomposed into a few simple loops described by the above process .
in this section we will illustrate the existence of causal loops in a specific geometry of momentum space .
@xmath83-poincar hopf algebra , a dimensionful deformation of the poincar group , describes a momentum space with de sitter metric , torsion and nonmetricity , which is the first well - studied example of the non - trivial geometry of momentum space in relative locality @xcite , @xcite , @xcite .
it is also one of the weakest deformation of poincar group as a hopf algebra@xcite .
the line element of the @xmath84poincar momentum space in comoving coordinates is given by : @xmath85 where @xmath83 is a large energy scale close to planck energy .
the mass - shell condition is given by the geodesic distance from point @xmath9 to the origin of momentum space@xcite : @xmath86 from which we can get e.o.m.([four velocity ] ) .
the addition rule of momenta on @xmath84poincar momentum space is as follows@xcite : @xmath87 as an example of the simplest causal loop , we look at two events @xmath88 , which non - linear conservation of momenta are @xmath89 , see fig .
[ loop ] .
the conservation of momenta can be writen out explicitly as ( [ addition rule ] ) .
@xmath90 now , we have all the necessary elements to calculate the condition ( [ general condition ] ) for this specific example .
it turns out that the matrix @xmath64 has rank three and an unknown @xmath91 drops out , which is due to the linear addition rule of the zero component of momenta , in ( [ addition rule ] ) . for simplicity ,
assume that the particles with momentum @xmath92 have same rest mass @xmath93 .
using @xmath53 to rescale the other unknowns and solve the above equation for @xmath94 and @xmath95 , we get the solution , in which the ratio of two proper times is @xmath96\!\!+\!2f \kappa^2 \sinh ( \frac{p^0_0}{\kappa})}{p^1_i e^{\frac{p^1_0}{\kappa}}\ ! \left[2 e^{\frac{k_0+r_0}{\kappa } } ( k_i\!-\!r_i ) -\!f p^1_i\right]\!\!-\!2f \kappa^2 \sinh ( \frac{p^1_0}{\kappa } ) } \label{solution } \vspace{-1 mm}\ ] ] where @xmath97 just for shorthand . thus as long as there exists physical values of momenta such that the solution for @xmath98 is finite and @xmath94 is positive
, this specific causal loop can occur .
we check it as follows . for simplicity and without lose of generality
, we set the last two spatial components for all the momenta to be zero . because the energy of single particle has to be smaller than @xmath83 , energies @xmath99 .
the requirement of timelike on - shell momenta @xmath100 constrains the spatial components , @xmath101 we can then plot the region in terms of @xmath102 that allow the equation to have physical solution for @xmath103 and @xmath104 under a specific relationship among other momenta .
see fig.[region ] surprisingly , for any low energy scale of momenta choice ( as long as @xmath83 is not infinity ) , there always exist solutions for positive @xmath105s and finite @xmath48 . at low energy ,
the solution of @xmath106 in eq .
( [ general condition ] ) is proportional to @xmath107 : @xmath108 m } \vspace{-0.5 mm}\ ] ] thus when the scale @xmath83 is much larger than the scale of momenta , @xmath48 has quite large scale compared with @xmath53 . at high energy ,
the scales of solutions are reasonable .
two random examples taken from two different energy scales are shown in table ( [ tab ] ) .
if the current form of relative locality were to be correct , then we would expect causal loops to be common at high energy .
.comparison of the results at different energy scales [ cols="^,^,^ " , ] [ tab ]
we have shown that relative locality allows causal loops solutions , and it is surprisingly simple to construct them .
this is a generic feature arising from the phase space structure of rl : there is no universal spacetime , and the configuration information lives on the different cotangent spaces of the momentum manifold .
the non - linear conservations of momenta at interaction events determine how the cotangent spaces @xmath29 of different particles are connected by the interaction , eq.([end points ] ) . due to the nontrivial relations between cotangent spaces ,
a particle can come back to the event that causes its own creation as one of the incoming participants of that interaction . in the last section , the causal loop s existence depends on specific choices of points on the cotangent space @xmath29 (
let us call itx dependence " ) . in general , after fixing the proper times , the solution set of @xmath109 ( if it exists ) is a lower dimensional subspace . in the example we calculated
, it was a line on @xmath54 .
the same unusual feature of x dependence " is also present in many of the usual loop processes without causal issues in relative locality , e.g.@xcite .
some nets of vertices lead to the dependence on `` where '' the events are on the cotangent space , while some do not .
we do not yet have fundamental reasons for choosing some forms of vertices rather than others just for the sake of x independence " .
this necessitates future work in understanding the physical meaning of the hamiltonian spacetimes and the choices of vertices
. in @xcite authors enforced the translation invariance on @xmath29 , which is a more strict symmetry than x independence " .
however , the approach there requires the use of very non - local interaction vertices to achieve the symmetry , which is not physical .
future work should address the following questions .
is x dependence " of many loops a generic feature ?
should translation invariance on @xmath29 be a fundamental symmetry of relative locality ?
how do we achieve x independence " or translation invariance on @xmath29 in a physical way ? can we remove the causal loops by enforcing those symmetries in the theory ? does the existence of causal loops imply that there is an inconsistency in the theory ?
in general relativity , einstein equations have closed - timelike - curve ( ctc ) solutions .
one way out of the grandfather paradox is the novikov self - consistency principle , which states that events in ctc influence each other in a self - adjusted , cyclical way and the only solutions can occur locally are those which are globally self - consistent @xcite .
however , it has been shown that at the quantum level , unitarity fails for interacting fields in ctc and the subjective probabilities of events can be different for different observers @xcite .
it probably points towards that spacetimes with ctc are unphysical . in relative locality
, a well - defined quantum theory has not been built yet , but there is ongoing research into constructing it .
the lesson from quantum field theory in curved spacetime shows that it would be non - trivial to have causal loops in a consistent unitary theory .
it is thus essential to study whether it is even possible to construct a unitary quantum field theory for relative locality .
if the current form of relative locality were to be experimentally established in the future , we need to rethink causality as a fundamental property of nature .
it would influence some approaches to quantum gravity which take discrete causal structure as basic assumption .
another possibility is that relative locality is an incomplete or a wrong theory in its current form , and the revision of it should exclude causal loops .
* acknowledgements : * i am very grateful to l. smolin , l. freidel , g. amelino - camelia , a. banburski , t. rempel , s. hossenfelder , j. kowalski - glikman , f. mercati , j. hnybida , t. wang and j. ziprick for helpful discussions and suggestions
. research at perimeter institute is supported by the government of canada through industry canada and by the province of ontario through the ministry of economic development & innovation .
the work was supported by nserc and fqxi .
s. hossenfelder , phys .
lett . * 104 * , 140402 ( 2010 ) [ arxiv:1004.0418 [ hep - ph ] ] . g. amelino - camelia , l. freidel , j. kowalski - glikman and l. smolin , phys . rev .
d * 84 * , 084010 ( 2011 ) [ arxiv:1101.0931 [ hep - th ] ] .
g. amelino - camelia , l. freidel , j. kowalski - glikman and l. smolin , gen .
. grav . * 43 * , 2547 ( 2011 ) , int .
d * 20 * , 2867 ( 2011 ) [ arxiv:1106.0313 [ hep - th ] ] .
l. freidel and l. smolin , arxiv:1103.5626 [ hep - th ] . g. amelino - camelia , m. arzano , j. kowalski - glikman , g. rosati and g. trevisan , class . quant .
* 29 * , 075007 ( 2012 ) [ arxiv:1107.1724 [ hep - th ] ] .
g. gubitosi and f. mercati , arxiv:1106.5710 [ gr - qc ] .
s. majid and h. ruegg , phys .
b * 334 * , 348 ( 1994 ) [ hep - th/9405107 ] .
j. r. camoes de oliveira , arxiv:1110.5387 [ gr - qc ] .
j. friedman , m. s. morris , i. d. novikov , f. echeverria , g. klinkhammer , k. s. thorne , u. yurtsever , grp-225 .
j. l. friedman , n. j. papastamatiou and j. z. simon , phys .
d * 46 * , 4456 ( 1992 ) . | we find that relative locality , a recently proposed planck - scale deformation of special relativity , suffers from the existence of causal loops .
a simple and general construction of such on - shell loop processes is studied .
we then show that even in one of the weakest deformations of the poincar group in relative locality , causality can be violated .
the search for quantum gravity has led to the idea that special relativity should be modified at high energies in such a way that ( modified ) lorentz transformations leave the planck scale invariant .
doubly special relativity ( dsr ) has been proposed as an embodiment of this idea @xcite .
however , it was shown in @xcite that this theory has observer dependent non - localities for distant interactions .
relative locality is a reincarnation of dsr that tries to clarify the issue by proposing a radically different way of thinking about physics .
the birth of relative locality came from the insight that we never directly observe the spacetime we postulate .
our usual picture of spacetime is constructed operationally from the measurements of energies , momenta and times of events@xcite .
the notions that everything shares a universal spacetime and that locality has absolute observer - independent meaning could be unwarranted assumptions . as a theory describing quantum gravity induced modifications to relativistic dynamics of particles , relative locality was proposed at the regime where @xmath0 , while their ratio @xmath1 is held fixed for every observer@xcite .
so effects due to the presence of the planck mass are expected .
some rl phenomology has been studied in @xcite relative locality takes momentum space @xmath2 as primary and formulates classical dynamics on the phase space @xmath3@xcite .
the geometry of momentum space is not preassumed to be that of a linear space but could have curvature , torsion and non - metricity in general , and should be tested by experiments .
there is no global projection that gives a description of processes in a universal spacetime .
the notion of absolute locality and universal spacetime is equivalent with the assumption that the conservation law of momenta is linear . @xcite .
we start with giving a brief review of the classical dynamics and phase space structure of relative locality . after defining the causal structure in rl
, we go on to show that the theory has solutions that are causal loops .
the general conditions allowing for such loops to happen are then studied .
we illustrate this construction when the geometry of momentum space is taken to be that of @xmath4poincar , which is the most well studied and the simplest non - trivial geometry of momentum space in rl @xcite .
the appearance of the causal loops is a result of the phase space structure of the theory and the causal loops vanish in the limit of special relativity.the existence of causal loops implies that it would be non - trivial to construct a quantum theory of rl with unitarity , a worry which should be addressed in future research . if rl were shown to be a well - tested theory in the future , causal loops will challenge our understanding of the fundamental role of causality .
more probably , this result implies that rl is incomplete or wrong . |
hadron physics in the charm and beauty sectors is booming , with mounting activity in experiments babar , cleo , belle , bes , lhc@xmath12 , cdf @xcite and theory @xcite .
one of the issues that has attracted much attention is the finding of hadronic states which can not be interpreted in the conventional picture of @xmath13 for mesons and @xmath14 for baryons .
multiquark states , hybrids or hadronic molecules have been suggested in several works @xcite .
the molecular picture stands on firm grounds once the use of chiral unitary theory in the light quark sector , or its extension through the local hidden gauge approach , has shown that many mesonic and baryonic resonances are dynamically generated from the interaction of more elementary hadron components @xcite .
concerning baryonic resonances with charm or hidden charm , work on molecules has been done in refs .
@xcite , while in the beauty sector , baryon states with beauty or hidden beauty have also been studied in refs .
@xcite . on the experimental side ,
@xmath7 excited states have been reported by the lhc@xmath12 collaboration in ref . @xcite .
two states , @xmath4 and @xmath5 are found in the experiment , with widths smaller than 0.66 mev in both cases .
although no direct spin and parity have been determined , the states are interpreted as orbitally excited states of the ground state of the @xmath15 .
one of the states , the @xmath5 , has been confirmed by the cdf collaboration in refs .
the association to the orbitally excited states of the @xmath15 seems most natural since predictions of quark models had been done for these states , as the orbitally excited @xmath7 states with @xmath16 and @xmath17 @xcite .
compared to the observed results , the @xmath7 masses , including that of the ground state , are only off by about 30 - 35 mev .
the closest work in spirit to the present one is that of ref .
@xcite where these states are dynamically generated from the interaction of mesons and baryons . in ref .
@xcite the heavy quark spin symmetry ( hqss ) is used as an underlying symmetry . according to it , the @xmath18 ,
@xmath19 states are degenerate in the heavy quark limit , as well as the @xmath20 baryon states , which are then considered together in a coupled channels approach .
an extrapolation of the weinberg tomozawa interaction in the light sector is then used @xcite , with elements of an su(6 ) spin - isospin symmetry @xcite . with suitable choices of the renormalization scheme for the loops , good agreement with the masses of the newly found @xmath7 states
is obtained .
our scheme takes advantage of the study done in refs .
@xcite . in those works
it was found that the use of the extended local hidden gauge approach to the heavy quark sector fully respects the hqss , but it provided a dynamics different from the one of ref .
in particular , the connection between @xmath18 and @xmath19 states ( or baryon states with @xmath20 ) requires pion exchange , or anomalous terms , which are found subleading in the large heavy quark mass counting , and numerically small .
similar conclusions are also found in ref .
@xcite . in some works @xcite ,
pion exchange is found relevant compared to vector meson exchange ( the dominant terms in the local hidden gauge approach ) , but as discussed in ref .
@xcite , this is in part due to the use of a type of form factor for vector mesons , not present when the equivalent chiral amplitudes are constructed , that suppresses the vector exchange . with the dynamics of the extended local hidden gauge approach ,
the number of coupled channels is small for each quantum number and we can get a good feeling of the basic building blocks in the dynamically generated states that appear . as we shall see , the two @xmath7 excited states are generated with masses close to the experimental ones , but more remarkable , the difference in the masses of the two excited states agrees well with experiment
. the widths obtained are zero within our basis of coupled channels , quite in agreement with the widths smaller than 0.66 mev found in the experiment .
we also make predictions in other isospin and spin sectors . on the formal aspects we show how the dominant terms correspond to having a heavy quark as spectator and present an easy way to implement the results of the heavy quark spin - flavour symmetry from the impulse approximation at the quark level .
we will look at the states @xmath21 , @xmath22 , @xmath23 , @xmath24 , @xmath25 which can couple to @xmath26 which we will investigate .
similarly , we shall look at @xmath27 and @xmath28 , @xmath29 , @xmath30 , @xmath31 , with @xmath32 and @xmath33 , belonging to a decuplet of @xmath34 states . in the local hidden gauge approach in su(3 ) @xcite the meson baryon interaction proceeds via the exchange of vector mesons as depicted in fig .
[ fig : f1 ] . as discussed in ref .
@xcite , when we exchange a light vector meson in diagram ( a ) , ( b ) of fig .
[ fig : f1 ] , the heavy quarks of the meson or the baryon are spectators and hence the interaction does not depend on their spin nor its flavour . from the technical point of view the interaction of the diagrams of fig .
[ fig : f1 ] can be obtained using su(3 ) symmetry considering @xmath35 quarks , since we do not consider states with strangeness or hidden strangeness .
thus , all matrix elements of the interaction are formally identical ( except for the mass or energy dependence ) to those found for the interaction of the analogous states @xmath36 , @xmath37 , @xmath38 , @xmath39 , @xmath40 , @xmath41 , @xmath42 , @xmath43 , @xmath44 , @xmath45 .
this interaction has been studied in ref .
@xcite and ref .
@xcite .
the transition potential from channel @xmath46 to channel @xmath47 is given by @xcite @xmath48 with @xmath49 the pion decay constant , @xmath50 ( @xmath51 ) the mass , energy of baryon of @xmath46 ( @xmath47 ) channel .
we take @xmath52 since we exchange light vector mesons .
the @xmath53 coefficients are evaluated in refs .
@xcite and we quote them below . for pseudoscalar mesons and @xmath54 baryons we have the coupled channels @xmath25 , @xmath21 , @xmath23 in @xmath8 and the @xmath53 coefficients are given in table [ tab : vij0 ] .
.@xmath53 coefficients for @xmath8 and @xmath55 . [ cols="^,^,^,^",options="header " , ] in summary , we predict 6 states with @xmath8 , two of them corresponding to the @xmath4 and @xmath5 , and 8 states with @xmath9 .
the energies of the states range from about 5800 mev to 6500 mev .
it is interesting to compare the results obtained here with those of ref .
@xcite . in this later work ,
the same interaction as here is used for the main diagonal channels , but the transition between different coupled channels is not obtained through vector or pion exchange as done here , but invoking a combined su(6 ) and heavy quark spin symmetry . in ref .
@xcite the states of @xmath9 are not investigated but for @xmath8 four states are obtained , two of them , with @xmath6 , are also associated to the @xmath4 and @xmath5 . in spite of the differences in the input ,
there are common features in the results .
the two states associated to the @xmath4 and @xmath5 exhibit , as here , a substantial coupling to @xmath2 .
there is also a @xmath56 state in ref .
@xcite at 5797 mev which we find at 5820 mev , only 33 mev higher , and another state at 6009 mev that we find at 5969 mev , 40 mev below .
the mostly @xmath57 state found here at 6316 mev , basically degenerate in @xmath6 , was either not found or not searched for in ref .
@xcite because of its higher mass .
the qualitative agreement between the results of the two approaches is remarkable and gives further support to the common predictions . in addition
, we have investigated states of @xmath9 and we find quite a few , some of them narrow enough for a clear experimental observation .
in this work we studied the interaction of @xmath0 , @xmath1 , @xmath2 and @xmath3 states with its coupled channels using dynamics mapped from the light quark sector to the heavy one . the starting point was to assume that the heavy quarks act as spectators in the dominant terms of the interaction .
then we studied pion exchange and vector exchange and obtained the results of pion exchange in the heavy quark spin symmetric approach combined with results of lattice qcd .
the same procedure led us to the standard interaction of the extended local hidden gauge approach from the exchange of light vector mesons . with these elements of the interaction , supplemented with the subleading terms , in the large heavy quark mass
counting , obtained from the exchange of heavy mesons in the local hidden gauge approach , we studied the interaction of the @xmath0 , @xmath1 , @xmath2 and @xmath3 with their coupled channels @xmath21 , @xmath22 , @xmath24 ( for the @xmath0 ) ; @xmath28 , @xmath29 ( for the @xmath1 ) ; @xmath57 , @xmath58 , @xmath59 , @xmath60 , @xmath61 , @xmath62 ( for the @xmath2 ) ; and @xmath60 , @xmath61 , @xmath62 ( for the @xmath3 ) , and we searched for poles of the scattering matrix in different states of spin and isospin .
we found six states in @xmath8 , with one of them degenerate in spin @xmath6 , and eight states in @xmath9 , less bound , two of them degenerate in spin @xmath6 , and two more degenerate in spin @xmath11 .
the coupling of the states to the different channels were evaluated , together with their wave function at the origin , in order to get a feeling of which are the largest building blocks in those molecular states .
in particular , when studying the interaction of @xmath2 with its coupled channels , we found two states in @xmath8 , degenerate in spin @xmath6 , which couple mostly to @xmath2 .
the degeneracy is broken when the pion exchange is considered , allowing a mixture with intermediate @xmath0 states .
then we find that the mass of the states is close to 5910 mev for natural values of the regularizing cutoff .
more important , the splitting between the two states when the pion exchange is considered is found of the order of 10 mev , rather independent of the cutoff used .
this feature was considered relevant in view of the existence of the @xmath4 and @xmath5 states , which are separated by this amount of energy . a fine tuning of the cut off
was then done to match the exact energy of one of these states and then it was used subsequently to make predictions in all the other sectors of spin , isopin , leading to the states reported above .
we think that the use of realistic dynamics , with strict fulfilment of heavy quark spin - flavour symmetry , make the results obtained here rather accurate and should serve as a guideline for future experimental searches of baryon states with open beauty .
we would like to thank j. nieves for useful discussions , and carmen garcia - recio , laura tolos , and jia - jun wu for useful comments .
this work is partly supported by the spanish ministerio de economia y competitividad and european feder funds under contract no .
fis2011 - 28853-c02 - 01 and the generalitat valenciana in the program prometeo , 2009/090 .
we acknowledge the support of the european community - research infrastructure integrating activity study of strongly interacting matter ( hadron physics 3 , grant no .
283286 ) under the seventh framework programme of the european union .
this work is also partly supported by the national natural science foundation of china under grant no .
11165005 and by a scientific research fund ( 201203yb017 ) of the education department of guangxi .
a. ali , pos beauty * 2011 * , 002 ( 2011 ) [ arxiv:1108.2197 [ hep - ph ] ] .
m. gersabeck , mod .
lett . a * 27 * , 1230026 ( 2012 ) [ arxiv:1207.2195 [ hep - ex ] ] .
s. l. olsen , prog .
suppl . *
193 * , 38 ( 2012 ) .
e. e. kolomeitsev and m. f. m. lutz , phys .
b * 582 * , 39 ( 2004 ) [ hep - ph/0307133 ] .
l. maiani , v. riquer , f. piccinini and a. d. polosa , phys .
d * 72 * , 031502 ( 2005 ) [ hep - ph/0507062 ] .
f. -k .
guo , p. -n .
shen , h. -c .
chiang , r. -g .
ping and b. -s .
zou , phys .
b * 641 * , 278 ( 2006 ) [ hep - ph/0603072 ] .
e. swanson , int . j. mod
. phys . a * 21 * , 733 ( 2006 ) [ aip conf .
proc . * 814 * , 203 ( 2006 ) ] [ hep - ph/0509327 ] . j. l. rosner , j. phys . conf
. ser .
* 69 * , 012002 ( 2007 ) [ hep - ph/0612332 ] .
d. gamermann , e. oset , d. strottman and m. j. vicente vacas , phys .
d * 76 * , 074016 ( 2007 ) [ hep - ph/0612179 ] .
m. nielsen , f. s. navarra and s. h. lee , phys .
rept . * 497 * , 41 ( 2010 ) [ arxiv:0911.1958 [ hep - ph ] ] .
t. branz , t. gutsche and v. e. lyubovitskij , phys .
d * 80 * , 054019 ( 2009 ) [ arxiv:0903.5424 [ hep - ph ] ] .
e. klempt and j. -m .
richard , rev .
phys . * 82 * , 1095 ( 2010 ) [ arxiv:0901.2055 [ hep - ph ] ] .
g. ortega , d. r. entem and f. fernandez , j. phys .
g * 40 * , 065107 ( 2013 ) [ arxiv:1205.1699 [ hep - ph ] ] . c. hidalgo - duque , j. nieves and m. p. valderrama , phys .
d * 87 * , 076006 ( 2013 ) [ arxiv:1210.5431 [ hep - ph ] ] .
j. a. oller , e. oset and a. ramos , prog . part .
* 45 * , 157 ( 2000 ) [ hep - ph/0002193 ] .
e. oset , a. ramos , e. j. garzon , r. molina , l. tolos , c. w. xiao , j. j. wu and b. s. zou , int .
j. mod .
e * 21 * , 1230011 ( 2012 ) [ arxiv:1210.3738 [ nucl - th ] ] .
j. hofmann and m. f. m. lutz , nucl .
a * 763 * , 90 ( 2005 ) [ hep - ph/0507071 ] .
t. mizutani and a. ramos , phys .
c * 74 * , 065201 ( 2006 ) [ hep - ph/0607257 ] .
l. tolos , a. ramos and t. mizutani , phys .
c * 77 * , 015207 ( 2008 ) [ arxiv:0710.2684 [ nucl - th ] ] .
j. -j .
wu , r. molina , e. oset and b. s. zou , phys .
* 105 * , 232001 ( 2010 ) [ arxiv:1007.0573 [ nucl - th ] ] . j. -j .
wu , r. molina , e. oset and b. s. zou , phys .
c * 84 * , 015202 ( 2011 ) [ arxiv:1011.2399 [ nucl - th ] ] .
o. romanets , l. tolos , c. garca - recio , j. nieves , l. l. salcedo and r. timmermans , nucl .
a * 914 * , 488 ( 2013 ) [ arxiv:1212.3943 [ hep - ph ] ] . c. garcia - recio , j. nieves , o. romanets , l. l. salcedo and l. tolos , phys
. rev .
d * 87 * , 074034 ( 2013 ) [ arxiv:1302.6938 [ hep - ph ] ] . o. romanets , l. tolos , c. garcia - recio , j. nieves , l. l. salcedo and r. g. e. timmermans , phys
. rev .
d * 85 * , 114032 ( 2012 ) [ arxiv:1202.2239 [ hep - ph ] ] . c. w. xiao ,
j. nieves and e. oset , phys .
d * 88 * , 056012 ( 2013 ) [ arxiv:1304.5368 [ hep - ph ] ] .
j. -j .
wu , l. zhao , b. s. zou , phys .
b * 709 * , 70 ( 2012 ) [ arxiv:1011.5743 [ hep - ph ] ] . c. garcia - recio ,
j. nieves , o. romanets , l. l. salcedo and l. tolos , phys .
d * 87 * , 034032 ( 2013 ) [ arxiv:1210.4755 [ hep - ph ] ] . c. w. xiao and e. oset , eur .
phys . j. a * 49 * , 139 ( 2013 ) [ arxiv:1305.0786 [ hep - ph ] ] .
g. ortega , d. r. entem and f. fernandez , phys .
b * 718 * , 1381 ( 2013 ) [ arxiv:1210.2633 [ hep - ph ] ] .
[ lhcb collaboration ] , phys .
lett . * 109 * , 172003 ( 2012 ) [ arxiv:1205.3452 [ hep - ex ] ] .
t. a. aaltonen _ et al .
_ [ cdf collaboration ] , phys . rev .
d * 88 * , 071101 ( 2013 ) [ arxiv:1308.1760 [ hep - ex ] ] .
p. palni [ for the cdf collaboration ] , arxiv:1309.6269 [ hep - ex ] .
s. capstick and n. isgur , phys .
d * 34 * , 2809 ( 1986 ) .
h. garcilazo , j. vijande and a. valcarce , j. phys .
g * 34 * , 961 ( 2007 ) [ hep - ph/0703257 ] .
j. nieves and m. p. valderrama , phys .
d * 84 * , 056015 ( 2011 ) [ arxiv:1106.0600 [ hep - ph ] ] .
z. -f .
sun , j. he , x. liu , z. -g .
luo and s. -l .
zhu , phys .
d * 84 * , 054002 ( 2011 ) [ arxiv:1106.2968 [ hep - ph ] ] .
m. bando , t. kugo , s. uehara , k. yamawaki and t. yanagida , phys .
lett . *
54 * , 1215 ( 1985 ) .
bando , t. kugo and k. yamawaki , phys .
rept . * 164 * , 217 ( 1988 ) .
u. g. meissner , phys .
rept . * 161 * , 213 ( 1988 ) .
e. oset and a. ramos , nucl .
a * 635 * , 99 ( 1998 ) [ nucl - th/9711022 ] .
s. sarkar , e. oset and m. j. vicente vacas , nucl .
a * 750 * , 294 ( 2005 ) [ erratum - ibid .
a * 780 * , 78 ( 2006 ) ] [ nucl - th/0407025 ] .
e. oset , a. ramos and c. bennhold , phys .
b * 527 * , 99 ( 2002 ) [ erratum - ibid .
b * 530 * , 260 ( 2002 ) ] [ nucl - th/0109006 ] .
k. u. can , g. erkol , m. oka , a. ozpineci and t. t. takahashi , phys .
b * 719 * , 103 ( 2013 ) [ arxiv:1210.0869 [ hep - lat ] ] .
l. roca , e. oset and j. singh , phys .
d * 72 * , 014002 ( 2005 ) [ hep - ph/0503273 ] .
k. p. khemchandani , a. martinez torres , h. kaneko , h. nagahiro and a. hosaka , phys .
d * 84 * , 094018 ( 2011 ) [ arxiv:1107.0574 [ nucl - th ] ] .
k. p. khemchandani , h. kaneko , h. nagahiro and a. hosaka , phys .
d * 83 * , 114041 ( 2011 ) [ arxiv:1104.0307 [ hep - ph ] ] .
s. sarkar , b. -x .
sun , e. oset and m. j. vicente vacas , eur .
j. a * 44 * , 431 ( 2010 ) [ arxiv:0902.3150 [ hep - ph ] ] .
m. b. wise , phys .
d * 45 * , 2188 ( 1992 ) .
f. mandl and g. shaw , _ quantum field theory _ ( wiley - interscience , new york , 1984 ) .
l. roca and e. oset , phys .
d * 82 * , 054013 ( 2010 ) [ arxiv:1005.0283 [ hep - ph ] ] .
m. bayar , j. yamagata - sekihara and e. oset , phys .
c * 84 * , 015209 ( 2011 ) [ arxiv:1102.2854 [ hep - ph ] ] .
j. m. flynn , p. fritzsch , t. kawanai , c. lehner , c. t. sachrajda , b. samways , r. s. van de water and o. witzel , arxiv:1311.2251 [ hep - lat ] . b. borasoy , phys . rev .
d * 59 * , 054021 ( 1999 ) [ hep - ph/9811411 ] .
r. c. carrasco and e. oset , nucl .
a * 536 * , 445 ( 1992 ) .
r. machleidt , k. holinde and c. elster , phys .
rept . * 149 * , 1 ( 1987 ) .
e. oset and m. j. vicente - vacas , nucl .
a * 446 * , 584 ( 1985 ) .
e. oset , h. toki and w. weise , phys .
rept . * 83 * , 281 ( 1982 ) . | in this paper we examine the interaction of @xmath0 , @xmath1 , @xmath2 and @xmath3 states , together with their coupled channels , using a mapping from the light meson sector . the assumption that the heavy quarks act as spectators at the quark level automatically leads us to the results of the heavy quark spin symmetry for pion exchange and reproduces the results of the weinberg tomozawa term , coming from light vector exchanges in the extended local hidden gauge approach . with this dynamics
we look for states dynamically generated from the interaction and find two states with nearly zero width , which we associate to the @xmath4 and @xmath5 states .
the states couple mostly to @xmath2 , which are degenerate with the weinberg tomozawa interaction .
the difference of masses between these two states , with @xmath6 respectively , is due to pion exchange connecting these states to intermediate @xmath0 states .
in addition to these two @xmath7 states , we find three more states with @xmath8 , one of them nearly degenerate in two states of @xmath6 .
furthermore we also find eight more states in @xmath9 , two of them degenerate in @xmath10 , and other two degenerate in @xmath11 . |
this work was supported by the state key programs for basic research of china ( 2005cb623605 and 2006cb921803 ) , and by national foundation of natural science in china grant nos .
10474033 and 60676056 .
p. zoller _ et al .
j. d * 36 * , 203 ( 2005 ) . v. giovannetti _
_ , science * 306 * , 1330 ( 2004 ) .
et al . _ ,
lett . * 78*,3221 ( 1997 ) .
et al . _ ,
lett . * 92*,127902 ( 2004 ) .
et al . _ ,
lett . * 81 * , 5932 ( 1998 ) .
a. ekert , phys .
* 67 * , 661 ( 1991 ) .
et al . _ ,
lett . * 73 * , 3081 ( 1993 ) .
m. zukowski _
et al . _ ,
71 * , 4287 ( 1993 ) .
l. childress _
et al . _ ,
a * 72 * , 052330 ( 2005 ) ; phys .
lett . * 96 * , 070504 ( 2006 ) ; j.i .
et al . _ , phys .
rev . lett . * 78 * , 3221 ( 1997 ) ; w. yao _
et al . _ ,
* 95 * , 030504 ( 2005 ) ; e. waks _ et al .
_ , phys . rev
. lett . * 96 * , 153601 ( 2006 ) ; c.h .
et al . _ ,
lett . * 76 * , 722 ( 1996 ) ; d. deutsch _
et al . _ ,
. lett . * 77 * , 2818 ( 1996 ) ; w. dr _
et al . _ ,
a * 59 * , 169 ( 1999 ) ; c.w .
et al . _ ,
nature * 438 * , 828 ( 2005 ) .
_ , science * 316 * , 1316 ( 2007 ) .
et al . _ ,
nature * 452 * , 67 ( 2008 ) .
n. sangouard _
et al . _ ,
a * 76 * , 050301 ( 2007 ) . c. simon _
lett . * 98 * , 190503 ( 2007 ) .
et al . _ ,
* 98 * , 240502 ( 2007 )
. p. van loock _
et al . _ ,
lett . * 96 * , 240501 ( 2006 ) .
et al . _ , new j. phys . * 8 * , 184 ( 2006 ) .
et al . _ ,
* 101 * , 040502 ( 2008 ) . c. cabrillo _
a * 59 * , 1025 ( 1999 ) .
_ , nature * 443 * , 671 ( 2006 ) .
et al . _ ,
science * 319 * , 1062 ( 2008 ) .
et al . _ ,
rev . lett . * 91 * , 043902 ( 2003 ) .
kane , nature * 393 * , 133 ( 1998 ) .
t. schaetz _
et al . _ ,
phys . rev .
* 94 * , 010501 ( 2005 ) . | based on the interaction between a three - level system and a microtoroidal resonator , we present a scheme for long - distance quantum communication in which entanglement generation with near 0.5 success probability and swaps can be implemented by accurate state detection via measuring about 100 photons . with this scheme the average time of successful entanglement distribution over 2500 km with high fidelity
can be decreased to only about 30 ms , by 7 orders of magnitude smaller compared with famous duan - lukin - cirac - zoller ( dlcz ) protocol [ l .- m .
duan _
et al .
_ nature ( london ) * 414 * , 413 ( 2001 ) ] . in quantum information
science @xcite , the entanglement distributed over quantum networks is a crucial requisite for metrology @xcite , quantum computation @xcite , and communication @xcite .
quantum communication promises completely secure transmission of keys with the ekert protocol @xcite and exact transfer of quantum states by quantum teleportation @xcite . because of losses and other noises in quantum channels , the communication fidelity dwindles exponentially with the channel length . in principle , this problem can be overcome by applying quantum repeaters @xcite , of which the basic principle consists of separating the full distance into shorter elementary links and entangling the links with quantum swaps @xcite .
one kind of quantum repeater protocols based on single or sub - photon coherent state transmission may have high initial fidelity of entanglement but have low successful entangling event @xcite .
a protocol of specific importance for long - distance quantum communication based on collective excitations in atomic ensembles has been proposed in a seminal paper of duan _ et al .
_ @xcite .
after that considerable efforts have been devoted along this line @xcite .
another kind of protocols using bright pulses may have high successful entangling rate but low initial entanglement fidelity@xcite .
does there exist a scheme through which we may have both high successful entangling rate and high initial fidelity of entanglement ? motivated by this consideration , we propose a protocol for efficient long - distance quantum communication . in this scheme
the entanglement between two qubits assisted by high - q microtoroidal resonators in the basic segments can be generated with a success probability near 0.5 by adopting interference effect and state projection from measurements @xcite .
the entanglement is then extended to a double length through entanglement swapping . in both of the two stages ,
the detection of qubit states is accomplished with high accuracy by measuring many photons using single - photon detectors . from this scheme
the efficient built - in entanglement purification and robustness against experimental imperfections can be achieved .
the obtained entanglement can be directly applied to perform entanglement - based quantum communication protocols like quantum teleportation , cryptography , and loophole - free bell inequality test .
the communication time increases near linearly rather than polynomially with the distance . for a distance of 2500 km
it may significantly decrease to about only 30 ms , by 7 orders of magnitude smaller than that through dlcz protocol .
this is in the same order of magnitude as the time scale for the light traveling over this distance .
this scheme can produce results comparable to those in @xcite but avoid its complex entanglement purification process .
recent advances in the observation of strong interaction between photons and single atoms through microscopic optical resonators @xcite lay the foundation for this scheme .
the schematic description of the generation of entanglement between two qubits in a basic segment is shown in fig .
[ fig:1 ] .
a microtoroidal resonator has two internal counterpropagating modes @xmath0 and @xmath1 with a common frequency @xmath2 .
these two modes are coupled owing to the scattering @xcite .
the intracavity fields are coupled to a tapered fiber with high efficiency @xmath3 @xcite .
the evanescent intracavity fields coherently interact with the ground state @xmath4 and the excited state @xmath5 with energy @xmath6 of a three - level atom which is near the external surface of the resonator with well - defined azimuthal phase @xmath7 @xcite .
a metastable @xmath8 of the atom does not interact with the field of modes @xmath0 and @xmath1 due to being off resonance with the field .
such a @xmath9 system may be provided by a donor atom in doped silicon basis @xcite , where the qubit states @xmath4 and @xmath8 are encoded onto electron zeeman sublevels and the excited state is provided by the lowest bound - exciton state , or other examples such as the hyperfine structure of trapped ions @xcite . in this paper
we consider the situation where the input field and the resonator are impedance - matched ( critical coupling ) , which can be reached through careful choice of the point of contact between the surface of the microtoroid and the tapered optical fiber and applying the input probe pulse with frequency @xmath10 @xcite . under this condition , there are two cases in the forward output of the tapered fiber : first , the output will drop to zero due to the interference between the cavity field @xmath0 and the input field @xmath11 when the atom is in state @xmath8 ( dark state ) ; second , many single - photons come through the resonator one by one with average interval time @xmath12 @xcite , when the atom is in state @xmath4 ( bright state ) and is in resonance with the cavity @xmath13 .
( a ) schematic illustration of entanglement establishment between two atoms l and r assisted by microtoroidal resonators .
( b ) the level diagram of three - level @xmath9 atoms .
( c ) schematic illustration of a microtoroidal resonator with two modes . ]
the qubits in nodes l and r are initialized in state @xmath14 , @xmath15 , then two same weak probe pulses with frequency @xmath10 are incident simultaneously into the resonators ( fig .
[ fig:1 ] ) .
if both of the pulses contain only a single photon , then the state of the system comprising of the atom and the forward output photon can be described by @xmath16 , @xmath15 .
the forward propagating photons are combined through such as a convex lens and measured at the midpoint of the optical length between nodes r and l. the mode of output photon measured by the single - photon detector d is @xmath17 , where @xmath18 is an unknown difference of the phase shift in the left and the right side channels and can be set to zero @xcite . a click in detector d
measures the photon @xmath19 @xcite . by applying @xmath20 to the state @xmath21 , we have the projected state of the atoms l and r : @xmath22 with @xmath23 and the conventional notation @xmath24 .
considering the presence of noise from the dark count of the photon detectors described by a vacuum " coefficient @xmath25 , the projected state becomes @xmath26 @xcite .
if the probe pulses of duration @xmath27 contain many photons , we will detect about @xmath28 single - photons arising from one atom bright , where @xmath29 is the length of a basic link , @xmath30 is the channel attenuation length .
these extra single - photons except the first one will further confirm the state of the system of atoms r and l , thus reduce the error count probability .
for example , if @xmath31 , according to @xcite , the number of background @xmath32 photons is only about @xmath33 , we may assume @xmath34 , and the number of photons from two atoms bright will be about @xmath35 .
assuming the probability of finding @xmath36 photons from @xmath37 @xmath38 atoms bright is given by a poisson distribution @xcite ( see fig .
[ fig:2 ] ) , we find that the probability @xmath39 , @xmath40 , and @xmath41 of finding @xmath42 $ ] photons due to zero , one , and two atoms bright is @xmath43 , 0.9773 , and @xmath44 , respectively .
thus , we can unambiguously distinguish the state of one atom bright from the other two states .
the dark count in this case is excluded and the projected state is @xmath45 provided that the recorded photon number @xmath46 $ ] . as for the lost part of the outgoing photons , since they are immediately lost , we can assume that the subspace u where the lost photons enter is approximately vacuum at all times ( born approximation ) @xcite .
thus the whole state can be written as @xmath47 .
after tracing the lost photons , we find that the state of the system of atoms r and l can be described by @xmath45 . the photons poisson distribution with average photon number @xmath48 , @xmath49 , and @xmath50 , respectively ] after establishing entanglement within the basic links ,
the entanglement can be extend to double length by entanglement swapping with the setup schematically illustrated in fig .
[ fig:3 ] .
two pairs of atom - resonator systems , l - i@xmath51 , and i@xmath52-r , are prepared in states @xmath53 and @xmath54 , respectively .
two identical probe pulses of duration @xmath55 simultaneously come into the microtoroidal resonators @xmath56 and @xmath57 , respectively .
the forward propagating photons are combined together and then detected by the single - photon detector d at the midpoint of the optical length between atoms i@xmath51 and i@xmath52 . if and only if d records @xmath58 $ ] photons , i.e. , one of the atoms i@xmath51 and i@xmath52 is bright , the entanglement swap succeeds with the success probability @xmath59
otherwise , the entanglement fails to extend , and the previous entanglement generating and swapping process needs to be repeated , till the protocol is successful at last .
the detector d measures @xmath60 with @xmath61 .
successful measurement will project the state @xmath62 into @xmath63 with @xmath64 .
then , atoms @xmath65 are manipulated to make a unitary transform @xmath66 due to spontaneous emission .
finally , we obtain the state @xmath67 . this protocol for entanglement swap
can be repeated to extend the communication length .
we have the success probability @xmath68 , @xmath69 for the @xmath37th entanglement swap .
considering the time for the signal traveling from d to l and r ( see fig .
[ fig:1 ] ) to tell the controller whether or not to start the next process , the average total time required for successful distribution of entanglement over distance @xmath70 is @xmath71 where @xmath72 is the time needed to establish entanglement within the basic links with @xmath73 being the light speed in the optical fiber .
note that @xmath74 increases near linearly with the channel length .
configuration for entanglement swapping . ] with the entangled state @xmath45 between two distant sites in hand , we can apply the entanglement to quantum communication protocols , such as quantum teleportation , cryptography , and bell inequality test directly . for quantum cryptograph and bell inequality test
( see fig .
[ fig:4]a ) , we first make a local phase shift to transform the state @xmath75 to state @xmath76 @xcite . then , two raman beams are simultaneously applied to atoms l and r to make @xmath77 and @xmath78 rotation about axis x , respectively . finally , two probe pulses of duration @xmath79 are applied simultaneously to the resonators to measure the states of atoms r and l with the results 0 for the forward propagating photon number @xmath80 recorded by detector @xmath81 and 1 for @xmath82 . according to the ekert protocol for quantum cryptography @xcite , @xmath77 and @xmath78
are chosen randomly and independently from the set @xmath83 , the measurement results becomes the shared secret key if the two sides get information through the classic communication that they have chosen the same rotation .
for the bell inequality test , we obtain the correlations @xmath84 @xcite , where @xmath85 ( @xmath86 ) denotes the number of measurements with two same ( different ) results , and the result @xmath87 violates the chsh inequality @xmath88 @xcite .
note that the outcome of every bell inequality experiment is used due to the unity state detection efficiency .
combined with the space - like measurements , this scheme can be used for a loophole - free test of the bell inequality @xcite . to faithful transfer unknown quantum states @xmath89 and @xmath90 satisfying @xmath91 , we can use quantum teleportation protocol through the entangled state @xmath92 ( see fig . [ fig:4]b ) .
two identical probe pulses of duration @xmath93 simultaneously enter the microtoroids , the forward propagating photons are mixed together and measured by the single - photon detector d , which records the photon @xmath94 with @xmath95 .
if the detector records @xmath46 $ ] photons , the protocol succeeds with a probability of @xmath96 and the state @xmath97 is projected into a state @xmath98 .
then , atom i and l are manipulated to make a unitary transformation @xmath99 and decay to the ground state @xmath100 .
finally , we obtain the state @xmath101 .
( a ) configuration for the realization of quantum cryptography and bell inequality test .
( b ) configuration for quantum teleportation . ] because of the unambiguous detection of the state of qubits with unity detection efficiency through measuring many single - photons rather than one arising from the atom bright , the errors from , e.g. , the dark count , low detection efficiency , and the imperfection that the single - photon detector may not distinguish between one and two photons can be overcome in this scheme .
this results in the increase of the quantum communication speed and the corresponding fidelity . under the born approximation ,
the influence arising from the lost photons in the optical fiber and from the spontaneous emission in stages of entanglement extension and teleportation may be negligible .
owing to the limited coherence time @xmath102 , atoms may decay from metastable state @xmath8 to the ground state @xmath4 . to solve this problem ,
we may store the spin state in nuclear memory which has the longest decoherence time in all quantum systems so far . fast electron - nuclear double resonance ( endor ) pulse techniques may be used for prompt storage and retrieve of the electron - spin state @xcite . if two atom @xmath56 and @xmath57 in fig.[fig:3 ] decay from @xmath8 to @xmath4 , detection d measures @xmath103 photons , thus , the entanglement swap fails . in a similar way
, we can show that in the whole process of distributing entanglement over long distance , there may exist only one atom decaying with at most probability @xmath104 that we can not exclude through detection of @xmath36 photons to ensure only one atom bright , except for the two atoms apart with a distance @xmath105 .
the decoherence arising from the spontaneous decay can be written as @xmath106 . assuming @xmath107 km , @xmath108 m / s , @xmath109 km @xcite , from equation [ eq6 ] , we have @xmath110 ms , which is on the same order of the time @xmath111 ms . compared with the corresponding time @xmath112 s using the dlcz protocol for expected fidelity @xmath113 provided that many atoms excitations are the only imperfection in the experiment @xcite , the total average time for successful distributing entanglement over that distance can be reduced by 7 orders of magnitude .
assuming @xmath114 ns @xcite , @xmath115 ms , we have @xmath116 . stationary and non - stationary phase shifts from stationary and non - stationary channels and set - up asymmetries , respectively , are main decoherence source in our scheme .
the non - stationary phase shifts increase with the length by the random - walk rule @xmath117 and can be reduced to a negligible degree @xcite . because the total average time @xmath74 decreases significantly by several orders of magnitude , this phase shifts in our scheme can be overcome much easier than that in dlcz protocol @xcite .
the stationary phase shifts are easier to handle than the non - stationary ones are .
the mechanism for photon blockade used in our scheme is very robust against many experimental imperfections @xcite . as a summary
, we present a microtoroidal resonator - based scheme for efficient long - distance quantum communication .
moreover our scheme shows high fidelity and good robustness against many experimental imperfections . to trap single atoms near the surface of the microtoroidal resonator
is still a technical challenge .
however , with the rapid advances in the relevant technologies it is no doubt that in near future the obstacle will be overcome . this scheme may open up realistic probability of efficient long - distance quantum communication . |
a significant number of studies were devoted to the estimation of power spectra and modelling of irregularly sampled time series ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . the descriptive statistics such as the mean , variance , higher central moments and robust equivalents
: they can summarize essential features of signals in a few numbers employing straightforward computations , which makes them excellent precursors of more detailed analyses like modelling and classification .
herein , a simple effects of irregular sampling on the characterization of a signal in terms of statistical parameters is presented .
separate works describe corrections of biases induced by small sample sizes and gaussian uncertainties in the calculation of weighted moments and cumulants @xcite .
time series sampled irregularly or with varying candences may lead to very different estimates of statistical parameters for the same signal .
for example , if a sinusoid is sampled mostly at maximum or minimum , the mean is offset by about the amplitude of the signal , the variance might be much smaller than expected ( since most points sample the same region of the signal ) , and the distribution appears very skewed ( the few non - clumped measurements form an asymmetric tail in the distribution of measurements ) .
if the same clump of data was in proximity of the average level , instead , the mean and skewness values could be close to the correct value ( by serendipity ) , but the variance would be much smaller than the true one .
sampling - induced biases do not arise from sparsely sampled data only and they might manifest independently of the number of measurements .
while time series with more measurements are generally associated with better coverage of signal features , the importance implicitly assigned to different parts of the signal by unweighted estimators is related to the relative frequency of measurements . in principle
, the most accurate and precise statistical parameters could be inferred from the model of a signal .
alas , models are often complicated , in the attempt to describe the signal features under many circumstances , they might require lengthy processing and their accuracy can not be guaranteed in all cases . for example , fourier series are well fitted to model periodic signals , but the description of sharp features , like those present in ea - type eclipsing binaries , requires a high number of harmonics , which can overfit smoother parts of the signal and cause unrealistic excursions in large intervals with no data @xcite .
linear interpolation is one of the simplest methods to approximate a model , assuming the signal features are sufficiently sampled
. this work estimates statistical parameters by averaging the linear interpolation of functions of time series measurements ( depending on the specific estimator ) .
if the signal can be recognized in time domain , interpolation can be performed in time without requiring further information .
if a sparsely sampled signal is primarily mono - periodic , it can be interpolated in phase by folding the time series with the corresponding period .
while the interpolated function might include profiles with spiky artefacts , its average is more robust and can be expressed as a weighted mean which assigns more relevance to scattered than clumped measurements .
statistical parameters weighted by such a scheme were tested on data from the _ hipparcos _ periodic catalogue and led to a significant general improvement in the accuracy and precision of estimator values and automated classification results , with respect to those obtained with the unweighted or error - weighted counterparts .
this paper is organized as follows .
after the definition of the notation and terminology in sec .
[ sec : notation ] , the description of weighted estimators as averages of linear interpolations is presented in sec .
[ sec : method ] , with weights defined in time and phase , including adaptations for low signal - to - noise regimes and small sample sizes .
the new weighting scheme is applied to time series from the _ hipparcos _ periodic catalogue in sec .
[ sec : hip ] , which includes a comparison of the values of statistical parameters computed with different weighting schemes and their effect on automated classification .
the conclusions are drawn in sec .
[ sec : concl ] , followed by a series of appendices with more details on the interpolation of ( mono-)periodic signals ( appendix [ app : periodicinterpolation ] ) , the illustration of modelled light curves to verify the accuracy of estimators ( appendix [ app : data ] ) , the definitions of the statistical parameters employed ( appendix [ app : def ] ) and additional scatter plots of estimators ( appendices [ app : scatter ] and [ app : scatterunweighted ] ) .
for a set of @xmath1 measurements @xmath2 , the following quantities are defined : * the population central moments of order @xmath3 around the mean @xmath4 are denoted by @xmath5 and respective cumulants @xmath6 , @xmath7 and @xmath8 ( e.g. , * ? ? ?
sample central moments @xmath9 , with respective cumulants @xmath10 , where @xmath11 and @xmath12 .
* standardized skewness and kurtosis are @xmath13 and @xmath14 , with population values @xmath15 and @xmath16 , respectively .
* noise - unbiased estimates ( sometimes called ` denoised ' for brevity ) of central moments and cumulants @xcite are denoted by an asterisk superscript . *
no systematics or instrumental errors are considered herein and uncertainties are often referred to as errors . *
the accuracy of an estimator is related to its distance from the true value and thus combines the concepts of bias and precision , while the classification accuracy rate is the ratio between the number of true positives and the total number of ( of a given type ) . * the precision of an estimator is quantified by its dispersion , while the classification precision rate is the ratio between the number of true positives of a given type and the total number of classified as such a type .
the proposed weighting scheme is derived from linear interpolation and thus all of its pros and cons are inherited by definition . strictly speaking , statistical estimators like the mean and moments
are not defined for deterministic signals .
however , such estimators can still give acceptable approximations when the time scale of features ( or period ) is much smaller than the time series duration .
in the case of mono - periodic signals , the proposed weighting scheme leads to estimator values of the signal as if the latter was measured on a complete cycle .
( population ) value of a statistical parameter @xmath17 of a continuous signal @xmath18 in time @xmath19 is computed by integrating function of such a signal : @xmath20 in case of regular sampling in fine intervals , sum of discrete elements tends to the population result for infinitely small intervals .
this limit is not necessarily satisfied when a continuous signal is sampled irregularly .
the weighting scheme described in sec .
[ sec : weights ] recovers the property that the sum of discrete ( weighted ) terms approaches the result of the continuous function , in the limit of very dense ( not necessarily uniform ) sampling .
linear interpolation is a simple method to approximate a function with a broken line connecting the data points , provided the function is sufficiently sampled in time or phase . in this section , such a function
is represented by the expression of the additive terms of a statistical estimator . for example , a central moment of order @xmath3 is defined by the average of terms of the form @xmath21 .
results from linearly interpolating such a function are equivalent to those obtained from an effectively infinite regular sampling of the interpolated function . in sec .
[ sec : weights ] , it is shown that the average of a linear interpolation can be expressed as a weighted mean .
central moments of a sinusoidal signal are illustrated in fig .
[ fig : illustration ] and they are related to the areas enclosed by @xmath22 , which are approximated by terms ( ) with weights ( ) adapting to the sampling density in phase . is sampled unevenly by 10 measurements @xmath23 ( red circles ) and shown in the panel on the top - left hand side .
the other panels illustrate different powers of deviations from the mean of the true signal ( blue curve ) and of the measurements ( red circles ) .
in particular , @xmath24 identifies terms associated with the central moments of order @xmath3 : these estimator values are related to the areas enclosed by the blue curves and the zero level ( shaded in blue ) .
the weights defined by eqs ( [ eq : pwi])([eq : pwn ] ) define the variable widths of the shaded bars , which reduce the contribution of clustered measurements and increase the one of scattered data . in this case , although most measurements sample the first half of the signal , the increased weights assigned to the remaining data provides a better estimate of the area enclosed by the second half of the signal than using unweighted schemes . ] since more importance is given to some measurements at the expense of other ones ( e.g. , a large sample with only a few important elements will be similar to one with only these few elements ) and exploiting correlations in the data with weights ( e.g. , assigning small weights to measurements separated by small intervals because their values are expected to be similar ) might introduce biases if applied to expressions assuming independent data .
however , small biases could be justified by significant improvements in precision and a mixed weighting scheme which balances precision and accuracy depending on the signal - to - noise ( @xmath25 ) level and sample size is described in sec .
[ sec : mixed ] . if the targeted features of signals are sufficiently sampled in time , as for the data from the _ kepler _ @xcite and _ corot _
@xcite missions , linear interpolation is expected to provide reliable approximations of and weights might be expressed in time domain .
if typical time intervals between measurements are larger than the resolution needed to sample the signal ( as it is often the case in the _ hipparcos _ and _ gaia _ surveys ) but the latter is mono - periodic or dominated by a single period , sampling can be improved by folding the light curve with the value of the fundamental period and computing interpolations ( or weights ) in phase . the interpolation - based weighting scheme .
if gaps in time or phase occasionally cover a relevant fraction of a signal , measurements at the gap boundaries might not provide sufficient information on the features of the signal and the importance of such data could be limited , for example , to some maximum weight value .
this strategy can extend the applicability of the weighting scheme to under - sampled signals , periodic or non - periodic , to mitigate statistical biases due to data clumps on scales smaller than the signal features if sampling in time is sparse but the data exhibit periodicity , the time series can be folded with the dominant period to increase the average sampling rate in phase ( by a factor the duration of the time series divided by the period ) . since interpolation of phase - sorted data
is carried over from the last to the first point in phase , weights in phase @xmath35 corresponding to eqs ( [ eq : twi_delta])([eq : twn_delta ] ) become ( see appendix [ app : periodicinterpolation ] ) : @xmath36 with @xmath37 . [ [ sec : mixed ] ] in order to avoid interpolating large noise fluctuations , weights could be set to reduce to inverse - squared uncertainties at low @xmath25 , which proved more precise in the simulations described by @xcite .
the transition between high and low @xmath25 regimes could be pursued with function of the form @xmath38 could be used to reduce the relevance of interpolation - based weights for small sample sizes @xmath1 @xmath39/n , \end{aligned}\ ] ] where @xmath40 is defined by eqs ( [ eq : twi])([eq : twn ] ) or ( [ eq : pwi])([eq : pwn ] ) . tuning parameters @xmath41 offer the possibility to reach a compromise solution between precision and accuracy at high and low values of @xmath1 and @xmath25 ratios , according to the specific set of estimators , signals , sampling , errors , sample sizes and their distributions in the data .
the effect of interpolation - based weights on statistical parameters was explored for a realistic distribution of signals , represented by time series from the _ hipparcos _ catalogue of periodic variable stars .
the _ hipparcos _
mission @xcite performed astrometric and photometric measurements of the brightest sources in the sky .
the full catalogue @xcite contains 118204 sources with photometry . among the 11597 stars identified as variables ,
a reliable period could be computed ( or was consistent with the one from literature ) for 2712 objects , which were published in the periodic catalogue ( vol .
* ; * ? ? ?
this set of sources , which contained light variations dominated mostly by single periods , was chosen to illustrate the application of one of the weighting schemes described in sec .
[ sec : method ] . in order to assess the accuracy of estimators with respect to the ( unknown ) real signal , the latter was assumed to be represented by a model of the time series and
were simulated employing uncertainties as gaussian random variables ( around the model ) .
only good quality measurements were accepted , flagged by the field ht4 as zero or one ( see vol . 1 of * ? ? ? * ) , which reduced the number of sources considered to 2683 .
only ) were : 1196 , 10027 , 17878 , 20570 , 24019 , 25673 , 39084 , 42715 , 42726 , 46502 , 52538 , 53937 , 58112 , 60904 , 61997 , 63125 , 69582 , 72583 , 88905 , 90026 , 93595 , 93724 , 96007 , 99675 , 102246 , 102409 , 112317 , 112470 and 118188 . ]
time series were folded with the period provided by the _ hipparcos _
catalogue and modelled by a cubic smoothing spline ( the ` smooth.spline ` function from the ` stats ` package in r , * ? ? ?
the condition of periodicity at the boundaries was approximated by replicating a whole cycle of folded data before and after the cycle considered as reference for the ` true ' statistical parameters and for the generation of simulated data .
the smoothing parameter was estimated from the data with a generalized cross - validation method ( e.g. , * ? ? ?
* ) and the degrees of freedom were adjusted in special cases , to mitigate over - fitting highly clumped data with large gaps as well as under - fitting the profile of eclipsing binaries . of the data , which identified light variations typical of eclipsing binaries , and the ratio @xmath42 between the median and the third largest gaps in phase , to better deal with clumped data with large gaps . denoting by @xmath1 the number of measurements in a time series , ` df ` equalled 15 , 24 and 36 for @xmath42 less than @xmath43 , @xmath44 and @xmath45 , respectively , provided @xmath46 . for greater values of @xmath47 , ` df ` was set to the smallest integer not less than @xmath48 .
the quoted values of ` df ` take into account the replicated data at each of the extremes of the folded light curve ( to induce quasi - periodic boundary conditions on the model of the central cycle ) . ]
the combination of large gaps and sharp features did not lead to accurate models and other methods might help avoid modelling artefacts in gapped data , although these are out of the scope of this application .
also , cases in which sparse sampling missed important features could not be improved .
nevertheless , the smoothed best - fitting curves seemed to capture the relevant shapes of true signals in most cases .
the resulting models are presented together with the original and simulated data in appendix [ app : data ] .
the difficulty to achieve accurate models for all sources and the need of more complex modelling techniques were confirmed . while models were sometimes not ideal , most of them were of sufficient quality to supply a realistic distribution of the relevant features of the _ hipparcos _ periodic variable stars .
differences from data of other surveys were expected to be greater than those due to modelling inaccuracies .
less than one per cent of all sources ( 23 time series ) with significant modelling artefacts in gaps with no data were removed ( many of them were ea - type eclipsing binaries with large data gaps ) , thus 2660 objects were included in the assessment of statistical parameters .
the number of measurements per time series ranged from 18 to 331 , with @xmath25 ratios from 0.2 to 116 , according to the definition in eq .
( [ eq : snr ] ) .
the _ hipparcos _ light curves were folded with the catalogue period and interpolation - based weights were computed in phase . for brevity , such weights are referred to as ` phase weights ' , while inverse - squared uncertainties are called ` error weights ' .
the estimators employed herein included weighted moments and cumulants corrected for biases from gaussian uncertainties ( called noise - unbiased or denoised estimators ; see * ? ? ?
* ) and some robust weighted measures , as defined in appendix [ app : def ] .
simulations in @xcite suggested that noise - unbiased phase - weighted sample moments can be more accurate for @xmath49 and sample sizes @xmath50 with respect to other schemes , while error weighting appeared the most appropriate option for noisy signals .
thus , the weighting scheme chosen for this application combined error and phase information as described by eqs ( [ eq : mixed ] ) and ( [ eq : pwi])([eq : pwn ] ) .
the balance between phase weights ( at high @xmath25 ) and error weights ( at low @xmath25 ) was controlled by the parameters @xmath51 and @xmath52 , which provided a satisfactory overall accuracy for the set of estimators and light - curve shapes of the _ hipparcos _ periodic variable stars .
the signal - to - noise ratio @xmath25 was estimated from the true ( model ) signal variance @xmath53 and the average of squared measurement uncertainties @xmath54 as follows : @xmath55^{1/2}. \label{eq : snr}\ ] ] the deviations of error - weighted sample statistical parameters from population values ( computed from a fine regular sampling of the models ) are compared to the noise - unbiased error - phase weighted counterparts as a function of @xmath25 ratio in figs [ fig : noise1][fig : noise2 ] .
error - phase weighted ( and noise - unbiased , when applicable ) estimators were generally more accurate than error - weighted estimators .
since larger light variations were correlated with higher @xmath25 levels , the effect of the correlation between errors and magnitudes ( e.g. , see * ? ? ?
* ) was visible when weighting by errors at high @xmath25 : the mean was biased towards brighter measurements , the variance tended to be smaller than the real value and the scatter of higher moments around the true values was greater .
such effects were significantly alleviated by error - phase weights . also , the accuracy of the noise - unbiased variance @xmath56 and kurtosis moment @xmath57 were much improved at @xmath58 . in the particular case of the standardized kurtosis
, the value of @xmath59 equalled the one of the cumulant @xmath60 by definition , for any weighting scheme ( unlike the noise - unbiased counterparts ) .
involves a term in @xmath61 , which does not simplify after the normalization by @xmath62 . ]
( a ) + + ( c ) + + ( e ) + ( b ) + + ( d ) + + ( f ) + ( a ) + + ( c ) + ( b ) + + ( d ) + estimators which provide similar information ( such as the mean and median , variance and interquartile range , skewness and kurtosis standardized by the estimated or true variance ) are compared in scatter plots in appendix [ app : scatter ] .
although contributions from different @xmath25 ratios are not distinguished , error - phase weighted ( and noise - unbiased , when applicable ) estimators are always associated with more strongly peaked distributions around the true values than the error - weighted analogues . in order to quantify the effect of error - phase weighting and additional denoising with respect to simple error weighting ,
the top panels in figs [ fig : quantile1][fig : quantile3 ] indicate the fraction of sources associated with estimators improved by phase weights and noise correction .
the accuracy of a generic estimator @xmath63 is assessed by its distance @xmath64 from the true value @xmath65 .
the abscissas in figs [ fig : quantile1][fig : quantile3 ] represent the difference between the distance of error - phase weighted ( and noise - unbiased , when applicable ) estimators and the one of error - weighted estimators from the correct values : negative differences indicate smaller distances to the true values and thus improved accuracy with respect to error - weighted estimators .
the accuracy of error - phase weighted and noise - unbiased estimators improved in 70-to-90 per cent of the cases and deteriorations were typically smaller in frequency and magnitude than improvements . in the singular case of the standardized skewness @xmath66 , denoising worsened almost 8 per cent of the estimates with respect to error - phase weighting . as apparent in fig .
[ fig : quantile2]a , the accuracy of the standardized noise - unbiased skewness worsened at low @xmath25 ratios .
this was expected because the noise - unbiased variance might be easily underestimated when uncertainties are of the same order as the signal , and the normalization of the skewness by a much smaller number than the correct one could rapidly degrade the precision of the estimator .
a similar ( less pronounced ) tendency at low @xmath25 ratios was also observed in the case of the standardized kurtosis .
( a ) + + ( c )
+ ( b ) + + ( d ) + ( a ) + + ( c ) + ( b ) + + ( d ) + ( a ) + ( b ) + as noted in @xcite , the noise - unbiased variance can become too small at low @xmath25 levels , overestimating the standardized noise - unbiased skewness and kurtosis , or leading to an undetermined skewness value in case of non - positive variance . on the other hand , when the variance is not corrected by noise biases , it is generally overestimated , and if the true values of skewness or kurtosis are sufficiently close to zero , the standardized noise-_biased _ skewness and kurtosis become more accurate ( by serendipity ) .
the effect of statistical parameters weighted by different schemes on automated classification was assessed by comparing the classification accuracy and precision as a function of variability type employing unweighted , error - weighted , phase - error weighted and noise - unbiased phase - error weighted estimators .
automated classification of stellar variability types was pursued with a set of attributes which characterized features of different classes .
some studies employed only information from light - curve modelling @xcite , with additional statistical parameters @xcite or colour information @xcite , while attributes from modelling , statistical and astrophysical quantities were used in @xcite .
the list of attributes employed for classification herein was restricted to descriptive statistics ( mean , median , variance , interquartile range , normalized and non - normalized skewness and kurtosis ) , in order to test the effect of different weighting schemes on classification .
only variability types which could be identified by the distribution of measurements were included in the training set , such as eclipsing binaries , rr lyrae , mira , @xmath67 cephei and alpha@xmath0 canum venaticorum stars , accounting for over 60 per cent of all sources in the _ hipparcos _ periodic catalogue . in these classification experiments , statistical parameters
did not depend on modelling and were computed on the original ( not simulated ) data , which thus included the sources previously excluded because of modelling issues .
the same quality flags , periods and weighting schemes were applied as described in sec .
[ sec : pars ] .
since the true signal variance was unknown , the @xmath25 ratio was estimated by substituting the unknown @xmath53 in eq .
( [ eq : snr ] ) with the noise - unbiased phase - weighted sample variance @xmath68 and by weighting the average of squared errors in the denominator with the same weights for consistency : @xmath69^{1/2}.\ ] ] the @xmath25 ratios of the sources selected for classification ( as described in sec .
[ sec : dataselection ] ) spanned a range from 0.4 to 140 .
sources from the _ hipparcos _ periodic catalogue were cross - matched with classifications from literature as available in the variable star index @xcite http://www.aavso.org/vsx/help/variablestartypedesignationsinvsx.pdf[aavso.org/vsx/help/variablestartypedesignationsinvsx.pdf ] ] of the american association of variable star observers ( aavso ) with the nearest object within 1 arcsec .
uncertain classifications ( with class labels followed by the mark ` :') were excluded , unless the uncertainty referred to properties of eclipsing binaries other than light - curve shapes , such as the physical characteristics , the luminosity class of the components , or the degree of filling of the inner roche lobes .
objects associated with combinations of different classes ( with labels joined by the symbol ` + ' ) were not addressed herein .
the subset of variability types chosen to test classification with statistical parameters are listed in table [ tab : ts ] and include 1605 sources associated with class labels ea , eb , ew , acv , m , rrab , rrc , dcep and dceps , defined in table [ tab : ts ] together with the corresponding sample sizes .
sources which poorly represented their class were not removed from the training set in order to avoid the introduction of selection biases ( e.g. , by favouring objects with higher @xmath25 ratios ) and because the focus of this application was related to the _ relative _ classification accuracy employing attributes with different weighting schemes .
llr variability type&label&number + eclipsing binary : algol type&ea&409 + beta lyrae type&eb&208 + w ursae majoris type&ew&170 + alpha@xmath0 canum venaticorum&acv&183 + mira ceti&m&218 + rr lyrae : asymmetric light curve&rrab&139
+ nearly symmetric light curve&rrc&28 + delta cephei & dcep&220 + first overtone pulsators&dceps&30 + + & total:&1605 + the distributions of a selection of error - phase weighted ( and noise - unbiased , when applicable ) statistical parameters as a function of variability type are presented in fig .
[ fig : scatter ] . the mean or median magnitudes of sources at broadly different distances from the observer do not provide information on the intrinsic source properties . however , the correlation between magnitude and noise , coupled with the interquartile range or variance ( as shown in fig .
[ fig : scatter]a ) , is related to the @xmath25 level and thus to an observational selection .
figure [ fig : scatter]b illustrates clearly the difference between robust and non - robust estimators for the same quantity , such as the interquartile range versus the variance , which proves effective at separating eclipsing binaries ( mostly ea and some fraction of eb types ) .
the interquartile range and standardized skewness are presented in fig .
[ fig : scatter]c : the skewness separates the eclipsing binary subtypes from most other classes , which are better distinguished by the interquartile range . a similar scatter plot is shown in fig .
[ fig : scatter]d in terms of normalized and non - normalized kurtosis moments , with the difference that some classes occupy different relative loci ( such as the ea types and the rrab with respect to dcep variables ) . the estimators and variability types illustrated in fig .
[ fig : scatter ] are also presented for the unweighted case in appendix [ app : scatterunweighted ] .
the distributions of most unweighted estimators in fig .
[ fig : scatterunweighted ] tend to be more scattered than the ones shown in fig . [
fig : scatter ] . in the case of acv stars , instead , the unweighted skewness and kurtosis are less scattered , because these sources have very low @xmath25 ratios ( typically @xmath70 ) and the correction of noise biases at such @xmath25 levels can decrease significantly the precision of higher moments @xcite .
another noticeable difference between figs [ fig : scatter ] and [ fig : scatterunweighted ] is related to the larger skewness @xmath66 of mira stars in the unweighted case , which can be understood by the systematic difference in the uncertainties associated with faint and bright measurements ( e.g. , * ? ? ? * ) .
such a difference is enhanced by the large amplitudes of light variations of mira types , so that greater uncertainties on the faint side of the signal generate more faint than bright ` outliers ' , biasing the unweighted skewness towards greater values .
( a ) + + ( c ) + ( b ) + + ( d ) + automated classification tests were performed employing random forests @xcite .
random forest is an accurate tree - based classification method ( e.g. , see * ? ? ?
* ) , quite robust to outliers and strongly correlated attributes .
the accuracy of the classifier was estimated from a subset of sources ( about one - third of all objects ) randomly omitted from the learning process and thus called ` out - of - bag ' sources .
the importance of an attribute for classification was measured by the mean decrease in accuracy after permuting the values of that attribute in the out - of - bag sources ( see * ? ? ? * for more details ) .
the random forest classifier was trained with estimators employing a single weighting scheme per run , i.e. , unweighted , error - weighted , error - phase weighted or noise - unbiased error - phase weighted . for each weighting scheme ,
random forest was executed 1000 times with 500 trees and the classification accuracy and precision rates were aggregated after each run in order to assess their mean and dispersion .
the mean classification accuracy and precision rates are listed as a function of variability type and attribute weighting scheme in tables [ tab : res1 ] and [ tab : res2 ] , and illustrated in figs [ fig : classresults]a and [ fig : classresults]b , respectively .
training - set sources were reclassified by sole statistical parameters with an overall accuracy from 77.0 to 82.6 per cent on average , depending on the weighting scheme adopted .
unweighted estimators led to the smallest accuracy level , closely followed by error - weighted estimators at 77.6 per cent .
the introduction of phase weights improved the accuracy to 82.3 per cent and noise - unbiased error - phase weighted estimators achieved the best overall classification accuracy of 82.6 per cent ( with uncertainty at the level of 0.2 to 0.3 per cent ) .
the overall trends were generally reflected by single variability types with only a few exceptions . in particular , error - weighted estimators of rrab types led to significantly worse classification accuracy than unweighted estimators .
this was explained by the relatively large amplitude of rrab variables , which enhanced the systematic difference of uncertainties between bright and faint measurements in the light curve : weighting by errors decreased the importance of faint measurements , increasing the similarity ( and thus confusion ) with dcep and m - type light curves . in the case of dceps variables ,
denoising was marginally counterproductive for accuracy , although still significantly better with respect to the unweighted or error - weighted schemes .
this was expected to be related to the increased confusion with the noise - unbiased estimators of ew types , which were more numerous and extended to lower @xmath25 ratios than the dceps stars : while the distributions of the standardized error - phase weighted kurtosis moments of dceps and ew types could just be separated , the additional denoising decreased the values of standardized kurtosis moments more for the ew than the dceps stars , leading to overlapping distributions dominated by the ew types . @l@ c@ c@ c@ c@ var . & & error & error - phase & err .- ph .
w. + type&unweighted&weighted&weighted & & denoised + ea & 87.7 @xmath71 0.3 & 87.7 @xmath71 0.4 & 88.9 @xmath71 0.4 & 89.0 @xmath71 0.3 + eb & 53.7 @xmath71 1.1 & 56.4 @xmath71 1.0 & 58.7 @xmath71 0.9 & 59.8
@xmath71 1.0 + ew & 59.2 @xmath71 1.4 & 60.9 @xmath71 1.2 & 68.5 @xmath71 1.0 & 70.2 @xmath71 1.0 + acv & 84.9 @xmath71 0.9 & 85.9 @xmath71 0.9 & 88.8 @xmath71 1.0 & 87.6 @xmath71 0.7 + m & 96.7 @xmath71 0.2 & 96.5 @xmath71 0.3 & 98.0 @xmath71 0.3 & 98.2 @xmath71 0.3 + rrab & 83.1 @xmath71 0.8 & 79.4 @xmath71 1.0 & 88.3 @xmath71 0.6 & 87.8 @xmath71 0.8 + rrc & 41.2 @xmath71 3.0 & 47.0 @xmath71 4.0 & 41.8 @xmath71 3.1 & 50.8 @xmath71 3.3 + dcep & 73.9 @xmath71 0.8 & 74.7 @xmath71 0.7 & 86.6 @xmath71 0.6 & 87.1 @xmath71 0.6 + dceps & 31.3 @xmath71 3.5 & 35.1 @xmath71 3.8 & 56.6 @xmath71 2.2 & 51.4 @xmath71 1.9 + + all & 77.0 @xmath71 0.3 & 77.6 @xmath71 0.3 & 82.3 @xmath71 0.2 & 82.6 @xmath71 0.2 + precision rates of classification per variability type showed successive improvements ( consistent within uncertainties ) from the unweighted to the error - weighted , error - phase weighted and noise - unbiased schemes .
a single exception was related to mira stars : their classification precision with error - weighted estimators was about one per cent worse than with unweighted attributes , as a consequence of the increased contamination by rrab stars ( as explained above ) .
@l@ c@ c@ c@ c@ var . & & error & error - phase & err .- ph .
w. + type&unweighted&weighted&weighted & & denoised + ea & 88.5 @xmath71 0.4 & 89.4 @xmath71 0.4 & 91.2 @xmath71 0.4 & 91.0 @xmath71 0.3 + eb & 58.4 @xmath71 1.0 & 59.9 @xmath71 1.0 & 63.1 @xmath71 0.9 & 64.5
@xmath71 0.8 + ew & 58.5 @xmath71 1.1 & 59.4 @xmath71 1.0 & 65.5 @xmath71 0.9 & 65.2 @xmath71 0.9 + acv & 72.4 @xmath71 0.5 & 74.8 @xmath71 0.6 & 76.3 @xmath71 0.5 & 76.6 @xmath71 0.5 + m & 98.2 @xmath71 0.2 & 97.2 @xmath71 0.2 & 98.6 @xmath71 0.3 & 98.8 @xmath71 0.3 + rrab & 80.8 @xmath71 0.7 & 80.5 @xmath71 0.9 & 89.8 @xmath71 0.8 & 90.5 @xmath71 0.9 + rrc & 52.5 @xmath71 3.2 & 56.5 @xmath71 3.3 & 82.4 @xmath71 4.9 & 88.4 @xmath71 4.3 + dcep & 74.7 @xmath71 0.8 & 74.7 @xmath71 0.8 & 85.3 @xmath71 0.6 & 85.4 @xmath71 0.6 + dceps & 37.9 @xmath71 3.5 & 39.7 @xmath71 3.3 & 55.4 @xmath71 1.9 & 56.2 @xmath71 1.8 + + all & 77.0 @xmath71 0.3 & 77.6 @xmath71 0.3 & 82.3 @xmath71 0.2 & 82.6 @xmath71 0.2 + and [ tab : res2 ] .
class labels are defined in table [ tab : ts ] and colour coded as shown in the legend of each panel .
, title="fig : " ] + and [ tab : res2 ] .
class labels are defined in table [ tab : ts ] and colour coded as shown in the legend of each panel .
, title="fig : " ] the importance of phase - weighted ( and noise - unbiased , when applicable ) attributes as a function of variability type from a run of random forest is depicted in fig .
[ fig : importance ] , with darker cells indicating more important attributes for a given class .
for example , the skewness confirmed to be particularly useful to distinguish eclipsing binaries and the asymmetric light curves of rrab stars .
also , large and small amplitude variables , such as acv and m types , could easily be separated by the variance . .
class labels are described in table [ tab : ts ] . ] the confusion matrix from a run of random forest , employing error - phase weighted ( and noise - unbiased , when applicable ) estimators , is shown in fig .
[ fig : cm ] .
the overall classification accuracy was 83 per cent and reached 92 per cent after aggregating the subtypes of eclipsing binary , rr lyrae and delta cephei into their superclasses .
the improvement in classification accuracy , with respect to results employing unweighted estimators , was about 6 per cent in both cases of separated and aggregated variability subtypes .
the level of confusion between eclipsing binaries of ea , eb and ew types was expected from the natural overlap in their definitions .
the misclassification of many acv types as eclipsing binaries was attributed to the restricted set of attributes employed herein ( e.g. , the attribute qso variations in * ? ? ? * proved to be useful in this context ) .
similarly , the confusion between rr lyrae and delta cephei variables would not have occurred if the period was included in the set of classification attributes . on the other hand , simple amplitude estimators such as variance and interquartile range were sufficient to separate mira stars from most other classes with high accuracy ( as apparent in fig .
[ fig : scatter ] ) . .
a simple weighting scheme based on linear interpolation , applicable in phase or time , was proposed in order to improve the accuracy of descriptive statistics of unevenly sampled time series .
weights represented by time intervals could be applied to well - sampled signals . for sparsely sampled but periodic time series ( dominated by a single period ) ,
light curves could be folded with the fundamental periods and weights expressed in terms of phase intervals .
the _ hipparcos _ catalogue of periodic variables represented a suitable test bed of unevenly sampled light curves with realistic distributions of variability types to investigate the accuracy of estimators weighted by different schemes .
noise - unbiased estimators weighted by phase intervals ( or inverse - squared uncertainties at low @xmath25 ratios ) improved the accuracy for 70-to-90 per cent of the sources with respect to the values of the error - weighted counterparts .
deteriorations in accuracy were observed in 10-to-30 per cent of the cases , with magnitudes typically smaller than the ones corresponding to improvements .
automated classification experiments confirmed that the best overall results were achieved employing the set of the most accurate attributes , i.e. , the noise - unbiased estimators weighted by phase intervals and inverse - squared uncertainties at high and low @xmath25 ratios , respectively .
the overall improvement in classification accuracy with respect to the result employing unweighted estimators was about 6 per cent and most of it was related to the introduction of phase weights which adapted to the varying sampling density of the light curve .
the author is grateful to m. sveges , p. dubath for useful comments on the original manuscript .
99 auvergne m. et al . , 2009 , a&a , 506 , 411 blomme j. et al . , 2010 ,
apj , 713 , l204 blomme j. et al . , 2011 , mnras , 418 , 96 borucki w.j .
et al . , 2010 , science , 327 , 977 breiman l. , 2001 , machine learning , 45 , 5 carbonell m. , oliver r. , ballester j.l . , 1992 , a&a , 264 , 350 debosscher j. , sarro l.m . , aerts c. , cuypers j. , vandenbussche b. , garrido r. , solano e. , 2007 , a&a , 475 , 1159 debosscher j. et al . , 2009 , a&a , 506 , 519 dubath p. et al . , 2011 , mnras , 414 , 2602 european space agency , 1997 , the hipparcos and tycho catalogues , esa sp-1200 eyer l. , 1998 , les toiles variables de la mission hipparcos , phd thesis n^o^ 3002 , universit de genve eyer l. , mignard f. , 2005 , mnras , 361 , 1136 eyer l. , grenon m. , falin j .- l .
, froeschle m. , mignard f. , 1994 , solar physics , 152 , 91 koen c. , 2005 , mnras , 361 , 887 liaw a. , wiener m. , 2002 , r news , 2 ( 3 ) , 18 perryman m.a.c .
et al . , 1997 , a&a , 323 , l49 perryman m.a.c .
et al . , 2001 ,
a&a , 369 , 339 r development core team , 2013 , r : a language and environment for statistical computing , r foundation for statistical computing , vienna , austria richards j.w .
et al . , 2011 ,
apj , 733 , 10 rimoldini l. et al . , 2012 ,
mnras , 427 , 2917 rimoldini l. , 2013a , preprint ( http://xxx.lanl.gov/abs/1304.6564[arxiv:1304.6564 ] ) rimoldini l. , 2013b , preprint ( http://xxx.lanl.gov/abs/1304.6715[arxiv:1304.6715 ] ) ruppert d. , wand m.p . ,
carroll r.j . , 2003 , nonparametric regression , cambridge university press sarro l.m .
, debosscher j. , aerts c. , lpez m. , 2009 , a&a , 506 , 535 stuart a. , ord j. , 1969 , kendall s advanced theory of statistics , charles griffin & co. ltd , london van leeuwen f. , 1997 , sp .
, 81 , 201 vio r. , strohmer t. , wamsteker w. , 2000 , pasp , 112 , 74 watson c. , henden a.a . , price a. , 2012 , vizier online data catalogue , aavso international variable star index , version 2012 - 09 - 23
light curves of the _ hipparcos _ periodic variables are illustrated in fig .
[ fig : lc ] .
panels on the left - hand side show the _ hipparcos _ data folded with the catalogue period , while the right - hand panels present data simulated around the model according to the measurement uncertainties ( assumed gaussian ) and the observed phases .
simulated data were employed to assess the accuracy of statistical estimators ( with respect to the model ) , while classification was performed on the original data .
figure [ fig : lc ] provides sample light curves for 4 of the 2683 _ hipparcos _ sources available online ( see supporting information ) .
only ) were not included in fig .
[ fig : lc ] . ]
the statistical estimators employed to test different weighting schemes are defined below .
sample noise - unbiased weighted moments and cumulants are defined following @xcite . denoting the sample weighted central moments of order @xmath3 by @xmath79 and representing the gaussian uncertainties by @xmath54 , the noise - unbiased estimators ( starred ) are defined as follows : @xmath80 -\frac{3}{w^4}\left(\sum_{i=1}^{n } w_i^2\epsilon_i^2\right)^2\\ ( m_2 ^ 2)^ * = & \left(m^*_2\right)^2 -\frac{4}{w^2}\sum_{i=1}^{n } w_i^2\epsilon_i^2 \left [ \left(x_i-\bar{x}\right)^2 -\frac{\epsilon_i^2}{2}\left(1-\frac{2w_i}{w } \right)\right]+\nonumber\\ & + \frac{2}{w^4}\left(\sum_{i=1}^{n } w_i^2\epsilon_i^2\right)^2\\
k^*_4 = & \,m^*_4 - 3\,(m_2 ^ 2)^ * \\ g^*_1=&\,k^*_3/(k^*_2)^{3/2 } \\ g^*_2=&\,k^*_4/(k^*_2)^2.\end{aligned}\ ] ] percentiles depend on the rank of sorted values , thus they are less sensitive to extreme values than moments and cumulants which involve powers of deviations from the mean and average over all elements . the @xmath81-th percentile @xmath82 is defined as the ( interpolated ) value such that @xmath81 per cent of the data are smaller than @xmath82 .
two common percentiles are the median @xmath83 and interquartile range @xmath84 . denoting the list of measurements @xmath23 sorted in increasing values by @xmath85 , associated with weights @xmath86 , respectively , the @xmath81-th weighted percentile @xmath82 is defined as follows : @xmath87 where @xmath88
figures [ fig : scatter1][fig : scatter2 ] compare statistical parameters which provide similar information , such as robust versus non - robust or normalized versus non - normalized estimators , as a function of the weighting scheme .
error - phase weighted ( and noise - unbiased , when applicable ) estimators correspond to the most strongly peaked distributions around the correct values .
\(a ) + + ( c ) + \(b ) + + ( d ) +
figure [ fig : scatterunweighted ] presents a selection of unweighted estimators for stars of different variability types .
this figure is intended to be compared to fig .
[ fig : scatter ] , which illustrates the same information for phase - error weighted and noise - unbiased estimators . | unevenly spaced time series are common in astronomy because of the day - night cycle , weather conditions , dependence on the source position in the sky , allocated telescope time , corrupt measurements , for example , or be inherent to the scanning law of satellites like _ hipparcos _ and the forthcoming _ gaia_. weighting scheme , adapting to the sampling density and noise level , time series from the _ hipparcos _ periodic catalogue led to significant improvements in the overall accuracy and precision of the estimators with respect to the unweighted counterparts and those weighted by inverse - squared uncertainties .
automated classification procedures employing statistical parameters weighted by the suggested scheme confirmed the benefits of the improved input attributes .
the classification of eclipsing binaries , mira , rr lyrae , delta cephei and alpha@xmath0 canum venaticorum stars employing exclusively weighted descriptive statistics achieved an overall accuracy of 92 per cent , about 6 per cent higher than with unweighted estimators .
[ firstpage ] methods : data analysis methods : statistical stars : variables : general . |
in recent years the field of ultracold atomic physics has attracted the attention of a great many condensed matter theorists , largely due to the prospect of finding novel realizations of many - body systems .
part of the appeal doubtless lies in the exquisite experimental control that may be exercised over the parameters of a system in which many of the complicating factors familiar from the solid state ( disorder , phonons , etc . )
are absent .
intrinsic _ aspects of ultracold atomic gases not dependent on the specifics of the experimental setting are qualitatively novel , however , and without antecedent in the study of condensed matter . in this latter category
we may place the possibility of spontaneous ordering of the spin degrees of freedom in a bose gas .
indeed , prior to the ` ultracold revolution ' the only bose superfluid that could be studied in the laboratory was @xmath0he , which has zero spin . with the advent of optical trapping of bose condensates of alkali atoms , which allows for a fully rotationally invariant setting , the experimental study of spin ordering within a hyperfine multiplet came within reach @xcite .
the earliest theoretical works motivated by these developments explored possible ordered phases using a mean field description , in both the spin 1 and spin 2 hyperfine multiplets @xcite . in this description the concept of spontaneous symmetry breaking plays a central role .
up until very recently , however , there were no experiments in which this spontaneous ordering was apparent .
the reason for this is that the simplest experimental protocol for the investigation of the hyperfine state of the gas is to apply a magnetic field gradient to split a gas cloud into different components in a stern - gerlach experiment @xcite .
the different components are subsequently imaged to determine their ( relative ) occupancies .
this technique naturally imposes a quantization axis , and any information concerning the _ coherence _ between different hyperfine levels is lost . thus magnetic alignment in the plane perpendicular to this axis , which depends on the relative phase of these different levels , can not be observed .
the characterization of magnetic ordering in atomic gases has taken a leap forward in the last few years thanks to the work of the berkeley group , who demonstrated _ in situ _ dispersive imaging of the transverse magnetization of a gas of @xmath1rb in the spin 1 multiplet @xcite , and subsequently employed this technique to investigate a number of fascinating aspects of this system , including the dynamics of spontanteous symmetry breaking , defect production , and the role played by magnetic dipole forces @xcite .
the above developments illustrate two important needs .
firstly , imaging of the spin order was necessary to bring much of this new physics out into the open .
secondly , the nonequilibrium character of most experiments requires that the mean field theory of equilibrium ordered states be supplemented with a _
dynamical _ description of the relevant order parameters .
it is our hope that the next few years will see the development of imaging techniques capable of detecting some of the spin orders to be discussed in section [ sec : spin ] , of which an average magnetization is only the simplest .
the aim of this work is to address the second need : uncovering the low energy description of the order parameter . for the case of the bose ferromagnet , which is appropriate to the spin 1 @xmath1rb system ,
this description was provided in an earlier paper @xcite . in this work
we will focus instead on states with vanishing average magnetization . for reasons that will be become clear in the following sections these cases are qualitatively different . in an ordered phase of matter we expect that the low energy degrees of freedom consist of variations of the order parameter on some manifold of symmetry broken states ( goldstone modes ) , together with any conserved quantities . in our earlier work on the bose ferromagnet @xcite the degrees of freedom were the local magnetic moment @xmath2 and the superfluid velocity @xmath3 . in the long wavelength limit
these were found to obey the coupled equations @xmath4 @xmath5 where @xmath6 is the convective derivative .
( [ lle ] ) is a modified landau - lifshitz equation and gives rise to quadratically dispersing spinwaves when linearized on around a solution @xmath7 .
the quadratic dispersion is a consequence of the two transverse deviations of the order parameter being canonically conjugate . for the phases with vanishing average magnetization that are the focus of this work ,
the conjugate variables involve deviations from the order parameter manifold .
this results in linearly dispersing goldstone modes .
the situation is analogous to the case of spin waves in an antiferromagnet , where the conjugate variables are the difference in magnetization on neighboring sites the nel order parameter and the sum .
it also follows from the vanishing of the average magnetization that the order parameter dynamics and superfluid flow are decoupled except for a global topological constraint . again , this is quite different from the case of the ferromagnet , where the two are coupled together in the equations of motion .
it follows that we can write a lagrangian for the spin degrees of freedom only .
this lagrangian is expressed in terms of a rotation matrix @xmath8 that specifies the local orientation of the spin state relative to some reference state . by expressing the matrix elements of @xmath8 in terms of an orthonormal triad @xmath9 , with @xmath10
, the spin lagrangian may be written @xmath11\ ] ] here @xmath12 are @xmath13 are some constants to be specified later , which depend on the ordered phase in question .
the lagrangian eq .
( [ l_genspin ] ) is the main result of this paper .
the only other phase for which the spin lagrangian was previously obtained is the polar phase of the spin 1 gas , to be discussed below , in which @xmath14 , @xmath15 , giving the familiar @xmath16 model @xcite .
the structure of the remainder of this paper is as follows . in section [ sec : basics ]
we review the basic description of a spinor condensate , and in particular the structure of the interaction hamiltonian .
section [ sec : spin ] describes in detail the ground state manifolds of the different phases of spinor condensates , beginning with the simplest non - trivial case , spin 1 , before introducing the majorana ( or stellar ) representation that is very useful in visualizing spin states .
the global structure and local geometry of the order parameter manifolds are then introduced , as well as a parametrization for the dynamical variables conjugate to the order parameter .
after this the derivation of the low energy lagrangian is a relatively simple matter , and is described in section [ sec : low ] along with the derivation of the equations of motion .
we have tried to keep the presentation pedagogical throughout , though at various points there are parenthetical technical comments that the casual reader should feel free to ignore . in a related work ,
barnett _ et al . _
@xcite have derived the full equations of motion of a spinor condensate in terms of the majorana representation and applied a group - theoretical analysis to the determination of all normal modes about an ordered ground state .
we pursue a complimentary goal of obtaining the full _ nonlinear _
lagrangian for the goldstone modes only .
one could take the point of view that the description of the dynamics of a dilute spinor gas is no different from its spinless counterpart , being governed by the time - dependent gross - pitaevskii equation , with lagrangian density @xmath17 ( we set @xmath18 from now on ) . for a spin @xmath19 gas @xmath20
is a @xmath21 component spinor and the hamiltonian density has the form @xmath22 where the first term is the kinetic energy .
the interaction part @xmath23 is quartic in @xmath20 and its form will be given below for @xmath24 and @xmath25 .
we will not discuss the influence of the trapping potential , save to assume that it preserves rotational symmetry .
for stationary solutions of the form @xmath26 the time - dependent description reduces to the time - independent gross - pitaevskii theory , with @xmath27 the chemical potential . instead of treating the action @xmath28 classically
, we can interpret it as the quantum action in a coherent state path integral .
little that we will have to say will depend upon this distinction .
in fact this superficial similarity between the spinless and spinful problems is quite misleading .
the ground state in a uniform system corresponds to some constant @xmath20 . in the spinless case
this fixes @xmath20 up to a phase once the density @xmath29 is specified .
but in the spinful case we still have to find the correct ` direction ' of @xmath20 in the complex @xmath21 dimensional spinor space , determined by minimizing the interaction hamiltonian @xmath23 .
the interactions will be assumed to respect rotational symmetry , so this minimum is only unique up to rotation ( specified by the three euler angles , say ) . choosing this rotation starting from some arbitrary reference state specifies the spontaneous breaking of rotational symmetry in the ground state .
the slow variation of this rotation in space and time constitutes the low energy dynamics of the system , the description of which is the subject of this work . to characterize these low energy manifolds
we must first specify the form of @xmath30 .
the structure of @xmath30 has been discussed in many works , starting with the first papers treating spinor bose condensates .
we will therefore keep the following discussion relatively brief .
low energy scattering between a pair of bosons occurs in the @xmath19-wave channel only , and can be treated as a @xmath31-function interaction , characterized by a set of interaction constants @xmath32 , @xmath33 for a pair of bosons with total spin @xmath34 .
bose symmetry dictates that the interaction vanishes for odd total spin . for spin 1
, the resulting interaction may be presented in the form @xcite @xmath35 where @xmath36 are the spin @xmath19 angular momentum matrices .
( [ spin1hint ] ) is the sum of a density - density and a spin - spin interaction . for spin 2
we have @xcite @xmath37 here @xmath38 is a scalar representing the amplitude of singlet pairs of spin s ( eq . ( [ spin1hint ] ) can be expressed using this operation instead of the spin - spin interaction . in the spin 2 case
both terms are needed ) .
our use of @xmath39 for two different quantities matches the notation of the works cited above , where one may find explicit expressions for @xmath40 @xmath41 in terms the constants @xmath32 defined above . in the spin 1 case
it is fairly clear how to minimize eq .
( [ spin1hint ] ) at fixed density . for the spin 2 case things
are less obvious . in the next section
we will discuss a method of parametrizing the spinor @xmath42 to make this operation transparent .
in this section we will introduce the majorana ( or _ stellar _ ) representation of spin states .
this provides a vivid way to picture spin ordering in higher spin condensates in which rotational symmetry is manifest .
none of the calculations of section [ sec : low ] depend upon this representation ; its use is rather in providing a concrete way to picture the ground state manifold . before beginning it is worth setting the scene with a more pedestrian discussion of the spin 1 case @xcite .
let us minimize eq .
( [ spin1hint ] ) with a spinor @xmath42 normalized to unity ( thus we are adopting units in which the density @xmath43 ) .
this is a matter of minimizing ( maximizing ) @xmath44 for @xmath45 ( @xmath46 ) .
one way to make the resulting states more clear is to write the complex spinor @xmath47 where @xmath48 and @xmath49 are two real vectors satisfying @xmath50 , and to work in cartesian components .
the relationship to the usual components @xmath51 @xmath52 is @xmath53 in this basis the angular momentum matrices take the form @xmath54.then we have @xmath55 for @xmath46 the interaction energy is minimized for @xmath48 and @xmath49 perpendicular and equal in magnitude . this state is termed _
ferromagnetic _ as it corresponds to maximal polarization of the spin .
the resulting order parameter manifold corresponds to the set of all configuration of a pair of orthogonal vectors , and is thus identified with the group of rotations @xmath56 for @xmath45 @xmath48 and @xmath49 are aligned
. the resulting _ polar _ state ( the name originates from an analogous state in superfluid @xmath57he ) can therefore be written as @xmath58 for @xmath59 a _ real _ unit vector .
note that this parametrization has some redundancy in that the points @xmath60 and @xmath61 are identified .
the resulting manifold is known as the _ mapping torus _ of the antipodal map of the sphere @xmath62 .
the global topology of the above ferromagnetic and polar order parameter manifolds naturally determines the character of the topological defects in the ordered phases , and certain features of the ordering transitions , some of which have already been discussed in the literature @xcite .
this is not the focus of the present work and topology will not be further discussed , even though the defect physics of the higher spin condensates promises to be highly non - trivial @xcite .
the polar state has @xmath63 .
nevertheless the above discussion makes it clear that polar ordering involves a choice of axis : the spinor @xmath64 is the @xmath65 state with respect to the axis @xmath59 .
it is natural to ask for an operator that acquires a non - zero expectation value in the polar state .
the obvious candidate is the spin 2 quadrupole tensor ( or _ nematicity _ ) @xmath66 with @xmath67 such expressions are familiar in the study of nematic liquid crystals , where the vector @xmath59 is known as the director . in the liquid crystal context the identification of @xmath59 and @xmath68 without the phase factor in eq .
( [ polar_parm ] ) ( the order parameter manifold is then the real projective plane @xmath69 ) makes for very different defect physics , however @xcite .
nematic ordering in solid state magnetic systems has been the subject of much experimental and theoretical work in recent years , with a good deal of uncertainty still remaining .
the observation of the polar state in the spin 1 bose gas would therefore be an important milestone .
searching for higher spin order parameters as the spin of the gas particles increases becomes arduous .
we now turn to a more convenient representation of the spin order .
the representation of a general spin @xmath19 state that ( sometimes ) bears his name was discovered by majorana in 1932 @xcite , and independently several times since @xcite , though it has antecedents in 19@xmath70 century mathematics @xcite .
a very nice discussion can be found in ref . .
the result is very simple to state , and represents a generalization of the bloch sphere for spin 1/2 to arbitrary spin .
up to normalization and a phase thus in more mathematical terms we are parametrizing the complex projective space @xmath71 an arbitrary spin @xmath19 state can be specified by locating @xmath72 indistinguishable points on the unit sphere ( see fig . [
fig : principal ] ) .
such a configuration is sometimes called a _ constellation _ , for reasons that will become clear .
spin state , with arrows representing the principal spinors .
[ fig : principal],title="fig:",scaledwidth=40.0% ] + there are two steps to understand why this is so .
first , imagine forming our spin @xmath19 from @xmath72 spin 1/2 in the totally symmetric subspace .
an arbitrary state may then be written as a totally symmetric spinor @xmath73 , where the round brackets denote the operation of symmetrization and each of the @xmath21 indices can take the value @xmath74 or @xmath75 .
the relationship between @xmath76 and the corresponding @xmath21 component spinor @xmath42 is @xmath77 next contract every index of @xmath76 with @xmath78 .
denoting by @xmath79 the resulting polynomial of order @xmath72 , the fundamental theorem of algebra tells us @xmath80 with @xmath81 some normalization .
thus @xmath76 may be written @xmath82 with the _ principal spinors _
@xmath83 , @xmath84 , etc . related to the @xmath85 by @xmath86 relations that are unchanged
if we normalize the spinors , in which case they correspond to @xmath72 points on the bloch sphere with coordinates @xmath87 , @xmath88 @xmath89 ( notice that @xmath76 in eq .
( [ root_spinors ] ) is not in general normalized when the principal spinors are ) .
then ( minus ) the roots of @xmath79 can be written @xmath90 and correspond to stereographic projection from the north pole to the plane tangent to the sphere at the south pole .
the beautiful feature of the majorana representation is that rotations act simply as rotations on the bloch sphere . of course , it is useful to have an explicit expression for the polynomial @xmath79 in terms of the @xmath21 components of the spin @xmath19 state @xmath42 .
it is easy to show @xmath91 if spinor indices are raised and lowered using the antisymmetric tensors @xmath92 and @xmath93 ( with @xmath94 and @xmath95 ) @xmath96 then rotational invariance upon contraction of indices is guaranteed ( in fact the result is invariant under the larger group @xmath97 , a result that will be useful later ) . if we denote by @xmath98 the result on @xmath42 of raising all indices of the corresponding symmetric spinor , then one can readily see that @xmath99 , and thus @xmath100 . after raising indices of the principal spinors we have @xmath101 , etc
.. under stereographic projection @xmath102 represents the antipodal map on the unit sphere .
thus we see that the spinor @xmath103 is represented by a set of points antipodal to those representing @xmath42 .
furthermore , the majorana representation of a normalized state with @xmath104 , corresponding to @xmath105 , consists of pairs of antipodal points ( and is thus only possible for integer spin ) .
this fact will be useful in minimizing the interaction energy ( recall the form of eq .
( [ spin2hint ] ) ) .
the transformation @xmath106 is in fact the ( anti - unitary ) operation of time reversal .
the use of the majorana representation to visualize spin ordering in a bose gas was suggested in ref . .
let us first see how the phases of the spin 1 gas discussed in section [ sec : spin1phases ] appear in this representation , before moving on to the spin 2 case . as mentioned in section [ sec : hint ] ,
( [ spin1hint ] ) , the interaction hamiltonian in the spin 1 case may be written @xmath107 for @xmath45 we should maximize @xmath108 .
based on the discussion of the previous section , this corresponds to placing the two points antipodally in the majorana representation .
it is evident that the corresponding spin 1 spinor is just the symmetric @xmath65 state with respect to the resulting axis .
this is just the polar state described before . for @xmath46 @xmath108
can be set to zero by making the two principal spinors equal since @xmath109 .
it is clear that this represents the ferromagnet , being a maximally polarized ( or _ coherent _ ) spin state , a result that generalizes to arbitrary @xmath19 .
suppose that in eq .
( [ spin2hint ] ) @xmath110 .
the corresponding term in the hamiltonian can be fully satisfied by states with @xmath111 . by inspecting the spin 2 angular momentum matrices @xmath112 it is easy to convince oneself that up to rotation the most general state satisfying these conditions is @xmath113 the principal spinors form a polyhedron with four identical triangular faces known as a _ disphenoid _ ( see fig .
[ fig : disphenoid ] ) .
, @xmath114 in eq .
( [ disphenoid ] ) .
bottom left : rectangular state with @xmath115 , @xmath114 .
bottom right : tetrahedron with @xmath116 .
[ fig : disphenoid],title="fig:",scaledwidth=30.0% ] + , @xmath114 in eq .
( [ disphenoid ] ) .
bottom left : rectangular state with @xmath115 , @xmath114 .
bottom right : tetrahedron with @xmath116 .
[ fig : disphenoid],title="fig:",scaledwidth=20.0% ] , @xmath114 in eq .
( [ disphenoid ] ) .
bottom left : rectangular state with @xmath115 , @xmath114 .
bottom right : tetrahedron with @xmath116 .
[ fig : disphenoid],title="fig:",scaledwidth=20.0% ] to fix the parameters @xmath117 , @xmath118 in eq .
( [ disphenoid ] ) we turn to the third term in eq .
( [ spin2hint ] ) . if @xmath46 , this term is minimized by placing four points in two antipodal pairs so that @xmath119 .
the principal spinors form a rectangle , corresponding to @xmath115 in eq .
( [ disphenoid ] ) .
the aspect ratio of the rectangle varies with @xmath117 , with @xmath120 being a square and @xmath121 corresponding to a pair of points at either pole .
the fact that the energy is minimized for _ any _ @xmath117 in this parameter regime is rather surprising , and we will return to it briefly below .
this state will be referred to as _ rectangular _ in the following ( in ref .
it was called nematic since @xmath122 . for @xmath45 @xmath123
should be minimized .
this can be done by taking @xmath116 .
the result is a regular tetrahedron .
note that this state has @xmath124 . for @xmath125
one must compare the energy of the rectangular state with that of the ferromagnet .
the resulting phase diagram can be found in refs . .
note that while the rectangular state maximizes the magnitude of the quadratic scalar @xmath126 the tetrahedral state maximizes the cubic invariant @xmath127 the inclusion of a sextic term proportional to @xmath128 in the interaction energy would lift the accidental degeneracy in the @xmath117 parameter discussed above .
a microscopic derivation of such a term is discussed in refs . , with a positive sign favoring the square state ( @xmath120 ) and a negative sign the uniaxial state in which two pairs of points coincide ( @xmath121 ) .
the inclusion of such a term in our formalism is a straightforward matter , and we will not discuss it further .
the rectangular and tetrahedral states evidently have certain discrete symmetries that are rather hard to discern by inspection of eq .
( [ disphenoid ] ) , and indeed went unnoticed in the earliest works on the spin 2 condensate @xcite .
it appears that the term ` cyclic ' used in several works to describe the tetrahedral phase is a consequence of a misidentification of the symmetry .
this illustrates the utility of the majorana representation in the visualization of spin order .
in section [ sec : spin1phases ] we identified the order parameter manifolds of the phases of the spin 1 gas . with the help of the majorana representation
we can now do the same for the spin 2 case .
roughly speaking , we expect the manifold to consist of all configurations related to those of section [ sec : spin2phases ] by rotation .
some rotations will leave the configuration of principal spinors unchanged , however , so the manifold can not simply be identified with the rotation group @xmath56 . in mathematical terms
the problem is to determine the _ orbits _ of a reference spinor under the action of the spin @xmath19 representation of the rotation group .
if we ignore the phase of the spinor for a moment , so that we are considering orbits in @xmath71 , this problem can be solved using the majorana representation by considering the manifold of constellations generated by all possible rotations @xcite .
for instance , the ferromagnetic spin 1 state has orbit in @xmath129 equal to @xmath62 ( and this is true in general for any spin @xmath19 coherent state ) , while for the polar state we have @xmath62 with antipodal points identified : the real projective plane @xmath69 . in general
if one finds a configuration of principal spinors unchanged under some subgroup of @xmath130 ( the _ stabilizer _ subgroup of a constellation ) , then the orbit is given by @xmath131 .
thus in the case of spin 1 , @xmath132 for the ferromagnet and @xmath133 in the polar phase , since in the latter case a parity transformation also leaves the points unchanged .
the reader will notice that these are _ not _ the order parameter manifolds identified in section [ sec : spin1phases ] for the spin 1 case .
we have neglected the phase of the spinor , which is a real degree of freedom .
we might then guess that any spinor on the order parameter manifold can be written @xmath134 where @xmath135 is the spin @xmath19 representation of the rotation @xmath8 and @xmath136 is some reference spinor corresponding to the phase in question .
this does not mean that the order parameter manifold is @xmath137 , because the stabilizer subgroups mentioned above leave the spinor corresponding to a particular constellation unchanged up to a phase .
we denote these phases as @xmath138 , with @xmath139 .
they must form a one - dimensional unitary representation of @xmath140 @xmath141 and allow us to make the identification @xmath142 showing that the order parameter manifold is @xmath143 , where the tilde is to denote the action of @xmath140 on @xmath137 : @xmath144 the simplest case to consider is the polar phase with @xmath133 , for which the only non - trivial phase is a @xmath145 associated with the parity transformation .
if @xmath140 were @xmath146 with only trivial phases , the ground state manifold would be @xmath147 , with the first factor coming from the @xmath148 and the second from @xmath149 .
the parity transformation and the associated minus sign are responsible for the identification @xmath150 already discussed in section [ sec : spin1phases ] . to turn to a less trivial example , let us see how this works for the case of the tetrahedral phase .
in this case @xmath151 , the symmetry group of the tetrahedron . a tetrahedron has 3 orthogonal 2-fold axes , and 4 3-fold axes .
for the representative spinor given earlier @xmath152 the @xmath153 axis is aligned with one of the 2-fold axes .
one easily verifies that @xmath153 axis rotations through @xmath154 leave the spinor unchanged .
alternatively , we can align one of the 3-fold axes with the @xmath153 axis with the choice ( see fig . [ fig:3axis ] ) @xmath155 ) .
[ fig:3axis],scaledwidth=40.0% ] ( this is most easily seen by considering the majorana polynomial , which has a root at @xmath156 ) .
now a rotation through @xmath157 is seen to reproduce the same spinor but with phase factors @xmath158 .
it is not hard to verify that these phases form a one - dimensional representation of @xmath159 ( the other non - trivial one - dimensional representation comes from changing the sense of the 3-fold axes ) .
the topological properties of the resulting space @xmath160 , and the implications for superfluid vortices in the tetrahedral phase , were discussed in ref . . for the rectangular phase @xmath161 in general , but @xmath162 for the square case ( @xmath120 ) .
here @xmath163 denotes the dihedral group .
the only non - trivial phases occur in the latter case , as may be seen by considering the value @xmath164 , when the four points lie on the equator of the bloch sphere .
then we have @xmath165 and a @xmath166 rotation about the @xmath153 axis is seen to give rise to a @xmath145 .
this construction generalizes readily to other ordered states of arbitrary spin once the corresponding stabilizer subgroups and phases are identified @xcite .
note that when @xmath140 is discrete , as for the spin 2 phases other than the ferromagnet , the order parameter manifold is 4 dimensional .
next we turn to the local properties of the order parameter manifold .
the inner product naturally endows this space with a metric @xcite .
using the parametrization eq .
( [ gen_param ] ) consider two states @xmath167 and @xmath168 related @xmath169 and @xmath170 with @xmath171 an infinitesimal rotation corresponding to @xmath172 we find the squared distance between these two states to be @xmath173 is @xmath174 eq . ( [ distance ] ) makes it clear that if @xmath175 , rotations and phase changes are coupled together .
there is an arbitrariness in the way the phase is apportioned between the two factors in eq .
( [ gen_param ] ) that amounts to a choice of gauge .
this gauge structure was discussed in ref . for the case of the ferromagnet , the only one of the phases discussed in section [ sec : spin ] with @xmath175 .
the discussion of the general case is relegated to appendix [ sec : gauge ] .
our main interest is in the other phases having @xmath176 , for which eq .
( [ distance ] ) decouples into separate contributions from the change in phase and the rotation , with the latter being characterized by the metric tensor @xmath177 .
note that @xmath177 is simply related to the nematicity @xmath178 in eq .
( [ nematicity ] ) .
the notion of distance described by the metric tensor is left invariant i.e. preserved if states @xmath167 are mapped by @xmath179 for some @xmath180 , but not right invariant , under which @xmath181 .
focusing now on the spin 2 case , we evaluate the metric for the state eq .
( [ disphenoid ] ) @xmath182 for the tetrahedral phase ( @xmath116 ) @xmath183 , showing that the order parameter manifold has a left and right invariant geometry .
this is perhaps not surprising given the highly symmetric arrangement of points in the majorana representation . for the rectangular phase ( @xmath115 )
, we see that in the case @xmath121 , @xmath184 .
the physical meaning is clear : because we have two points at either pole the stabilizer subgroup is @xmath185 , so that the order parameter manifold is only three dimensional .
the same holds true for the spin 1 polar phase .
we will see that the metric plays a crucial role in fixing the dynamics on the order parameter manifold .
having characterized the order parameter manifold for the spinor condensates , we are almost ready to study the dynamics on that manifold .
it remains to identify the _
conjugate variables_. we expect these to be coupled by the first term of eq .
( [ gp_action ] ) , which expresses the conjugacy of @xmath42 and @xmath186 . substituting the parametrization eq .
( [ gen_param ] ) into that term gives @xmath187 where @xmath188 are the components of the angular velocity .
by analogy with rigid body dynamics , we refer to this as the ` body frame ' angular velocity : eq .
( [ conj_param ] ) shows that it is conjugate to the ` body frame ' angular momentum @xmath189 , i.e. that of the unrotated state . as is well known , the time integral of the term @xmath190 has an alternative interpretation as the berry phase associated with the time evolution of @xmath20 . in this context
the formula eq .
( [ conj_param ] ) appears in ref . .
thus in the cases of interest where @xmath191 the variables conjugate to the rotations are nonzero only as one deviates from the order parameter manifold . to account for these deviations
we generalize the parametrization eq .
( [ gen_param ] ) to @xmath192 where @xmath193 is a state with @xmath194 .
note that @xmath195 @xmath196 is angular momentum in the ` lab frame ' . to find the deviation corresponding to @xmath193 ,
let us consider the state @xmath197 where @xmath198 is some normalization factor , and @xmath199 it is not hard to see that to quadratic order the normalization takes the form @xmath200 and that to this order @xmath194 , as required .
the effect of the @xmath201-distortion on the tetrahedral state is shown in fig .
[ fig : distort ] . due to boosts in the @xmath153-direction [ fig : distort],title="fig:",scaledwidth=23.0% ] due to boosts in the @xmath153-direction [ fig : distort],title="fig:",scaledwidth=23.0% ] using the parametrization eq .
( [ gen_param_conj ] ) in eq .
( [ conj_param ] ) gives then @xmath202 to first order in @xmath201 .
we make the following aside .
the matrix @xmath203 that acts on @xmath136 is a polar decomposition of the @xmath204 representation of an element of @xmath97 @xcite .
now @xmath205 is isomorphic to the connected lorentz group .
this isomorphism has a beautiful physical counterpart .
the lorentz transformations have a natural action on the celestial sphere , the space of light rays on the past ( say ) light cone .
further , the elements of @xmath97 @xmath206 have a natural action on the bloch sphere : @xmath207 , corresponding to a mbius transformation on the stereographic coordinates @xmath208 @xmath209 ( the fact that @xmath210 and @xmath211 correspond to the same mbius transformation accounts for the @xmath212 quotient above ) . remarkably , group elements related by the isomorphism mentioned above correspond to identical transformations of the sphere @xcite ! the deformations of eq .
( [ l_deform ] ) are just those generated on the celestial sphere by a boost of the frame of reference , which in particular determines the aberration of the fixed stars . thus the name ` stellar representation ' sometimes used to describe the majorana picture of spin states is more than picturesque . with the parametrization eq .
( [ gen_param_conj ] ) we have accounted for six spin degrees of freedom ( plus the superfluid phase ) : three rotation variables parametrizing the order parameter manifold and three conjugate variables .
the remaining @xmath213 variables required to specify the spin state do not correspond to any broken symmetries and must describe gapped modes . for the spin 2 case ,
the missing two degrees of freedom can be readily found by considering the ratio of the two ( complex ) @xmath97 invariants in eqns .
( [ quad_inv],[cubic_inv ] ) .
@xmath214 the powers are chosen so that the normalization in eq .
( [ l_deform ] ) drops out . by construction @xmath27
is unchanged for all @xmath201 and @xmath8 starting from the state eq .
( [ disphenoid ] ) , and could be expressed in terms of @xmath117 and @xmath118 .
this concludes our discussion of the spin degrees of freedom in a spinor condensate . with this background
, we will see that the derivation of the low energy lagrangian is extremely straightforward .
we are going to use the parametrization eq .
( [ gen_param_conj ] ) in the lagrangian eq .
( [ gp_action ] ) . in doing so
, we are treating the problem as as _ constrained _ dynamical system , in which deviations from the ground state manifold associated with the gapped modes described in the previous section are assumed to be infinitely stiff .
the lagrangian for the spin degrees of freedom then takes the form @xmath215 where @xmath216 @xmath217 , in analogy to the earlier definition of @xmath218 .
we could additionally allow for a variation @xmath219 in the density of the gas , which is conjugate to the phase @xmath220 , to describe the density modes , but will not do so here . in eq .
( [ l_exp ] ) we have not included the gradient of the conjugate variables ( including @xmath219 ) in the part arising from the kinetic energy , as such terms can be neglected in the long wavelength limit of interest . the conjugate variables do however appear in the interaction term @xmath221 the notation is of course chosen to emphasize the rigid body analogy .
the precise form of the ` inertia tensor ' will depend on the phase under consideration ; we discuss the spin 2 phases for definiteness . in that case
the interaction hamiltonian has the form eq .
( [ spin2hint ] ) . computing the quadratic variation of this expression with the conjugate variables
is facilitated by the @xmath97 invariance of the third term : its variation is determined solely by the normalization in eq .
( [ norm ] ) .
we obtain @xmath222 it is then straightforward to eliminate the @xmath201 degrees of freedom using the equation of motion , obtained from eq .
( [ l_exp ] ) @xmath223 to obtain the final result @xmath224 eq .
( [ l_exp_l_elim ] ) represents the main conclusion of this work , being the low energy lagrangian for the spin degrees of freedom of the condensate .
it takes the form of a sigma model in @xmath225 dimensions with target space @xmath56 ( if we ignore the possibility of vortices in the superfluid phase @xmath220 the subtle global structure of the ground state manifold discussed in section [ sec : global ] can be ignored ) . in the case of the tetrahedral phase , the metric tensor @xmath226 , and the lagrangian eq .
( [ l_exp_l_elim ] ) becomes that of the _ principal chiral model _ , having independent left and right @xmath56 symmetries .
an alternative form for eq .
( [ l_exp_l_elim ] ) follows from noting that , if @xmath227 @xmath228\ ] ] with @xmath229 thus we have @xmath230\ ] ] by expressing the matrix elements of @xmath8 in terms of an orthonormal triad @xmath9 , with @xmath10 this may be written @xmath231\ ] ] recall that for the spin 1 polar phase and for the special value @xmath121 in the spin 2 rectangular phase the metric tensor has one zero eigenvalue and two equal non - zero eigenvalues . as a result
both @xmath232 and @xmath233 have _ two _ vanishing eigenvalues , and eq . ( [ l_alt2 ] )
reduces to the usual @xmath234 sigma model .
we find the equations of motion corresponding to eq .
( [ l_exp_l_elim ] ) by writing the variation @xmath235 this gives @xmath236 substitution into eq .
( [ l_exp_l_elim ] ) leads to the equations of motion . @xmath237
it follows from their definition that @xmath238 satisfy the ( maurer - cartan ) equation @xmath239 note that if @xmath238 is interpreted as a non - abelian gauge field , the above condition corresponds to vanishing field strength , and to the absence of topological defects .
the equations of motion can be linearized by ignoring the right hand side of eq .
( [ mc_eqn ] ) , allowing us to write @xmath240 .
the linear equations of motion following from eq .
( [ eom ] ) are then @xmath241 a wave equation describing the propagation of three spin wave modes with velocities @xmath242 for the spin 2 case eq . (
[ spin2_inertia ] ) gives @xmath243 with @xmath244 given by the diagonal elements of eq .
( [ disphenoid_metric ] ) with @xmath115 .
for the square case ( @xmath245 ) we have @xmath246 and @xmath247 these results check with ref . , which also includes the normal phonon mode as well as the mode associated with variations of the @xmath117 and @xmath118 parameters in eq .
( [ disphenoid ] ) . in the case of the rectangular phase this latter mode appears gapless in mean field theory , but as explained in section [ sec : spin2phases ] this is the result of an accidental degeneracy that does not persist in the next order of approximation .
we have achieved our goal of providing a framework in which the parameters entering the low energy spin lagrangian of an arbitrary ordered state of a spinor condensate ( with @xmath191 ) may be easily calculated .
though we focused on the spin 2 states , any other state can be treated by the same method once the problem of minimizing the mean field energy is solved .
the extension of the present formalism to spin ordered mott insulating phases in which the phase variables are quantum disordered does not present any particular difficulties .
perhaps the most interesting problem that we have not addressed in detail relates to the character of topological defects in these systems .
the occurrence of nonabelian stabilizer subgroups means that vortices have very novel characteristics @xcite .
we mention one consequence of our work for the _ quantum _ description of such vortices .
the phase factors associated with elements of the stabilizer subgroups that were discussed in section [ sec : global ] will appear in the path integral when vortices are present , as may be seen from eq .
( [ conj_param ] ) .
consider an imaginary time path integral with fields obeying the boundary condition @xmath248 if , as we go from @xmath249 , the field at a point @xmath250 is subject to a rotation that evolves from @xmath251 , for @xmath252 , the @xmath220 variable must increase @xmath253 in order to ensure periodicity of the fields , leading to a phase factor @xmath254 in the path integral ( @xmath255 is the density ) .
ref . discusses the effect of these phases for the simplest case of the spin 1 polar phase ( or rather the mott insulating phase based upon it ) , where they are @xmath256 and the defects are abelian .
the nonabelian case remains unexplored .
the support of the nsf under grant dmr-0846788 is gratefully acknowledged , as well as several useful conversations with ryan barnett and gil refael .
in the general case the distance between states on the ground state manifold has the form eq .
( [ distance ] ) .
as mentioned in the text , there is an arbitrariness in the apportioning of phase between the two factors of eq .
( [ gen_param ] ) that belies a natural gauge structure . in other words , eq .
( [ gen_param ] ) is unchanged under @xmath258 for some arbitrary function @xmath259 .
let us define a ( berry ) vector potential @xmath260 where @xmath261 denotes an expectation in the state @xmath136 , @xmath262 , and @xmath263 denotes the exterior derivative ( we find it convenient to use the language of differential forms ) .
the vector potential allows us to define a covariant derivative @xmath264 .
the metric tensor @xmath265 can then be cast in the form the two terms in eq .
( [ gen_metric ] ) are manifestly gauge invariant under the transformation eq .
( [ gauge_trans ] )
. the gauge invariant metric tensor @xmath267 that appears in the first term takes the form @xcite @xmath268 in the second term , @xmath269 by itself is gauge dependent , but the associated field strength is not @xmath270 where in the second step we have used the maurer - cartan equation eq .
( [ mc_eqn ] ) .
( [ field_strength ] ) has a more familiar form , as may be seen by introducing a unit vector @xmath271 parallel to @xmath272 .
then we have @xmath273 where @xmath274 , and in the penultimate line we have used the fact that the determinant of the rotation matrices is unity . as a result @xmath275 which generalizes the mermin - ho relation in eq .
( [ v_fix ] ) to an arbitrary spin state .
an important example is provided by the ferromagnet , for which the state @xmath276 is a fully spin polarized ( coherent ) state , and we have @xmath277 then the first term of eq .
( [ gen_metric ] ) takes the form @xmath278 the resulting metric sets the form of the hamiltonian in ref . . | we present a derivation of the low energy lagrangian governing the dynamics of the spin degrees of freedom in a spinor bose condensate , for any phase in which the average magnetization vanishes .
this includes all phases found within mean - field treatments except for the ferromagnet , for which the low energy dynamics has been discussed previously .
the lagrangian takes the form of a sigma model for the rotation matrix describing the local orientation of the spin state of the gas .
= 1 |
# 1 ps . @headings oddfootevenfoot oddhead evenheadoddhead # # 1 ps .
@headings = 12 # 1@xmath4
= 11 # 1 ps . @headings oddfootevenfoot oddhead evenheadoddhead # # 1 ps .
@headings # 1#2#3_phys .
lett . _ * # 1*(19#2)#3 # 1#2#3_z .
phys . _ * # 1*(19#2)#3 # 1#2#3_phys .
lett . _ * # 1*(19#2)#3 # 1#2#3_rev .
phys . _ * # 1*(19#2)#3 # 1#2#3_phys .
rep . _ * # 1*(19#2)#3 # 1#2#3_phys .
rev . _ * # 1*(19#2)#3 # 1#2#3_nucl .
phys . _ * # 1*(19#2)#3 # 1#2#3_sov . j. nucl
. phys . _ * # 1*(19#2)#3 # 1#2#3_acta phys .
polon . _ * # 1*(19#2)#3 @ 11 # 1@fileswauxout @ 12 # 1#2#3#4#5#6 = # 1#2#1 # 2 0b_0 .5 cm *
stefano catani * + .1 cm infn , sezione di firenze and univ . di firenze , florence , italy + .5 cm * michelangelo l. mangano , * * paolo nason , * + 0.1 cm cern , th division , geneva , switzerland + .5 cm * luca trentadue@xmath5 * + .1 cm univ . di
parma and infn , gruppo collegato di parma , parma , italy cern - th/96 - 21 +
we present in this letter a theoretical reassessment of the evaluation of the top quark production cross section in high - energy hadronic collisions .
the top quark having been found @xcite , the comparison between its observed production properties and those expected from the standard model will be an important probe of the possible existence of new phenomena .
one of the most important tests to be performed concerns the total production cross section .
this is the most inclusive quantity available , and is a priori the least sensitive to a detailed understanding of the higher order corrections which influence the evolution of the initial and final states . within qcd ,
one expects the perturbative expansion in powers of @xmath6 to be well behaved and to provide an accurate estimate of the total cross section already at low orders . in particular ,
the first estimates of the total production cross section using the full next - to - leading - order ( nlo ) matrix elements @xcite gave an increase ( relative to the born result ) of the order of 30% for masses above 100 gev .
the residual perturbative qcd uncertainty , evaluated by varying the renormalization and factorization scales , was shown to be no larger than 10% .
the choice of parametrization for the input parton densities was also shown to give effects of this order of magnitude , by using the available sets .
it was pointed out in ref .
@xcite that logarithmic contributions associated to the emission of soft gluons from the initial state could significantly enhance the nlo result .
an independent study of the soft gluon resummation in top production has appeared recently @xcite .
these studies are based on the resummation formulae obtained in the pioneering works of refs .
@xcite . using these formulae for explicit calculations
requires the choice of a prescription in order to bypass the landau - pole singularity which is exposed by the integration of the qcd running coupling over gluon energies of the order of @xmath7 .
in particular it was shown in ref .
@xcite that , within the proposed formalism , one can get sensible results only by cutting off the soft gluon emission at scales below @xmath8 . the choice made in ref .
@xcite follows instead a suggestion put forward in ref .
@xcite . according to this prescription the landau pole appearing in the mellin transform of the resummed hard cross section
should be integrated over in a principal value sense .
it was argued in ref .
@xcite that this prescription results in non - perturbative power - suppressed ambiguities scaling like @xmath9 in the hard scale @xmath10 typical of the process under consideration .
the net effect of gluon resummation evaluated in ref .
@xcite amounts to an increase by approximately 10% of the nlo cross section .
in a companion paper @xcite we will show that the approaches of refs .
@xcite overestimate both the gluon resummation contribution and the associated non - perturbative residual uncertainty .
this is a consequence of the fact that their resummation formulae introduce unjustified factorially growing terms in the perturbative expansion . in ref .
@xcite we will also show that a more natural prescription exists , in which these terms are not present . in the following , we will briefly summarize the basis of our criticism and the definition and main properties of the new proposed resummation prescription .
we will then present a numerical study indicating that the impact of resummation on the total top production cross section is of the order of a percent , much smaller than previously thought .
we will also show that threshold corrections due to higher order coulomb effects are likewise negligible .
we will conclude this study with an updated analysis of the current theoretical uncertainties coming from the scale dependence , from the choice of partonic densities and from the value of .
it should be pointed out that this study is performed within the strict domain of the standard model .
corrections to the top cross section much larger than the standard model qcd uncertainties can be obtained in the presence of new phenomena . for a partial list of specific examples , see _
e.g. _ refs .
the heavy flavour production cross section in the born approximation is given by the formula [ hvqcrosssection ] = _ i , j = q|q,|qq , gg _
0 ^ 1 dx_1dx_2 f^(1,i)(x_1 ) f^(2,j)(x_2)^(ij ) ( ) , = , where @xmath11 is the mass of the heavy quark , and @xmath12 is the square of the total centre - of - mass energy .
the functions @xmath13 are the parton densities for parton @xmath14 in hadron @xmath15 , and are evaluated at a scale @xmath16 of the order of the heavy quark mass @xmath11 .
explicit formulae for the partonic cross section @xmath17 can be found for example in ref .
the corresponding @xmath18 space formula is [ mphq ] = _ i , j = q|q,|qq , gg _ c - i^c+i f^(1,i)_n+1f^(2,j)_n+1 ^(ij)_n ^-n dn , where the mellin transforms are defined in the following way f_n=_0 ^ 1 x^n f(x ) , _
n=_0 ^ 1 z^n(z ) .
the integration contour lies to the right of all singularities .
we use the following resummed formula [ sigmares ] ^(res)=_i , j = q|q,|qq , gg _ c_mp - i^c_mp+i f^(1,i)_n+1f^(2,j)_n+1 ^(ij)_n+1 ^(ij)_n ^-n dn .
where , in the @xmath19 scheme [ deltahf ] ^(ij)_n = n g_ij , 1(0n ) + o(^k ^kn ) where [ g1hf ] g_q|q , 1 ( ) = g_1^ ( ) , g_gg , 1 ( ) = g_1^ ( ) [ g0ms ] g_1^ ( ) = + , and b_0=c_a=3,c_f=. in formula ( [ deltahf ] ) the strong coupling constant @xmath20 is evaluated at a scale @xmath16 of the order of the heavy quark mass @xmath11 .
the factor @xmath21 resums all the leading logarithmic terms @xmath22 due to soft gluon emission . the contour in eq . ( [ sigmares ] ) is chosen between the cuts on the negative @xmath18 axis and the singularity at @xmath23 , which is a landau pole
. we will call this the minimal prescription ( mp ) . in ref .
@xcite we will show that this prescription enjoys the following remarkable properties : * it is accurate at the leading log level in the threshold limit . *
if we expand the mp formula in powers of @xmath24 , the expansion is asymptotic to the full formula .
* the power expansion of the mp formula does not have any factorially growing coefficients , and therefore is free of ambiguities of the order of powers of @xmath25 .
the ambiguity associated with the asymptotic expansion of the mp resummation is exponentially suppressed , being of the order @xmath26 .
we claim that previous calculations of resummation effects @xcite have a corresponding power expansion which does have factorially growing terms .
these terms are spurious .
they arise either from certain approximations performed when going from the @xmath18 space to the @xmath27 space formulae , or from the integration of the running coupling constant down to the landau singularity .
our result is consistent with the findings of ref .
@xcite , indicating that no power - like @xmath28 ambiguity emerges from the resummation of multiple gluon emission in drell yan processes
. a full discussion of these issues will appear in a companion paper @xcite , where we will also study the implications of this prescription for drell yan , heavy quark and jet production in hadronic collisions . in the present work we will report on results obtained with our mp formula of eq .
( [ sigmares ] ) , which was implemented without any further approximation in a numerical program .
in this section we present the results obtained for a very heavy quark , and in particular for the top cross section .
first of all , we remind the reader that large soft gluon effects are present not only in the production cross section , but also in the deep - inelastic processes that are used to determine the structure functions . in order to perform a complete study of the resummation one should therefore use resummed formulae also when fitting deep - inelastic scattering data , direct photon production data , and in general all phenomena which are used to constrain the parton densities . in the present work we will assume that the structure functions have been properly extracted from data ( including the resummation effects ) and assess the significance of the resummation for heavy flavour production .
effects of comparable size may arise from refitting the structure functions using resummed formulae .
our conclusions will mostly be based upon the fact that the resummation effects we find are in fact very small , and can be neglected .
the importance of the resummation effects is illustrated in fig .
[ frtev ] where we plot the quantities , , where @xmath29 is equal to our mp resummed hadronic cross section ( [ sigmares ] ) in which the terms of order @xmath30 and @xmath31 have been subtracted , and @xmath32 is the full hadronic nlo cross section @xcite . in other words , what we display is the relative contribution to the total cross section coming from corrections of order higher than those accounted for by the full nlo result .
we show the results for the @xmath33 and @xmath34 production channels separately , as well as for the total .
the results are shown at @xmath35tev as a function of the top mass in the range @xmath36 .
the wide mass range is chosen for the sole purpose of illustration .
we chose a common renormalization and factorization scale @xmath37 , and parton densities from the cteq1 set @xcite .
the plot clearly indicates that the contribution of soft - gluon resummation is negligible , unless the top mass approaches the total hadronic energy .
this result is consistent with the expectation that the soft gluon enhancement of the production cross sections should be relevant only very close to threshold . for a top mass of @xmath38
gev the increase due to this effect is below 1% .
we also note that the effect of resummation is potentially much larger for the @xmath34 channel , because of the @xmath39 enhancement in the exponent of the resummation formula ( see eq .
[ g1hf ] ) .
it turns out , however , that when the @xmath34 subprocess becomes important ( _ i.e. _ for small top masses ) we are so far from threshold that the resummation effect is very small .
conversely , for large masses , the quark component , which has smaller resummation effects , dominates .
we see that at the lhc , where production is dominated by the @xmath34 initial state , we are far enough from the hadronic threshold , so that the resummation effects are again small .
we conclude therefore that the effects of gluon resummation can be neglected at the present stage , as they are much smaller than the uncertainties present in the nlo determination .
note that , due to the dominance of the @xmath34 subprocess , the relative importance of the resummation effects at @xmath40 tev is larger in @xmath41 collisions than in @xmath42 . beside the soft gluon emission effects , also coulomb effects may enhance or deplete the cross section near threshold @xcite .
we have calculated these effects in the following way .
we have separated the partonic born cross section formulae into their colour singlet and octet components ^(gg)&=&_(8)^(gg)+_(1)^(gg ) + ^(q|q)&=&_(8)^(q|q ) where @xmath43 can be found in refs .
@xcite , and _
( 1)^(gg)()= , where @xmath44 .
the coulomb - resummed cross section is given as ^coul ( ) = _
( 8) ( ) + _ ( 1 ) ( ) .
we do not include bound state effects , which , as shown in ref .
@xcite , are much smaller . in fig .
[ topcoul ] we have plotted the hadronic quantity which ( similarly to the case of soft gluon effects ) represents the relative correction due to the resummation of coulomb effects not already included into the nlo results .
we see that these effects are similar in magnitude to the soft - gluon effects , and fully negligible at the tevatron .
since at the lhc we are further away from the threshold region , we conclude that also there they will be negligible
. therefore , the following estimates of the top cross section will not include neither soft gluon effects , nor coulombic ones .
electroweak corrections to top production in hadronic collisions have been considered in refs .
they range from -0.97% to -1.74% of the born cross section , for a higgs mass of 60 and 1000 gev respectively @xcite .
they will also not be included in the following estimates .
we now proceed to review the current theoretical uncertainties that arise at nlo .
uncertainties due to unknown higher - order effects are usually accounted for by varying the renormalization ( @xmath45 ) and factorization ( @xmath46 ) scales . in principle , independent variations of the two scales should be considered . in fig .
[ scales ] we show the scale dependence of the top cross section .
note that there is compensation of the factorization scale dependence when the different subprocesses are added up .
in particular , the @xmath34 and @xmath47 factorization scale dependence have opposite behaviour .
in fact , it is only the combined scale dependence that can be considered an estimate of the neglected subleading corrections . as a second point
, we observe that the maximum of the cross section is reached around @xmath48 .
thus , the usual choice of the range @xmath49 appears to be particularly justified in this case . as a third point , we observe that the scale dependence in the cross section goes in the same direction for the two scales , so that , for the purpose of estimating the associated uncertainty , it is sufficient to consider the simultaneous variation of @xmath45 and @xmath46 . aside from the scale uncertainties , which reflect the limitations of the perturbative qcd calculation ,
there are uncertainties associated to our imprecise knowledge of the physical parameters involved .
in particular , the strong coupling constant is determined within a certain accuracy . in the case of top production at the tevatron ,
larger strong couplings tend to give larger partonic cross section . however , for larger strong couplings , the altarelli
parisi evolution , which occurs along the wide span in @xmath50 from the values at which deep - inelastic fits are performed ( @xmath51 ) to the top mass , softens the quark parton densities , thereby decreasing the cross section . in order to perform a fair estimate of the uncertainty due to @xmath52 , we need sets of parton densities fitted with different values of the strong coupling . the sets of ref .
@xcite meet our purpose that accompany the fortran program for the structure function sets of ref .
@xcite are not consistent with the values of @xmath20 quoted there , the differences being of the order of 1% . in the present work ,
we extract the values of @xmath53 from their quoted values of @xmath20 using the standard two - loop formula . ] . in fig .
[ lambdavar ] we show the cross section as a function of the strong coupling . for comparison
we also show the @xmath20 dependence if the parton densities are kept fixed .
we see the remarkable reduction in the @xmath20 dependence due to the compensation of the rise of the partonic cross section and the decrease of the quark parton densities .
our results for the top cross section are collected in tables [ top175 ] , [ top160 ] and [ top190 ] , for values of the top mass that span the current uncertainties .
we also show results obtained with with the recent parametrization derived in ref .
@xcite by including the cdf jet data .
we stress that the numbers in the table are obtained with a standard two - loop calculation .
no resummation effects are included .
our ranges for the top cross section are thus given by @xmath54 pb , @xmath55 pb , @xmath56 pb for = 160 , 175 , 190 gev respectively , and are adequately fitted in this range by the expression ( t|t)=e^ ( ) . as a central value for our determination we have chosen the mrsa@xmath57 @xcite result with @xmath58 , in association with a value of @xmath59gev ( which corresponds to @xmath60 , according to the standard two - loop formula @xcite ) . for the mrs sets with variable @xmath61 , we have used @xmath62gev ( which corresponds to @xmath63 ) . in @xmath42 collisions at @xmath64tev
we obtain , with the same method , a top cross section of @xmath65 pb , @xmath66 pb , @xmath67 pb for = 160 , 175 , 190 gev respectively .
the cross section bands are also shown in fig
. [ bandtop1 ] , for @xmath35tev and fig .
[ bandtop2 ] for @xmath64tev .
because of the high precision of this theoretical prediction , the measurement of the top cross section could become a sensitive probe of new physics @xcite @xcite . for reference , we also quote the cross section for top production at the lhc .
we get @xmath68nb , for @xmath69gev and @xmath70tev .
the error is obtained with the same scale and @xmath20 variations we used for the tevatron cross section .
we wish to thank r.k .
ellis for useful comments .
we also wish to thank w. hollik and d. wackeroth for computing the value of the electroweak corrections to top production quoted above .
two of us ( sc and lt ) acknowledge the hospitality and the partial support received by the cern th division while this work was carried out .
99 # 1#2#3_nucl .
_ * b#1 * ( # 2 ) # 3 # 1#2#3_phys .
lett . _ * # 1b * ( # 2 ) # 3 # 1#2#3_phys . rev .
lett . _ * # 1 * ( # 2 ) # 3 # 1#2#3_phys .
* # 1 * ( # 2 ) # 3 # 1#2#3_phys .
* d#1 * ( # 2 ) # 3 # 1#2#3_phys .
* # 1 * ( # 2 ) # 3 # 1#2#3_z .
* # 1 * ( # 2 ) # 3 f. abe et al .
, cdf collab . , ; + f. abe et al . ,
cdf collab . , .
s. abachi et al . , d0 collab .
p. nason , s. dawson and r. k. ellis , ; w. beenakker , h. kuijf , w.l .
van neerven and j. smith , .
g. altarelli , m. diemoz , g. martinelli and p. nason , .
e. laenen , j. smith and w.l .
van neerven , ; .
e. berger and h. contopanagos , ; + preprint anl - hep - cp-95 - 85 , hep - ph/9512212 .
dokshitzer , d.i .
dyakonov and s.i .
troyan , ; + g. parisi , ; + g. curci and m. greco , ; + d. amati et al . , ; + m. ciafaloni and g. curci , ; + p. chiappetta , t. grandou , m. le bellac and j.l .
meunier , .
g. sterman , .
s. catani and l. trentadue , ; . g. sterman and h. contopanagos , ; + l. alvero and h. contopanagos , .
s. catani , m.l .
mangano , p. nason and l. trentadue , in preparation .
e. eichten and k. lane , ; + c.t .
hill and s.j .
parke , ; + r. casalbuoni et al . ,
cpt-95/p.3176 , ugva - dpt 1995/04 - 887 , hep - ph/9505212 ; + b. holdom , m.v .
m. beneke and v.m .
braun , ; + yu.l .
dokshitser , g. marchesini and b.r .
webber , preprint cern - th-95 - 281 , hep - ph/9512336 .
j. botts et al . , .
a. sommerfeld `` atombau und spektrallinien '' , bd .
2 ( vieweg , braunschweig , 1939 ) ; a.d . sakharov , jetp * 18 * ( 1948 ) 631 .
v. fadin , v. khoze and t. sjstrand , .
s. willenbrock and a. stange , ; + w. beenakker , a. denner , w. hollik , r. mertig , t. sack and d. wackeroth , ; + c. kao , g.a .
ladinsky , c.p .
yuan fsu - hep-930508 , hep - ph/9305270 ; + r. harlander , m. jezabeck and j.h .
khn , ttp 95 - 25 , hep - ph/9506292
. w. hollik and d. wackeroth , private communication .
martin , r.g .
roberts and w.j .
stirling , .
l. montanet et al . , and 1995 off - year partial update for the 1996 edition available on the pdg www pages ( url : http://pdg.lbl.gov/ ) . j. huston et al . , michigan state preprint msu - hep-50812 , hep - ph/9511386 .
martin , r.g .
roberts and w.j .
stirling , .
yang and c.s .
li , ; + c.s .
hu , j.m . yang and c.g .
a. caner , for the cdf collaboration , presented at the 10@xmath71 les rencontres de physique de la valle daoste , la thuile , val daosta , march 3 - 9 , 1996 ; m. narain , for the d0 collabration , presented at the 10@xmath71 les rencontres de physique de la valle daoste , la thuile , val daosta , march 3 - 9 , 1996 . | # 6 plus 1000pt minus 1000pt # 1 # 1= to # 1#2#1#2 # 1#2#3 ^ 2 # 1#2 # 3 # 1#1| # 1| # 1 # 1#1 # 1/ # 1| # 1| # 1#1^ = # 1#2#3 @xmath0 # 1#2#3#4 @xmath1 # 1#2 @xmath2 # 1#2#3#4#5 @xmath3 # 1 # 1@xmath4 = 1.5ex plus 1pt = 11 addtoresetequationsection normalsizesetsizexiipt plus3pt minus3pt@ plus3pt plus3.5pt minus0pt twosidetrue
@=12 0.0 in 0.0 in 6.0 in 8.5 in .7 cm -1 cm .75 in = 11 |
our statistical study of @xmath11350 edge - on galaxies derived from the rc3 ( _ third reference catalogue of bright galaxies _ , de vaucouleurs et al .
@xcite ) ( ltticke et al .
@xcite , hereafter paper i ) has revealed that 45% of all bulges ( s0-sd galaxies ) are box- and peanut - shaped ( b / p ) .
this high frequency must be considered in a theory of bulge formation and especially in scenarios explaining the formation process of b / p bulges .
currently discussed theories of bulge formation are primordial or secular evolution scenarios . while monolithic collapse ( eggen et al . @xcite ( els ) ) , clumpy collapse ( kauffmann et al .
@xcite , baugh et al .
@xcite ) , and inside - out formation ( van den bosch @xcite , kepner @xcite ) are proposed for primordial bulge formation , merging processes ( wyse et al . @xcite and references therein ) and dynamical evolution triggered by bars ( e.g. combes et al .
@xcite ) are the basis of secular evolution scenarios .
however , combinations of both scenarios are also possible ( combes @xcite ) . in the last 20 years several working hypothesis about the origin of b / p bulges have been developed .
external cylindrically symmetric torques can act on an initially non - rotating triaxial spheroidal stable system so that it becomes box - shaped ( may et al .
@xcite ) . however , such torques are `` problematical to relate to some astrophysical counterparts '' ( combes et al .
mergers of two disk galaxies are proposed by binney & petrou ( @xcite ) and rowley ( @xcite ) as a possible formation process of b / p bulges .
however , such a scenario requires very special conditions such as the precise alignment of the spin and orbital angular momenta of the two galaxies ( bureau @xcite ) .
a more common process to form b / p bulges could be the accretion of satellite galaxies , although an oblique impact angle of the satellite is needed ( binney & petrou @xcite , whitmore & bell @xcite ) and the mass of the accreted satellite is constrained .
if it is too large , the stellar disk would be disrupted ( barnes @xcite , hernquist @xcite ) . the origin of b / p bulges by bars was first proposed by combes & sanders ( @xcite ) .
this theory was strongly supported ten years later by the studies of combes et al .
( @xcite ) , raha et al .
( @xcite ) , and pfenniger & friedli ( @xcite ) dealing with n - body simulation of disk galaxies . the connection between b / p bulges and the presence of a bar is difficult to prove by surface photometry , because b / p bulges are observable only in almost edge - on galaxies
. however , there is photometric evidence in a few edge - on galaxies from cuts parallel to the major axis ( de carvalho & da costa @xcite , dettmar & ferrara @xcite ) and in two intermediately inclined galaxies ( bettoni & galletta @xcite , quillen et al .
@xcite ) pointing to bars .
additionally , kinematic observations of around 20 b / p bulges with characteristic bar signatures in velocity fields and in orbit analyses ( e.g. kuijken & merrifield @xcite ) are reported .
in addition , our statistics ( paper i ) have shown that the overall frequency of barred galaxies ( @xmath1 55% , paper i , knapen et al .
@xcite ) is able to explain the high fraction of b / p bulges . on the other side observational evidence for accreted material in b / p bulges
is found by bureau & freeman ( @xcite ) for a few b / p bulges .
indirect evidences are substructures and asymmetries in some special b / p bulges ( dettmar @xcite , dettmar & barteldrees @xcite , ltticke et al .
@xcite , hereafter paper iii ) .
combes et al .
( @xcite ) and shaw ( @xcite ) conclude from their investigation of b / p bulges , listed by jarvis ( @xcite ) , de souza & dos anjos ( @xcite ) , and shaw ( @xcite ) , that they show less evidence for significant merging and are likely not the result of recent accretion events .
however , b / p bulges can indirectly originate from interaction and finally accretion by triggering the formation of bars ( noguchi @xcite , gerin et al .
@xcite , walker et al .
@xcite ) , which then actually build up the b / p structures ( fisher et al .
@xcite , mihos et al .
therefore the search for bars is one of the best ways to discriminate between the proposed evolutionary scenarios of b / p bulges .
the essential advantage of investigations in the nir is the small extinction by dust at these wavelengths ( rieke & lebowsky @xcite , knapen et al .
therefore the influence of the dust lane in the plane of edge - on galaxies is largely reduced by nir observations , especially in the @xmath2-band . in this way
the structure of a galaxy near the plane in radial as well as in vertical direction can be analysed and thereby bars detected ( dettmar & ferrara @xcite ) .
in addition it is even possible to determine parameters of bars , e.g. their thickness and projected length .
therefore it is possible to investigate the connection between bars and shapes of bulges .
observations were carried out at the 1.5 m tirgo / gornergrat in 1994/95 ( run identification : ti ) with nicmos3 arnica ( 0.92@xmath3/pixel ) , the 1.2 m telescope at calar alto in june 1996 ( ca4 ) and june 1997 ( ca2 ) with magic ( rockwell nicmos3 array ) using the high resolution mode ( 1.2@xmath3/pixel ) , the 2.2 m telescope at calar alto during two runs in january 1997 ( ca3 ) and march 1998 ( ca1 ) also equipped with magic ( hr mode : 0.64@xmath3/pixel ) , and at eso / la silla with the 2.2 m telescope in may 1998 ( e1 ) with irac-2b ( nicmos3 array ) using the lc objective ( 0.49@xmath3/pixel ) . to the resulting sample of 56 galaxies
four galaxies kindly provided by emsellem ( e2 : eso / la silla 2.2 m telescope , april 1998 , irac-2b , lc objective ) are added .
the sample was selected from the list of edge - on galaxies with bulge classification ( paper i ) according to the allocated observing time , matching the field size , and covering all bulge types .
filters and typical on - source integration times for each observing run are listed in table [ obs ] and the final sample is given in table [ gal ] .
at eso / la silla and tirgo the @xmath4- resp .
@xmath2-band filters were used . due to large thermal emission in @xmath2 for magic
we observed at calar alto mainly in @xmath5 . using different filters
we can show that the bulge shape in the nir is independent from the observed wavelengths ( paper i ) .
.observational data of the nir [ cols="<,^,^,^,^ " , ] notes : + col .
( 1 ) : bulge type as defined in paper i. + col . ( 2 ) : inclination of the simulated galaxy . + col .
( 3 ) : aspect angle of the bar ( @xmath6 = edge - on and @xmath7 = end - on ) .
( 4 ) : frequency of the bulge type of the simulated galaxy , viewed under different aspect angles , at a constant inclination .
( 5 ) : modified frequency of the bulge type of the simulated galaxy , viewed under different aspect angles , at a constant inclination taking into account that only 55% of all galaxies are barred and the remaining galaxies without a bar ( 45% ) have a bulge of type 4 .
( 6 ) : as col .
( 5 ) , but with 70% barred and 30% unbarred galaxies + along the bar major axis of the simulated galaxy , measured in the face - on case ( @xmath8 ) , the decrease of the surface density is exponential ( pfenniger & friedli @xcite ) and the length of the bar can be determined to @xmath120 kpc .
this length exactly fits inside the radius of corotation which occurs at 10 kpc ( pfenniger & friedli @xcite ) .
the axis ratio of the bar is @xmath9 ( fig . [ sim ] , top right ) and is inside the range of observed values ( 0.2 - 1 ) ( martin @xcite ) .
regarding the galaxy edge - on ( @xmath10 ) with an edge - on orientated bar ( @xmath11 ) is the best way to derive the quantitave bar parameters ( introduced in section 3 ) of the simulated galaxy , because bar and b / p structures are most prominent at this geometry .
however , the cuts along and parallel to the major axis do not reveal the typical bar signature ( fig .
[ simcut ] ) which is visible in the nir observations .
the three components bulge , bar , and central part of the bulge can not be distinguished as it is possible in the observations ( cf .
[ n2654 ] , [ n7332 ] , and [ n1175 ] ) . due to the lack of a bar signature the bar thickness as well as the central bulge length ( cbu ) can not be determined .
the ratio of bal / bul @xmath12 , taking into account bal @xmath13 kpc and bul @xmath14 kpc ( fig .
[ simcut ] ) , is not consistent with the observed ratios for peanut bulges ( @xmath15 ) ( fig .
[ balbul ] ) , while bal / bpl @xmath16 ( bpl @xmath18 kpc ; fig .
[ simcut ] ) is consistent with observed ratios in peanut bulges ( bal / bpl @xmath17 ) . while in observations
the positions of the maxima of the b / p distortion coincide with the limits of the central bulge , in the n - body simulation the positions of the maxima are not related to any other features in the profiles . the differences between the galaxy derived from the n - body simulation and the observed galaxies likely result from the initial conditions of the simulated model .
pfenniger & friedli ( @xcite ) are just using two merged disks with the same rotation properties instead of an independent bulge component in their simulation .
the inner disk represents the `` bulge '' and is chosen with the main purpose `` to have sufficient freedom to shape the initial rotation curve '' .
therefore the initial resemblance with a real bulge is only approximate .
using such a bulge component in the simulation could also be the reason for the lack of the bar signature when the galaxy is seen edge - on , since the inner disk could cover the signature .
another explanations for the fact that no bar signature is visible could lay in the limited spatial resolution or in the lack of gas in the simulation , which influences the bar structure .
it is also possible that the exponential decrease of the bar surface brightness and the smooth end of the bar prevent the detection of a bar signature .
additionally , the investigated simulated galaxy is only a single , although representative , example of a barred galaxy . variations of the initial conditions such as the bulge - to - disk ratio , q parameter , gas fraction and distribution , and halo structure end up in different bar strenghts , shapes , thickness , lenghts , and density profiles in n - body simulations ( e.g. friedli & benz @xcite ) .
therefore conclusions drawn from comparisons between our statistics of observed bulge types ( paper i ) and results from the simulated galaxy , which is investigated in this study , are limited .
future studies have to analyse recently published simulations of barred galaxies ( e.g. pfenniger @xcite ) to get more relevant comparisons with observations .
the strong correlation of the bar signature with b / p bulges in our study shows that b / p bulges are well explained by dynamical processes in bar potentials .
the ratio of the bar length to the length of the b / p distortion ( bpl ) ( @xmath18 ) in the simulated barred galaxy and the observed ratio of bal / bpl ( @xmath19 ) in galaxies with peanut bulges is consistent with such a scenario .
the correlation of the bar signature with b / p bulges is previously claimed , using the same detection method , by the investigations of individual galaxies , namely ngc 1381 ( de carvalho & da costa @xcite ) , ngc 5170 ( dettmar & barteldrees @xcite ) , and ngc 4302 ( dettmar & ferrara @xcite ) .
many further evidences for the connection between bars and b / p bulges are reported in the literature : bettoni & galletta ( @xcite ) and quillen et al .
( @xcite ) find them in their investigation of the intermediately inclined galaxies ngc 4442 resp .
ngc 7582 . furthermore ,
n - body simulations ( e.g. combes et al .
@xcite , raha et al .
@xcite , pfenniger & friedli @xcite ) show that the evolution of a b / p bulge is coupled with the development of a bar .
independent evidence is obtained by observing the gas kinematics in nii and h@xmath20 . the characteristic `` figure - of - eight '' rotation curve which is a strong signature for the presence of a bar and is detected in many b / p bulges : e.g. ngc 5746 , ngc 5965 ( kuijken & merrifield @xcite ) , ngc 2683 ( merrifield @xcite ) , and additional ten further galaxies by bureau & freeman ( @xcite ) and five by merrifield & kuijken ( @xcite ) .
control samples of non - b / p bulges show no kinematical signature of a bar ( bureau & freeman @xcite , merrifield & kuijken @xcite ) . in two b / p bulges (
eso 597- 36 and ngc 3390 ) without kinematic bar signature bureau & freeman ( @xcite ) find indications for accretion events .
however , since the positioning of the slit was difficult ( bureau & freeman @xcite ) , dust in the galactic plane could affected their observations .
therefore bars can not be excluded in these two galaxies . a merger - induced bar formation with a resulting b / p bulge ( mihos et al .
@xcite ) could be a possible scenario for these galaxies .
the four detected non - b / p bulges : ngc5907 ( miller & rubin @xcite ) , ic5096 , eso240 - 11 ( bureau & freeman @xcite ) , and ngc 681 ( merrifield & kuijken @xcite ) with a bar signature in the gas kinematics can be explained by our analysis of the n - body simulation with bars which are nearly end - on ( see next subsection ) .
signatures of the gas kinematics in barred potentials are also detected in retrograde orbits of ngc 128 ( emsellem & arnault @xcite ) , the galaxy with one of the most prominent b / p bulges ( donofrio et al .
@xcite ) . direct kinematic evidence for bar streaming motions in two galaxies with b / p bulges ( ngc 3079 and ngc 4388 ) is presented by veilleux et al . ( @xcite ) . furthermore ,
in hi the presence of a bar is also verified in one galaxy with a peanut bulge analysing the position velocity diagram ( ic 2531 : bureau & freeman @xcite ) .
the gas kinematics in the central region obtained in co ( @xmath21 line ) in the case of ngc 3628 ( reuter et al .
@xcite ) , ngc 891 ( garcia - burillo & guelin @xcite ) , and ngc 4013 ( garcia - burillo et al .
@xcite ) again point to the presence of a bar . the conclusion ( often given in the literature , e.g. pfenniger & friedli @xcite ) , that b / p bulges are thick bars , can not be supported . from the observational point of view bars and bulges , if b / p - shaped or not , are obviously individual components with different lengths ( e.g. fig .
[ n2654 ] ) .
stellar orbits in a barred potential build up b / p structures and the bar itself .
therefore b / p bulges are the composition of a spheriodal bulge and a b / p structure representing the thick part of the bar .
the non - detection of bars ( table [ connec ] ) in all observed tbbs ( @xmath15% of all b / p bulges belong to this group , paper iii ) is an indication for the hypothesis that the boxiness of these bulges is not related to the the presence of a bar .
this idea gets more convincing support from the fact that the extent of the box shape is too large to result from an usual bar potential ( paper iii ) .
however , the lack of a bar signature does not necessarily mean that there is no bar , since weak bars , which have exponential surface brightness profiles ( elmegreen & elmegreen @xcite ) , could remain undetecable by the here presented search for a bar signature .
in this way also the missing bar signature in the simulated barred galaxy is explainable , because the surface density distribution along the simulated bar is exponential and additionally , the bar axis ratio of this galaxy points to a weak bar ( @xmath22 , martinet & friedli @xcite ) .
the connection between b / p bulges and bars can be understood as interplay between resonant bending fed by vertical diffusion of orbits and instabilities similar to the fire - hose one ( bureau @xcite , friedli @xcite ) .
while there is evidence for this connection in the most galaxies with b / p bulges , the question for the different shape of the b / p bulges remains to be answered .
a correlation between a strong signature of a bar , marked by constant surface brightness in radial direction ( cf . section 3 ) , and prominent b / p bulges is found in our study .
thereby the strong bar signature can be associated with a flat bar , which is known to be strong ( elmegreen & elmegreen @xcite ) .
therefore the correlation can be interpreted by the assumption that strong barred galaxies are responsible for prominent b / p bulges , while galaxies with a weak bar have bulges with a lower degree of boxiness .
n - body simulations by friedli ( @xcite ) support this explanation for different bulge shapes . in this way
it is also possible to explain partly the observed correlation of bal / bul with the bulge type ( cf .
section 4 ) by larger bars in galaxies with prominent b / p bulges , because flat bars tend to be larger ( elmegreen & elmegreen @xcite , elmegreen et al .
since this correlation holds for early as well as for late types , it can not be explained by the fact that the same bar will produce more pronounced b / p bulges , the smaller the pre - existing spheroidal bulge is .
however , having this in mind it is not surprising that the weak bar of the here investigated simulated galaxy can produce a peanut bulge .
furthermore , one could expect that peanut bulges more often occur in late type spirals , but the observed fraction of peanut bulges in galaxies in the range of s0-scd is rather constant and equal to zero in spirals later than sc ( paper i ) .
this result could be interpreted by a dependence of the bar strength on the hubble type .
such a correlation of bars is found by elmegreen & elmegreen ( @xcite ) and elmegreen et al .
( @xcite ) .
however , in contrast to them seigar & james ( @xcite ) do not find that flat ( strong ) bars are more common in early hubble types and exponential ( weak ) bars in later types .
this statement is in agreement with our results for the profiles of bar signature concerning the hubble type , but it is uncertain , if their results concerning bar morphology detected in face - on galaxies can be transformed to bar signatures in edge - on galaxies .
additionally , for the explanation of bulge shapes it has to be taken into account that gas mass and a central black hole effect the shape of bulges . a significant fraction of gas mass ( berentzen et al .
@xcite ) or a supermassive central black hole can weaken the b / p structure ( friedli @xcite ) . explaining bulge shapes by aspect angles of bars was introduced by n - body simulations ( combes et al . @xcite , pfenniger & friedli @xcite ) .
kinematical bar diagnostics in edge - on spiral galaxies using simulations of families of periodic orbits and hydrodynamical simulations are also consistent with observations .
they confirm the dependence of the boxiness on the aspect angle of the bar ( bureau & athanassoula @xcite , athanassoula & bureau @xcite ) .
an exact determination of aspect angles is not possible , because the envelope of the signature of the outer bar region in the position - velocity diagram is too faint ( athanassoula & bureau @xcite ) .
the strong correlation of bal / bul with the bulge type and the agreement of the observed ratio bal / bul only of the most prominent b / p bulges , @xmath15 , with that ratio obtained for face - on galaxies , @xmath23 ( athanassoula & martinet @xcite , baumgart & peterson @xcite , martin @xcite ) , shows that the differences in bal / bul can be interpreted by different aspect angles of bars .
therefore our analysis points also to a dependence of the shape of a bulge on the aspect angle of the bar . in this way and
consistent with our investigation of the simulated barred galaxy peanut bulges result from bars seen nearly edge - on ( large projected bar length , bal ) , boxy bulges from intermediately inclined bars , and elliptical ones from end - on bars ( small bal ) or they have no bar at all .
thereby bars detected in non - b / p bulges can be explained as well as differences between the fraction of barred galaxies ( 70% in the nir ) and galaxies with b / p bulges ( 45% ) , although also other parameters might play a role ( weak bar , large gas mass , or dominant bulge component ) .
the dependence of the bulge shape on the aspect angle of the bar is also supported in our study by the modified frequencies of the bulge types derived in the simulation .
these frequencies are consistent with the observed frequencies of the bulge types , if one keeps in mind that only a single simulated barred galaxy is compared with all morphological types of disk galaxies .
larger bulge components in the simulation would weaken the b / p structures and thereby reduce the modified frequency of b / p bulges .
stronger bars would have the contrary effect on the modified frequency .
additionally , many galaxies of our investigated sample with peanut bulge and strong bar signature show features at the ends of the bumps , which can be associated with spiral arms extending outward from the ends of a bar ( baggett & mackenty @xcite ) .
therefore it can be suggested that bars of these galaxies are indeed nearly edge - on .
there exists only a few rough estimates of aspect angles of bars in edge - on galaxies from observations .
wakamatsu & hamabe ( @xcite ) adopt for ngc 4762 ( bulge type 3 in the optical , type 4 in the nir ; paper i ) an aspect angle of the bar of @xmath24 ( although a lower limit of 21@xmath25 is also possible from their argumentation ) and garcia - burillo & guelin ( @xcite ) for ngc 891 ( bulge type 2 ) @xmath26 .
these results are in agreement with a dependence of the bulge shape on the bar aspect angle .
the coincidence of the positions of the maxima of the b / p distortion with the limits of the central bulge , which is found here for the first time , can not finally be explained ( @xmath27 ) .
selection effects due to different aspect angles of the bar , which would influence the length of the b / p distortion ( bpl ) , can be neglected , since we have shown that peanut bulges , for which bpl is defined only , can be associated with nearly edge - on bars .
the coincidence points also to a possible connection between the bar length and the length of the central bulge ( cbu ) , since bpl is connected with bal . therefore possible explanations could be related to secondary bars ( e.g. friedli @xcite , greusard et al .
this idea is supported by the result from n - body simulations ( friedli & martinet @xcite ) that the corotation of the inner bar roughly coincides with the location of the horizontal inner lindblad resonance ( ilr ) . at this position
also the maximum of the elevation of the peanut structure occurs , because the vertical ilr is very close to the horizontal ilr and the conjunction of these two resonances build up the peanut structure ( combes et al .
@xcite ) . however , the observed ratio of the bar lengths of the primary to the secondary , which ends near corotation ( greusard et al .
@xcite ) , varies between 4.0 and 13.4 ( friedli et al .
@xcite ) , respectively 2.7 and 7.1 ( greusard et al .
@xcite ) , while the ratio of bar length to length of the central bulge is much smaller and varies only between 2.4 and 3.1 .
the central bulge could also be associated with a primordial bulge , which influences the bar structure and the position of the horizontal ilr . since at this location also
the peanut structures is most prominent , the extent of the b / p distortion would depend on the mass and length of the primordial bulge .
such a hypothesis is supported by n - body simulations in which the size of the evolving bar is directly related to the initial size of the bulge ( pfenniger @xcite ) . in this way
the nearly constant ratio of bar length or bpl to cbu would be a natural result .
the correlation between cbu and bpl and especially the cbu / bpl ratio of 1 could also be explained by nuclear rings .
they correspond to the horizontal ilr of bars ( buta & crocker @xcite , buta & combes @xcite ) and therefore occur at the same radial position as the maximum of the elevation of the peanut structure
. a nuclear ring could be responsible for the increase of the surface brightness , which we observe in radial cuts near the plane ( fig .
[ n2654 ] , [ n7332 ] , [ n1175 ] ) , or the ring is only the border for the central bulge .
however , we have not seen any structure inside the central bulge pointing to rings , but it is likely that our resolution is too low to detect them . since we know now with high probability the fact that bars in peanut bulges are edge - on orientated , it is for the first time possible to determine the thickness of bars by observations . from the mean value of bal / bat for the galaxies with peanut bulges a ratio of bar length to thickness of @xmath0 ( taking the error in bal and bat into account ) is derived .
this value is at the upper limit of the values estimated in the literature ( ranging from 4 to 15 , sellwood & wilkinson @xcite ) .
it must be mentioned that bars are more extended in the nir than in the optical ( friedli et al .
@xcite , chapelon et al .
@xcite ) . however , from a sample of nine barred galaxies ( wozniak et al .
@xcite , friedli et al .
@xcite ) investigated in the @xmath28- and @xmath2-band we have computed that this effect is only of the order of @xmath15% . therefore our rather large value for the ratio is not an effect of the observed wavelengths .
we have analysed a sample of edge - on disk galaxies in the nir and images derived from a n - body simulation of a barred galaxy using cuts parallel to the major axis and our classification of bulge shapes introduced in paper i. our main conclusions are : * cuts along and parallel to the major axis of edge - on galaxies show a strong correlation between bars and b / p bulges .
this correlation is for the first time verified in a larger sample of galaxies ( n @xmath29 ) . *
the investigation of the bar signatures in the nir reveals a correlation between prominent b / p bulges and strong bar signatures pointing to a dependence of the boxiness of a galaxy on the bar strength . *
the ratio of the projected bar length ( bal ) to the length of the bulge ( bul ) correlates with the bulge types , i.e. with the bulge shape .
this correlation is the first evidence from surface photometry for an explanation of the different degree of boxiness by the aspect angle of a bar .
the maximum of this ratio for galaxies with peanut bulges is consistent with ratios observed in face - on galaxies . *
the ratio of bal to the length of the b / p distortion ( bpl ) has a mean value for the observed peanut bulges of @xmath19 .
this value is consistent with the investigated n - body simulation where a ratio of @xmath18 can be derived for an edge - on galaxy with an edge - on bar . *
the frequencies of bulge types in the simulated galaxy , seen under different aspect angles of a bar and under high inclinations , are consistent with the observed frequencies corrected for the total fraction of barred galaxies .
these results support the interpretation that bars viewed under different angles in edge - on galaxies play a major role for the bulge shape .
however , the viewing angle is not the only parameter , since the size of the pre - existing spheroidal bulge and various bar shapes and types also influence the bulge shape . * interpreting galaxies with peanut bulges as galaxies with a bar seen edge - on the ratio of bar length to thickness , @xmath30 , can be directly measured for the first time . * from the observational point of view bars and b / p bulges are different components revealed by their different lengths .
the inner part of barred disk galaxies can be described by three components : the spheroidal bulge ( including the central bulge ) , the bar ( = thin bar ) , and the b / p structure ( = thick bar ) .
* we have found a new size relation of the positions of the maxima of the b / p distortion with the extent of the central bulge ( @xmath31 ) .
this size relation is most likely related to the in - plane ilr of the bar .
part of this work was supported by the _ deutsche forschungsgemeinschaft , dfg_. furthermore the authors want to thank daniel pfenniger who provided the particle position file , andreas schrer for helping with idl programming , and the referee f. combes for constructive comments . combes f. , 2000 , bulge formation . in : hammer f. , thuan t.x .
, cayatte v. , guiderdoni b. , tran thanh van j. ( eds . ) , building galaxies : from the primordial universe to the present , proceedings of rencontres de moriond , ed .
frontieres , p. 413
dettmar r .- j .
, barteldrees a. , 1990b , observational evidence for a relation of box- and peanut - shaped bulges to bars . in : jarvis b.j .
, terndrup d.m .
( eds . ) , eso / ctio workshop on bulges of galaxies , eso , p. 259 | we have observed 60 edge - on galaxies in the nir in order to study the stellar distribution in galaxies with box / peanut - shaped bulges .
the much smaller amount of dust extinction at these wavelengths allows us to identify in almost all target galaxies with box / peanut - shaped bulges an additional thin , central component in cuts parallel to the major axis .
this structure can be identified with a bar .
+ the length of this structure scaled by the length of the bulge correlates with the morphologically classified shape of the bulge .
this newly established correlation is therefore mainly interpreted as the projection of the bar at different aspect angles .
galaxies with peanut bulges have a bar seen nearly edge - on and the ratio of bar length to thickness , @xmath0 , can be directly measured for the first time .
in addition , the correlation of the boxiness of bulges with the bar strength indicates that the bar characteristic could partly explain differences in the bulge shape .
furthermore , a new size relation between the box / peanut structure and the central bulge is found .
our observations are discussed in comparison to a n - body simulation for barred galaxies ( pfenniger & friedli @xcite ) .
we conclude that the inner region of barred disk galaxies are build up by three distinct components : the spheroidal bulge , a thin bar , and a b / p structure most likely representing the thick part of the bar . |
in this paper we are concerned with the problem of boundary stabilization for a @xmath0 system of first - order hyperbolic _ quasilinear _ pdes , with actuation at only one of the boundaries .
the quasilinear case is of interest since many relevant physical systems are described by @xmath0 systems of first - order hyperbolic quasilinear pdes , such as open channels @xcite , transmission lines @xcite , gas flow pipelines @xcite or road traffic models @xcite .
this problem has been considered in the past for @xmath0 systems @xcite and even @xmath3 systems @xcite , using the explicit evolution of the riemann invariants along the characteristics .
more recently , an approach using control lyapunov functions has been developed , for @xmath0 systems @xcite and @xmath3 systems @xcite .
these results use only static output feedback ( the output being the value of the state on the boundaries ) .
however , they do not deal with the same class of systems considered in this work ( which includes an extra term in the equations ) ; with this term , it has been shown in @xcite that there are examples ( even for linear @xmath0 system ) for which there are no control lyapunov functions of the `` diagonal '' form @xmath4 ( see next section for notation ) which would allow the computation of a static output feedback law to stabilize the system , even if feedback is allowed on both sides of the boundary .
several other authors have also studied this problem .
for instance , the linear case has been analyzed in @xcite ( using a lyapunov approach ) and in @xcite ( using a spectral approach ) .
the nonlinear case has been considered by @xcite and @xcite using a lyapunov approach , and in @xcite , @xcite , and @xcite using a riemann invariants approach .
the basis of our design is the backstepping method @xcite ; initially developed for parabolic equations , it has been used for first - order hyperbolic equations @xcite , delay systems @xcite , second - order hyperbolic equations @xcite , fluid flows @xcite , nonlinear pdes @xcite and even pde adaptive designs @xcite .
the method allows us to design a full - state feedback law ( with actuation on only one end of the domain ) making the closed - loop system locally exponentially stable in the @xmath1 sense .
the gains of the feedback law are the solution of a 4 x 4 system of first - order hyperbolic linear pdes , whose well - posedness is shown .
the proof of stability is based on @xcite ; we construct a strict lyapunov function , locally equivalent to the @xmath1 norm , and written in coordinates defined by the ( invertible ) backstepping transformation .
the paper is organized as follows . in section [ sect - prob ]
we formulate the problem . in section [ sect - linear ]
we consider the linear case and formulate a backstepping design that globally stabilizes the system in the @xmath5 sense . in section [ sect - main ] we present our main result , which shows that the linear design locally stabilizes the nonlinear system in the @xmath1 sense .
the proof of this result is given in section [ sect - proof ] .
we finish in section [ sect : conclusions ] with some concluding remarks .
we also include an appendix with the proof of well - posedness of the kernel equations , and some technical lemmas .
consider the system @xmath6,\ , t\in[0,+\infty),\label{eqn - z}\ ] ] where @xmath7\times[0,\infty)\rightarrow \mathbb{r}^2 $ ] , @xmath8 \rightarrow \mathcal{m}_{2,2}(\mathbb{r})$ ] , @xmath9 \rightarrow \mathbb{r}^2 $ ] , with @xmath10 denoting the set of @xmath0 real matrices .
we assume that @xmath11 is twice continuously differentiable with respect to @xmath12 and @xmath13 , and we assume that ( possibly after an appropiate state transformation ) @xmath14 is a diagonal matrix with nonzero eigenvalues @xmath15 and @xmath16 which are , respectively , positive and negative , i.e. , @xmath17 where @xmath18 denotes the diagonal matrix with @xmath19 in the first position of the diagonal and @xmath20 in the second .
we also assume that @xmath21 , implying that there is an equilibrium at the origin , and that @xmath22 is twice continuously differentiable with respect to @xmath12 .
denote @xmath23,\ ] ] and assume that @xmath24\right)$ ] . denoting @xmath25^t$ ] , we study classical solutions of the system under the following boundary conditions @xmath26 which are consistent ( see @xcite ) with the signs of ( [ eqn - speeds ] ) , at least for small values of @xmath12 .
we assume that @xmath27 is twice differentiable and vanishes at the origin . in ( [ eqn - bcz12 ] ) , @xmath28 is the actuation variable , and our task is to find a feedback law for @xmath28 to make the origin of system ( [ eqn - z]),([eqn - bcz12 ] ) locally exponentially stable .
_ the case with @xmath29 in ( [ eqn - z ] ) was addressed in @xcite and @xcite by using control lyapunov functions to design a static output feedback law ; this approach has been shown to fail in @xcite for some cases with @xmath30 , at least for a `` diagonal '' lyapunov function of the form @xmath31 . _
next , we present a new design , based on the backstepping method , to stabilize a @xmath0 hyperbolic linear system ; this procedure will be used later to _ locally _ stabilize system ( [ eqn - z ] ) , ( [ eqn - bcz12 ] ) .
consider the system @xmath32,\ , t\in[0,+\infty),\ ] ] where @xmath33\times[0,\infty)\rightarrow \mathbb{r}^2 $ ] , @xmath34 \rightarrow \mathcal{m}_{2,2}(\mathbb{r})$ ] , where the matrices @xmath35 and @xmath36 are respectively diagonal and antidiagonal , as follows : @xmath37 where @xmath38 are @xmath39)$ ] and @xmath40 are @xmath41)$ ] functions , verifying that @xmath42 , and with boundary conditions @xmath43 where @xmath44 and the components of @xmath45 are @xmath46 ^t$ ] . our objective is to design a full - state feedback control law for @xmath28 to ensure that the closed - loop system is globally asymptotically stable in the @xmath5 norm , which is defined as @xmath47 .
there are two cases , depending on whether @xmath48 in ( [ eqn - bculinear ] ) is nonzero or @xmath49 .
we first analyze the first case , thus assuming @xmath50 .
our approach to designing @xmath28 , following the backstepping method , is to seek a mapping that transforms @xmath45 into a _
target _ variable @xmath51 with asymptotically stable dynamics as follows : @xmath52 with boundary conditions @xmath53 where the components of @xmath51 are denoted as @xmath54^t.\end{aligned}\ ] ] system ( [ eqn - gammalinear ] ) , ( [ eqn - bc1beta ] ) verifies the properties expressed in the following proposition . [ pr - target ]
consider system ( [ eqn - gammalinear ] ) , ( [ eqn - bc1beta ] ) with initial condition @xmath55)$ ] .
then , for every @xmath56 , there exists @xmath57 such that @xmath58 in fact , the equilibrium @xmath59 is reached in finite time @xmath60 , where @xmath61 is given by @xmath62 define @xmath63,\ ] ] where @xmath64 will be computed later .
select : @xmath65 notice that @xmath66 defines a norm equivalent to @xmath67 .
computing the derivative of @xmath68 and integrating by parts , we obtain @xmath69_0 ^ 1,\label{eqn - udotlinear1}\end{aligned}\ ] ] where we have used that @xmath70 and @xmath71 commute . since @xmath72>0,\ ] ] and , on the other hand , @xmath73_0 ^ 1 = -a \alpha^2(1,t)\mathrm{e}^{-\mu } -(b - q^2a)\beta^2(0,t),\end{aligned}\ ] ] choosing @xmath74 , @xmath75 , and @xmath76 , where @xmath77}\left\{\frac{1}{\epsilon_1(x ) } , \frac{1}{\epsilon_2(x ) } \right\}$ ] , we get that @xmath78 , therefore : @xmath79 where @xmath80 can be chosen as large as desired .
this shows exponential stability of the origin for the @xmath51 system . to show finite - time convergence to the origin
, one can find the explicit solution of ( [ eqn - gammalinear ] ) as follows .
define first @xmath81 noting that they are monotonically increasing functions of @xmath13 , and thus invertible .
note that the components of @xmath51 verify the differential equations @xmath82 which can be rewritten as follows @xmath83 the solution of these equations is @xmath84 and @xmath85 , where @xmath86 and @xmath87 are arbitrary functions . now , if @xmath88 , @xmath89 are the initial condition for the states , one obtains @xmath90 ( valid for @xmath91 ) and @xmath92 ( valid for @xmath93 ) . using the boundary conditions ( [ eqn - bc1beta ] )
one finds the remaining values of @xmath86 and @xmath87 , and thus the solution of the system , as follows : @xmath94 thus , after @xmath60 , where @xmath95 one has that @xmath96 . to map the original system ( [ eqn - wlinear ] ) into the target system ( [ eqn - gammalinear ] ) , we use the following transformation : @xmath97 where @xmath98 is a matrix of kernels . defining @xmath99
the original and target boundary conditions ( respectively ( [ eqn - bculinear ] ) and ( [ eqn - bc1beta ] ) ) can be written compactly ( omitting dependences in @xmath13 and @xmath100 ) as @xmath101 introducing ( [ eqn - tran ] ) into ( [ eqn - gammalinear ] ) , applying ( [ eqn - wlinear ] ) , integrating by parts and using the boundary conditions , we obtain that the original system ( [ eqn - wlinear ] ) is mapped into the target system ( [ eqn - gammalinear ] ) if and only if one has the following three matrix equations : @xmath102 expanding ( [ eqn - k1 ] ) , we get the following kernel equations : @xmath103 with boundary conditions obtained from ( [ eqn - k2])([eqn - k3 ] ) @xmath104 the equations evolve in the triangular domain @xmath105 .
notice that they can be written as two separate @xmath0 hyperbolic systems , one for @xmath106 and @xmath107 and another for @xmath108 and @xmath109 . by theorem [ th - wp ] ( see the appendix ) , one finds that , for @xmath50 , under the assumption that @xmath38 are @xmath39)$ ] , @xmath40 are @xmath41)$ ] and that @xmath42 , there is a unique solution to ( [ eqn - kuu])([eqn - bc4 ] ) , which is in @xmath110 . to study the invertibility of transformation ( [ eqn - tran ] ) , we look for a transformation of the the target system ( [ eqn - gammalinear ] ) into the original system ( [ eqn - wlinear ] ) as follows : @xmath111 where @xmath112 introducing ( [ eqn - traninv ] ) into ( [ eqn - wlinear ] ) , applying ( [ eqn - gammalinear ] ) , integrating by parts and using the boundary conditions , we obtain as before a set of kernel equations : @xmath113 with boundary conditions @xmath114 again by theorem [ th - wp ] ( see the appendix ) , one finds that there is a unique solution to these equations , which is @xmath110 . from the transformation ( [ eqn - tran ] ) evaluated at @xmath115 , one gets @xmath116 with control law ( [ eqn - linearcontrol ] ) the following result holds . [ th - control ] consider system ( [ eqn - wlinear ] ) with boundary conditions ( [ eqn - bculinear ] ) , control law ( [ eqn - linearcontrol ] ) , and initial condition @xmath117)$ ] .
then , for every @xmath56 , there exists @xmath57 such that @xmath118 in fact , the equilibrium @xmath119 is reached in finite time @xmath60 , where @xmath61 is given by ( [ eqn - tf ] ) .
since the transformation ( [ eqn - tran ] ) is invertible , when applying control law ( [ eqn - linearcontrol ] ) the dynamical behavior of ( [ eqn - wlinear ] ) is the same as the behavior of ( [ eqn - gammalinear ] ) , which is well - posed from standard results and whose explicit solution and stability properties we know from proposition [ pr - target ] .
thus , we obtain the explicit solutions of @xmath45 from the direct and inverse transformation , as follows : @xmath120 where @xmath121 is the explicit solution of the @xmath122 , @xmath123 system , given by ( [ eqn - expalpha])([eqn - expbeta ] ) , with initial conditions : @xmath124 in particular , we know that @xmath51 goes to zero in finite time @xmath60 , therefore @xmath45 also shares that property . finally , since the origin of the @xmath51 system is @xmath5 exponentially stable with an arbitrary large exponential decay rate , we conclude , using the inverse transformation , that the origin of the @xmath45 system is also @xmath5 exponentially stable with an arbitrary large exponential decay rate . equation ( [ eqn - l2normth ] ) follows by using the inverse and direct transformations to relate the @xmath5 norms of @xmath45 and @xmath51 ( using the fact that the kernels of the transformations are continuous , and thus bounded , functions ) . if the coefficient @xmath48 is zero in ( [ eqn - bculinear ] )
, the method presented in the paper is not valid since ( [ eqn - bc1 ] ) would require the value of one of the control kernels to be infinity at the boundary of the domain @xmath125 .
similarly , if the coefficient is close to zero one still gets very large values for the kernels close to the boundary , resulting in potentially large control laws . the method can be modified to accommodate zero or small values of @xmath48 by setting a slightly different target system ( [ eqn - alpha])([eqn - beta ] ) , as follows : @xmath126 where @xmath127 is to be obtained from the method ; regardless of the value of @xmath127 , this is a cascade system which is still @xmath5 exponentially stable and converges in finite time by the same arguments of proposition [ pr - target ] , since now , using the same lyapunov function @xmath68 defined in ( [ eqn - lyaplinear ] ) , we obtain @xmath69_0 ^ 1 \nonumber \\ & & + 2 \beta(0,t ) \int_0
^ 1\alpha(x , t ) a \frac{\mathrm{e}^{-\mu x}}{\epsilon_1(x ) } g(x ) dx , \label{eqn - udotlinearq0}\end{aligned}\ ] ] the new term ( which is the last one ) can be controlled by slightly modifying the coefficients of @xmath71 in the proof of proposition [ pr - target ] , obtaining the same result as before .
the kernel equations resulting from the transformation are still the same ( [ eqn - kuu])([eqn - kvv ] ) , with the same boundary conditions ( [ eqn - bc2])([eqn - bc4 ] ) for @xmath107 , @xmath108 , and @xmath109 ( which reduces to @xmath128 when @xmath49 ) , but one obtains an _ undetermined _ boundary conditions for @xmath106 : @xmath129 where @xmath130 can be chosen as desired ; by choosing at least a continuous function , one can apply theorem [ th - wp ] and thus the kernel equations are well - posed .
after @xmath130 has been chosen and the kernels have been computed , one obtains the value of @xmath127 as @xmath131 invertibility of the transformation follows as before , thus one obtains the same result of theorem [ th - control ] . the non - uniqueness in ( [ eqn - bc1q0 ] ) gives the designer some freedom in shaping the input function @xmath127 from @xmath123 to @xmath122 . also note
that this has no impact in the feedback law as the kernels @xmath108 and @xmath109 ( which are the ones appearing in ( [ eqn - linearcontrol ] ) ) are uniquely defined and independent of the non - unique @xmath106 and @xmath107 .
we wish to show that the linear controller ( [ eqn - linearcontrol ] ) designed using backstepping works _ locally _ for the nonlinear system , in terms that will be made precise . for that , we write our quasilinear system ( [ eqn - z ] ) in a form equivalent ( up to linear terms ) to ( [ eqn - wlinear ] ) . define @xmath132 we obtain a new state variable @xmath45 from @xmath12 using the following transformation : @xmath133=\left [ \begin{array}{cc } \varphi_1(x ) & 0\\ 0 & \varphi_2(x ) \end{array } \right ] \left [ \begin{array}{c } z_1(x , t)\\ z_2(x , t ) \end{array } \right]=\phi(x ) z(x , t),\label{eqn - zw}\ ] ] so that @xmath134 w(x , t)=\phi^{-1}(x ) w(x , t).\ ] ] it follows that @xmath45 verifies the following equation : @xmath135 where @xmath136 w.\end{aligned}\ ] ] it is evident that @xmath137 and that @xmath138 .
also , @xmath139.\label{eqn - nonlinearc}\ ] ] thus , it is possible to write ( [ eqn - w ] ) as a linear system with the same structure as ( [ eqn - wlinear ] ) plus nonlinear terms : @xmath140 where @xmath141 and @xmath142 computing the boundary conditions of ( [ eqn - w2 ] ) by combining ( [ eqn - bcz12 ] ) with the transformation ( [ eqn - zw ] ) , and defining @xmath143 and @xmath144 , one obtains @xmath145 where @xmath146 . in what follows
we will consider the case @xmath147 ; the case @xmath49 is analogous ( see remark [ rem - qzero ] ) .
notice that the linear parts of ( [ eqn - w2 ] ) and ( [ eqn - bcu2 ] ) are identical to ( [ eqn - wlinear ] ) and ( [ eqn - bculinear ] ) , and that the coefficients @xmath148 and @xmath70 verify the assumptions of section [ sect - linear ] . also , it is clear that the nonlinear terms verify @xmath149 , @xmath150 , and @xmath151 therefore , we consider using the feedback law : @xmath152 which implies , in terms of the original @xmath12 variable : @xmath153 where the kernels are computed from ( [ eqn - kuu])([eqn - bc4 ] ) using the coefficients @xmath148 and @xmath70 from ( [ eqn - nonlinearc ] ) and ( [ eqn - nonlinearsigma ] ) .
next , we show that the control law ( [ eqn - nonlinearcontrol2 ] ) , which is computed for the linear part of the system , asymptotically stabilizes the nonlinear system , although locally
. however , the right space to prove stability of the closed - loop system is @xmath1 , instead of the space @xmath5 that was used in section [ sect - linear ] for the linear system . denoting : @xmath154,g(z)=g_0(z_2),q_1=\left[\begin{array}{c } 0
\\ 1 \end{array } \right],k(x)= \left[\begin{array}{c } \frac{\varphi_1(x)k^{vu}(1,x)}{\varphi_2(1 ) } \\
\frac{\varphi_2(x)k^{vv}(1,x)}{\varphi_2(1 ) } \end{array } \right],\end{aligned}\ ] ] the boundary conditions of the closed loop system would be written as : @xmath155 a necessary condition for system ( [ eqn - z ] ) with boundary conditions ( [ eqn - bcz ] ) to be well - posed in the space @xmath1 is that the initial conditions verify the corresponding second - order compatibility condition .
these are @xmath156 while ( [ eqn - cc1 ] ) and ( [ eqn - cc3 ] ) are natural compatibility conditions , the conditions ( [ eqn - cc2 ] ) and ( [ eqn - cc4 ] ) are artificial ( since they show up due to the feedback law that has been designed ) and rather stringent , as they require very specific values of the initial conditions .
thus , we modify our control law in a way that , without losing its stabilizing character , does not require any specific values in the initial values beyond the natural conditions ( [ eqn - cc1 ] ) and ( [ eqn - cc3 ] ) .
the modification in the boundary conditions consists in adding a dynamic extension to the controller as follows : @xmath157 where @xmath158 is one of the states of the following system : @xmath159 where the constants @xmath160 and @xmath161 can be chosen as desired with the only conditions that @xmath162 and @xmath163 .
it is evident that with positive values of these constants , ( [ syst - a - b ] ) is always stable .
the initial conditions of @xmath158 and @xmath164 are an additional degree of freedom that can be used to eliminate the compatibility conditions ( [ eqn - cc2 ] ) and ( [ eqn - cc4 ] ) . with the modification of the control law , these compatibility conditions are now @xmath165 call @xmath166 selecting @xmath167 the compatibility conditions are automatically verified . we are now ready to state our main result .
define the norms @xmath168 and @xmath169 .
[ thm - main ] consider system ( [ eqn - z ] ) and ( [ syst - a - b ] ) with boundary conditions ( [ eqn - bcz2 ] ) and initial conditions
@xmath170^t \in h^2([0,1])$ ] , and @xmath171 and @xmath172 verifying ( [ eqn - b0 ] ) , with the kernels @xmath108 and @xmath109 obtained from ( [ eqn - kuu])([eqn - bc4 ] ) where the coefficients @xmath148 and @xmath70 are computed from ( [ eqn - nonlinearc ] ) and ( [ eqn - nonlinearsigma ] ) .
then , under the assumptions of smoothness for the coefficients stated in section [ sect - prob ] , for every @xmath56 , there exist @xmath173 and @xmath57 such that such that , if @xmath174 and if the compatibility conditions ( [ eqn - cc1 ] ) and ( [ eqn - cc3 ] ) are verified , then : @xmath175
we first establish some definitions and notation .
for @xmath176 with components @xmath177 and @xmath178 denote @xmath179 , and @xmath180 } \vert \gamma(x )
\vert,\quad \vert \gamma \vert_{l^1}= \int_0 ^ 1 \vert \gamma(\xi ) \vert d\xi.\ ] ] in what follows , for a time - varying vector @xmath181 , we denote @xmath182 and @xmath183 to simplify our notation . for a @xmath184 matrix @xmath185 ,
denote : @xmath186 for the kernel matrices @xmath187 and @xmath188 denote @xmath189 for @xmath190)$ ] , recall the following well - known inequalities , that will be used later : @xmath191\leq c_4 \vert \gamma \vert_{h^1},\label{eqn - sobolev1}\\ \vert \gamma_x \vert_{\infty } & \leq & c_5\left [ \vert \gamma_x \vert_{l^2}+\vert \gamma_{xx } \vert_{l^2}\right]\leq c_6 \vert \gamma \vert_{h^2}.\quad\,\,\label{eqn - sobolev2}\end{aligned}\ ] ] define the following linear functionals , the first two of which are , respectively , the inverse and direct transformations ( [ eqn - tran ] ) and ( [ eqn - traninv ] ) : @xmath192(x ) & = & \gamma(x , t)-\int_0^x k(x,\xi ) \gamma(\xi , t ) d\xi,\\ \mathcal l[\gamma](x ) \label{eqn - call } & = & \gamma(x , t)+\int_0^x l(x,\xi ) \gamma(\xi , t ) d\xi,\\
\mathcal k_{1}[\gamma](x ) & = & -k(x , x)\gamma(x , t)+\int_0^x k_{\xi}(x,\xi ) \gamma(\xi , t ) d\xi,\qquad\\
\mathcal k_{2}[\gamma](x ) & = & -k(x , x)\gamma(x , t)-\int_0^x k_{x}(x,\xi ) \gamma(\xi , t ) d\xi,\qquad\\
\mathcal l_{1}[\gamma](x ) & = & l(x , x)\gamma(x , t)+\int_0^x l_x(x,\xi ) \gamma(\xi , t ) d\xi,\qquad\\
\mathcal l_{11}[\gamma](x ) & = & ( l_x(x , x)+l_\xi(x , x))\gamma(x , t)+l_x(x , x)\gamma(x , t ) \nonumber \\ & & + \int_0^x l_{xx}(x,\xi ) \gamma(\xi , t ) d\xi.\label{eqn - callx}\end{aligned}\ ] ] for simplicity , in what follows we drop writing the @xmath13 dependence in functionals and the @xmath100 dependence in the variables . using ( [ eqn - calk ] ) and
( [ eqn - call ] ) , we define @xmath193 $ ] and @xmath194 $ ] as : @xmath195,x\right),\quad f_2=f_{nl}\left(\mathcal l[\gamma],x\right).\end{aligned}\ ] ] to prove theorem [ thm - main ] , we notice that if we apply the ( invertible ) backstepping transformation ( [ eqn - tran ] ) to the nonlinear system ( [ eqn - w2 ] ) we obtain the following transformed system : @xmath196 and using the inverse transformation ( [ eqn - traninv ] ) the equation can be expressed fully in terms of @xmath51 as : @xmath197+f_4[\gamma]=0,\end{aligned}\ ] ] where the functionals @xmath198 and @xmath199 are defined as @xmath200\gamma_x\right ] , \label{eqn - f3 } \\ f_4&=&\mathcal k\left [ f_1[\gamma ] \mathcal l_{1 } \left[\gamma \right ] + f_2[\gamma]\right].\label{eqn - f4}\qquad\,\end{aligned}\ ] ] the boundary conditions are @xmath201 by the assumptions on the coefficients and applying theorem [ th - sm ] , the direct and inverse transformations ( [ eqn - tran ] ) and ( [ eqn - traninv ] ) have kernels that are @xmath202 functions . differentiating twice with respect to @xmath13 in these transformations , it can be shown that the @xmath1 norm of @xmath51 is equivalent to the @xmath1 norm of @xmath12 ( see for instance @xcite ) . thus , if we show @xmath1 local stability of the origin for ( [ eqn - gammanl2])([eqn - gammabc ] ) , the same holds for @xmath12 .
we proceed by analyzing ( using a lyapunov function ) the growth of @xmath203 , @xmath204 and @xmath205 . relating these norms with @xmath206
, we then prove @xmath1 local stability for @xmath51 .
define @xmath207 for @xmath71 as in ( [ eqn - ddef ] ) .
proceeding analogously to ( [ eqn - udotlinear1])([eqn - udotlinear2 ] ) , we get some extra nonlinear terms : @xmath208_0 ^ 1 \nonumber \\ & & - 2 \int_0 ^ 1\gamma^t(x , t ) d(x)\left(f_3[\gamma,\gamma_x]+ f_4[\gamma ] \right ) dx .\end{aligned}\ ] ] let us analyze first the last term : @xmath209 + f_4[\gamma ] \right ) dx \right|
\leq k_1 \int_0 ^ 1 \vert \gamma \vert \left ( \vert f_3[\gamma,\gamma_x]\vert + \vert f_4[\gamma ] \vert\right ) dx.\ ] ] applying lemma [ lem - f3f4bound ] ( see the appendix ) , we obtain that there exists a @xmath210 , such that for @xmath211 , @xmath212 \vert dx & \leq & k_2 \vert \gamma_x \vert_{\infty}\vert \gamma \vert^2_{l^2},\quad\\ \int_0 ^ 1 \vert \gamma \vert \vert f_4[\gamma ] \vert dx & \leq & k_3 \vert \gamma \vert_{\infty } \vert \gamma \vert_{l^2}^2,\end{aligned}\ ] ] and using inequality ( [ eqn - sobolev1 ] ) and noting that @xmath213 , we obtain @xmath212 \vert dx & \leq & k_5 \vert \gamma_x \vert_{\infty}v_1,\quad\\ \int_0 ^ 1 \vert \gamma \vert \vert f_4[\gamma ] \vert dx & \leq & k_6 \vert \gamma_x \vert_{\infty}v_1+k_7 v_1^{3/2}.\end{aligned}\ ] ] now , @xmath73_0 ^ 1 & = & b a^2(t)\mathrm{e}^{\mu } -a \alpha^2(1,t)\mathrm{e}^{-\mu } -b \beta^2(0,t ) \nonumber \\ & & + a(q\beta(0,t)+g_{nl}(\beta(0,t)))^2,\end{aligned}\ ] ] and for @xmath211 , @xmath214 , and @xmath215 , we obtain @xmath73_0 ^ 1 & \leq & -a \alpha^2(1,t)\mathrm{e}^{-\mu } + ( a(\vert q\vert + k_8)^2-b ) \beta^2(0,t ) \nonumber \\ & & + b a^2(t)\mathrm{e}^{\mu}\end{aligned}\ ] ] thus , choosing @xmath216 and @xmath217 and @xmath218 as in the proof of proposition [ pr - target ] , we obtain the following proposition : [ pr - gamma ] there exists @xmath210 such that if @xmath211 then @xmath219 where @xmath220 , @xmath221 , @xmath222 , @xmath223 and @xmath224 are positive constants .
define @xmath225 .
notice that the norms of @xmath226 and @xmath227 are related ( see lemma [ lem - gammaetaequiv ] in the appendix ) . taking a partial derivative in @xmath100 in ( [ eqn - gammanl ] )
we obtain an equation for @xmath226 as follows : @xmath228-\sigma(x)\right ) \eta_x + f_5[\gamma,\gamma_x,\eta]+f_6[\gamma,\eta]=0,\end{aligned}\ ] ] where @xmath229 and @xmath230 are defined as @xmath231\eta \right ] + \int_0^x k(x,\xi ) f_{12}[\gamma,\gamma_x]\eta(\xi ) d\xi + k(x,0 ) \lambda_{nl}\left(\gamma(0),0\right ) \eta(0 ) \nonumber \\ & & + \mathcal k \left [ f_{11}[\gamma,\eta]\gamma_x\right ] , \label{eqn - f5 } \\
f_6&=&\mathcal k\left[f_{11}[\gamma,\eta ] \mathcal l_x \left[\gamma\right ] \right ] + \mathcal k\left[f_{1}[\gamma]\mathcal l_x \left[\eta\right]\right ] + \mathcal k[f_{21 } [ \gamma,\eta]],\label{eqn - f6}\end{aligned}\ ] ] where @xmath232,x\right)\mathcal l[\eta],\quad\,\,\ , \\
f_{12}&= & \frac{\partial \lambda_{nl}}{\partial \gamma } \left(\mathcal l[\gamma],x\right ) \left(\gamma_x+
\mathcal l_x [ \gamma ] \right ) + \frac{\partial \lambda_{nl}}{\partial x } \left(\mathcal l[\gamma],x\right ) , \quad\,\,\ , \\ f_{21}&=&\frac{\partial f_{nl}}{\partial \gamma } \left(\mathcal l[\gamma],x\right)l[\eta].\quad\end{aligned}\ ] ] the boundary conditions for @xmath233^t$ ] are @xmath234 to find a lyapunov function for @xmath226 , we use the next lemma : [ lem - lyapeta ] there exists @xmath173 such that , for @xmath235 , there exists a symmetric matrix @xmath236>0 $ ] verifying the identity : @xmath237 \left(\sigma(x)-f_1[\gamma]\right)- \left(\sigma(x)-f_1[\gamma]\right)^t r[\gamma]=0,\label{eqn - symm}\ ] ] and the following bounds : @xmath238(x)&\leq&c_1+c_2 \vert \gamma \vert_{\infty},\label{eqn - rpos2}\\ \vert \left ( \left(r[\gamma]-d(x)\right ) \sigma(x)\right)_x \vert&\leq & c_2 \vert \gamma \vert_{\infty } \left(1 + \vert \gamma _ x \vert_\infty \right),\quad\label{eqn - thetax}\\ \vert \left ( r[\gamma ] \right)_t \vert & \leq & c_3 \left(\vert \eta\vert+ \vert \eta \vert_{l^1 } \right),\label{eqn - thetat}\end{aligned}\ ] ] where @xmath239 are positive constants .
we explicitly construct @xmath236 $ ] as @xmath240=d(x)+\theta[\gamma],\ ] ] with @xmath241=\left [ \begin{array}{cc } 0 & \psi[\gamma ] \\
\psi[\gamma ] & 0 \end{array } \right],\label{eqn - thetadef}\ ] ] where @xmath242 $ ] is defined as : @xmath243=\frac{d_{11}(x)\left(f_1[\gamma]\right)_{12}-d_{22}(x)\left(f_1[\gamma ] \right)_{21}}{\epsilon_2(x)+\epsilon_1(x)+\left(f_1[\gamma]\right)_{11}-\left(f_1[\gamma]\right)_{22}},\label{eqn - psidef}\ ] ] where @xmath244\right)_{ij}$ ] denotes the coefficient in row @xmath245 and column @xmath246 in the matrix @xmath193 $ ] .
identity ( [ eqn - symm ] ) follows by using the construction of @xmath236(x)$ ] in ( [ eqn - rdef])([eqn - psidef ] ) , and the fact that @xmath71 and @xmath70 are diagonal and commute .
to ensure that the denominator of ( [ eqn - psidef ] ) is different from zero , denote @xmath247 } \left(\epsilon_1(x)+\epsilon_2(x)\right)>0 $ ] . applying ( [ eqn - f1bound ] ) from lemma [ lem - f3f4bound ] ,
there exists @xmath210 for which , if @xmath248 , one gets : @xmath249\right)_{22}-\left(f_1[\gamma]\right)_{11 } \geq k_1 - k_2 \vert \gamma \vert_{\infty},\ ] ] thus if @xmath250 with @xmath251 , we obtain @xmath249\right)_{22}-\left(f_1[\gamma]\right)_{11 } \geq \frac{k_1}{2},\label{dem - bound}\ ] ] thus @xmath242 $ ] is well - defined .
applying again ( [ eqn - f1bound ] ) in the numerator of ( [ eqn - psidef ] ) to bound ( [ eqn - thetadef ] ) , we obtain : @xmath252 \vert \leq k_3 \vert \gamma \vert_{\infty},\label{eqn - thetabound}\ ] ] and noting @xmath253 , we obtain directly the bound ( [ eqn - rpos2 ] ) , and by choosing @xmath254 , with @xmath255 , we show @xmath236>0 $ ] . inequality ( [ eqn - thetax ] ) is equivalent to showing : @xmath256 \sigma(x)\right)_x \vert \leq c_2 \vert \gamma \vert_{\infty } \left(1 + \vert \gamma _ x \vert_\infty \right).\ ] ] we first use ( [ eqn - thetabound ] ) to bound @xmath257(x ) \sigma_x(x ) \vert$ ] , and for @xmath258(x ) \sigma(x ) \vert$ ] we take a derivative in @xmath13 in ( [ eqn - psidef ] ) , use the bound ( [ dem - bound ] ) and use the fact that @xmath259= f_{12}[\gamma,\gamma_x]$ ] and using lemma [ lem - f5f6bound ] , there exists @xmath260 such that if @xmath261 , @xmath262 \vert & \leq & k_1 \left ( \vert \gamma \vert_{\infty}+\vert \gamma_x \vert_{\infty}\right).\end{aligned}\ ] ] to show ( [ eqn - thetat ] ) we use @xmath263= \vert f_{11}[\gamma,\eta]$ ] and apply lemma [ lem - f5f6bound ] . setting @xmath264
the lemma follows .
define : @xmath265(x ) \eta(x , t)dx.\ ] ] computing @xmath266 , applying lemma [ lem - lyapeta ] , and integrating by parts , we find @xmath267\left(\sigma(x ) -f_1[\gamma ] \right)\right)_x \eta(x , t)dx \nonumber \\ & & + \left[\eta^t(x , t)r[\gamma](x)\left(\sigma(x ) -f_1[\gamma](x ) \right ) \eta(x , t ) \right]_{x=0}^{x=1 } + \int_0 ^ 1\eta^t(x , t ) \left(r[\gamma]\right)_t \eta(x , t)dx \nonumber \\ & & -2\int_0 ^ 1\eta^t(x , t ) r[\gamma]f_5[\gamma,\gamma_x,\eta,\eta_x,]dx -2\int_0 ^ 1\eta^t(x , t ) r[\gamma]f_6[\gamma,\eta ] dx.\label{eqn - vdot}\end{aligned}\ ] ] the first three terms of ( [ eqn - vdot ] ) are analyzed using lemma [ lem - lyapeta ] .
thus , there exists @xmath210 such that , for @xmath235 , we find , for the first term : @xmath268\left(\sigma(x ) -f_1[\gamma ] \right)\right)_x \eta(x , t)dx \nonumber \\ & \leq & -\lambda_1 v_2 + k_1 \vert \eta \vert_{l^2}^2\left ( \vert \gamma \vert_{\infty } + \vert \gamma _ x \vert_\infty\right ) .
\quad \,\,\label{eqn - vdott1}\end{aligned}\ ] ] the second term of ( [ eqn - vdot ] ) is bounded using the boundary conditions , ( [ eqn - udotlinear1])([eqn - udotlinear2 ] ) , and lemma [ lem - lyapeta ] , as : @xmath269(x)\left(\sigma(x ) -f_1[\gamma](x ) \right ) \eta(x , t ) \right]_{x=0}^{x=1 } \nonumber \\ & \leq & -\lambda_2 \left ( \eta_1 ^ 2(1,t)+ \eta_2 ^ 2(0,t ) \right ) + k_2 \vert \gamma \vert_{\infty } \left ( \eta_2 ^ 2(0,t ) + \eta_1 ^ 2(1,t)\right ) \nonumber \\ & & + k_3(1 + \vert \gamma \vert_{\infty})(a(t)^2+b(t)^2).\end{aligned}\ ] ] finally , we bound the third term of ( [ eqn - vdot ] ) applying lemma [ lem - lyapeta ] as follows : @xmath270\right)_t \eta(x , t)dx & \leq & k_3\int_0 ^ 1 \vert \eta\vert^2 \left(\vert \eta\vert+ \vert \eta \vert_{l^1 } \right ) dx \leq k_4\vert \eta \vert^2_{l^2 } \vert \eta \vert_{\infty}.\qquad\label{eqn - vdott2}\end{aligned}\ ] ] applying lemmas [ lem - lyapeta ] and [ lem - f5f6bound ] to the last terms of ( [ eqn - vdot ] ) , we get , for @xmath271 , @xmath272f_5[\gamma,\gamma_x,\eta,\eta_x]dx \right| \leq
k_5 \int_0 ^ 1 \vert \eta \vert \vert f_5[\gamma,\eta ] \vert dx \nonumber \\ & \leq & k_6 \vert \eta \vert^2_{l^2 } \left ( \vert \gamma \vert_{\infty}+\vert \gamma_x \vert_{\infty}\right ) + k_{7 } \vert \eta \vert_{l^2}\vert \eta(0,t)\vert \vert \gamma(0,t ) \vert , \end{aligned}\ ] ] and @xmath272f_6[\gamma,\eta]dx \right| \leq k_{8 } \int_0 ^ 1 \vert \eta \vert \vert f_6[\gamma,\eta ] \vert dx \leq k_{9 } \vert \eta \vert^2_{l^2 } \vert \gamma \vert_{\infty}.\end{aligned}\ ] ] thus , it is clear that by choosing @xmath273 small enough , using lemma [ lem - gammaetaequiv ] to bound @xmath274 by @xmath275 , and noting @xmath276 , we obtain the following proposition : [ pr - gammat ] there exists @xmath277 such that if @xmath278 @xmath279 for @xmath280 positive constants .
define @xmath281 . notice that the norms of @xmath282 and @xmath283 are related ( see lemma [ lem - thetaetaequiv ] in the appendix ) . taking a partial derivative in @xmath100 in ( [ eqn - etanl2 ] )
we obtain an equation for @xmath282 : @xmath284-\sigma(x)\right ) \theta_x + f_7[\gamma,\gamma_x,\eta,\eta_x,\theta]+f_8[\gamma,\eta,\theta]=0,\label{eqn - thetanl2}\end{aligned}\ ] ] where @xmath285 and @xmath286 are defined as @xmath287\eta \right]+ \int_0^x k(x,\xi ) f_{12}[\gamma,\gamma_x]\theta(\xi ) d\xi + \mathcal k_{1 } \left [ f_1[\gamma]\theta \right ] \nonumber \\ & & + \int_0^x k(x,\xi ) f_{14}[\gamma,\gamma_x,\eta,\eta_x]\eta(\xi ) d\xi + k(x,0)\frac{\partial \lambda_{nl}}{\partial \gamma}\left(\gamma(0),0\right)\eta(0 ) \eta(0 ) \nonumber \\ & & + k(x,0 ) \lambda_{nl}\left(\gamma(0),0\right ) \theta(0 ) + \mathcal k \left [ f_{11}[\gamma,\eta]\eta_x\right ] + \mathcal k \left [ f_{13}[\gamma,\eta,\theta]\gamma_x\right ] , \label{eqn - f7 } \\ f_8&=&2\mathcal k\left[f_{11}[\gamma,\eta ] \mathcal l_x \left[\eta\right ] \right ] + \mathcal k\left[f_{1}[\gamma]\mathcal l_x \left[\theta\right]\right ] + \mathcal k\left[f_{13}[\gamma,\eta,\theta ] \mathcal l_x \left[\gamma\right ] \right ] \nonumber \\ & & + \mathcal k[f_{22 } [ \gamma,\eta,\theta]],\label{eqn - f8}\end{aligned}\ ] ] where @xmath288,x\right ) \mathcal l[\eta ] \mathcal l[\eta ] + \frac{\partial \lambda_{nl}}{\partial \gamma } \left(\mathcal l[\gamma],x\right ) \mathcal l[\theta],\\ f_{14}&= & \frac{\partial^2 \lambda_{nl}}{\partial \gamma^2 } \left(\mathcal l[\gamma],x\right ) \mathcal l[\eta ] \left(\gamma_x+\mathcal l_{1}[\gamma]\right ) + \frac{\partial \lambda_{nl}}{\partial \gamma } \left(\mathcal l[\gamma ] , x\right ) \left(\eta_x+ \mathcal l_{1}[\eta]\right ) \nonumber \\ & & + \frac{\partial^2 \lambda_{nl}}{\partial x \partial \gamma } \left(\mathcal l[\gamma],x\right ) \mathcal l[\eta ] , \quad\,\,\ , \\ f_{22}&=&\frac{\partial^2 f_{nl}}{\partial \gamma^2 } \left(\mathcal l[\gamma],x\right)\mathcal l[\eta]\mathcal l[\eta ] + \frac{\partial f_{nl}}{\partial \gamma } \left(\mathcal l[\gamma],x\right)\mathcal l[\theta].\quad\end{aligned}\ ] ] the boundary conditions for @xmath289^t$ ] are @xmath290 since ( [ eqn - thetanl2 ] ) has the same structure as ( [ eqn - etanl2 ] ) , we define : @xmath291(x ) \theta(x , t)dx,\ ] ] where @xmath236(x)$ ] was defined in lemma [ lem - lyapeta ] .
computing @xmath292 , and proceeding exactly as in ( [ eqn - vdot ] ) , we find : @xmath293\left(\sigma(x ) -f_1[\gamma ] \right)\right)_x \theta(x , t)dx \nonumber \\ & & + \left[\theta^t(x , t)r[\gamma](x)\left(\sigma(x ) -f_1[\gamma](x ) \right ) \theta(x , t)\right]_{x=0}^{x=1 } + \int_0 ^ 1\theta^t(x , t ) \left(r[\gamma]\right)_t \theta(x , t)dx \nonumber \\ & & -2 \int_0 ^ 1\theta^t(x , t ) r[\gamma]f_7[\gamma,\gamma_x,\eta,\eta_x,\theta]dx -2 \int_0 ^ 1\theta^t(x , t ) r[\gamma]f_6[\gamma,\eta,\theta]dx.\qquad \label{eqn - wdot}\end{aligned}\ ] ] the first three terms of ( [ eqn - wdot ] ) are analyzed as in ( [ eqn - vdott1])([eqn - vdott2 ] ) : @xmath294f_6[\gamma,\eta,\theta]dx\right| + 2 \left|\int_0 ^ 1\theta^t(x , t ) r[\gamma]f_7[\gamma,\gamma_x,\eta,\eta_x,\theta]dx\right| \nonumber \\ & & + k_3\vert \theta \vert^2_{l^2 } \vert \eta \vert_{\infty } + k_4(\eta^4_2(0,t)+(1+\vert \gamma \vert_{\infty})(a^2(t)+b^2(t ) ) ) .\qquad\label{eqn - wdot2}\end{aligned}\ ] ] finally , applying lemmas [ lem - lyapeta ] and [ lem - f7f8bound ] in the last two terms of ( [ eqn - wdot2 ] ) , there exists a @xmath295 , such that for @xmath271 , @xmath296f_7[\gamma,\gamma_x,\eta,\eta_x,\theta]dx \right| \leq k_5 \int_0 ^ 1 \vert \theta \vert \vert f_7[\gamma,\gamma_x,\eta,\eta_x,\theta ] \vert dx \nonumber \\ & \leq & k_6 \vert \theta \vert_{l^2}^2 \left ( \vert \gamma \vert_{\infty}+\vert \gamma_x \vert_{\infty}\right)+ k_7\vert \theta \vert_{l^2 } \vert \eta \vert_{l^2}^2 + k_8\vert \theta \vert_{l^2 } \vert \eta_x \vert_{l^2}\vert \eta \vert_{\infty } + k_9 \vert \theta \vert_{l^2 } \nonumber \\ & & + k_{10 } \vert \theta \vert_{l^2 } \left ( \vert \eta \vert_{l^2}\vert \eta \vert_{\infty}^2+\vert\eta(0,t)\vert^2 + \vert \gamma(0,t)\vert \vert \theta(0,t ) \vert\right),\qquad\end{aligned}\ ] ] and @xmath297f_8[\gamma,\eta,\theta]dx \right| \leq k_{11 } \int_0 ^ 1 \vert \theta \vert \vert f_8[\gamma,\eta,\theta ] \vert dx \nonumber \\ & \leq & k_{11 } \vert \theta \vert^2_{l^2 } \vert \gamma \vert_{\infty}+k_{12 } \vert \eta \vert_{l^2 } \vert \theta \vert_{l^2 } \vert \eta \vert_{\infty } + k_{12 } \vert \eta \vert_{l^2 } \vert \theta \vert_{l^2}^2 + k_{13 } \vert \eta \vert_{l^2}^2 \vert \theta \vert_{l^2}.\qquad\,\,\end{aligned}\ ] ] thus , by choosing @xmath273 and @xmath298 small enough to apply lemma [ lem - thetaetaequiv ] , we finally obtain the following proposition : [ pr - gammatt ] there exists @xmath299 such that if @xmath300 then @xmath301 where @xmath302 are positive constants . defining @xmath303 , and combining propositions [ pr - gamma ] , [ pr - gammat ] , and [ pr - gammatt ] ,
there exists @xmath295 such that if @xmath304 @xmath305 for @xmath306 . to compensate the last term
, we augment this lyapunov function and define @xmath307 .
then , @xmath308 and choosing @xmath309 , one obtains @xmath310 for some positive @xmath221 . following @xcite and noting @xmath311
, then for sufficiently small @xmath312 , it follows that @xmath313 exponentially .
given that @xmath314 ( by proposition [ prop - equiv ] ) is equivalent to the @xmath1 norm of @xmath51 when @xmath315 is sufficiently small , and since by construction @xmath316 verifies the required second - order compatibility conditions , there exists @xmath173 and @xmath57 such that if @xmath317 , then : @xmath318 since , as we argued , for small enough @xmath319 the @xmath1 norms of @xmath12 and @xmath51 are equivalent , this proves theorem [ thm - main ] . _
[ rem - qzero ] the proof has been carried out for the case @xmath50 .
if @xmath49 , we have to modify the target system following section [ sect - q0 ] and this implies the appearance of a linear boundary term ( a coefficient times @xmath320 ) in the @xmath51 system ; similarly , in the @xmath226 and @xmath282 systems , @xmath321 and @xmath322 terms will appear .
these terms can be controlled using the same lyapunov function by following the strategy outlined in section [ sect - q0 ] _
we have solved the problem of full - state boundary stabilization for a @xmath0 system of first - order hyperbolic quasilinear pdes with actuation on only one boundary .
we have shown , using a strict lyapunov function , @xmath1 local exponential stability of the state .
it is possible to extend this result to design an observer , as shown in @xcite , and combining both results one obtains an output - feedback controller with similar properties ( see @xcite )
. it would be of interest to extend the method to @xmath3 systems .
for instance , a @xmath323 first - order hyperbolic system of interest is the saint - venant - exner system , which models open channels with a moving sediment bed @xcite ; the extension is shown ( for the linear case ) in @xcite . while extending the lyapunov analysis to @xmath324 systems has been done @xcite , considerable extra effort is required to extend backstepping to a general @xmath324 system , even in the linear case . in general
, the method needs @xmath325 kernels resulting in a @xmath326 system of coupled first - order hyperbolic equations , whose well - posedness depends critically on the exact choice of the transformation and target system .
the extension has been shown possible , for the linear case , if the system has @xmath327 positive and one negative transport speeds , with actuation only on the state corresponding to the negative velocity @xcite .
we show well - posedness of the following hyperbolic @xmath2 system , which is generic enough to contain all the kernel equation systems that appear in the paper : @xmath328 evolving in the domain @xmath329 , with boundary conditions : @xmath330 this type of system has been called `` generalized goursat problem '' by some authors @xcite .
however the boundaries of the domain @xmath125 are characteristic for ( [ eqn - hypf1 ] ) and ( [ eqn - hypf4 ] ) , thus the general results derived in @xcite can not be applied .
the following theorems discusses existence , uniqueness and smoothness of solutions to the equations .
[ th - wp ] consider the hyperbolic system ( [ eqn - hypf1])([eqn - bcf4 ] ) . under the assumptions @xmath331),\,g_i , c_{ji}\in \mathcal c(\mathcal t),\,\,i , j=1,2,3,4 $ ] and @xmath332)$ ] with @xmath42
, there exists a unique @xmath110 solution @xmath333 , @xmath334 .
[ th - sm ] consider the hyperbolic system ( [ eqn - hypf1])([eqn - bcf4 ] ) . under the assumptions of theorem
[ th - wp ] , and the additional assumptions @xmath335),\,g_i , c_{ji}\in \mathcal c^n(\mathcal t)$ ] , there exists a unique @xmath336 solution @xmath333 , @xmath334 .
next we prove the theorems ; the proof is based on transforming the equations into integral equations and then solving them using a successive approximation method . the equations can be transformed into integral equations by the method of characteristics .
for that , it is necessary to define : @xmath337 and @xmath338 . note that all the @xmath339 functions are monotonically increasing and thus invertible , due to positivity of the @xmath340 coefficients . under the assumptions of theorem
[ th - wp ] , it also holds that @xmath341)$ ] .
define , for @xmath342 , the characteristic lines along which ( [ eqn - hypf1])([eqn - hypf4 ] ) evolve : @xmath343 where the argument @xmath344 that parameterizes @xmath345 and @xmath346 belongs to the interval @xmath347 $ ] , with @xmath348 defined as @xmath349 the following holds if @xmath342 and @xmath350 $ ] , it holds that @xmath351 , for @xmath352
. also , under the assumptions of theorem [ th - wp ] , @xmath345 , @xmath346 , and @xmath348 are continuous in their domains of definition since they are defined as compositions of continuous functions .
moreover , the following inequalities are verified @xmath353 using these definitions , ( [ eqn - hypf1])([eqn - hypf4 ] ) are integrated to : @xmath354(x,\xi),\end{aligned}\ ] ] where we have denoted @xmath355,\quad g_j(x,\xi)=\int_0^{s^f_j(x,\xi ) } g_j\left(x_j(x,\xi , s),\xi_j(x,\xi , s)\right ) ds , \\
i_j[f](x,\xi)&= & \sum_{i=1}^4 \int_0^{s^f_j(x,\xi ) } \hspace{-10pt } c_{ji}\left(x_j(x,\xi , s),\xi_j(x,\xi , s)\right ) f^i\left(x_j(x,\xi , s),\xi_j(x,\xi , s)\right ) ds,\label{eqn - integrat } \qquad\end{aligned}\ ] ] for @xmath356 . substituting the boundary conditions ( [ eqn - bcf1])([eqn - bcf4 ] ) and expressing the terms in ( [ eqn - bcf1 ] ) and ( [ eqn - bcf4 ] ) containing @xmath357 and @xmath358 in terms of the solution ( [ eqn - integrat ] ) we get four integral equations which have the following structure : @xmath359(x,\xi)+i_j[f](x,\xi),\qquad\end{aligned}\ ] ] where @xmath360 , @xmath361 is has the values @xmath362 and @xmath363 and the values of the @xmath364(x,\xi)$ ] are @xmath365 and @xmath366&=&q_1(x_1(x,\xi,0))i_2[f](x_1(x,\xi,0),0 ) + q_2(x_1(x,\xi,0))i_3[f](x_1(x,\xi,0),0),\qquad \,\,\\
q_4[f]&=&q_3(x_4(x,\xi,0))i_2[f](x_4(x,\xi,0),0 ) + q_4(x_4(x,\xi,0))i_3[f](x_4(x,\xi,0),0).\qquad \,\,\end{aligned}\ ] ] in this form , the equations are amenable to be solved using the successive approximation method .
this is explained next .
the successive approximation method can be used to solve the integral equations .
define first the following functional acting on @xmath367 : @xmath368(x,\xi)=q_j[f](x,\xi)+i_j[f](x,\xi),\ ] ] and the vectors : @xmath369,\ , \phi[f]=\left[\begin{array}{c}\phi_1[f]\\\phi_2[f]\\\phi_3[f]\\\phi_4[f]\end{array}\right].\,\ ] ] define then @xmath370(x,\xi).\quad \label{eqn - fn}\end{aligned}\ ] ] finally define for @xmath371 the increment @xmath372 , with @xmath373 by definition .
it is easy to see that , since @xmath374 is a linear functional , the equation @xmath375(x,\xi)$ ] holds .
if @xmath376 exists , then @xmath377 is a solution of the integral equations ( and thus solves the original hyperbolic system ) . using the definition of @xmath378
, it follows that if @xmath379 converges , then @xmath380 first , define : @xmath381 } \left| q_i(x ) \right| , \bar c= \left(1+\sum_{i=1}^4\bar q_{i } \right)\left(\sum_{j=1}^4\sum_{i=1}^4\bar c_{ji } \right).\end{aligned}\ ] ] next , we prove the following two lemmas : [ lem - int ] for @xmath334 , @xmath371 , @xmath342 , and @xmath382 , @xmath383 defined as in ( [ eqn - x1])([eqn - s4f ] ) , it follows that @xmath384 we show the result for @xmath385 .
it follows for @xmath386 by switching @xmath387 and @xmath388 , respectively , for @xmath389 and @xmath390 . for @xmath391 we can write : @xmath392^n ds.\ ] ] to prove the inequality , change the variable of integration to @xmath393 .
then , taking into account @xmath394=\frac{1}{\phi_1'(z)}=\epsilon_1(z),\ ] ] the integral can be bounded as follows : @xmath395^n ds \nonumber \\ & = & \int_{\phi_1^{-1}\left(\phi_1(x)-\phi_1(\xi)\right)}^{x } z^n/ \epsilon_1(z ) dz \leq k_{\epsilon } \int_{0}^{x } z^n dz = k_{\epsilon } \frac{x^{n+1}}{n+1}.\end{aligned}\ ] ] for @xmath396 the integral can be written as : @xmath397^n ds . \ ] ] as before , change the variable of integration to @xmath398 .
then one has that @xmath399 , thus the integral can be bounded as follows : @xmath400^n ds \nonumber \\
& = & \int_{\phi_3^{-1}\left(\phi_1(x)+\phi_2(\xi)\right)}^{x } z^n/ \epsilon_1(z ) dz \leq k_{\epsilon } \int_{0}^{x } z^n dz = k_{\epsilon } \frac{x^{n+1}}{n+1},\qquad\,\end{aligned}\ ] ] which concludes the proof .
[ lemma - bound ] for @xmath334 , @xmath371 and @xmath342 , assume that @xmath401 then it follows that @xmath402 .
we show it for @xmath385 ; the structure of the equations is the same for @xmath386 . for @xmath396 :
i_2[\delta f^n](x,\xi)\right| \leq \sum_{i=1}^4 \bar c_{2i } \int_0^{s^f_2(x,\xi ) } \left| \delta f^n_i\left(x_2(x,\xi , s),\xi_2(x,\xi , s ) \right ) ds \right| \nonumber \\ & \leq & \bar \phi \frac{k_{\epsilon}^n \bar c^{n}}{n ! } \sum_{i=1}^4 \bar c_{2i } \int_0^{s^f_2(x,\xi ) } x_2^n(x,\xi , s ) ds \leq \bar \phi \frac{k_{\epsilon}^{n+1 } \bar c^{n+1 } x^{n+1 } } { ( n+1 ) ! } , \end{aligned}\ ] ] where lemma [ lem - int ] has been applied .
similarly , for @xmath391 : @xmath404(x,\xi ) & \leq & \left| q_1[\delta f^n](x,\xi)\right|+\left| i_1[\delta f^n](x,\xi)\right| \nonumber \\ &
\leq & \bar q_1 \bar \phi \frac{\bar c^n k_{\epsilon}^n}{n ! } \sum_{i=1}^4 \bar c_{2i } \int_0^{s^f_2(x_1(x,\xi,0),0 ) } x_2^n(x_1(x,\xi,0),0,s ) ds \nonumber \\ & & + \bar q_2 \bar \phi \frac{\bar c^n k_{\epsilon}^n}{n ! } \sum_{i=1}^4 \bar c_{3i } \int_0^{s^f_3(x_1(x,\xi,0),0 ) } x_3^n(x_1(x,\xi,0),0,s )
ds \nonumber \\ & & + \bar \phi \frac{\bar c^n k_{\epsilon}^n}{n ! } \sum_{i=1}^4 \bar c_{1i } \int_0^{s^f_1(x,\xi)}x_1^n(x,\xi , s)ds \nonumber \\ & \leq & \bar q_1 \bar \phi \frac{\bar c^n k_{\epsilon}^{n+1}}{n ! } \sum_{i=1}^4 \bar c_{2i } \frac{x_1(x,\xi,0)^n}{n+1 } + \bar q_2 \bar \phi \frac{\bar c^n k_{\epsilon}^{n+1}}{n ! } \sum_{i=1}^4 \bar c_{3i } \frac{x_1(x,\xi,0)^n}{n+1 } \nonumber \\ & & + \bar \phi \frac{\bar c^n k_{\epsilon}^{n+1}}{n ! } \sum_{i=1}^4 \bar c_{1i } \frac{x^n}{n+1 } \leq \bar \phi \frac{\bar c^{n+1 } k_{\epsilon}^{n+1}x^{n+1}}{(n+1 ) ! } , \end{aligned}\ ] ] since @xmath405 .
thus the lemma is proved .
next we show that ( [ eqn - fseriesdef ] ) converges .
[ prop - conv ] for @xmath406 , @xmath334 , one has that @xmath407 the result follows if we show that @xmath408 .
we prove the bound by induction . for @xmath409 , the result follows from ( [ eqn - fn ] ) .
assume that the bound is correct for all @xmath245 in @xmath410 .
then , we get for @xmath411 that @xmath412 ( x,\xi ) \right| \leq \bar \phi \frac{\bar c^{n+1 } k_{\epsilon}^{n+1 } x^{n+1}}{(n+1)!},\ ] ] where we have used lemma [ lemma - bound ] .
thus the proposition follows . from proposition [ prop - conv ]
we conclude that the successive approximation series is bounded and converges uniformly .
thus , a bounded solution to equations ( [ eqn - hypf1])([eqn - bcf4 ] ) exists .
this proves the existence part of theorem [ th - wp ] . to prove uniqueness ,
let us denote by @xmath413 and @xmath414 two different solutions to ( [ eqn - hypf1])([eqn - bcf4 ] ) .
defining @xmath415 .
by linearity of ( [ eqn - hypf1])([eqn - bcf4 ] ) , @xmath416 also verifies ( [ eqn - hypf1])([eqn - bcf4 ] ) , with @xmath417 for all @xmath245 .
then @xmath418 for @xmath419 , and proposition [ prop - conv ] we conclude @xmath420 , which implies that @xmath421 . to prove that the solution is continuous , note that since ( [ eqn - fseriesdef ] ) converges uniformly , one only needs to prove continuity of each term .
first , @xmath422 since the @xmath423 are defined as a sum of compositions of continuous functions .
similarly , since @xmath424 is defined as the integral ( with continuous limits ) of continuous functions times the previous @xmath425 composed with continuous functions , by induction it can be shown that @xmath426 .
thus @xmath427 and theorem [ th - wp ] is proved . next we sketch the proof of theorem [ th - sm ] .
we only consider @xmath428 ; for @xmath429 the result can be proven by induction .
denote @xmath430 and @xmath431 . by differentiating with respect to @xmath13 and @xmath432 in ( [ eqn - hypf1])([eqn - hypf4 ] )
, we find two uncoupled @xmath433 hyperbolic systems ( for @xmath434 and for @xmath435 ) @xmath436 now , differentiating the boundary conditions in ( [ eqn - bcf2 ] ) it is found that : @xmath437 and setting @xmath438 in ( [ eqn - hypf2])([eqn - hypf3 ] ) @xmath439 we can find a set of boundary conditions for @xmath440 and @xmath441 , @xmath442 , at the boundary @xmath438 : @xmath443 similarly , differentiating boundary conditions ( [ eqn - bcf1 ] ) and ( [ eqn - bcf4 ] ) we find boundary conditions for @xmath444 and @xmath445 at @xmath446 : @xmath447 and setting @xmath446 in ( [ eqn - hypf1])([eqn - hypf4 ] ) we can also find two sets of boundary conditions for @xmath441 , @xmath448 , at the boundary @xmath446:@xmath449 thus , both the @xmath450 s and @xmath451 s verify equations formally equivalent to ( [ eqn - hypf1])([eqn - bcf4 ] ) , with derivatives of the old equations coefficients as new coefficients , and the @xmath333 s as additional terms .
if these equations have solutions , then the solutions must be the partial derivatives of the @xmath333 functions .
now , under the assumptions of theorem [ th - sm ] , by theorem [ th - wp ] there is a ( at least ) continuous solution @xmath452 . plugging that solution into the equations we just derived for the @xmath450 s and @xmath451 s , one obtains equations whose coefficients and boundary conditions are ( at least ) continuous .
hence theorem [ th - wp ] can be applied implying that the @xmath450 s and @xmath451 s are continuous .
thus @xmath453 , proving the result .
next we give some technical lemmas used throughout the paper .
the first lemma follows from the fact that the control direct and inverse kernels are @xmath202 functions .
@xmath454 \vert & \leq & c_1 \left ( \vert \gamma\vert+\vert \gamma \vert_{l^1 } \right),\\ \vert \mathcal l[\gamma]\vert & \leq & c_2 \left ( \vert \gamma\vert+\vert \gamma \vert_{l^1 } \right),\\ \vert \mathcal k_{1}[\gamma]\vert & \leq & c_3 \left ( \vert \gamma\vert+\vert \gamma \vert_{l^1 } \right),\\ \vert \mathcal k_{2}[\gamma]\vert & \leq & c_4 \left ( \vert \gamma\vert+\vert \gamma \vert_{l^1 } \right),\\ \vert \mathcal l_{1}[\gamma]\vert & \leq & c_5 \left ( \vert \gamma\vert+\vert \gamma \vert_{l^1 } \right),\\ \vert \mathcal l_{11}[\gamma]\vert & \leq & c_6 \left ( \vert \gamma\vert+\vert \gamma \vert_{l^1 } \right).\end{aligned}\ ] ] the next lemma is based on the fact that , since @xmath455 is twice differentiable with respect to @xmath456 and @xmath13 , and since we have @xmath149 , it follows that there exists a @xmath457 and @xmath458 , @xmath459 , @xmath460 such that if @xmath461 , then , for any @xmath462 , it holds that @xmath463 similarly , since @xmath464 is twice differentiable with respect to @xmath456 and once with respect to @xmath13 , and @xmath465 , there exists a @xmath466 and @xmath467 , @xmath468 , @xmath469 such that if @xmath470 , then for any @xmath471 , @xmath472 then , the following lemma holds .
first , from ( [ eqn - gammanl2 ] ) we see that @xmath481(x)+f_4[\gamma](x)=0.\ ] ] therefore , calling @xmath210 the value of @xmath295 in lemma [ lem - f3f4bound ] and assuming @xmath211 , we can compute a bound on @xmath482 as follows : @xmath483\vert_{\infty}+\vert f_4[\gamma]\vert_{\infty } \nonumber \\ & \leq & k_2 \left ( \vert \gamma_x \vert_{\infty } + \vert \gamma_x \vert_{\infty } \vert \gamma \vert_{\infty } + \vert \gamma \vert_{\infty}^2 \right ) \leq k_3 \left ( \vert \gamma_x \vert_{\infty } + \vert \gamma \vert_{\infty } \right).\end{aligned}\ ] ] proceeding similarly with the @xmath5 norm , @xmath484\vert_{l^2}+\vert f_4[\gamma]\vert_{l^2 } \nonumber \\ & \leq & k_2 \left ( \vert \gamma_x \vert_{l^2 } + \vert \gamma_x \vert_{l^2 } \vert \gamma \vert_{\infty } + \vert \gamma \vert_{\infty } \vert \gamma \vert_{l^2 } \right ) \leq k_3 \left ( \vert \gamma_x \vert_{l^2 } + \vert \gamma \vert_{l^2 } \right).\end{aligned}\ ] ] for the last two inequalities , we solve for @xmath227 : @xmath485(x)+f_4[\gamma](x)\right).\ ] ] remembering the definition of @xmath70 , @xmath77}\left\{\frac{1}{\epsilon_1(x ) } , \frac{1}{\epsilon_2(x ) } \right\}>0 $ ] , and assuming that @xmath211 we obtain@xmath486 therefore if we choose @xmath487 , we reach the third inequality . proceeding similarly with the @xmath5 norm : @xmath488 so choosing @xmath489 , we reach the fourth inequality .
therefore , choosing @xmath490 , all inequalities are verified and the lemma is proven .
taking an @xmath13-derivative in ( [ eqn - gammanl2 ] ) : @xmath494 \gamma_{xx}+ f_{32}[\gamma,\gamma_x]+f_{42}[\gamma,\gamma_x]=0,\ ] ] where @xmath495 and @xmath496 are defined as : @xmath497\gamma_x\right ] + f_{12}[\gamma,\gamma_x]\gamma_x,\\ f_{42}&=&\mathcal k_{2 } \left [ f_1[\gamma ] \mathcal l_{1 } \left[\gamma \right ] + f_2[\gamma ] \right ] + f_{12}[\gamma,\gamma_x]\mathcal l_{1 } \left[\gamma \right ] \nonumber \\ & & + f_1[\gamma ] \mathcal l_{11 } \left[\gamma \right]+f_1[\gamma ] l(x , x ) \gamma_x+f_{23}[\gamma,\gamma_x ] \gamma_x , \end{aligned}\ ] ] where @xmath498 = \frac{\partial f_{nl}}{\partial x}\left(\mathcal l[\gamma],x\right)+\frac{\partial f_{nl}}{\partial \gamma}\left(\mathcal l[\gamma],x\right)\left(\gamma_x+\mathcal l_{1}[\gamma]\right).\ ] ] these functionals verify the following bound , similar to the bounds developed in lemma [ lem - f3f4bound ] , if @xmath473 : @xmath499 therefore , using lemma [ lem - gammaetaequiv ] , and inequality ( [ eqn - f1bound ] ) , and making @xmath492 small enough , we can compute the bounds as in the proof of lemma [ lem - gammaetaequiv ]
. we can write ( [ eqn - etanl2 ] ) analogously to ( [ eqn - gammanl2 ] ) : @xmath501(x)+f_6[\gamma,\eta](x)=0,\ ] ] where @xmath502 is defined as : @xmath503\eta_x + f_{11}[\eta]\gamma_x \right].\end{aligned}\ ] ] the functional @xmath502 verifies the following bound , similar to the bounds developed in lemma [ lem - f3f4bound ] , if @xmath473 : @xmath504 therefore , using lemma [ lem - f5f6bound ] , lemma [ lem - gammaetaequiv ] , and inequality ( [ eqn - f1bound ] ) , and making @xmath315 small enough , we can compute the bounds as in the proof of lemma [ lem - gammaetaequiv ] . | in this work , we consider the problem of boundary stabilization for a quasilinear @xmath0 system of first - order hyperbolic pdes . we design a new full - state feedback control law , with actuation on only one end of the domain , which achieves @xmath1 exponential stability of the closed - loop system .
our proof uses a backstepping transformation to find new variables for which a strict lyapunov function can be constructed .
the kernels of the transformation are found to verify a goursat - type @xmath2 system of first - order hyperbolic pdes , whose well - posedness is shown using the method of characteristics and successive approximations .
once the kernels are computed , the stabilizing feedback law can be explicitly constructed from them .
nonlinear hyperbolic systems , boundary conditions , stability , lyapunov function , backstepping , method of characteristics , integral equation , goursat problem tbd |
when a gas of atoms or molecules is subjected to a strong laser field , high - harmonic generation ( hhg ) takes place @xcite .
this process converts many of the laser photons into a single high - frequency photon in the extreme ultraviolet ( xuv ) or soft x - ray regime .
the generation process can be understood in terms of the three - step model @xcite .
first , the electron tunnels out under the influence of the laser field , and then it propagates freely in the laser field and can be driven back to the nucleus , where it can finally recombine .
hhg has been used to generate coherent xuv radiation @xcite .
concerning hhg from molecules , there has been a lot of attention of the community recently toward the relationship between the emitted radiation and molecular characteristics .
one can for instance determine the internuclear distance in the diatomic molecules h@xmath0 , h@xmath1 and to a lesser extent co@xmath1 and o@xmath1 from a destructive interference minimum in the harmonic spectrum @xcite or reconstruct molecular orbitals from the emitted radiation @xcite .
the correct explanation of the minimum in co@xmath1 , however , appears to require a multiorbital treatment @xcite .
recent experiments on n@xmath1 indicate multiorbital contributions as well @xcite . preparing molecules to control
the harmonic emission is also possible : ring current states were used theoretically to generate circularly polarized high harmonics @xcite . in the three - step model @xcite or its quantum - mechanical formulation in terms of the strong - field approximation ( sfa ) @xcite , due to wave - packet
spreading the on - axis continuum wave packet carries practically no signature of the ground - state wave function at the moment of recombination .
therefore , the molecular characteristics imprinted on the emitted spectra must come from the recombination step ( or from multiorbital interference ) . in the existing practical molecular imaging techniques
the influence of the laser field on the recombination step and often also the effect of the coulomb potential on the propagation step are ignored . to study the effects of these approximations ,
we perform a numerical comparison between the harmonics emitted in a normal laser - induced hhg process and harmonics emitted when an artificially prepared wave packet collides with the molecular ion in the absence of any laser pulse @xcite .
additionally , we compare the harmonics generated by both physical and artificial laser pulses to disentangle the influence of the different harmonic trajectories .
we show that elliptically polarized hhg radiation from linearly polarized generating pulses occurs near the two - center interference .
elliptical polarization of harmonics from linearly polarized generating pulses was also found in recent experiments @xcite .
in this article we solve the time - dependent schrdinger equation ( tdse ) numerically in two dimensions for a molecular ion with a single electron .
we consider two dimensions because in 2d many of the 3d characteristics of hhg are already present , such as the existence of directions perpendicular and parallel to the laser polarization axis . on the other hand
, the tdse can be solved very quickly , allowing the tdse to be solved for many alignment angles of the molecule in the laser field .
the tdse is solved using the split - operator method @xcite , and the ground - state wave function is found by imaginary - time propagation @xcite .
atomic units are used throughout this article unless indicated otherwise .
we will focus on 2d h@xmath0 with fixed nuclei .
the tdse reads @xmath2 where @xmath3 is the momentum operator , @xmath4 , @xmath5 is a softcore potential , and @xmath6 is the time - dependent electric field of the laser pulse . the internuclear axis @xmath7 makes an angle @xmath8 with respect to the laser polarization axis @xmath9 .
using the softcore parameter @xmath10 and an internuclear distance of @xmath11 , the ionization potential is @xmath12 ev . the total harmonic emission spectrum including @xmath9- and @xmath13-polarization is calculated from the numerical solution of the tdse as @xmath14 [ eq : tdse harmonics ] where @xmath15 is the dipole acceleration and @xmath16 is proportional to the intensity of the emitted radiation at frequency @xmath17 . here
@xmath18 is a standardly used temporal window that prevents high - frequency artifacts at the boundaries of the integration . for h@xmath0 , one observes a minimum in the spectrum of emitted radiation polarized in the @xmath9-direction because of interference between the two centers of the molecule @xcite .
using the plane - wave approximation for the returning electron , the ( first and usually only observable ) minimum occurs when the @xmath9-projection of the internuclear distance as seen by the returning wave packet is equal to half the de broglie wavelength . therefore the minimum will shift toward higher harmonics with higher angles between the laser polarization and molecular axis .
since it is a structural minimum that depends only on the geometry of the bound state , one expects to see no shifts in the location of the minimum when different laser pulses are used or when instead gaussian wave packets are used to generate harmonics in a laser - field - free electron - ion collision @xcite . for the wave - packet simulation without laser field ,
the center of the potential is placed in the middle of the grid at @xmath19 , and the gaussian wave packet is introduced with its center at position @xmath20 , with the grid size denoted as @xmath21 .
the initial wave packet @xmath22 is given by the superposition @xmath23 where @xmath24 is the ground - state wave function , and @xmath25 , @xmath26 quantify the momentum spread of the gaussian wave packet in the @xmath9- and @xmath13-directions .
the wave packet moves with a central momentum @xmath27 toward the molecular core .
the norm of the gaussian wave packet @xmath28 should be set small to mimic the situation of hhg at the typically used intensities .
we use @xmath29 .
the momentum - spread parameter in the @xmath13-direction is chosen as @xmath30 where a tuning parameter @xmath31 is used to study the effects of the different types of gaussian wave packets and can be set to simulate the character of the continuum wave packet as generated by a laser pulse .
the momentum - spread parameter in the @xmath9-direction @xmath25 is set relatively large to allow for many harmonics to be probed by one wave packet .
the propagation time is chosen such that a classical particle with momentum @xmath32 moves from @xmath33 to @xmath34 during the propagation . as a result , the strongest emission is expected at the middle of the propagation , such that little distortion is introduced when using a window function in the temporal fourier transform for obtaining the power spectrum . as an example , for @xmath35 and @xmath36 , the grid dimensions are @xmath37 a.u . and @xmath38 a.u .
here we use @xmath39 spatial grid points and 2000 time steps .
the propagation time equals the optical period of a laser field with a 780-nm wavelength . for the simulation of the laser - induced hhg process we use a laser pulse
linearly polarized in the @xmath9-direction .
the initial state is set to the ground state @xmath24 .
the time - dependent wave function is propagated for the laser pulse duration and two additional cycles after the end of the laser pulse to minimize distortions from the dipole acceleration window and to allow the wave packets to return to the nucleus . for the laser - induced calculations we use a grid measuring @xmath40 a.u . with @xmath41 grid points and 2000 time steps per optical cycle .
the laser wavelength for the calculations of laser - induced hhg is 780 nm . for the gaussian - wave - packet collisions , the same energy scale in units of harmonics of a 780 nm laser pulse
is used . in fig .
[ fig : intensity1 ] we show the alignment dependence of the emitted radiation for harmonic 49 polarized along the @xmath9-direction . we compare a gaussian wave packet with @xmath36 and @xmath35 to laser pulses with an intensity of @xmath42 w/@xmath43 and different lengths .
the laser pulses have a @xmath44 envelope of either three or five cycles , or a trapezoidal envelope of fifteen cycles length with five - cycle ramps .
the carrier - envelope phase , i.e. , the phase between the carrier wave and the envelope , is @xmath45 for the shortest pulse and 0 for the five - cycle @xmath44-pulse .
the trapezoidal pulse has a sinusoidal carrier wave .
the intensity data are integrated over one harmonic order .
the figure shows that the different laser pulses give rise to minima that are near , but not exactly at , the same position as the minimum from the gaussian wave packet . for emission polarized along the @xmath9-direction . a gaussian wave packet ( black solid line )
is compared to a three - cycle @xmath44-pulse ( red dashed line ) , a five - cycle @xmath44-pulse ( blue dotted line ) and a fifteen - cycle trapezoidal pulse ( green dot - dashed line ) . ] in fig .
[ fig : intensity2 ] we plot the positions of the minima @xmath46 in the alignment dependence versus harmonic order for different laser pulses and the gaussian wave packet from fig .
[ fig : intensity1 ] . also indicated in the figure are the curves that are predicted for the two - center minimum ( @xmath47 ) @xcite using either the energy - conserving relationship @xmath48 from the lewenstein model @xcite , or using the @xmath49-corrected relationship @xmath50 that has been adopted previously for molecular imaging @xcite .
the physical argument for the @xmath49-correction is that , when describing the returning electron as a plane wave , one should take into account that at the moment of recombination its wave number is modified by the absorption of @xmath51 into the kinetic energy .
( left ) blue crosses are for the gaussian wave packet and red plusses for the three - cycle pulse of fig .
[ fig : intensity1 ] .
( right ) green squares are for the fifteen - cycle pulse of fig .
[ fig : intensity1 ] and violet triangles are for a ten - cycle trapezoidal pulse .
the black solid line displays the two - center interference based on the sfa relation @xmath48 and the black dashed line is the @xmath52-corrected result based on the relation @xmath50 . ] from fig .
[ fig : intensity2 ] it is clear that the gaussian wave packet gives rise to a very smooth shift of the minimum as a function of alignment angle @xmath8 .
the laser pulses produce minima that follow the same trend as the gaussian wave packet , but are scattered around the general trend .
the results for short and longer pulses are scattered differently , but not less or more .
the differences between relatively long ten- and fifteen - cycle pulses pulses are small .
this is expected , since both pulses are effectively almost cw - like .
even for the fifteen - cycle pulse the depletion of the ground state remains below 6% .
the results suggest that when using the plane - wave approximation for the returning electron in molecular imaging applications , a dispersion relationship in between the lewenstein and @xmath49-corrected relationships should be used @xcite .
we compare the positions of interference minima for different types of gaussian wave packets in fig .
[ fig : intensity3 ] .
we vary the momentum spread of the wave packet in the perpendicular direction , the central momentum of the wave packet and the position at which the wave packet starts . for a broad gaussian wave packet starting far from the nucleus ( blue crosses ) and a broad gaussian wave packet starting close to the nucleus ( red plusses ) .
( right ) a narrow gaussian wave packet starting close to the nucleus ( green squares ) and a broad gaussian wave packet starting far from the nucleus but with less energy ( violet triangles ) . the blue crosses and red plusses have @xmath53 and @xmath36 , the violet triangles have the same @xmath31 but @xmath54 .
the red plusses and green squares start at @xmath55 a.u . , the blue crosses correspond to an initial position of @xmath56 a.u . and the green squares to @xmath57 a.u . ]
the striking and important observation is that all curves lie very close together .
apparently the position of the minimum is insensitive to the momentum distribution of the continuum wave packet . only for the lowest harmonics we can observe some difference between the different kinds of continuum wave packets .
small differences appear there between wave packets starting far away and those starting close to the nucleus , due to the effect of the long - range binding potential .
our findings suggest that in terms of the three - step model , the propagation step has little effect on the observed position of the minimum and can not account for the big fluctuations observed in fig .
[ fig : intensity2 ] . in the three - step model @xcite ,
the laser field during the recombination step and possible interferences between different parts of the continuum wave packet are ignored . in reality
, the electronic wave packet does not recombine under laser - field - free conditions , and different trajectories recombine at different times with different phases . to study the effect on the recombination process , we resort to a comparison of the minimum positions using artificial pulses .
the pulses are four - cycle sinusoidal pulses with a constant envelope corresponding to an intensity of @xmath42 w/@xmath43 , i.e. , a section of a cw laser field . at @xmath58 ,
the electric field is @xmath59 .
optionally , we employ either or both of two methods to influence the recombination step : ( i ) setting the dipole acceleration to 0 after some point in time during the propagation and ( ii ) turning off the laser field for the inner region near the nuclei after some point in time . every half laser cycle ,
both a short and long classical electronic trajectory contribute to every harmonic peak @xcite .
the distinction between short and long trajectories is based on whether the electron spends shorter or longer than @xmath60 in the continuum , where @xmath61 is the laser period . setting the dipole acceleration to 0 beyond @xmath62 using a temporal width @xmath63 @xcite , eq .
becomes @xmath64 [ eq : alpha filter ] where @xmath65 is the propagation length , @xmath66 and @xmath67 .
we use @xmath68 .
we set the dipole acceleration to 0 at either @xmath69 , the return time of the most energetic trajectory , or at @xmath70 , the time at which the return momentum of the first half - cycle s long trajectory matches that of the second half - cycle s short trajectory .
thus with @xmath69 we take into account only the short trajectories from the first half - cycle , and @xmath70 is the optimal point in time for selecting only a single pair of short and long trajectories . additionally , we optionally turn off the laser in the inner region at time @xmath71 . to prevent artifacts ,
the field is turned off gradually in both space and time .
the laser interaction is completely turned off for @xmath72 , undisturbed for @xmath73 , and we use a @xmath44-transition between these two extremes . in the time domain , we use a smoothened step function ( convolution of a gaussian with a step function ) with a width of @xmath74 . in formula , the hamiltonian in eq .
is replaced by @xmath75 when @xmath76 is used , this special setup allows us to compare near - physical harmonics to those generated in an identical setup where only trajectories starting during the first half - cycle contribute and with the laser field completely turned off during all recombinations .
this method could be easily extended to filter out either the short or long trajectories . in fig .
[ fig : intensity4 ] we compare the scattering around the gaussian - wave - packet results from the sinusoidal pulse ( red plusses ) with those from setting the dipole acceleration to 0 at @xmath70 ( green circles ) and from setting the dipole acceleration to 0 at @xmath69 ( violet points ) . for a broad gaussian wave packet starting far from the nucleus ( blue crosses ) and for harmonics generated from a sinusoidal laser pulse ( red plusses ) .
( right ) harmonics from sinusoidal pulses with @xmath70 ( green circles ) and with @xmath69 ( violet points ) . ] with @xmath70 the interference between the long and short trajectories leads to a strong , but regular oscillation of the laser - induced results around the gaussian - wave - packet results . the same interference between the short and long trajectories can be seen in an associated harmonic spectrum as the top , black solid line in fig .
[ fig : intensity5 ] . when every harmonic peak is caused by a single trajectory ( violet points in fig .
[ fig : intensity4 ] ) , the interference disappears completely and the result is almost as smooth as that from the gaussian wave packet . under normal circumstances , additional later returns from the same trajectories
contribute to the spectrum . for a finite pulse length
, different half - cycles also contribute differently because of the pulse envelope .
additionally , for a numerical calculation the dipole acceleration window @xmath18 also changes the contributions between different half - cycles .
all of these together then smoothen but irregularize the oscillation of the green circles in fig .
[ fig : intensity4 ] , leading to the scattering of the laser - induced data points observed in fig .
[ fig : intensity2 ]
. the interference between the short and long trajectories can probably also explain the scattering of the two - center minimum as a function of intensity as found by gonoskov and ryabikin @xcite .
-direction for the green circles from fig .
[ fig : intensity4 ] for @xmath77 ( black solid line ) and @xmath78 ( red dashed line ) .
also for the violet points from fig .
[ fig : intensity6 ] for @xmath77 ( blue dotted line ) and @xmath78 ( green dot - dashed line ) .
solid and dashed arrows indicate spectral minimum positions predicted using @xmath48 and @xmath50 , respectively . ] the fact that the results for a single harmonic trajectory ( violet points in fig .
[ fig : intensity4 ] ) lie so close to the gaussian - wave - packet result , means that the laser field has no significant influence on the amplitude of the recombination matrix elements .
this supports using hhg for molecular imaging @xcite , as in a typical experimental setup only short trajectories contribute to the harmonic spectrum .
interestingly , however , in fig .
[ fig : intensity6 ] we show that turning off the laser field during the recombination does have a significant effect on the interference between the short and long trajectories . for harmonics from a sinusoidal laser pulse with @xmath70 ( green circles ) and additionally with @xmath76 ( violet points ) . ] in the figure , the green circles are copied from fig .
[ fig : intensity4 ] .
additionally , we show the case where there is only a single set of short and long trajectories with additionally the laser pulse turned off in the inner region at @xmath76 ( violet points ) . the strong reduction in scattering amplitude for the violet points in fig .
[ fig : intensity6 ] can be understood from fig .
[ fig : intensity5 ] , where we plot the harmonic spectra for molecular alignment angles @xmath79 and @xmath80 for the case of the green circles ( violet points ) in fig .
[ fig : intensity6 ] as the top ( bottom ) two curves . for the unmodified laser pulse
we observe a significant shift of the trajectory interference positions in the harmonic spectrum when going from alignment at @xmath79 to @xmath80 .
although at first sight the bottom two curves in fig .
[ fig : intensity5 ] look more distinct from one another , a closer look reveals that the alignment dependence of the trajectory interference minima is actually a lot smaller with the laser field turned off in the inner region , as there is no shift visible . the strong scattering at low harmonics for the violet points in fig .
[ fig : intensity6 ] is caused by the fact that the finite temporal widths of the filters @xmath81 and @xmath82 suppress the complete lower end of the spectrum .
the two - center minimum in the harmonic spectrum is accompanied by a phase jump in the harmonic phase . using the plane - wave approximation , this should be a sharp @xmath83-phase jump @xcite .
however , in experiments a smaller and smoother phase jump is observed @xcite .
such deviations can be attributed to nonclassical momenta @xcite and to effects of the coulombic potential @xcite .
similarly , a phase jump is observable when one considers a fixed harmonic as a function of @xmath8 . the phase of harmonic 49 for emission polarized along @xmath9 is shown in fig .
[ fig : harmonicphase ] .
the same set of laser pulses and gaussian wave packet is used as in fig .
[ fig : intensity1 ] .
the curves have been shifted such that for @xmath84 the phase is 0 . for harmonic 49 generated by a gaussian wave packet ( black solid line ) and generated by a three - cycle @xmath44-pulse ( red dashed line ) , a five - cycle @xmath44-pulse ( blue dotted line ) , and a fifteen - cycle trapezoidal pulse ( green dot - dashed line ) . ]
the figure shows that both the gaussian wave packet and the extremely short three - cycle laser pulse give rise to a mostly constant phase as a function of @xmath8 with a phase jump slightly smaller than @xmath83 at the location of the minimum .
the longer pulses have a more smeared - out phase jump . in the neighborhood of the minimum their jump
is a lot smaller than @xmath83 but over the complete @xmath8-range the jump seems to be bigger than @xmath83 .
this behavior for the longer pulses is probably an effect of more , and longer , trajectories contributing to the harmonics .
different trajectories are associated with different coulomb corrections and therefore the harmonic phase becomes smeared out .
this is in accordance with the shallower intensity minima in fig .
[ fig : intensity1 ] for the longer pulses .
it is interesting to investigate the phase jump for the different gaussian wave packets of fig .
[ fig : intensity3 ] .
this is plotted in fig .
[ fig : harmonicphase2 ] for a smaller range of @xmath8 for clarity . for harmonic 49 generated by a broad gaussian wave packet starting far from the nucleus ( solid black line ) , a broad gaussian wave packet starting close to the nucleus ( red dashed line ) , a narrow gaussian wave packet starting close to the nucleus ( green dot - dashed line ) , and a broad gaussian wave packet starting far from the nucleus but with less energy ( blue dotted line ) . ]
the broad gaussian wave packet starting closing to the nucleus ( red dashed line ) starts out with small perpendicular momentum components .
because of the short propagation time before the interaction with the core , both coulomb effects on the momentum distribution and perpendicular momentum components will be relatively small in this scenario . as a result
we observe a sharp almost-@xmath83-phase jump . to some extent
, the difference with the narrow gaussian wave packet ( green dot - dashed line ) is that the latter experiences roughly the same coulomb effects , but starts out with much larger perpendicular momentum components .
this leads to a much smoother phase jump .
independent of the central momentum of the wave packet , a broad wave packet starting far away from the nucleus ( black solid and blue dotted lines ) experiences a smoother phase jump because of coulomb effects and the associated increased nonparallel momentum components .
high harmonic radiation is coherent with well - defined polarization @xcite .
we can write the emitted radiation field @xmath85 as @xmath86 where @xmath87 is a complex function and @xmath88 is the ellipticity as a function of frequency @xmath17 . here @xmath88 can be either positive or negative , @xmath89 represents linearly polarized light , and @xmath90 represent positively and negatively circularly polarized light , depending on the choice of propagation direction .
furthermore , @xmath91 is the unit vector in the main polarization direction , and @xmath92 is the unit vector in the perpendicular direction .
the angle between the main polarization direction of the emitted radiation and the polarization axis of the laser pulse is @xmath93 .
similarly to the angle @xmath8 between the molecular axis and the laser polarization axis , a counterclockwise rotation corresponds to a positive angle .
however , we limit the definition of @xmath93 to the range @xmath94 , whereas the complex number @xmath87 covers the full complex domain . at a given @xmath17 ,
all possible types of coherent plane - wave radiation can be uniquely described by the parameters @xmath95 .
numerically the acceleration of the dipole moment is identified with the far - field harmonic field ( up to an overall phase )
. therefore the fourier transformed dipole acceleration is given by @xmath96 for an experimentalist the most practical way to measure the polarization angle and ellipticity of the emitted radiation is to pass the harmonics through a polarization filter and measuring the emitted intensity for many polarization directions .
the direction of greatest emission corresponds to the main polarization direction , and the emission in the orthogonal direction is a measure for the ellipticity of the emitted radiation @xcite . only under considerable effort for the experimentalist , the harmonic phase can be measured interferometrically as a function of alignment angle @xcite or as a function of harmonic order @xcite .
theoretically one has easy access to the harmonic phase . using the phase information , eq .
can be inverted as described in the following to deduce the polarization parameters @xmath93 , @xmath88 and the complex number @xmath87 from the complex - valued @xmath97 and @xmath98 . as a measure of the ellipticity
we define a cross term @xmath99 as ( we omit the @xmath17-dependence for simplicity ) @xmath100 where and we used that @xmath101 .
the following equalities can be straightforwardly derived for the polarization parameters in terms of @xmath102 and @xmath99 , @xmath103 [ eq : polarizationconversion ] where for @xmath87 we picked the numerically most stable expression .
numerically one will run into problems using the above conversion if the emitted radiation is either linearly or circularly polarized .
therefore one should check beforehand if one of these conditions applies and use appropriate simplified conversion equations instead .
the two - center interference minimum can also be observed in the polarization direction @xmath93 of the harmonics . because the emission in the direction parallel to the laser polarization direction is strongly suppressed at the minimum
, we expect a @xmath104-jump toward the minimum .
the @xmath83-jump for the harmonic phase in the @xmath9-direction in fig .
[ fig : harmonicphase ] translates to a full @xmath83-rotation for @xmath93 .
this is exactly what is observed for a gaussian wave packet and harmonics generated by different laser pulses in fig .
[ fig : polarization1 ] .
the polarization direction of the emitted radiation was averaged over one harmonic order using the total emitted intensities as weights . for harmonic 49 generated by a gaussian wave packet ( black solid line ) and generated by a three - cycle @xmath44-pulse ( red dashed line ) , a five - cycle @xmath44-pulse ( blue dotted line ) , and a fifteen - cycle trapezoidal pulse ( green dot - dashed line ) . ]
the phase jump in the @xmath9-direction at the two - center interference minimum becomes smoother for low harmonics @xcite . in the following
we investigate how the jump in the polarization direction depends on harmonic order .
we plot the main polarization direction for different harmonics in fig .
[ fig : polarization2 ] . for harmonic 31 ( black solid line ) , harmonic 41 ( red dashed line ) , harmonic 51 ( blue dotted line ) , and harmonic 61 ( green dot - dashed line ) . the black crosses are the phase difference @xmath105 for harmonic 31 . a three - cycle @xmath44-pulse was used . ] here we plot the polarization data points corresponding to the exact harmonics , i.e. , no averaging was done .
the right - most curves for high harmonics show a simple polarization - direction jump around the two - center minimum . for the lowest harmonics ( on the left - hand side )
the behavior becomes more complicated .
to explain this finding , we also plot the phase difference @xmath105 between @xmath98 and @xmath97 for harmonic 31 ( black crosses ) .
the intensity ratio @xmath106 equals 1 for @xmath107 and @xmath108 , and it reaches a maximum of @xmath109 at @xmath110 . for a given ratio @xmath106 , the main polarization direction is aligned more along the laboratory - frame ( @xmath9 or @xmath13 ) direction with the higher amplitude if the phase difference @xmath105 between the laboratory - frame directions is far from 0 or @xmath83 .
we can observe this effect clearly in fig .
[ fig : polarization2 ] : at @xmath8 equal to @xmath111@xmath112 for harmonic 31 , where @xmath105 is around @xmath104 and @xmath113 is bigger than @xmath114 , the relatively slow increase in the polarization direction shows the tendency that the polarization is clamped toward the @xmath13-direction ( @xmath115 ) .
we fit the jump observed in fig .
[ fig : polarization1 ] with a smoothened step function to determine the location @xmath116 and the width @xmath117 of the polarization - direction jump .
when one plots the location @xmath116 as a function of harmonic order , one obtains a result very similar to that shown in fig .
[ fig : intensity2 ] . in fig .
[ fig : polarization3 ] we plot the width @xmath117 as a function of harmonic order for a gaussian wave packet and different laser pulses . .
blue crosses are for the gaussian wave packet , red plusses for the three - cycle pulse , and green squares for the fifteen - cycle pulse .
the brown line shows the curve expected from purely geometric considerations . ] again , the results for the different laser pulses are scattered around the gaussian - wave - packet result .
the width of the jump for @xmath93 as a function of harmonic order does not depend on the laser pulse length .
one observes that the width of the jump decreases with increasing harmonic order .
an important part of this decrease is due to a purely geometric effect : with increasing harmonic order , the minimum moves to higher @xmath8 leading to a narrower interference pattern as a function of @xmath8 . in the absence of coulomb effects ,
i.e. , when @xmath97 and @xmath98 have the same phases apart from @xmath83-jumps , the polarization direction @xmath93 is given by @xmath118 .
we assume that @xmath98 is alignment independent over the range of the two - center minimum , as also indicated by numerical tests that we have performed , and that the alignment dependence of @xmath97 comes purely from the two - center interference with the phase lag @xmath119 , i.e. , @xmath120 @xcite .
then the variation @xmath121 of the polarization direction on varying @xmath8 in the vicinity of @xmath116 is proportional to @xmath122 .
thus the width @xmath123 should be such that @xmath124 is independent of harmonic frequency , provided that @xmath125 is frequency independent .
the brown line in fig .
[ fig : polarization3 ] is obtained for @xmath123 if we set @xmath124 arbitrarily equal to @xmath126 using @xmath127 .
a comparison between the brown line and the other curves shows that at the high end of the spectrum the decrease in @xmath117 can not be explained any more exclusively by the geometric effect .
since @xmath116 varies very slowly in this range , we expect only a slow variation in @xmath125 .
this suggests that decreasing coulomb effects play a role , in accordance with fig .
[ fig : intensity2 ] , where for harmonics 5080 we observe a transition toward the curve predicted by the sfa dispersion relationship , also indicating decreasing coulomb effects in this range . in fig .
[ fig : ellipticity1 ] we plot the ellipticity @xmath88 of the emitted radiation for harmonic 49 as a function of @xmath8 . again
the polarization data was averaged over one harmonic order using the intensities as weights .
the plot shows that both a gaussian wave packet and different laser pulses give rise to both significant and varying elliptical emission .
nonzero ellipticity means that the harmonics in the @xmath9 and @xmath13-directions are emitted with different phases .
using the plane - wave approximation for the returning electron , one would not expect to see any ellipticity for a symmetric molecule @xcite .
since the ionization and propagation step are identical for the two components of the radiation , the ellipticity must come from the recombination step .
therefore , this result confirms that the coulomb effects can lead to significant ellipticity .
the ellipticity for parallel or perpendicular alignment is zero , because at these alignment angles , the perpendicular component of the emitted radiation vanishes .
-pulse ( red dashed line ) , a five - cycle @xmath44-pulse ( blue dotted line ) , and a fifteen - cycle trapezoidal pulse ( green dot - dashed line ) . ]
the ellipticity goes through zero at some intermediate alignment angle .
if we plot the angle of zero ellipticity as a function of harmonic order , we arrive at fig .
[ fig : ellipticity2 ] .
this plot shows that in the close vicinity of the two - center interference minimum , the ellipticity goes through zero .
this is as expected , because at the location of the minimum , the @xmath9-component of the emitted radiation is very small . because the @xmath9-component has opposite signs before and after the minimum , the ellipticity changes handedness through the minimum . for each harmonic
, we can also plot the extrema of the ellipticity that can be reached and the alignment angles at which those extrema are reached .
the results are shown in figs . [
fig : ellipticity3 ] and [ fig : ellipticity4 ] , respectively . .
same wave packets and pulses as in fig .
[ fig : ellipticity3 ] . ]
we observe that for sufficiently high harmonic orders , the ellipticity extrema become smaller in absolute value and move closer to the two - center interference minimum with increasing harmonic order .
since the ellipticity is an indicator of non - plane - wave character , the decreasing ellipticity is another signature of decreasing coulomb effects for higher harmonic orders , which correspond to higher return momenta .
when a gaussian wave packet collides with an aligned h@xmath0-molecule , the resulting harmonic spectrum has a structural minimum from the two - center interference between the two lobes of the orbital .
the position of this minimum is reproduced using an effective plane - wave momentum that transitions from the @xmath52-corrected @xmath50 at low harmonics to the sfa - based @xmath128 at high harmonics .
a laser - induced hhg spectrum shows the same behavior if only a single electronic trajectory contributes per harmonic , as is the case for a typical experimental setup .
this justifies using hhg for molecular imaging as the laser field has no significant effect on the amplitude of the recombination matrix element .
when a single set of short and long electron trajectories contributes to the spectrum , the interference between the two trajectories causes a large but regular oscillation around the general trend .
introducing more and longer trajectories by using longer pulses has the effect of averaging out the oscillations to a smaller scattering around the gaussian - wave - packet result and leads to a smoother interference minimum as a function of @xmath8 .
our results show that the effect of the coulomb potential can lead to significant ellipticity of the emitted radiation . around the interference minimum
, the main polarization angle makes a @xmath83-jump and the ellipticity goes through zero .
the coulomb effects are less important at higher harmonics .
therefore we observe decreasing overall ellipticity and a relatively sharp jump in the polarization direction at the high end of the spectrum . in the wave - packet calculations ,
the coulomb effects could be investigated in more detail by changing the strength of the potential for the evolution of the continuum part .
this may be subject of future work .
finally , we mention that for randomly oriented molecules , the perpendicular harmonic components and thereby also the ellipticity vanish due to the cylindrical symmetry of the system around the laser polarization axis .
the behavior of the phase is more complicated . since , however , hhg is dominated in our case by the large orientation angles for geometrical reasons , we do not expect a clear signature of the phase jump for randomly oriented molecules .
the authors thank the deutsche forschungsgemeinschaft for funding the _
centre for quantum engineering and space - time research _
( quest ) .
we acknowledge the support from the european marie curie initial training network grant no .
ca - itn-214962-fastquast . | we present a theoretical investigation into the two - center interference in aligned h@xmath0 .
the influence of the laser field on the recombination step is investigated by comparing laser - induced harmonic generation with harmonic generation from field - free collisions of gaussian wave packets with the core .
we find that for different gaussian wave packets colliding with the molecule , the interference minimum occurs at the same alignment angle .
the same result is obtained for the laser - induced spectrum when only a single electronic trajectory per harmonic contributes . when multiple electronic trajectories contribute , we find an effect on the minimum position because the interference between short and long trajectories is alignment - dependent . the two - center interference and the influence of the coulombic potential
are clearly seen not only in the harmonic intensity and phase but also in the polarization direction and ellipticity .
we observe significant ellipticity of the emitted radiation around the two - center interference minimum . |
* probabilistic vs. deterministic domain decomposition . * in the solution of large boundary value problems ( bvps ) arising in realistic applications ( such as airfoil simulation or weather forecast ) , the discretization of the bvp on a domain @xmath0 leads to algebraic systems of equations that can only be solved on a parallel computer with @xmath1 processors .
not only does parallelization require multiple processors but also parallel algorithms .
the classical schwarz alternating method @xcite is the best known of the state - of - the - art algorithms which we will refer to as deterministicand serves to illustrate the difficulty that they all face .
the idea is to divide @xmath0 into a set of @xmath2 overlapping subdomains , @xmath3 , and have processor @xmath4 solve the restriction of the partial differential equation ( pde ) to the subdomain @xmath5 .
but because the solution is not known in the first place , an initial guess has to be made at start , in order to give processor @xmath4 a well - posed ( yet incorrect ) problem .
the boundary conditions ( bcs ) along the fictitious interfaces of @xmath5 are then updated from the solution of the surrounding subdomains in an iterative way , hopefully leading to convergence . however , since the inter - processor communication involved in this updating procedure is intrinsically sequential , it will eventually set an upper limit to the scalability of the algorithm according to amdahl s law .
summing up , the matrix algebra arising in the discretization of bvps does not , in general , lend itself optimally to parallellization .
+ an alternative to deterministic methods which is specifically designed to circumvent the scalability issue is the probabilistic domain decomposition ( pdd ) method , which has been successfully applied to elliptic @xcite , parabolic @xcite and hyperbolic bvps @xcite .
pdd consists of two stages . in the first stage ,
the solution is calculated only on a set of interfacial nodes along the fictitious interfaces , by solving the probabilistic representation of the bvp with the monte carlo method .
more precisely , the pointwise solution of the bvp is @xmath6 $ ] , i.e. the expected value of a functional @xmath7 of a given stochastic process @xmath8 conditioned to @xmath9 .
it is then possible to reconstruct ( approximately ) the solution on the interfaces , so that the pde restricted to each of the subdomains is now well posed , and can be independently solved the second stage of pdd .
note that both stages in pdd are embarrasingly parallel by construction
. moreover , pdd is naturally fault - tolerant .
subdomains and @xmath10 interfaces with @xmath11 nodes each ( @xmath12 ) . *
b ) * fem mesh on subdomain @xmath13 .
* c ) * the nodal values interpolated with rbfs along the interface @xmath14 , making up a dirichlet bc for @xmath13 and @xmath15 . ] * nomenclature of pdd .
* in this paper , we shall exclusively deal with elliptic bvps .
let us introduce some terminology ( see figure [ i : figura1 ] ) .
the domain @xmath16 ( not necessarily simply connected ) on which the bvp is being solved is partitioned into @xmath17 _ nonoverlapping _ subdomains @xmath18 .
the boundary @xmath19 of a subdomain @xmath20 contains several ( @xmath21 ) _ artificial interfaces_each of which is shared between @xmath20 and another adjacent subdomain which are labeled @xmath22 ( note that this labeling is not unique ) .
a subdomain boundary @xmath19 may or not contain some portion of the actual boundary . in sum , @xmath23 .
artificial interfaces are discretized into _ interfacial nodes _ ( or simply , nodes ) uniquely labeled @xmath24 .
assume @xmath17 functions @xmath25 defined on @xmath17 domains @xmath26 such that @xmath27 if @xmath28 .
then , their direct sum is defined as @xmath29 the solution of the bvp on the nodes is calculated by resorting to the probabilistic formulation of the bvp with a monte carlo method , yielding the _ nodal values _ ( or _ nodal solutions _ ) @xmath30 .
consider a subdomain @xmath20 .
a dirichlet bc can then be provided on every @xmath31 by interpolation of the nodal values @xmath32 .
along with the actual bcs which apply on @xmath33 , the bvp on @xmath20 now is well posed and can be solved right away , yielding @xmath34 . once the subdomain solutions @xmath35 are available , they are put together to form a global pdd solution : @xmath36 .
( we reserve @xmath37 for the exact solution of the bvp and denote the global pdd approximations with @xmath38 . ) since adjacent subdomains @xmath39 and @xmath5 share a dirichlet bc on their common interface , the pdd solution is continuous in @xmath0although not necessarily differentiable .
* the cost of computing the nodal solutions . * the convenience of pdd depends on whether a suitable stochastic representation for the bvp under consideration is available , and on the cost involved in numerically solving it . due to the poor accuracy of the monte carlo method ( compared with deterministic ones ) ,
the bulk of the cost of pdd falls on the calculation of the nodal solutions to within a required accuracy .
more precisely , given a _
nodal error tolerance _ @xmath40 and a _ confidence interval _ @xmath41 , the cost of solving the bvp on an interfacial node scales as @xmath42 , where @xmath43 is the weak convergence rate of the numerical integration scheme ( also called the _ bias _ )
. this poor rate of convergence is due to the slow convergence of both the statistical error and the bias , which have to be tackled simultaneously . for bvps with dirichlet bcs there quite a few linear integrators ( i.e. with @xmath44 ) . regarding the statistical error , replacing the mean by a multilevel estimator of the expected value of feynman - kac functionals
has recently been shown to dramatically reduce the cost to @xmath45 @xcite .
when the bias law is sharp , extrapolation la talay - tubaro or regression methods in the spirit of @xcite can further improve the accuracy at virtually no extra cost . * using rougher numerical solutions to reduce the variance . * by construction , pdd offers an additional device to accelerate the monte carlo simulation of the nodal values , namely the possibility of calculating and exploiting rougher estimates of the global solution of the bvp .
assuming that a numerical solution @xmath46 with a _ target nodal error tolerance _
@xmath47 is required , it may be worth to calculate before a rougher approximation @xmath48 , with an @xmath49 tolerance ; and then use it to draw the stochastic pathwise nodal control variate @xmath50 alongside the monte carlo realizations of the feynman - kac functional .
( in ( [ f : intro_xi ] ) , the integrals are ito s , @xmath51 is the bvp potential , and @xmath52 and @xmath53 are the diffusion matrix and first - exit time from @xmath0 of the stochastic process ( [ f : sde ] ) driven by a wiener process @xmath54 . )
this allows one to construct afterwards an estimator of the feynman - kac functional involving the control variate ( [ f : intro_xi ] ) , which has the same expected value but a much smaller variance .
this notion is what we call iterpdd . in order to fix ideas ,
let us introduce the following notation : @xmath55 means that @xmath38 is a pdd approximation obtained with tolerance @xmath40 and no variance reduction ; while @xmath56= \textrm{iterpdd$(a_0,a_1)$ , or simply iterpdd$(a_0,a_1)$}\ ] ] means that first @xmath57 is calculated without variance reduction , then differentiated in order to construct @xmath58 according to ( [ f : intro_xi ] ) , which in turn is used as control variate in order to reduce the variance in calculating @xmath59 with a target tolerance @xmath47 , which is the ultimate goal . because the nodal values of @xmath59 can now be calculated with much less variance , statistical errors are smaller , and the time ( or cost ) it takes the computer to hit the tolerance @xmath47 is also less .
in fact , @xmath60\big)\times$cost plainpdd$(a_0)$},\end{aligned}\ ] ] where @xmath61 $ ] is pearson s correlation . as ( [ f : cost_of_iterpdd ] ) indicates , there is a tradeoff between the effort invested in calculating @xmath62 , and the reduction of variance yielded by @xmath63 , which depends on the quality of @xmath62 .
the most straightforward procedure is to simply guess some @xmath49 .
while numerical tests indicate that the iterpdd strategy can be quite successful , a poorly chosen @xmath64 may well result in an overall cost of iterpdd@xmath65 _ larger _ than that of plainpdd@xmath66 .
therefore , at the heart of iterpdd lies an optimization problem for @xmath64 . in order to make educated guesses of @xmath64 given @xmath47 ,
two questions must be tackled : i ) how does the cost of a @xmath67plainpdd@xmath68 simulation depend on @xmath40 ; and ii ) how does @xmath69 $ ] depend on @xmath40 .
the former requires that the sde integrator have a predictable and sharp order of weak convergence .
deriving a _ sensitivity formula _ for ii ) is one of the main points of this paper . * a multigrid - like pdd algorithm .
* with such a sensitivity formula in place which can predict an optimal ( or more realistically , good enough ) initial tolerance @xmath64 for @xmath70=$]iterpdd@xmath65 , it is natural to try and compute @xmath62 faster by finding @xmath71 which minimizes the cost of @xmath72=$]iterpdd@xmath73 . much like in the multigrid method ,
a number @xmath74 and an _ optimal sequence _ of nested iterpdd simulations can be envisioned with tolerances @xmath75 which fully exploits the potential of control variates for a given bvp . however , in iterpdd , given a target tolerance @xmath47 , the number @xmath74 and the sequence @xmath76 must be determined _ before _ actually running one single pdd simulation . to compound matters , all of @xmath74 , @xmath77 , and in general any result provided by any such _ scheduling algorithm _
will be affected by the randomness introduced by monte carlo , making its performance meaningful only in terms of its expected value and variance .
* outline of the paper .
* we start by revisiting the probabilistic formulation of elliptic bvps with dirichlet bcs in section [ s : representation ] , and identifying the pathwise control variates .
section [ s : theory ] formally poses the problem namely , acceleration of pdd with the iterpdd scheme and presents our own theoretical results concerning the aforesaid sensitivity @xmath69 $ ] .
section [ s : globalerror ] links the nodal target error tolerance , @xmath47 , to the global pdd error tolerance @xmath78and discusses the stability of pdd . in section [ s : approximations ] , the formal restrictions in section [ s : theory ] are relaxed leading to practical , but partially heuristic , scheduling algorithm / sensitivity formula ( algorithm [ a : scheduling ] ) and final iterative , multigrid - like pdd loop ( algorithm [ a : iterpdd ] ) .
they are numerically tested in section [ s : experiment ] , and section [ s : conclusions ] concludes the paper .
consider the linear elliptic bvps with dirichlet bcs in @xmath79 : @xmath80 where the _ differential generator _ @xmath81 is @xmath82 and the functions @xmath83 are regular enough that the solution to ( [ f : ellipticbvp ] ) exists and is unique . in particular , if @xmath84 and the matrix @xmath85 $ ] is such that @xmath86 the pointwise solution to ( [ f : ellipticbvp ] ) admits the following probabilistic representation @xmath87:= e\big[\ , g({\bf x}_{\tau})e^{\int_0^{\tau}c\big({\bf x}_s\big)ds } \ , - \ , \int_0^{\tau}f\big({\bf x}_t\big)e^{\int_0^t c\big({\bf x}_s\big)ds}dt\,\big],\ ] ] where @xmath88 $ ] stands for the expected value , the functional @xmath7 is called _ score _ , and @xmath89 is the value at @xmath90 of the stochastic process @xmath91\rightarrow\mathbb{r}^d$ ] , driven by the stochastic differential equation ( sde ) @xmath92 where the diffusion matrix @xmath52 is obtained from @xmath93 ( by lu decomposition , by instance ) , and @xmath54 is a standard @xmath94-dimensional wiener process .
the _ first exit time _
@xmath53 is defined as @xmath95 , i.e. the time when a solution of the sde ( [ f : sde ] ) first touches @xmath96 at the _ first exit point _ @xmath89 .
the process @xmath97 can be thought of as a non - differentiable trajectory inside @xmath0 .
equation ( [ f : dynkin ] ) is dynkin s formula , a particular case of the more general feynman - kac formula for parabolic bvps
see @xcite for more details . in practice , solving ( [ f : dynkin ] ) involves two levels of discretization .
first , the sde ( [ f : sde ] ) has to be integrated numerically according to a numerical scheme @xmath98 ( which we will call _ integrator _ ) with a timestep @xmath99 , yielding a discretized score @xmath100 .
this results in @xmath101 $ ] being a biased estimator of @xmath102 $ ] , with _ signed _ bias @xmath103\big):=e[\phi_h]-e[\phi].\ ] ] as suggested by the notation , the bias depends on both the timestep @xmath104 and the specific integrator @xmath98 .
it takes asymptotically the form of a power law : @xmath105\big)\shortrightarrow \beta h^{\delta } \textrm { as } h\shortrightarrow 0^+,\ ] ] where @xmath106 is a signed constant independent of @xmath104 .
second , the expected value in ( [ f : dynkin ] ) is replaced by an estimator over @xmath107 independent realizations @xmath108 of the sde ( [ f : sde])which is the essence of the monte carlo ( mc ) method .
typically , that estimator is the mean , which , according to the central limit theorem , introduces a _ statistical error _
@xmath109 - \frac{1}{n}\sum_{j=1}^n \phi_h^{(j)}\big|\leq q\sqrt{\frac{v[\phi_h]}{n}}\ ] ] with probability @xmath110 , and @xmath111 for @xmath112 .
moreover , @xmath113:=e\big[\big(\phi_h - e[\phi_h]\big)^2\big]\leq b_{\xi}^2(e[\phi ] ) + v[\phi]/n , \ ] ] where mse stands for mean square error . since @xmath114|]\leq e[(\phi_h - e[\phi_h])^2]$ ] , @xmath101=u({\bf x}_0)$ ] , and @xmath102=u({\bf x}_0)$ ] , it holds @xmath115)|v[\phi]/n}\leq \\\nonumber \leq |b_{\xi}(e[\phi_h])|+\sqrt{v[\phi]/n } \leq |b_{\xi}(e[\phi_h])|+q\sqrt{v[\phi]/n},\end{aligned}\ ] ] with probabilities approaching @xmath41 as @xmath116
. looking at ( [ f : mse_error ] ) or ( [ f : bound_of_error ] ) , the two ways to speed up monte carlo simulations of ( [ f : dynkin ] ) are : using an integrator @xmath98 with the highest possible @xmath43 for the bvp under consideration ; and/or replacing the mean in ( [ f : statistical_error ] ) by another estimator of the expected value which has less variance , thus achieving the same statistical error with a smaller @xmath107 . in this paper
, we investigate a technique of variance reduction based on pathwise control variates , which is difficult to implement in other contexts , but suits the framework of pdd ideally . in order to discuss it ,
we adopt the same ansatz as in . consider the system of sdes : @xmath117dt + \sigma({\bf x}_t)d{\bf w}_t , & { \bf x}_0={\bf x}_0 , \\
dy_t= c({\bf x}_t)y_tdt + { \bm \mu}^t({\bf x}_t)y_td{\bf w}_t , & y_0=1,\\ dz_t= -f({\bf x}_t)y_tdt + { \bf f}^t({\bf x}_t)y_td{\bf w}_t , & z_0= 0 , \end{array } \right.\ ] ] where @xmath118 are smooth arbitrary fields , and let @xmath119 be the evaluation of the processes @xmath120 at @xmath121 . if @xmath122 , the system ( [ f : vr_milstein_sys ] ) is just another way of writing down the functional in ( [ f : dynkin ] ) : @xmath123 it turns out see for details that @xmath124=e[{\tilde\phi}]=e[\phi]\ ] ] regardless of the arbitrary functions @xmath125 and @xmath126 , but the variance @xmath127\neq v[{\tilde\phi}]= e\big[\ , \int_0^{\tau}y_t^2\parallel \sigma^t\nabla u+u{\bm\mu}+{\bf f}\parallel_2 ^ 2\,dt \big]\ ] ] _ does _ depend on them ; and in fact , making the choice @xmath128 then @xmath129=0$]meaning that one single realization would yield the exact pointwise solution @xmath130 deterministically .
certainly this is a moot point , for in order to do so , the exact solution would be required in the first place ; and the system ( [ f : vr_milstein_sys ] ) would still have to be integrated numerically so that the variance of the discretized approximation would be small , but finite .
let @xmath131 .
choosing @xmath132 in ( [ f : combining_method ] ) , @xmath133 leads to the pathwise equivalent of importance sampling , while if @xmath134 , @xmath135 can be interpreted as the stochastic equivalent of control variates . both importance sampling and
control variates are well - known variance reduction techniques in statistics @xcite .
there are several reasons why the second method is the more convenient : * control variates do not need to store @xmath38 . * for inadequate @xmath136 and @xmath137 , importance sampling
may actually lead to _ increased _ variance .
this is much less likely so with control variates ( as it will be seen in a moment ) .
* from the point of view of numerical integration , introducing a non - zero @xmath138 in ( [ f : vr_milstein_sys ] ) leads to a system of stochastic equations with multiplicative noise : @xmath139 .
such sdes can be unstable .
* finally , the trajectories are not affected by control variates , while if @xmath140 , the particles are actually pushed around depending on @xmath137 .
the first fact facilitates the analysis of iterpdd .
therefore , we will settle for ( [ f : exactf ] ) henceforth . in statistics ,
a control variate @xmath58 for @xmath7 is a variable which is drawn alongside @xmath7 and has a large correlation with it ( i.e. close to @xmath141 ) .
then , the variable @xmath142 ) \ ] ] is such that @xmath143=e[\phi]$ ] , but the variance is not . at the unique critical point @xmath144/v[\xi]$ ] , @xmath129 $ ] is minimal and yields a reduction of variance @xmath145}{v[\phi ] } = 1-\rho^2[\phi,\xi]\ ] ] ( see @xcite for details ) , where @xmath146 is pearson s correlation : @xmath147:= \frac{cov[\phi,\xi]}{\sqrt{v[\phi]v[\xi]}}\qquad\textrm{(such that $ \rho^2[\phi,\xi]\leq 1$).}\ ] ] by comparison of ( [ f : statistics_vr ] ) and ( [ f : vr_milstein_sys ] ) , the pathwise control variate is ito s integral @xmath148 ( note that @xmath149 , with @xmath150 , yields the same @xmath151 $ ] as @xmath58 and thus the same reduction of variance . )
when @xmath152 , we will refer to @xmath58 as the _ exact control variate_. in that case , the correlation is perfect and the variance is zero ( in the limit @xmath116 ) . if @xmath153 , @xmath58 is close to the minimizer @xmath154 of ( [ f : statistics_vr ] ) and the equality in ( [ f : variance_reduction ] ) still holds . but even if @xmath136 is so poor that ( [ f : variance_reduction ] ) breaks down becoming a mere lower bound , any reasonable substitute of @xmath155 will still produce a @xmath58 with sufficient correlation as not to increase the variance .
summing up , the method of pathwise control variates is intrinsically robust unlike pathwise importance sampling ( [ f : exactmu ] ) .
the control variate @xmath58 can be drawn by a quadrature of ito s integral ( [ f : xicontrolvariate ] ) , or alternatively , by enlarging system ( [ f : vr_milstein_sys ] ) with one extra equation : @xmath156 importantly , the same pathwise control variate can be used for parabolic bvps and bcs involving derivatives .
we close this section by adressing the choice of integrator @xmath98 for solving ( [ f : vr_milstein_sys ] ) with @xmath157 and @xmath158 .
the simplest integrator for bounded sdes is the euler - maruyama scheme plus a naive boundary test .
this yields @xmath159 , which is much poorer than the linear rate of weak convergence of the euler - maruyama integrator in free space . however , if the solution , coefficients , and boundary of the bvp are smooth enough , then there are a variety of methods which manage to raise @xmath43 to one
see @xcite for a recent review .
in particular , the integrator of gobet and menozzi restores weak linearity of the euler - maruyama scheme by simply shrinking the domain ( algorithm [ a : gobet_menozzi ] ) .
@xmath160 [ a : gobet_menozzi ] in algorithm [ a : gobet_menozzi ] , @xmath161 is the signed distance from @xmath162 to @xmath96 with the convention that it is negative inside @xmath0 ( see @xcite for further details ) ; @xmath163 is the approximation to @xmath53 ; @xmath164 is a @xmath165dimensional standard normal distribution ; and @xmath166 is the projection of @xmath167 on @xmath96 along the outward normal @xmath168 ( @xmath0 is supposed smooth enough that such projection exists and is unique ) .
[ d : balanced_mc_simulation ] a balanced mc simulation with accuracy @xmath40 and confidence interval @xmath41 is such that : statistical error = @xmath169 = absolute value of signed bias . a balanced mc simulation is thus one guaranteed to attain a total @xmath170 ( [ f : mse_error ] ) smaller than @xmath40 ( with a probability @xmath41 ) with the least computational complexity ( cost ) , because it takes the largest possible @xmath104 and the fewest possible realizations compatible with @xmath171 and @xmath41 , namely @xmath172}{a^2}\qquad\textrm { and } \qquad h=\big ( \frac{a}{2|\beta| } \big)^{1/\delta}.\ ] ] ( we remark in passing that , while splitting the total error between bias and variance looks natural enough , it is definitely suboptimal @xcite . )
the average number of steps before hitting the boundary is @xmath173/h,\ ] ] where mean first exit time , @xmath174 $ ] , is well defined for a given bvp and @xmath175 . on average , every trajectory makes @xmath176 visits to the integrator @xmath98 . for a balanced mc simulation and a tolerance @xmath40 , the expected
cost is , then : @xmath177/h= 4q^2e[\tau](2|\beta|)^{1/\delta}\frac{v[\phi]}{a^{2 + 1/\delta}}:= k\frac{v[\phi]}{a^{2 + 1/\delta}}.\ ] ] using control variates to reduce the variance , the trajectories remain the same , but the cost per timestep is increased by a factor @xmath178 .
this is due to the extra cost of interpolating the control variates from a lookup table and evaluating the extra term @xmath126 in ( [ f : vr_milstein_sys ] ) . inserting ( [ f : variance_reduction ] ) ,
the pointwise cost with control variates is @xmath179(1-\rho^2[\phi,\xi])}{a^{2 + 1/\delta}}.\ ] ] the cost of a global pdd approximation @xmath180 with plainpdd(@xmath47 ) is then @xmath181}{a_0^{2 + 1/\delta}},\ ] ] @xmath182 is the cost ( independent of @xmath47 ) involved in solving all the subdomains once the interfacial values are available . finally , @xmath183 \bigg ( \big(\frac{a_0}{a_1}\big)^{2 + 1/\delta } + \kappa\big(1-\rho_i^2[\phi,\xi]\big ) \bigg).\ ] ] above
, @xmath184 is the cost ( independent of @xmath47 and @xmath64 ) of constructing and storing @xmath185 with @xmath186plainpdd@xmath187 , and @xmath188 $ ] depends on @xmath64 only as long as there are no quadrature errors ( ie as @xmath116 ) .
the results in this subsection are exact ( within the assumptions ) in the limit @xmath116 ; approximations are left to section [ s : approximations ] . to the best of our knowledge
, they are also new .
let us introduce the following notation : * @xmath190 , with @xmath191 means that @xmath192 is some realization of the distribution @xmath193 .
the notation @xmath194 denotes a specific realization , labeled with the chance variable @xmath195 .
thus , @xmath194 is the same as @xmath190 , but the former considers @xmath196 fixed and treats @xmath192 as a scalar ( or vector ) .
* @xmath197 is an @xmath198-dimensional standard normal distribution , with @xmath199 ( by default ) , or known from the context .
for instance , @xmath200 is the realization labeled by @xmath195 . *
the notation @xmath201 means that the stochastic variable @xmath192 ( which may depend on several parameters represented by the first argument ) is constructed based on a pdd simulation with nodal statistical errors labeled with @xmath202 , where @xmath203 labels the statistical error on node @xmath204 .
( an arbitrary node is denoted by @xmath175 and @xmath205 . )
[ th : lema1 ] let @xmath7 and @xmath192 be stochastic variables , @xmath58 an exact control variate for @xmath7 , and assume that @xmath206,v[\xi]$ ] and @xmath207 $ ] are finite .
it holds : 1 .
@xmath208=v[\phi]$ ] .
@xmath209=-v[\phi]\leq 0.$ ] 3 .
@xmath210=-cov[\phi,\eta]$ ] . _ proof .
_ we will make use of the cauchy - schwarz inequality @xmath211|\leq \sqrt{v[a]v[b]}$ ] for @xmath40,@xmath212 with finite variances .
notice first that @xmath213=0=v[\phi]+v[\xi]+cov[\phi,\xi]$ ] , so that @xmath209 $ ] can not be positive .
for the first result , note that @xmath214|=-cov[\phi,\xi]$ ] , and using the cauchy - schwarz inequality , @xmath215v[\xi]}-2cov[\phi,\xi]\leq 0 $ ] . summing this and @xmath206+v[\xi]+2cov[\phi,\xi]=0 $ ] yields @xmath216 + v[\xi ] - 2\sqrt{v[\phi]v[\xi]}= \big(\sqrt{v[\phi ] } - \sqrt{v[\xi ] } \,\big)^2\ ] ] which can only happen if @xmath206=v[\xi]$ ] .
then , it is clear that @xmath217=-2v[\phi]$ ] .
finally , @xmath218|\leq \sqrt{v[\phi+\xi]v[\eta]}=0 $ ] , so that @xmath219=cov[\phi,\eta]+cov[\xi,\eta]=0 $ ] and the third result follows .
@xmath220 the central limit theorem behind the estimates for the statistical error in ( [ f : statistical_error ] ) assumes that monte carlo errors are normally distributed .
moreover , they are biased due to discretization .
lemma [ th : lema3 ] makes this assumption explicit .
[ th : lema3 ] let @xmath221/n}{\cal n}$ ] .
then @xmath222=\beta h^{\delta}$ ] and @xmath223=\beta^2h^{2\delta}+v[\phi]/n$ ] .
_ the first moment is trivial . for the mse ,
just notice that @xmath223= e[(\beta h^{\delta}+\sqrt{v[\phi]/n}{\cal n})^2]= \beta^2h^{2\delta } + ( v[\phi]/n)e[{\cal n}^2 ] + \beta\sqrt{v[\phi]/n}e[{\cal n}]$ ] .
the distribution of @xmath224 defined in lemma [ th : lema3 ] results from combining the central limit theorem in ( [ f : statistical_error ] ) with the @xmath225 in ( [ f : mse_error ] ) .
it is natural enough and further justified by the match of the first two moments of the error , but rigourous only asymptotically as @xmath116 . in order to track errors from the nodal values into the subdomains
, one needs to know how they are propagated by the interfacial interpolators .
definition [ d : linear_interpolator ] introduces the relevant notation and properties .
[ d : linear_interpolator ] let @xmath226 be real constants , @xmath227 a set of @xmath228 distinct points in @xmath229 ; @xmath230 and @xmath231 two sets of scalars associated to the points in @xmath232 , and @xmath233 a function .
let @xmath234({\bf x}):\gamma\mapsto{\mathbb r}$ ] be a smooth approximation to @xmath235 obtained by interpolation of @xmath236 .
we say that @xmath237 is a linear interpolator if @xmath238({\bf x})= \lambda r[u_i\,|\,{\bf x}_i\in\gamma]({\bf x } ) + \mu r[v_i\,|\,{\bf x}_i\in\gamma]({\bf x } ) , \qquad { \bf x}\in{\gamma}.\ ] ] for some norm @xmath239 in @xmath232 we call the interpolation error @xmath240-z||.\ ] ] the most common interpolation schemes are linear in the sense of definition [ d : linear_interpolator]for instance , rbf interpolation @xcite .
[ d : error_propagation_function ] the error propagation function @xmath241 for the elliptic bvp ( [ f : ellipticbvp ] ) in the subdomain @xmath242 is defined as @xmath243 where @xmath244 and @xmath245 are respectively the solution of the deterministic and stochastic bvps @xmath246({\bf x } ) \\
{ \tilde w}_k({\bf x};{\bm\omegaup})= r[\,{\cal n}(\omega_i)\,|\,{\bf x}_i\in\gamma_j^k\,]({\bf x } ) \end{array } \right\ } & \textrm { if } & { \bf x}\in\gamma_j^k,\,1\leq j \leq m_k . \end{array } \right.\ ] ] [ d : definition_of_psi ] let @xmath53 , @xmath52 , @xmath247 , @xmath248 , and @xmath54 be the same as in ( [ f : xicontrolvariate ] ) .
for a fixed @xmath249 , @xmath250 thanks to the smoothness of the iterfacial interpolator , gradients inside @xmath20 are well defined , and thus the integral in ( [ f : definition_of_psi ] ) is also well defined regardless of the continuity of @xmath251 across the interfaces .
we are now prepared to state the main theoretical result .
[ th : teorema ] assume an elliptic bvp with dirichlet bcs like ( [ f : ellipticbvp ] ) with exact solution @xmath37 , and let @xmath252 be a pdd simulation of it in the limit @xmath116 , with accuracy @xmath253 , nodal statistical errors labeled by @xmath196 , balanced mc simulations , and smooth linear interpolator @xmath237 . let @xmath254 be the control variate at a given interfacial node @xmath255 constructed from @xmath252 according to ( [ f : xicontrolvariate ] )
. then @xmath256 } v[\xi(a;{\bm\omegaup } ) ] = v[\phi]-cov[\phi,\psi(\bm\omegaup)]\,a+v[\psi(\bm\omegaup)]\big(\frac{a}{2}\big)^2 + { \cal o}\bigg(\sum_{k=1}^m\sum_{j=1}^{m_k}\epsilonup_u|_{\gamma_j^k}\bigg ) + { \cal o}\bigg(\sum_{k=1}^m\epsilonup_{\omega_k}(a)\bigg)\ ] ] and @xmath257 = -\sqrt{\frac{v[\phi]}{v[\xi(a;{\bm\omegaup } ) ] } } + \frac{cov[\phi,\psi({\bm\omegaup})]}{\sqrt{v[\phi]v[\xi(a;{\boldsymbol\omega})]}}\frac{a}{2 } + { \cal o}\bigg(\sum_{k=1}^m\sum_{j=1}^{m_k}\epsilonup_u|_{\gamma_j^k}\bigg ) + { \cal o}\bigg(\sum_{k=1}^m\epsilonup_{\omega_k}(a)\bigg),\ ] ] where @xmath258 is the auxiliar variate defined by ( [ f : definition_of_psi ] ) , @xmath259 the error of interpolating @xmath155 along an interface @xmath232 , and @xmath260 is the error of the pdd subdomain solver . _
_ let us consider the integral representation of the solution of ( [ f : ellipticbvp ] ) @xmath261 where @xmath262 and @xmath263 is green s function , defined as the solution of @xmath264 under the adequate smoothness requirements on @xmath81 and @xmath96 the solution @xmath263 to the homogeneous bvp ( [ f : green_g ] ) exists and is unique , which ensures the validity of ( [ f : green_representation ] ) @xcite .
consider the solution restricted to subdomain @xmath20 . since @xmath263 does not depend on the boundary data , @xmath265 can also be represented as @xmath266 on the other hand , the pdd subdomain approximation is affected by the error of the subdomain solver , @xmath267 , which depends on @xmath40 due to the bc along the interfaces : @xmath268 in the limit @xmath116 , the nodal values are given by the distribution in lemma [ th : lema3 ] .
running a pdd simulation with balanced mc simulations , tolerance @xmath40 and probability @xmath41 , is then equivalent to fixing @xmath196 and taking @xmath269 ( note that the statistical error , and hence @xmath196 , are only known after running the pdd simulation . )
the dirichlet bc condition on an interface @xmath232 is then @xmath270({\bf x})= u|_{\gamma}({\bf x})+\frac{a}{2}r\big[\,sign\big(\beta({\bf x}_i)\big)+\frac{1}{q}{\cal n}(\omega_i)\,|\,{{\bf x}_i\in\gamma}\,\big]({\bf x } ) + { \cal o}(\epsilonup_u|_{\gamma}).\ ] ] inserting ( [ f : v_restricted_to_gamma ] ) into ( [ f : greens_representation_of_vk ] ) : @xmath271({\bf
y})d^{d-1}{\bf y}+\\\nonumber + { \cal o}\bigg(\sum_{j=1}^{m_k}\epsilonup_u|_{\gamma_j^k}\bigg ) + \epsilonup_{\omega_k}(a).\end{aligned}\ ] ] since @xmath237 is a smooth interpolator , gradients are well defined inside @xmath20 and along the interfaces , but not necessarily across them . in order to circumvent this issue
we take the direct sum of subdomain gradients , @xmath272 where @xmath273({\bf y})d^{d-1}{\bf y}\bigg).\ ] ] note that the quantity in parentheses in ( [ f : aux1 ] ) is the green representation of the bvp with solution @xmath274 ( [ f : error_propagation_function ] ) , by linearity . now , according to definitions ( [ f : xicontrolvariate ] ) and ( [ f : definition_of_psi ] ) for node @xmath175 , @xmath275 in ( [ f : xi(a ) ] ) , we have used the fact that @xmath276 except on the interfaces .
since the interfaces are smooth and the random paths are not , the trajectories cross the interfaces at isolated points which make no contribution to the integral , so that @xmath277 .
taking the variance of ( [ f : xi(a ) ] ) and using lemma [ th : lema1 ] yields ( [ f : v[xi(a , w ) ] ] ) .
moreover , @xmath278= cov[\phi,\xi ] + \frac{a}{2}\,cov[\phi,\psi({\boldsymbol\omega ) } ] + { \cal o}\bigg(\sum_{k=1}^m\sum_{j=1}^{m_k}\epsilonup_u|_{\gamma_j^k}\bigg ) + { \cal o}\bigg(\sum_{k=1}^m\epsilonup_{\omega_k}(a)\bigg).\ ] ] finally , equation ( [ f : rho(a , w ) ] ) follows from @xmath279=cov[\phi,\xi(a;\boldsymbol\omega)]/\sqrt{v[\phi]v[\xi(a;{\boldsymbol\omega})]}$ ] recalling that the correlation with the exact control variate is @xmath280 by lemma [ th : lema1 ] .
@xmath281 + the point of lemma [ th : teorema ] is to predict the correlation between the score @xmath7 and an approximate control variate @xmath282 without actually running a pdd simulation to produce the latter but rather `` simulating '' it . in exchange ,
the variable @xmath283 must be computed , but it can be constructed on a subdomain - per - subdomain basis , i.e. in a fully parallellizable way .
the drawback is that the formulas derived so far are only rigourous at @xmath284 , and that they depend on quite a few problem - dependent constants many of them node - dependent as well .
these difficulties will be addressed in section [ s : approximations ] .
but before that , we shall examine the issues of global error and stability of pdd , and introduce some more results which will later be useful in assessing the necessary simplifications .
the main tool is theorem 3.7 in gilbarg and trudinger : [ th : gandt ] let @xmath285 be the solution of an elliptic bvp like ( [ f : ellipticbvp ] ) with @xmath84 , such that @xmath285 is continuous on @xmath286 and twice differentiable on @xmath96 .
then @xmath287 where @xmath288 is a constant depending only on @xmath289 and @xmath290 .
in particular , if @xmath0 lies between two parallel planes a distance @xmath291 apart , then ( [ f : gandt ] ) is satisfied with @xmath292}-1 $ ] .
recall that @xmath293 $ ] , and @xmath289 is the largest distance between two points in @xmath0 .
it is convenient to introduce the following parameter : [ d : overshoot ] let @xmath237 be a linear interpolator and @xmath232 a subdomain interface with nodes @xmath294 .
let @xmath295 be @xmath296 scalars such that @xmath297 .
the overshoot constant of @xmath237 with respect to the discretization @xmath298 of @xmath232 is defined as @xmath299({\bf x})\big|.\ ] ] analogously , let @xmath300 and @xmath301 .
the overshoot constant measures the excess of the reconstructed function over any of the interpolated values which are @xmath302 in absolute value . for piecewise interpolation , @xmath303 , but for more accurate interpolators
, @xmath304 due to the runge and gibbs phenomena @xcite see figure [ i : figura3 ] ( right ) for illustration .
we are interested in bounding the largest pdd error throughout @xmath0 .
assuming as always balanced mc simulations , and neglecting the error of the subdomain solver , the pdd error in subdomain @xmath20 obeys @xmath305({\bf x } ) & \textrm{if $ { \bf x}\in\gamma_j^k,\,1\leq j\leq m_k.$ } \end{array } \right.\ ] ] therefore , @xmath306hence the name of error propagation function . applying theorem [ th : gandt ] to ( [ f : error_pde ] ) and by linearity of @xmath237 , @xmath307({\bf x})\big| \leq \frac{aq_k\gamma_r^{\omega_k}}{2}\big ( 1 + \frac{1}{q}\sup_{{\bf x}_i\in\omega_k}|{\cal n}(\omegaup_i)|\big),\ ] ] where @xmath308 depends on the size and shape of @xmath20 and @xmath309 .
thanks to the symmetry of the standard normal distribution around zero , @xmath310 , so that the _ global _ pdd error ( neglecting the error of the subdomain solver ) can be bounded by @xmath311 equation ( [ f : global_pdd_error_bound ] ) reflects that the global pdd error depends on pde coefficients in ( [ f : ellipticbvp ] ) , on the shape and size @xmath312 of the subdomains ( typically , @xmath313 ) , and on the number of nodes per subdomain ( typically , around @xmath314 ) , rather than on @xmath315 and @xmath316 . therefore , pdd is intrinsically stable provided that the number , shape and discretization of the subdomains are so chosen that the subdomain errors are controlled
. we will now derive the nodal tolerance @xmath317 required to enforce a set global pdd error tolerance @xmath318 . recall that @xmath319 is the total number of interfacial nodes sitting on @xmath19 .
the distribution @xmath320 is an _ extreme value _ distribution .
the value @xmath319 for which its maximum is less than @xmath321 with probability @xmath41 is to be extracted from @xmath322^{-1}(p_q ) \textrm { ( such that } pr[\sup_{1\leq j\leq n_k } { \cal n}(\omega_j ) < x ] = p_q.)\ ] ] where @xmath323 $ ] and @xmath323^{-1}$ ] stand for the cumulative distribution function ( cdf ) of a given distribution and its inverse , respectively .
let @xmath324^{-1}(p_q)\big)\ ] ] and define @xmath325 since the @xmath198 standard normals in @xmath326 are i.i.d . , the cdf is @xmath327(x)= \int_{-\infty}^x s_s(t ) dt= pr[\big({\cal n}_1\leq x\big)\cap \ldots \cap \big({\cal n}_s\leq x\big)]= \big ( cdf[{\cal n}](x ) \big)^s . \ ] ] for the standard normal distribution @xmath328(x)= \frac{1}{2\pi}\int_{-\infty}^xe^{-t^2}dt= \frac{1}{2}\big ( 1+\textrm{erf}(x/\sqrt{2 } ) \big),\ ] ] where erf@xmath329 is the error function .
the nodal target tolerance @xmath47 can then be related to the global pdd error tolerance by @xmath330 as figure [ i : figura2 ] shows , @xmath331 depends very mildly on the typical number of nodes per subdomain .
anyway , the bound ( [ f : a(eps ) ] ) will be a large overestimation in many cases , for the effect of the statistical errors decays fast away from the interfaces .
if the moments of @xmath332 were required , a useful fact is that as @xmath198 grows , @xmath326 tends to the gumbel distribution @xmath333(x)= g\big(\frac{x - l_s}{b_s}\big):= \exp[-e^{-x}]\ ] ] with location and scaling parameters @xmath334^{-1}(1/s)$ ] and @xmath335 . from ( [ f : a(eps ) ] ) as a function of @xmath198 for several values of @xmath336 and @xmath337 ( notice the different scales ) . ]
in order to construct an implementable multigrid - like iterpdd algorithm , several approximations are needed , which are listed as heuristics @xmath338 through @xmath339 below .
this assumption is used on several occasions in * h2 * -*h5*. interfacial interpolation errors in ( [ f : v[xi(a , w ) ] ] ) and ( [ f : rho(a , w ) ] ) will be dropped .
then , inserting ( [ f : v[xi(a , w ) ] ] ) into ( [ f : rho(a , w ) ] ) yields , after some manipulation , @xmath341 \gtrsim \frac{\frac{v[\psi({\bm\omegaup})]}{v[\phi]}\big(1-\rho^2[\phi,\psi(\bm\omegaup)]\big)(a/2)^2 } { 1 -\sqrt{\frac{v[\psi(\bm\omegaup)]}{v[\phi]}}\rho[\phi,\psi(\bm\omegaup)]a + \frac{v[\psi(\bm\omegaup)]}{v[\phi]}(a/2)^2},\ ] ] where the sign @xmath342 has replaced the equality to make up for dropping the interpolation and subdomain - solver errors . next , note that @xmath343}{v[\phi ] } \gtrsim 1-\rho^2[\phi,\xi(a;{\bm\omegaup})],\ ] ] since ( [ f : variance_reduction ] ) only strictly holds for the minimizer of ( [ f : statistics_vr ] ) , and with @xmath253 , the pdd solution @xmath136 yields a control variate off the minimizer , regardless of @xmath104 . note also that the denominator in ( [ f : aux2 ] ) is positive , since @xmath344|\leq 1 $ ] : @xmath345}{v[\phi]}}\rho[\phi,\psi(\bm\omegaup)]a + \frac{v[\psi(\bm\omegaup)]}{v[\phi]}(a/2)^2 \geq \bigg(1-\sqrt{\frac{v[\psi({\bm\omegaup})]}{v[\phi]}}(a/2)\bigg)^2.\ ] ] for small @xmath40 ( see * h1 * ) , the denominator in ( [ f : aux2 ] ) can be dropped .
more precisely , since the mc simulations are balanced , @xmath346}{v[\phi]}}\frac{a}{2}=q\sqrt{\frac{v[\psi({\bm\omegaup})]}{n}},\ ] ] which is negligible if @xmath347<<n$]recall that this @xmath348 is meant without variance reduction , and @xmath347={\cal o}(v[\bar\psi])$ ] is bounded by ( [ f : v[barpsi ] ] ) in * h4 * below . assuming this and putting all together , it holds @xmath349}{v[\phi]}\gtrsim \frac{v[{\psi({\bm\omegaup})}]a^2}{4v[\phi]}\big ( 1-\rho^2[\phi,{\psi({\bm\omegaup})}]\big).\ ] ] the importance of ( [ f : the_key ] ) is that the two factors affecting iterpdd namely @xmath40 and the random statistical errors on the nodes have been separated .
moreover , the latter has been expressed in terms of @xmath347 $ ] and @xmath350 $ ] . also , * if the score @xmath7 and the auxiliar variate @xmath283 were perfectly correlated ( i.e. @xmath351=1 $ ] ) , then @xmath352=0 $ ] , meaning that @xmath353 . since this can only happen
if @xmath354 , the solution would be @xmath355 .
* the opposite limit , @xmath356=0 $ ] , yields asymptotically @xmath357}{v[\phi]}= \frac{v[\psi({\bm\omegaup})]}{v[\phi]}(a/2)^2.\ ] ] the propagation of nodal statistical errors ( labeled with @xmath196 ) onto the subdomain solutions in critical in pdd . in section [ s : globalerror ] , it was shown that the effect of @xmath196 on the pdd aggregate error can be controlled on a subdomain - per - subdomain basis . in iterpdd(@xmath359 ) , there are two further aspects to @xmath196 .
first , how much the variance reduction produced by @xmath360 depends on the chance variable .
second , how it affects the decrease predicted by ( [ f : the_key])for @xmath64 will be determined based on that formula .
in particular , the nodal simulations will in turn be `` simulated '' themselves by randomly drawing a chance variable
say @xmath361and computing @xmath362 $ ] and @xmath363 $ ] .
[ th : e[w ] ] let @xmath364 $ ] be the expected value relative to @xmath358 and assume that interpolation errors are negligible .
then , @xmath365=0 $ ] ( @xmath366 ) . _ proof . _
the green function representation of @xmath367 is @xmath368({\bf y})d^{d-1}{\bf y},\ ] ] where @xmath263 is determined by ( [ f : green_g ] ) and is deterministic . by linearity of the integral , the interpolator , and the expected value , @xmath369= \sum_{j=1}^{m_k}\int_{\gamma_j^k } \frac{\partial g({\bf x},{\bf y})|_{\omega_k}}{\partial { \bf n}}r\big[e_{\bm\omegaup}[{\cal n}({\omegaup_i})]\,|\,{\bf x}_i\in\gamma_j^k\big]({\bf y})d^{d-1}{\bf y}= \\\nonumber \sum_{j=1}^{m_k}\int_{\gamma_j^k } \frac{\partial g({\bf x},{\bf y})|_{\omega_k}}{\partial { \bf n } } \big ( 0 + \epsilon_0|_{\gamma_j^k } \big)d^{d-1}{\bf y } , \end{aligned}\ ] ] where @xmath370 is the error in the reconstruction of the constant function @xmath371 on the interface @xmath31 with @xmath237 , and which is zero by hypothesis .
@xmath372 as a consequence of lemma [ th : e[w ] ] , @xmath373=\nabla e_{\bm\omegaup}[{\tilde w}_k]=0 $ ] , and thus @xmath374= \int_0^{\tau}y_t\sigma^t\sum\limits_{k=1}^m\nabla\big({\bar w}_k+\frac{1}{q}e_{\bm\omegaup}[{\tilde w}_k]\big)d{\bf w}_t= \int_0^{\tau}y_t\sigma^t\sum\limits_{k=1}^m\nabla{\bar w}_kd{\bf w}_t= : { \bar \psi}.\ ] ] the variance can be calculated by ito s isometry : @xmath375= e\big[\,\int_0^{\tau } y_t^2({\bf x}_t)||\sigma^t({\bf x}_t)\sum\limits_{k=1}^m\nabla w_k({\bf x}_t,{\bm\omegaup})||^2_2\,dt\,\big],\ ] ] where @xmath376 . by the triangular inequality and the inequalities @xmath377 }
e_{\bm\omegaup}\big[\ , v[\psi({\bm\omegaup } ) ] \,\big]\leq v[{\bar \psi } ] + \frac{2\lambda\,e\big[\int_0^{\tau}y_t^2({\bf x}_t)dt\,\big]}{q^2 } \,e_{\bm\omegaup}\big[\,\big(\sup_{{\bf x}\in\omega}||\sum\limits_{k=1}^m{\nabla\tilde w}_k({\bf x};{\bm\omegaup})||_2 ^ 2\big)\,\big].\end{aligned}\ ] ] to the best of the authors knowledge , there are no interior , a priori estimates of the gradient of the solution of ( [ f : ellipticbvp ] ) ( with @xmath378 ) as sharp as the bound provided by ( [ th : gandt ] ) for the solution itself . therefore , based on more particular results such as and
, we make the reasonable assumption that for the pde @xmath379 in @xmath20 with dirichlet bcs on @xmath19 @xmath380 where @xmath381 and @xmath382 are positive and may depend on anything but the value of the dirichlet bc .
then , there exist positive constants @xmath383 and @xmath198 such that @xmath384 the values @xmath385 are the @xmath386 of the subdomain @xmath387 with the largest gradient estimate , and @xmath198 its number of nodes
. the @xmath388 shows up due to the interpolation along the interfaces ; and @xmath389 .
this leads to the bounds @xmath390 \leq v[{\bar\psi } ] + 2\lambda e[\int_0^{\tau } y_t^2dt]\frac{1}{q^2}\big ( k ' + 2\gamma_r k''\sup\limits_{1\leq j \leq s } { \cal n}(\omegaup_j ) \big)^2 = : v[{\bar\psi } ] + { { \tilde v}({\bm\omegaup})},\ ] ] @xmath391 } v[{\bar\psi}]\leq 2\lambda e[\int_0^{\tau } y_t^2dt](k'+\gamma_r k'')^2 = : { \bar v}.\ ] ] as the final preparatory step , let us calculate the noise - to - signal ratio ( nsr)defined as @xmath392=\sqrt{v[\cdot]}/e[\cdot]$]of the variable @xmath393 ( [ f : nsr ] ) : @xmath394}}{e_{{\bm\omegaup}}[{\bar v}+{\tilde v}({\bm\omegaup})]}= \frac{\sqrt{v_{\bm\omegaup}\big[\big(1 + 2\gamma_r\frac{\textrm{k'}}{\textrm{k''}}\sup\limits_{1\leq j \leq s}{\cal n}(\omegaup_j)\big)^2\big ] } } { q^2\big(1+\gamma_r\frac{\textrm{k'}}{\textrm{k''}}\big)^2+e_{\bm\omegaup}\big [ \big(1 + 2\gamma_r\frac{\textrm{k'}}{\textrm{k''}}\sup\limits_{1\leq j \leq s}{\cal n}(\omegaup_j)\big)^2 \big]}.\ ] ] @xmath395 we are finally in a position to precisely estate our claim and the supporting heuristic .
the ultimate goal is to use the deterministic value @xmath396 $ ] instead of a random @xmath347 $ ] yielded by one simulation ( note that @xmath396\leq e_{{\bm\omegaup}}\big[v[\psi({\bm\omegaup})]\big]$ ] ) .
therefore , it is important that the ratio @xmath397\big]}/e_{{\bm\omegaup}}\big[v[\psi({\bm\omegaup})]\big]$ ] be small so that @xmath398:=v[{\bar\psi}]\approx v[\psi({\bm\omegaup})]$ ] .
this ratio is problem - dependent but , in order to provide a rough estimate , we substitute @xmath347 $ ] by its upper bound @xmath393 .
then , we simulate the nsr of the bound ( table [ t : nsr ] ) for realistic values @xmath399 and over a broad range of @xmath400 ( which captures the effect of the @xmath81 , @xmath51 , the geometry @xmath0 , and the pdd partition @xmath401 ) , and the typical number of nodes per subdomain , @xmath198 .
since the nsr of the proxy is negligible in most of the scenarios ( and specially as @xmath198 grows ) , we argue that the same should hold for @xmath347 $ ] .
obviously , specific problems would allow for sharper estimates .
given @xmath47 and its corresponding timestep @xmath402 from ( [ f : n_y_h ] ) , the correlation between @xmath7 and @xmath403 is better than between their discretized counterparts @xmath404 and @xmath405 .
this leads to an overestimation of the predicted decrease of variance and has a significant effect , especially if @xmath40 and @xmath47 are comparable , or if @xmath151\gtrsim.99 $ ] .
we invoke * h1 * to justify the following perturbative ansatz : @xmath406\approx cov[\phi,\xi(a;{\bm\omegaup})]+b_{\xi}\big ( cov[\phi_{h_0},\xi_{h_0 } ] \big),\ ] ] where @xmath407 \big)$ ] is the discrete covariance bias . since @xmath408=e[\phi_{h_0}\xi_{h_0}]-u_{h_0}e[\xi_{h_0}])]$ ] and @xmath402 is small enough , @xmath407 \big)= { \cal o}(h_0^{\delta})$ ] .
that covariance bias does not seem accessible without having @xmath409 , but we can use the fact that @xmath209=-v[\phi]$ ] ( lemma [ th : lema1 ] ) to argue that @xmath410 \big)\big|\approx \big|b_{\xi}\big ( v[\phi_{h_0 } ] \big)\big|=:|\alpha|h^{\delta},\ ] ] which can be extracted from a fit ( see * h5 * ) . then , assuming further that @xmath411}\approx \frac{1}{v[\phi ] } \qquad\textrm { and } \qquad \frac{1}{v[\xi_{h_0}(a;{\bm\omegaup})]}\approx \frac{1}{v[\xi(a)]},\ ] ] the squared correlation of the discretized variables is @xmath412 \approx \rho^2[\phi,\xi(a;{\bm\omegaup } ) ] - \frac{2\big|\alpha\rho[\phi,\xi(a;{\bm\omegaup})]\big|}{v[\phi]}h_0^{\delta},\ ] ] and since @xmath413 , the effective variance reduction in iterpdd(@xmath414 ) is @xmath415}{v[\phi_{h_0}]}\approx \frac{v[\phi+\xi(a;{\bm\omegaup})]}{v[\phi ] } + \big|\frac{\alpha\rho[\phi,\xi(a)]}{\beta v[\phi ] } \big|a_0.\ ] ] for @xmath47 small enough , as @xmath416 , @xmath417|\shortrightarrow 1 ^ -$ ] , so ( [ f : final_formula ] ) is capped by @xmath418}{v[\phi_{h_0 } ] } = \big| \frac{\alpha}{\beta v[\phi ] } \big|a_0= \frac{2|b_{\xi}(v[\phi])|}{v[\phi]}. \end{aligned}\ ] ] this makes intuitively sense , because the variance of the score can hardly drop below its discretization error , even with an exact control variate . in order to apply the sensitivity formula ( [ f : sensitivity_formula ] ) and algorithms [ a : iterpdd ] and [ a : scheduling ] ,
a number of constants must be estimated . here
, we describe a fast way of accomplishing this .
we stress the fact that any monte carlo simulation ( whether or not related to pdd ) also would require @xmath419 and @xmath206 $ ] in order to enforce a set error tolerance . * global constants ( @xmath43 and @xmath420 ) .
* as already discussed , with smooth bvps , integrators with at least approximately known @xmath43 should be available @xcite .
the constant @xmath420 can easily be found by comparing the time taken by the computer to complete a number of visits to the implementations of ( [ f : vr_milstein_sys ] ) with and without @xmath126 . * nodal constants related to first moments .
* they are @xmath174 $ ] ( needed for @xmath421 ) , @xmath422 $ ] , and @xmath106 .
we consider a generic discretized random variable @xmath423 , and assume that its first moment obeys the noisy model @xmath424\sim e[\eta_0 ] + { \cal n}(b^{(1)}h^{\delta},v[\eta_0]).\ ] ] let @xmath425 be a cloud of @xmath426 equispaced timesteps . set a number @xmath427 and let @xmath428 $ ] be the mean of @xmath427 realizations of @xmath429 . after computing @xmath426 mc independent simulations ,
the cloud of data @xmath430\big),\ldots,\big(h_{\hat m},{\hat e}[\eta_{h_{\hat m}}]\big)$ ] is fitted to the noisy model ( [ f : noisy_model_moment1 ] ) in order to extract @xmath431 $ ] and @xmath432 .
( in order to do so , @xmath433 $ ] in ( [ f : noisy_model_moment1 ] ) can be replaced by the mean of the sample variances @xmath434 $ ] . ) here , we provide a rougher recipe to carry out the fit with matlab ( see also @xcite ) . for this purpose , it is convenient to think of model ( [ f : noisy_model_moment1 ] ) as a member of the generalized linear model ( glm ) family .
since @xmath435)]=b^{(1)}h^{\delta}$ ] , the link is the identity and the fit is readily carried out by issuing the matlab command .... coeff= glmfit((h.^delta),eh,'normal','link','identity ' ) .... where ` h ` and ` eh ` above are matlab arrays with respectively @xmath436 and @xmath437, ... ,{\hat e}[\eta_{h_{\hat m}}]\,)$ ] ; and the components of the output , ` coeff(1 ) ` and ` coeff(2 ) ` , are the fitted values to @xmath431 $ ] and @xmath432 , respectively ( check the matlab documention for getting error bounds alongside ) . by applying this recipe ,
one gets approximations to the following quantities : if @xmath438 , to @xmath422\approx { \hat cov}[\phi,{\bar\psi}]= \verb|coeff(1)|$ ] ( and to its bias as a byproduct ) ; and if @xmath439 , to @xmath440 and to @xmath102\approx{\hat e}[\phi]=\verb|coeff(1)|$ ] as a byproduct .
the values @xmath441, ...
,{\hat e}[\tau_{h_{\hat m}}]$ ] are obtained along the means @xmath442, ...
,{\hat e}[\phi_{h_{\hat m}}]$ ] ( see last line in algorithm [ a : gobet_menozzi ] ) , and are used to fit @xmath174 $ ] .
* nodal constants related to second moments . * regarding the fitting of variances , it is well - known that variances of i.i.d .
gaussian distributions obey a scaled chi - squared pdf . accommodating the discretization bias ,
the appropriate noisy model is @xmath443\sim b^{(2)}h^{\delta } + \frac{v[\eta_0]}{{\hat
n}-1}\chiup^2_{{\hat n}-1}= b^{(2)}h^{\delta } + \gamma\big ( \frac{{\hat n}-1}{2},\frac{2v[\eta_0]}{{\hat n}-1}\big)\ ] ] where @xmath444 is the chi - squared distribution with @xmath445 degrees of freedom , and @xmath446 is a gamma distribution with shape parameter @xmath447 and scale parameter @xmath448/({\hat n}-1)$ ] .
then , ( [ f : noisy_model_moment2 ] ) can be identified with a glm where the noise is gamma and the link function is the identity , since @xmath449\big]= b^{(2)}h^{\delta } + e\big[\gamma\big ( \frac{{\hat n}-1}{2},\frac{2v[\eta_0]}{{\hat n}-1}\big)\big]= b^{(2)}h^{\delta}+v[\eta_0].\ ] ] let @xmath450= \sum_{j=1}^{{\hat n}}(\eta_h^{(j)}-{\hat e}[\eta_h])^2\big/({\hat n}-1)$ ] be the sample variance .
after filling the matlab array ` vh ` with @xmath451,\ldots,{\hat v}[\phi_{h_{\hat m}}]$ ] , issuing the command .... coeff= glmfit((h.^delta),vh,'gamma','link','identity ' ) .... yields ` coeff(1)`@xmath452/({\hat n}-1)$ ] and ` coeff(2)`@xmath453 .
particularizing to @xmath439 allows to estimate @xmath454\approx v[\phi]$ ] and @xmath455 ; and to @xmath456 yields the fitted @xmath454\approx v[{\bar\psi}]$ ] ( and its bias ) .
[ a : fitting ] the full fitting procedure is sketched as algorithm [ a : fitting ] .
note that the same sets of trajectories can be used for all the poinwise constants it is only the functionals that change . combining heuristics
* h1 * through * h5 * we put forward the sensitivity formula ( [ f : sensitivity_formula ] ) , based on ( [ f : the_key ] ) and ( [ f : final_formula ] ) with @xmath457 replacing @xmath283 : @xmath458\approx \rho^2[\phi,\xi]-\big|\frac{\alpha\rho[\phi,\xi]}{\beta}\big|a_0 , \textrm { with } \rho^2[\phi,\xi]\approx \frac{v[{\bar\psi}]a^2}{4v[\phi]}\big ( 1-\rho^2[\phi,{\bar\psi}]\big).\ ] ] with it , a scheduling algorithm could be as algorithm [ a : scheduling ] .
there , the stopping criterion should take into account the quality of the available estimates and thus be problem - dependent .
( we discuss this aspect further on section [ s : experiment ] . )
[ a : scheduling ] finally , we summarize the new version of pdd in algorithm [ a : iterpdd ] . [ a : iterpdd ]
we consider a bvp like ( [ f : ellipticbvp ] ) with on the two - dimensional domain sketched in figure [ i : figura1 ] with @xmath459 subdomains , three interfaces and @xmath12 nodes .
the pde is @xmath460 with @xmath461 such that @xmath462 is the exact solution , as well as the dirichlet bc
see figure [ i : figura3 ] .
the coefficients of the stochastic representation of ( [ f : pde_ejemplo ] ) are : @xmath463 ( @xmath464 is the two - dimensional identity matrix ) , @xmath465\,/\big(1.1+\sin{(x+y)}\big)$ ] , and @xmath466 .
the integrator is gobet and menozzi s ( algorithm [ a : gobet_menozzi ] ) , for which we assume that @xmath44 ; @xmath237 is an rbf interpolator with multiquadrics @xcite , and the subdomain solver is fem .
the parameters of @xmath237 and fem were so chosen that their errors are negligible compared with the nodal errors . ) . *
* the bias - related function @xmath467 ( [ f : error_propagation_function ] ) used to generate @xmath457 ( [ f : definition_of_psi ] ) on the subdomain singled out in figure [ i : figura1 ] .
notice the various signs of @xmath468 and the overshoots of the rbf interpolator . in this problem , @xmath469 .
] with just four small subdomains , it obviously does not take a parallel computer to solve this toy problem but it serves to test the idea , theory and approximations introduced in this paper in a controlled environment . with that purpose ,
the algorithms have been coded in fully vectorized matlab , and the nodal values and subdomain bvps solved on single processors sequentially .
@xmath470 & 0.16 & 0.16 & 0.16 & 0.17 & 0.19 & 0.47\\ \hline \textrm{cost of plainpdd$(a_1)$ } & & & 1.60\times 10 ^ 9 & 1.47\times 10 ^ 7 & 8.43\times 10 ^ 5 & 65989 \\ v[\phi_{h_1 } ] & & & 4.92 & 5.10 & 5.45 & 4.08 \\ \big ( \sum_{i=1}^n |u({\bf x}_i)-{\tilde u}_1({\bf x}_i)| \big)/n & \multicolumn{2}{c|}{\textrm{use exact $ u$ instead of $ { \tilde u}(a_1)$ } } & 0.012 & 0.06 & 0.11 & 0.77 \\
\sup_{\omega}|u-{\tilde u}_1| & & & 0.03 & 0.12 & 0.24 & 0.34 \\ \sup_{\omega}||\nabla { \tilde u}_1||_2 & & & 0.87 & 0.94 & 2.14 & 6.61 \\ \hline s(a_0,a_1 ) & 128.16 & 72.61 & 7.37 & 58.14 & 46.39 & 22.67 \\ \hline \hline \end{array}\ ] ] as a preparatory step , we carefully compute @xmath420 and the nodal constants . in particular , the precise values of @xmath471 ensure that the pointwise mc simulations are balanced .
then , we pick @xmath472 ; @xmath473 ; and run a set of @xmath474=$]iterpdd@xmath65 simulations that will serve as reference .
table [ t : illustration ] illustrates some typical quantities . on the column labeled exact , the control variate from the exact solution
@xmath475 has been used , so that all of the error is quadrature error .
this represents the maximum benefit that could possibly be extracted .
otherwise , the approximation @xmath476 is stored on a lookup table ( here , a grid ) , and interpolated from there to compute @xmath63except on the column lookup , where the lookup table has been filled with the exact @xmath477 in order to gauge purely the effect of interpolation .
since the cost of solving the subdomains and filling the lookup tables is negligible , the cost of iterpdd(@xmath359 ) is measured as : @xmath478 where @xmath479 is the number of trajectories from node @xmath204 at @xmath480 , and so on . in table
[ t : illustration ] , @xmath481 is the mean over the @xmath316 nodes of @xmath482\,\big|$ ] ; and the speedup of iterpdd over plainpdd is defined as @xmath483 over plainpdd@xmath66 .
the symbols denote the reference ( balanced ) pdd simulations . in order to show the trend ,
they have been approximated with smoothing splines ( solid color curves ) .
* dashed curves * : predicted speedup , based on the sensitivity formula ( [ f : sensitivity_formula ] ) and fast estimates of constants ( algorithm [ a : fitting ] ) .
although the latter do not lead to balanced mc simulations as with the reference simulations , their maxima calculated with algorithm [ a : scheduling ] ( vertical lines ) are coincident see also ( * ) on table [ t : schedula ] . ] on figure [ i : figura4 ] , the values of @xmath484 for the full set of reference simulations are depicted with symbols ( and joined with smoothing splines ) . as @xmath64 grows , the effect of randomness is more significant .
now , we proceed to test algorithm [ a : iterpdd ] proper .
first , we estimate the nodal constants with algorithm [ a : fitting ] , taking ( without trying to optimize in any sense ) , @xmath485 equispaced timesteps @xmath104 in @xmath486 $ ] and @xmath487 .
( the cost of this is comparatively very small . )
a good estimate @xmath488 is straightforward to produce .
based on the resulting fitted constants , we run the scheduling algorithm [ a : scheduling ] in order to get the optimal sequences @xmath489 for the same set @xmath472 as with the reference simulations before .
the minimization ( line @xmath490 in algorithm [ a : scheduling ] ) is carried out with matlab s ` fmincon ` , and stopped as soon as the predicted @xmath491 .
the results , both those predicted by algorithm [ a : scheduling ] , and those actually attained ( using the same set of @xmath492 ) are listed on table [ t : schedula ] .
the cumulative speedup is defined as the cost of iterpdd@xmath493plainpdd@xmath66 .
the speedup curves predicted by algorithm [ a : scheduling ] for @xmath472 are also plotted with dashed lines on figure [ i : figura4 ] , and their maxima highlighted with vertical lines .
a few comments are in order .
@xmath494 even though algorithm [ a : scheduling ] relies on fast estimates of the constants , non - balanced simulations , and neglects the effects of randomness , the predicted speedup curves resemble quite well those obtained from the reference simulations .
in particular , the position of the maxima of both sets are nearly coincident .
moreover , for @xmath359 small enough , algorithm [ a : scheduling ] provides estimates of the costs , speedup and mean correlation ( on figure [ i : figura4 ] and table [ t : schedula ] ) which are consistently conservative , as predicted by the theory in section [ s : approximations ] .
later , when the tolerances @xmath495 are not so small , heuristics * h1-h4 * break down , and the fact that @xmath396<e_{{\bm\omegaup}}\big[v[\psi({\bm\omegaup})]\big]$ ] begins to tell , so the estimates may not be conservative , but still they are acceptable . in any case , since most of the cumulative speedup is attained on the last iterpdd simulation i.e .
on iterpdd@xmath65 , as shown on table [ t : schedula]there is not much to gain from iterating iterpdd more than once or twice . finally , let us mention that running the scheduling algorithm with nodal constants based on @xmath496 instead of @xmath457 ( as we did ) leads to very similar results to those reported .
in fact , in this bvp , it seems that the effect of discretization on @xmath497 $ ] ( [ f : vr_cap ] ) outweighs the effect of randomness in the range of @xmath47 studied . this is why the speedups in table [ t : schedula ] grow linearly with @xmath47 .
importantly , this means that one order of magnitude has been knocked down from the monte carlo cost estimate @xmath498 given in the introduction see also ( [ f : pointwise_cost ] ) .
therefore , the nodal cost goes now as @xmath499 , which is still larger than for the deterministic subdomain solver but now in a fully scalable way .
in this paper we have laid out the theoretical foundations of a much improved version of pdd , which we have called iterpdd .
we have analyzed in detail the ingredients of iterpdd and their relative importance .
while the numerical results on the bvp used for illustration are very encouraging , substantially larger and more challenging problems must be tackled in order to assess the real potential of iterpdd . with this goal ,
the pdd programme is currently being further developed in three main directions : i ) blending it with giles multilevel method along the lines of @xcite ; ii ) extending it to parabolic bvps , mixed bcs , and processors with insufficient memory ; and iii ) producing a c++/mpi code which allows us to perform realistic comparisons with deterministic domain decomposition methods . while nonlinear bvps could be accommodated into that framework by linearization , of far greater interest
are probabilistic representations of nonlinear equations , such as in @xcite .
this work was supported by portuguese national funds through fct under grants uid / cec/50021/2013 and ptdc / eia - cco/098910/2008 .
fb also acknowledges fct funding under grant sfrh / bpd/79986/2011 .
f. bernal , j.a .
acebrn and i. anjam , _ a stochastic algorithm based on fast marching for automatic capacitance extraction in non - manhattan geometries_. siam journal on imaging sciences * 7*(4 ) , 2657 - 2674 ( 2014 ) .
s. mancini , f. bernal and j.a .
acebrn , _ an efficient algorithm for accelerating monte carlo approximations of the solution to boundary value problems_. accepted in the journal of scientific computing ( 2015 ) . g. n. milstein and m. v. tretyakov , _ stochastic numerics for mathematical physics _ , springer - verlag , berlin , 2004 . | we present an iterative scheme , reminiscent of the multigrid method , to solve large boundary value problems with probabilistic domain decomposition ( pdd ) . in it ,
increasingly accurate approximations to the solution are used as control variates in order to reduce the monte carlo error of the following iterates resulting in an overall acceleration of pdd for a given error tolerance .
the key ingredient of the proposed algorithm is the ability to approximately predict the speedup with little computational overhead and in parallel .
besides , the theoretical framework allows to explore other aspects of pdd , such as stability .
one numerical example is worked out , yielding an improvement of between one and two orders of magnitude over the previous version of pdd .
* keywords : * pdd , domain decomposition , scalability , high - performance supercomputing , variance reduction , feynman - kac formula . |
the five stars considered here are pulsating variables and show the sorts of light curve irregularities that are generally associated with classification as semiregular variables .
the measurements section of the simbad database lists v cvn , r umi , and rs cyg as sra stars , ux dra as srb , and sx her as srd , although conflicting classifications exist .
rs cyg and ux dra are also classified as carbon stars . sx her is earlier in spectral type ( k2 ) than the others ( late m or c ) .
v cvn has been reported to have two closely spaced periods ( loeser et al .
1986 , kiss et al .
1999 , kiss et al .
2000 ) and exhibits time - dependent polarization ( boyle et al .
1986 , magalhes et al .
sx her is a metal - poor star ( preston and wallerstein 1963 ) that shows no signs of an infrared excess ( carney et al .
its period is apparently changing slowly ( percy and kolin 2000 ) .
in 3 we argue on physical grounds that the irregular pulsations of these stars can not be multi - periodic as has been suggested ( kiss 1999 ) , nor can they be of a stochastic nature as proposed in konig ( 1999 ) .
this means that , by default , they therefore must be chaotic .
if the light curve is indeed unsteady because of a chatic dynamics , then it is not astonishing that an interpretation within a traditional periodic framework would lead to a spurious conclusion that there are closely spaced peaks or that the period is changing ( buchler , j. r. & kollth , z. , 2001 ) .
in 4 we apply a global flow reconstruction ( serre 1996 ) to the data .
this tool has been specifically designed for the study of chaotic signals which have only a few degrees of freedom ( low dimensional chaos ) .
we show that this reconstruction can capture the pulsation dynamics of these stars .
the fact that the reconstruction allows us to generate synthetic lcs with properties similar to the observed lcs thus provides additional , _ a posteriori _ evidence for low dimensional chaos .
the global flow reconstruction also yields quantitative information about the underlying dynamics , such as its dimension .
we can use this information to uncover the physical nature of the irregular lcs of these stars .
in particular , it allows us to address the question of whether the same resonant mechanism is operative here as in the case of the rv tau - type star r scuti where the large amplitude irregular pulsations were shown to be a consequence of the resonant excitation of an overtone with close to double the frequency of the basic pulsation mode under highly nonadiabatic conditions ( buchler 1995 , 1996 ) .
the fundamental mode is self - excited and wants to grow , whereas the entrained , resonant overtone is stable and wants to decay , from which a chaotic motion of alternating growth and decay ensues , a scenario known in nonlinear dynamics under the name of shilnikov ( berg 1986 ) .
our analysis is based on high quality v band light curve ( lc ) data from the observational program at grinnell college that span close to 18 years .
the observations were made with the 0.61 m telescope at the grant o. gale observatory using a conventional photoelectric photometer .
the data were reduced differentially relative to the following comparison stars ( in parentheses after the associated variable star ) : r umi ( hd 150747 ) , rs cyg ( hd 192422 ) , v cvn ( hd 116531 ) , ux dra ( hd 178089 ) , sx her ( hd 145226 ) . the comparison stars were monitored relative to additional check stars and found to exhibit no significant variability . the reduction included corrections for atmospheric extinction and transformation to the ubv system using b band data that are not presented here .
the precision of the data is determined primarily by atmospheric conditions but can be estimated to be about 0.01 magnitude from the repeated measurements that make up a single data point as well as from related data .
because continuity of the data is important for the sort of analysis presented here , efforts were made to essentially eliminate the gaps between observing seasons by taking some data ( a small fraction of the total ) at larger airmasses and in brighter twilight than is desirable for photometric observations .
the quality of the data obtained in twilight was improved by modeling the nonlinear time variation in the sky brightness .
although the quality of the measurements obtained under these conditions is a bit worse than that for data taken under more favorable conditions , it is still good enough for the purposes of this paper and is far preferable to gaps in the data .
the data are available from rrc .
the observed lc data are displayed as dots , jointly with the smoothed fits , in fig .
the sampling is irregular with an average spacing of 10 d. the light curves of r umi , v cvn , and ux dra exhibit sizeable vertical shifts that could be due to dust formation during the pulsation but , as the next section suggests , they could also be part of the intrinsic pulsation dynamics of the stars .
the lc of sx her shows large fluctuations in overall amplitude .
rs cyg displays distinctive dips at the maxima of its lc .
figure [ vcvnzk ] , we present a comparison of the photoelectric cadmus data and the combined visual amateur astronomer data ( aavso , afoev , baavs , vsjso ) of v cvn for approximately the same time - intervals .
these smoothed time - series therefore have a different zero points and different amplitudes , and they also have quite a different quality .
( the smoothings that we apply to these data will be discussed below . )
we display the smoothed lcs ( top ) , the amplitude fourier spectra ( fs ) ( right ) and the time - frequency ( tf ) plots ( kollth & buchler 1996 ) , also known as spectrograms .
for the latter we have used the zhao - atlas - marks ( 1990 ) kernel that belongs to a class of tfs known under the name of reduced interference distributions ( rids ) .
these are designed to give much sharper images than the older transforms , such as gbor , for example ( e.g. cohen 1994 ) . in the tf plots
we have enhanced the power at the harmonics to the same level as the fundamental to make it visible ( @xmath0 , where @xmath1 for @xmath2 , where @xmath3 is the frequency of the fundamental peak , @xmath4 , for example , for @xmath5 , and a linear increase in between ) .
a comparison of the tf plots with the fs ( on the right ) clearly shows the size of this enhancement @xmath6 . despite the fact that the cadmus and amateur lcs are appreciably different at times , the tf plots are remarkably similar , showing their insensitivity to observational errors and to the smoothing procedure . in figs .
[ vcvnzk ] through [ sxherzk ] we display the tfs of rs cyg , r umi , ux dra and sx her .
the tfs of all five stars show some amplitude ( power ) at twice the fundamental peak , and both rs cyg and v cvn also have appreciable amplitude at @xmath7 . from the very appearance of the observed lcs one
finds it hardly astonishing that for all five stars the instantaneous amplitudes in the dominant peaks waver in time , but more interestingly the instantaneous frequencies change as well .
in addition , the first harmonic frequency peak does not change synchronously with that of the fundamental frequency .
this asynchrony between the fundamental and the harmonic peaks rules out that all of the harmonic structure is due to the anharmonicity of the motion .
furthermore it can not be explained by dust or spots , by binarity , by evolution or by stochasticity , even though all of these effects may be present and affect the lcs .
we shall first argue that the pulsations of these stars are neither periodic nor multi - periodic ( see also buchler & kollth 2003 ) .
then we will give physical , energetic arguments that also eliminate a stochastic nature for these pulsations .
this then , by default , leaves a low dimensional chaotic nature as the only alternative . because the multi - periodic label is sometimes used with different connotations , to be specific , we mean that the signal is such that ( a ) the fs contains a few discrete , basic frequencies , and all the other peaks correspond to linear combinations of these basic frequencies that arise because of nonlinear couplings , and ( b ) the amplitudes and frequencies of all these peaks are constant , or , at most , vary very slowly in time .
the beat cepheids and rr lyrae are examples of this type of systems .
multi - periodicity would thus imply that the major frequency peaks @xmath8 in the fourier spectra are real and should correspond to actual pulsation frequencies or linear combinations thereof . using these frequency peaks one
can then make a multi - periodic fit @xmath9 to the observational data .
for the stars that we are considering we find that we need from 50 to 100 frequencies for a decent fit .
this however poses a physical problem , namely the origin of these frequencies .
indeed , stellar modelling shows that there simply do not exist enough radial modes in the frequency range of interest for these types of star .
this remains true even if allowance is made for nonradial modes with @xmath10 low enough ( say @xmath11 ) to be observable photometrically and assuming that the latter can indeed be excited to such large amplitudes .
furthermore , typically only a few modes are found to be linearly unstable , and one would have to explain why so many other stable modes get entrained to such high amplitudes . we thus conclude that these lcs are not generated by a large set of excited pulsation modes .
there is still the possibility of a multi - periodic , but slowly evolving star .
the structure of the peaks should thus changes slowly as well , but for our five stars the timespan of the observational data is too short compared to the evolutionary timescale .
we note however , that for the rv tau type star r sct ( kollth 1990 ) , the frequencies in the fs appear randomly variable from one section of the data to another which thus also eliminates an explanation in terms of slowly evolving multi - periodic stars .
a recent paper ( konig 1999 ) claim to explain the irregular behavior of r sct as the superposition of two stochastically excited damped linear oscillators .
typically stochastic excitation of vibrationally stable modes , such as the solar oscillations , leads to much smaller amplitudes .
it seems more likely that pulsations as large as those observed in r sct are due to a self - excited mode , such as in the cepheid and rr lyrae variables .
in fact this brings us to our main rejection of stochastic excitation for r sct .
while such a stochastic explanation is perhaps correct from a mathematical point of view , it is not meaningful on physical grounds because no mechanism is proposed that could excite damped modes to such large pulsation amplitudes ( up to factors of 40 in the lc of r sct , for example ) . in order to be more quantitative ,
let us consider a 0.7 model with l=1000 and = 5300k ( and x=0.726 , z=0.004 ) . in our turbulent convective hydro - code ( kollth 2002 )
this model is unstable in the fundamental mode and leads to chaotic pulsations of alternating large and small amplitudes .
the average pulsational kinetic energies in such typical cycles are @xmath12ke@xmath13 = 0.31@xmath14 and 0.23@xmath14 ergs , respectively .
the average turbulent energies are @xmath12te@xmath13 = 0.15@xmath14 and 0.12@xmath14 ergs , somewhat smaller , but of the same order within the accuracy of mixing - length models . the alleged stochastic excitation of konig ( 1999 ) can only be associated with the convective motions .
our numbers show that even if _ all _ available turbulent energy were converted into pulsational kinetic energy , there is barely enough energy .
the pulsation is closely associated with a single ( most likely radial ) pulsation mode , or perhaps at most two of them , thus has a very directed form of kinetic energy .
the amount of energy transferred stochastically from the kinetic energy associated with the largely disordered turbulent convection to these modes depends on the overlap integrals between the spatial distribution of the turbulent energy ( eqs .
6 and 7 of buchler , goupil & kovcs 1993 ) .
this overlap must be extremely small because of the very different nature of the two types of motion .
furthermore the power in the fourier spectrum of turbulent convection at the pulsation frequency is very limited .
this rules out stochasticity as the cause of the observed irregular behavior . however , the amount of energy flowing through the pulsating envelope , the product of the luminosity and the period @xmath156.1@xmath14 ergs , is almost 100 times the pulsation kinetic energy .
this highly nonadiabatic situation certainly allows the observed changes in pulsational kinetic energy .
a perhaps better estimate of the efficiency of conversion of heat into pulsation energy , or the percentage change of pulsation kinetic energy per cycle , is given by the linear growth - rates of the modes ( @xmath16 for the stellar model above ) .
this is precisely what a chaotic dynamics is about : the nonlinear interaction of a small number of strongly nonadiabatic ( dissipative ) pulsation modes .
we conclude that it is therefore ultimately physics that rules out stochasticity for these stars .
this is not to say , however , that superposed on the low dimensional radial chaotic dynamics , there can not be stochastic motions , which furthermore are nonradial .
there are additional reasons for believing that the pulsations of these five stars are chaotic and have a low dimension : first , numerical hydrodynamic simulations , albeit of w virginis type stellar models , uncovered the chaotic nature of the pulsations ( buchler & kovcs 1987 , aikawa 1987 , buchler , goupil & kovcs 1987 , kovcs & buchler 1988 , letellier 1996 ) .
second , the analysis of the lc of the rv tau - type star r scuti clearly demonstrated the presence of a low dimensional chaotic pulsation dynamics ( buchler , serre , kollth & mattei 1995 , buchler , kollth , serre & mattei 1996 ) .
more recent work by other authors gives corroborating evidence for low dimensional chaos in large amplitude variabel stars ( kiss & szatmry 2002 , ambika 2003 ) .
linear analyses are very useful for multi - periodic signals , but it is well known , at least outside astronomy , that for irregular , intrinsically nonlinear signals they yield limited information about the dynamics ( weigend & gershenfeld 1994 , abarbanel 1993 ) . as we have already mentioned , a phenomenologically motivated multiperiodic fit will always be successful as an interpolation , but will give us not much information about the stellar pulsation aside from one or perhaps several average periods. for the study such chaotic signals and for understanding the underlying dynamics a different approach is required .
a popular test for nonlinear behavior is the construction of first return maps ( ott 1993 ) . for a dynamics with a fractal attractor with fractal dimension @xmath172 , such as the rssler band ,
the first return maps show a tight parabolic structure , but for our stars they turn out to be essentially scatter diagrams .
this negative result implies that , if this attractor is indeed fractal , the fractal dimension of the dynamics substantially exceeds a value of 2 , and could even be greater than 3 .
we recall that for the star r sct it was found that @xmath18 ( buchler 1996 ) .
llrrrccccr & @xmath19 & @xmath20 & @xmath21 & @xmath22 & @xmath23 & @xmath24 & @xmath25 & @xmath26 & @xmath27 + & 0.01 & 0 & 23 & 1.064 & 2.093e03 & 3.543e03 & 3.275e03 & 7.340e03 & 2.07 + & 0.01 & 0 & 25 & 1.072 & 2.442e03 & 3.504e03 & 3.635e03 & 7.254e03 & 2.07 + & 0.01 & 8 & 19 & 1.063 & 1.508e03 & 3.532e03 & 2.712e03 & 7.764e03 & 2.20 + & 0.02 & 0 & 17 & 1.081 & 1.296e03 & 3.511e03 & 2.112e03 & 7.898e03 & 2.25 + & 0.02 & 4 & 15 & 1.083 & 1.319e03 & 3.523e03 & 2.146e03 & 8.039e03 & 2.28 + & 0.02 & 4 & 17 & 1.087 & 1.516e03 & 3.539e03 & 2.188e03 & 7.890e03 & 2.23 + & 0.03 & 0 & 21 & 1.096 & 1.233e03 & 3.554e03 & 2.696e03 & 6.759e03 & 1.90 + & 0.03 & 8 & 21 & 1.091 & 1.420e03 & 3.547e03 & 3.231e03 & 6.554e03 & 1.85 + & @xmath19 & @xmath20 & @xmath21 & @xmath22 & @xmath23 & @xmath24 & @xmath25 & @xmath26 & @xmath27 + & 0.025 & 1 & 40 & 0.631 & 3.368e03 & 2.793e03 & + 1.981e03 & 5.248e03 & 1.88 + & 0.025 & 7 & 40 & 0.622 & 3.149e03 & 2.793e03 & + 1.830e03 & 5.365e03 & 1.92 + & 0.03 & 4 & 35 & 0.581 & 3.539e03 & 2.701e03 & 5.122e04 & 5.438e03 & 2.01 + & 0.04 & 4 & 33 & 0.489 & -1.142e-03 & 6.245e-03 & 2.398e-03 & 2.724e-03 & 2.29 + & 0.04 & 4 & 40 & 0.500 & 1.258e03 & 2.708e03 & 9.522e04 & 6.444e03 & 2.38 + & 0.04 & 5 & 34 & 0.491 & 2.214e03 & 2.724e03 & 1.062e03 & 6.290e03 & 2.31 + & @xmath19 & @xmath20 & @xmath21 & @xmath22 & @xmath23 & @xmath24 & @xmath25 & @xmath26 & @xmath27 + & 0.01 & 0 & 22 & 0.279 & 4.665e03 & 5.102e03 & 4.098e03 & 1.119e02 & 2.19 + & 0.01 & 6 & 22 & 0.278 & 4.514e03 & 5.087e03 & 4.005e03 & 1.105e02 & 2.17 + & 0.01 & 6 & 23 & 0.263 & 4.933e03 & 5.065e03 & 4.108e03 & 1.047e02 & 2.07 + & 0.01 & 12 & 21 & 0.279 & 4.665e03 & 5.102e03 & 4.098e03 & 1.119e02 & 2.19 + & 0.02 & 0 & 22 & 0.279 & 4.672e03 & 5.076e03 & 4.694e03 & 1.108e02 & 2.18 + & 0.02 & 6 & 21 & 0.287 & 4.448e03 & 5.124e03 & 4.309e03 & 1.155e02 & 2.25 + & 0.03 & 0 & 21 & 0.292 & 4.615e03 & 5.115e03 & 4.999e03 & 1.152e02 & 2.25 + & @xmath19 & @xmath20 & @xmath21 & @xmath22 & @xmath23 & @xmath24 & @xmath25 & @xmath26 & @xmath27 + & 0.018 & 0 & 6 & 0.39 & 1.236e02 & 3.963e03 & 8.458e03 & 5.975e03 & 1.51 + & 0.016 & 0 & 5 & 0.43 & 9.825e03 & 3.478e03 & 5.684e03 & 7.398e03 & 2.13 + & 0.017 & 0 & 6 & 0.41 & 1.155e02 & 3.720e03 & 7.904e03 & 6.423e03 & 1.73 + & 0.018 & 0 & 6 & 0.39 & 1.236e02 & 3.963e03 & 8.458e03 & 5.975e03 & 1.51 + & 0.019 & 0 & 5 & 0.39 & 1.522e02 & 3.633e03 & 1.058e02 & 5.548e03 & 1.53 + [ tab ]
over the last decade we have developed a special tool for the analysis of irregular time - series , namely the _ global flow reconstruction technique _
( reconstruction technique for short ) .
our primary motivation has not been phenomenological , but rather has aimed for an understanding of the physical mechanism behind the irregular lcs of variable stars on the one hand , and for deriving qualitative and quantitative properties of this dynamics ( serre 1996 ; reviewed in buchler 1997 , buchler & kollth 2001 ; a similar approach was followed in other research areas , gouesbet 1997 , hegger 1999 ) .
only two very simple working assumptions underlie the reconstruction technique : ( 1 ) _ the lc is produced by the ( deterministic ) nonlinear interaction of a small number of pulsation modes _ , ( 2 ) _ the system is autonomous _ , explicit time - dependence , such as evolution over the span of the data , can be ignored .
it may appear astonishing that one should be able to derive basic physical properties of the star from the mere knowledge of a _
scalar _ time - series , such as the lc .
yet , in the case of the star r scuti it was possible to demonstrate that the irregular pulsations are the result of the nonlinear interaction between a self - excited pulsation mode that entrains a vibrationally stable mode through an approximate 2:1 resonance ( buchler 1995 , 1996 ) .
the technique introduces a ( euclidean ) _ reconstruction space _ of dimension @xmath28 , in which one uses the observed time - series @xmath29 to construct successive position vectors @xmath30 the quantity @xmath21 is called the delay .
assumption ( 1 ) allows us to use the observed time - series to construct the best mathematical relationship @xmath31 , ( called a map or a flow in the case of a differential description ) , that evolves the system that produces the lc from one time - step to the next : @xmath32 a dimension @xmath28 that is large enough to capture the dynamics of the pulsation is called an _ embedding dimension_.
our goal is to find the
_ minimum embedding dimension _
this dimension @xmath33 is not known and it is one of the fundamental quantities that the analysis aims to determine .
this dimension is important because it is closely related to the dimension @xmath34 of the actual physical phase space of the dynamics .
for example , for a single oscillatory mode , @xmath34 would be equal to 2 , for 2 coupled oscillatory modes @xmath35 , and the involvement of a secular mode would add 1 to the dimension . practice has shown that a multivariate polynomial functional form is suitable for the map @xmath36 and that polynomes up to order of 4 are generally sufficient ; the coefficients of these polynomials are computed with a least - squares fit to the observed time - series ( serre 1996 ) .
the reconstruction technique also introduces a _ delay _
@xmath21 which we consider a free parameter .
the reconstruction requires a time series with equal time intervals .
our data preparation is done in several steps : ( a ) the data are binned and averaged over a time interval @xmath37 . the typical time interval for the cadmus data has a very broad peak around 10d . in the following we have chosen values for @xmath37 = 10d , except for ux dra where we chose 1d .
( b ) the data are next smoothed with a cubic spline ( reinsch 1967 ) with a smoothing @xmath19 which results in an equally spaced time series , with time intervals chosen to be 1d .
( c ) finally a moving gaussian filter of width @xmath20 is applied to the splined time series .
results of a preliminary analysis of these stars were reported in buchler , kollth & cadmus 2001 . here
we perform a much broader analysis ; in particular our search covers a wider range of smoothings , @xmath19 and @xmath20 . as a result
we obtain firmer conclusions than were obtained in that paper . -1.5 cm in this section we apply the reconstruction to the observed lc of r umi , and concurrently we discuss some of the technical details of the reconstruction . in figure [ rumisyn ] on top
we display a fit to the r umi data , with the smoothing parameters @xmath38 and @xmath39 .
two observational points , at jd 46039.5095 and 46613.8169 , give an unnatural looking wiggle in the smoothed data ( in fig . 1 of buchler 2001 )
, and have therefore been omitted in the reconstructions .
we have verified though that the reconstructions with all the observational points give essentially the same results , indicating a relative insensitivity of the r umi reconstructions to individual points . in practice ,
the reconstructions proceed as follows : from a smoothed light curve ( with given @xmath19 and @xmath20 ) a map @xmath40 is constructed for a given dimension @xmath28 and a given value of delay @xmath21 .
this map @xmath40 is then iterated from a seed value ( chosen along the observed time - series ) to generate what we call a _
synthetic _ lc . actually , the map @xmath40 generates @xmath28 scalar sequences , namely the components of @xmath41 , shifted with respect to each other by @xmath21 days , but are essentially identical when superposed .
this can be considered an _ a posteriori _ confirmation that the reconstruction is successful . for r
umi as for the other stars we have chosen 15000 d long iterations of which we typically discard the first three thousand days , because they may be transients in the evolution toward the chaotic attractor .
because the data preparation obviously will have an effect on the reconstructed attractors , we therefore perform them for a range of smoothing and filtering values @xmath19 and @xmath20 .
ideally , we would also expect the results of the reconstructions to be independent of these smoothing parameters in some range of values ( for a discussion of these issues we refer to serre 1996 ) . in our surveys
we perform the reconstructions with delays typically ranging from @xmath42 to 40 d. in a robust reconstruction the results should come out similar for some range of neighboring @xmath21 values .
it is obvious that for a successful reconstruction the data have to sample sufficiently well the dominant features of the dynamics in phase space .
when the lc is very short there are only 20 pulsation cycles for r umi the iteration of the constructed maps is not always stable and the synthetic lcs blow up .
there can be several causes for such behavior , boundary crises ( ott 1993 ) . in practice , we iterate a given map successively with different seeds until we obtain a synthetic lc of sufficient length and quality .
how do we judge the quality of the reconstruction ?
there is no point in comparing chaotic signals point by point , but rather they need to be compared in a global or statistical sense .
however , we are seriously hampered by the shortness of both the observed and synthetic lcs .
we are therefore forced to rely on more subjective criteria .
our first acceptance criterion relies on _ a visual comparison _ of the observed and synthetic lcs .
when we attempt reconstructions for r umi in 3d ( with @xmath43 ) , we find that the synthetic lcs bear no resemblance to the observed lc for any @xmath19 or @xmath20 .
it is therefore immediately obvious that @xmath44 .
in contrast , the 4d reconstructions show a relatively large insensitivity to the smoothing .
we have made our survey with @xmath45 , 0.02 , to 0.05 and @xmath46 , 4 and 8 ( @xmath20=0 indicates that no gaussian filter was applied ) .
good synthetic lcs have been obtained with between 3 and 9 delay values which , depending on the amount of smoothing , range from @xmath47 to 27d . with larger
smoothing , ( @xmath48 ) , the reconstructions deteriorate , with the synthetic lcs becoming more symmetric about the average and more regular .
figure [ rumisyn ] displays a sample of our best segments of 4d synthetic lcs , with the delays @xmath21 , and the smoothings @xmath19 and @xmath20 indicated on the right .
the fact that our 4d reconstructions appear to capture the dynamics of the pulsations of r umi suggests that @xmath49 .
-5pt a second comparison can be made at the level of the _ amplitude fourier spectra ( fs)_. because the signals are unsteady , we do not want to compare individual peaks , but rather the fss should have the same overall structure .
figure [ rumifs ] displays successively from top to bottom , the fs of the smoothed and of our best synthetic lcs , indicating excellent agreement .
the relative broadness of the fundamental peak for the observed lc is due to its shorter time span .
finally , a third comparison can be performed with _ bk projections _
( broomhead & king 1987 ) which are the projections of the temporal signals on the first few eigenvectors of the auto - correlation matrix which represent an optimal visual way of spreading out the attractor . in practice
, we first generate the projection matrix with the smoothed observed lc with a chosen delay @xmath21 , and then use this matrix to make the projections of all the lcs .
the bk projections onto the first three eigenvectors , labelled x , y and z , are displayed in fig .
[ rumifs ] , indicating that the structure of the reconstructed attractor is very similar to that of the observations .
the three tests indicate that , considering the extreme shortness of the data , the reconstructions for r umi are quite good and reasonably robust . in table [ tab ]
we present * the linear stability eigenvalues * ( @xmath50 ) of the fixed points of the maps .
they should correspond to the frequencies of the modes that are excited in the pulsation and therefore to the stability of the equilibrium state of the star .
note that the value of the lowest frequency @xmath24 is close to the frequency of the lowest peak in the fs ( [email protected]@xmath52 ) .
the frequency @xmath26 is not as well determined , although the frequency ratios are reasonably close to a 2:1 resonance .
table [ tab ] indicates that the growth rates are comparable to the frequencies , indicating the presence of extreme nonadiabaticy , consistent with the numerical modelling of such stars ( kovcs & buchler , 1988 ) .
column 4 represents the values of the magnitude of the fixed point , the center of the motion , which are close to the average magnitude of the observed lc , 0.95 , but not equal to it presumably because of the skewedness of the light curve .
the growth rate @xmath23 of the lower frequency mode is positive ( unstable ) whereas the second mode is damped ( @xmath53 ) .
we see that the reconstructions consistently reproduce the linear part of the maps , despite the fact that the amplitude of the lc never gets very small , and that therefore it does not explore the neighborhood of the fixed point .
we note in passing that in nonlinear dynamics this scenario of two resonant modes , one unstable and the other stable , goes under the name of modified shilnikov scenario ( glendenning & tresser 1985 ) .
lyapunov exponents and the ( generally non - integer ) _ lyapunov dimension @xmath54 _ ( ott 1993 , serre 1996 ) can give us additional quantitative information about the attractor .
in particular @xmath54 sets a lower limit on the dimension @xmath34 of the physical phase space , because an attractor of fractal dimension @xmath54 must live in a euclidean space of dimension @xmath55 .
for example , a dynamical system of two coupled pulsation modes ( @xmath56 ) can have an attractor with a fractal dimension @xmath57 . in order to extract accurate lyapunov exponents and a lyapunov dimension @xmath54 one needs lcs containing at least 500 cycles , @xmath15 200,000d . the observed lc with 6500d
is much too short for that purpose .
furthermore the maps for r umi are not stable enough to generate sufficiently long synthetic lcs to compute reliable lyapunov exponents , but they yield a rough estimate . for 12000d long synthetic lcs with @xmath58 the @xmath54 fall in a broad range from 2 .
to 3.2 with an average of 2.8 .
for the best few synthetic lcs @xmath54 averages to 2.9 . a couple of rather rare , 130,000d long 4d iterations , with @xmath45 , @xmath59 , @xmath60 gave @xmath61 and for @xmath45 , @xmath59 , @xmath62 gave @xmath63 .
we believe that these latter values may be closer to the actual value of @xmath54 , but these values should be taken with a grain of salt . in principle
, 5d reconstructions can be used to confirm that a 4d space is large enough , because the fractal dimension should be independent of the dimension of the reconstruction space .
unfortunately , such reconstructions are not very robust because of the paucity of pulsation cycles in the observed lc .
we note however that the @xmath54 for the few decent 5d synthetic lcs that we could make cluster around 3.5 , they essentially do not increase much as @xmath28 is increased from 4 to 5 .
this is a corroboration that the embedding dimension is indeed @xmath49 . in the study of r sct ( buchler 1995 , 1996 ) , with the help of @xmath64
, we were able to obtain a _
lower limit _ of 4 for the dimension @xmath34 of the physical phase space of the dynamics , and very strong evidence for a minimum embedding dimension @xmath49 gave an _ upper limit _ of @xmath35 .
we thus could infer that @xmath35 for r sct . here
, for r umi we can state the minimum @xmath49 and thus the upper limit on @xmath65 with only some confidence . also , because of the uncertainties in @xmath54 just mentioned we obtain a lower limit of either 3 or 4 for @xmath34 , depending whether @xmath66 or @xmath67 .
however , if @xmath34 were only equal to 3 , then one would have to invoke the coupling of one oscillatory mode ( 2d ) with a thermal ( secular ) mode ( 1d ) , rather than with a second oscillatory pulsation mode ( 2d ) . while this is possible we deem it unlikely , and it appears that @xmath35 , as for r sct . in this 4d physical phase space , the complex amplitudes of the two pulsation modes may be considered the natural generalized coordinates for this dynamics .
furthermore , the linear stability analysis of the 4d maps gives us results that are very similar to those of r sct ( buchler 1996 ) .
we can therefore conclude that the same pulsation mechanism is operative in r umi , namely that its irregular lc is generated by two nonlinearly interacting , resonant modes in which the low frequency mode is linearly unstable and the harmonic mode is linearly stable , resulting in a continual and unsteady energy exchange .
fourier amplitude spectrum top : of the smoothed lc , below : of synthetic lcs .
bk projections , top : smoothed lc , below : sections of synthetic lcs with various smoothings and delays @xmath21 .
[ sxherfs ] -1.5 cm in figure [ rscygsyn ] on top we display the 15 cycles long lc for rs cyg , superposed on a lc fitted with the smoothings @xmath38 , @xmath39 .
we find again that 3d reconstructions are not satisfactory at all , with the periods , in particular , being quite bad .
the 4d reconstructions show some insensitivity to the smoothing , as long as it is large enough : good lcs have been obtained from a range of typically 2 5 delay values which , depending on the amount of smoothing , range from @xmath68 to 40d .
synthetic lc obtained from reconstructions with a smoothing of @xmath69 and 0.020 were too regular , and the fixed points of the map are about 0.20 to 0.25 too large .
the displayed synthetic lcs come from smoothing values @xmath70 .
typically there are from 2 to 6 @xmath21 values for which there are good synthetic lcs .
[ rscygsyn ] displays a sample of our best 12000 d long segments of 4d .
this suggests that the embedding dimension for rs cyg is also @xmath49 .
this conclusion is corroborated by the amplitude fs of fig .
[ rscygfs ] and the bk projections in fig .
[ rscygfs ] .
the linear stability eigenvalues of the fixed points of the maps are shown in table 2 .
the frequency ratios are approximately to 2:1 for the reconstructions with the larger values of @xmath71 that we have used .
the growth rate @xmath23 of the first mode is again positive ( unstable ) and that of the second mode @xmath53 , but not for @xmath72 . in all cases ,
the lowest frequency @xmath24 is reasonably close to the frequency of the fundamental peak in fig .
[ rscygfs ] . that the linear stability eigenvalues of the fixed points are not very consistent is not very astonishing , because after all the amplitude of the lc is never small , and therefore it does not at all explore the neighborhood of the fixed point where the stability properties are determined .
the lyapunov dimensions @xmath54 of all our good synthetic lcs are found to run from 2.1 to 3.2 with an average of @xmath73 .
the same average @xmath54 is obtained for the best few .
this imposes a lower limit of 3 or 4 on @xmath34 , but the latter can still be , and we believe it to be 4 for the same reasons we have given in the discussion on r umi above .
there are unfortunately too few data points to attempt a reconstruction in 5d which could corroborate that @xmath49 .
we conclude that the analysis of the rs cyg lc is consistent with , and perhaps suggests again that two interacting , highly nonadiabatic modes are at work , which in addition are in resonance . in other words
rs cyg seems to have the same underlying 4d dynamics that we have found for r sct and r umi . -1.5
cm the smoothed v cvn lc is displayed in fig .
[ vcvnsyn ] ( @xmath45 , @xmath74 ) .
we have not been able to make any satisfactory 3d reconstructions for v cvn . however , the 4d reconstructions show insensitivity to the smoothing . for each smoothing set , good synthetic lcs
have been obtained for typically 2 8 delay values which , depending on the amount of smoothing , range from @xmath75 to 27d .
based on the visual appearance of the synthetic lcs the 4d reconstructions in fig .
[ vcvnsyn ] are good , suggesting @xmath49 . however , for some reason , the dominant frequencies of all the synthetic lcs are systematically smaller ( fig . [ vcvnfs ] ) , and the overtone frequencies are somewhat larger . the bk projections , displayed in fig .
[ vcvnfs ] , are a little tamer for the synthetic lcs .
the same difference is already apparent in fig .
[ vcvnsyn ] , where the lcs show fluctuations in the average magnitude that are stronger than those of any of the synthetic ones .
it could be that these fluctuations of the average magnitude are due to dust , or it could be that the dynamics is more complicated in the sense that a higher dimensional phase space ( more modes ) is required to capture it . in order to answer that question we would likely need a longer lc .
the lyapunov dimensions @xmath54 of the good 4d synthetic lcs range from 3.4 to 3.6 , substantially larger than for r umi and rs cyg .
good 5d reconstructions are also possible , but they occur only for one or 2 @xmath21 values that run from 15 to 26d , depending on the smoothing .
significantly , the fractal dimensions of the good synthetic lcs run from @xmath76 to 4.02 , they do not increase with @xmath28 , corroborating that indeed @xmath49 . another interesting difference with r sct or r
umi appears in the linear stability eigenvalues of the fixed point displayed in table [ tab ] .
the frequencies associated with the linear stability eigenvalues of the fixed point are again in a close resonance ratio of 2:1 , and the modes are highly nonadiabatic . however , here it is the higher frequency mode that is unstable , and the lower one that is stable . in the study of buchler 2001
reconstructions were unsuccessful because the range of smoothing parameters was very limited .
this shows that a survey with a range of smoothings is important . -1.5
cm the smoothed lc of sx her is shown on top of fig .
[ sxhersyn ] with ( @xmath45 , @xmath46 ) . underneath ,
are our best segments of synthetic lcs in several reconstruction dimensions @xmath28 .
as for the previous stars , for sx her the 3d reconstructions are not good at all . in 4d
they are a little better , but still not very satisfactory , having more symmetry than the lc .
as already found in buchler ( 2002 ) they are very sparse in @xmath21 space , and hence not robust .
5d reconstructions , while capturing some of the features of the lcs appear unsatisfactory . on the other hand , in 6d the synthetic lcs display many of the characteristic features of the lc , such as asymmetric bursts .
the fourier spectra are shown in fig .
[ sxherfs ] for the same lcs displayed in fig .
[ sxhersyn ] .
the first frequency is recovered relatively well , although the fundamental peak shows more structure than for the lc .
the bk projections shown in fig .
[ sxherfs ] are reasonably good , but they reflect the tamer nature of the lcs and their smaller amplitudes . we conclude that the dynamics which underlies the pulsations of sx her may be more complicated than that of the other stars ( to witness , for example , the intermittency in the central part of the analyzed lc ) and therefore higher dimensional ( @xmath77 , perhaps @xmath78 ) . however , this conclusion should be taken with some caution because the observational coverage of the lc is sparse in some places for this shorter period star , as fig .
[ cad ] shows .
it is also possible that we have missed a range of smoothing parameters where the lower - d reconstructions are successful . -1.5
cm figure [ uxdrasyn ] on top displays a smoothed fit to the lc for ux dra ( @xmath79 , @xmath46 ) . in a 3d reconstruction space
the synthetic lcs definitely do not look like the lc . in 4d good looking synthetic lcs
can be generated , although the reconstructions are not particularly robust .
they appear possible only for a relatively narrow range of @xmath19 and only for a few @xmath21 values in the range 3 6d .
this relative lack of robustness is not very astonishing considering that , compared to its complexity , the lc is very short . for a good reconstruction
we clearly need lc features to repeat over the span of the lc .
in addition the amplitude of this star is only half that of the other 4 stars , so that the noise level is higher . for the same reasons we have had difficulty making reconstructions in 5d and 6d .
however , the synthetic lcs are consistent with @xmath49 .
figure [ uxdrasyn ] displays a sample of our best segments of 4d ynthetic lcs .
the dominant frequencies of all the synthetic lcs are not very good as seen in the amplitude fs in fig .
[ uxdrafs ] .
the bk projections of the synthetic lcs , although they bear a good resemblance to that of the lc , but are tamer .
the lyapunov dimensions for the synthetic lcs of fig .
[ uxdrasyn ] range from 2.4 to 3.1 . the linear stability analysis , shown in the table [ tab ] , as for v cvn yields a _
stable _ low frequency , @xmath24 , mode and an unstable @xmath26 mode , the stability being the reverse of that of r sct and r umi , but the same as for v cvn .
however , the @xmath24 frequencies are not particularly close to the low frequency peak @xmath80d@xmath52 in the fs , and the frequency ratio has a larger scatter around 2:1 than for the other stars . we are forced to conclude that , either the observed lc of ux dra are too short to make robust reconstructions , or perhaps the dynamics could be contaminated by dust formation and obscuration .
-1.5 cm in our survey we have only varied the dimension @xmath28 , the smoothing parameters @xmath19 and @xmath20 and the delays @xmath21 . the polynomial order has been fixed at @xmath81 . for the binning and averaging time interval we have chosen @xmath82d for all stars except ux dra , where we obtained better results with @xmath83d . in general
, we have not attempted to optimize the values of @xmath84 which might lead to more robust reconstructions .
each vector component of our polynomial maps in 4d contains 70 coefficients , and one might be inclined object that it is not astonishing that such a map fits the data well , and that one would obtain an equally good fit with a ( linear ) multi - periodic fourier sum with the same number of coefficients .
however , for the latter is just a fit with limited predictive power and which tells us very little about the actual dynamics .
in contrast , the constructed ( nonlinear ) maps have remarkable predictive power , the long synthetic lcs that we generate from the map do capture the dynamics as our comparisons have shown .
-1.5 cm the reader may be curious why in contrast to the cepheids , which are the close relatives of the semiregular variables , pulsate in a regular , periodic or multiperiodic fashion .
the fundamental physical difference is in the size of the relative linear growth rates ( the growth rates @xmath85 periods ) . for the classical variable stars , the latter are very small , of the order of a percent , whereas those of the population ii cepheids , semiregulars and mira variables are of order unity .
the reason for this difference lies in the fact that the luminosity / mass ratio of the latter is about ten times larger .
this causes a large increase in the coupling between pulsation and heat flow and concomitantly in the growth rates .
when the relative growth rates of the excited modes are small , as in the classical cepheids , there exists near center manifold in the parlance of nonlinear dynamics , buchler 1993 ) .
it can then be shown that the most complicated motions will be periodic pulsations ( limit cycles ) or multiperiodic pulsations ( beat cepheids ) .
no such manifold exists in the highly nonadiabatic semiregular stars , and more complicated , irregular ( chaotic ) pulsations can arise . intuitively , large growth rates are a prerequisite for irregular pulsations , for amplitude modulation on the time scale of a period , although this is not sufficient .
it was shown unambiguously that the hertzsprung progression of the fourier phases of the bump cepheids has its origin in a 2:1 resonance between the linearly unstable fundamental mode and the linearly stable second overtone ( buchler 1993 ) .
it is again the presence of a center manifold that gives rise to a synchronization of the two modes , and a periodic pulsation results in which the presence of the entrained second mode merely gives rise to a bump on the light curve .
in contrast , although the same 2:1 resonance occurs in the semiregulars , the absence of a center manifold allows the pulsation to be chaotic .
we have analyzed the 6500day long observed lcs of the semiregular / mira stars r umi , v cvn , rs cyg , sx her and ux dra .
first , with a time - frequency analysis we have shown that in each case the dominant instantaneous frequency peak is not steady , and that the harmonic peak does not vary synchronously with the dominant one .
this has led to the conclusion that these stars are not multi - periodic in the usual sense of the word , but that they are more likely to be generated by a low dimensional chaotic dynamics that involves a small number of pulsations modes .
_ multimode _ might therefore be a more suitable label for these stars than multiperiodic .
the results of a global flow reconstruction technique give strong evidence that the lcs of all five stars are generated by a low dimensional chaotic pulsation dynamics .
the observed lcs of all 5 stars are really very short in terms of the number of pulsation cycles and their complexity , and our results must be taken with a little caution .
however , interestingly , the evidence is very strong that for r umi , rs cyg and v cvn , and probably for ux dra as well , the pulsation takes place in a 4d phase space .
the lc of sx her appears to be more complicated and perhaps a higher dimensional space , 6d , may be required .
our analysis shows that the lcs of r umi and rsc cyg are generated by nonlinear dynamics that involve two resonant pulsation modes , a linearly self - excited lower frequency mode which entrains a second , linearly stable mode through a 2:1 resonance .
the irregular pulsation occurs as a result of continual exchange of energy between the two resonant modes .
this is the same scenario as was found for the irregularly pulsating rv tau - type star r sct ( buchler 1995 , 1996 ) , but we caution that the reconstructions are not as robust as they are for r sct because of the shortness of the observed light curves . in particular , we were not able to construct long synthetic lcs and extract a reliable lyapunov dimension from these as for r sct .
the analysis of rs cyg strongly suggests the presence of the same pulsation mechanism is operative .
interestingly , the analysis shows that the lc of v cvn may similarly be due to a resonant interaction of two pulsation modes in a 2:1 resonance , but here it appears that the higher frequency mode is self - excited , whereas the lower frequency mode is stable .
the analysis of ux dra is much less conclusive , but is compatible with the same resonant 2-mode scenario . to summarize then
, we find good evidence for a common physical explanation for the irregularity that characterizes the pulsations of r umi , rs cyg , v cvn ( and probably ux dra ) as we had found earlier for r sct , namely that the underlying low dimensional chaotic pulsation dynamics arises from the nonlinear interaction of two resonant modes , and their high nonadiabaticity causes the pulsations to be chaotic .
this work has been supported by the national science foundation ( ast03 - 07281 ) at uf and ( ast-8420901 and ast-8718533 ) at grinnell college , and by the hungarian otka ( t-038440 , t-038437 ) .
we thank the aavso , afoev , baavs and vsolj for the data that were used in fig .
[ vcvnzk ] .
this research has made use of the simbad database , operated at cds , strasbourg , france .
we also wish to thank an anonymous referee for his comments .
buchler , j. r. 1997 , in _ international school of physics `` enrico fermi '' , course cxxxiii , past and present variability of the solar - terrestrial system : measurement , data analysis and theoretical models _ , eds . g. cini castagnoli & a. provenzale , societa italiana de fisica , bologna , italy .
275 buchler , j.r .
, goupil m.j .
, kovcs g. 1993 , aa 280 , 157 buchler , j.r .
, goupil m.j . , kovcs g. 1987 , phys lett a 126 , 177 buchler , j. r. & kovcs , g. 1987 , apjlett , 320 , l57 buchler , j. r. & kollth , z. , 2001 , _ nonlinear analysis of irregular variables _ , in _ stellar pulsation : nonlinear studies _
, eds . m. takeuti & d.d .
sasselov , assl ser .
257 , 185 buchler j. r. & kollth , z. , 2003 _ nonlinear properties of the semiregular variable stars _ in _ mass - losing pulsating stars and their circumstellar matter _ , ed .
y.nakada , m.honma & m. seki , assl 283 , 59 .
buchler , j. r. , kollth , z. , cadmus , r. r. , jr.2001 , _ chaos in the music of the spheres _ , proceedings of chaos 2001 , potsdam , germany , aip conf proc 622 , p. 61 ; buchler j. r. , kollth , z. , serre , t. , mattei , j. 1996 , a&a 311 , 833 buchler , j.r . & kovcs , g. 1987 , apj lett 320 , l57 buchler , j. r. , serre , t. , kollth , z. , mattei , j. 1995 , phys rev lett 74 , 842 carney .
b.w . , latham , d.w . ,
stefanik , r.p . , morse , j.a .
2003 , aj 125 , 293 gouesbet , g. , lesceller , l. , letellier , c. , brown , r. , buchler , j.r . ,
kollth , z. 1997 , in _ nonlinear signal and image analysis _ , ann .
808 , p. 25
hegger , r. , kantz , h. , schreiber , t. 1999 , chaos 9 , 413 kovcs , g. & buchler , j. r. 1988 , apj , 334 , 971 letellier , c. , gouesbet , g. , soufi , f. , buchler , j.r . ,
kollth , z. , 1996 , chaos , 6 , 466 loeser , j.g .
, baliunas , s.l . ,
guinan , e.f .
, mattei , j.a . ,
wacker , s. 1986 , in cool stars , stellar systems and the sun , proceedings of the fourth cambridge workshop on cool stars , stellar systems , and the sun , zeilik , m. & gibson , d.m .
( springer - verlag , new york ) , p.460 | an analysis of the photometric observations of the light curves of the five large amplitude , irregularly pulsating stars r umi , rs cyg , v cvn , ux dra and sx her is presented .
first , multi - periodicity is eliminated for these pulsations , they are not caused by the excitation of a small number of pulsation modes with constant amplitudes .
next , on the basis of energetics we also eliminate stochasticity as a cause , leaving low dimensional chaos as the only alternative .
we then use a global flow reconstruction technique in an attempt to extract quantitative information from the light curves , and to uncover common physical features in this class of irregular variable stars that straddle the rv tau to the mira variables .
evidence is presented that the pulsational behavior of r umi , rs cyg , v cvn and ux dra takes place in a 4-dimensional dynamical phase space , suggesting that two vibrational modes are involved in the pulsation .
a linear stability analysis of the fixed points of the maps further indicates the existence of a two - mode resonance , similar to the one we had uncovered earlier in r sct : the irregular pulsations are the result of a continual energy exchange between two strongly nonadiabatic modes , a lower frequency pulsation mode and an overtone that are in a close 2:1 resonance .
the evidence is particularly convincing for r umi , rs cyg and v cvn , but much weaker for ux dra .
in contrast , the pulsations of sx her appear to be more complex and may require a 6d space
. psfig # 1 2 2 |
the solution of transfer equation in spherical geometry remains a classic problem even 75 years after the first attempts by @xcite , who used eddington approximation . in later decades
more accurate methods were given ( see * ? ? ?
* ; * ? ? ?
* for historical reviews ) .
@xcite , @xcite developed a variable eddington factor method , and computed the solution on rays of constant impact parameter ( tangents to the discrete shells and parallel to line of sight ) in 1d spherical geometry .
this is a very efficient differential equation based technique which uses feautrier solution along rays of constant impact parameter .
an integral equation method was developed by @xcite to solve the problem , again based on tangent rays approach .
@xcite presented a highly accurate finite difference method based on the first order form of the transfer equation .
all these methods were later extended to expanding , and highly extended atmospheres . however , in this paper we confine our attention to static , 1d spherical atmospheres . in a next epoch in the development of spherical radiative transfer , the integral operator techniques were proposed .
the idea of operator splitting and the use of approximate operators in iterative methods was brought to the astrophysical radiative transfer in planar media by @xcite .
@xcite extended his work with a new definition of the approximate operator .
the application of integral operator technique to the spherical transfer started with the work of @xcite and @xcite .
they used approximate operators that are diagonal , constructed from core saturation approach .
the @xmath0 operator contains the non - local coupling between all the spatial points .
@xcite showed that the diagonal part ( local coupling ) of the actual @xmath0 operator itself is an optimum choice for the ` approximate operator ' .
these methods are known as approximate lambda iteration ( ali ) methods .
the ali methods which are based on the concept of operator splitting and the use of jacobi iterative technique , were widely used in the later decades in radiative transfer theory ( see * ? ? ?
* for historical reviews ) .
@xcite used an implicit integral method to solve static spherical line transfer problems . the most recent and interesting work on spherical radiative transfer
are the papers by @xcite and @xcite both of which are based on gauss - seidel ( gs ) and successive over relaxation ( sor ) iterative techniques .
@xcite were the first to use bicg technique in astrophysics .
they use bicg with incomplete lu decomposition technique in their double splitting iterative scheme along with orthomin acceleration .
they applied it to multi - dimensional line transfer problem .
@xcite describes a variant of orthomin acceleration which uses ` minimization with respect to a set of conjugate vectors ' .
he uses a set of n ( usually n=2 or n=4 ) conjugate direction vectors which are orthogonal to each other , constructed using the residual vectors with a purpose to accelerate the convergence sequence .
@xcite developed gmres ( actually its variant called generalized conjugate residuals gcr ) method to solve the spherical transfer problem .
it is based on an application of the idea of krylov subspace techniques .
they applied it to a more general time - dependent transport with velocity fields in a medium which scatters anisotropically .
they apply gmres method to the neurino transfer .
it can also be used for radiation transfer problem , including the simple problem of 2-level atom line transfer discussed in this paper . the preconditioned bi - conjugate gradient method ( hereafter pre - bicg ,
see eg . , * ? ? ?
* ) was first introduced to the line transfer in planar media , by @xcite who describe the method and compare it with other prevalent iterative methods , namely gs / sor . in this paper
we adopt the pre - bicg method to the case of spherical media .
we also show that the ` stabilized preconditioned bi - conjugate gradient ( pre - bicg - stab ) ' is even more advantageous in terms of memory requirements but with similar convergence rate as pre - bicg method .
it is well known that the spherical radiative transfer in highly extended systems , despite being a straight forward problem , has two inherent numerical difficulties namely ( i ) peaking of the radiation field towards the radial direction , and ( ii ) the @xmath1 dilution of radiation in spherical geometry . to handle these
, it becomes essential to take a very large number of angle ( @xmath2 ) points and spatial ( @xmath3 ) points respectively .
the existing ali methods clearly slow down when extreme angular and spatial resolutions are demanded ( for example see table [ table_1 ] ) .
therefore there is a need to look for a method that is as efficient as ali methods , but faster , and is relatively less sensitive to the grid resolution . the pre - bicg method and a variant of it
provide such an alternative as we show in this paper .
governing equations are presented in sect .
[ transfer ] . in sect .
[ geo ] , we define the geometry of the problem and the specific details of griding . in sect .
[ model ] , the benchmark models are defined .
we briefly recall the jacobi , and gs / sor methods in sect .
[ ali ] . in sect .
[ bicg_sphere ] we describe the pre - bicg method .
the computing algorithm is presented in sect .
[ prebicg_alg ] . in sect .
[ prebicg - stab ] we describe the pre - bicg - stab method briefly , and we give the computing algorithm in sect .
[ stab_alg ] . in sect .
[ bicg_ali ] we compare the performance of pre - bicg with the jacobi , and gs / sor methods . in sect .
[ results ] we validate this new method , by comparing with the existing well known benchmark solutions in spherical line radiative transfer theory .
conclusions are presented in sect .
[ conclusions ] .
in this paper we restrict ourselves to the case of a 2-level atom model .
further , we assume complete frequency redistribution ( crd ) .
the transfer equation in divergence form is written as @xmath4\nonumber \\ & & \times [ \bm{s}(x , r)-\bm{i}(r , \mu , x ) ] .
\label{transfer_div}\end{aligned}\ ] ] here , @xmath5 is the specific intensity of radiation , @xmath6 - the source function , @xmath7 - the radial distance , @xmath2 - the direction cosine , @xmath8 - the frequency measured in doppler width units from line center , @xmath9 - the line profile function , and @xmath10 , @xmath11 - line center and continuum opacities respectively .
the differential optical depth element is given by @xmath12 there are several methods which use the above form of the transfer equation ( see * ? ? ?
in our paper we solve the transfer equation on a set of rays tangent to the spherical shells .
it is written as @xmath13 \nonumber \\ & & \times [ \bm{s}(x , r)-\bm{i}^{\pm}(z , p , x ) ] , \label{transfer_red}\end{aligned}\ ] ] for the outgoing ( + ) and incoming ( - ) rays respectively . here
@xmath14 is the distance along the tangent rays and @xmath15 is the distance from the center to the points on the vertical axis ( the mid - line ) , where the tangent rays intersect it ( see fig .
[ fig_geometry ] ) .
the direction cosines @xmath16 are related to @xmath15 by @xmath17 for a shell of radius @xmath7 .
the optical depth scale along the tangent rays are now computed using @xmath18 . in practical work , due to the symmetry of the problem
, it is sufficient to perform the computations on a quadrant only .
the source function is defined as @xmath19 @xmath20 is the continuum source function taken as the planck function @xmath21 throughout this paper .
the monochromatic optical depth scale @xmath22 $ ] , with @xmath23 along the tangent rays . for simplicity , hereafter we omit the subscript @xmath14 from @xmath24 and write @xmath3 to denote @xmath24 .
the line source function is given by @xmath25 with the thermalization parameter defined in the conventional manner as @xmath26 , where @xmath27 and @xmath28 are collisional and radiative de - excitation rates .
the intensity along the rays is computed using the formal solution integral @xmath29+\nonumber \\ & & \int_{\tau}^{t } \exp[-\delta \tau'_{x } ] \bm{s}(\tau')[\phi(x)+\beta_{\rm c}]{\rm d}\tau ' . \label{int_plus}\end{aligned}\ ] ] the corresponding integral for the incoming rays is @xmath30+\nonumber \\ & & \int_{0}^{\tau } \exp[-\delta \tau'_{x } ] \bm{s}(\tau')[\phi(x)+\beta_{\rm c}]{\rm d}\tau ' .
\label{int_minus}\end{aligned}\ ] ] here , @xmath31 represents the inner boundary condition imposed at the core and along the mid - vertical line ( see fig .
[ fig_geometry ] ) .
@xmath32 is the outer boundary condition specified at the surface of the spherical atmosphere .
when the above formal integral is applied to a stencil of short characteristic ( mop ) along a tangent ray , it takes a simple algebraic form @xmath33 + \nonumber \\ & & \bm{\psi}^{\pm}_{\rm m}(\tau , p , x)\bm{s}_{\rm m}+\bm{\psi}^{\pm}_{\rm o}(\tau , p , x)\bm{s}_{\rm o}+ \nonumber \\ & & \bm{\psi}^{\pm}_{\rm p}(\tau , p , x)\bm{s}_{\rm p } , \label{int_sc}\end{aligned}\ ] ] where @xmath34 are the source function values at m , o and p points on a short characteristic .
the coefficients @xmath35 are calculated following the method described in @xcite . in fig .
[ fig_geometry ] , we show the geometry used for computing the specific intensity @xmath36 along rays of constant impact parameter . in a spherically symmetric medium , we first discretise the radial co - ordinate @xmath7 ( @xmath37 ) , where @xmath38 is the core radius , and @xmath39 is the outer radius of the atmosphere . the radial grid is given by @xmath40 , where @xmath41 is the radius of the outer most shell , and @xmath42 is that of the inner most shell .
@xmath43 is the probability that the direction of propagation of an emitted photon lies within an element of solid angle @xmath44 . in the azimuthally symmetric case ,
it is @xmath45 . to calculate the mean intensity @xmath46 in plane parallel geometry
, we integrate the intensity over the angular variable @xmath47 itself . in spherical medium
, we have one to one correspondence between the ( @xmath2,@xmath7 ) and the ( @xmath15,@xmath7 ) system . in ( @xmath15,@xmath7 ) system , the probability that a photon is emitted with its impact parameter between @xmath15 and @xmath15+@xmath48 , propagating in either positive @xmath2 or negative @xmath2 direction is @xmath49 .
the direction cosines made by the rays in the ( @xmath2,@xmath7 ) space , with a tangent ray of constant @xmath15 value , are given by @xmath50 at different radii @xmath51 .
therefore the angular integration factor @xmath45 can be changed to @xmath49 ( see * ? ? ?
_ the @xmath15 - grid construction _
: if @xmath52 is the number of core rays , then the @xmath15-grid for the core rays is computed using : @xmath53 the number of lobe rays equals the number of radial points . for lobe rays ,
the @xmath15-grid is same as radial @xmath7-grid .
it is constructed using : @xmath54 + where @xmath55=the number of radial points .
thus , the total number of impact parameters is @xmath56 .
we have followed @xcite in defining the @xmath15-grid in this manner .
geometrical distances along the rays of constant impact parameter are constructed as follows : @xmath57 for spherical shells we perform several tests using power - law type variation of density . for such atmospheres ,
the line and continuum opacities also vary as a power law given by @xmath58 let @xmath59 and @xmath60 denote the proportionality constants for @xmath10 and @xmath11 respectively . the constant @xmath59 can be determined using the optical depth at line center @xmath61 . for a power law with index @xmath62 , @xmath63 using the given input value of @xmath64 we can compute the constant @xmath65 .
we use voigt profile with damping parameter @xmath66 or the doppler profile for the results presented in this paper .
the spherical shell atmosphere is characterized by the following parameters : ( @xmath39 , @xmath62 , @xmath61 , @xmath66 , @xmath67 , @xmath68 , @xmath69 ) .
we recall that @xmath39 is the outer radius of the spherical atmosphere surrounding a hollow central cavity of radius @xmath38 .
when @xmath70 we recover the plane parallel limit .
for the spherical shell atmospheres , we take @xmath71 as the unit of length to express the radial co - ordinate .
the boundary conditions are specified at the outer boundary ( @xmath72 ) and the inner boundary .
there are two types of inner boundary conditions : + ( a ) _ emitting core _ : + core rays : @xmath73 lobe rays : @xmath74 @xmath75 along the mid - vertical .
+ ( b ) _ hollow core : _ + for both the core and the lobe rays : @xmath76 the hollow core boundary condition is also called ` planetary nebula boundary condition ' ( see * ? ? ?
it is clear that a spherical shell with a hollow core is equivalent to a plane parallel slab of optical thickness @xmath77 with symmetry about the mid - plane at @xmath78 .
we use spherical shell atmospheres for most of our studies .
the ali methods have been successfully used for the solution of transfer equation in spherical shell atmospheres ( see eg . , * ? ? ? * and references therein ) .
these authors use the jacobi iterative methods ( first introduced by @xcite ) for computing the source function corrections .
recently the gs method has been proposed to solve the same problem ( see eg .
* ; * ? ? ?
@xcite proposed the gmres method for solving spherical radiative transfer problem .
gmres and pre - bicg both belong to krylov subspace technique . in this paper
we compute the spherical transfer solutions by jacobi and gs / sor methods , and compare with the solutions computed using the pre - bicg method . for the sake of clarity
, we recall briefly the steps of jacobi and gs / sor methods .
+ _ jacobi iteration cycle _ : the source function corrections are given by @xmath79 } , \label{deltas_jacobi}\ ] ] for the @xmath80th iterate . here
@xmath81 is the depth index .
the @xmath82 is the approximate operator which is simply taken as the diagonal of the actual @xmath0 operator defined through @xmath83=\bm{\bar{j } } ; \label{j_bar}\ ] ] @xmath84 _ gs / sor iteration cycle _ : the essential difference between the jacobi and gs / sor methods is the following : @xmath85 here the parameter @xmath86 is called the relaxation parameter which is unity for the gs technique .
the sor method is derived from the gs method by simply taking @xmath87 ( see * ? ? ? * for details ) .
the source function correction for the gs method is given by @xmath88 } , \label{deltas_gs}\ ] ] where the quantity @xmath89 denotes the mean intensity computed using new values of the source function as soon as they become available .
for those depth points for which the source function correction is not yet complete , gs method uses the values of the source function corresponding to the previous iteration ( see * ? ? ? * ) . for clarity
we explain how the gs algorithm works in spherical geometry , on rays of constant impact parameter .
+ begin loop over iterations + begin loop over radial shells with index @xmath90 + begin loop over impact parameters ( or directions ) with increasing @xmath15 + for the @xmath80th iteration : * for the incoming rays ( @xmath91 ) : * + ( * reverse sweep along radial shells * ) \(a ) this part of the calculations start at the outer boundary for all impact parameter rays .
+ ( b ) @xmath92 are first calculated for a given radial shell @xmath90 using @xmath93 , @xmath94 and @xmath95 .
+ ( c ) the partial integral @xmath96 are calculated before proceeding to the next shell .
this part of the calculations is stopped when the core ( for the core rays ) and the mid - vertical line ( for the lobe rays ) are reached .
+ * for outgoing rays ( @xmath97 ) : * + ( * forward sweep along radial shells * ) \(d ) this part of the calculations start at the inner boundary .
first , for the radial shell with @xmath98 + @xmath99 is calculated , using boundary conditions @xmath100 .
+ ( e ) @xmath101 is computed and the source function is updated using @xmath102 . + ( f ) for the next radial shell @xmath103 : + to calculate @xmath104 by applying the short characteristic formula , @xmath105 , @xmath106 and @xmath107 are needed .
already @xmath108 , @xmath109 and @xmath110 are available .
gs takes advantage of the available new source function at @xmath98 .
@xmath104 is calculated with this set of source functions .
+ ( g ) then @xmath111 are calculated using @xmath104 .
+ ( h ) note that , @xmath112 was calculated using @xmath113 , @xmath109 , and @xmath110 whereas @xmath111 used the `` updated '' source function @xmath108 . therefore @xmath114 is corrected by adding the following correction : @xmath115 \(i ) @xmath116 and @xmath117 are now calculated .
+ ( j ) since `` updated '' @xmath118 at @xmath103 is also available now , before going to the next radial shell it is appropriate to correct the intensity at the present radial shell by adding to it , the following correction term @xmath119 end loop over impact parameters ( or directions ) + end loop over radial shells + end loop over iterations .
in this section we first describe the essential ideas of the pre - bicg method .
the complete theory of the method is described in @xcite .
we recall that the 2-level atom source function with a background continuum is given by @xmath120 it can be re - written as @xmath121 where @xmath122 from equations ( [ source_line ] ) , ( [ j_bar ] ) and ( [ jbar ] ) , we get @xmath123 + \epsilon \bm{b}_{\nu}(r)\}+ \nonumber \\ & & ( 1-\tilde p(x , r))\bm{s}_{\rm c}(r ) . \label{source_total1}\end{aligned}\ ] ] therefore the system of equations to be solved becomes @xmath124\bm{s}(x , r)= \tilde p(x , r ) \epsilon \bm{b}_{\nu}(r)+ \nonumber \\ & & ( 1-\tilde p(x , r))\bm{s}_{\rm c}(r ) , \label{system}\end{aligned}\ ] ] which can be expressed in a symbolic form as @xmath125;\,\bm{y}=\bm{s}(x , r ) .
\label{system_gen}\end{aligned}\ ] ] the vector @xmath126 represents quantities on the rhs of equation ( [ system ] ) .
now we describe briefly , how the pre - bicg method differs from ali based methods .
let @xmath127 denote the @xmath80-dimensional euclidean space of real numbers .
* definition * : the pre - bicg algorithm is a process involving projections onto the @xmath128-dimensional subspace ( @xmath129 ) of @xmath127 @xmath130 and also being orthogonal to another @xmath128-dimensional subspace of @xmath127 @xmath131 here @xmath132 is taken as the initial residual vector @xmath133 with @xmath134 the initial guess for the solution of equation ( [ system_gen ] ) .
the vector @xmath135 is taken as arbitrary such that the inner product @xmath136 .
the method recursively constructs a pair of bi - orthogonal bases @xmath137 and @xmath138 for @xmath139 and @xmath140 respectively , such that they satisfy the bi - orthogonality condition @xmath141 . for the purpose of application to the radiative transfer theory it is convenient to write the pre - bicg steps in the form of an algorithm for simplicity we drop the explicit dependence on variables .
our goal is to solve equation ( [ system_gen ] ) . in this section
the symbols @xmath142 and @xmath143 are used to be in conformity with the standard notation of residual and conjugate direction vectors .
they should not be confused with the radius vector @xmath51 and impact parameter @xmath144 which appear in spherical radiative transfer theory .
\(a ) the very first step is to construct and store the matrix @xmath145 ( which does not change with iterations , for the cases considered here , namely 2-level atom model ) .
details of computing @xmath145 efficiently is described in appendix [ appendix_lambda ] .
we follow the preconditioned version of the bicg method .
preconditioning is a process in which the original system of equations is transformed into a new system , which has faster rate of convergence .
for example , this can be done by solving the new system @xmath146 where @xmath147 is an appropriately chosen matrix , called the `` preconditioner '' ( see also equation 2 of * ? ? ? * ) .
this preconditioner is chosen in such a way that , + ( i ) the new system should be easier to solve , + ( ii ) @xmath148 itself should be inexpensive to operate on an arbitrary vector , + ( iii ) the preconditioning is expected to increase the convergence rate . +
the choice of the preconditioner depends on the problem at hand . when an appropriate @xmath149 is chosen such that the amplification matrix @xmath150 $ ] has as small a maximum eigen value as possible ( see * ? ? ?
* ) , the convergence rate is enhanced .
what enables the convergence of ali , that satisfies the above property , and simplest to manipulate , is the diagonal of the @xmath0 itself .
therefore the amplification matrix @xmath151 $ ] with a diagonal form for @xmath149 is a simple and natural choice as a ` preconditioner ' .
we construct the preconditioner matrix @xmath147 by taking it as the diagonal of @xmath152 .
+ ( b ) an initial guess for the source function is @xmath153 where the thermal part @xmath154 is taken as an initial guess for @xmath155 .
+ ( c ) the formal solver is used with @xmath134 as input to calculate @xmath156 .
+ ( d ) the initial residual vector is computed using @xmath157 \(e ) the initial bi - orthogonal counterpart @xmath158 for @xmath159 is chosen such that we have @xmath160 .
one can choose @xmath161 itself .
such an initial choice of @xmath158 vector is necessary , as the method is based on the construction of bi - orthogonal residual vectors @xmath142 and @xmath162 recursively , for @xmath163 , where @xmath128 is the number of iterations required for convergence . the process of constructing the bi - orthogonal vectors gets completed , once we reach the convergence . in other words , the number of bi - orthogonal vectors necessary to guarantee a converged solution represents the actual number of iterations itself .
it is useful to remember that when we refer to ` bi - orthogonality ' hereafter , say eg .
, of the residual vectors @xmath142 , @xmath164 we simply mean that @xmath165 for @xmath166 , but @xmath167 need not be unity .
+ ( f ) the bi - orthogonalization process makes use of conjugate direction vectors @xmath168 and @xmath169 for each iteration .
they can be constructed during the iterative process , again through recursive relations .
an initial guess to these vectors is made as @xmath170 and @xmath171 .
+ ( g ) the preconditioned initial residual vectors @xmath172 are computed using @xmath173 \(h ) for @xmath174 the following steps are carried out until convergence : + ( i ) using the formal solver with @xmath143 as input ( instead of actual source vector @xmath175 ) , @xmath176 $ ] is obtained .
+ ( j ) @xmath177 $ ] is computed using @xmath178=\bm{p}_{\rm i}-(1-\epsilon)\tilde p \,\bm { \bar j}[\bm{p}_{\rm i } ] .
\label{a_l } \nonumber\ ] ] \(k ) the inner products @xmath179 , \bm{p}_{\rm i}^ * \ra \quad \textrm{and } \quad \la \bm{r}_{\rm i } , \bm{\zeta}_{\rm i}^ * \ra , \label{a_l_lstar}\ ] ] are computed and used to estimate the quantity @xmath180 , \bm{p}_{\rm i}^ * \ra}.\ ] ] \(l ) the new source function is obtained through @xmath181 _ test for convergence _
: let @xmath182 denote the convergence criteria . if @xmath183 then iteration sequence is terminated .
otherwise it is continued from step ( m ) onwards .
the convergence criteria @xmath182 is chosen depending on the problem . +
( m ) following recursive relations are used to compute the new set of vectors to be used in the @xmath184th iteration : @xmath185 , \\ & & \bm{r}_{\rm i+1}^*=\bm{r}_{\rm i}^*-\alpha_{\rm i } \hat { a}^t [ \bm{p}_{\rm i}^*],\\ & & \bm{\zeta}_{\rm i+1}^*=\hat{m}^{-1 } \bm{r}_{\rm i+1}^*.\end{aligned}\ ] ] \(n ) the quantity @xmath186 is computed using @xmath187 \(o ) the conjugate direction vectors for the @xmath184th iteration are computed through @xmath188 \(p ) the control is transferred to step ( g ) . the converged source function @xmath175 is finally used to compute the specific intensity everywhere within the spherical medium .
in spite of higher convergence rate , computation and storage of the @xmath189 matrix is a main dis - advantage of the pre - bicg method . to avoid this , and to make use of only the ` action ' of @xmath152 matrix on an arbitrary vector , a method called ` bicg - squared '
was developed ( see * ? ? ? * for references and details ) , which is based on squaring the residual polynomials .
later it was improved by re - defining the residual polynomial as a product of two polynomials and obtaining a recursive relation for the new residual polynomial .
this product involves residual polynomial of the pre - bicg method and a new polynomial which ` smoothens ' the iterative process . in this section
we give the computing algorithm of the pre - bicg - stab method as applied to a radiative transfer problem .
as described below , we can avoid computing and storing of the @xmath189 matrix in the pre - bicg - stab method .
however we would now need to call the formal solver twice per iteration unlike in pre - bicg method , where it is called only once .
this results in an increase in number of operations per iteration when compared to pre - bicg method , causing a slight increase in the cpu time per iteration . in spite of
these the pre - bicg - stab method turns out to be always faster than the regular pre - bicg method in terms of convergence rate ( lesser number of iterations for convergence ) .
now we give the algorithm of pre - bicg - stab method to solve the system @xmath190 . here
@xmath147 is a suitably chosen preconditioner matrix .
the computing algorithm is organized as follows : + ( a ) first initial preconditioned residual vectors and conjugate direction vectors are defined through @xmath191 @xmath192 \(b ) for @xmath193 the following steps are carried out until convergence .
+ ( c ) using @xmath194 instead of the source function a call to the formal solver is made to compute @xmath195 .
( d ) the coefficient @xmath196 can be evaluated now as @xmath197 \(e ) another vector @xmath198 is calculated as @xmath199 \(f ) using @xmath198 in place of the source function a call to the formal solver is made to obtain @xmath200 .
+ ( g ) the coefficient @xmath201 is estimated as @xmath202 \(h ) the updated new source function is calculated as @xmath203 \(i ) test for convergence is made as in the pre - bicg algorithm . + ( j ) before going to the next iteration a set of recursive relations are used to compute residual vectors @xmath204 and conjugate direction vectors @xmath205 for the next iteration , where the coefficient @xmath206 is @xmath207 \(k ) the control is now transferred to the step ( b ) .
there are two characteristic quantities that define iterative techniques .
they are ( a ) convergence rate , which is nothing but the maximum relative change ( mrc ) defined as @xmath208 and ( b ) the total cpu time @xmath209 required for convergence .
@xmath209 is the time taken to reach a given level of convergence , taking account only of the arithmetic manipulations within the iteration cycle .
we also define a quantity called the true error @xmath210 and use it to evaluate these methods . in this section
we compare @xmath211 and @xmath209 for the jacobi , gs , sor , pre - bicg and the pre - bicg - stab methods . the sor parameter used is 1.5 .
it is worth noting that the overrates ( the time taken to prepare the necessary set up , before initiating the iterative cycle ) are expected to be different for different methods .
for instance , in jacobi and gs / sor this is essentially the cpu time required to set up the @xmath149 matrix . in the pre - bicg method this involves the time taken to construct the @xmath189 matrix , which is a critical quantity of this method .
the pre - bicg method is described in this paper in the context of a 2-level atom model , because of which , we do not need to update the @xmath189 matrix at each iteration .
for the pre - bicg - stab method it is the time taken to construct the preconditioner matrix @xmath147 . fig .
[ fig_mrc ] shows a plot of @xmath211 for different methods .
we can take @xmath211 as a measure of the convergence rate .
@xcite show that it always becomes necessary to use high resolution grids , to achieve high accuracy of the solution ( see also sect .
[ intro ] of this paper ) .
this is especially true in the case of spherical radiative transfer where a spatial grid with a large number of points per decade becomes necessary to achieve reasonable accuracy . in the following
we discuss how different methods respond to the grid refinement .
it is a well known fact with the ali methods , that the convergence rate is small when the resolution of the depth grid is very high .
in contrast they have a high convergence rate in low resolution grids .
on the other hand the @xmath211 of pre - bicg and pre - bicg - stab methods have higher convergence rate even in a high resolution grid . fig . [ fig_mrc](a ) shows @xmath211 for different methods when a low resolution spatial grid is used ( 5pts / d in the logarithmic scale for @xmath3 grid ) .
the jacobi method has a low convergence rate . in comparison
, gs has a convergence rate which is twice that of jacobi .
sor has a rate that is even better than that of gs .
however pre - bicg and the pre - bicg - stab methods have the higher convergence rate . fig .
[ fig_mrc](b ) and [ fig_mrc](c ) are shown for intermediate ( 8 pts / d ) and high ( 30 pts / d ) grid resolutions .
the essential point to note is that , as the grid resolution increases , the convergence rate decreases drastically and monotonically for the jacobi and the gs methods .
it is not so drastic for the sor method which shows non - monotonic dependence on grid resolution .
the pre - bicg and pre - bicg - stab methods exhibit again a monotonic behaviour apart from being relatively less sensitive to the grid resolution . in table
[ table_1 ] we show what happens when we set convergence criteria to progressively smaller values ( @xmath212@xmath213 , @xmath214 , and @xmath215 for tables [ table_1a ] , [ table_1b ] and [ table_1c ] respectively ) for various grid resolutions .
the model used to compute these results is ( @xmath62 , @xmath39 , @xmath61 , @xmath66 , @xmath67 , @xmath68 , @xmath216)= ( 0 , 10 , @xmath217 , @xmath218 , @xmath219 , 0 , 1 ) .
the idea is to demonstrate that for a given grid resolution ( corresponding rows of the tables [ table_1a ] , [ table_1b ] and [ table_1c ] ) , all the methods show a monotonic increase in the number of iterations for convergence , as we decrease the @xmath220 . on the other hand pre - bicg and pre - bicg - stab require much less number of iterations to reach the same level of accuracy .
_ cpu time considerations _ :
table [ table_2 ] shows the cpu time requirements for the methods discussed in this paper .
the model used to compute these test cases is ( @xmath62 , @xmath39 , @xmath61 , @xmath67 , @xmath68 , @xmath69 , @xmath220)= ( 0 , 300 , @xmath221 , @xmath219 , 0 , 1 , @xmath214 ) .
the grid resolution considered is 30 pts / d .
the cpu time for convergence can be defined as the computing time required to complete the convergence cycle and reach a fixed level of accuracy .
we recall that the overrates in computing time is the time taken to prepare the necessary set up before initiating the iterative cycle .
total computing time is the sum of these two . in appendix [ appendix_lambda ]
we discuss in detail how to construct @xmath149 matrix for jacobi , gs / sor methods and @xmath152 and @xmath147 matrices for pre - bicg and pre - bicg - stab methods respectively with an optimum effort .
construction of these matrices constitutes the overrates in computing time of each method .
the first row of table [ table_2 ] shows that pre - bicg is the fastest to complete the convergence cycle .
the reason why pre - bicg - stab takes slightly longer time than pre - bicg is explained at the end of sect .
[ prebicg - stab ] .
the second row of table [ table_2 ] shows that all methods except pre - bicg take nearly 8 seconds as overrates for the chosen model .
pre - bicg takes additional 3 - 4 seconds as explicit integrals are performed for computing off - diagonal elements also ( unlike the other methods where such integrals are performed only for diagonal elements ) .
the last row of table [ table_2 ] shows that in terms of total cpu time requirement , the other methods fall behind the pre - bicg and the pre - bicg - stab .
pre - bicg seems to be a bit faster compared to pre - bicg - stab for the particular model chosen .
however it is model dependent .
for instance , as the contribution towards overrates increases , pre - bicg - stab clearly stands out as the fastest method of all , discussed in this paper .
we now study the true errors in these methods ( see figure [ fig_true ] ) .
the model parameters are ( @xmath62 , @xmath39 , @xmath61 , @xmath67 , @xmath68 , @xmath216)= ( 0 , 10 , @xmath217 , @xmath219 , 0 , 1 ) . a coherent scattering limit is used . to define a true error
, we need a so called ` exact solution ' . except for highly idealized cases ,
exact solutions do not exist . for practical purposes , the exact solution can be defined as a solution obtained on a spatial grid of resolution that is three times larger than the grid resolution of the model that we are interested in .
also , we extend the iteration until @xmath211 reaches an extremely small value of @xmath222 . the source function computed in this way
can be called @xmath223 ( fully converged solution on an infinite resolution ) ( see * ? ? ?
the source function at the @xmath80th iterate is denoted by @xmath224 .
we define the true error as @xmath225 following @xcite . in fig .
[ fig_true](a ) we show @xmath210 computed for the pre - bicg method using three grid resolutions , namely 10 pts / d , 14 pts / d and 20 pts / d .
the plateau of each curve represents the minimum value of the true error reached for a given grid resolution .
we notice that as the resolution increases , @xmath210 gradually decreases in magnitude as expected . in fig .
[ fig_true](b ) we show @xmath210 computed for the pre - bicg - stab method .
the model parameters are same as in fig .
[ fig_true](a ) . clearly , pre - bicg - stab shows a smooth decrement of true error compared to pre - bicg , because of the smoothing polynomial used to define the residual vectors . in fig .
[ fig_true](c ) we compare the decrement of true errors for different iterative methods .
the grid resolution chosen is 14 pts/ d with other model parameters being same as in figs .
[ fig_true](a ) , ( b ) .
the decrease of the true errors follows the same pattern in all the iterative methods , although the number of iterations required for @xmath210 to reach a constant value ( plateau ) depends on the method . to reach the same level of true error ,
the pre - bicg and pre - bicg - stab methods require considerably less number of iterations , when compared to the other three .
suppose that @xmath152 is an @xmath226 matrix .
the solution to the problem @xmath227 is a vector of length @xmath55 . in an @xmath55-dimensional vector space @xmath228 the maximum number of linearly independent vectors is @xmath55 .
hence , there can at the most be @xmath55 orthogonal vectors in @xmath228 .
the pre - bicg method seeks a solution by constructing orthogonal vectors .
we recall that the residual counterpart vectors @xmath229 constructed during the iteration process are orthogonal to the initial residual vector @xmath230 .
thus , when we reach convergence after @xmath231 iterations , we will have a set of @xmath232 orthogonal vectors @xmath233 .
from the arguments given above , it is clear that @xmath234 , namely in the pre - bicg method , ` the convergence must be reached theoretically in at the most @xmath55 steps ( or iterations ) ' .
this sets an upper limit to the number of iterations to reach convergence ( see also * ? ? ?
for example when the dimensionality of a problem is high ( very large value of @xmath55 ) , the pre - bicg method ensures convergence in at the most @xmath55 iterations . a theoretical upper bound on the number of iterations also exists for the pre - bicg - stab method , whereas the other methods do not have such a theoretical upper bound . in practice
we find that pre - bicg and pre - bicg - stab methods actually require much less number of iterations than @xmath55 , even when @xmath55 is large .
the main purpose of this paper is to propose a new method to solve the line transfer problems in spherically symmetric media . in this section
we show some illustrative examples in order to compare with the famous benchmarks for spherical transfer solutions presented in the papers by @xcite . in fig .
[ fig_source ] we show source functions for different test cases . fig .
[ fig_source](a ) shows the source functions for @xmath235 and @xmath236 for @xmath237 and @xmath238 .
other model parameters are ( @xmath62 , @xmath68 , @xmath69)=(2 , 0 , 1 ) . we use a doppler profile to compare with the results of @xcite .
plane parallel result is also shown for comparison .
when @xmath235 we observe that the thermalization is reached at the thermalization length for the doppler profile namely @xmath239 .
when @xmath236 thermalization does not occur .
clearly the minimum value of the source function is @xmath240 . for the large values of @xmath237 and opacity index @xmath241 , as @xmath7 increases the opacity decreases steadily and the source function indeed approaches this minimum value near the surface layers . for this case ,
the departure of the source function from planar limit is severe near the surface .
it can be shown ( dashed line of fig .
[ fig_source](b ) , see also fig . 3 of @xcite ) that this departure is not so acute when @xmath242 , but is more acute when @xmath243 ( dash triple - dotted line in fig . [ fig_source](b ) ) .
[ fig_source](b ) we plot source function for the same model as fig .
[ fig_source](a ) but for various values of @xmath244 . for negative @xmath244 , the distinction between @xmath245 vs. @xmath3 curves for different @xmath62 is small . for positive @xmath244 ,
the effects are relatively larger ( see dot - dashed and long dashed curves in fig .
[ fig_source](b ) ) . in fig .
[ fig_source](c ) , we show source function variation for a range of spherical extensions @xmath39 . we have chosen an effectively optically thin model ( @xmath61 , @xmath67)=(@xmath246 , @xmath215 ) because in such a medium , thermalization effects do not completely dominate over the effects of sphericity .
other parameters are same as in fig .
[ fig_source](b ) . clearly , the decrease in the value of source function throughout the atmosphere is monotonic , with an increase in the value of @xmath39 from 1 to @xmath247 . in fig .
[ fig_intensity1 ] , we show effects of limb darkening in spherical atmospheres for @xmath248 and @xmath249 . the other model parameters for fig . [
fig_intensity1](a ) are ( @xmath62 , @xmath61 , @xmath67 , @xmath68 , @xmath216)= ( 2 , @xmath250 , @xmath219 , 0 , 1 ) . a doppler line profile is used . from the figure , we notice absorption in the line core and emission in the near line wings ( @xmath251 ) for @xmath252 and @xmath253 .
this is the characteristic self reversal observed in spectral lines formed in extended spherical atmospheres .
the self reversal decreases gradually as @xmath254 increases , and finally vanishes for large values of @xmath254 .
indeed for extreme value of @xmath255 , we observe a pure emission line . in fig .
[ fig_intensity1](b ) we show line profiles formed in a semi - infinite spherical medium .
the model parameters are same as in fig .
[ fig_intensity1](a ) except for ( @xmath39 , @xmath61)=(300 , @xmath256 ) .
the profiles for a range of @xmath257 are shown . for the core rays ( @xmath258 )
we see a pure absorption line due to thermalization of source function . for other angles ,
as expected we see chromospheric type self - reversed emission lines , formed in the lobe part of the spherical medium .
in this paper we propose a robust method called preconditioned bi - conjugate gradient ( pre - bicg ) method to solve the classical problem of line transfer in spherical media .
this method belongs to a class of iterative methods based on the projection techniques . we briefly present the method , and the computing algorithm .
we also present a transpose - free variant called the stabilized preconditioned bi - conjugate gradient ( pre - bicg - stab ) method which is more advantageous in some of its features .
the pre - bicg and pre - bicg - stab methods are validated in terms of its efficiency and accuracy , by comparing with the contemporary iterative methods like jacobi , gs and sor . to calculate the benchmark solutions we use spherical shell atmospheres .
few difficult test cases are also presented to show that the pre - bicg and pre - bicg - stab are efficient numerical methods for spherical line transfer .
l. s. anusha likes to thank dr .
han uitenbroek , dr . a. asensio ramos and dr
. m. sampoorna for useful discussions .
asensio ramos , a. , & trujillo bueno , j. 2006 , in eas pub .
18 , radiative transfer and applications to very large telescopes , ed .
stee , 25 auer , l. h. 1984 , in methods in radiative transfer , ed .
kalkofen , w. ( cambridge : cambridge university press ) , 237 auer , l. h. 1991 , in stellar atmospheres : beyond classical models , ed .
crivellari , l. , hubeny , i. , & hummer , d. g. ( dordrecht : kluwer academic publishers ) , 9 auer , l. h. , fabiani bendicho , p. , & trujillo bueno , j. 1994 , , 292 , 599 cannon , c. j. 1973 , jqsrt , 13 , 627 chandrasekhar , s. 1934 , mnras , 94 , 522 chevallier , l. , paletou , f. , & rutily , b. 2003 , , 411 , 221 daniel , f. , & cernicharo , j. 2008 , , 488 , 1237 gros , m. , crivellari , l. , & simonneau , e. 1997 , , 489 , 331 hamann , w - r .
1985 , , 145 , 443 hamann , w - r .
2003 , in asp conf .
288 , stellar atmosphere modeling , ed .
hubeny , i. , mihalas , d. , & werner , k. ( san francisco : asp ) , 171 hestenes , m. r. , & stiefel , e. 1952 , journal of research of the national bureau of standards , 49(6 ) , 409 hubeny , i. , & burrows , a. 2007 , , 659 , 1458 hubeny , i. in asp conf .
288 , stellar atmosphere modeling , ed .
hubeny , i. , mihalas , d. , & werner , k. ( san francisco : asp ) , 17 hummer , d. g. , & rybicki , g. b. 1971 , mnras , 152 , 1 klein , r. i. , castor , j. i. , greenbaum , a. , taylor , d. , & dykema , p. g. 1989 , jqsrt , 41 , 199 kosirev , n. a. 1934 , mnras , 94 , 430 kunasz , p. b. , & auer , l. h. 1988 , jqsrt , 39 , 67 kunasz , p. b. , & hummer , d. g. 1973 , mnras , 166 , 57 kunasz , p. b. , & hummer , d. g. 1974 , mnras , 166 , 19 mihalas , d. 1978 , stellar atmospheres ( 2nd ed . ; san francisco : freeman ) olson , g. l. , auer , l. h. , & buchler , j. r. 1986 , jqsrt , 35 , 431 paletou , f. , & anterrieu .
2009 , arxiv:0905.3258 , http://arxiv.org/abs/0905.3258 peraiah , a. 2002 , an introduction to radiative transfer ( cambridge university press ) peraiah , a. , & grant , i. p. 1973 , jima , 12 , 75 saad , y. 2000 , iterative methods for sparse linear systems ( 2nd ed . )
scharmer , g. b. 1981 , , 249 , 720 schmid - burgk , j. 1974 , , 32 , 73 trujillo bueno , j. , & fabiani bendicho , p. 1995 , , 455 , 646 werner , k. , & husfeld , d. 1985 , , 148 , 417 crrrrr & jacobi & gs & sor & pre - bicg & pre - bicg - stab + cpu time for convergence & 7 min 49 sec & 4 min 4 sec & 1 min 18 sec & 27 sec & 42 sec + overrates in computing & 6 sec & 6 sec & 6 sec & 9 sec & 6 sec + total computing time & 7 min 55 sec & 4 min 10 sec & 1 min 24 sec & 36 sec & 48 sec +
in pre - bicg method , it is essential to compute and store the @xmath189 matrix . a brute force - fully numerical way of doing
this is as follows .
suppose that the dimension of @xmath152 matrix is @xmath226 , where @xmath55 is the number of depth points . by sending a @xmath259-source function @xmath55 times , to a formal solver subroutine , @xmath55 columns of @xmath152 matrix can be calculated . but
this takes a large amount of cpu time especially for large values of @xmath55 .
instead , there is a semi - analytic way of calculating the @xmath152 matix . by substituting the @xmath259-source function in the expression for the intensity on a short - characteristic stencil of 3-points ( mop in standard notation )
we can obtain recursive relations for intensity matrix elements @xmath260 , which can then be integrated over frequencies and angles to get the @xmath0 matrix . finally , @xmath261 $ ] .
the diagonal of @xmath0 is @xmath149 and the diagonal of @xmath152 is the preconditioner matrix @xmath147 . for plane parallel full - slab problem , this is given in @xcite .
in radiative transfer problems with spherical symmetry , it is sufficient to compute the solution on a quadrant .
however this causes a tricky situation , in which we have to define a mid - line ( see fig . [ fig_geometry ] ) on which a non - zero boundary condition @xmath262 has to be specified . for the outgoing rays ,
the mid - vertical line is the starting grid point for a given ray . since the intensity at the starting point is non - zero ( @xmath262 )
, intensity at any interior point depends on the intensity at all the previous points .
recall that @xmath263 and @xmath264 and so on until we reach the mid - line .
it is easy to see from above equations that intensity calculation at a short - characteristic stencil mop is not confined only to the intensity on mop , but also on all previous points , through spatial coupling .
this is specific to performing radiative transfer on a spherical quadrant .
note that even for the construction of a diagonal @xmath0 , all the elements @xmath265 of the intensity matrix has to be computed .
we present below the recursive relations to compute @xmath266 .
+ * for the incoming rays ( @xmath91 ) - reverse sweep * + do @xmath267 + consider an arbitrary spatial point @xmath268 .
the delta - source vector is specified as + @xmath269 do @xmath270 , where @xmath271 is the total number of impact parameters .
+ for the inner boundary points , define @xmath272 for the core rays and @xmath273 for lobe rays .
the index @xmath274 represents the total number of points on a given ray of constant impact parameter @xmath15 .
the external boundary condition has to be taken as zero for constructing integral operators like @xmath0 .
@xmath275 else + if ( @xmath291 ) @xmath292 else @xmath293 end if + end if + end if + end do + for those rays for which @xmath294 + do @xmath278 + @xmath292 end do + end do + end do + * for the outgoing rays ( @xmath97 ) - forward sweep * + let @xmath295 + @xmath296 do @xmath270 + for @xmath297 ( inner boundary point ) +
@xmath298 else @xmath303 end if + end do + for those rays for which @xmath294 + do @xmath304 + @xmath292 end do + end do + end do + the algorithm given above saves a great deal of computing time by cutting down the number of calls to the formal solver -2 instead of @xmath55 -the first call to store the @xmath35 and @xmath305 at all depth points , and the second call to compute @xmath265 | a robust numerical method called the preconditioned bi - conjugate gradient ( pre - bicg ) method is proposed for the solution of radiative transfer equation in spherical geometry . a variant of this method called stabilized preconditioned bi - conjugate gradient ( pre - bicg - stab ) is also presented .
these are iterative methods based on the construction of a set of bi - orthogonal vectors .
the application of pre - bicg method in some benchmark tests show that the method is quite versatile , and can handle hard problems that may arise in astrophysical radiative transfer theory . |
let @xmath11 be an integer and @xmath4:=\{1,2,\dots , n\}$ ] . by @xmath12}$ ] we denote the family of all subsets of @xmath4 $ ] and by @xmath13}k$ ] the family of all @xmath14-subsets of @xmath4 $ ] .
a family @xmath15}$ ] is an _ antichain _ if @xmath5 for all distinct @xmath16 .
an antichain @xmath17 is called _ flat _ if it is contained in two consecutive levels of @xmath12}$ ] , i.e. if @xmath18}{k}\cup\tbinom{[n]}{k+1}$ ] for some @xmath14 . more generally , for a subset @xmath19 $ ] , we call @xmath17 a _ @xmath1-antichain _ if it contains only @xmath14-sets with @xmath20 , i.e. @xmath21 where @xmath22}{k}$ ] . the _ dual _ @xmath23 of a family of sets @xmath24 is the collection @xmath25 of subsets of @xmath26 $ ] given by @xmath27\ : \ i \in a_j\}$ ] for @xmath28 $ ] .
it is well known that a family of sets @xmath17 is an antichain if and only if its dual @xmath23 is a _ completely separating system _ ( css ) : a css is a collection @xmath29 of blocks of @xmath4 $ ] such that for each @xmath30 $ ] there are blocks @xmath31 with @xmath32 and @xmath33 .
the dual of a flat antichain on @xmath13}{k } \cup \binom{[n]}{k-1}$ ] is a _ fair css _ :
a css in which each point occurs exactly @xmath14 or @xmath34 times .
the consideration of minimum size csss led to a conjecture which subsequently became the _
flat antichain theorem _ , which follows from results of lieby @xcite ( see also @xcite ) and kisvlcsey @xcite .
this theorem greatly reduced the computational search space for csss .
it says that for every antichain @xmath17 there is an equivalent flat antichain @xmath35 , where the equivalence relation is defined by : @xmath35 is equivalent to @xmath17 if and only if @xmath36 and @xmath37 .
griggs et al .
@xcite showed that the flat antichains minimize the _ blym - values _
@xmath38 within their equivalence classes , and more generally , they minimize ( maximize ) @xmath39 for every convex ( concave ) function @xmath40 .
csss have applications in various areas including coding theory , search theory , and topology .
see @xcite and @xcite for applications and further references for csss .
a @xmath1-antichain @xmath17 is called a _ maximal @xmath1-antichain _ if there is no @xmath1-antichain @xmath35 with @xmath41 . for @xmath42 ,
any @xmath43-antichain is also a @xmath1-antichain , and if it is a maximal @xmath1-antichain , then it is also a maximal @xmath43-antichain .
the converse is not true as can be seen by the family @xmath44 for @xmath45 , which is a maximal @xmath46-antichain but not a maximal @xmath47-antichain , since @xmath48 is an antichain properly containing @xmath17 .
this example is illustrated in figure [ fig : maximality ] , showing a graph on 9 vertices whose edge set is the complement of @xmath49 in @xmath50}{2}$ ] , while the set of 4-cliques is precisely @xmath51 .
-antichain which is not maximal as a @xmath47-antichain , and thus also not strongly maximal . ]
we also call a @xmath1-antichain @xmath17 _ strongly maximal _ if it is maximal as an antichain , not just as @xmath1-antichain , i.e. if there is no antichain @xmath35 with @xmath41 .
equivalently , @xmath17 is a strongly maximal @xmath1-antichain if it is also a maximal @xmath52-antichain , where @xmath14 and @xmath53 are the smallest and the largest element of @xmath1 , respectively . in this paper
, we always assume @xmath54 and @xmath55 , and we study the problem of determining the smallest possible cardinality of a maximal @xmath1-antichain .
our approach to the minimum size of a maximal antichain is based on the graph interpretation which was used in @xcite to completely resolve the problem for the case @xmath56 . to prove our asymptotic results we adapt arguments from @xcite where lower bounds for the size of maximal @xmath57-antichains were proved .
the main tool in this part is the graph removal lemma .
let @xmath58 be a graph on @xmath59 vertices and let @xmath60
. then there exists @xmath61 such that every graph @xmath62 on @xmath63 vertices containing at most @xmath64 copies of @xmath58 can be made @xmath58-free by removing @xmath65 edges .
this was proved , using szemerdis regularity lemma , first for @xmath58 being a triangle @xcite and later in general @xcite .
see @xcite for a recent proof avoiding the regularity lemma , and thus improving the bounds on @xmath66 significantly .
the paper is organized as follows . in section [ sec : graph ] the problem is reformulated as a question in extremal graph theory . in section [ sec : construction ] we describe a construction for graphs , which we believe to correspond to optimal antichains . some evidence for this conjecture is provided by exhaustive search results for @xmath10 and small @xmath63 which are presented in section [ sec : small_n ] .
section [ sec : asymptotics ] contains our asymptotic results : if @xmath7 our construction yields the correct leading term , and for @xmath10 we prove a first nontrivial bound . finally , in section
[ sec : problems ] we suggest some open problems for further investigations .
as proposed in @xcite , we associate a graph @xmath67 with an antichain @xmath17 . the vertex set is @xmath68 $ ] and the edge set @xmath69}{2}\setminus\mathcal a_2.\ ] ] we start by making some simple observations .
[ lem : simple_observations ] let @xmath17 be a @xmath1-antichain on @xmath63 points and let @xmath70 be the associated graph . 1 .
every @xmath71 induces a clique in @xmath62 .
if @xmath17 is a maximal @xmath1-antichain and @xmath72 is the vertex set of a @xmath14-clique in @xmath62 for some @xmath20 then there is an @xmath73 with @xmath74 or @xmath75 . in particular , every edge of @xmath62 is contained in a @xmath14-clique for some @xmath76 .
the first statement is just the antichain property applied to @xmath77}{2}\setminus e$ ] and @xmath71 .
the second statement follows from maximality , because if no such @xmath78 exists then we can add @xmath72 to @xmath17 to obtain a strictly larger antichain .
the elements of the antichain are the non - edges of the graph and some cliques whose sizes are in @xmath79 .
we say that a graph is _
@xmath1-saturated _ if every edge is contained in a @xmath14-clique for some @xmath76 .
being @xmath1-saturated is a necessary and sufficient condition for a graph to be @xmath80 for some maximal @xmath1-antichain @xmath17 . in general
, it is not possible to reconstruct the maximal @xmath1-antichain uniquely from the graph : if @xmath81 then a 4-clique can ( among others ) come from a single 4-set or from four 3-sets in @xmath17 . for a @xmath1-saturated graph @xmath70 with @xmath68 $ ] let @xmath82 denote the set of all @xmath1-maximal antichains @xmath17 with @xmath83 . in other words
@xmath82 is the collection of all families @xmath84}$ ] such that 1 .
@xmath85}{2}\setminus e$ ] , 2 . for all @xmath76
every @xmath86 induces a @xmath14-clique in @xmath62 , and 3 . for all @xmath76 and every @xmath14-clique @xmath72 in @xmath62
, there is a @xmath87 and an @xmath88 with @xmath74 or @xmath89 . a greedy choice among all the maximal @xmath1-antichains corresponding to a given @xmath1-saturated graph @xmath62 is based on the intuition to take the sets in @xmath17 as large as possible for the given graph @xmath62 .
more precisely , for @xmath20 , let @xmath90 denote the set of @xmath14-cliques in @xmath62 that are not contained in a @xmath91-clique for any @xmath87 with @xmath92 .
note that for @xmath1-saturated graphs @xmath93 .
we put @xmath94 and define the maximal @xmath1-antichain @xmath95 by @xmath96}{2}\setminus e\right)\cup\mathcal m.\ ] ] it is not necessarily the smallest maximal @xmath1-antichain corresponding to the graph @xmath62 : for @xmath97 and @xmath62 the complete graph on @xmath63 vertices we obtain @xmath98}{4}$ ] , while @xmath99}{3}$ ] is a smaller maximal @xmath1-antichain corresponding to the same graph .
but under the following sparseness condition on @xmath62 the antichain @xmath100 has minimum size .
let @xmath101 be the minimal element of @xmath79 .
we say that a @xmath1-saturated graph is _
@xmath1-sparse _ if for every @xmath102 , there exists a @xmath101-clique @xmath103 such that @xmath72 is the unique element of @xmath104 containing @xmath105 .
the next lemma asserts that in this situation the antichain @xmath100 has minimum size among all @xmath1-antichains with corresponding graph @xmath62 .
[ lem : canonical_ac ] let @xmath70 be a @xmath1-sparse , @xmath1-saturated graph , and let @xmath106 .
then @xmath107 for all maximal @xmath1-antichains @xmath108 .
let @xmath35 be any maximal @xmath1-antichain with @xmath109 .
clearly @xmath110}{2}\setminus e$ ] .
let @xmath111 be the set of @xmath101-cliques in @xmath62 that are contained in exactly one element of @xmath104 . since @xmath62 is @xmath1-sparse there is a mapping @xmath112 with @xmath113 for every @xmath102 . by the definition of @xmath111 ,
the mapping @xmath114 is injective . by the second part of lemma [ lem : simple_observations ] every element of the image of @xmath114
is contained in some member of @xmath35 .
if the sets @xmath115 and @xmath116 are contained in the same @xmath117 , then @xmath118 for @xmath119 because @xmath120 is contained in a unique element of @xmath104 .
this , in turn , implies @xmath121 for the same reason . hence @xmath122 is a lower bound for @xmath123 , and this proves the statement .
the following lemma provides a characterization of the @xmath1-saturated graphs @xmath62 such that @xmath100 is a strongly maximal @xmath1-antichain .
[ lem : strong_maximality ] let @xmath62 be a @xmath1-saturated graph for @xmath124 . then @xmath106 is strongly maximal if and only if the following holds . for every @xmath125 , every @xmath126 and every @xmath127-clique @xmath105 in @xmath62
, we have that either 1 . [ cond_1 ]
@xmath105 contains a @xmath128-clique that is not contained in any @xmath129-clique , or 2 .
[ cond_2 ] @xmath105 is contained in a @xmath129-clique .
suppose @xmath130 satisfies the condition and @xmath17 is not strongly maximal .
then there is a set @xmath131 that can be added such that @xmath132 is still an antichain .
the set @xmath105 must induce a clique in @xmath62 and by lemma [ lem : simple_observations ] this implies @xmath133 and @xmath134 .
hence @xmath135 for some @xmath136 .
but then either of the conditions [ cond_1 ] ( with lemma [ lem : simple_observations ] ) and [ cond_2 ] ( using the definition of @xmath100 ) implies that @xmath105 can not be added without violating the antichain condition .
for the converse , suppose @xmath62 is a @xmath1-saturated graph containing a @xmath127-clique @xmath105 , @xmath137 , such that @xmath105 is not contained in any @xmath129-clique but every @xmath138-subset of @xmath105 is contained in a @xmath129-clique . by construction of @xmath106 , for every @xmath128-subset @xmath139 , the antichain @xmath17 contains a @xmath140-set @xmath141 for some @xmath142 .
but this implies that @xmath132 is still an antichain , hence @xmath17 is not strongly maximal .
to summarize , we have proved the following . 1 . for every maximal @xmath1-antichain @xmath17
there is a unique @xmath1-saturated graph @xmath80 .
2 . minimizing the size of a maximal @xmath1-antichain is equivalent to @xmath143 where the maximum is over the set of all @xmath1-saturated graphs @xmath62 and the minimum is over the set @xmath144 .
if the graph @xmath62 is @xmath1-sparse then the antichain @xmath106 defined by ( [ eq : canonical_ac ] ) achieves the minimum in ( [ eq : reformulation ] ) .
in this section , we describe a family of @xmath1-saturated graphs with many edges and few cliques ( corresponding to small maximal @xmath1-antichains ) .
our construction is inspired by the optimal graphs for @xmath145-antichains in @xcite : we start with a complete bipartite graph with almost equal parts and add a few edges to make the graph @xmath1-saturated without producing too many cliques with sizes in @xmath1 .
let @xmath53 be the maximal element of @xmath1 .
the vertex set is partitioned into three disjoint sets @xmath72 , @xmath105 , @xmath146 with @xmath147 and the edge set is determined as follows . * the induced subgraph on @xmath72 is the disjoint union of @xmath148-cliques . *
the induced subgraph on @xmath105 is the disjoint union of @xmath149-cliques . *
every pair @xmath150 with @xmath151 and @xmath152 is an edge . * the vertices in @xmath146
are isolated . clearly every clique of at least two vertices is contained in an @xmath53-clique , so @xmath104 is the set of all @xmath53-cliques .
any @xmath53-clique contains vertices from @xmath72 and vertices from @xmath105 , and every @xmath101-clique that has nonempty intersection with @xmath72 and with @xmath105 is contained in a unique @xmath53-clique .
thus the graph @xmath62 is @xmath1-sparse and using lemma [ lem : canonical_ac ] the antichain @xmath100 , defined by ( [ eq : canonical_ac ] ) , has minimum size among the antichains in @xmath82 and contains only @xmath153- and @xmath53-sets .
we get @xmath154 and @xmath155 thus @xmath156 the corresponding maximal @xmath1-antichain has size @xmath157 note that in this construction every @xmath127-clique for @xmath158 is contained in an @xmath53-clique , thus by lemma [ lem : strong_maximality ] the antichains @xmath100 are strongly @xmath1-maximal . in the above construction we wasted the vertices in the set @xmath146 , and the solution can be slightly improved by connecting @xmath146 to the rest of the graph .
the best way to do this depends on the remainder of @xmath63 modulo @xmath53 .
here we describe what to do for @xmath159 and @xmath160 ( @xmath161 ) .
: : in this case @xmath163 $ ] , @xmath164 $ ] and @xmath165 and the edge set is @xmath166,\,j\in[2m+1,4m]\right\}\cup\{\{2i-1,2i\}\ : \
i\in[2m]\},\ ] ] and we have @xmath167 @xmath168 .
: : in this case @xmath163 $ ] , @xmath164 $ ] and @xmath169 and the edge set is @xmath170,\,j\in[2m+1,4m+1]\right\}\cup\left\{\{2i-1,2i\}\ : \ i\in[2m]\right\}\\ \cup\{\{4m,4m+1\}\},\end{gathered}\ ] ] and we have @xmath171 @xmath172 .
: : in this case @xmath163 $ ] , @xmath173 $ ] and @xmath165 and the edge set is @xmath166,\,j\in[2m+1,4m+2]\right\}\cup\{\{2i-1,2i\}\ : \
i\in[2m+1]\},\ ] ] and we have @xmath174 @xmath175 . : : in this case @xmath163 $ ] , @xmath173 $ ] and @xmath176 and the edge set is @xmath170\cup\{4m+3\},\,j\in[2m+1,4m+2]\right\}\cup\{\{2i-1,2i\}\ : \ i\in[2m+1]\}\\ \cup\left\{\{2m,4m+3\}\right\},\end{gathered}\ ] ] and we have @xmath177 the construction is illustrated in figure [ fig : examples ] .
-saturated graphs for different values of @xmath178 .
all pairs of vertices on different sides of the dashed line are joined by an edge , but these edges are omitted in the pictures . ] -saturated graphs for different values of @xmath178 .
all pairs of vertices on different sides of the dashed line are joined by an edge , but these edges are omitted in the pictures . ] -saturated graphs for different values of @xmath178 .
all pairs of vertices on different sides of the dashed line are joined by an edge , but these edges are omitted in the pictures . ] -saturated graphs for different values of @xmath178 .
all pairs of vertices on different sides of the dashed line are joined by an edge , but these edges are omitted in the pictures . ]
we conjecture that for large @xmath63 these graphs are best possible .
[ conj : graph_conjecture ] if @xmath63 is large , in any @xmath46-saturated graph @xmath62 on @xmath63 vertices , we have @xmath179 in terms of antichains the equivalent statement is that for any maximal @xmath46-antichain @xmath17 on @xmath63 points , we have @xmath180
in this section we present some computational results for @xmath181 and @xmath182 .
these results , summarized in table [ tab : small_cases ] , were obtained by solving binary programs for maximizing @xmath183 and @xmath184 , respectively , over all pairs @xmath185 of a @xmath1-saturated graph @xmath62 on @xmath63 vertices and an antichain @xmath186 .
the binary programs were solved with the commercial mip solver cplex .
.computational results for small values of @xmath63 .
we report the size of the optimal antichain @xmath17 that was found and its profile vector @xmath187 where @xmath188 is the number of @xmath14-sets in @xmath17 .
the last column contains the size of the antichain corresponding to the graphs constructed in section [ subsec : l_4 ] .
[ cols= " > , > , > , > , > , > " , ] we make the following observations . 1 . for @xmath182 ,
the construction in section [ subsec : l_4 ] yields optimal @xmath47-saturated graphs .
so the bound in conjecture [ conj : graph_conjecture ] might be true for @xmath47-graphs even without the restriction to large @xmath63 .
2 . if @xmath189 , i.e. when the construction in section [ subsec : l_4 ] yields a regular graph , it is also optimal for @xmath46-saturation .
3 . for @xmath97 all optimal examples that have been found do not contain any 3-set .
a particularly interesting case is @xmath190 .
the unique optimal @xmath46-saturated graph has the set of 4-cliques @xmath191 defining a @xmath192-regular graph in which every vertex is contained in exactly two 4-cliques , and every edge is contained in a unique 4-clique .
this can not generalize to large values of @xmath63 , since by the removal lemma in a @xmath46-saturated graph on @xmath63 vertices with the property that every edge is contained in at most a constant number of 4-cliques , the number of edges is @xmath8 .
the removal lemma argument that was used in @xcite to derive asymptotic lower bounds for the size of a maximal @xmath57-antichain in terms of turn numbers can be adapted to establish that for sets @xmath1 containing @xmath145 the construction in section [ subsec : general_construction ] gives the correct leading term .
[ thm : strong_max_asymptotics ] let @xmath1 be a fixed set containing 2 and 3 , and let @xmath53 be its maximal element .
for @xmath1-saturated graphs @xmath62 on @xmath63 vertices and @xmath186 , we have for @xmath193 , @xmath194 let @xmath185 be a pair of a @xmath1-saturated graph @xmath70 and an antichain @xmath186 which maximizes @xmath195 . in particular , @xmath196 , and
since every triangle is contained in an element of some @xmath197 , and every @xmath14-clique contains @xmath198 triangles , the number of triangles in @xmath62 is bounded by @xmath199 by the removal lemma , we can make @xmath62 triangle - free by removing @xmath8 edges , in other words there is a partition @xmath200 such that @xmath201 and @xmath202 is triangle - free , and thus @xmath203 by turn s theorem . for @xmath204 and @xmath20 , let @xmath205 denote the number of elements of @xmath197 containing @xmath206 . using the @xmath1-saturation of @xmath62 and observing that @xmath202 , being triangle - free , contains at most @xmath207 edges of any @xmath14-clique in @xmath62
, we obtain @xmath208 and consequently @xmath209 [ cor : k_contains_3 ] let @xmath1 be a fixed set containing 2 and 3 with maximal element @xmath53 , and let @xmath17 be a maximal @xmath1-antichain of minimum size on @xmath63 points .
then @xmath210 the lower bound follows from theorem [ thm : strong_max_asymptotics ] and the upper bound comes from the construction in section [ subsec : general_construction ] .
[ cor : strongly_maximal_asymptotics ] let @xmath1 be a fixed set containing 2 with maximal element @xmath53 , and let @xmath17 be a strongly maximal @xmath1-antichain of minimum size on @xmath63 points .
then @xmath210 the lower bound follows from corollary [ cor : k_contains_3 ] , because @xmath17 is also a maximal @xmath211-antichain , and the upper bound comes again from the construction in section [ subsec : general_construction ] .
we expect that the statement of theorem [ thm : strong_max_asymptotics ] remains true if the condition @xmath7 is dropped , but in this case we can prove only a weaker bound .
the main difficulty comes from the fact that we are not able to deduce that an optimal graph is essentially triangle - free ( meaning the number of triangles is @xmath212 ) .
this weakens both bounds that are essential in the proof of theorem [ thm : strong_max_asymptotics ] : the upper bound for the number of edges , and the lower bound for the number of cliques in terms of the edge number .
the obvious starting point for deriving bounds is that an optimal graph is almost @xmath213-free for every @xmath20 , which by turn s theorem yields an asymptotic upper bound for the number of edges .
unfortunately , for @xmath9 this bound is far away from the conjectured truth @xmath214 .
we now consider @xmath10 and derive a bound for @xmath215 using lower bounds for the number of triangles in graphs with more than @xmath214 edges .
note that for sets @xmath1 of size @xmath153 , the antichain @xmath100 is the only element of @xmath82 , so maximizing @xmath215 is equivalent to ( [ eq : reformulation ] ) . for notational convenience , in the following we write @xmath216 for @xmath217 .
proceeding as in the proof of theorem [ thm : strong_max_asymptotics ] , we obtain a first bound .
[ lem : first_bound ] let @xmath62 be a @xmath46-saturated graph on @xmath63 vertices . then @xmath218 as @xmath219 , @xmath62 can be made @xmath220-free by removing @xmath8 edges , and we obtain a partition @xmath200 with @xmath221 and @xmath202 @xmath220-free , in particular any 4-clique in @xmath62 contains at most 5 edges from @xmath202 . using the 4-saturation of @xmath62 , we obtain @xmath222 and finally @xmath223 . note that this proof together with turn s theorem already implies @xmath224 for @xmath46-saturated graphs . in order to improve the inequality
, we bound the number of triangles .
first we have a lower bound in terms of the number of edges . in any graph @xmath62 with @xmath63 vertices and @xmath225 edges ,
the number of triangles is at least ( see @xcite ) @xmath226 on the other hand we get an upper bound for the number of triangles in terms of the number of 4-cliques .
[ lem : triples_and_quads ] let @xmath62 be a @xmath46-saturated graph on @xmath63 vertices
. then the number of triangles is at most @xmath227 let @xmath228 denote the set of triangles that are contained in a 4-clique in @xmath62 , and let @xmath229 be the set of all other triangles .
note that @xmath230 , so the number of triangles is dominated by @xmath231 . consider four vertices
@xmath232 inducing a 4-clique in g , and let @xmath233 be another vertex . there is at most one triangle in @xmath229 of the form @xmath234 with @xmath235 . consequently , there are at most @xmath236 triangles in @xmath229 having two vertices in @xmath237 , and by summation over all 4-cliques the claim follows . combining the two bounds for the number of triangles we obtain a second bound for @xmath215 in terms of @xmath216 .
[ lem : second_bound ] let @xmath62 be a @xmath46-saturated graph on @xmath63 vertices
. then @xmath238 now we can combine ( [ eq : first_bound ] ) and ( [ eq : second_bound ] ) to eliminate @xmath216 and obtain an absolute bound for @xmath215 .
[ thm : weak_max_asymptotics ] let @xmath62 be a @xmath46-saturated graph on @xmath63 vertices .
then @xmath239 the right hand sides of ( [ eq : first_bound ] ) and ( [ eq : second_bound ] ) are increasing and decreasing in @xmath216 , respectively . equating them and solving for @xmath216 yields @xmath240 and the result follows .
let @xmath17 be a maximal @xmath46-antichain of minimum size on @xmath63 points .
then @xmath241 the lower bound is a consequence of theorem [ thm : weak_max_asymptotics ] , and the upper bound comes from the construction in section [ subsec : l_4 ] .
an obvious next step is to improve the asymptotic bounds with the aim to replace the coefficient of @xmath242 in theorem [ thm : weak_max_asymptotics ] by @xmath243 , which gives the leading term of the inequality in conjecture [ conj : graph_conjecture ] .
it seems that arguments based on the removal lemma are not sufficient to reach this goal .
one reason for this failure could be that the removal lemma only requires that there are only @xmath244 4-cliques , while in our situation we can use the stronger information that the numbers 4-cliques is @xmath245 .
another problem is to characterize the extremal graphs ( resp .
antichains ) up to isomorphism . for the case
@xmath56 this was done in @xcite , and it would be nice to prove that for @xmath246 and large @xmath63 ( some slight variations of ) the constructions in section [ subsec : l_4 ] yield all extremal graphs . in view of the computational results in table [ tab : small_cases ] one might also study the question if any maximal @xmath1-antichain of minimum size contains only @xmath14-sets and @xmath53-sets , where @xmath14 and @xmath53 are the smallest and the largest element of @xmath1 , respectively .
d. gerbner , b. keszegh , n. lemons , c. palmer , d. plvlgyi , c. palmer , and b. patks . saturating sperner families . , pages 110 , online first : http://dx.doi.org/10.1007/s00373-012-1195-6[`doi:10.1007/s00373-012-1195-6 ` ] , 2012 .
preprint http://arxiv.org/abs/1105.4453v1[`arxiv:1105.4453v1 ` ] .
m. grttmller , i.t .
roberts , and l.j . rylands .
antichains and completely separating systems - a catalogue and applications . ,
2012 . in press , http://dx.doi.org/10.1016/j.dam.2012.04.016[`doi:10.1016/j.dam.2012.04.016 ` ] . | let @xmath0 be a natural number , and let @xmath1 be a set @xmath2:=\{1,2,\dots , n\}$ ] .
we study the problem to find the smallest possible size of a maximal family @xmath3 of subsets of @xmath4 $ ] such that @xmath3 contains only sets whose size is in @xmath1 , and @xmath5 for all @xmath6 , i.e. @xmath3 is an antichain .
we present a general construction of such antichains for sets @xmath1 containing 2 , but not 1 . if @xmath7 our construction asymptotically yields the smallest possible size of such a family , up to an @xmath8 error .
we conjecture our construction to be asymptotically optimal also for @xmath9 , and we prove a weaker bound for the case @xmath10 .
our asymptotic results are straightforward applications of the graph removal lemma to an equivalent reformulation of the problem in extremal graph theory which is interesting in its own right .
* keywords : * extremal set theory , sperner property , maximal antichains , flat antichains |
the origin and evolution of structures in the universe can be suitably understood within the hierarchical scenario . in such schemes ,
present - day systems are the result of the aggregation of several sub - structures , and in particular , satellite galaxies can contribute significantly as building blocks of galaxy formation .
however , present - day systems of satellites have evolved since their formation ; and therefore the observations of such systems can provide invaluable information on the formation and evolution of galaxies .
there are several works that explore the characteristics of systems of satellites .
holmberg ( 1969 ) found that the distribution of galaxy satellites presents a lack of objects within @xmath2 of the primary disk plane ( @xmath3 ) .
he searched for companions of nearby bright disk galaxies ( the primaries ) within 50 kpc of projected separation ( @xmath4 ) .
his sample was composed of 218 satellites including both physical systems and interlopers that contaminated his photometrically selected sample .
he found 45 objects within @xmath3 while 173 had @xmath5 .
if the position angles relative to the primary plane are homogeneously distributed , then the number expected within @xmath3 would be @xmath6 73 . after removing contamination by presumed interlopers ( which are expected to be distributed isotropically ) he concluded that there were no physical objects with @xmath3 .
the results remained controversial since valtonen et al .
( 1978 ) found an excess of compact satellites with small angles relatives to the primary disk .
based on this work , byrd et al .
( 1987 ) argued that the anisotropy of the holmberg sample was mainly due to selection effects for he may have not considered these compact dwarf objects .
the authors proposed a dynamical stripping of the gas content of the satellite as a possible explanation for the excess of such objects near the primary plane .
some years later , the problem was investigated again in a statistical way by zaritsky et al .
( 1997 ) using redshift data .
they selected a sample of 115 satellites of 69 nearby primaries , these satellites have a maximum projected separation @xmath7 kpc .
these authors could not confirm the holmberg effect in the inner 50 kpc since they had few objects at this distance , but they found a similar anisotropy effect with reliable statistical significance at larger separations , @xmath8 kpc .
one simple starting point to approach this problem is to assume an isotropic initial distribution of satellites and that evolution generates the observed anisotropies .
this scenario has been explored by quinn @xmath9 goodman ( 1986 ) , who found that anisotropy in the distribution of satellite positions is not efficiently produced by dynamical evolution , even at small separation , @xmath10 kpc . on the other hand , numerical simulations like those carried out by abadi et al .
( 2002 ) indicate that , in a hierarchical scenario , the formation of disks involve substantial accretion of satellites which strongly influence the angular momentum vector and therefore the primary plane .
thus , observed anisotropies in the distribution of the remaining satellites could be a direct consequence of this process .
in addition to this work , pearrubia et al .
( 2002 ) also based on numerical simulations , established that the time interval before the satellite disruption strongly depends on the dark matter halo shape , the initial orbital eccentricities and inclinations , and on the mass fraction of satellites relative to the primary .
our work aims to search for anisotropies of the angular distribution of satellites relative to the primary plane as well as exploring the possible dependence on satellites properties .
our analysis is based on the study of an unprecedentedly large statistical sample extracted from the 2df galaxy redshift survey . in section 2
we present the data and selection criteria , in section 3 we show the main results , and in section 4 we summarise our conclusions .
the largest complete redshift survey available to date is the 2dfgrs which comprises information of objects on approximately 1500 square degrees of the sky in the north ( ngc ) and south ( sgc ) strips , and in several random fields .
this catalogue has been obtained with the multifibre technology at the anglo - australian observatory , and provides photometric and spectroscopic data for 232155 galaxies with magnitudes @xmath11 .
it contains positions , magnitudes ( in @xmath12 and r band ) , eccentricities , position angles , redshifts ( well determined for 221414 galaxies ) and the spectral classification through the @xmath1 parameter ( which has a high confidence for objects with @xmath13 ) .
the selection criteria adopted in this work are the usual in the literature . from 2dfgrs
we have selected 1567 primary galaxies with @xmath13 restricted to absolute magnitude @xmath14 .
we have imposed an isolation criterion by considering primaries with no companion galaxies within projected distance @xmath15 kpc and relative radial velocity difference @xmath16 km / s , and with absolute magnitude difference @xmath17 .
the value of the hubble constant in this work is assumed to be @xmath18 km / s @xmath19 . after the selection of the primary sample , we searched for satellite galaxies around each primary .
we define as satellites galaxies with @xmath7 kpc and @xmath20 km / s which are at least 2 magnitudes fainter than their primary , @xmath21 .
with all these restriction our final sample consists of 3079 satellites .
a few primaries have too many satellites ( up to @xmath22 objects ) .
these objects are likely to be contaminated by group - like systems with the primary as the brightest group member . in order to exclude these spurious systems
we have restricted the sample to 1498 primaries with @xmath23 out of the 1567 original primaries .
furthermore , we notice that the removed systems show systematically high mean velocity dispersion , close to typical value of groups of galaxies , supporting our conservative cut - off in the number of members per system .
we have also imposed an additional restriction to the eccentricities of the primaries in order to deal with well determined relative angles @xmath24 of satellite position with respect to the primary plane projected into the first quadrant .
thus , we have not considered primaries with @xmath25
the sample selected under the criteria described in the previous section allows us to investigate the satellite position angle distribution relative to the primary plane disk ( @xmath24 ) .
km / s and the dotted line satellites with @xmath0 km / s .
the adopted threshold @xmath26 km / s is approximately the mean value of the satellite relative velocity distribution ( 163.9 km / s).,width=317 ] we have divided the sample into objects with high or low @xmath27 compared to @xmath28km / s ( a value close to the mean @xmath29 km / s ) .
the distribution of satellite position angles for these two subsamples are shown in fig .
[ vel ] . from this figure
one can notice the deficiency of satellites with low @xmath24 values in the subsample of satellites with relative velocity @xmath30 km / s .
on the contrary , the high velocity subsample is roughly isotropic .
we may quantify the departure from isotropy defining the ratio @xmath31 of objects in planar ( @xmath3 ) to polar ( @xmath32 ) angular positions @xmath33 , for isotropy @xmath34 .
the resulting values are : @xmath35 for @xmath0 km / s and @xmath36 for @xmath37 km / s . quoted uncertainties in this work
were calculated based on 20 bootstrap re - samplings of the data .
other useful statistical measures of this anisotropy effect can be obtained by fitting a function to the @xmath24 distribution .
we have adopted a double cosine function @xmath38 , and a linear fit @xmath39 where @xmath40 is the mean number of objects and a and @xmath41 are the anisotropy parameter in each fit . for an isotropic distribution on @xmath24
we expect @xmath42 .
either function , the double cosine and the linear relation , are suitable to fit the data , see for instance fig . [ comp ] .
we notice , however that the double cosine function is less sensitive to noise fluctuations at the extremes due to poor number statistics and therefore might provide more conservative estimate of the deviation from isotropy . for this reason we will mainly refer to the a parameter although both a and
@xmath41 values are listed in the tables corresponding to the different subsamples analysed .
@xmath43 & a & @xmath44 + @xmath45 & @xmath0 km / s & 0.4 & 0.1 & 1276 & 0.12 & 0.04 + @xmath46 & @xmath37 km / s & 0.0 & 0.1 & 916 & 0.01 & 0.04 + + @xmath47 & @xmath0 km / s , + & @xmath48 kpc @xmath49 & 0.2 & 0.1 & 713 & 0.11 & 0.05 + @xmath50 & @xmath0 km / s , + & @xmath8 kpc @xmath49 & 0.2 & 0.1 & 691 & 0.12 & 0.06 + [ tvel ] in table [ tvel ] we list the number of satellites in each subsample ( @xmath0 km / s and @xmath37 km / s ) and the corresponding @xmath41 and a values .
the a parameter for subsample @xmath45 departs from isotropy by more than @xmath51 which contrasts with the behaviour of subsample @xmath46 where @xmath52 .
this reflects the trends shown in fig .
[ vel ] , and is consistent with our previous estimate of @xmath31 .
the fact that the anisotropy signal is not significant in the high relative velocity subsample ( @xmath37 km / s ) suggests a higher contamination due to interlopers for these objects ( see for instance zaritsky et al .
1993 ) .
the number of objects per bin is @xmath53 , where @xmath54 and @xmath55 are the number of physical satellites and interlopers as function of angle , respectively .
false satellites should have a homogeneous distribution of @xmath24 angles , so we take @xmath56 . on the other hand ,
the distribution of real satellites respect to the primary plane may present anisotropies .
the observed anisotropy parameter @xmath57 is related to the real anisotropy parameter @xmath58 by : @xmath59 .
thus , even a modest observed anisotropy @xmath60 , can imply a substantially large real anisotropy . as suggested by zaritsky ( 1992 ) , the relative fraction @xmath61 of objects with @xmath62 for a given sample can provide a useful measure of the degree of contamination given the different volumes involved .
if a sample is seriously contaminated by interlopers , we should obtain @xmath63 .
therefore , we calculated @xmath61 for two subsamples , satellites with @xmath0 km / s and satellites with @xmath37 km / s .
for these two subsamples @xmath61 values are @xmath64 and @xmath65 respectively , so that from this particular test we can not argue for substantial contamination in the high velocity subsample .
thus the different behaviour of high and low relative velocity satellites could be a real effect .
we show two possible fit to the data . in solid curve
is the linear relation and in long - dashed ones is the double cosine function .
, width=317 ] in his pioneer work , holmberg found an excess of polar satellites within 50 kpc from the primary ; nevertheless , zaritsky et al . (
1996 ) could not reproduce this result , but they found such an excess at larger separations ( 250 kpc@xmath66500 kpc ) from the primaries . we have searched for possible dependences of the detected anisotropies on projected separations to the primaries .
the resulting anisotropy parameter a are displayed in table [ tvel ] , where it can be appreciated the lack of strong dependence on projected separation . from now on ,
we restrict our sample of satellites to those objects with @xmath0 km / s .
distribution , in the bottom is represented the spectral type @xmath1 of the satellites .
the error bars correspond to poisson uncertainties .
, width=317 ] in order to investigate if there are any preferred satellite characteristics which show a more significant anisotropy signal , we have analysed the distributions of absolute magnitude and spectral type of satellites in polar ( @xmath32 ) and planar ( @xmath3 ) positions .
the objects with @xmath3 and @xmath32 are 385 and 477 respectively .
we show in fig [ fig2 ] @xmath67 and @xmath68 , the corresponding distributions of absolute magnitude @xmath69 and spectral type parameter @xmath1 for these two cases .
the error bars correspond to poisson uncertainties .
it can be appreciated in these figures that the excess of objects with @xmath32 , is distributed quite uniformly along the explored range of luminosities and spectral types . in order to quantify the anisotropy for these subsamples we show in table [ tsat ] the corresponding values of @xmath41 and a parameters .
& @xmath41&@xmath43 & a & @xmath44 + s1 & @xmath70 & 633 & 0.3 & 0.1 & 0.15 & 0.04 + s2 & @xmath71 & 643 & 0.1 & 0.1 & 0.09 & 0.04 + + s3 & @xmath72 & 580 & 0.3 & 0.1 & 0.16 & 0.04 + s4 & @xmath73 & 643 & 0.1 & 0.1 & 0.08 & 0.04 + + s5 & @xmath74 & 644 & 0.3 & 0.09 & 0.15 & 0.05 + s6 & @xmath75 & 632 & 0.2 & 0.09 & 0.09 & 0.06 + [ tsat ] the adopted thresholds in absolute magnitude @xmath69 , spectral type @xmath1 and colour index @xmath76 were chosen in order to divide the samples into approximately equal number subsets . by inspection to table 2
it can be seen that the anisotropy of sample s2 , although less statistically significant that in sample s1 , still shows an excess of polar satellites at a @xmath77 level .
thus , this result does not support the hypothesis of byrd et al .
( 1987 ) where the lack of satellites with @xmath3 in holmberg s data would correspond to a combination of dynamical stripping of gas and selection effects , where faint compact satellites should be found preferentially near the primary plane .
the results for subsample s3 indicate that the satellites with low star formation activity present the strongest anisotropy signal compared to those of high star formation activity ( subsample s4 ) .
subsamples s5 and s6 indicate that the satellites characterized by low colour index values show a stronger excess near the primary poles .
this can be appreciated from fig [ fig2]@xmath68 , where is evident that the intermediate spectral type satellites ( @xmath78 ) show a strong deficiency at @xmath3 .
we have also explored for possible dependencies of the anisotropy on properties of the primaries , namely luminosity , spectral type and colour index . in table
[ tpri ] we list n , @xmath41 , @xmath43 , a and @xmath44 for subsamples selected according to primary properties ( regardless of satellites characteristics ) .
@xmath43 & a & @xmath44 + p1 & @xmath79 & 630 & 0.2 & 0.1 & 0.10 & 0.07 + p2 & @xmath80 & 643 & 0.3 & 0.1 & 0.14 & 0.05 + + p3 & @xmath81 & 686 & 0.4 & 0.1 & 0.18 & 0.03 + p4 & @xmath82 & 581 & 0.1 & 0.1 & 0.03 & 0.06 + + p5 & @xmath83 & 645 & 0.1 & 0.1 & 0.08 & 0.05 + p6 & @xmath84 & 631 & 0.3 & 0.1 & 0.16 & 0.05 + [ tpri ] @xmath45 , @xmath68 s3 , @xmath85 p2 and @xmath86 p3.,width=317 ] different ranges of absolute magnitudes of the primaries show different @xmath24 distributions ( see first two lines in table [ tpri ] ) . for faint primaries ( p2 )
the angular positions of satellites are well reproduced by our double cosine fit , where the anisotropy parameter differs from zero by @xmath87 .
however , for bright primaries ( p1 ) this effect is not statistical significant . in a recent paper ,
prada et al . (
2003 ) have shown that the luminosity and halo mass of a galaxy are closely related .
this results suggest that the anisotropy of the satellite distribution would not have a tidal origin , since in this case we would expect satellites of bright primaries to show the strongest anisotropy , contrary to our findings .
on the other hand , we detect a tight dependence of the satellite anisotropy distribution on the spectral type of the primary . in table
[ tpri ] it can be seen that satellites of passive star forming primaries ( p3 ) show a departure from isotropy of at least @xmath88 .
a different behavior is found for the sample of primaries with significant star formation activity ( p4 ) which show a remarkable isotropic distribution of satellites . regarding the colour indexes of primaries we find a more significant effect for primaries with larger values of @xmath76 colour index ( sample p6 ) . in fig .
[ fig3 ] we show four different subsamples with the high anisotropy signal : @xmath67 satellites with @xmath0 km / s ( @xmath45 ) , @xmath68 satellites with @xmath72 ( s3 ) , @xmath85 and @xmath86 primaries with absolute magnitude @xmath89 ( p2 ) and @xmath81 ( p3 ) respectively .
a useful test to check the real presence of the signal in our analysis is to restrict to primaries with high eccentricity values ( @xmath90 ) .
we have performed this test for subsample p3 ( primaries with low star formation activity ) which show a large anisotropy effect . in table
[ tem ] we show the results obtained for this sample for different threshold values of @xmath90 .
it can be seen that the strongest signal corresponds to primaries which are nearly edge - on ( @xmath91 ) which gives substantial confidence to our findings .
.the anisotropy parameters @xmath41 and a for samples of passively star forming primaries ( @xmath81 ) with different eccentricities . [ cols="^,^,^,^,^,^,^ " , ] [ tem ]
we have carried out an analysis of the anisotropy of satellite angular positions with respect to the primary plane .
the sample was selected in redshift space from the 2dfgrs and is large enough for studying possible dependencies of anisotropies on both primary and satellite properties .
we confirm the existence of the effect first claimed by holmberg , that is , the preference of the satellites to lie near the pole of the primary .
we note however that this effect is present up to relative projected separations @xmath7 kpc .
we use an anisotropy parameter a suitable to measure the departure from isotropy in the different subsamples analysed . nevertheless , this parameter indicates a lower limit to the real anisotropy , given the contamination by false satellites expected to be uniformly distributed with respect to the primary plane .
the deviation from isotropy is detected at least at @xmath51 level for objects with low values of relative velocity @xmath0 km / s while the high velocity subsample is consistent with isotropy .
although in principle this effect could be related to different levels of contamination by interlopers , zaritsky 1992 test provides comparable values of contamination in both high and low relative velocity subsamples . by inspection to the luminosity of the primaries
, we conclude that the anisotropy of the distribution of satellites around faint primaries shows a high statistical significance ( @xmath87 ) .
the most important signal was detected in the sample of satellites corresponding to passive star forming primaries , where the distribution of @xmath24 angles departs from isotropy at the @xmath92 level ; while the opposite behaviour is present in primaries with @xmath93 , for which the a parameter results consistent with zero .
moreover , we also notice that the signal is even higher when we raise the eccentricity threshold of the primaries .
we also have considered the different spectral types of the satellites , finding the strongest anisotropy for satellites with @xmath94 ( objects with little present day star formation activity ) .
we notice that primaries with a low value of @xmath1 in our sample also have satellites with poor star formation activity .
the isolation criteria imposed to our sample lowers the probability of including a significant number of elliptical galaxies .
thus , most primaries in our sample are expected to be of late type morphology . due to the morphology density relation
we expect only 10@xmath95 of our isolated sample of primaries to be ellipticals ( whitmore et al .
1993 ) .
we could in principle interpret the holmberg effect as being mainly caused by evolutionary processes .
systems dominated by old stellar population objects have formed early , so the physical processes that may have removed satellites near the primary plane have been present for a large time . by contrast , in younger primary and satellite systems , we would expect a larger fraction of objects that have not been accreted yet on to the primary so that they would show an isotropic distribution of satellites .
however , as discussed by quinn @xmath9 goodman ( 1986 ) , it may be difficult to produce anisotropies in the distribution of satellites from pure dynamical evolution , specially at large scale @xmath7 kpc .
more likely , we suggest that according to abadi et al .
( 2002 ) results , such anisotropies could be the remnants of the formation of the disks of the primaries , where substantial orbital momentum is transfered into primary rotation during the early stages of galaxy formation .
thus , the anisotropy effects we detect at large projected separations , could be the relics of such mechanism .
we also argue that the significant dependence of the anisotropy effect on the relative velocity between the satellite and the primary could be an important point of disagreement between the results of previous works .
we thank the referee for helpful comments and suggestions which greatly improved the previous version of this paper .
this work was partially supported by the consejo nacional de investigaciones cientficas y tcnicas , agencia de promocin de ciencia y tecnologa , fundacin antorchas and secretara de ciencia y tcnica de la universidad nacional de crdoba . | we have analysed a sample of satellite and primary galaxies in the 2df galaxy redshift survey . in our study
we find a strong statistical evidence of the holmberg effect ( that is a tendency for satellites to avoid regions along the line defined by the primary plane ) within 500 kpc of projected distance to the primary .
this effect is present only when we restrict to objects with radial velocity relative to the primary @xmath0 km / s which correspond approximately to the mean of the distribution .
we explore the dependence of this anisotropy on spectral type ( @xmath1 ) , colours , and luminosities of both primaries and satellites , finding that objects with a low present - day star formation rate present the most significant effect .
cosmology : theory - galaxies : formation - galaxies : evolution . |
amorphous materials have a great potential for application in technological devices , but their uses are limited due to several factors .
one of them is the difficulty to obtain information about their atomic structures .
since most chemical and physical properties of these materials are defined by the first neighbors , knowledge of their structures becomes necessary .
due to the lack of translational symmetry , the determination of their atomic structures involves the combination of different diffraction and spectroscopic techniques as well as simulations and modeling .
the structure of an amorphous binary alloy is described by three pair correlation functions @xmath6 , which are the fourier transformation of the three partial structure factors @xmath7 .
the total structure factor @xmath1 , which can be derived from scattering measurements , is a weighted sum of these three @xmath7 factors @xcite .
thus , to determine the three partial @xmath7 at least three independent @xmath1 factors are needed .
usually , the following methods have been used to obtain these @xmath1 factors : 1 .
isomorphous substitution : in this method one component of the amorphous alloy is substituted by a ( expected ) chemically similar one .
for example , in amorphous ni - zr alloys the zr atoms are replaced by hf atoms or the ni atoms are partially substituted by co or fe atoms @xcite .
this method is not very accurate since one can not be sure of the chemical similarity of the two elements involved .
isotope substitution : in this method one component of the amorphous alloy is substituted by an isotope or by an isotopic mixture and the scattering measurements are carried out using neutron sources .
for example , in amorphous ni - zr alloys the ni atoms are replaced by a ni isotope @xcite .
the advantages of this method are limited by the availability of adequate isotopes . with the development of the synchrotron radiation sources , anomalous wide angle x - ray scattering ( awaxs ) and differential anomalous scattering ( das ) techniques became available for structural study of multicomponent disordered materials .
awaxs utilizes an incident radiation that is tuned close to an atomic absorption edge so that it interacts resonantly with the electrons of that particular atom .
the atomic scattering factor of each chemical component can therefore be varied individually and the chemical environment about each component in the material can be investigated .
thus , in the case of an amorphous binary alloy , by using only one sample three independent @xmath1 factors can be obtained .
however , the matrix formed by the weights of the three @xmath1 factors is ill - conditioned , becoming difficult the determination of the three @xmath7 factors .
fuoss@xcite tried to overcome this difficulty by implementing the das approach , which was proposed by schevick @xcite .
the das approach consists in taking the difference between the scattering patterns measured at two incident photon energies just below the edge of a particular atom , so that all correlation not involving this atom subtract out since only the atomic scattering factor of this atom changes appreciably .
later , de lima _ et al_. @xcite , following a suggestion made by munro @xcite , combined the differential scattering factors @xmath2 and the @xmath1 factors .
they observed that this combination reduces the conditioning number of the matrix formed by the weights of these factors , allowing more stable values of @xmath7 to be obtained . the reverse monte carlo ( rmc ) simulation technique @xcite has been successfully used for structural modeling of amorphous structures .
@xmath1 factors or their fourier transformations , the total reduced radial distribution functions @xmath8 , can be used as input data .
applications of this technique to polymeric @xcite , crystalline @xcite and amorphous @xcite materials are described in the literature . in this paper , we report the determination of the three @xmath7 factors obtained for amorphous nizr@xmath0 alloy by two independent ways : by making a combination of awaxs and das techniques and also by combining awaxs and rmc simulations techniques .
according to faber and ziman @xcite , the @xmath1 factor is obtained from the scattered intensity per atom @xmath9 as follows : @xmath10}{\langle f(k , e)\rangle^2 } \nonumber\,,\\ & = & \sum_{i=1}^n{\sum_{j=1}^n{w_{ij}(k , e ) { \cal s}_{ij}(k ) } } \ , , \label{eqstructurefactor}\end{aligned}\ ] ] where @xmath11 is the transferred momentum , @xmath12 is the incident photon energy , @xmath7 are the partial structure factors and @xmath13 are given by @xmath14 and @xmath15 ^ 2 \nonumber\,.\end{aligned}\ ] ] here , @xmath16 is the concentration and @xmath17 is the atomic scattering factor of the component of type @xmath18 and @xmath19 and @xmath20 are the anomalous dispersion terms . the total reduced distribution function @xmath8 is related to the @xmath1 factor through the fourier transformation @xmath21 \sin ( kr)\ , dk}\ , , % \label{eqg } \nonumber\ ] ] while the partial reduced distribution functions @xmath6 are related to the @xmath7 by means of the fourier transformation @xmath22 \sin ( kr)\ , dk}\ , .
% \label{eqpartialg } \nonumber\ ] ] from @xmath6 the partial radial distribution functions @xmath23 can be calculated by @xmath24 interatomic distances are obtained from the maxima of the @xmath6 functions and coordination numbers are calculated by integrating the peaks of the @xmath23 functions . by using the faber and ziman formalism @xcite , the @xmath2 factor around the component @xmath18
is obtained as follows @xmath25 - \bigl[ls(k , e_m ) - ls(k , e_n)\bigr]}{\langle f(k , e_m)\rangle^2 - \langle f(k , e_n)\rangle^2 } \nonumber\,,\\ & = & \sum_j{u_{ij}(k , e_m , e_n ) { \cal s}_{ij}(k)}\ , , \label{eqdsfpartial}\end{aligned}\ ] ] where @xmath26 is the laue scattering term and @xmath27 } { \langle f(k , e_m)\rangle^2 -\langle f(k , e_n)\rangle^2 } \ , .
\label{eqwdsf}\ ] ] for @xmath28 the weight @xmath29 must be multiplied by 2 .
the differential distribution function @xmath30 can be calculated by @xmath31 \sin ( kr)\ , dk } \ , .
% \label{tddf } \nonumber\ ] ] for a binary amorphous alloy , as compared to the total radial distribution function @xmath32 , which is a sum of three @xmath23 functions , this function is the sum of only two @xmath23 functions and is sensitive only to the environment around the atoms of type @xmath18 . in this way it is similar to exafs , though there are important differences due to the different sections of @xmath11-space measured @xcite . in order to determine the three @xmath7 factors using the eight @xmath1 and four @xmath2 factors ,
a least - squares method was used .
first of all , we define the function @xmath33 ^ 2 } \nonumber\,.\ ] ] for @xmath34 , @xmath35 indicates the @xmath36 factors and for @xmath37 it stands for the @xmath38 factors , as defined by eqs . [ eqstructurefactor ] and [ eqdsfpartial ] . in the same way , for @xmath34 , @xmath39 indicates the @xmath40 weights and for @xmath37 it stands for the @xmath41 weights , as seen in eqs .
[ eqw ] and [ eqwdsf ] .
@xmath42 is a weight which allows us to vary the contribution of any @xmath35 factor . here , @xmath43 , @xmath44 , and @xmath45 .
by taking the derivatives of the @xmath46 function with respect to the @xmath7 factors and putting them equal zero a linear system formed by three equations can be found .
these three linear equations can be written as a matrix equation through @xmath47 and by inverting this matrix equation the @xmath7 can be found .
there are several methods to invert a matrix , and we have adopted the singular value decomposition ( svd ) method @xcite . the basic idea and the algorithm of the standard rmc method are described elsewhere@xcite and its application to different materials is reported in the literature @xcite . in the rmc procedure , a three - dimensional arrangement of atoms with the same density and chemical composition of the alloy is placed into a cell ( usually cubic ) with periodic boundary conditions and the @xmath48 functions corresponding to it are directly calculated through @xmath49 where @xmath50 is the number of atoms at a distance between @xmath51 and @xmath52 from the central atom , averaged over all atoms . by allowing the atoms to move ( one at each time ) inside the cell , the @xmath53 functions can be changed and , as a consequence , the @xmath7 and @xmath54 factors are changed .
thus , the @xmath55 factor is compared to the @xmath1 factor in order to minimize the differences between them .
the function to be minimized is @xmath56 ^ 2}\ , , \nonumber%\label{eqpsi}\ ] ] where the sum is over @xmath57 experimental points and @xmath58 is related to the experimental error in @xmath1 .
if the movement decreases @xmath59 , it is always accepted .
if it increases @xmath59 , it is accepted with a probability given by @xmath60 ; otherwise it is rejected . as
this process is iterated @xmath59 decreases until it reaches an equilibrium value .
thus , the atomic configuration corresponding to equilibrium should be consistent with the experimental total structure factor within the experimental error . by using the @xmath53 functions and the @xmath61 factors
the coordination numbers and interatomic distances can be calculated .
in addition , the bond - angle distributions can also be determined .
amorphous nizr@xmath0 thin films were prepared on a nacl cooled substrate by the sputtering triode technique at the argonne national laboratory . in order to eliminate the substrate a metallic holder with a central rectangular hole
was glued on the thin film , and the set was immersed in water to dissolve the nacl substrate .
the thickness and density of the thin films were 2 @xmath62 m and 7.32 g/@xmath63 , respectively .
the awaxs scattering experiments were performed some years ago on the d4 beamline at lure ( orsay - france ) using the dci synchrotron radiation source .
all measurements were performed using a two - circles diffractometer in the vertical plane described elsewhere@xcite equipped with a two crystal si ( 220 ) or ge ( 400 ) monochromator and a si : li energy sensitive detector which is sufficient to discriminate the large k@xmath64 resonant raman / fluorescence signal when incident photon energy is tuned close to the ni and zr k edges but insufficient to distinguish the small elastic , k@xmath65 and compton signals .
the si ( 220 ) and ge ( 400 ) crystal monochromators were used to collect data around the ni and zr edges , respectively .
the sample was sealed , under vacuum , into a cell containing a large kapton window fixed around the diffractometer horizontal rotation axis to reduce the air scattering . in order to take into account the time - decrease of the beam , the incident intensity @xmath66 was measured by placing a nai photomultiplier tube between the monochromator and the diffractometer .
absorption measurements were performed on the ni and zr k edges for energy calibration . for each incident
photon energy listed in table [ tabf ] , the energy windows for k@xmath64 , resonant raman / fluorescence , k@xmath65 , compton , and elastic signals were defined by measuring the scattered intensities at two different angular positions : the beginning of the windows was defined at the lowest scattering angle and their ends at the highest one .
scattering patterns were collected using the symmetrical reflection geometry . to improve the quality of the data ,
at least four scattering measurements were performed at the incident photon energies shown in table [ tabf ] .
.[tabf ] @xmath67 and @xmath68 values used here . [ cols="^,^,^,^,^ " , ] in order to interpret the scattering data correctly , the real and the imaginary parts @xmath69 and @xmath70 of the atomic scattering factor were determined accurately following a procedure described by dreier _
et al_.@xcite x - ray absorption coefficients were measured over a broad energy range near the ni and zr k edges on the sample and @xmath70 was calculated using the optical theorem .
the absorption measurements were performed on the xas beamline at lure , in transmission mode , and two standard ionization chambers were used as the detecting system . outside the region of measurement ,
theoretical values @xcite of @xmath70 were used to extend the experimental data set over a larger energy range and @xmath69 was calculated using the kramers - kronig relation . for the measurements away from the k edges ,
the @xmath69 and @xmath70 values were taken from a table compiled by sasaki @xcite .
the obtained values for the incident photon energies 8330 ev and 17987 ev are listed in table [ tabf ] together with sasaki values .
the atomic scattering factor away from the k edge , @xmath71 , of neutral ni and zr atoms were calculating according to the analytic function given by cromer and mann @xcite .
each scattering pattern was treated separately . due to the diffractometer characteristics
, the polarization correction was disregarded .
the measured scattered intensities away from the ni and zr k edges were corrected for detector non - linearities and reabsorption effects being then put on a per atom scale and the compton scattering eliminated . for those measured close to the ni and zr edges , besides the corrections above , the contribution of the k@xmath65 fluorescence to the elastic scattering was eliminated .
this contribution was evaluated , after corrections for reabsorption effects , by using the k@xmath65/k@xmath64 ratio , which was measured at an incident photon energy above the edge , because they were completely separated .
two method were tested to put the scattered intensities on a per atom scale : that described by krogh - moe - norman @xcite and the one based on the high - angle procedure @xcite . for the elastic scattered intensity around the zr edge ,
for which the @xmath11 data range is larger ( @xmath72 @xmath73 ) , both method gave the same result .
the @xmath11 data range was reduced and the krogh - moe - norman method was then applied .
the normalized signal was identical to the one obtained previously .
so , we have kept the krogh - moe - norman method to put the scattered intensities on a per atom scale including those around the ni edge , for which the @xmath11 data range is smaller ( @xmath74 @xmath73 ) .
the good quality of the total structure factors shown in fig .
[ fig1 ] and [ fig2 ] corroborates the accuracy of the procedure used .
the compton scattering contribution was calculated according to the analytic approximation given by plinkas @xcite .
figures [ fig1 ] and [ fig2 ] show the @xmath1 factors for the incident photon energies described in table [ tabf ] .
the @xmath1 factors for the energies greater than 16101 ev ( see fig .
[ fig2 ] ) show a pre - peak located at about @xmath75 @xmath73 , which reaches the maximum intensity close to the zr edge .
it is absent in the @xmath1 factors for the smallest energies ( see fig .
[ fig1 ] ) . since @xmath1 is a weighted sum of @xmath7 , we calculated the weights @xmath13 for the @xmath3 , @xmath4 and @xmath5 factors .
their contributions for the @xmath1 factor at 8330 ev are about 4% , 32% and 64% while for the @xmath1 factor at 17987 ev they are about 10% , 43% and 47% , respectively .
thus , the determination of ni - ni first neighbors is more difficult than for the ni - zr and zr - zr first neighbors .
@xmath1 factors for energies @xmath76 ev ( a ) , @xmath77 ev ( b ) , @xmath78 ev ( c ) and @xmath79 ev ( d ) . ]
@xmath1 factors for energies @xmath80 ev ( e ) , @xmath81 ev ( f ) , @xmath82 ev ( g ) and @xmath83 ev ( h ) . ]
we have used the scattered intensities on a per atom scale obtained at the incident photon energy pairs ( 8210 , 8330 ev ) , ( 8330 , 9659 ev ) , ( 16101 , 17987 ev ) and ( 17987 , 19200 ev ) to calculate four @xmath2 factors around the ni and zr atoms .
figures [ fig3 ] and [ fig4 ] show them and their @xmath30 functions . figure [ fig3 ] shows that the chemical environment around ni and zr atoms is very different .
for instance , it is seen a broad shoulder on the right side of the first peak in the @xmath84 factors which is absent in the @xmath85 factors .
the @xmath86 functions show the first coordination shell split into two subshells , located at about @xmath87 and 3.23 , while the @xmath88 functions display the first coordination shell located at a mean distance @xmath89 .
the contribution of the ni - ni and ni - zr pairs to them is not resolved .
this can be related to the small @xmath11 data range of the measured scattered intensities .
the weight @xmath90 to the @xmath84 factor is about 25% , and it is almost four times larger than the weight @xmath91 for the @xmath1 factors , allowing a more realistic determination of the number of ni - ni first neighbors . @xmath92 factors . @xmath84 for energies ( @xmath93 ) ( a ) and ( @xmath94 ) ( b ) . @xmath85 for energies ( @xmath95 ) ( c ) and ( @xmath96 ) ( d ) . ]
@xmath97 functions .
@xmath88 for energies ( @xmath98 ) ( a ) and ( @xmath99 ) ( b ) .
@xmath86 for energies ( @xmath95 ) ( c ) and ( @xmath96 ) ( d ) .
] by considering the atomic radius of ni and zr atoms , the subshells located at the distances @xmath100 and 3.23 in the @xmath86 functions were directly attributed to zr - ni and zr - zr correlations .
this fact allowed us to deconvolute their first shells and , consequently , also the first shells of the @xmath88 functions , by assuming a gaussian shape for them . to do this
, we explicitly took into account the @xmath11 dependence by fourier transforming into the @xmath11 space the gaussian distributions multiplied by the weights @xmath29 and back fourier transforming over the same @xmath11 range used to calculate the @xmath30 . the structural parameters extracted from the deconvolution
are listed in table [ tab2 ] .
ccccccccccc + + & & & & + bond type & & & & + @xmath101 & & & & + @xmath51 ( ) & & & & + + + & & & & + bond type & & & & + @xmath101 & & & & + @xmath51 ( ) & & & & + + + & & & & + bond type & & & & + @xmath101 & & 6.9 & 2.3 & 3.5 & 1.1 & + @xmath51 ( ) & & 2.73 & 3.62 & 2.73 & 3.62 & + + + & & & & + bond type & & & & + @xmath101 & & & & 1.0 & 2.0 & 4.0 & 4.0 + @xmath51 ( ) & & & & 2.82 & 3.17 & 3.30 & 3.47 + + + & & & & + bond type & & & & + @xmath101 & & & & + @xmath51 ( ) & & & & + + + & & & & + bond type & & & & + @xmath101 & & & & + @xmath51 ( ) & & & & by considering the eight @xmath1 and the four @xmath2 factors , a reduction of a factor of 15 ( at @xmath102 @xmath73 ) and 7 ( at @xmath103 @xmath73 ) in the conditioning number of the matrix @xmath104 was reached when compared to the use of the eight @xmath1 factors only . to obtain the three @xmath7 factors ,
the @xmath1 and @xmath2 factors were restricted to @xmath105 @xmath73 .
the @xmath3 , @xmath4 and @xmath5 factors ( thin lines ) obtained are shown in fig .
[ fig5 ] . from this figure ,
the @xmath4 and @xmath5 factors seem to be of good quality , specially between @xmath106 @xmath73 and @xmath107 @xmath73 , where the conditioning number presents the largest changes . on the other hand
, the @xmath3 factor displays a minimum apparently without physical meaning in this region .
the @xmath4 factor shows a minimum at @xmath108 @xmath73 , which was already observed in the @xmath84 factors , while the @xmath5 factor displays a shoulder at this @xmath11 value .
although the @xmath3 factor is very noisy , it is possible to see a weak pre - peak at about @xmath106 @xmath73 which was already observed in the @xmath1 factors around the zr edge and also some weak halos beyond @xmath109 @xmath73 .
@xmath3 , @xmath4 and @xmath5 factors obtained from the matrix inversion ( thin lines ) and rmc simulations ( thick lines ) techniques . ]
figure [ fig6 ] shows the @xmath23 functions ( thin lines ) obtained by fourier transformation of the @xmath7 factors . from this picture
one can see that the rdf@xmath110 function shows a very weak first coordination shell , located at about @xmath111 , while the second and third ones , located at about @xmath112 and 6.7 , respectively , are very intense .
the rdf@xmath113 function displays a well isolated first coordination shell , located at about @xmath114 ; a small shoulder at about @xmath115 , and well defined and intense second and third coordination shells , located at about @xmath116 and 7.9 .
the rdf@xmath117 function shows the first coordination shell well isolated , located at about @xmath118 ; the second one splits into two subshells , located at about @xmath119 and 6.35 , which are very weak . the interatomic distances and coordination numbers for the first neighbors obtained from these functions are listed in table [ tab2 ] .
rdf@xmath110 , rdf@xmath113 and rdf@xmath117 functions obtained from the matrix inversion ( thin lines ) and rmc simulations ( thick lines ) techniques . ] in order to perform the simulations we have considered a cubic cell with 1800 atoms ( 600 ni and 1200 zr ) , @xmath120 , and a mean atomic number density @xmath121 atoms / @xmath122 .
the minimum distances between atoms were fixed at the beginning of the simulations at @xmath123 , @xmath124 and @xmath125 . to make the simulations we used the rmc programs available at the internet @xcite .
the @xmath1 factors obtained at the photon energies 8330 ev , 17398 ev and 19200 ev were used as input data .
these @xmath1 factors were cut at the same @xmath11 value . the @xmath1 ( solid lines ) and @xmath55 factors ( square lines )
are shown in fig .
[ fig7 ] and they show a good agreement . @xmath1 ( full lines ) and @xmath55 ( squares ) factors for the amorphous nizr@xmath126 alloy . ] the @xmath127 factors ( thick lines ) are also shown in fig .
[ fig5 ] , together with those obtained from the matrix inversion method .
the agreement among them is very good , mainly for the @xmath5 factor .
for instance , the shoulder previously observed in @xmath5 factor and the peaks located around @xmath128 @xmath73 , 4.3 @xmath73 and 6.2 @xmath73 are also seen in the @xmath129 factor ; the minimum at @xmath130 @xmath73 and peaks around @xmath131 @xmath73 , 5.2 @xmath73 and 7.6 @xmath73 observed in the @xmath4 are also seen in the the @xmath132 factor ; and it is interesting to note that the minimum previously observed in the @xmath3 factor is now replaced by a well defined maximum in the @xmath132 factor .
other peaks observed before in the @xmath3 factor are also seen in the @xmath133 factor .
the rdf@xmath134 ( thick lines ) functions are also shown in fig .
the interatomic distances and coordination numbers for the first neighbors extracted from the @xmath53 and rdf@xmath134 functions are also listed in table [ tab2 ] .
it is interesting to note that although we have used eight independent @xmath1 and four @xmath2 factors to obtain the three @xmath7 factors by the matrix inversion method , the @xmath3 factor shows a minimum between @xmath106 @xmath73 and 3.5 @xmath73 which have no physical meaning . on the other hand , using only three @xmath1 factors as input data in the rmc method we have obtained excellent results , even for the @xmath3 factor . in a previous study , lee _
et al_. @xcite and mizoguchi _
et al_. @xcite reported the @xmath7 factors obtained for the amorphous ni@xmath135zr@xmath136 and ni@xmath137zr@xmath138 alloys , respectively .
et al _ used alloys containing mixtures of co and fe in small quantities while mizoguchi _
et al _ used the isotopic substitution method .
a comparison among the @xmath127 , @xmath53 and rdf@xmath134 functions obtained in this work , by using only one sample , with those reported by them shows that they are in a good agreement . by defining the partial bond - angle distribution functions @xmath139 where @xmath140 is the atom in the corner
, we calculated the angular distribution of the bonds between first neighbor atoms .
the six @xmath139 functions found are shown in fig .
this kind of information can not be obtained by the matrix inversion method because the partial distribution functions give only a one - dimensional description of the atomic arrangement .
@xmath139 functions obtained by the rmc simulations . ] hausleitner and hafner @xcite investigated several amorphous alloys formed by transition metals using molecular dynamics simulations .
they obtained the structure factors , coordination numbers , interatomic distances and also the bond - angle distribution functions . according to them ,
if the components of an alloy have a large difference in the number of @xmath141 electrons there is a pronounced non - additivity of the pair interactions and a strong interaction between unlike atoms .
therefore , in certain amorphous alloys the interatomic distance of heteropolar bonds can be shorter than the distance of homopolar ones @xcite .
this also favors the formation of trigonal - prismatic units .
the shortening effect was not observed in this study .
the bond angle values of triangular and square faces of a distorted prism are found around @xmath142 , @xmath143@xmath144 , @xmath145 and @xmath146 .
thus , the bond angle values shown in fig .
[ fig8 ] suggest that distorted polyhedral units with triangular faces are present in the atomic structure of the amorphous nizr@xmath0 alloy and if there are square - faced polyhedral units , they are found in a very small quantity .
the presence of this kind of units was already reported in amorphous ni@xmath147ti@xmath148 @xcite , ni@xmath149ti@xmath150 @xcite , pd@xmath151si@xmath152 @xcite and ni@xmath136b@xmath135 @xcite alloys .
havinga _ et al_. @xcite reported crystallographic data for several compounds with cual@xmath0 ( c16)-type structure .
we have used their data for the nizr@xmath0 compound in the crystal office98 software@xcite to obtain the interatomic distances and coordination numbers for the first neighbors listed in table [ tab2 ] .
its structure has angles around @xmath15356@xmath154 , @xmath155 , @xmath156@xmath157 and @xmath158@xmath159 .
for the amorphous nizr@xmath0 alloy the @xmath160 functions ( @xmath18 and @xmath161 can be ni or zr atoms ) found are very similar to those present in the nizr@xmath0 compound .
on the other hand , the @xmath162 functions show some important differences .
ni sequence in the compound is linear ( @xmath163 ) while in amorphous nizr@xmath0 alloy it can also be found at triangular angles ( @xmath164 and @xmath157 ) ; the zr ni ni and zr ni zr sequences found in the amorphous nizr@xmath0 alloy show bond angles around @xmath165 that are not found in the compound .
these data suggest that in the amorphous alloy the local structure around a ni atom can be more distorted than around a zr atom .
the chemical short - range order parameter is a quantitative measurement of the degree of chemical short - range order .
there are several definitions of chemical short - range order parameters @xcite .
we have adopted the warren chemical short - range order ( csro ) parameter @xmath166 @xcite given below to compare the amorphous nizr@xmath0 alloy and the nizr@xmath0 compound .
@xmath167 here , @xmath168 and @xmath169 and @xmath170 are the total coordination numbers in the first shell . by using the coordination numbers given in table [ tab2 ] , we found @xmath171 and for the compound @xmath172 .
the chemical short - range order in the amorphous nizr@xmath0 alloy is similar to that found in a solid solution , for which @xmath173 .
the same behavior is observed in the nizr@xmath0 compound .
the amorphous nizr@xmath0 alloy was investigated by using awaxs , das and rmc simulations techniques .
the three @xmath7 factors were determined by using two independent ways : a combination of awaxs and das techniques and by combining awaxs and rmc simulations techniques .
the @xmath4 and @xmath5 factors obtained by both combinations showed an excellent agreement , but the simulations gave the best result for the @xmath3 factor .
the coordination numbers and interatomic distances calculated from the rdf@xmath174 functions obtained by these two combinations show a good agreement among themselves and with those extracted from the deconvolution of the @xmath88 and @xmath86 functions .
the results achieved in this study suggest that the combination of the rmc simulations method and awaxs can be used to obtain more reliable @xmath7 factors than those obtained by the matrix inversion method , in particular for the ni - ni pairs , which have the smallest weighting factor .
this was already observed in a neutron diffraction study performed in a cubr alloy @xcite , which is also an ill - conditioned problem similar to ours .
in addition , this combination gives us some evidence of the presence of distorted polyhedral units with triangular faces in the atomic structure of the amorphous nizr@xmath0 alloy .
the calculated warren csro parameter indicated that the chemical short - range order found in the amorphous nizr@xmath0 alloy is similar to that found in a solid solution , and there is some resemblance with that found in the nizr@xmath0 compound . | the local atomic structure of an amorphous nizr@xmath0 alloy was investigated using the anomalous wide - angle x - ray scattering ( awaxs ) , differential anomalous scattering ( das ) and reverse monte carlo ( rmc ) simulations techniques .
the awaxs measurements were performed at eight different incident photon energies , including some close to the ni and zr k edges . from the measurements
eight total structure factor @xmath1 were derived . using the awaxs data
four differential structure factors @xmath2 were derived , two about the ni and zr edges . the partial structure factors
@xmath3 , @xmath4 and @xmath5 were estimated by using two different methods .
first , the @xmath1 and @xmath2 factors were combined and used in a matrix inversion process .
second , three @xmath1 factors were used as input data in the rmc technique .
the coordination numbers and interatomic distances for the first neighbors extracted from the partial structure factors obtained by these two methods show a good agreement . by using the three - dimensional structure derived from the rmc simulations ,
the bond - angle distributions were calculated and they suggest the presence of distorted triangular - faced polyhedral units in the amorphous nizr@xmath0 structure . we have used the warren chemical short - range order parameter to evaluate the chemical short - range order for the amorphous nizr@xmath0 alloy and for the nizr@xmath0 compound .
the calculated values show that the chemical short - range order found in these two materials is similar to that found in a solid solution . |
cadmium zinc telluride ( czt ) has emerged as the material of choice for the detection of hard x - rays and soft gamma - rays with excellent position and energy resolution and without the need for cryogenic cooling . the high density of czt ( @xmath4 5.76 g/@xmath1 ) and high average atomic number ( @xmath5 50 ) result in high stopping power and in a large cross section for photoelectric interactions .
the main application of czt detectors is the detection of photons in the 10 kev to @xmath61 mev energy range .
czt has a major impact in various fields including medical imaging , homeland security applications , and space - borne x - ray and gamma - ray astronomy . to give a few examples of the latter , the swift mission launched in 2004 carries a wide field of view x - ray telescope for the discovery of gamma - ray bursts ( grbs ) in the energy range from 15 kev to 150 kev @xcite .
this burst alert telescope makes use of the coded mask technique to localize grbs with an accuracy of 1 - 2 arcmin .
it is built out of an array of 32,768 co - planar 2@xmath04@xmath04 mm@xmath7 czt detectors covering a total area of 0.5 m@xmath8
. the proposed exist ( energetic x - ray imaging survey telescope ) mission @xcite also uses the coded mask approach . with 15,000 pixelated czt detectors , each with a volume of [email protected]@xmath00.5 @xmath1 and with 16@xmath016 pixels , an angular resolution of 5 arcmin will be achieved and the energy range between 10 kev and 600 kev will be covered .
+ in this paper , we present a detailed study of czt grown with the modified horizontal bridgman ( mhb ) process by the company orbotech medical solutions ltd .
@xcite ( previously imarad ) .
this growth technique gives uniform substrates at high yields and thus modest costs .
one of the disadvantages of the mhb technique is a somewhat lower bulk resistivity of @xmath9 cm compared to @xmath10 cm obtained with the more conventional high pressure bridgman ( hpb ) process .
this results in higher leakage currents and therefore degrades the energy resolution at lower energies ( @xmath11 100 kev ) .
several authors recognized that it is possible to reduce dark currents in mhb detectors by using _ p_-type - intrinsic-_n_-type ( pin ) and metal - semiconductor - metal ( msm ) contacts @xcite .
both schemes , pin and msm , can indeed reduce the dark currents by factors @xmath1210 and improve on the energy resolution of the detectors at low ( @xmath11 100 kev ) energies .
encouraged by the result of these authors , we experimented with different cathode and anode materials on several mhb substrates , optimizing for the first time the cathode and anode contacts separately .
we discussed a comparison of different cathode materials in an earlier paper @xcite .
gold ( au ) was among the metals yielding the best performance in terms of dark current suppression and energy resolution as cathode contact material . with the objective to finalize our studies on contact materials , we present in this paper the results on different anode contact materials with an au cathode and added a study of the performance at lower temperatures which is especially important for space borne applications .
we derived @xmath2-products of electrons and holes needed for detector simulations .
these reproduce the measured data well , and can be used to further improve detector performance and design .
the paper is structured as follows : + in section [ sec : detectorfabrication ] , origin and some general properties of our substrate are presented .
we explain how we used this substrate to fabricate detectors suited for our measurements and give a short description of our experimental setup .
then in section [ sec : char ] , we discuss crystal inhomogeneities and present the results of photoluminescence mapping of the zn content and of infrared transmission microscopy . subsequently , in section [ sec : contact ] , the choice of the contact materials is explained , and the design of the detectors we used for comparison measurements is described in detail .
section [ sec : performance ] presents the results of our comparison experiments and a discussion of the influence of anode contact materials on the performance . in section [ sec : simulation ] , we show that we can reproduce experimental data with computer simulations . for this , @xmath2-products of electrons and of holes need to be known .
we explain how we measured the @xmath2-products of electrons and used simulations to derive the @xmath2-products of holes . in section [ sec : t ] , we present studies of the temperature dependence on the detector response and the electronic properties of czt .
as czt substrate , we used a detector from orbotech medical solutions ltd . @xcite grown by the modified high - pressure bridgman ( mhb ) process , delivered with a monolithic indium ( in ) cathode , and 8@xmath08 indium anodes ( pixels ) .
the size of the wafer was [email protected]@xmath00.5 @xmath1 , the pitch of the pixels ( anode contacts ) 2.46 mm , and the gap between the pixels 0.6 mm . in order to minimize unknown parameters , we used a newly delivered czt detector from which we removed the original contacts and deposited new ones using the following procedure . in a first step ,
the original contacts were removed by polishing with abrasive paper , followed by fine polishing with 0.5 @xmath13 m particle size alumina suspension , and then by rinsing clean with pure methanol . during the polishing process ,
the quality of the polished surface was constantly monitored with an optical microscope .
afterwards , the detector was etched for two minutes in a 1% bromine - methanol solution and subsequently rinsed with methanol .
metalization was performed with an electron beam evaporator keeping the original geometry : pixel pitch 2.46 mm , gap between pixels 0.6 mm .
a temperature sensor was used to ensure that the substrate temperature never exceeded 100@xmath3c . for the performance measurements
, we used a custom - designed pc card with 8@xmath08 contact pads , which had a delrin @xcite plastic fixture to mount the detector .
gold - plated , spring - loaded pogo - pins were used to connect the anode pixels to the pc card , which fans out the contacts to the readout electronics .
the cathode was biased at -1000 v , while the anode pixels were held at ground potential .
a hybrid electronic readout was used to read out the central 16 pixels of the detector .
three of the pixels in the center of the crystal and the cathode ( 4 channels ) were read out with a fast amptek 250 amplifier .
the post - amplified signals were digitized with a 4-channel 500 mhz oscilloscope , and transferred via ethernet to a personal computer .
this set - up allows us to measure the pulse length , and thus the drift times of the charge carriers with an accuracy of 10 ns at 662 kev .
the signals from the other 13 pixels were read out with an asic and a custom designed vme board .
therefore , we used only information of the central 16 pixels .
figure [ fig : pixelpattern ] shows a schematic view of the anode side in which these pixels are highlighted gray .
additionally , we monitored the noise of the readout electronics with a pulser : both readout chains have a fwhm noise of @xmath64 kev .
energy spectra were taken at 59 kev ( @xmath14am ) , 122 kev ( @xmath15co ) , and 662 kev ( @xmath16cs ) .
the zn - fraction ( x ) in a cd@xmath17zn@xmath18te sample determines the bandgap and therefore the energy required for the production of electron - hole pairs .
a 1% variation of the zinc fraction corresponds roughly to a 0.5% shift of the photopeak position . in the case of pixelated detectors , a variation in zn fraction
does not necessarily influence the performance , since each individual pixel s spectra can be corrected based on simple calibration measurements .
but , when we compare the performance of individual pixels on the same substrate , we have to exclude the possibility that local differences in crystal structure and composition are the cause of the observed effects .
therefore , the group at fisk university used a photoluminescence method to map the zinc content , and infrared transmission microscopy to determine crystal defects of the czt substrate .
photoluminescence ( pl ) mapping techniques @xcite normally use laser light having an energy above the bandgap of the crystal to illuminate the sample .
photons are absorbed and electronic excitations are created .
if through radiative relaxation light is emitted , the process is called photoluminescence . for each spot on the crystal wafer ,
this emitted pl light is collected and fed into a spectrophotometer .
the acquired pl spectra are fitted and the peak wavelength determined . from the peak position , the zn - fraction ( x ) is calculated .
@xcite .
our pl mapping system operated at room temperature and used a 20 mw he - ne laser ( @xmath19= 633 nm ) , a [email protected] cm@xmath8 x - y stage , and a s2000 miniature fiber optic spectrometer from ocean optics .
labview - based gui software was developed by the group at fisk university to perform automatic scanning , pl spectra acquisition , pl peak position analysis as well as to display in false color the zn index ( x ) or zn concentration on the surface of czt sample .
sample preparation consisted of fine polishing without etching .
the results are shown in fig .
[ fig : znmapping ] .
the zn fraction varies on the anode side between 0.110 and 0.135 , on the cathode side between 0.120 and 0.135 .
toney et al .
@xcite performed similar room temperature pl mappings of hpb samples , which were 2 mm thick and had an area of 1 - 2 cm@xmath8 .
these samples were etched in a 0.5% bromine - methanol solution for 7 minutes .
they showed a zn fraction fluctuation of 1 - 2% within one sample , and 5%-10% between the top and the heel of a 13 cm long boule .
this variation is smaller than in our substrate .
but , this may be due to the fact that our samples were not etched , since li et al .
@xcite reported that etching the substrates improves surface quality and reduces the zn - fraction variations .
microscopy in infrared ( ir ) light of a wavelength , where czt is sufficiently transparent , was used to localize crystal defects . in the case of czt
these are mostly tellurium ( te ) inclusions and grain boundaries .
we used unpolarized ir light with a wavelength of 960 nm to reveal the general condition of the crystal .
polarized light with a wavelength of 1150 nm was used to determine the stressed regions in the czt substrate . with a polarizer oriented at right angles with respect to the plane of polarization , unstressed material produces a dark field of vision . because stress generates birefringence , stressed regions are seen bright on a dark background . in fig .
[ fig : defects ] only fig .
3.2 was taken with polarized light . in figure
[ fig : defects ] , left hand side , the image of the whole substrate is shown .
there are some defects visible mainly in the outer region of the detector which we did not use for evaluation of detector performance .
these defects are surface scratches caused by the detector holder and the gold plated pogo - pins ( fig .
[ fig : defects ] : 1 , 2 ) . in the central region
relevant for our measurements , only one major defect is visible .
this defect is shown with higher resolution in fig .
[ fig : defects ] : panels 3.1 .
( unpolarized ) and 3.2 ( polarized ) .
the defect is most likely a crack with te inclusions . using fig .
[ fig : defects ] : 3.2 one finds that it covers a region of about 1 mm@xmath21 .
it is localized about 200 micron below the anode side .
we will show later that the pixel in which the defect is located has an inferior performance compared to the others .
we excluded this pixel from the analysis .
as already mentioned in section [ introduction ] , one of the disadvantages of the mhb method is the lower bulk resistivity , compared to crystals grown with the hpb method , causing high leakage currents that contribute to the intrinsic noise and deteriorate the performance of detectors at low energies ( @xmath11 100 kev ) .
the possibility to reduce bias currents by using high work - function ( wf ) blocking contacts on the cathode side of the n - type czt was discussed by various authors ( e.g. @xcite ) .
nemirovski et al .
@xcite reported on the characterization of three types of contact configuration : first , ohmic in contacts on both , anode and cathode side , second , rectifying au contacts on the cathode side with in contacts on the anode side , third , non - ohmic contacts on both , anode and cathode side ( au and aluminum ( al ) ) . at 122 kev
, he obtained an energy resolution of 5.2% , 4.6% , and 7% for the three configurations respectively .
vadawale et al .
@xcite presented results of 20 czt substrates .
the contacts were either standard in , platinum ( pt ) , au or hybrid contacts , i.e. pt on the cathode side , and in , or al on the anode side .
their best energy resolution was 5.0% at 122 kev with pt contacts followed by 5.7% with au contacts .
they found also that the detector uniformity depends on the anode contact material : low work - function material on the anode side produces a more homogeneous response throughout the detector .
while these studies give interesting hints , it is still unknown which is the best choice of anode contacts . therefore , we decided to study anode contact materials with a broad range of work - functions .
we chose in , ti , cr , and au with work - functions of 4.12 ev , 4.33 ev , 4.50 ev and 5.10 ev respectively for our comparison experiment .
we excluded pt ( work - function : 6.35 ev ) because substrates with pt contacts showed poor performance when the deposition was done with the help of an electron beam system ( for more details see @xcite ) .
as cathode contact material , we used au , since our earlier study ( see @xcite ) has shown that au reduces the dark current and improves the energy resolution .
the surface and detector treatment before and during contact deposition influences the detector performance .
therefore we deposited the four different metals on a single detector substrate using a quarter of the 64 pixels for each material as shown in fig .
[ fig : contactmaterial ] .
this design ( 4m - detector ) ensures that all pixels have the same history .
there remains another important point .
one has to make sure that performance differences between different contact materials are not caused by substrate inhomogeneities . in section [ sec :
char ] , we described already some of our precautions .
additionally we fabricated a detector from the same substrate chip , where instead of four different materials only one , ti , was used ( ti - detector ) . in this way we were able to compare pixels of the four sectors under identical conditions .
the results discussed in this section relate to only 15 of the 16 central pixels .
one pixel was removed from the analysis , because infrared microscopy ( see section [ sec : char ] ) has shown a major crystal defect in the area of this pixel leading to very poor performance in both contact configurations .
energy resolutions discussed in this section are resolutions of single pixels . signals in neighboring pixels were ignored . for higher energies
( 122 kev and 662 kev ) the anode signal was corrected for the depth of the photon interaction ( doi ) based on the cathode - to - anode signal ratio .
table [ tab : am241 ] summarizes the results obtained with both contact configurations for an @xmath14am source .
the energy resolution is averaged over the three or four pixels of the 4m - detector with identical contact material and compared to an average over the corresponding pixels of the ti - detector .
the numbers are corrected for electronic noise . in parentheses
, we added the uncorrected data .
both sets of data present the same picture .
therefore , we discuss the uncorrected data .
ti with resolutions of ( [email protected] ) kev ( 4m - detector ) as well as ( [email protected] ) kev , ( [email protected] ) kev , ( [email protected] ) kev and ( [email protected] ) kev ( ti - detector ) had obviously the best performance followed closely by in ( [email protected] ) kev .
au and cr as anode contact materials are inferior by @xmath68% : ( [email protected] ) kev ( au ) , ( [email protected] ) kev ( cr ) . table [ tab : meanfwhm ] summarizes our results at 122 kev , and 662 kev .
the important result is that all four different anode materials show rather similar energy dependence on the resolution , and that the order of performance remains unchanged . in fig .
[ fig : spectra ] , the energy spectra of a single ti pixel are shown for three different photon energies . in case of the higher energies 122 kev ( @xmath15co ) and 662 kev ( @xmath16cs ) ,
the anode signal was corrected for the depth of the photon interaction ( doi ) based on the cathode - to - anode signal ratio .
given the large pixel pitch of 2.46 mm , we obtained the excellent energy resolution of 1.4 kev ( 4.0 kev ) , 2.0 kev ( 4.1 kev ) , and 7.2 kev ( 9.9 kev ) for photon energies of 59 kev , 122 kev , and 662 kev , respectively .
again , the values given are corrected for electronic noise with the uncorrected data added in parentheses . the peak - to - valley ratio defined as the ratio of the maximal counts in the photopeak to the mean of the channels between 560 kev and 580 kev , was @xmath615.5 for 662 kev .
performance measurements for czt crystals grown with the hpb process have been reported in the literature @xcite .
the energy resolution was 0.73% and 0.93% at 662 kev for two different detectors . besides a different production process
, there are other important features in which our setup varies from theirs . the substrate measured [email protected]@xmath01.0 @xmath1 with 11@xmath011 anode pixels and a pitch 1.27 mm .
a steering grid between pixel anodes was used which was biased at negative voltage to focus the electrons to the anodes .
the grid electrode was 100 @xmath13 m wide and had a 200 @xmath13 m gap between grid and pixel .
these differences , one has to keep in mind when comparing both experiments , because these parameters have a strong influence on the energy resolution . a smaller pixel pitch and a larger thickness enhances the small pixel effect .
using simulations we have shown , that an increase in the thickness from 0.5 cm to 1 cm for a czt detector with pixel pitch 1.64 mm will improve the energy resolution by @xmath630% ( for more details see @xcite ) .
additionally we have compared simulated detectors with our geometry and the one used by zhang et al . @xcite but without a steering grid and found that the energy resolution is @xmath640% better in the latter case . taking this into account and the fact that a steering grid will further improve the resolution
, we conclude that the resolution , we obtained , is comparable to the energy resolutions reported by zhang et al.@xcite of 5.0 kev and 6.2 kev at 662 kev .
in this section , we first describe 3-d simulations of the detector response .
subsequently , we present measurements and detailed comparisons of the simulated and experimentally measured detector response which allow us to determine the @xmath2-products for electrons and holes .
we simulated a [email protected]@xmath00.5 @xmath1 detector with 8@xmath08 pixels with a pixel pitch of 2.46 mm and a monolithic cathode .
the simulation package consists of three components ( for further details see @xcite ) .
the first component calculates with a finite - difference method the 3-d electric field as well as the weighting potentials of the pixel and cathode contacts . in the second
, the geant 4.0 code @xcite is used to simulate the interactions of the incident @xmath23rays and the secondaries produced in the detector .
the third component uses the information , where the primary photons deposit energy in the detector to generate electrons and holes , and tracks them through the detector .
this component uses the electric field and the weighting potentials to compute the charge induced on the contacts .
electrons reaching the substrate surface in the gap between the pixels were left at the position , where they hit the surface , i.e. the surface conductivity was set to infinity .
the detector material was assumed to be uniform .
we measured electron mobility as well as electron lifetime and used these values as input parameters in our simulations .
hole mobilities and lifetimes were determined from comparisons of simulated and experimental data .
the weighting potentials of a central pixel and the cathode contact are presented in figure [ fig : wp ] .
the graphs show a small - pixel - effect @xcite , i.e. the pixel weighting potential has a steep gradient close to the pixel . as the pixel pitch ( 2.46 mm ) is roughly comparable to the detector thickness ( 5 mm ) ,
the small - pixel - effect is not very pronounced .
therefore , electrons and holes induce charge on the anode , even if they move deep inside the detector .
the cathode signal does not show any small - pixel - effect and depends almost linearly on the distance from the cathode contact .
the first step in the determination of electron mobility and lifetime is the measurement of the electron @xmath2-product ( @xmath24 ) , which can be calculated applying formula [ mutau ] : @xmath25 where @xmath26 is the detector thickness , and @xmath27 and @xmath28 are the photopeak centroids for two different bias voltages @xmath29 and @xmath30 @xcite .
a low energy source has to be used to ensure that most of the interaction occurs near the cathode , and that hole contributions can be neglected .
the left panel of fig .
[ fig : simparameter ] shows the @xmath31-product for the central 16 pixels .
the mean is @xmath32 cm@xmath8 v@xmath33 and the rms is 0.95 .
+ the @xmath31-product of hpb czt substrates are given by f. zhang et al .
@xcite and by s. barthelmy et al .
f. zhang et al .
@xcite reported the @xmath31-product of two different hpb czt substrates with means of @xmath34 cm@xmath8 v@xmath33 and of @xmath35 cm@xmath8 @xmath36 and rms values of @xmath37 cm@xmath8 @xmath36 and @xmath38 cm@xmath8 @xmath36 respectively , which shows that czt detectors grown with the mhb process have a similar mean , but a wider spread .
+ the @xmath39-products of all 32,768 planar czt hpb detectors ( 4@xmath04@xmath02 mm@xmath7 ) used in the burst alert telescope(bat ) on - board the swift gamma - ray burst explorer ( see section [ introduction ] ) were measured @xcite , and they vary between 5.0@xmath010@xmath40 and 1.0@xmath010@xmath41 cm@xmath8 v@xmath33 . the variation within a single detector was not determined .
these measurements show that the two detectors used by zhang et al .
have a @xmath31-product above the average value of hpb czt detectors .
the electron mobilities can be inferred from the widths ( durations ) of the induced anode pulses , which are mainly determined by the drift times of the electrons . the right side of fig .
[ fig : simparameter ] shows the correlation of the pulse widths and the induced anode charge . with a detector bias of -1000 v ,
the longest observed drift times were @xmath50.35 @xmath13sec .
thus , given that electrons drift up to 0.5 cm through the detector , we determined an electron mobility of @xmath42 cm@xmath8 v@xmath33 s@xmath33 and a mean electron lifetime of @xmath43 s using the previously measured mean @xmath31-product . with these electron @xmath44 and @xmath45-values , we deducted the corresponding values for holes by comparing the measured anode to cathode correlation with simulated correlations for different @xmath46 and @xmath47 .
[ fig : vergleichsim_data ] right side displays the simulated correlation with @xmath48 cm@xmath8 v@xmath33 s@xmath33 and @xmath49 s and for comparison the corresponding correlation of our experimental data ( left side ) .
even the finer details of the experimental data plot are reproduced well .
we used this set of parameters in all simulations .
the following paragraphs will show that simulations based on these parameters reproduce the experimental data .
having determined the hole parameters , we can study separately the contributions of electrons and holes to the anode and cathode signals ( see fig .
[ fig : cathodesignal ] ) .
the anode signal dependency on the distance of energy deposition from the anode is given in the left panel of fig .
[ fig : cathodesignal ] .
the signal drops close to the anode due to the steep weighting potential increase . in this area
the hole contribution is maximal and decreases to almost zero close to the cathode .
the cathode signal dependency is given on the right panel .
it is almost linear with the distance from the anode side .
the deviation close to the cathode is caused by holes reaching the cathode before being trapped .
the approximately linear dependence of the induced cathode charge on the depth of the photon interaction ( doi ) can be used as a diagnostic tool to test , whether the substrate is uniform .
we compared the behavior of the signal dependency on the doi by dividing the total cathode signal into five different signal ranges ( 0.8 - 1.2 v , 1.2 - 1.6 v , 1.6 - 2.0 v , 2.0 - 2.4 v and 2.4 - 2.8 v ) . for each interval ,
the position of the photopeak and the number of events in the photopeak was determined separately .
the number of photopeak events is defined as the number of events lying between @xmath50 around the photopeak position .
the correlation of the anode versus the cathode signals depends on where in the crystal the charge is generated and to where it drifts .
figure [ fig : doidependenc1 ] shows the measured and simulated mean photopeak positions for different intervals of the cathode signal . while the measured data consists of four different sets for each of the four materials , there is only one set of simulated data , because the contact material was not taken into account in our simulations .
the agreement between data and simulations is excellent for all but one pixel .
this pixel is the one which was excluded from the analysis , because infrared transmission microscopy showed the presence of a crystal defect close to the anode ( section [ subsec : defects ] ) .
figure [ fig : doidependenc2 ] compares the measured and the simulated relative numbers of photopeak events for different ranges of the cathode signal .
the measured photopeak events distribute as in the simulations , one pixel deviating more than the others .
for this pixel , infrared transmission microscopy shows the presence of a crystal defect close to the anode ( see sect .
[ subsec : defects ] ) .
because the simulation matches the experimental data well , we conclude that in both cases for a given cathode signal interval the charge generation occurs in the same range of dois for simulation and experiment . for the deviating pixel ( the same as above )
, the mean photopeak position varies less than for the others , confirming that the crystal defect affects the charge induced at the cathode and/or the anode .
we conclude that the full volume under all but one pixel is active detector volume and produces proper cathode signals .
in satellite experiments ambient air temperature is well below @xmath51c .
therefore we studied the dependence of the electronic properties and the performance of the czt detector over a broad temperature range from room temperature down to -30@xmath3c .
one important parameter is the @xmath39-product .
it influences the detector performance .
decreasing product means inferior resolution .
therefore we measured the electron @xmath2-product and the spectra as function of temperature with an @xmath14am source as described in section [ sec : simulation ] .
figure [ fig : tempabhaengigkeit ] gives the results .
the product shows the highest values at and above 0@xmath3c and decreases at lower temperatures by up to @xmath630% .
similar behavior has been reported for hpb substrates @xcite which had a maximum at 5@xmath3c with @xmath24@xmath6 9@xmath010@xmath52cm@xmath21 v@xmath33 with a drop of about @xmath630% at -20@xmath3 to @xmath24@xmath6 6@xmath010@xmath52cm@xmath8 v@xmath33 . in fig .
[ fig : temp_spectra ] 122 kev energy spectra are presented .
they are measured at 10@xmath3c and at -30@xmath3c together with their anode to cathode signal correlations .
the top two panels show that the energy resolution deteriorates markedly , when the detector is operated at lower temperatures and all other conditions ( bias voltage ) are kept constant .
the anode - cathode correlation shows that the width of the photopeak increases for all cathode signal amplitudes , and thus for all dois . in the measurements that just have been discussed ,
we kept the bias voltage constant at -1000 v. this voltage has been chosen , because we found that at temperatures above 10@xmath3c a change of voltage from -1000 v to -1500 v will increase the dark current by several orders of magnitude , so much , that no useful spectra could be obtained . if the detector is cooled , however , this electrical breakdown phenomenon is suppressed , and higher bias voltages become possible . the bottom panel of fig .
[ fig : temp_spectra ] shows an energy spectrum measured at -30@xmath3c and a cathode bias of -1500 v. with the increased bias voltage most of the room temperature performance is recovered : energy resolution of 5.5% at -30@xmath3c compared to 4.4% at 10@xmath3c , and a somewhat lower signal amplitude ( 0.28 v and to 0.29 v respectively ) .
while most of the performance degradation at lower temperatures will be due to the reduction of the @xmath39-product , another effect may have influenced our measurements to some extent : vibrations of the temperature chamber which was used to cool the czt detector .
in this paper , we present the results of a detailed study of the performance of a mhb czt detector from the company orbotech medical solutions ltd . @xcite .
+ using photoluminescence , we mapped the spatial distribution of the zinc content , which showed a variation of @xmath63% .
this variation does not affect the energy resolution of the detector if information about the location of the interaction is available ( e.g. pixelated detectors ) , because in this case the effect can be corrected for . with infrared transmission microscopy
, we could detect some crystal defects and could correlate poor pixel performance with such defects .
+ we tested various anode contact materials and found that ti showed the best performance of 2.0 kev , 1.9 kev and 7.3 kev at 59 kev , 122 kev and 662 kev , respectively .
we are using this metal now to contact all our mhb detectors .
we showed detailed comparisons of experimentally measured and simulated detector response . after adjusting the hole mobility and lifetime
, our simulation gives a good description of the measured data and shows that we have a deep understanding of our data .
therefore , the simulation can be used to optimize the detector design as the pixel width and pitch .
+ we studied the detector properties as a function of temperature . the electron @xmath54-product drops by about 30% , when cooling the substrate from 20@xmath3c to -30@xmath3c with constant bias voltage , and the energy resolution deteriorates accordingly , a behavior that has been noted already for mhb substrates @xcite . for hpb substrates , a corresponding decrease of the electron @xmath54-product has been observed @xcite .
we were able to recover almost all of the room temperature performance by increasing the bias voltage at low temperatures , which is possible , because at lower temperatures the breakdown voltage is increased . taking everything into account ,
we conclude , that the overall performance of czt detectors , produced with the cost - effective mhb process , is comparable to the performance of czt detectors which are produced with the expensive hpb process .
+ _ acknowledgments : _ we thank uri el hanany from orbotech inc . for several free czt detectors .
we acknowledge s. komarov , l. sobotka , d. leopold , and j. buckley for helpful discussions .
thanks to electrical engineer p. dowkontt , and electrical technician g. simburger for their support .
this work is supported by nasa under contracts nng04wc176 and nng04gd70 g , and the nsf / hrd grant no .
0420516 ( crest ) .
the authors at fisk university gratefully acknowledge financial support from the national science foundation through the fisk university center for physics and chemistry of materials ( cpcom ) , cooperative agreement ca : hrd-0420516 ( crest program ) , and from us doe through the national nuclear security administration ( nnsa ) , office of nonproliferation research and engineering ( na-22 ) , grant no .
de - fg52 - 05na27035 .
-product for the 16 central pixels .
the right panel displays the anode charge as function of pulse width ( duration ) for a cathode bias of -1000 v. the pulse width is largely given by the drift time of the electrons through the detector.,width=529 ] co source ) and the anode signal dependence on cathode signal with a cathode bias voltage of -1000 v for two different temperatures ( 10@xmath3c , top , and -30@xmath3c , middle ) .
the signals were measured with the amptek 250 readout chain and are therefore given as voltages . at the bottom the spectrum with a bias of -1500 v at a temperature of -30@xmath3c , as in the center plot .
the energy resolutions given are 4.4% , 10.6% and 5.5% fwhm ( top to bottom).,width=566 ] l l l l + + & + ti & [email protected] ( [email protected] ) & au & [email protected] ( [email protected] ) + ti & [email protected] ( [email protected] ) & cr & [email protected] ( [email protected] ) + ti & [email protected] ( [email protected] ) & in & [email protected] ( [email protected] ) + ti & [email protected] ( [email protected] ) & ti & [email protected] ( [email protected] ) + l l l + & 122.1 kev & 662 kev + au & [email protected] ( [email protected] ) & [email protected] ( [email protected] ) + cr & [email protected] ( [email protected] ) & [email protected] ( [email protected] ) + in & [email protected] ( [email protected] ) & [email protected] ( [email protected] ) + ti & [email protected] ( [email protected] ) & [email protected] ( [email protected] ) + | the detector material cadmium zinc telluride ( czt ) achieves excellent spatial resolution and good energy resolution over a broad energy range , several kev up to some mev .
presently , there are two main methods to grow czt crystals , the modified high - pressure bridgman ( mhb ) and the high - pressure bridgman ( hpb ) process .
the study presented in this paper is based on mhb czt substrates from the company orbotech medical solutions ltd .
@xcite .
former studies have shown that high - work - function materials on the cathode side reduce the leakage current and therefore improve the energy resolution at lower energies .
none of the studies have emphasized on the anode contact material .
therefore , we present in this paper the result of a detailed study in which for the first time the cathode material was kept constant and the anode material was varied .
we used four different anode materials : indium , titanium , chromium and gold , metals with work - functions between 4.1 ev and 5.1 ev .
the detector size was [email protected]@xmath00.5 @xmath1 with 8@xmath08 pixels and a pitch of 2.46 mm .
the best performance was achieved with the low work - function materials indium and titanium with energy resolutions of 2.0 kev ( at 59 kev ) and 1.9 kev ( at 122 kev ) for titanium and 2.1 kev ( at 59 kev ) and 2.9 kev ( at 122 kev ) for indium .
taking into account the large pixel pitch of 2.46 mm , these resolutions are very competitive in comparison to those achieved with detectors made of material produced with the more expensive conventional hpb method .
we present a detailed comparison of our detector response with 3-d simulations .
the latter comparisons allow us to determine the mobility - lifetime - products ( @xmath2-products ) for electrons and holes .
finally , we evaluated the temperature dependency of the detector performance and @xmath2-products .
for many applications temperature dependence is important , therefore , we extended the scope of our study to temperatures as low as -30@xmath3c .
there are two important results . the breakdown voltage increases with decreasing temperature , and electron mobility - life - time - product decreases by about 30% over a range from 20@xmath3c to -30@xmath3c .
the latter effect causes the energy resolution to deteriorate , but the concomitantly increasing breakdown voltage makes it possible to increase the applied bias voltage and restore the full performance .
cdznte , czt detectors , temperature dependency 95.55.ka |
let @xmath7 be a finite , simple , connected , and undirected graph of order @xmath8 . for a vertex @xmath9 ,
the _ open neighborhood of @xmath10 _ is the set @xmath11 , the _ closed neighborhood of @xmath10 _ is the set @xmath12=n_g(v ) \cup \{v\}$ ] , and the _ degree _ of a vertex @xmath10 is @xmath13 ; we drop @xmath1 if it is clear in context .
+ let each vertex of a graph @xmath1 be given either the color black or the color white .
denote by @xmath14 the initial set of black vertices of @xmath1 .
the color change rule " changes the color of a vertex @xmath10 from white to black if @xmath10 is the only white neighbor of a black vertex @xmath15 ; in this case , we say that _ @xmath15 forces @xmath10 _ and write @xmath16 . the initial black set @xmath14 is said to be zero forcing " if , after finitely many applications of the color change rule , all vertices of @xmath1 are forced to black . the zero forcing number of @xmath1 " , @xmath0 , is defined as the minimum of the cardinalities of all zero forcing sets of @xmath1 .
+ the notion of a zero forcing set , as well as the associated zero forcing number , of a graph was introduced by the aim minimum rank special graphs group " ( see @xcite ) to bound the minimum rank for numerous families of graphs .
zero forcing parameters were then studied and applied to the minimum rank problem in @xcite .
more recently , viewing zero forcing as a dynamical process unfolding on a graph , the four authors of @xcite and , independently , hogben et al in @xcite have studied the number of steps ( or units of time ) it takes for an initial zero forcing set to force all vertices of a graph to black .
it is also noteworthy that , independent of the forgoing efforts and in connection with the control of quantum systems , this parameter has been studied by some physicists who called it the graph infection number " @xcite . on the other hand , probability has been playing a central role in graph theory ever since the seminal paper by erds and renyi @xcite . in the 2005 _ nature _
article entitled evolutionary dynamics on graphs " @xcite , lieberman , hauert , and nowak studied the question : what is the probability that a newly introduced mutant generates a lineage that takes over the whole population ? " + let us henceforth use the term classical zero forcing " to refer to the notion of zero forcing already defined .
taking the viewpoint that classical zero forcing is a kind of graph infection " or
spread of a mutation " , it s then natural indeed compelling to formulate a notion of probabilistic zero forcing " wherein classical zero forcing is obtained as a special case . to this end , let @xmath14 be an initial black set of a graph @xmath1 ; here @xmath17 , but @xmath14 is not necessarily zero forcing in the classical sense . with respect to @xmath14 , we define a probabilistic color change rule " ( hereafter also referred to as pccr " ) in which a black vertex forces each of its white neighbors independently to the color black with probability given as follows .
let @xmath18 denote the probability of the event @xmath19 ( which we think of as the forcing probability of @xmath15 on @xmath10 " ) that the vertex @xmath15 forces the color of a white vertex @xmath10 to black .
we adopt the following + @xmath20\subseteq z \\ \frac{\mid n[u]\cap z\mid}{\deg(u ) } & \mbox { otherwise . }
\end{array } \right . } \ ] ] we will assume that a vertex @xmath15 forces each of its white neighbors independently ; i.e. , we think of the action of a black vertex on each of its @xmath21 white neighbors as a binomial ( or bernoulli ) experiment with probability of success ( where by a success we mean the conversion of one of the @xmath21 white neighbors to black ) given by the formula above .
since @xmath22 when @xmath23 and @xmath24 are both white neighbors of @xmath15 , we will simply write @xmath25 to specify the forcing probability of @xmath15 when ambiguity is not a concern . let @xmath26 denote the probability ( conversion probability " ) that a vertex @xmath10 will be black after one global application ( iteration ) of pccr , given a set of black vertices @xmath27 .
we thus define @xmath28 if @xmath29 and @xmath30 if @xmath31 , where @xmath32 ; i.e. , we assume that each white vertex is being acted upon independently by its black neighbors .
+ observe that classical zero forcing where a black vertex @xmath15 forces its sole white neighbor @xmath10 is recovered as a pair , a black vertex @xmath15 and a white vertex @xmath10 , with forcing probability @xmath33 and conversion probability @xmath28 . from the perspective of modeling a real world "
situation , our formulation is sensible : it only takes one affected entity to spread a condition to an entire population .
we want to mention here that curtis miller has independently formulated a number of probabilistic variants of zero forcing @xcite .
but our present definition , as well as our approach and emphasis , is distinct from the ones he considered ( personal communication ) . notice that , once a classical zero forcing ( black ) set @xmath34 containing an initial black set @xmath35 of @xmath1 is attained through probabilistic zero forcing , classical zero forcing will be in force as a part of probabilistic zero forcing ; before then , some combination of classical and probabilistic zero forcing occurs .
in the rest of this paper , we will at times abbreviate probabilistic zero forcing as pzf " .
let s call a graph @xmath1 , with vertices labeled @xmath36 through @xmath37 , a colored graph if each of its vertices receives the color black or the color white , and let s regard as distinct two colored graphs @xmath38 and @xmath39 with the same underlying graph @xmath1 when there is a vertex @xmath40 which has one color in @xmath38 but has the other color in @xmath39 . for a colored graph @xmath1 , denote by @xmath41 the set of all colored graphs which can be obtained with positive probability from @xmath1 in one global application of pccr . let a graph @xmath42 , with initial black set @xmath14 , be given .
then pzf generates a countable collection of probability sample spaces @xmath43 , one sample space @xmath44 at each step @xmath45 which can be inductively defined as follows : @xmath46 , where @xmath47 is @xmath42 with @xmath48 colored black and @xmath49 colored white ; once @xmath50 is defined , @xmath51 .
let @xmath52 be the probability function defined on @xmath44 which is derived from the pccr .
it follows that @xmath53 for each @xmath54 .
the @xmath55 just defined represents a discrete dynamical system on the underlying graph @xmath1 , and its evolution is the central problem of the pzf theory .
+ let @xmath56 and @xmath57 ; if @xmath58 , we ll write @xmath59 . define @xmath60 . if @xmath61 , then it follows that @xmath62 , since @xmath63 .
it s clear that , relative to an initial black set @xmath14 , @xmath64 , where @xmath65 is previously defined .
also notice that @xmath66 by _ color change asymmetry _ ; namely , if a vertex @xmath10 is black in a colored graph @xmath1 , then @xmath10 is black for each member of @xmath41 .
suppose for some @xmath67 , there is no @xmath68 such that @xmath1 contains a black vertex @xmath10 which forces with probability one ( i.e. , @xmath10 has a lone white neighbor ) , then @xmath69 ( as sets ) for @xmath70 .
since there are at most @xmath71 colored graphs on the underlying graph @xmath42 , the sequence @xmath44 is eventually constant ; i.e. , there is some @xmath72 such that @xmath73 for any @xmath74 .
note that this does not mean that @xmath75 for @xmath74 .
+ the evolution of our discrete dynamical system is a markov process . here is the gross view : let @xmath76 be an ordered listing of the @xmath77 possible colored graphs ( possible states of the system ) underlied by a graph @xmath1 of order @xmath37 : each @xmath78 can be identified with the positive integer @xmath79 . for each @xmath54
, we associate a _ probability vector _ @xmath80 of length @xmath77 whose @xmath79-th entry denotes the probability that our system is in the state @xmath79 at time @xmath45 .
the stationary transition matrix @xmath6 for the system has size @xmath81 , and each of its entries @xmath82 provides the probability that a system in state @xmath83 at time @xmath45 will transition into state @xmath79 at time @xmath84 ; @xmath82 is thus determined by the pccr applied to the state @xmath83 .
+ we observe that , starting with a colored graph @xmath47 of order @xmath37 in @xmath85 , by the fact that the black set is expanding ( or non - contracting ) from any @xmath86 to each @xmath87 , the number of colored graphs ( and hence the size of the transition matrix ) requiring consideration is typically and tremendously reduced from @xmath77 .
recall that , in the theory of markov processes , an _ absorbing state _ is one that can not be exited once entered ; i.e. , it is a state @xmath79 such that @xmath88 ; a state @xmath79 is _ transient _ ( also called _ non - recurrent _ ) if there is a state @xmath83 reachable from @xmath79 , but state @xmath79 is not reachable from @xmath83 . if there is only one absorbing state @xmath89 , then @xmath90 for each state @xmath79 ( see p.247 and p.241 of @xcite , also see pp.58 - 66 of @xcite for proofs ) .
since our dynamical system has transient states except for one absorbing state ( the colored graph @xmath91 where all vertices of @xmath91 are black ) , we have the following that is , a colored graph will turn entirely black with probability one as the number of steps @xmath45 goes to infinity . the following is an instructive example .
+ consider the complete bipartite graph @xmath94 , where @xmath95 , which we ll also call the star with @xmath21 pendants ; let the center @xmath96 be the only black vertex of @xmath47 .
we have @xmath46 and @xmath97 for @xmath98 ; i.e. , all possible colored graphs ( or states ) occur in every sample space @xmath44 for @xmath98 .
let @xmath91 denote the state where all vertices of the star are black .
then proposition [ limiting state ] says that @xmath99 , which can also be seen via a direct , inductive argument on the number of pendants @xmath21 via the following observation : the center @xmath96 is the only vertex capable of forcing ; each @xmath100 is obtained from @xmath101 through a binomial experiment wherein each white pendant is turned black by the forcing strength of @xmath96 ; the average forcing strength of @xmath96 is monotonically increasing toward @xmath36 as @xmath45 goes to infinity .
@xmath102 + in order that probabilistic zero forcing be a natural extension of classical zero forcing , there ought to be a probability @xmath5 assigned to each pair @xmath103 which indicates the likelihood that the graph @xmath1 will be forced entirely black given the initial black set @xmath104 under finitely many applications of the pccr ; moreover , this probability should satisfy the following conditions : 1 ) @xmath105 , 2 ) @xmath106 if and only @xmath6 is a zero forcing set of @xmath1 , 3 ) @xmath107 for @xmath108 .
given a pair @xmath109 and the associated @xmath55 , define @xmath110 to be the subset of @xmath44 such that @xmath111 if and only if the black set of @xmath112 contains a classical zero forcing set of @xmath1 ; note that @xmath113 for some @xmath114 by the definition of pccr .
it s clear that the following definition of @xmath5 meets the three requirements . + let @xmath115 , where @xmath67 is the least @xmath45 such that @xmath113 .
we think of @xmath5 as the probability that a graph @xmath1 with an initial black set @xmath104 will turn entirely black in finite time " .
+ let @xmath116 , @xmath117 , with initial black set @xmath118 be given ; also let @xmath119 .
recall two adjacent vertices form a classical zero forcing set for @xmath1 . then , in going from @xmath85 to @xmath120 , the only forcing action is given by @xmath121 .
only one ( call it @xmath122 ) of the four colored graphs of @xmath120 does not contain a classical zero forcing set , and thus @xmath123 . note the equivalence , trivially here , of @xmath5 to @xmath124 .
@xmath102 + let @xmath1 be the tree given in fig [ tree ] , with initial black set @xmath125 . the forcing action propagates from @xmath96 to the three pendants ( vertices of degree @xmath36 ) . in fig [ tree ] ,
the number next to each edge @xmath126 is @xmath18 ; notice that forcing can only occur in one direction for each edge in this pair @xmath109 .
for example , the number @xmath127 is next to the edge @xmath128 because @xmath129 given that the vertex @xmath130 is already ( forced to ) black .
likewise , the number next to each vertex indicates the probability that each vertex will be forced black given the initial black vertex @xmath96 . + we determine @xmath5 by starting with the observation that a colored graph containing a classical zero forcing set first occurs in @xmath131 , since a black set @xmath132 containing @xmath96 is classically zero forcing if and only if @xmath133 . the top two rows of figure [ tree - samplespaces ] show @xmath85 and @xmath120 , whereas the third row only shows @xmath134 , namely the colored graphs of @xmath131 which contain classical zero forcing sets . from left
to right , the first three graphs of the third row form @xmath135 and last three graphs of the third row form @xmath136 , where @xmath39 and @xmath137 are the second and the fourth graphs , respectively , of the second row .
now , it s clear that @xmath138 .
thus , @xmath139=\frac{4}{9}$ ] + we can also deduce @xmath5 in this case without explicitly enumerating elements of the sample spaces @xmath43 in the following way .
first , still notice that a black set @xmath132 containing @xmath96 is classically zero forcing if and only if @xmath133 ; also notice that @xmath140 , but @xmath141 and @xmath142 .
thus , the probability that @xmath1 will be turned entirely black by @xmath143 under pzf is given by @xmath144 , which must be the same as the probability @xmath145 of the event @xmath146 that @xmath147 or @xmath148 .
let @xmath149 denote the event that @xmath150 .
observe @xmath151 ; i.e. , if @xmath146 happens , then @xmath149 already happened .
thus , we have @xmath152 . @xmath102 +
the following is a basic question in our pzf theory at this juncture one which we ll solve for the graph depicted in figure [ tree ] in the case the initial black set is a singleton . +
[ the question ] let @xmath153 , where @xmath6 is an initial black set of @xmath1 and @xmath83 is an integer between @xmath36 and @xmath0 .
what is @xmath154 for each @xmath155 and what are the initial black sets @xmath6 attaining @xmath154 for each @xmath83 ? + let s answer question [ the question ] for @xmath156 when @xmath1 is the star @xmath94 with @xmath21 pendants , and let @xmath157 for simplicity .
label the center of the star @xmath96 and the pendants @xmath23 through @xmath158 .
given the symmetry group of the star , it s enough to consider two cases : 1 ) @xmath125 , 2 ) @xmath159 . in case 1 ) , @xmath160 and @xmath161 ; thus , @xmath162 . in case 2 ) , we have @xmath163 and @xmath164 since @xmath157 by hypothesis . in passing from @xmath120 to @xmath131
, we have @xmath165 and @xmath166 ; thus , @xmath167 .
obviously , @xmath168 when @xmath157 .
we note that it s a straightforward matter to determine @xmath154 for each @xmath169 since , for each @xmath6 of cardinality @xmath83 , we only need to consider the case @xmath170 and the case @xmath171 . @xmath102 + returning to the graph depicted in figure [ tree ] , let s consider all cases @xmath172 where @xmath173 .
there are four distinct cases to consider : 1 ) @xmath174 , 2 ) @xmath175 , 3 ) @xmath176 , 4 ) @xmath177 .
having found in case 1 ) that @xmath178 in the previous example , let s consider case 2 ) .
by the pccr , we have @xmath179 , and we have in @xmath180 that @xmath181 ; we get @xmath182 . in case 3 ) , @xmath161 and @xmath183 . in case 4 ) , since @xmath184 , @xmath120 contains only the colored graph where the only black vertices are @xmath185 and @xmath130 , and @xmath164 . since @xmath166 , @xmath186 .
therefore , we conclude that @xmath187 @xmath102 + _ acknowledgement .
_ the authors thank kiran chilakamarri for a number of very helpful discussions .
in particular , they thank him for explaining to them the idea of iterated sample spaces and for locating an informal reference on the theory of absorbing markov chains .
the authors also thank curtis miller for the aforementioned preprint , which gave additional impetus for them to finish this paper .
lastly , the authors thank the anonymous referee for some helpful comments and corrections .
aim minimum rank - special graphs work group ( f. barioli , w. barrett , s. butler , s.m .
cioab , d. cvetkovi , s.m .
fallat , c. godsil , w. haemers , l. hogben , r. mikkelson , s. narayan , o. pryporova , i. sciriha , w. so , d. stevanovi , h. van der holst , k.v .
meulen , a.w .
zero forcing sets and the minimum rank of graphs .
_ linear algebra appl . _
* 428/7 * ( 2008 ) 1628 - 1648 .
f. barioli , w. barrett , s.m .
fallat , h.t .
hall , l. hogben , b. shader , p. van den driessche , and h. van der holst , zero forcing parameters and minimum rank problems .
_ linear algebra appl . _
* 433 * ( 2010 ) 401 - 411 . | the _ zero forcing number _ @xmath0 of a graph @xmath1 is the minimum cardinality of a set @xmath2 of black vertices ( whereas vertices in @xmath3 are colored white ) such that @xmath4 is turned black after finitely many applications of the ( classical ) color change rule " : a white vertex is converted to a black vertex if it is the only white neighbor of a black vertex .
zero forcing number was introduced and used to bound the minimum rank of graphs by the
aim minimum rank special graphs work group " .
we introduce here a probabilistic color change rule ( pccr ) which is a natural generalization of the classical color change rule .
we introduce a theory of probabilistic zero forcing arising out of the pccr ; the theory yields a quantity @xmath5 , which can be viewed as the probability that a graph @xmath1 with an initial black set @xmath6 will be converted entirely to the color black .
we also interpret the evolution of the sample spaces of this theory as a markov process .
we end with a few basic examples illustrating this theory .
zero forcing , probabilistic zero forcing , probabilistic color change rule , markov process , absorbing chains 05d40 , 60j10 , 05c50 , 05c05 + |
entangled coherent states ( ecss ) in free - traveling fields @xcite have been found to be useful for various applications such as bell inequality tests @xcite , tests for non - local realism @xcite , quantum teleportation @xcite , quantum computation @xcite , precision measurements @xcite , quantum repeater @xcite and quantum key distribution @xcite .
the ecss can be realized in various systems that can be described as harmonic oscillators and numerous schemes for their implementing have been suggested @xcite .
an ecs in a free - traveling field was experimentally generated using the photon subtraction technique on two approximate superpositions of coherent states ( scss ) @xcite .
a proof - of - principle demonstration of quantum teleportation using an ecs as a quantum channel was performed @xcite .
so far , most of the studies on ecss have considered symmetric types of two - mode states where the local amplitudes have the same value .
since ecss are generally sensitive to decoherence due to photon losses in testing bell - type inequalities @xcite , it would be worth investigating the possibility of using an asymmetric type of ecs to reduce decoherence effects .
in fact , asymmetric ecss can be used to efficiently teleport an scs @xcite and to remotely generate symmetric ecss @xcite in a lossy environment .
in addition to the asymmetry of the ecss , as a closely related issue , it would be beneficial to study strategies for distributing the ecss over asymmetrically lossy environments . in this paper
, we investigate asymmetric ecss as well as asymmetric entanglement distribution strategies , and find their evident advantages over symmetric ones for testing the bell - clauser - horne - shimony - holt ( bell - chsh ) inequality @xcite under various conditions .
the remainder of the paper is organized as follows . in sec .
[ sec : bell_test_prefect ] , we discuss the bell - chsh inequality tests using asymmetric ecss with ideal detectors .
we consider photon on - off detection and photon number parity detection , respectively , for the bell - chsh inequality tests .
section [ sec : bell_test_deco ] is devoted to the study of decoherence effects on bell inequality violations with an asymmetric strategy to share ecss .
we then investigate effects of inefficient detectors in sec .
[ sec : ineff ] , and conclude with final remarks in sec .
[ sec : remarks ] .
in this paper , we are interested in a particular form of two - mode ecss @xmath0 where @xmath1 are coherent states of amplitudes @xmath2 for field mode @xmath3 , and @xmath4^{-1/2}$ ] are the normalization factors .
we note that amplitudes @xmath5 and @xmath6 are assumed to be real without loss of generality throughout the paper .
the ecss show noticeable properties as macroscopic entanglement when the amplitudes are sufficiently large and these properties have been extensively explored @xcite .
it is worth noting that there are studies on other types of ecss such as multi - dimensional ecss @xcite , cluster - type ecss @xcite , multi - mode ecss @xcite and generalization of ecss with thermal - state components @xcite while we focus on two - mode ecss in free - traveling fields in this paper .
we call @xmath7 ( @xmath8 ) the even ( odd ) ecs because it contains only even ( odd ) numbers of photons .
when @xmath9 , we call the states in eq .
( [ ecs ] ) symmetric ecss , and otherwise they shall be called asymmetric ecss .
it is straightforward to show that an ecs can be generated by passing an scs in the form of @xmath10 @xcite , where @xmath11 , through a beam splitter . a beam splitter with an appropriate ratio
should be used to generate an asymmetric ecs with desired values of @xmath5 and @xmath6 . in this paper , asymmetric ecss with various values of @xmath5 and @xmath6
are compared for given values of @xmath12 .
the average photon numbers of the ecss are solely dependent on the values of @xmath12 and are simply obtained as @xmath13 where @xmath14 and @xmath15 is the number operator for field mode @xmath3 .
we first investigate a bell - chsh inequality test using photon on - off measurements with the displacement operations .
a bell - chsh inequality test requires parametrized measurement settings of which the outcomes are dichotomized to be either @xmath16 or @xmath17 @xcite . in the simplest example of a two - qubit system , a parametrized rotation about the @xmath18 axis followed by a dichotomized measurement in the @xmath19 direction
is used . in our study , the displacement operator that is known to well approximate the qubit rotation for a coherent - state qubit @xcite is used for parametrization .
the displacement operator can be implemented using a strong coherent field and a beam splitter with a high transmissivity .
together with the displacement operator , the photon on - off measurement is experimentally available using current technology with an avalanche photodetector @xcite in spite of the issue of the detection efficiency .
we shall analyze the effects of the limited detection efficiency in sec .
the photon on - off measurement operator for mode @xmath3 is defined as @xmath20 where @xmath21 $ ] is the displacement operator and @xmath22 denotes the fock state .
the correlation function is defined as the expectation value of the joint measurement @xmath23 and the bell - chsh function is @xmath24 in any local realistic theory , the absolute value of the bell - chsh function is bounded by 2 @xcite .
we calculate an explicit form of the correlation function @xmath25 using eqs .
( [ eq : o ] ) and ( [ e_onoff ] ) , of which the details are presented in in appendix [ sec : corr_func ] .
we then find the bell - chsh function @xmath26 using eq .
( [ bell_onoff ] ) and its absolute maximum values @xmath27 over the displacement variables @xmath28 , @xmath29 , @xmath30 and @xmath31 together with @xmath32 , @xmath5 and @xmath6 .
it requires a numerical multivariable maximization method , and we use the broyden - fletcher - goldfarb - shanno ( bfgs ) algorithm @xcite throughout this paper .
+ the optimized bell - chsh functions against the average photon numbers @xmath32 for @xmath33 are presented in fig .
[ onoff_n ] , while fig .
2 shows how asymmetric the ecss become in order to maximize the bell violations .
the solid curves show the results for symmetric ecss while the dotted curves correspond to general cases ( asymmetric ones ) .
the bell - chsh functions are numerically optimized over all displacement variables and amplitudes under the condition of eq . .
the bell violations occur both for the even and odd ecss regardless of the values of @xmath32 that are consistent with the results in ref .
the odd ecs , @xmath34 , violates the inequality more than the other one , @xmath35 for a given average photon number .
the bell - chsh function for state @xmath34 reaches up to @xmath36 while the maximum bell - chsh function for @xmath35 is @xmath37 .
this is due to the fact that the odd ecs is maximally entangled regardless of the value of @xmath38 @xcite and the on - off measurement can effectively reveal nonlocality of the ecss when the amplitudes are small @xcite .
it is obvious from fig .
[ onoff_n ] that one can increase the amount of violations by using the asymmetric ecss for certain regions of @xmath39 . in the case of state @xmath34 , this difference appears for @xmath40 .
this difference in the optimized bell - chsh function reaches its maximum value @xmath41 for @xmath42 . for this value of @xmath38
, the symmetric ecs gives @xmath43 for @xmath44 while the asymmetric ecs yields @xmath45 for @xmath46 and @xmath47 .
the even ecs also shows a small increase of the violation when an asymmetric ecs is used in place of the symmetric ecs for @xmath48 .
the maximum difference is @xmath49 when @xmath50 . in order to further investigate the advantages of the asymmetric ecs
, we plot the optimized bell - chsh function @xmath27 as a function of @xmath5 and @xmath6 for @xmath51 in fig .
[ onoff_minus_ab ] .
the blue line indicates the point for the maximum bell - chsh function for each value of @xmath38 .
the unsmooth change in the blue line at @xmath52 ( @xmath53 ) results from the numerical optimization process where _ local _ maximum values are compared with changes of related parameters , i.e. , the displacement variables and amplitudes .
we note that such comparisons among local maxima at a number of different parameter regions lead to unsmooth changes in several plots throughout this paper @xcite .
in fact , the blue line splits to two symmetric curves from the point of @xmath53 , according to our numerical calculation ( which is obvious because @xmath5 and @xmath6 are simply interchangeable ) , while we plot only one of the curves for convenience . the results in figs .
[ onoff_n ] and [ onoff_minus_ab ] shows that when the average photon number is relatively large , the asymmetric ecs outperforms the symmetric one in testing the bell - chsh inequality with photon on - off measurements and displacement operations . on the other hand ,
when the average photon number is small , the symmetric ecs gives larger bell violations .
+ we now consider the photon number parity measurements with the displacement operators for both modes . here
we use the displacement operator again in order to approximate a parametrized rotation for a coherent - state qubit in the bell - chsh test .
the displaced parity measurement is given by @xmath54 where @xmath55 denotes each mode and the correlation function is @xmath56 with which the bell - chsh function @xmath57 can be constructed using eq . .
we have obtained and presented an explicit form of the correlation function in appendix [ sec : corr_func ] .
the optimized bell - chsh functions are plotted for varying @xmath39 using parity measurements in fig .
[ nvsb_parity ] .
in contrast to the on - off measurement case , the optimized bell - chsh functions increase monotonically toward cirelson s bound @xcite , @xmath58 , for both even and odd ecss as shown in the figure .
these results are consistent with previous studies @xcite . as explained in ref .
@xcite , the reason that the bell - chsh violation increases monotonically in the case of the parity measurement but not in the case of the on - off measurement can be explained as follows .
when the average photon number of the symmetric ecs is sufficiently large , it can be represented as a maximally entangled two - qubit state in a @xmath59 hilbert space spanned by the even and odd scs basis ( @xmath10 ) and the displacement operator well approximates the qubit rotation .
therefore , in this limit , the violation approaches the maximum value , @xmath58 , with the parity measurement that perfectly discriminates between the even and odd scs .
in contrast , the photon number on - off measurement can not cause bell violations for large amplitudes because the vacuum weight in the state almost vanishes in the limit of @xmath60 @xcite . as implied by the apparent overlaps between the cases of symmetric ecss ( solid curve ) and those of the asymmetric ecss ( dotted ) ,
our numerical investigations show that unlike the case with on - off measurements , the asymmetric ecss do not show any larger bell violations .
figure [ parity_plus_ab ] shows the optimized bell - chsh function of @xmath61 for amplitudes @xmath5 and @xmath6 . the blue line in the figure indicates the states with which the maximum violations are obtained for the same average photon numbers .
the figure shows that the asymmetry of the amplitudes does not improve the amount of violations in the case of the parity measurement .
we conjecture that the symmetric structure of the parity measurement is closely related to this result .
in this section , we consider decoherence effects on bell nonlocality tests using an ecs .
we will study the bell - chsh inequality violation using a general form of ecs and compare it to the result with a symmetric ecs for both on - off and parity measurements . in the presence of photon loss ,
the time evolution of a density operator @xmath62 is described by the master equation as @xcite @xmath63 where @xmath64 is the interaction time , and the lindblad superoperators @xmath65 and @xmath66 are defined as @xmath67 and @xmath68 with a decay rate @xmath69 .
we consider two different strategies , i.e. symmetric ( a ) and asymmetric ( b ) ones , when distributing an ecs over a distance to alice and bob as illustrated in fig .
[ photon_loss ] . in the case of strategy
a , photon loss occurs symmetrically for both modes of the ecs during the decoherence time @xmath64 . on the other hand , in strategy b , an ecs
is first generated in the location of alice and one mode of the ecs is sent to bob at a distance .
in the latter case , photon losses occur only in one of the field modes but the decoherence time becomes @xmath70 .
assuming a zero temperature bath and a decay rate @xmath69 for both cases , a direct calculation of the master equation leads to the states @xmath71 \\ & + \ket{-\sqrt{t}\alpha_1}\bra{-\sqrt{t}\alpha_1}\otimes\ket{-\sqrt{t}\alpha_2}\bra{-\sqrt{t}\alpha_2 } \bigr\ } , { \addtocounter{equation}{1}\tag{\theequation}}\label{rhoa}\\ \rho^\pm_b(\alpha_1,\alpha_2,t ) & = \
\mathcal{n_\pm}^2 \bigl\ { \ket{\alpha_1}\bra{\alpha_1}\otimes\ket{t\alpha_2}\bra{t\alpha_2 } \\ & \pm e^{-2(1-t^2 ) \alpha_2 ^ 2 } \bigl [ \ket{\alpha_1}\bra{-\alpha_1}\otimes\ket{t\alpha_2}\bra{-t\alpha_2 } \\ & + \ket{-\alpha_1}\bra{\alpha_1}\otimes\ket{-t\alpha_2}\bra{t\alpha_2 } \bigr]\\ & + \ket{-\alpha_1}\bra{-\alpha_1}\otimes\ket{-t\alpha_2}\bra{-t\alpha_2 } \bigr\ } { \addtocounter{equation}{1}\tag{\theequation}}\label{rhob}\end{aligned}\ ] ] for strategies a and b , respectively , where @xmath72 . for convenience , we define the normalized time @xmath73 which has the value of zero when @xmath74 and increases to @xmath75 as @xmath64 increases to the infinity .
if the cross terms in eqs . and vanish , the states become classical mixtures of two distinct states and quantum effects generally disappear .
we observe that the cross term in @xmath76 is proportional to @xmath77 and it is proportional to @xmath78 in @xmath79 .
this implies that one may reduce decoherence effects using strategy b by making the amplitude smaller for the mode in which loss occurs ( i.e. , by making the field mode sent to bob in fig .
[ photon_loss ] to have the smaller amplitude ) .
such a strategy was also applied to the tele - amplification protocol @xcite and the distributed generation scheme for ecss @xcite .
+ + we first consider bell - chsh tests with on - off measurements in comparing strategies a and b regarding robustness to the decoherence effects .
an explicit form of the correlation functions calculated using eqs . and can be found appendix [ sec : corr_func ] .
we numerically optimize the corresponding bell - chsh function , @xmath80 , over all displacement variables and amplitudes @xmath5 and @xmath6 for given @xmath81 .
the numerically optimized bell - chsh functions against the normalized time are presented for both @xmath82 and @xmath83 in fig .
[ onoff_t ] where the decrease of violations due to the decoherence effects is apparent .
the optimizing values of amplitudes @xmath5 and @xmath6 are found between @xmath84 and @xmath85 , where relatively smaller values correspond to large values of @xmath81 .
we observe that strategy b leads to larger violations than strategy a for @xmath86 when we use the even ecs @xmath87 .
for the smaller value of @xmath81 , strategy a gives slightly larger violations where the difference is up to @xmath88 . on the other hands ,
the odd ecs @xmath89 shows larger violations for strategy b than strategy a regardless of the value of @xmath81 . as explained earlier in sec .
iia , the discontinuities of the first derivative at @xmath90 for @xmath91 and @xmath92 for @xmath93 in fig .
[ onoff_t](a ) emerge from the numerical optimization process where local maxima for different parameter regions are compared .
+ the bell violations of the odd ecss for varying @xmath5 and @xmath6 for @xmath94 , numerically optimized for the displacement variables , are shown in fig .
[ onoff_ab ] .
figure [ onoff_ab](a ) clearly shows the asymmetric behavior of the optimized bell - chsh function under strategy a. figure [ onoff_ab](b ) is for the case of asymmetric decoherence ( strategy b ) . for a given average photon number ,
the optimizing value of @xmath6 in strategy b is smaller than that of strategy a. it shows that an asymmetric ecs can reduce the asymmetric decoherence effects by lessening the amplitude of the mode which decoherence occurs and adjusting the other mode which is free from decoherence .
these results are consistent with previous studies for bell inequality tests using hybrid entanglement @xcite and quantum teleportation @xcite , where the amplitude of the mode that suffers decoherence should be kept small in order to optimize bell violations or teleportation fidelities .
it is important to note from fig .
[ onoff_ab ] that strategy b shows significantly larger violation than that of strategy a. in strategy a ( fig .
[ onoff_ab](a ) ) , the maximum value of the bell - chsh function is @xmath95 for @xmath96 and @xmath97 . on the other hands ,
the maximum value for strategy b ( fig . [ onoff_ab](b ) ) is @xmath98 for @xmath99 and @xmath100 . noting that @xmath101 is the maximum value of the bell function by a local realistic theory , the absolute value of the bell function with strategy b , @xmath102 , is significantly larger than that of strategy a , @xmath103 .
it is a remarkable advantage of using asymmetric ecss with the asymmetric distribution scheme in testing the bell - chsh inequality .
+ figure [ parity_t ] presents the numerically optimized bell - chsh function @xmath104 with parity measurements against the normalized time @xmath81 ( see appendix [ sec : corr_func ] for details ) . for both the even and odd ecss , strategy b give significantly larger violations over strategy a. the optimizing values of @xmath5 and @xmath6 approach infinity for @xmath105
this is because , when we use the photon number parity measurements and the photon loss is absence in one mode , the optimized bell - chsh functions increase monotonically with repect to the amplitude of that mode .
however , we are interested in values that can be obtained within a realistic range ( e.g. @xmath106 ) .
we thus plot in fig .
[ parity_ab ] the optimization results for the even ecss under a normalized decoherence time ( @xmath107 ) for varying @xmath5 and @xmath6 .
in contrast to the case of on - off measurements , as shown in fig .
[ parity_ab](a ) , the asymmetry of the amplitudes of the ecs does not improve the amount of bell violations for strategy a where decoherence occurs symmetrically .
similar to the results of the on - off measurements , the asymmetry of the amplitudes in ecss increases the amount of bell violation for strategy b ( fig . [ parity_ab](b ) ) . in this case , the maximum value of optimized bell - chsh function is @xmath108 for @xmath109 and @xmath110 .
a similar value @xmath111 can be obtained for @xmath112 and @xmath113 where the average photon number is given by @xmath114 .
this value is much larger than that of strategy a , which only shows @xmath115 for @xmath116 .
here we investigate the effects of detection inefficiency in the bell - chsh inequality tests .
an imperfect photodectector with efficiency @xmath117 can be modeled using a perfect photodetector together with a beam splitter of transmissivity @xmath118 in front of it @xcite
. meanwhile , decoherence by photon loss ( eqs . and ) in the entangled state can also be modeled using a beam splitter .
the only difference between the effect of detection inefficiency and that of photon loss in the entangled state is the order of photon loss and displacement as shown in fig .
[ scheme1 ] .
we compare these two cases in appendix [ sec : equiv ] where the results show that the two cases lead the same bell - chsh function except for the coefficients of the displacement variables which disappear during the optimization process .
we have numerically investigated both cases of on - off and parity measurements with limited efficiencies , @xmath119 for mode 1 and @xmath120 for mode 2 , for both even and odd ecss . as we find that the even ecs is better for on - off measurements and the odd ecs is better for parity measurements , as already implied in figs .
6 and 8 , we present the two corresponding cases in figure [ inefficient ] .
it shows that the on - off measurement scheme generally gives much larger violations compared to the parity measurement scheme .
this is consistent with the results of the case with photon loss effects studied in the previous section ( see figs .
6(b ) and 8(a ) ) . in detail , for the case of the displaced on - off measurements with the odd ecs ( figs .
[ inefficient](a ) and [ inefficient](b ) ) , optimizing values of @xmath5 and @xmath6 lie between @xmath84 to @xmath85 , which is experimentally feasible @xcite .
we find that the asymmetric ecs ( fig [ inefficient](b ) ) shows larger violations compared to the symmetric ecs ( fig [ inefficient](a ) ) .
when @xmath121 , a detection efficiency of @xmath122 is required for violation of @xmath123 with the symmetric ecs while a smaller efficiency @xmath124 is sufficient for the same amount of violation with the asymmetric ecs . when the efficiencies for modes 1 and 2 are @xmath125 and @xmath126 , the asymmetric ecs shows the optimized bell quantity of @xmath127 which is larger than @xmath128 for the symmetric ecs . in the case of the parity measurements using the even ecss ,
the improvement by the asymmetric ecs is even larger .
for example , comparing @xmath129 line in fig .
[ inefficient](c ) and ( d ) , we find that there is a violation @xmath130 for @xmath131 if we use the asymmetric form of ecs .
however , much larger efficiency @xmath132 is needed when we use the symmetric ecs .
the optimizing values of @xmath5 and @xmath6 lie between @xmath133 to @xmath134 if @xmath135 and @xmath136 .
we also note that if one of the detectors is perfect , the optimizing value of the amplitude of that mode goes to infinity .
this behavior can be inferred from the results in sec .
[ subsec : parity_loss ] with fig .
9(b ) for the case of decoherence , noting that the decoherence and the detection inefficiency give qualitatively the same effects ( appendix b ) .
when both the detectors are perfect , of course , the larger amplitude gives the larger violation for each mode as implied in fig .
4 ; the maximum violation appears when both @xmath5 and @xmath6 become infinity .
in this paper , we have studied asymmetric ecss as well as asymmetric lossy environments for bell - chsh inequality tests .
we first investigated the violations of bell - chsh inequality using perfect on - off detectors and ideal ecss .
we have shown that the asymmetric form of ecs could give larger violations of the bell - chsh inequality in some region of the averaged photon numbers of the ecss .
on the other hand , in the case of photon number parity measurements , we could not find apparent improvement of the bell violations using the asymmetric form of ecs with perfect detectors .
we then studied bell - chsh violations under the effects of decoherence on the ecss .
we considered two different schemes for distributing entanglement under photon loss , i.e. , symmetric and asymmetric schemes . in the symmetric scheme ,
the photon losses occur in both modes of the ecss . on the other hands ,
photon loss occurs only in one of the two modes in the case of the asymmetric scheme .
we showed that the asymmetric form of ecs can increase the amount of violations significantly under the asymmetric scheme compared to the case of the symmetric scheme .
for example , when the normalized time is @xmath94 , the bell - chsh function using photon on - off detectors for the asymmetric loss scheme shows the maximum value of @xmath102 , which is much larger than that for the symmetric case , @xmath103 .
a similar improvement can be made in the case of photon number parity measurements ; when @xmath107 with the ecs of average photon number @xmath137 , the optimized bell - chsh function under the symmetric loss scheme is about @xmath138 but it is @xmath139 under the asymmetric scheme . we have also investigated effects of inefficient detectors .
we show that the asymmetric form of ecss lowers the detection efficiency required for violation of the bell - chsh inequality .
for example , a detection efficiency of @xmath122 is required for a bell - chsh violation of @xmath123 with the symmetric ecs , but a smaller efficiency @xmath124 is sufficient for the same amount of violation with the asymmetric ecs .
this improvement is even more significant for the case of parity measurements particularly when the detection efficiencies of two detectors differ much .
when the detection efficiency for mode 1 is @xmath129 , the required detection efficiency for mode 2 is @xmath140 to show the violation of @xmath141 using the symmetric ecs is used , but it is @xmath142 when the asymmetric ecs is used .
the optimizing amplitudes are found between @xmath84 to @xmath85 in the case of the on - off measurements and between @xmath133 to @xmath134 for the parity measurements .
it is worth noting that these values of amplitudes for ecss are within reach of current technology @xcite . in summary
, our extensive study reveals that the asymmetric form of ecss and the asymmetric scheme for distributing entanglement enable one to effectively test the bell - chsh inequality with the same resources .
this work was supported by the national research foundation of korea ( nrf ) grant funded by the korea government ( msip ) ( no .
2010 - 0018295 ) and by the center for theoretical physics at seoul national university .
the numerical calculations in this work were performed using chundoong cluster system in the center for manycore programming at seoul national university .
here , we present the explicit forms of the correlation functions defined in eqs . and under lossy effects . instead of computing the correlation function for every case
, we calculate the results for an ecs under the beam splitters with transmissivities @xmath119 and @xmath120 for mode 1 and mode 2 , respectively , and show that these results are applicable to all the cases discussed in this paper .
the beam splitter operator with transmissivity @xmath118 between modes @xmath143 and @xmath144 can be represented by @xmath145 $ ] @xcite .
if a coherent - state dyadic @xmath146 for mode @xmath147 is mixed with the vacuum for mode @xmath148 , the initial coherent - state dyadic becomes @xmath149 \\ & = \exp\bigl[-\frac{1}{2}(1-\eta)({\left|\alpha\right|}^2+{\left|\beta\right|}^2 - 2\alpha\beta^*)\bigr]\ket{\sqrt{\eta}\alpha}\bra{\sqrt{\eta}\beta}.\label{dkb}{\addtocounter{equation}{1}\tag{\theequation}}\end{aligned}\ ] ] we now apply this result to a pure ecs @xmath150 mixed with the vacuum modes through two beam splitters with transmissivities @xmath119 and @xmath120 , and the result is @xmath151\\ & = \mathcal{n_\pm}^2 \bigl\ { \ket{\sqrt{\eta_1}\alpha_1}\bra{\sqrt{\eta_1}\alpha_1}\otimes\ket{\sqrt{\eta_2}\alpha_2}\bra{\sqrt{\eta_2}\alpha_2 } \\ & \pm e^{-2[(1-\eta_1)\alpha_1 ^ 2+(1-\eta_2)\alpha_2 ^ 2]}&\\ & \bigl[\ket{\sqrt{\eta_1}\alpha_1}\bra{-\sqrt{\eta_1}\alpha_1}\otimes\ket{\sqrt{\eta_2}\alpha_2}\bra{-\sqrt{\eta_2}\alpha_2 } \\ & + \ket{-\sqrt{\eta_1}\alpha_1}\bra{\sqrt{\eta_1}\alpha_1}\otimes\ket{-\sqrt{\eta_2}\alpha_2}\bra{\sqrt{\eta_2}\alpha_2 } \bigr ] \\ & + \ket{-\sqrt{\eta_1}\alpha_1}\bra{-\sqrt{\eta_1}\alpha_1}\otimes\ket{-\sqrt{\eta_2}\alpha_2}\bra{-\sqrt{\eta_2}\alpha_2 } \bigr\}.\end{aligned}\ ] ] it is straightforward to check that if we let @xmath152 then the state will be exactly the same to eq . .
in addition , @xmath153 with @xmath154 will give eq . .
using the result , the correlation functions for on - off measurement and parity measurement are obtained as @xmath155\\ & = \frac{2}{2 \pm 2 e^{-2 \alpha_1 ^ 2 - 2 \alpha_2 ^ 2 } } \bigl[1\mp2 e^{-2 \alpha_1 ^ 2 - 2 \alpha_2 ^ 2-\chi_i^2-\chi_r^2-\xi_i^2-\xi_r^2 } \bigl(e^{\alpha_1 ^ 2 \eta_1 + \chi_i^2 + \chi_r^2 } \cos \bigl(2 \alpha_1 \sqrt{\eta_1 } \xi_i\bigr ) \\ & + e^{\alpha_2 ^ 2 \eta_2+\xi_i^2+\xi_r^2 } \cos \bigl(2 \alpha_2 \chi_i \sqrt{\eta_2}\bigr ) \bigr ) -2 e^{\alpha_1 ^ 2 \eta_1 + \alpha_2 ^ 2 \eta_2 } \cos \bigl(2 \bigl(\alpha_1 \sqrt{\eta_1 } \xi_i+ \alpha_2 \sqrt{\eta_2}\chi_i \bigr)\bigr)\\ & \pm e^{-2 \bigl(\alpha_1 ^ 2+\alpha_2 ^ 2\bigr ) } + 2 e^{-\bigl(\xi_r-\alpha_1 \sqrt{\eta_1}\bigr)^2-\bigl(\chi_r-\alpha_2 \sqrt{\eta_2}\bigr)^2-\chi_i^2-\xi_i^2 } + 2 e^{-\bigl(\alpha_1 \sqrt{\eta_1}+\xi_r\bigr)^2-\bigl(\alpha_2 \sqrt{\eta_2}+\chi_r\bigr)^2-\chi_i^2-\xi_i^2}\\ & -e^{-\bigl(\xi_r - \alpha_1 \sqrt{\eta_1}\bigr)^2-\xi_i^2}-e^{-\bigl(\alpha_1 \sqrt{\eta_1}+\xi_r\bigr)^2-\xi_i^2 } -e^{-\bigl(\alpha_2 \sqrt{\eta_2}+\chi_r\bigr)^2-\chi_i^2}-e^{-\bigl(\chi_r-\alpha_2 \sqrt{\eta_2}\bigr)^2-\chi_i^2}\bigr ] , \\
e_\pi(\xi,\chi ) & = \operatorname{tr}[\rho^\pm \hat{\pi}_1(\xi)\otimes\hat{\pi}_2(\chi)]\\ & = \frac{1}{2\pm 2e^{-2 \alpha_1 ^ 2 - 2 \alpha_2 ^ 2}}\exp \left(-2 \left(\alpha_1 ^ 2 ( \eta_1 + 1 ) + \alpha_2 ^ 2 ( \eta_2 + 1)+\chi_i^2+\chi_r^2+\xi_i^2+\xi_r^2\right)\right)\\ & \left(\pm2e^{4 \left(\alpha_1 ^ 2 \eta_1 + \alpha_2 ^ 2 \eta_2\right ) } \cos \left(4 \alpha_1 \sqrt{\eta_1 } \xi_i+4 \alpha_2 \chi_i \sqrt{\eta_2}\right)+e^{2 \left(\alpha_1 ^ 2 -2\alpha_1\xi_r\sqrt{\eta_1 } + \alpha_2 ^ 2 -2 \alpha_2 \chi_r \sqrt{\eta_2}\right ) } \left(e^{8 \alpha_1 \sqrt{\eta_1 } \xi_r+8 \alpha_2 \chi_r \sqrt{\eta_2}}+1\right)\right),\end{aligned}\ ] ] where @xmath156 and @xmath157 , and @xmath158 and @xmath159 were defined in eqs . and .
by substituting @xmath160 , we can get a correlation functions for the case of perfect detectors .
likewise , @xmath152 gives the correlation functions for distribution strategy a and @xmath153 , @xmath154 gives the correlation function for strategy b. as we will discuss below , we also can use these results for the cases of imperfect detectors even though they differ in the order of the displacement operators and the beam splitters .
in this section , we show that ecss give the same optimized value of the bell - chsh function independent to the order of the displacement and the photon loss by a beam splitter .
note that it is sufficient to only consider two coherent - state dyadics of the form @xmath161 and @xmath162 .
this is because ecss only contain this kind of dyadics and a photon loss by a beam splitter is superlinear(linear in the space of density matrices ) and a displacement operator is linear .
first , suppose that the states undergo a beam splitter first and displacement second . from the above , the dyadics after the transmission to the beam splitter with transmissivity @xmath118 would be @xmath163\ket{\sqrt{\eta}\gamma}\bra{-\sqrt{\eta}\gamma}.\end{aligned}\ ] ]
now we apply the displacement operator @xmath164 $ ] to photon lost dyadics .
then the states would be @xmath165\\ & \quad\ket{\sqrt{\eta}\gamma-\xi}\bra{-\sqrt{\eta}\gamma-\xi}.{\addtocounter{equation}{1}\tag{\theequation}}\label{disloss}\end{aligned}\ ] ] next , let us consider the case we employ the operators to opposite order , displacement first and photon loss second .
after the displacement , the dyadics would be @xmath166\ket{\gamma-\xi}\bra{-\gamma-\xi}.\end{aligned}\ ] ] applying photon loss to these dyadics using eq .
, we finally obtain @xmath167\\ & \quad\ket{\sqrt{\eta}(\gamma-\xi)}\bra{\sqrt{\eta}(-\gamma-\xi)}.{\addtocounter{equation}{1}\tag{\theequation}}\label{lossdis}\end{aligned}\ ] ] these results show that if we replace @xmath168 with @xmath169 in eq . , it will be exactly the same to eq . .
this means that the optimized values of the bell - chsh functions for all displacement variables do not depend on the order of loss and displacement .
a. mann , b. c. sanders , and w. j. munro , phys .
a * 51 * , 989 ( 1995 ) .
r. filip , j. rehcek , and m. dusek , j. opt .
b : quantum semiclassical opt . * 3 * , 341 ( 2001 ) .
d. wilson , h. jeong , and m. s. kim , j. mod . opt . * 49 * , 851 ( 2002 ) .
h. jeong , w. son , m. s. kim , d. ahn , and .
brukner , phys .
a * 67 * , 012106 ( 2003 ) .
j. wenger , m. hafezi , f. grosshans , r. tualle - brouri , and p. grangier , phys .
a * 67 * , 012105 ( 2003 ) .
h. jeong and t. c. ralph , phys .
* 97 * , 100401 ( 2006 ) .
m. stobinska , h. jeong and t. c. ralph , phys .
a * 75 * , 052105 ( 2007 ) .
h. jeong , phys .
a * 78 * , 042101 ( 2008 ) .
c .- w . lee and h. jeong , phys .
a * 80 * , 052105 ( 2009 ) . c. c. gerry , j. mimih , and a. benmoussa , phys .
a * 80 * , 022111 ( 2009 ) .
h. jeong , m. paternostro , and t. c. ralph , phys .
lett . * 102 * , 060403 ( 2009 ) .
y. lim , m. paternostro , m. kang , j. lee , and h. jeong , phys .
a * 85 * , 062112 ( 2012 ) .
kirby and j.d .
franson , phys .
a * 87 * , 053822 ( 2013 ) .
kirby and j.d .
franson , phys .
rev . a * 89 * , 033861 ( 2014 ) .
x. wang , phys .
a * 64 * , 022302 ( 2001 ) .
s. j. van enk and o. hirota , phys .
a * 64 * , 022313 ( 2001 ) .
h. jeong , m. s. kim , and j. lee , phys .
a * 64 * , 052308 ( 2001 ) .
h. jeong and m. s. kim , quantum info . comp .
* 2 * , 208 ( 2002 ) .
n. b. an , phys .
a * 68 * , 022321 ( 2003 ) .
p. t. cochrane , g. j. milburn , w. j. munro , phys .
a * 59 * , 2631 ( 1999 ) . h. jeong and m. s. kim , phys .
a * 65 * , 042305 ( 2002 ) .
t. c. ralph , a. gilchrist , g. j. milburn , w. j. munro , and s. glancy , phys .
a * 68 * , 042319 ( 2003 ) .
a. p. lund , t. c. ralph , and h. l. haselgrove , phys .
lett . * 100 * , 030503 ( 2008 ) .
p. marek and j. fiur , phys .
a * 82 * , 014304 ( 2010 ) . c. r. myers and t. c. ralph , new .
j. phys * 13 * 115015 ( 2011 ) .
c. c. gerry and r. a. campos , phys .
a * 64 * , 063814 ( 2001 ) . c. c. gerry , a. benmoussa , and r. a. campos , phys .
a * 66 * , 013804 ( 2002 ) .
w. j. munro , k. nemoto , g. j. milburn , and s. l. braunstein , phys .
a * 66 * , 023819 ( 2002 ) .
r. a. campos , c. c. gerry , and a. benmoussa , phys .
a * 68 * , 023810 ( 2003 ) .
j. joo , w. j. munro , and t. p. spiller , phys .
rev . lett . * 107 * , 083601 ( 2011 ) .
o. hirota , k. kato , and d. murakami , arxiv:1108.1517 ( 2011 ) .
j. joo , k. park , h. jeong , w. j. munro , k. nemoto , and t. p. spiller , phys .
a * 86 * , 043828 ( 2012 )
. y. m. zhang , x. w. li , w. yang , and g. r. jin , phys .
a * 88 * , 043832 ( 2013 ) .
b. yurke and d. stoler , phys .
a * 35 * , 4846 ( 1987 ) .
p. tombesi and a. mecozzi , j. opt .
b * 4 * , 1700 ( 1987 ) .
c. c. gerry , phys .
a * 55 * , 2478 ( 1997 ) b. c. sanders and d. a. rice , opt .
electronics * 31 * , 525 ( 1999 ) .
b. c. sanders and d. a. rice , phys . rev .
a * 61 * , 013805 ( 1999 ) .
s. takeuchi , j. kim , y. yamamoto , h. h. hogue , appl .
. lett . * 74 * , 1063 ( 1999 ) .
r. fletcher , _ practical methods of optimization ( second edition ) _ ( wiley , 1987 ) .
o. hirota and m. sasaki , arxiv : quant - ph/0101018 ( 2001 ) .
one may ask whether this unsmooth change implies that our results around @xmath170 are not the global maxima .
we can not exclude such a possibility because we do not have an analytical evidence to claim that all the results obtained using the bfgs algorithm @xcite throughout the paper are the global maxima .
this is a typical problem in an optimization process with multiple parameters as we do in the present paper with 8 to 10 parameters ( assuming @xmath5 and @xmath6 are real without loss of generality ) .
b. s. cirelson , lett .
phys . * 4 * , 93 ( 1980 ) . | we study an asymmetric form of two - mode entangled coherent state ( ecs ) , where the two local amplitudes have different values , for testing the bell - clauser - horne - shimony - holt ( bell - chsh ) inequality .
we find that the asymmetric ecss have obvious advantages over the symmetric form of ecss in testing the bell - chsh inequality .
we further study an asymmetric strategy in distributing an ecs over a lossy environment and find that such a scheme can significantly increase violation of the inequality . |
according to the most promising cosmological scenarios , structure formation is traced back to the evolution of primordial density fluctuations .
these fluctuations , originated from quantum fluctuations@xmath7@xmath8@xmath9@xmath10@xmath11@xmath8@xmath12 in an inflationary phase of the early universe , grew up through gravitational instability to a maximum radius @xmath13 . at the time @xmath14 corresponding to maximum expansion , perturbations broke away from the general expansion and at @xmath15 them began to collapse .
hence the collapse of perturbations onto local density maxima of the primordial density field has a key role in structure formation and several studies deal with this problem .
the problem of collapse has been investigated from two points of view , namely : \i ) that of the statistical distribution of the objects formed@xmath12@xmath8@xmath16 \ii ) that of the structure of these objects and its dependence on the statistical properties of the primordial density field@xmath17@xmath8@xmath18@xmath10@xmath19@xmath8@xmath20 @xmath21@xmath8@xmath22@xmath10@xmath23@xmath8@xmath24 @xmath25@xmath8@xmath26@xmath10@xmath27@xmath8@xmath28 @xmath29@xmath8@xmath30@xmath10@xmath31@xmath8@xmath32 @xmath33 . in the spherical accretion model introduced by hoyle and narlikar@xmath34 and applied to clusters of galaxies by gunn & gott@xmath35 , the matter around the core of the perturbation is a homogeneous fluid with zero pressure . if the density inside the perturbation is greater than the critical density , it is bound and shall expand to a maximum radius @xmath36 : @xmath37 where @xmath38 is the initial radius and @xmath39 is the overdensity inside the radius @xmath40 .
such matter shall collapse in a time @xmath41 where @xmath42 is the hubble parameter at the initial time @xmath43 .
+ this model was introduced in order to overcome the problem of the excessively steep density profiles , @xmath44 , obtained in numerical experiments of simple gravitational collapse .
some authors were able to produce shallower profiles@xmath17@xmath8@xmath45@xmath8@xmath46 , @xmath47 , through the @xmath48 @xmath49 process .
several years later , observational evidences for secondary infall in the outskirts of clusters of galaxies has been reported in ref .
+ even if this model and in general the original form of sim ( secondary infall model ) is able to explain better than previous models the structure of clusters of galaxies it has some drawbacks , ( e.g. it does not predict the right structure of density profiles ) , that can be overcome with some improvements in the original model@xmath50 .
two noteworthy limits by the gunn & gott@xmath35 model are the following : _ a _ ) it neglects tidal interaction of the perturbation with the neighboring perturbation ; _ b _ ) it neglects substructure existing in the outskirts of proto - galaxies or proto - clusters and in the background . the effect of the tidal field was studied in a recent paper@xmath51 . in that paper , we showed that the gravitational interaction of the quadrupole momentum of a proto - structure with the tidal field of the neighboring proto - structures introduces another potential energy term in the equation describing the collapse which acts in the sense to delay the collapse .
this effect has revealed as one of noteworthy importance in solving several of the drawbacks of the cdm model@xmath51@xmath8@xmath52@xmath8@xmath53@xmath8@xmath54 , namely : \1 ) lack of a mechanism originating the bias@xmath51 ; \2 ) discrepancies between : * a1 ) x - ray temperature function of clusters and observations@xmath52 . * a2 ) two - point correlation function of clusters and observations@xmath52 ; * a3 ) mass function and velocity dispersion function of clusters and observations@xmath54 ; * a4 ) shape of clusters and observations@xmath55 ; * a5 ) angular two - point correlation function of galaxies and observations@xmath53 . for
what concerns point `` @xmath56 '' , we may say that during the last two decades , astronomers have discovered that at least a third of galaxy clusters are not dynamically relaxed systems but contain substructure on scales of the same order as the cluster itself@xmath55@xmath8@xmath57@xmath8@xmath58 .
this implies that clusters are currently forming or have formed recently enough that they have not had time to undergo significant degree of violent relaxation and phase mixing .
when one studies the structure of the velocity field around a peak one finds that at radii @xmath40 larger than the filtering scale @xmath59 , ( @xmath0 ) , a sensitive fraction of the matter of the peak is gravitationally unbound to it@xmath60 .
this is not surprising since a non zero fraction of unbound matter should be expected even in the case of a uniform system with a maxwellian velocity distribution ( see section 2.2 for a discussion ) .
the encounters between collapsing clumps and the unbound high velocity clumps influence the collapse of the proto - cluster .
we want to study this point and to compare its effect on cluster collapse with that produced by the effects of tidal interaction of the protocluster with the neighboring ones .
+ in the following we show how the unbound substructures influences the collapse of the perturbation .
the plan of this work is the following : in section 2 we deal with the unbound substructure effects . in section
3 we give our results and finally section 4 is devoted to the conclusions .
our cosmological setting is described by a hierarchical clustering scenario , a cdm model ( see ref .
38 ) in which structure formation goes _ bottom - up_. local density extrema are assumed to be the sites of structure formation@xmath61@xmath8@xmath62 .
the perturbations grows linearly until the linearly evolved dispersion @xmath63 of the density contrast is @xmath64 , while it enters in the non - linear phase when @xmath65 .
denoting this epoch by @xmath66 one has : @xmath67 ( liddle & lyth@xmath68 , equation 8.2 ) , where as always the subscript 0 denotes the present value of the linearly evolved quantity and @xmath69 is related to the power spectrum @xmath70 of density fluctuations by : @xmath71 being @xmath72 the filtering scale and @xmath73 the window function : @xmath74}{(k r_{\rm f})^{3}}\ ] ] a subgalactic mass scale collapses at a redshift : @xmath75 ( silk & stebbins@xmath76 ) , where @xmath77 is the biasing parameter .
in particular a perturbation of @xmath78 collapses at a redshift @xmath79 while perturbations in the range @xmath80 collapse almost at the same @xmath81 because the mass variance in this region varies only of a factor 3 ( ref .
a part of this last perturbations have a cross section too little for gravitational merging to be destroyed@xmath82 . therefore a mass perturbation @xmath83 ( see ref .
38 ) can survive until the cluster enters non - linear phase at @xmath84 .
+ in order to study the dynamical evolution of the proto - cluster , we use the shell method@xmath51@xmath8@xmath85@xmath8@xmath86@xmath8@xmath87 . if we consider a shell of matter around the centre of the cluster this is made of galaxies , gas and substructure . during the collapse ,
the shell encounters clouds of substructure some of which have high speed .
supposing that the systems subject to the encounter ( the shell of matter and the high speed clumps ) have , respectively , median radii @xmath88 and @xmath89 , velocity dispersions of order @xmath90 , @xmath91 , that at the instant of closest approach their centres are separated by @xmath92 and have relative speed v the condition@xmath93 : @xmath94 ensures that in the course of the encounter the objects constituting the encountering systems barely move from their initial locations .
the result of the encounter is to generate perturbations in the velocity and energy of the objects constituting the shell ( here we are only interested in the effect of the encounter on the shell ) .
in fact the effects of high speed encounters on the internal structure of `` stellar '' systems decrease as the encounter speed increases and consequently high speed encounters generate only small perturbations of otherwise steady - state systems .
+ the conditions expressed in equation ( [ eq : impuls ] ) is known as impulse approximation ( hereafter ia)@xmath95 and yields accurate results often even when the condition is not strictly satisfied . as stressed by gnedin & ostriker@xmath86 ,
the ia works well in the outer region of clusters , wherein galaxies are assumed to move little during the interaction , while the ia is not appropriate in the cluster core . before going on ,
we want to analyze the applicability of the ia as expressed by equation ( [ eq : impuls ] ) to our problem .
a first rough analysis can be performed remembering that the escape velocity in the cluster is fixed by the virial theorem according to equation ( see for example ref .
44 , ( equation ( 8.3 ) ) ) : @xmath96 where @xmath97 is the escape velocity and @xmath98 is the 3-d dispersion velocity , and @xmath99 being @xmath90 the velocity dispersion of the objects constituting the shell , and then @xmath100 the relative velocity , v , is the sum of the infall velocity of the shell and the velocity of clumps moving outwards : @xmath101 imposing that the clumps move at speed larger than @xmath97 , we get : @xmath102 the infalling shells are marginally bound , hence they travel at a velocity close to the local escape velocity @xmath103 this last statement can be easily proven using for example an hernquist model for the cluster . in general , conservation of energy yields : @xmath104 where @xmath105 is the ( negative ) potential and @xmath106 is the maximum ( apocentric ) radius of the infalling particle .
for example , for a hernquist cluster , one has @xmath107 and @xmath108 = v_{\rm esc}^2(r ) [ 1 - ( r + a ) / ( r_a + a ) ] = v_{\rm esc}^2(r ) ( r_a - r ) / ( r_a + a)\ ] ] for a hernquist model , @xmath109 is a few tenths of the virial radius .
the infalling particles are between the virial and turnaround radii , i.e. between @xmath110 and roughly 2 @xmath110 .
so @xmath111 and @xmath112 which means that @xmath113 is a little smaller than @xmath97 . using the previous results ( equation ( [ eq : vee])and ( [ eq : vee1 ] ) )
, we have : @xmath114 if @xmath115 , in eq . ( [ eq : impuls ] ) , we are dealing with the shell then : + 1 ) @xmath116 + if @xmath117 , we are dealing with the clump : @xmath118 ( see ref .
39 ) and then : + 2 ) @xmath119 . + both the relations 1 ) and 2 ) satisfy the condition : @xmath120 value that according to ref .
44 ( see page 440 ) is considered good to use the ia .
+ we want to remark that the previous analysis does not take account of the reduction of escape velocity with radius : given the potential of the sistem , @xmath121 , the escape velocity is @xmath122 .
this means that going outwards the ia is more strictly satisfied .
the energy change in the shell , due to an encounter with high velocity clumps , can be computed as follows .
+ suppose to have a perturber ( a high speed clump of substructure , member of the cluster ) characterized by a speed larger than the escape speed from the cluster .
this high velocity clump , that we stress is inside and part of the cluster , moves away from it .
we use cylindrical coordinates ( r , z ) such that the z - axis coincides with the perturber s trajectory given by : @xmath123 and that @xmath124 is the centre of the perturbed system ( shell of matter ) . note that we are interested only in clumps not contained in the core and moving from the centre in the outward direction , so that they do not go through the cluster centre .
the choice of not taking into account the effect of clumps in the core area is due to gnedin & ostriker@xmath86 observation that the ia works well in the outer region of clusters , wherein galaxies are assumed to move little during the interaction .
the velocity increment @xmath125 of a test particle at the position ( r , z ) is given by : @xmath126 where @xmath127^{1/2}\ ] ] and @xmath128 is the potential of the perturber .
assuming that the perturber is an isothermal sphere it can be easily shown that : @xmath129 where @xmath91 is the perturber internal velocity dispersion .
+ the energy , gained by the shell of matter during the encounter with a high velocity clump , is given by : @xmath130 ^ 2 \sigma(r ) r dr \label{eq : ene}\ ] ] ( see ref .
44 ) where @xmath131 is the surface density of the perturbed system .
we assume that the perturbed system ( shell ) is characterized by an average density given by : @xmath132 where , according to ref
. 2 notation , @xmath133 is the expansion factor of the shell at time @xmath134 , @xmath38 its initial radius and @xmath135 its mass .
we are studying the collapse of a shell of matter then @xmath136 and @xmath38 coincide with the turn - around epoch and radius of the shell , respectively . assuming that no shell crossing occurs , the total mass inside each shell stays constant and then : @xmath137 the surface density is given by : @xmath138 at this point we are ready to calculate the energy change in the shell by using equations ( [ eq : iso ] ) , ( [ eq : ene ] ) , ( [ eq : surf ] ) and remembering that @xmath139 : @xmath140 where @xmath141 is the maximum value of the expansion parameter , corresponding to the turn - around epoch , and @xmath142 the minimum value of the expansion parameter corresponding to the time of full collapse of the perturbation . during infall a shell of matter experiences several encounters with all the unbound clumps located between its initial position and the final one .
each one of the encounters produces an increase of the energy of the shell . in order to calculate the total energy , @xmath143 , acquired during infall
we must calculate the number of unbound clumps `` internal '' to its orbit .
the number of perturbers can be calculated observing that : aa ) they are peaks of the density field having dimension much less than that of the cluster but large enough to survive till the cluster enters the non - linear phase ; bb ) they must satisfy the ia condition .
clumps in the scdm scenario are not distributed randomly but are clustered and so tend to be more bound to their nearby neighbors .
this may make them more subject to mergers in some cases soon after they form , but may also prevent mergers by placing them in a high velocity dispersion environment . for survival , we must require a large density contrast to have developed .
if the smaller scale fluctuations have a density contrast of @xmath144 , they should be able to survive tidal disruption or dynamical friction drag when its environment undergoes non - linear collapse .
this translates into a minimum mass scale of @xmath145 ( see ref .
39 ) . in order to determine the number density and the characteristics of these peaks we follow the same choice made by antonuccio - delogu & colafrancesco@xmath93 and del popolo & gambera@xmath87 .
we restrict ourselves to consider a subset of peaks of the density field , having the central height @xmath5 larger than a threshold @xmath146 .
the threshold is chosen by satisfying the condition its radius @xmath147 is smaller than the average peak separation , @xmath148 : @xmath149 where the radius of a peak @xmath150 is given by : @xmath151^{\frac{1}{3}}\ ] ] ( see ref .
45 ) , where : @xmath152 and here @xmath153 is the i - th momentum of the variance of the fluctuation field : @xmath154 where @xmath155 is the filtered power spectrum .
the power spectrum that we adopt is @xmath156 with the transfer function @xmath157 given in ref .
7 ( equation ( g3 ) ) : @xmath158}{2.34 q } \cdot [ 1 + 3.89q+ ( 16.1 q)^2+(5.46 q)^3+(6.71)^4]^{-1/4 } % % t^2(k ) & = & [ \ln \left ( 1 + 4.164k\right)]^2 \cdot ( 192.9 + 1340k+ \nonumber \\ % & + & 1.599\cdot 10 ^ 5k^2 + 1.78\cdot 10 ^ 5k^3 + 3.995\cdot % 10 ^ 6k^4)^{-1/2 } % \label{eq : ma5}\ ] ] where @xmath159 is the normalizing constant and @xmath160 . here
@xmath161 represents the ratio of the energy density in relativistic particles to that in photons ( @xmath162 corresponds to photons and three flavors of relativistic neutrinos ) . for the previous cdm transfer function , normalized as @xmath163 and filtering radius @xmath164 , ( corresponding to a gaussian filtered mass @xmath165 )
one has @xmath166 , @xmath167 and then @xmath168 . remembering that the number density of peaks , @xmath169 , having @xmath170 , is given by : @xmath171 where @xmath172 being @xmath173 defined in ref . 7
( equation ( 4.4 ) ) together with @xmath174 and @xmath175 , we obtain @xmath176 the average mass of the peaks with @xmath170 is given by : @xmath177 where @xmath178 ( see ref .
this mass satisfies both condition `` aa ) '' and silk & stebbins@xmath76 prescription for surviving clumps .
+ the fraction of this subpeaks satisfying the ia , and consequently satisfying condition `` bb ) '' , can be calculated using the conditional probability distribution @xmath179 giving the peculiar velocity around a peak of a gaussian density field having central height @xmath5 : @xmath180 ^ 2}{\lambda_{\rm i}}}\ ] ] ( see ref .
37 ) ( @xmath181 denotes ensemble averages ) , where @xmath182\ ] ] @xmath183 , ( ref
. 7 ) @xmath184 @xmath185 is the autocorrelation function and @xmath186 is the central curvature of the peak . @xmath187 , @xmath188 , @xmath189 are three of the four eigenvalues of the covariance matrix @xmath135 described in antonuccio - delogu & colafrancesco@xmath60 .
these three eigenvalues represent the components of the peculiar velocity dispersions along directions parallel ( @xmath187 ) and perpendicular ( @xmath190 ) with respect to the radial direction of a density peak .
the fraction of peaks having @xmath191 can be obtained by means of the conditional probability distribution @xmath192 : @xmath193 the fraction of peaks satisfying the condition @xmath194 can be easily calculated to be @xmath195 for a maxwellian velocity distribution , @xmath196 , characterized by a uniform velocity dispersion @xmath197 . in the case of a top - hat perturbation of radius @xmath198 , @xmath199 can be as large as @xmath200 for a typical velocity dispersion , and even larger for larger values of @xmath201 . using the conditional probability distribution @xmath202 , in the case of a @xmath203 peak
we get @xmath204 within @xmath205
. can be as large as @xmath200 . ] in fig . 1
, we plot the spatial variation of @xmath194 for the cdm spectrum , introduced in section 2 , smoothed on a scale @xmath206 mpc and for @xmath203 . in any case in the following , we ll make a conservative calculation using the lower estimates for @xmath207 . finally , the total number of subpeaks verifying the above conditions `` aa ) '' , `` bb ) '' is given by : @xmath208 while the total energy , @xmath209 , injected in the shell by the impulsive encounters with all high speed subpeaks is given by : @xmath210 the energy for unit mass of a collapsing shell is a modified version of that given by ref .
47 ( equation ( 19.9 ) ) and can be written similarly to ref .
48 in the form : @xmath211 where @xmath212 is the mass of the shell of matter . + integrating equation ( [ eq : ener ] ) , using equations ( [ eq : mass ] ) , ( [ eq : mas ] ) and remembering that @xmath213 where @xmath214 , with @xmath215 and @xmath216 respectively the critical mass density and the hubble constant at the time @xmath136 , the time of collapse is given by : @xmath217 -\frac{4 g}{h_{\rm i}^2 ( 1+\overline{\delta } ) r_{\rm i}^5 } \delta e_{\rm t } } } \label{eq : tco}\ ] ] integrating analytically equation ( [ eq : tco ] ) , we find the solution : @xmath218 having assumed @xmath219 . in the case
@xmath220 the solution of this equation is that given in ref . 2 : @xmath221
the results found by means of the model introduced in the previous section are shown in the fig . 1 - 5 .
as previously reported , fig .
1 shows the spatial variation of @xmath222 for the cdm spectrum , introduced in section 2 , smoothed on a scale @xmath206 mpc and for @xmath203 .
the fraction of clumps with @xmath223 within @xmath224 is @xmath225 , and increases with increasing radius .
this behavior is due to the fact that the potential of the structure becomes shallower going towards its outskirts , and consequently @xmath226 is substantially lower . in fig .
2 we show the total energy injected in the shell , because of the impulsive encounters of collapsing shells with the high speed clumps , in function of the initial radius of the shell .
the dashed line represents @xmath227 for a cdm model with spectrum given by equation ( [ eq : ma5 ] ) for @xmath203 .
as shown , shells collapsing from larger distances from the centre of the cluster acquire more energy .
this is due to two different reasons : firstly the energy , @xmath228 , acquired in an encounter with a single clump increases with increasing turn - around radius of the shell of matter .
secondly , @xmath143 , is proportional to the number of unbound clumps that increases with radius ( see equation ( [ eq : num ] ) ) and then a shell falling from a larger distance encounters more unbound clumps .
the solid line represents @xmath143 for @xmath229 . the value of @xmath143 is smaller with respect to the previous case
the reason is due to the fact that the fraction of the unbound peaks decreases for higher @xmath5 peaks because the potential of these protostructures is deeper and hence the value of @xmath97 is systematically higher . + in fig .
3 we show the collapse time versus the distance from the centre of the cluster .
the figure plots the collapse time from 1 @xmath230@xmath231 on , because of the limits of the ia , stressed by ref .
42 , in the central part of the cluster .
the solid line represents , @xmath232 , for @xmath203 and taking account of the unbound clumps effect , while the dashed line is the collapse time , @xmath233 , given by the model in ref .
this figure shows that the collapse time is accelerated , with respect to the model in ref . 2 , because of the energy acquired by the shells during encounters with unbound clumps .
4 we show @xmath232 , for @xmath229 . as in the previous figure the solid line represents @xmath232 taking account of the unbound clumps effect , while the dashed line represents @xmath233 given by the model in ref .
2 . comparing fig . 3 and fig . 4
we see that the effect of the unbound peaks on the acceleration of the collapse time is larger in the case @xmath203 .
this is in agreement to what observed for @xmath143 .
+ finally , in fig .
5 we make a comparison between the acceleration in collapse time due to unbound clumps to the slowing down effect produced by the gravitational interaction of the quadrupole moment of the system with the tidal field of the matter of the neighboring proto - clusters , described in ref .
3 . as shown in fig . 5
, the magnitude of the two effects is comparable at @xmath234 , but the accelerating effect produced by the high speed clumps decreases with increasing distance from the cluster centre , with respect to the slowing down effect produced by the tidal field . + the mechanism considered in this paper is different from that described in ref . 3 for two reasons : * the delay of the collapse produced by unbound clumps is due to encounters of infalling shells of matter and unbound clumps .
the effect described in ref .
3 is produced by tidal interaction of the matter of a given protocluster with that of the neighboring ones ; * the substructure responsible for collapse delay in this paper is that contained inside the infalling shell while in ref .
3 it is external to the shell .
the two effects have a different physical origin , and act in the opposite direction : unbound clumps accelerate the collapse , the effect of the tidal field is to delay the shell collapse .
it is obvious that during the infall of a shell both effects are present with the final result that at small radii ( @xmath235 ) they counterbalance while at larger radii tidal field is dominant producing a slowing down of the shell collapse .
we want to stress that the idea that high velocity clumps can influence the collapse and secondary infall of matter in protoclusters was introduced ( but not developed ) in a paper by antonuccio & colafrancesco@xmath60 . in the conclusion of that paper , regarding the velocity correlation tensor and unbound substructure around density peaks
, they noted that high velocity clumps can exchange momentum with the collapsing or infalling clumps and generate a drag force onto the collapsing material with a consequent delay of the collapse . according to their paper the drag force is due to dynamical friction .
unfortunately this last considerations are not quite right because : + 1 ) the effects of the high velocity clumps on the collapse shell of matter can not be calculated using the dynamical friction approximation since a necessary condition to apply this last is that the encounter of two systems is characterized by @xmath236 ( where @xmath237 is the internal velocity of the system ) ( see ref .
44 ) . to be more precise ,
according to ref .
49 , a body of mass @xmath135 moving through a homogeneous and infinite distribution of bodies of mass @xmath238 , that we call field particles , is subject to the deceleration : @xmath239 where @xmath240 is the coulomb logarithm and @xmath241 is the phase - space number density field particles . as clearly stated by equation ( [ eq : dyn ] ) only field particles moving slower than @xmath135 contribute to the force . in our case , the role of the field particles is played by the high speed clumps that have speeds much larger than those characterizing the motion of the shell particles ( when ia is satisfied ) or at least a bit larger .
+ 2 ) the effects of the high velocity clumps is to accelerate the shell collapse and not to delay it .
+ we note that the erroneous conclusion of antonuccio & colafrancesco@xmath60 paper is not connected with their analysis of the conditional probability distribution @xmath242 developed in the paper , that we used to calculate the number of subpeaks satisfying the ia .
the erroneous conclusion is due to the way they thought to calculate the interaction between the high velocity clumps and the infalling shells .
in this paper we have studied the role of unbound high velocity clumps on the collapse of density peaks in a scdm model ( @xmath243 , @xmath244 , @xmath245 ) . we have shown that : + 1 ) the encounters of collapsing shells with unbound clumps produce an acceleration in the collapse of density peaks .
the effect is larger for peaks having a lower value of @xmath246 ( see fig . 2 @xmath247 3 ) .
+ 2 ) a comparison between the collapse acceleration due to encounters of shells with unbound clumps and the slowing down effect produced by tidal interaction between a protocluster and the neighboring ones ( see ref .
3 ) has shown that these two effects act in opposite direction , namely the first accelerate the shell collapse , while the second delay the protocluster collapse , especially in its outskirts .
the magnitude of the two effects is comparable at @xmath234 but the effect of the tidal field dominates at larger radii .
+ 3 ) the acceleration in the collapse is due to the additive action of several high speed clumps with the matter of the infalling shells .
we are grateful to e. spedicato , e. recami , g. mamon and o. gnedin for stimulating discussions during the period in which this work was performed and whose comments and suggestions helped us to improve the quality of this work .
a. del popolo and e.nihal ercan would like to thank bo@xmath248i university research foundation for the financial support through the project code 01b304 .
+ 0 aguilar l.a . ,
white s.d.m . , 1985 , apj , 295 , 374 gunn j.e . ,
gott j.r . , 1972 ,
apj , 176 , 1 ( gg ) del popolo a. , gambera m. , 1998 , a&a , 337 , 96 ( dg ) guth a.h . , pi s.y . , 1982 , phys.rev.lett . ,
49 , 1110 hawking s.w . , 1982 , phys.lett b , 115 , 295 starobinsky a.a . , 1982 , phys.lett .
b , 117 , 175 bardeen j.m . , bond j.r . , kaiser n. , szalay a.s . , 1986 ,
apj , 304 , 15 ( bbks ) kaiser n. , 1984 , apj , 284 , l9 filmore j.a . , goldreich p. , 1984
, apj , 281 , 1 bertschinger e. , 1985 , apjs 58 , 39 hoffman y. , shaham j. , 1985 , apj , 297 , 16 quinn p.j . , salmon j.k . , zurek w.h . , 1986 ,
nature , 322 , 329 west m.j .
, dekel a. , oemler a. , 1987 , apj , 316 , 1 ryden b.s . , gunn j.e . , 1987 ,
apj 318 , 15 ryden b.s . , 1988 , apj , 329 , 589 hoffman y. , 1988 , apj , 328 , 489 efstathiou g. , frenk c.s .
, white s.d.m .
, davis m. , 1988 , mnras , 235 , 715 white s.d.m . , zaritsky d. , 1992 , apj , 394 , 1 evrard , a.e . , mohr , j.j .
, fabricant , d.g . ,
geller , m.j .
, 1993 , apj , 419 , l9 crone m.m .
, evrard a.e . ,
richstone d.o . , 1994 ,
apj , 434 , 402 navarro j.f . , frenk c.s . , white s.d.m . , 1995 , mnras , 275 , 720 navarro j.f . , frenk c.s . , white s.d.m . , 1996 ,
apj , 462 , 563 navarro j.f .
, frenk c.s . , white s.d.m . , 1997 , apj , 490 , 493 avila - reese v. , firmani c. , hernandez x. , 1997 , _ preprint sissa astro - ph/9710201 _ hoyle f. , narlikar , j.v . , 1966 , proc .
gott j.r . , 1975 , apj , 201 , 296 gunn j.e . , 1977 ,
apj , 218 , 592 regs e. , geller m.j . , 1989 ,
aj , 98 , 755 briel u.g . ,
henry j.p , schwarz r.a . ,
del popolo a. , gambera m. , spedicato e. , recami e. , 2000a , a&a , 353 , 427 del popolo a. , gambera m. , 1999 , a&a , 344 , 17 del popolo a. , takahashi y. , kiuchi h. , gambera m. , 1999 , a&a , 348 , 667 del popolo a. , gambera m. , 2000b , a&a , 357 , 809 beers t.c .
, geller m.j . , 1983 , apj , 274 , 491 jones c. , forman w. , 1984 , apj , 276 , 38 dressler a. , shectman s.a . , 1988 ,
aj , 95 , 985 antonuccio - delogu v. , colafrancesco s. , 1995 , _ preprint sissa astro - ph/9507005 _ liddle a.r .
, lyth d.h .
, 1993 , phys .
, 231 , n 1 , 2 silk j. , stebbins a. , 1993 , apj , 411 , 439 rees m.j . , 1986 , mnras , 218 , 25p henon , m. , 1964 , ann .
dap . , 27 , 83 gnedin , o.y .
, ostriker j.p . , 1999 , astro - ph/9902326 del popolo a. , gambera m. , 1997 , a&a , 321 , 691 binney j. , tremaine s. , 1987 , galactic dynamics , in princeton series in astrophysics , princeton university press . antonuccio - delogu v. , colafrancesco s. , 1994 , apj , 427 , 72 peacock j.a . , heavens a.f . , 1990 , mnras , 243 , 133 peebles p.j.e . , 1980 , the large scale structure of the universe , princeton university press peebles p.j.e . , 1993 , principles of physical cosmology , princeton university press chandrasekar s. , 1943 , apj , 97 , 255 | in the cold dark matter ( hereafter cdm ) scenario even isolated density peaks contain a high fraction of small scale clumps having velocities larger than the average escape velocity from the structure .
these clumps populate protoclusters , especially in the peripheral regions , @xmath0 ( where @xmath1 is the filtering scale ) . during the cluster collapse and the subsequent secondary infall , collapsing or infalling clumps ( having @xmath2 )
interact with the quoted _ unbound _ clumps ( or _ high speed clumps _ , as we also call them ) having @xmath3 .
we study the interaction between these two kind of clumps by means of the impulse approximation@xmath4 and we find that the collapse of _ bound _ clumps is accelerated with respect to the homogeneous case ( gunn & gott s model , ref .
2 ) .
the acceleration of the collapse increases with decreasing height of the peak , @xmath5 .
we finally compare the acceleration produced by this effect to the slowing down effect produced by the gravitational interaction of the quadrupole moment of the system with the tidal field of the matter of the neighboring proto - clusters studied in del popolo & gambera@xmath6 . we find that the magnitude of the slowing down effect is larger than the acceleration produced by the effect studied in this paper , only in the outskirts of the cluster .
we want to stress that the one which we study in this paper is also present in an isolated protocluster , being produced by the interaction of the collapsing clumps with the _ unbound _ substructure internal to the collapsing clumps itself while that studied in ref .
3 is produced by substructure external to the density peak . |
corruption seems to be an unavoidable part of human social interaction , prevalent in every society at any time since the very beginning of human history till today . in sharp contrast to the high prevalence of corruption in many countries and the rather large literature on political , social and economical aspects of corruption
there is only a small number of attempts to model the dynamics of corruption in a mathematically quantified way .
the modelling approach in these few attempts essentially follows two paths .
the first is in the sense of microeconomics and incorporates game theoretic aspects ( for a recent model in this direction see the book by steinrcken @xcite and the references therein ) or rules for maximizing a certain economically based profit functional ( @xcite@xcite )
. then a set of differential equations for the evolution of the mean corruption is derived and a stability analysis done on that basis . in these models
one usually makes rather detailed assumptions about the underlying organization structure on which the individuals interact .
the second line of approach is more in the sense of cellular automata ( ca ) models with rather simple state variables and local interaction dynamics .
for example in the article by wirl @xcite a simple 1-dimensional deterministic cellular lattice automata model is used to describe the propagation of corruption . nevertheless , as is well known in ca - modelling , the global dynamical picture can be highly complex and nontrivial .
up to now all these attempts did not take into account the complex network of social relationships as the underlying structure for the spread of corruption . in this article we will present a model for the spread of corruption on complex networks in the spirit of epidemiology .
the model describes aspects of the evolution of corruption in a virtual population and incorporates some basic universal features of corruption .
the local interaction dynamics of the model is similar to cellular automata but `` lives '' not on a lattice type graph like most of the ca - models but on complex networks . considering corruption as a nonstandard epidemic process relies on the plausible assumption that corruption rarely emerges out of nothing but is usually related to some already corrupt environment which may `` infect '' susceptibles
of course the spontaneous decision of somebody to act corruptly is possible and could easy be handled in the model as an external weak source of infection .
one of the very special features in corruption propagation which differs from what is used in describing classical epidemic processes is the threshold like dependence of the local transition probabilities . by this
we mean that a noncorrupt individual gets infected with high probability if the number of corrupt individuals in the group of his direct social contacts ( encoded as the set of neighbors in a `` friendship '' or acquaintance graph ) exceeds a certain threshold number .
otherwise if the number of corrupt individuals in somebodies social neighborhood is below that threshold value there is only a small probability to get corrupt via such `` local '' interactions .
the second main difference to classical epidemic processes is the mean field dependence of the corruption process . by this
we mean that an individual can get corrupt just because there is a high prevalence ( or believed prevalence ) in the society even when there is no corruption in the local neighborhood .
there is another interesting mean field term entering the game , namely the society strikes back to corruption with an efficiency proportional to the fraction of the noncorrupt people
. both mean field terms are nonlinear and together with the local propagation mechanisms they give rise to a rather complex dynamical picture .
corruption is a substructure of human social interaction .
common sense associates corruption mainly with a deviation from fair play interaction in the development of social relations .
clearly what is meant be fair play depends on the cultural context of a given population / society .
this vague description of corruption is in the spirit of sociology and psychology and differs from the more narrow corruption concepts usually considered in economics or political sciences .
there , corruption is mainly seen as a misuse of public power to gain profit in a more or less illegal way . in any case
, corruption has many different faces in its concrete appearance and no single model approach will be able to describe the whole picture in an adequate way .
but this does not at all imply that mathematical models are useless in this situation .
they can provide a substantial improvement in our understanding of corruption as long as one clearly defines the aim and limitations of the taken approach . for the model approach developed in this article we will use the notion of corruption in the more general , first sense .
more precisely our intention is to describe changes in mind ranging from damming of corruption as a criminal act to accepting corruption as an attractive option .
therefore in this paper we do not introduce the group of state representatives or officials since we assume that the essential changes in mind which allow corrupt acts happen long before an individual is in the position to act corruptly .
empirical investigations about motives and `` typology '' of corrupt actors ( see @xcite for results from case studies in germany ) have shown that the majority of individuals involved in corruption affairs are highly educated , well positioned with respect to social status and do not think to have done something wrong , indicating the importance of mind changes prior to corrupt acts .
there is a notorious problem in finding good empirical data which would allow to estimate the prevalence of corruption .
probably the greatest effort over the last years to measure the degree of corruption in various countries was made by `` transparency international '' ( ta ) , a non profit group of individuals and organizations which are highly concerned by the lack of sound data . since 1995
they publish a yearly corruption report and a so called corruption perception index ( cpi ) @xcite .
ta is well aware of the subjectivity in peoples perception of corruption but hopes that the large number of cases involved in the cpi averages out most of the bias .
figure [ 62 ] gives a cpi - rank plot of the 2004 date from ta .
note that a value of @xmath0 for the cpi corresponds to the absence of corruption . for 2004 finland holds the
top ranking and germany is on place @xmath1 with an index of @xmath2 .
[ ptb ] corrpap1.eps it is not our aim to explain the values of the cpi or other corruption data sets , since this would require a semirealistic modelling of the social and economical structure of individual countries which is completely illusionary at the present stage of research .
rather we want to demonstrate which scenarios are dynamically possible and whether there are phase transitions .
in this section we first describe the basic setting for our model structure .
refinements and more detailed aspects will be discussed later on . due to the common view
, corruption is first of all a property of the relations between individuals irrespectively which definition of corruption one uses . since an act of corruption requires that at least one of the participants in a corrupt relation has a mental state which tolerates or even assigns a positive value to ( his personal view of ) corruption we will focus mainly on the spread of this mental state change ( from not accepting to accepting corrupt acts as an option for one s own activities ) .
therefore to discuss corruption as an epidemic process in the afore mentioned sense it is useful to assign a corruption property to the individuals themselves . in the simplest case
we just have a time dependent @xmath3 state variable @xmath4 assigned to each individual , encoding whether the vertex is corrupt @xmath5 or not @xmath6 at time @xmath7 ( of course more refined scales for the degree of corruption are possible and will be discussed in a forthcoming paper ) .
the underlying structure on which corruption spreads is a given finite graph @xmath8 from a random graph space @xmath9 with fixed vertex set @xmath10 furthermore we consider in this article only stationary graphs with no changes in time on the underlying graph structure ( the study of corruption on evolutionary graphs requires a paper in its own ) .
the dynamics is specified by conditional transition probabilities @xmath11 which depend mainly on the states on @xmath12 and a meanfield term reflecting the influence of the total prevalence of corruption in the society . here
@xmath13 is the usual graph metric on @xmath14 and @xmath15 is the degree of @xmath16 .
we define @xmath17 as the density of corruption at time @xmath7 .
the standing assumptions on @xmath11 are the following : @xmath18\end{aligned}\ ] ] in other words the probability to become corrupt depends only on the local prevalence of corruption among the neighbors and the mean corruption in the society and individuals who became corrupt can cure from corruption with a rate proportional to the density of the noncorrupt individuals in the society . in classical i.i.d .
epidemics one would have the following as functional dependence for the local part of the conditional probabilities : @xmath19 which is for small @xmath20 and @xmath21 proportional to @xmath22 for corruption the function @xmath23 is more like in voter models , that is below a critical value @xmath24 of the number of corrupt individuals in @xmath25 the value of @xmath23 is close to zero and above @xmath24 it is a number @xmath26 much larger then zero . due to this property local clustering can force the epidemics to spread whereas in classical epidemic processes high clustering slows down the spread of an infection due to reinfection of the already infected .
we want to illustrate this by two simple examples .
the simplest , almost trivial example is the @xmath27 lattice with additional edges to the next - nearest neighbors . setting @xmath28 and @xmath29 for @xmath30
it is easy to see that there is a nonzero probability for infecting all vertices starting with one infected individual at time @xmath31 the infection function @xmath23 will be the same as in example 1 .
we start with a regular tree of degree 3 .
replacing each vertex by a triangle and gluing the triangles along the former edges of the regular tree gives a regular graph of degree 4 where the triangle corners act now as the new vertices . in each neighbor pair of triangles @xmath32 ( that are the triangles which have a common vertex )
we form an edge randomly between the set of vertices lying in @xmath33 and @xmath34 ( see fig .
[ ptb ] corrpap2.eps once a triangle is infected the corruption jumps to all the three neighbor triangles due to the extra random edge present between each neighbor pairs of triangles .
hence again we have a nonzero probability that the whole graph becomes infected . in the above examples we have used a very simple and somehow extreme form of the infection function @xmath35 in the following we will investigate the situation for two canonical subclasses of infection functions .
we say that @xmath23 is a vertex independent , fixed threshold infection function if there is a @xmath36 such that @xmath37 for @xmath38 and @xmath39 for @xmath40 for the second class of functions we assume the threshold to be degree dependent .
namely we call @xmath41 a vertex dependent , relative threshold function if for some @xmath42 we have @xmath43 for @xmath44 and @xmath45 for @xmath46 .
furthermore we say that @xmath23 is a voter - type infection function if @xmath47 for @xmath48 and @xmath49 for @xmath50 .
in this paper we will mainly investigate the spread of corruption for the fixed threshold case . to distinguish between the different ways in which an individual can become corrupt we will speak about the @xmath51 or @xmath52 process . for convenience of the reader
we give in tabular 1 a summary of the different processes .
note that in contrast to standard voter models we do not have the possibility of a locally induced backflip from the corrupt state to the noncorrupt .
a kind of quenched disorder could easily be introduced by randomizing the relevant parameters individually but this will be the subject of a forthcoming paper .
generalizations of classical epidemic dynamics to processes with a local threshold have recently also been studied in the context of models of contagion ( see @xcite and references therein ) but not yet been mixed with global mean field processes .
in the last 10 years there has been an enormous progress in understanding the fine structure of social networks .
this is mostly due to the availability of large data sets for some special social networks like e - mail correspondence , coauthorship network in scientific publications and movie actor networks to name just a few prominent examples .
all these networks of social relations share three remarkable properties of the associated graph which are : 1 ) the diameter scales at most logarithmically in size 2 ) the graphs have a very high , asymptotically non - vanishing clustering coefficient- in other words the graphs are locally far from being tree - like 3 ) the degree distribution follows a power law ( scale - free graphs ) . properties 2 and 3 have striking consequences for the spread of corruption as will be discussed later on .
there exists meanwhile a large collection of algorithms to generate complex networks with the above mentioned properties . a widely used quantity to measure
the local clustering is the triangle number @xmath53 and it s averaged value @xmath54 .
a natural generalization is the @xmath55 clique number @xmath56 defined as the number of complete graphs of order @xmath21 containing @xmath57 in social network graphs @xmath58 is usually proportional to @xmath15 and @xmath54 becomes independent of the population size for large @xmath59 and stays bounded away from zero .
another very remarkable property of real networks is the power - law distribution for the degree . by an asymptotic power - law distribution for a discrete random variable @xmath60
we denote every functional behavior of the form @xmath61 with exponent @xmath62 most real networks have exponents between 2 and 4 ( see @xcite for an excellent overview ) .
classical epidemic processes on such graphs have been studied by many authors , and perhaps the most astonishing result in this context is the absence of an epidemic threshold in case the exponent is below @xmath63 ( @xcite ) .
this phenomenon is related to the existence of a massive center of size independent diameter induced by the high number of hubs ( vertices with an exceptional large degree ) .
hubs play also a significant role for @xmath64 - process as will be explained in the next section .
one of the main differences between corruption epidemics and classical epidemics is the different effect of clustering on the epidemic threshold and the total number of infected individuals . in the classical situation any epidemics
will be slowed down by the presence of local cycles due to the high probability of reinfection . in corruption epidemics
local clustering may speed up the propagation of corruption due to the nonlinear dependence of the infection probability on the number of infected neighbors as was already demonstrated in example 1 in the previous section . in the next section we will give two further examples where the strength of this effect can explicitly be computed .
in this section we want to look at some threshold properties associated to the @xmath65 process .
we are still far from a good understanding of the quantitative picture of this kind of processes for a given type of graph , which is mainly a consequence of our lack of knowledge how to handle graphs with high local clustering in a mathematical satisfactory way . in this section
we want to state just some general observations and numerical results concerning the spread of threshold - like dynamics .
furthermore we will analyze two examples of tree - like graphs which might serve as an illustration .
a more careful mathematical analysis of @xmath65 processes requires a paper in its own .
one of the remarkable differences between a classical epidemic process and a process based on local threshold dynamics is the dependence on the initial number of `` infected '' vertices in the latter case .
classical epidemics does not know such things- either an epidemic process is overcritical ( reproduction number @xmath66 ) and a single initial infected vertex infects with with positive probability a positive fraction of the whole population , or the process is below criticality ( @xmath67 ) and all infected will die out respectively become healthy . in corruption epidemics
both parts - the mean field process as well the local @xmath65 process - can have phase transitions with respect to the initial number of corrupt vertices .
that means , there is critical initial density of corrupt vertices @xmath68 such that for initial densities below @xmath68 the number of infected stays as it is or goes down to zero . above @xmath68
the entire population becomes corrupt with high probability . as an illustration we give in fig .
[ 52 ] the dependence of @xmath68 on the edge density @xmath69 on a classical random graph space @xmath70 with @xmath59 vertices and @xmath71 edges .
[ ptb ] corrpap3.eps although in this paper we mainly concentrate on the case of absolute threshold values @xmath36 we give for comparison in fig.[55 ] the edge density dependence result for a relative , degree dependent threshold @xmath72 .
there is still a critical density but its value increases with the edge density since the mean threshold increases now proportional to the mean degree .
[ ptb ] corrpap4.eps as already mentioned in section 3 one expects that the presence of clustering ( respectively many triangles ) decreases the critical density @xmath68 since the @xmath64- process can propagate more easily . in fig .
[ 99 ] the effect of the increase of the triangle number is clearly to see .
here we used a modified @xmath70 random graph space where randomly triangles are added ( keeping the total number of edges constant ) .
the threshold value @xmath36 was chosen to be @xmath73 since for higher @xmath36 one has to add higher order complete subgraphs instead of triangles .
[ ptbptb ] corrpap5.eps the next figure ( fig .
[ 300 ] ) shows the dependence of the critical density on @xmath74 the two curves represent the threshold values for an end - prevalence of @xmath0 respectively @xmath75 percent .
since the mean degree in this simulation is about @xmath76 one has a vanishing contribution of the @xmath64- process above @xmath77 .
the critical threshold @xmath68 stays than essentially at a value given by the mean field process ( see next section for details ) .
[ ptb ] to get an impression of the contribution of the different kind of processes ( local @xmath64 and @xmath20 , global @xmath78 and @xmath79 - for details see next section ) to the end - prevalence we give in fig .
[ 400 ] the accumulated number of state changes caused by each of the subprocesses till saturation . for small values of @xmath36
the @xmath64- process dominates all others .
[ ptbptb ] corrpap7.eps we turn now to a more theoretical consideration , namely which type of vertices ( type in the sense of degree and local clustering ) are especially well suited for the propagation of corruption via the @xmath65process .
assume we have given a random scale free graph space @xmath9 with @xmath59 vertices .
we further assume that there are two types of edges ( according to the way they where generated ) the independent ones , generated at random with just preferences to the degree ( like the preferential attachment rule by albert&barabasi or the `` cameo - principle '' in @xcite ) and local ones which are relevant for the creation of triangles .
let further the ( asymptotic ) degree distribution given by @xmath80 with @xmath81 and @xmath82 the normalization constant .
the independent edges are generated with probability @xmath83 and each individual generates @xmath84 edges by himself . from @xcite
one knows that the triangle number @xmath85 is proportional to @xmath86 with a proportionality constant @xmath87 .
a basic quantity in highly clustered networks is the probability @xmath88 that two random chosen elements from @xmath89 have a common edge . since @xmath90 one obtains for the conditional probability @xmath91 . assuming that the generation of triangles is a sufficiently independent process one obtains for the conditional @xmath92 clique number @xmath93 here @xmath94 is the number of @xmath95- cliques ( complete graphs of order @xmath95 ) containing @xmath16 . for
@xmath96 the power in the @xmath55dependence gets negative and hence the high degree vertices contribute almost nothing to the clique - clustering .
of course all this consideration rely on the assumption of some kind of independence in the triangle - formation process . in any case
this results indicate that highly clustered medium degree vertices are especially well suited for the spread of corruption .
a similar kind of analysis can be carried out for random graph models which have an intrinsic high probability to generate local cliques e.g. intersection graphs ( for an introduction to random intersection graphs and comparison with erds - renyi random graphs see @xcite and @xcite ) .
the above arguments seem to support the conjecture that in corruption epidemics the vertices from the tail of the degree distribution play a less dominant role .
this is indeed true in the case of a relative , degree - dependent threshold where hubs are much more difficult to infect than medium or low degree vertices . for absolute thresholds in the @xmath64- process
the situation is more complex since for scale free degree distributions with small exponents ( @xmath97 ) there are other mechanisms than local clustering which can cause a radical dropdown of the critical initial density . in fig.[56 ] and fig.[57 ] we give numerical results for the relation between the critical density @xmath68 and the exponent @xmath98 keeping the edge density fixed .
there is a clear phase transition around @xmath99 for @xmath100 and @xmath101 for @xmath102 .
[ ptb ] corrpap8.eps [ ptbptb ] corrpap9.eps the explanation of this observation is closely related to a structural phase transition in scale free random graphs at @xmath103 - namely that for most vertices @xmath16 an asymptotically positive fraction of all vertices has bounded distance to @xmath16 . to link this property with the @xmath64 - process one has to look more closely on the degree - degree correlation in scale - free graphs . depending on the choice of the model one can have very different correlations like:@xmath104 .
formula [ 111 ] holds for instance for the cameo - model ( @xcite ) whereas formula [ 113 ] is valid for scale - free graphs generated via the molloy&reed algorithm ( the later one represents the random graph space containing all graphs with a given scale - free degree distribution equipped with the uniform measure and was used for the simulations in fig.[56 ] and [ 57 ] ) .
evolutionary graphs like the albert&barabasi model have usually asymmetric and more complicated correlations .
since a detailed analysis of the @xmath64 - process for scale - free graphs is beyond the scope of this paper we just give a heuristic outline why in graphs with a correlation as in formula [ 113 ] the threshold density @xmath68 tends to zero as @xmath105 for exponents @xmath97 .
for fixed @xmath106 and @xmath107 ( note that the typical maximal degree is about @xmath108 it is obvious that vertices @xmath16 with @xmath109 get almost surely infected ( as @xmath105 ) via the @xmath64 - process as soon as @xmath110 .
let @xmath111 be the set of such vertices .
one the other side it follows from [ 113 ] that a vertex @xmath112 with @xmath113 is linked to the set @xmath114 with probability@xmath115 . since @xmath116 is close to @xmath117 for @xmath118 one has an almost sure multiple linkage of vertices @xmath112 with @xmath119 to the set @xmath114 .
these vertices get now again infected via the @xmath64 - process . by iterating this procedure one may arrive at an positive @xmath59- independent infection density @xmath120 such that the @xmath78 - process is overcritical and finally the whole vertex set becomes corrupt .
the mechanism described requires @xmath59 to be large and therefore we conjecture that the difference to the numerical results depicted in fig.[56 ] ( phase transition at @xmath121 instead of @xmath63 ) is due to finite size effects . in the case of @xmath102 ( fig.[57 ] )
the finite size effects are smaller and the phase transition is closer to @xmath63 .
a similar kind of arguments shows , that the expected path - length is finite for @xmath97 .
namely since the expected number @xmath122 of vertices at distance @xmath95 from a vertex @xmath16 with degree @xmath123 is approximately given by @xmath124 @xmath125 for @xmath97 ( note that this expression is only valid for @xmath95 s.t.@xmath126 ) .
the essential diameter @xmath127 ( a large fraction if the whole vertex set is within a ball of diameter @xmath127 ) is then given by the smallest @xmath95 such that @xmath128 ( for a more extensive discussion of the notion of essential diameter see @xcite ) .
for @xmath129 one obtains therefore @xmath130 . for @xmath131
the essential diameter is no longer bounded but growths logarithmically in @xmath59 .
it is interesting that the jump in the critical density at @xmath132 in fig.[56 ] coincides with a jump in diameter from @xmath133 to @xmath134 .
a small essential diameter can have fatal consequences for corruption epidemics since most vertices are closely linked to hubs and , as was outlined above , hubs are with high probability corrupt .
a precise estimation of the dependence of @xmath68 from @xmath135 and @xmath98 requires a careful discussion of the involved constants . for scale - free graphs with additive degree correlation like
cameo - graphs one still has a bounded essential diameter for exponents less than @xmath63 .
but the first argument about chains of almost sure linkages from high degree to low degree vertex sets can not be adopted .
one expects therefore a higher value of the critical density @xmath68 . in fig.[58 ] and fig . [ 999 ] we give numerical results for a scale - free graph with additive degree - correlation generated via a modified molloy&reed algorithm virtual vertices for each vertex @xmath16 and makes than a random matching between the virtual vertices .
two vertices @xmath16 and @xmath112 are connected by an edge if there is an edge between two corresponding virtual vertices . to generate an additive degree correlation we mark @xmath71 virtual vertices as red such that each vertex @xmath16 has at least one red and at most @xmath136 @xmath137 red associated virtual vertices ( if there are note too many vertices with very small degree , the constant @xmath136 can be choosen as @xmath137 ) .
than the marked red virtual vertices are randomly matched with the unmarked ones . ]
[ ptbptbptb ] corrpap10.eps [ ptbptbptb ] corrpap11.eps to compare with the multiplicative case we have chosen the same parameters and degree - distribution as for fig.[57 ] .
there is a clear increase of @xmath68 to observe but , although unlikely , it remains open wether there is a vanishing threshold in the limit @xmath138 . for intermediate couplings
we still expect @xmath139 as @xmath59 diverges for @xmath140 where @xmath141 depends on thew concrete model .
it is remarkable that low @xmath98 and a tendency to multiplicative correlation is mainly expected to hold in societies with strong hierarchical structures of social dependencies e.g. dictatorships ( see @xcite for details ) , whereas democracies are characterized by less strong degree correlation .
finally we will discuss two examples of graph structures where the critical infection density can be explicitly be computed .
the first one is a regular infinite tree of degree 4 where of course no triangles are present ( see fig .
[ 12 ] ) .
[ h ] corrpap12.eps .
the second structure is a regular infinite graph of again of degree 4 with positive local cluster coefficient ( @xmath142 ) and a global tree - like structure ( see fig.[13 ] ) .
[ hptb ] corrpap13.eps in both cases an exact computation of the critical infection density is possible .
we give a short outline for the case of threshold value @xmath102 and @xmath143 ( the case @xmath144 requires more lengthy computations but can be done in a similar fashion ) and start with the case of the regular 4 tree .
an initial configuration is given by marking each vertex with probability @xmath145 as noncorrupt ( black ) and with probability @xmath146 as corrupt ( white ) .
we ask for the critical probability @xmath147 such that for @xmath148 almost surely the entire tree becomes white ( corrupt ) and for @xmath149 there remains an infinite cluster of noncorrupt ( black ) vertices with probability one .
note that no finite cluster of black vertices -that is a finite black subgraph surrounded by white vertices- can survive so there are either infinite black clusters or none .
we call an invariant infinite black cluster immune .
since @xmath102 any vertex in an immune cluster must have at least three black neighbors from that cluster . denote by @xmath150 the rooted tree with outdegree @xmath63 ( fixing a root gives a canonical direction to the edges of the tree so it makes sense to speak about the outdegree of a vertex ) .
every vertex has degree @xmath133 except the root which has degree @xmath63 .
let @xmath16 be the @xmath151 dependent probability that the root is contained in an immune cluster ( as a subgraph of @xmath150 ) conditioned that the root vertex is initially black .
by arguments from the general theory of branching processes @xmath16 equals the largest solution of the following recursion equation @xmath152 figure [ 32 ] displays the different situations which enter the above equation .
the solutions are @xmath153 and @xmath154 . since @xmath155 is needed to have a positive nonzero solution we get for the critical probability @xmath156 [ ptb ] corrpap14.eps in a similar fashion one can derive a recursion equation for the generalized tree case .
for that let @xmath157 be the rooted generalized tree shown in fig.[42 ] .
[ ptb ] corrpap15.eps to every vertex is attached an outgoing triangle , hence the degree of a vertex is @xmath133 except the root which has degree @xmath73 . to settle the question about @xmath147 for the original generalized tree it is enough to analyze the corresponding problem for @xmath157 .
again let @xmath16 be the probability that the root vertex is in an immune cluster conditioned that the root is initially black .
one gets the following recursion equation@xmath158 ( see fig.[42 ] ) .
the solutions are @xmath159 and @xmath154 . again
since @xmath160 is needed to get a positive nonzero solution we get for the critical probability @xmath161 that means the presence of clustering in this example lowers the critical initial density needed to infect the whole graph by almost a factor of @xmath162 the study of the regular @xmath163tree generalizes easily to the case of regular @xmath164 trees ( @xmath165 ) .
the recursion equation in this case is @xmath166 a straightforward but lengthy computation gives for the critical probability@xmath167 in the special case of a @xmath168 tree ( @xmath169 ) one obtains @xmath170 . for completeness
we give without proof the formula for the computation of the critical probability in case of a rooted random tree with arbitrary outdegree distribution .
let @xmath171 be the generating function for the outdegree ; that is @xmath172 is the probability that a random chosen vertex has outdegree @xmath173 ( and hence total degree @xmath174 ) .
the critical probability @xmath147 is given by the smallest @xmath145 such that the equation @xmath175 has a positive real solution .
a careful reader may have noticed that there is a big structural difference between the generalized tree in example 2 of section 3 and the generalized tree just discussed above .
namely the graph of the first example has the property , that any two vertices can be linked by a chain of triangle where neighbor triangles always have a common edge .
the graphs in the examples of this section do not have this property since neighboring triangles have only a common vertex . for threshold values
@xmath176 one has to consider chains of @xmath177 cliques .
we say that a graph is well @xmath21 -linked if any pair of vertices can be linked by a chain of complete graphs of order @xmath21 such that all neighboring @xmath55 cliques have a @xmath178- clique in common . for well @xmath21- linked graphs the critical density @xmath179 is zero ( a finite number of initially infected vertices
can already infected a positive fraction of the vertex set ) for @xmath65processes with @xmath180 whereas for graphs which are not well linked one needs a positive critical density .
the above study on trees or generalized trees is insofar important as in most random graph models used for complex networks one has as a tree or generalized tree as the typical local structure around a random chosen vertex .
furthermore the dependence of the corruption dynamics on graph properties like edge density or degree distribution is in large parts of the parameter space entirely caused by the @xmath65 process .
in this section we want to have a closer look at the mean field dependence of the corruption process . to gain some insight in the possible type of behavior we start with some simple assumptions which will be refined later on .
again we will argue in a discrete time model but the transition to continuos time makes no problem and gives the same results .
let @xmath182 the density of corrupt people at time @xmath183 we assume that the affinity for an individual to change its behavior from noncorrupt to corrupt increases proportional to the corruption prevalence .
furthermore to become really corruptly minded an individual has to overcome some fear which we put proportional to @xmath184 .
formally this reads as @xmath185 with @xmath186 $ ] .
corrupt individuals can recover due to state and police effects ( uncovering , fear etc . ) .
again it seems reasonable to assume that the probability to recover is proportional to @xmath187 since only the noncorrupt part of a society is willing to fight corruption .
formally we will assume that @xmath188 with @xmath189 $ ] .
this gives @xmath190 with the two obvious fixed points @xmath154 and @xmath191 for @xmath78 @xmath192 there is a third intermediate fixed point @xmath193 an interesting phenomena happens for parameter pairs @xmath194 s.t .
@xmath195 since under this conditions both fixed points at @xmath154 and @xmath117 are locally stable .
hence there are two basins of attraction- one for @xmath154 and one for @xmath196 with @xmath197 as the boundary point .
in other words , if the initial percentage of corruption is less @xmath197 corruption stays under control whereas for an initial value larger @xmath197 things run out of control and a corruption collapse takes place .
of course this mean field part of the model is still very simplistic and one should not expect any quantitative fit with empirical data .
but the qualitative statement seems to be quite stable with respect to modifications .
for instance there are good reasons to believe that neither the mean field infection nor the mean field recover process are linear in @xmath198 we want to end this section with a small modification of the mean field `` ansatz '' where we include social weights .
this is a natural and meanwhile very common approach in network dynamics and can easily be adopted to the corruption model . in the above argumentation on the attraction of becoming corrupt it is plausible to assume that corrupt individuals with high social influence have a stronger influence on the mean field probability to get corrupt than individuals with low social importance .
a similar argument holds for the recover probability . as a simple measure for social strength we use the degree of the vertices since high degree vertices are more likely to play a dominant social role than low degree vertices .
formally we introduce the weighted density @xmath199at time @xmath7 as @xmath200 where @xmath201 is the number of vertices with degree @xmath21 and @xmath202 the number of corrupt ( state @xmath117 ) vertices with degree @xmath21 at time @xmath7 . the mean field equation for group @xmath21
is now given by@xmath203 .
multiplying the last equation by @xmath204 and summing over @xmath21 gives @xmath205 which is the same as equation ( [ 100 ] ) .
therefore the introduction of social weights does not add anything new to the dynamical picture .
there is of course a difference in the interpretation since a small real initial prevalence of corruption can give rise to a high initial value of @xmath206 as soon as the corruption is concentrated at the high degree vertices . here
also a difference between scale free networks and classical random networks is seen since in the scale free case high degree vertices ( hubs ) are much more frequent than in the classical case .
in this paragraph we will investigate some aspects of the interplay between the mean field process described in the previous section and the local , threshold dependent , corruption propagation .
for @xmath208 there is a core infected component generated via the @xmath65 process . to gain some insight how such a core infected part of the population changes the mean field dynamics
we will assume that a certain fraction , say @xmath209 of the population is permanently infected and resistant to the @xmath52deletion process .
denoting by @xmath210 the density in the noncore part of the population ( the normalization here is still with respect to the total population size ) we get the following mean field dynamics:@xmath211 . since the state where all individuals are infected is stationary we get the following set of fixed points:@xmath212 . for @xmath213
there are no real fixed points except @xmath214 which becomes globally stable under this condition .
since we have a polynomial of degree 3 we get @xmath215 as the condition for the fixed point at @xmath216 to be locally stable .
furthermore in this case also the fixed point at @xmath217 becomes a local attractor .
this is for instance the case when @xmath218 becomes very small and @xmath219 - being back essentially in the situation of the previous section . in case
when @xmath220 it is easy to show that the fixed point at @xmath221 becomes a global attractor ( to see this just note that the derivative at @xmath222 is always positive for the relevant parameter intervals ) .
the above considerations show that the possible dynamical evolution scenarios are the same for @xmath223 and @xmath224 but there is a very strong influence of @xmath218 on the parameter regimes of @xmath78 and @xmath79 for which one has a corruption collapse . whereas in case @xmath223
one is always in the basin of attraction of zero for @xmath225 sufficiently small and @xmath226 ( in other words @xmath227 is never a global attractor ) one can now have the phenomenon that only the complete saturation with corruption is stable ( @xmath228 ) . as an example lets look at the case where @xmath229 . for @xmath223
there is a fixed point at @xmath230 and hence for an initial infection density @xmath231 the pure mean field dynamics converges to zero . in the case @xmath232 one
has for @xmath233 only the stable fixed point @xmath234 at @xmath235 there is a phase transition since a new indifferent ( slope 1 ) fixed point at @xmath236 emerges . for @xmath237 this fixed point bifurcates into two fixed points where the first one at @xmath238 becomes locally stable with a basin of attraction given by @xmath239 @xmath240 we close this section by presenting a numerical result showing the different contributions to the overall infection ( end - prevalence ) of the local and mean field processes as a function of the edge density in the random graph space @xmath241 .
fig.[200 ] gives the accumulated number of state changes ( divided by @xmath59 ) caused by the @xmath242 and @xmath20- process at initial density values slightly above the critical one .
[ ptb ] corrpap16.eps up to an edge density of @xmath73 ( corresponding to a mean degree of @xmath133 ) the @xmath181 process gives the major contribution to the end prevalence in the overcritical situation .
parallel to the increase in the edge density increases the contribution of the @xmath64- and @xmath20- process ( in the intermediate phase of density between @xmath73 and @xmath63 dominated by the @xmath64- process ) till a sharp peak at edge density @xmath243 where the @xmath244 process outperforms all the others ( at the same time the critical initial corruption density @xmath68 drops down and becomes almost zero ) . the peak is easy to understand since for the chosen parameters we have at an edge density of @xmath133 an equality between the recover rate @xmath79 and the expected number of new corruptions caused by a single corrupt vertex via the @xmath244 process ( which is @xmath245 ) . in terms of classical epidemic
processes this corresponds to the case of reproduction number @xmath246 above this value single initial corrupt vertex is already enough to cause in conjunction with the mean field process a total infection of the network .
in this section we want to present some simulation results of the corruption process taking place on some medium size complex networks .
small graph sizes are interesting as they are typical for communities in highly social structured populations . as a simple to generate random graph space with high clustering and power law degree distribution
we have chosen so - called intersection graphs .
intersection graphs can easily be defined as follows .
first one forms random sets from a finite base set of @xmath59 elements ( random means in this context that the set elements are chosen uniform i.i.d . from the base set ) .
these sets constitute the vertices of a random graph .
edges will be defined via the set intersection property , namely there is an edge between @xmath173 and @xmath247 if the associated sets @xmath248 and @xmath249 have nonempty intersection .
the size ( cardinality ) @xmath250 of a set @xmath251 is itself a random variable drawn i.i.d . from a pre - given probability distribution @xmath252 to get interesting graph spaces one furthermore requires @xmath253 for theoretical results about the structure of random intersection graphs see @xcite@xcite@xcite .
it is worth noting that intersection graphs have a high clustering by definition ( if an element is contained in say @xmath21 sets simultaneously this @xmath21 sets form a complete subgraph ) .
most simulations were done for the case when @xmath254 is an asymptotic power law distribution with exponent @xmath63 or when @xmath254 is singular ( all sets have the same size ) .
random intersection graphs have a multiplicative degree correlation and therefore the critical threshold should be very low for exponents less than @xmath63 be the arguments from section 5 . above that value the form of the degree distribution has only little influence on the corruption propagation .
[ c]|l|l|l|**graph characteristic * * & * fp2 ( 1987 - 1991 ) * & * fp3 ( 1990 - 1994 ) * + # vertices & @xmath255 & @xmath256 + # edges & @xmath257 & @xmath258 + mean degree & @xmath259 & @xmath260 + maximal degree & @xmath261 & @xmath262 + # vertices wiht degree @xmath263 @xmath134 & @xmath264 & @xmath265 + size of largest component & @xmath266 & @xmath267 + mean # triangles per vertex & @xmath268 & @xmath269 + exponent of degree distribution & @xmath270 & @xmath271 + besides random intersection graphs generated according to some degree specifications we used also a collection of real collaboration graphs .
these graphs come from a database about research and development projects funded by the european community ( fp2 - 3 ) .
it s vertices are organizations involved in european research projects .
two organizations are linked if they have a joint project ( see table ii for the main graph characteristics ) . in total
the data base contains about 8000 projects and 13000 participating organizations .
in essence the network shows all the main characteristics that are known from other complex network structures like scale free degree distribution ( with exponent between @xmath73 and @xmath63 ) , small diameter and high clustering and vertex correlation .
the initial fraction of infected individuals was either distributed at random over the vertex set or clumped together in a sufficiently large ball with a random chosen vertex as center . in the following we want to give a small sample of simulations on the just mentioned graphs and
try to discuss its main features .
fig.[2 ] displays the prevalence of corruption on the real network fp2 .
the absolute threshold value @xmath272 is very high and does not allow for a big outbreak of corruption .
but there is a metastable small community of individuals , highly linked and almost resistant to the @xmath52process .
it took more then 800 complete updates till this structure broke down .
[ ptb ] corrpap17.eps the next figure [ 3 ] presents a similar situation on an almost twice as large real graph ( fp3 ) .
[ ptb ] corrpap18.eps in contrast to the previous case we have a much smaller @xmath65 value and an only slightly reduced threshold @xmath36 .
[ ptbptb ] corrpap19.eps [ ptbptbptb ] corrpap20.eps [ ptbptbptbptb ] corrpap21.eps [ ptbptbptbptbptb ] corrpap22.eps [ ptbptbptbptbptbptb ] corrpap23.eps [ ptbptbptbptbptbptbptb ] corrpap24.eps [ ptbptbptbptbptbptbptbptb ] corrpap25.eps [ ptbptbptbptbptbptbptbptbptb ] corrpap26.eps the network fp3 is extremely high clustered ( mean degree @xmath273 , mean triangle number @xmath274 and a total of @xmath256 nodes and @xmath275 edges ) and stays metastable with a very small corruption cluster for about 200 updates till it jumps by a factor @xmath0 to another metastable state .
fig.[4 ] gives a more detailed view of the accumulated contributions by the different processes for a time interval around the jump in prevalence . in the initial phase
the @xmath244process was dominating the @xmath181process and vice versa in the second phase .
the next pictures show a situation where after an initial phase of slow growth a corruption collapse happened .
it seems that the absolute threshold value @xmath276 is well below the critical value where the system can still stabilize .
it is surprising that the system semi - stabilizes after an initial rapid increase in the prevalence for a rather long time ( fig.[6 ] ) . to a certain extend the results
can be explained along the argumentation in section 7 . in fig.[800 ] the accumulated infection processes for the initial phase are shown . here the @xmath20- process , although undercritical , causes a redistribution of infection till a clustered configuration is reached such that the @xmath64- process can start .
than the systems stays in almost complete balance till the @xmath181 process ( which is slow in all our examples ) wins ( fig.[801 ] ) .
note the difference to fig.[4 ] , where the @xmath181 process never really contributes to the infection .
finally we show two simulations for a sample of a random set graph model with about 1000 vertices ( fig.[20 ] and fig.[21 ] ) .
although both prevalence curves look similar there is a clear difference in the process fine - structure ( fig.[22 ] and fig.[23 ] ) .
in the first instance the @xmath244 and @xmath181 process are causing the collapse whereas in the second case the @xmath65 process in conjunction with the @xmath244 process is the main booster .
the few examples of single simulation runs given in this section show already , that there are many different routes to obtain high prevalence in corruption typically interrupted by long phases of metastability .
similar to other complex systems with hidden phase transitions ( e.g. the climate ) there can be an unnoticed small accumulation of infection till a critical density- a point of no return- of corruption is reached from which on an almost complete saturation of the society ( or a corresponding subsystem ) by corruption becomes the normality .
one of the basic question in classical epidemics as well as in corruption dynamics is : what can be done to slow down the `` infection '' propagation or prevalence . knowing the different phase transitions and their dependency on structure properties and social parameters is of great help in designing proper prevention scenarios . in the following
we will try to relate some of the findings from our model to what is considered by practicians as useful in corruption reduction .
first we would like to emphasize again that the present model deals in a rather abstract way with the propagation of mental willingness to be corrupt and not so much with realized corruption which always requires a specific environment and additional structural assumptions . hence concerning corruption control
, we only will be able to support certain prevention scenarios in the sense , that they go into the right direction and that there effect is strong or weak but without being able to make quantitative statements .
the model presented in this paper contains , besides structural parameters for the underlying network , @xmath134 relevant parameters : @xmath64- characterizing the strength of the local threshold process , @xmath78- characterizing the strength of the mean field attraction on becoming corrupt , @xmath79- the strength of the `` society strikes back '' term , @xmath20- the strength of the classical epidemic process ( assumed to be very small ) and @xmath36- the height of the local threshold .
three of the parameters- @xmath277 and @xmath20- are positively correlated to the spread of corruption whereas @xmath73 parameters- @xmath36 and @xmath79- are negatively correlated .
as is well known from classical epidemic control for infectious diseases it is very hard if not impossible to change basic social parameters in a short time .
this can only be achieved in a long running educational process .
therefore not much can be done in avoiding high clustering in certain relevant areas of the society in order to prevent the emergence of highly connected corruption nets .
as the name already indicates , transparency international favours as an effective tool to decrease corruption especially the increase of transparency in all forms of administrative decision making as well as transparency in financial affairs of socially exposed persons , institutions and companies .
the effect of an increase of transparency translates into our model as an increase of the value of @xmath36 and a decrease of the values of @xmath278 and @xmath20 .
strengthening of justice , police and similar instruments to fight and uncover corruption has again the effect of lowering @xmath78 ( via increase of fear ) but may also increase the value of @xmath79 ( uncovering rate ) . since an increase of @xmath79 above the value of @xmath78 and @xmath64 would perhaps require a total police state
, it is illusionary to overcome corruption just by means of law , justice and police .
besides necessary long term educational efforts in school and public to strengthen the moral resistance against corruption ( increase of @xmath36 and decrease of @xmath64 ) it seems a good strategy to make administrative and political decision hierarchies as independent and decentralized as possible to avoid high clustering .
we would like to end these short remarks by a few comments on the role of hubs - the very high degree vertices typically present in scale free graphs - in corruption dynamics .
while a priori not especially well suited to transmit corruption via the @xmath65 process due to the local tree like structure around the hubs ( compared with low degree vertices ) they nevertheless are more often exposed to corruption and have therefore a higher probability to get corrupt .
if the hub density is sufficiently high ( as is the case for scale - free degree distributions with exponent @xmath97 ) and the degree correlation is stronger than additive many vertices are linked to the hubs via social dependencies and in turn also can get corrupt .
furthermore they may play a fatal role in increasing the weighted corruption density relevant for the mean field process as was explained at the end of section 7 .
the described situation is probably typical for strongly hierarchically organized countries or regional substructures e.g. systems with a dictatorial or monarchical tendency . in such societies
a high prevalence of corruption seems almost unavoidable since the threshold @xmath68 is close to zero . for democratic societies
it seems therefore wise , to watch the behavior of hubs- whatever their social interpretation might be- more intensively than the `` normal '' part of the society .
in this article we have presented a first study of the spread of corruption on scale free and highly clustered networks .
one of the main observations so far is the strong dependence of the asymptotic dynamics on the initial number of corrupt individuals .
this holds as well for the mean field process as for the local dynamics .
second there is a fatal resonance effect between global and local dynamics lowering dramatically the critical density of initial infection .
as expected there is a positive correlation between clustering and spread of corruption respectively the critical initial density .
scale - freenes seems to play an important role for the corruption process for distributions with small exponent ( @xmath97 ) and multiplicative degree correlation due to the high prevalence of infected hubs and the strong linkage of medium and low degree vertices to them . for higher exponents the dynamics
is rather insensitive to the degree distribution .
the strength of the degree correlation ( from weak - additive till strong - multiplicative or even higher powers ) in networks of social acquaintances seems to be related to the political and institutional structure of a society which favours liberal organization forms as being less vulnerable to corruption .
there is a whole bunch of natural continuations or generalizations which have to be investigated next .
clearly a deeper understanding of the pure @xmath65 process and its phase transitions is necessary .
the mathematical problem is already highly nontrivial on trees .
the following short list gives a selection of natural generalizations and refinements : besides the specific context of corruption dynamics there is a multitude of topics where the model presented in this paper could easily be adopted to .
this includes so different themes as political opinion formation , social disorder processes , strategies for advertisement , doping usage , the spread of prejudices , migration dynamics , global terrorist networks and innovation processes . in all these examples
one has a local and global dynamics very similar to the one described here .
of course there are differences .
for instance in many mind formation problems the state space of individuals is rather complex and the local dynamics allows for many transitions not just 0 - 1 as in the corruption model .
furthermore aging phenomena and limits of resources could be included .
but besides this addition of structure and complexity and the various interpretations there remains a good part of the findings of this work to be true .
there will be phase transitions in the initial density of certain properties and there can be resonance effects between the nonlinear global and local dynamics - both making the prediction of future difficult and challenging
. we would like to thank for the support of the volkswagen foundation , the dfg - research group 399 `` spectral analysis , asymptotic distributions and stochastic dynamics '' and the austrian research center seibersdorf for providing us with the data set of eu - funded research projects .
p. martin especially thanks klaus geppert , detlef leenen , christian pestalozza and albrecht randelzhofer for stimulating discussions .
blanchard , c. chang , t. krueger : _ epidemic thresholds on scale - free graphs : the interplay between exponent and preferential choice _ , annales henri poincar vol . 4 , suppl.2 , ( 2003 ) , arxiv : cond - mat/0207319 j. fill , e. scheinerman , k. singer - cohen : _ random intersection graphs when m=(n ) : an equivalence theorem relating the evolution of the g(n , m , p ) and g(n , p ) models _ , random structures and algorithms * 16 * , 2 ( 2000 ) | we study corruption as a generalized epidemic process on the graph of social relationships .
the main difference to classical epidemic processes is the strong nonlinear dependence of the transmission probability on the local density of corruption and the mean field influence of the overall corruption in the society .
network clustering and the degree - degree correlation play an essential role in corruption dynamics .
we discuss phase transitions , the influence of the graph structure and the implications for epidemic control .
structural and dynamical arguments are given why strongly hierarchically organized societies like systems with dictatorial tendency are more vulnerable to corruption than democracies .
a similar type of modelling can be applied to other social contagion spreading processes like opinion formation , doping usage , social disorders or innovation dynamics . |
topological phases of matters have attracted a lot of attention in recent years due to discovery of graphene @xcite and topological insulators @xcite .
also , the search for conventional and unconventional topological superconductors , hosting topological phases @xcite , has become one of the growing topics in condensed - matter physics @xcite . as long as underlying symmetries in the topological systems
are preserved , the symmetry protected gapless edge or surface states within gapped bulk states are robust against many forms of perturbations , making relevant requirement for quantum electronic devices and , especially , topological quantum computations @xcite .
one - dimensional ( 1d ) heterostructures @xcite consisting of ingredients such as superconductors , topological insulators , spin - orbit coupled semiconductors and ferromagnets can support topologically nontrivial phases hosting , for instance , non - abelian majorana bound states at zero energy levels @xcite and fractionally charged fermion bound states @xcite . furthermore
, 1d majorana chain has been proposed to include majorana fermions @xcite at its boundaries .
in contrast , one of the simplest 1d topological insulators is known as the su - schrieffer - heeger ( ssh ) model with the bdi symmetry class which is proposed for polyacetylene @xcite .
ssh model has been shown to exhibit a diversity of nontrivial topological phases @xcite . in this context ,
most of the recent studies have been devoted to investigating the effects of complex boundary potentials @xcite , next - nearest - neighbor hopping @xcite , superconducting correlations @xcite , disorders @xcite , time - dependent potential @xcite , magnetic flux @xcite , modulated on - site potentials @xcite , and hopping amplitudes @xcite on topological properties of the 1d topological insulators . in theoretical and experimental studies ,
both magnetic field @xcite and spin - orbit coupling @xcite are important keys for the existence of nontrivial phases .
it has been recently shown that a spatially varying magnetic field can be served as an effective spin - orbit interaction leading to creating flat bands of majorana states @xcite and fractionalization of charged bound states @xcite .
it has also been studied the effect of modulated spin - orbit coupling @xcite on trivial and nontrivial regimes of ssh model @xcite . however , the interplay effect of zeeman fields and spin - orbit coupling on topological properties of 1d superlattices @xcite and , in particular , ssh model has received only limited attention
. in experimentally realizable cases , fine - tuning of parameters of a system to certain values through external fields is not easily reachable , particularly , when dealing with nontrivial topological phases in meso- and nanoscopic platforms @xcite .
so , it is important to provide a much wider range of required model parameters into reach that facilitates experimental observations of desired phases . generally speaking ,
flexibility in increasing or decreasing the range of less important nontrivial or trivial topological phases in favour of a desired topological phase is practically one of the most important requirements of topological quantum computations such that the topological class of the system remains intact .
therefore , an interesting question is how to increase the nontrivial topological region of 1d dimerized lattice in space of parameters with the combined effects of spin - orbit coupling and zeeman field respecting underlying symmetries . in this work
, we analyze the effect of perpendicular and parallel zeeman fields on topological phases of a ssh chain with modulated spin - orbit coupling @xcite .
if the zeeman fields are absent , trivial regions , where the system is an ordinary insulator , can be found in weak and strong modulated spin - orbit coupling regimes .
furthermore , there exist two types of topologically nontrivial phases characterized by one or two pairs of zero - energy edge states .
we show that in the presence of uniform ( staggered ) perpendicular zeeman field for weak ( strong ) modulated spin - orbit coupling , the trivial region would be turned into a nontrivial topological one through a topological phase transition by increasing the magnitude of the zeeman field that results in the emergence of a pair of zero - energy edge states
. moreover , the perpendicularly applied uniform ( staggered ) zeeman field in the weak ( strong ) modulated spin - orbit coupling regime causes the nontrivial topological phase with two pairs of zero - energy edge states to change its topology and turn into the topologically nontrivial phase hosting one pair of zero - energy states .
symmetry arguments show that the system in the presence of such perpendicular zeeman fields possesses inversion , chiral , particle - hole and effective time - reversal symmetries which still falls in the bdi class with topological number @xmath1 .
we also calculate analytically the winding number through bulk properties of quantum states and show that the results of periodic boundary conditions ( pbcs ) are in good agreement with those of open boundary conditions ( obcs ) , according to bulk - edge correspondence .
further , we explore the robustness of the symmetry protected edge states by adding local perturbations such as parallel zeeman field and spin - dependent on - site potential .
these perturbations break the symmetries implying that the nontrivial topological phases are protected by inversion symmetry fundamentally .
this paper is organized as follows : in sec .
[ section1 ] , we introduce a tight - binding model of the system , and obtain bulk energy bands .
we also investigate symmetries of the system . in sec .
[ section2 ] , topological phase diagrams without the zeeman fields are studied . in secs .
[ section3 ] and [ section4 ] , the effects of uniform and staggered perpendicular zeeman fields , respectively , on the topological phases are studied by calculating of the energy spectrum and topological number of a finite system . for the winding number an analytic formula
is derived via bulk states in sec . [ winding ] .
we discuss the effect of symmetry breaking perturbations on the robustness of edge states in sec .
[ section symmetry ] and finally conclusions are presented in sec . [ section5 ] .
we consider a 1d dimerized lattice along the @xmath2-axis which contains two sublattices @xmath3 and @xmath4 in each unit cell with a lattice constant @xmath5 . the lattice is subjected to a modulated spin - orbit coupling @xcite and also both perpendicular and parallel
zeeman fields are applied .
thus , the total tight - binding hamiltonian describing the system is the sum of the hamiltonians of the ssh model , @xmath6 , the spin - orbit coupling , @xmath7 , and the zeeman fields , @xmath8 , as @xmath9 with @xmath10,\nonumber \\ \hat{h}_{so}&=&\sum\limits_{n,\sigma}[\lambda\hat{a}^\dagger_{n,\sigma}\hat{b}_{n,-\sigma}-\lambda^\prime\hat{a}^\dagger_{n,\sigma}\hat{b}_{n-1,-\sigma}+h.c.],\nonumber\\
\hat{h}_b&=&\sum_{n,\sigma,\sigma^\prime}[\hat{a}^\dagger_{n,\sigma}(\mathbf m_{n , a}\cdot\mathbf{\tau})_{\sigma\sigma^\prime}\hat{a}_{n,\sigma^\prime}\nonumber\\ & + & \hat{b}^\dagger_{n,\sigma}(\mathbf m_{n , b}\cdot { \bf\tau})_{\sigma\sigma^\prime}\hat{b}_{n,\sigma^\prime}],\nonumber\end{aligned}\ ] ] where @xmath11 and @xmath12 ( @xmath13 ) are the fermion creation ( annihilation ) operators of electrons with spin @xmath14 on the sublattices @xmath3 and @xmath4 of the _ n_th unit cell , respectively . in addition , @xmath15 and @xmath16 ( @xmath17 and @xmath18 ) denote the hopping ( spin - orbit coupling ) amplitudes in the unit cell and between two adjacent unit cells , respectively , with the dimerization strength @xmath19 , @xmath20 is the pauli vector acting on the spin subspace and the zeeman field vector on the _
n_th unit cell is @xmath21 with sublattice index @xmath22 . here ,
since the lattice is invariant under rotations about the @xmath2-axis , without loss of generality , we apply the perpendicular zeeman field along the @xmath23-direction , _ i.e. _ , @xmath24 . furthermore , for simplicity , the zeeman field is supposed to be identical in each unit cell , so we drop its unit cell index @xmath25 hereafter .
we shall later investigate the effect of the zeeman field in the @xmath2-direction , therefore we include the zeeman field as @xmath26 .
we choose @xmath27 as the energy unit and the lattice constant @xmath28 .
@xmath29 indicates inter - cell ( intra - cell ) hopping energy.,width=321 ] adopting pbcs and fourier transforming , the total hamiltonian @xmath30 , eq .
( [ full hamiltonian ] ) , can be easily written as @xmath31 where @xmath32 , and @xmath33 with @xmath34 we diagonalize the hamiltonian in the momentum space , eq .
( [ hk - matrice ] ) , to obtain the eigenvalues as @xmath35 with @xmath36 where @xmath37 indicates the conduction ( valance ) band and @xmath38 represents upper ( lower ) subband .
it is well studied that a topological phase transition accompanying by closing and reopening of bulk band gap always determines the boundaries between different topological phases in the absence of interactions @xcite .
the gap closing conditions of @xmath39 can be obtained by calculating @xmath40 . in the absence of both the perpendicular zeeman field and the modulated spin - orbit coupling
, the energy spectrum reduces to @xmath41 and the energy gap closes at the momentum @xmath42 if @xmath43 . on the other hand , in the presence of the modulated spin - orbit coupling and @xmath44 , the energy gap closure conditions are @xcite @xmath45 and @xmath46 at the boundaries of the brillouin zone @xmath47 and momentum @xmath48 , respectively .
but in the presence of both the @xmath23-component of zeeman field and the modulated spin - orbit coupling , we find that the gap closes at the momenta @xmath47 and @xmath48 with the corresponding conditions , respectively , given by @xmath49 and @xmath50 in fact , the above relations are the generalized versions of eqs .
( [ gap clo k=0 ] ) and ( [ gap clo k = pi ] ) .
this means that the boundaries between topological phases can be changed drastically , as a result of the perpendicular zeeman field .
importantly , the sign of right hand sides of eqs .
( [ gap closure k=0 ] ) and ( [ gap closure k = pi ] ) can be either positive , namely , @xmath51 or negative , namely , @xmath52 depending on the values of the parameters .
we refer to the former ( latter ) case , _
i.e. _ , relation ( [ weakspin - orbit ] ) ( ( [ strongspin - orbit ] ) ) as the weak ( strong ) modulated spin - orbit coupling case .
consequently , the weak ( strong ) modulated spin - orbit coupling requires @xmath53 ( @xmath54 ) .
in addition , there are some parameter regions where only one of the inequalities from each relations ( [ weakspin - orbit ] ) and ( [ strongspin - orbit ] ) can be fulfilled .
we use the phrase ` mixed regime ' to refer to these cases . therefore , the perpendicular zeeman fields of the sublattices @xmath3 and @xmath4 should be either parallel or antiparallel with respect to each other in the unit cell as depicted in fig .
[ fig1fig2 ] .
subsequently , one may anticipate that uniform and staggered zeeman fields not only satisfy the energy gap closing conditions , [ see eqs.([gap closure k=0 ] ) and ( [ gap closure k = pi ] ) ] but also preserve some symmetries of the system which will be discussed below .
the system shows particle - hole and chiral symmetries defined respectively as @xmath55 and @xmath56 where @xmath57 and @xmath58 with @xmath59 , @xmath60 , and @xmath61 being the identity matrix , the complex conjugation , and the @xmath2-component of the pauli vector @xmath62 , respectively . also , @xmath63 is the @xmath64-component of the pauli vector @xmath65 acting on sublattice space . because of the fact that @xmath66 , the anti - unitary effective time - reversal operator can be determined as @xmath67 .
these symmetry operators also exhibit the features that @xmath68 , @xmath69 , and @xmath70 . as a result of the last property
, there is no kramer s degeneracy related to effective time - reversal symmetry .
therefore , according to the standard altland - zirnbauer classification @xcite , the symmetry class is bdi in the presence of either uniform , @xmath71 , or staggered , @xmath72 , zeeman field .
moreover , when @xmath71 , the hamiltonian ( [ hk - matrice ] ) also satisfies the inversion symmetry relation @xmath73 , where @xmath74 with @xmath75 being the @xmath2-component of @xmath65 . on the other hand ,
if @xmath76 , it can be clearly seen that @xmath39 fulfils @xmath77 under the inversion symmetry operation @xmath78 .
notice that due to inversion and reflection symmetries , topological classifications fall beyond the standard classification @xcite which are based on the global symmetries , _
i.e. _ , time - reversal , particle - hole , and sublattice ( chiral ) symmetries . in cases satisfying to the modified topological classifications @xcite , the inversion symmetry operator commutes with all the global symmetries @xcite .
in contrast , in our case , interestingly , the inversion operators @xmath79 and @xmath80 anti - commute with both chiral and particle - hole operators but commute with effective time - reversal operator . note that a one - dimensional system belonging to the bdi class has either @xmath1 or @xmath81 topological index depending on algebraic relations between inversion operator and other global symmetry operators @xcite . according to the modified periodic table @xcite , the algebraic relations of our system
imply that topological index is @xmath1 .
in order to understand the effect of the perpendicular zeeman fields on the topological properties of the system , the topological phase diagram in the absence of such fields is first investigated by calculating topological invariant @xcite numerically .
the relevant topological invariant for 1d system can be expressed by the topological number @xmath82 where the zak phase @xcite is @xmath83 with @xmath84 being the occupied bloch states @xcite . both quantized zak phase @xcite and symmetry - protected states localized on the boundaries of the system @xcite characterize the existence of nontrivial topological phase , emerging from bulk states properties and symmetry configurations .
the quantized zak phase has also been measured experimentally in 1d periodic potentials using ultracold atoms in optical lattices @xcite and periodic tubes in phononic crystals @xcite . without zeeman fields for parameters : ( a ) @xmath85 , ( b ) @xmath86 .
capital letters m , s , and w denote mixed , strong , and weak modulated spin - orbit coupling regimes .
topologically distinct regions @xmath87 and @xmath88 are depicted by gray ( green ) , light ( yellow ) , and dark ( red ) colors , respectively .
the relevant part of energy spectrum for different values of spin - orbit couplings under obcs with 50 unit cells is shown for ( c ) @xmath85 and ( d ) @xmath86 . in the middle of the energy gap ,
the spectrum contain zero- , two- , or four - fold degenerate states corresponding to topologically distinct insulating phases.,width=321 ] since the value of zak phase is gauge dependent , we follow the choice of unit cell so that the zak phase of bloch bands takes the values @xmath89 , @xmath90 or @xmath91 @xcite .
thus , the topological number @xmath92 corresponds to the existence of one pair ( two pairs ) of zero - energy states at the boundaries under obcs and indicates that the system is in the topologically nontrivial phase .
furthermore , when the number @xmath93 , the system is topologically trivial without zero - energy states .
the resulting topological phase diagrams in the plane @xmath94 for the case @xmath95 and @xmath96 are shown in figs . [ fig2](a ) and [ fig2](b ) , respectively .
the black solid lines are phase boundaries where the bulk band gap closes . topologically distinct regions @xmath97 and @xmath88 are indicated by gray ( green ) , light ( yellow ) , and dark ( red ) colors , respectively .
the different regimes of mixed , weak , and strong modulated spin - orbit couplings are denoted by @xmath98 " , @xmath99 " , and @xmath100 " , respectively , as well .
as one can see in fig .
[ fig2](a ) , there are three trivial regions , @xmath101 . while one of them is located in the weak modulated spin - orbit regime , the others are in the the strong modulated spin - orbit regime .
also , in the strong modulated spin - orbit and mixed regimes , there are nontrivial topological phases with different topological number @xmath1 . for the case of @xmath96 , there are two trivial regions in the strong modulated spin - orbit regime and the nontrivial topological phase with topological number @xmath102 can be found not only in the strong modulated spin - orbit regime but also in the weak modulated spin - orbit regime [ see fig . [ fig2](b ) ] . in both cases @xmath95 and @xmath96 , the topological phase with number @xmath103 is only restricted to the mixed regime .
notice that if @xmath104 , the phase diagrams are not symmetric with respect to the parameters @xmath17 and @xmath105 . to show the number of zero - energy states in connection with the distinct topological phases , we calculate the energy spectra of the 1d lattice dimerization with 50 unit cells under obcs for the cases @xmath95 and @xmath96 as presented in figs .
[ fig2](c ) and [ fig2](d ) , respectively . in either cases , in the middle of the energy gap , depending on the appropriate choice of parameters @xmath17 and @xmath105 in the three different regions of panels ( a ) and ( b ) , one can observe zero , one , or two pairs of zero - energy states related to topologically distinct insulating phases .
we consider the @xmath23-components of zeeman field terms on sublattices @xmath3 and @xmath4 which vary cyclically with the parameter @xmath106 as , @xmath107 where the amplitude @xmath108 and phase factor @xmath109 are parameters to control the strength and sign of the zeeman field on sublattice @xmath3 ( @xmath4 ) , respectively . without loss of generality , we assume @xmath110 . also , we take @xmath111 and @xmath112 or @xmath113 with @xmath114 an integer , as required by the inversion symmetry . as discussed already above , when @xmath95 , in the absence of zeeman fields , the system is always gapped and subsequently , a trivial insulator in the weak modulated spin - orbit coupling regime .
thus , in order to close the energy gap and change the topology of the band structure , we must apply a uniform perpendicular zeeman field [ see eqs .
( [ gap closure k=0 ] ) and ( [ gap closure k = pi ] ) ] .
it s straightforward to see that for @xmath115 , which establishes the uniform zeeman field , closing and reopening of the bulk gap take place at two certain values of the zeeman field magnitude given by @xmath116 within which this uniform zeeman field drives the system into topologically nontrivial phase .
therefore , the topological phases caused by zeeman field can be characterized by the @xmath0 such that the quantized zak phase @xmath90 ( @xmath117 ) manifests the existence of a pair of zero - energy edge states under obcs .
the dependence of energy spectrum on @xmath118 is shown in fig . [ fig3](a ) for @xmath95 with @xmath119 and @xmath120 , _ i.e. _ , weak modulated spin - orbit coupling regime .
low energy states associated with edge states are indicated by thick solid ( red ) lines .
one can see that within a certain range of @xmath118 the lowest energy states become zero - energy states . in this region ,
the value of zak phase is @xmath90 as shown in fig .
[ fig3](b ) indicating the establishment of nontrivial topological phase .
outside of this region , the zak phase takes @xmath89 value and the energy gap is open as well .
the induced nontrivial topological phase , due to perpendicular zeeman field , stems from spin - splitting of the helical basis projected into a 1d bipartite lattice . under obcs in presence of uniform zeeman field with 20 unit cells .
( b ) probability distribution for two different values @xmath121 and @xmath122 . here , @xmath85 , @xmath119 , @xmath123 , and @xmath124 .
] the energy spectrum as a function of @xmath106 is plotted in fig . [ fig3](c ) for @xmath95 with parameters @xmath125 and @xmath120 .
there is an energy gap for small @xmath106 .
as we increase @xmath106 in magnitude , a phase transition occurs , consequently , the gap closes and flat bands appear . with further increase of @xmath106 ,
flat bands disappear and gap reopens .
naturally , the spectrum is symmetric about @xmath126 arising from reflection symmetry with respect to the @xmath127-plane . the corresponding zak phase is shown in fig . [ fig3](d ) .
one can see that in the parameter region in which the flat bands are appeared , the zak phase takes @xmath90 value indicating the existence of zero - energy edge states , whereas , in the other region , the zak phase is 0 exhibiting a topologically trivial phase . upon applying the uniform perpendicular zeeman field , both pairs of edge states for the case of @xmath96 in the weak spin - orbit coupling become unstable and
gap opens , as shown in fig .
[ fig3](e ) . with the increase of the @xmath118 , while one of the pair states remains unstable , interestingly , the other one becomes stable ; as a result , the system undergoes a topological phase transition from a normal insulator to a topological insulator .
moreover , at large @xmath118 , since only one of the spin species will be dominated , the possibility of establishing the edge states containing both types of spins states vanishes and subsequently gap opens for the second time .
the corresponding zak phase as a function @xmath118 is illustrated in fig .
[ fig3](f ) .
one can see that if the zero - energy states are present , then the zak phase shows a nontrivial topology .
it is worthwhile noting that , in the case of weak spin - orbit coupling for both @xmath95 and @xmath96 , if @xmath128 , one obtains @xmath129 ( @xmath93 ) , thus , the system is always gapped . in fig .
[ fig4](a ) , the energy spectrum versus @xmath118 is shown for mixed regime with parameters @xmath130 , @xmath123 , and @xmath119 ( the result of case @xmath131 is the same and not shown ) . in this regime , for both cases @xmath96 and @xmath95 , the uniform perpendicular zeeman field preserves the two - fold degenerate zero - energy edge state from zero up to a certain value of the zeeman field strength .
this can be understood as follows . in the mixed region ,
equal strengths of inter- and intra cell spin - orbit couplings with almost large magnitudes cause that the band structure around the fermi surface can be affected slightly by small or mediate zeeman field strengths without losing its topological features .
as described above , increasing the zeeman strength beyond the critical value , completely splits opposite spin states resulting in the disappearance of zero - energy states . in order to confirm that the zero - energy states exhibit themselves as localized states at sample boundaries , the probability distribution of the lowest energy states for two different values @xmath132 and @xmath133 with the same parameters as fig . [
fig4](a ) is depicted in panel ( b ) . in the presence of zero - energy states , for @xmath132 ,
the probability distribution has maxima at the two boundaries of the chain , while for @xmath134 , minimum values of probability distribution occur at such points .
in this section we will first investigate the effect of perpendicular zeeman field on different topological phases occurred in the strong modulated spin - orbit coupling regime . as discussed in sec .
[ section2 ] , in the absence of zeeman fields , for both cases @xmath95 and @xmath96 , when inter - cell spin - orbit coupling is larger than intra - cell one the system is a trivial insulator . to change the topology of these regions from trivial phase into a nontrivial one ,
it is necessary to close and then reopen the energy gap with the help of staggered perpendicular zeeman field [ see eqs .
( [ gap closure k=0 ] ) and ( [ gap closure k = pi ] ) ] .
it can be easily checked that for @xmath128 , providing the staggered zeeman field , within the range @xmath135 the phase of trivial region turns into the nontrivial . here ,
we have defined , @xmath136 the evolution of energy spectrum as functions of @xmath118 and @xmath106 is depicted in figs .
[ fig5](a ) with @xmath119 and [ fig5](c ) with @xmath137 , respectively , for @xmath86 , @xmath138 , @xmath139 , and @xmath140 .
the bulk states of the eigenvalues , depicted by thin solid black lines , have a different structure compared to the uniform perpendicular zeeman field case [ see figs . [ fig3](a ) and [ fig3](c ) ] , whereas the lowest energy states of eigenvalues ( thick solid red lines ) including the boundary states reveal a broadly similar behavior .
also , the evolution of zak phase as functions of @xmath118 and @xmath106 with the same parameters as figs . [ fig5](a ) and [ fig5](c ) is presented in figs . [ fig5](b ) and [ fig5](d ) , respectively .
they show that zero - energy states have nontrivial characteristics .
also , the @xmath118 dependence of energy spectrum in the strong modulated spin - orbit coupling regime containing four - fold degenerate boundary states is shown in fig . [ fig5](e ) . as can be seen
, there is only a four - fold degenerate zero - energy state at @xmath141 .
similar to fig .
[ fig3](e ) , as we increase the strength of staggered perpendicular zeeman field , the system goes to a trivial phase such that the lower and upper bulk bands separate with an energy gap in the parameter regime @xmath142 . by further increasing the strength of zeeman field ,
two branches of low energy states ( thick solid red lines ) merge together again in the parameter regime @xmath143 as zero - energy flat bands ( @xmath144 . thus , the unstable four - fold degenerate nontrivial topological phase in the strong modulated spin - orbit coupling regime for @xmath95 ( even for @xmath96 ) is changed to the two - fold degenerate topologically nontrivial phase by @xmath118 though undergoing a topological phase transition . also , in fig .
[ fig5](f ) , the @xmath118 dependence of zak phase of panel ( e ) is shown .
notice that when @xmath115 , the system is always gapped and a trivial insulator in both cases @xmath95 and @xmath96 . under obcs with 20 unit cells in the presence of
staggered perpendicular zeeman field .
( b ) probability distribution for two different values @xmath145 and @xmath146 . here , @xmath85 , @xmath119 , @xmath123 , and @xmath140 .
] in the mixed regime , for @xmath147 , the two - fold degenerate nontrivial phase still remains nontrivial in the presence of staggered perpendicular zeeman field . the behavior of energy states is depicted in fig .
[ fig6](a ) which is similar to the one described in the case of uniform zeeman field .
however , in contrast to the case presented in fig .
[ fig4](a ) , one can observe that zero - energy states can sustain larger staggered zeeman field strengths than that of the uniform zeeman field case .
this can be traced back to the specifically staggered magnetic field , having a zero net magnitude in each unit cell , that only deforms the spectrum such that a large value of strength can make a significant effect .
figure [ fig6](b ) shows the probability distributions of lowest energy states for different values of @xmath118 in the presence of staggered perpendicular zeeman field for the mixed regime .
in addition to the zak phase , a useful quantity for discussing the bulk features of quantum states is a winding number .
it can be used to distinguish different topological phases determining the number of pairs of zero - energy edge states .
the definition of the winding number is relevant in the case of block off - diagonal hamiltonians .
a well - studied feature of systems with the chiral symmetry class , arising from the sublattice or chiral symmetry , is that hamiltonians describing of such systems can be brought into block off - diagonal form in the basis of chiral operator @xcite . obviously , in this basis , the chiral operator @xmath148 must be diagonalized , @xmath149 , via the unitary matrix , @xmath150 it is possible to transform the hamiltonian ( [ hk - matrice ] ) by the unitary transformation @xmath151 , yielding @xmath152 with @xmath153 therefore , the block off - diagonal representation of the hamiltonian , eq .
( [ off - diagonal ] ) , allows for the definition of the chiral index as @xmath154 with @xmath155cos(k)\nonumber \\ & -2i[(t+\delta t)\lambda+(t-\delta t)\lambda^\prime]sin(k)\nonumber \\ & -2(t^2+\delta t^2)+m_{y , a}m_{y , b}+\lambda^2+\lambda^{\prime 2}.\end{aligned}\ ] ] @xmath99 is the winding number of @xmath156 characterizing topologically distinct phases .
the integration over @xmath157 in eq .
( [ winding number ] ) can be performed with the use of cauchy s residue theorem , leading to a simple analytical formula for the winding number . for weak and strong modulated spin - orbit coupling regimes as well as
mixed regime in the presence of uniform ( staggered ) perpendicular zeeman field , we find @xmath99 as @xmath158 where @xmath159 is the heaviside theta function and @xmath160 means the real part .
@xmath161 and @xmath162 are defined by eqs .
( [ unizeeman1 ] ) and ( [ unizeeman2 ] ) ( eqs .
( [ stagzeeman1 ] ) and ( [ stagzeeman2 ] ) ) .
@xmath163 denotes nontrivial topological region which manifest itself by a pair of degenerate zero - energy boundary states under obcs and @xmath164 denotes trivial region in which the system is an ordinary insulator .
notice that , the analytical results support the above - obtained numerical topological phases quite well .
now , let us discuss the stability of the topological phase by considering local perturbations which remove the symmetries of the system . in the nontrivial topological region , in addition to either uniform or staggered @xmath23-component of zeeman field ,
if we introduce the zeeman field along the @xmath2-axis with the same direction and strength on the sublattices , @xmath165 , then the topological phases fall to the ai class .
this is because of breaking of chiral and particle - hole symmetries and preserving of effective time - reversal symmetry .
subsequently , the mid - gap edge states under obcs are no longer stable and shift away from zero energy value as shown in fig .
[ fig10](a ) .
moreover , in the absence of @xmath23-component of zeeman field , the uniform zeeman field along the @xmath2-direction causes the chain to be a nontrivial insulator with zak phase @xmath166 only in the mixed regime under the conditions @xmath167 as discussed in the previous secs .
[ section3 ] and [ section4 ] , the system is an ordinary insulator in the presence of uniform ( staggered ) zeeman field along the @xmath23-axis in the weak ( strong ) spin - orbit coupling regime within the range @xmath168 .
interestingly , in this range , we can close and reopen the energy gap to change the topology of the system with the aid of @xmath2-component of uniform zeeman field .
the energy gap closure happens in the presence of uniform zeeman field along the @xmath2-axis and @xmath23-component of staggered zeeman field with conditions @xmath169 hence , the chain is nontrivial insulator @xmath103 within the range @xmath170 . in fig .
[ fig10](b ) the energy spectrum is plotted versus @xmath171 for @xmath165 and @xmath172 . at @xmath173 ,
the chain is an ordinary insulator .
as we increase @xmath171 , the gap closes at @xmath174 and low energy states merge together .
if the @xmath2-component of zeeman field exceeds the certain value @xmath175 , the bulk states join to the mid - gap states and then gap reopens .
similarly , the energy gap closure conditions in the presence of @xmath2- and @xmath23-component of uniform zeeman fields take place at @xmath176 the chain has a nontrivial topology with the condition @xmath177 . for this case , the energy spectrum is plotted in fig .
[ fig10](c ) which exhibits a nontrivial phase in the parameter space @xmath178 .
notice that for zero value of the @xmath2-component of zeeman field , the relations ( [ x1 ] ) and ( [ x2 ] ) reduce to the ones presented in eqs .
( [ gap closure k=0 ] ) and ( [ gap closure k = pi ] ) . in order to investigate the effect of effective time - reversal symmetry on the stability of the edge states , we add the following perturbation to the total hamiltonian @xmath179 where @xmath180 ( @xmath181 ) is the strength of on - site potential .
this perturbation breaks not only both particle - hole and chiral symmetries but also effective time - reversal symmetry . as it is shown in fig .
[ fig10](d ) the @xmath2- and @xmath23-component of zeeman fields with perturbation @xmath182 preserve the mid - gap edge states but shift them away from zero energy . in this case ,
due to absence of effective time - reversal , chiral and particle - hole symmetries , the topological class is the a. + remarkably , it is easy to obtain that nonuniform zeeman field along the @xmath2-axis , @xmath183 , breaks the inversion symmetry even in the presence of either uniform or staggered perpendicular zeeman field . in fig . [ fig10](e)((f ) ) , the energy spectrum ( zak phase ) is depicted versus @xmath171 for different magnitudes of @xmath2-component of zeeman fields on each sublattice , _
i.e. _ , @xmath184 . in this case
, the edge states separate from each other and , subsequently , the zak phase does not have quantized values .
therefore , the above inspections insure that the topological edge states in the studied system , belonging to the bdi , ai , and a classes , are protected by the inversion symmetry .
we have investigated the mutual effects of zeeman field and modulated spin - orbit coupling on the topological characteristics of 1d dimerized lattice .
it is shown that 1d lattice dimerization is a trivial insulator in different parameter regions of weak and strong modulated spin - orbit coupling .
we have also showed that with the aid of uniform perpendicular zeeman field for weak modulated spin - orbit coupling and staggered perpendicular zeeman field for strong modulated spin - orbit coupling , the trivial phases change to a nontrivial topological phase with a topological phase transition .
therefore , as compared with the bare ssh model , the region of nontrivial topological phase in the parameter space can be extended by adding the modulated spin - orbit coupling and the perpendicular zeeman field to the ssh model .
this nontrivial topological phase exhibits one pair of zero - energy edge states obeying non - abelian statistics that is a key requirement in topological quantum computing .
the topological phases of the system are classified with @xmath0 number and belongs to the bdi class .
we also have derived an analytical formula for the winding number through bulk quantum states to confirm our numerical results .
finally , the stability of the edge states to externally applied zeeman field along the chain and on - site spin - dependent potential has also been investigated .
the inclusion of these effects breaks chiral , particle - hole , effective time - reversal , and inversion symmetries resulting in that the edge states are topologically protected and robust against the perturbations as long as the inversion symmetry requirement is satisfied .
we remark that the discussion presented here is an example of the general issue implying that required experimentally accessible regions in parameter space may be engineered , in order to reach desired phases efficiently .
the authors would like to thank h. guo , y .- m .
lu , and zh .
yan for valuable comments .
we are also grateful to a. najafi for reading the manuscript carefully .
this work is partially supported by iran science elites federation under grant no .
11/66332 .
y. s. hor , a. j. williams , j. g. checkelsky , p. roushan , j. seo , q. xu , h. w. zandbergen , a. yazdani , n. p. ong , and r. j. cava , phys .
lett . * 104 * , 057001 ( 2010 ) ; s. sasaki , m. kriener , k. segawa , k. yada , y. tanaka , m. sato , and y. ando , phys .
lett . * 107 * , 217001 ( 2011 ) ; j. j. he , j. wu , t .-
choy , x .- j .
liu , y. tanaka , and k. t. law , nat .
commun . * 5 * , 3232 ( 2014 ) ; v. s. pribiag , a. j. a. beukman , f. qu , m. c. cassidy , ch .
charpentier , w. wegscheider , and l. p. kouwenhoven , nat .
* 10 * , 593 ( 2015 ) ; y. kitaev , ann . phys .
( amsterdam ) * 303 * , 2 ( 2003 ) ; s. das sarma , m. freedman , and c. nayak , phys .
94 * , 166802 ( 2005 ) ; c. nayak , a. stern , m. freedman , and s. d. sarma , rev . mod
. phys . * 80 * , 1083 ( 2008 ) ; j. k. pachos and s. h. simon , new j. phys . * 16 * 065003 ( 2014 )
. j. d. sau , r. m. lutchyn , s. tewari , and s. das sarma , phys .
lett . * 104 * , 040502 ( 2010 ) ; j. alicea , phys .
b * 81 * , 125318 ( 2010 ) ; r. m. lutchyn , j. d. sau , and s. das sarma , phys . rev
. lett . * 105 * , 077001 ( 2010 ) ; y. oreg , g. refael , and f. von oppen , phys .
lett . * 105 * , 177002 ( 2010 ) ; j. d. sau , s. tewari , r. m. lutchyn , t. d. stanescu , and s. das sarma , phys .
b * 82 * , 214509 ( 2010 ) ; a. cook and m. franz , phys . rev .
b * 84 * , 201105(r ) ( 2011 ) ; s. b. chung , h .- j .
zhang , x .-
qi , and s .- c .
zhang , phys .
b * 84 * , 060510(r ) ( 2011 ) ; m. duckheim and p. w. brouwer , phys . rev .
b * 83 * , 054513 ( 2011 ) ; v. mourik , k. zuo , s. m. frolov , s. r. plissard , e. p.
a. m. bakkers , and l. p. kouwenhoven , science * 336 * , 1003 ( 2012 ) ; a. das , y. ronen , y. most , y. oreg , m. heiblum , and h. shtrikman , nat .
* 8 * , 887 ( 2012 ) ; m. t. deng , c. l. yu , g. y. huang , m. larsson , p. caroff , and h. q. xu , nano lett .
* 12 * , 6414 ( 2012 ) .
s. nadj - perge , i. k. drozdov , b. a. bernevig , and a. yazdani , phys .
b * 88 * , 020407(r ) ( 2013 ) ; s. nadj - perge , i. k. drozdov , j. li , h. chen , s. jeon , j. seo , a. h. macdonald , b. a. bernevig , and a. yazdani , science * 346 * , 602 ( 2014 ) .
f. wilczek , nat . phys . * 5 * , 614 ( 2009 ) ; m. franz , physics * 3 * , 24 ( 2010 ) ; e. m. stoudenmire , j. alicea , o. a. starykh , and m. p. a. fisher , phys . rev .
b * 84 * , 014503 ( 2011 ) ; j. alicea , rep .
. phys . * 75 * , 076501 ( 2012 ) .
s. kivelson and j. r. schrieffer , phys .
b * 25 * , 6447 ( 1982 ) ; r. rajaraman and j. s. bell , phys .
b * 116 * , 151 ( 1982 ) ; c. y. hou , c. chamon , and c. mudry , phys .
98 * , 186809 ( 2007 ) ; b. seradjeh , j. e. moore , and m. franz , phys . rev .
lett . * 103 * , 066402 ( 2009 ) ; l. santos , y. nishida , c. chamon , and c. mudry , phys .
b * 83 * , 104522 ( 2011 ) ; s. gangadharaiah , l. trifunovic , and d. loss , phys . rev . lett . *
108 * , 136803 ( 2012 ) .
h. takayama , y. r. lin - liu , and k. maki , phys .
b * 21 * , 2388 ( 1980 ) ; a. j. heeger , s. kiverson , j. r. schrieffer , and w. p. su , rev .
mod . phys . * 60 * , 781 ( 1988 ) ; j. ruostekoski , g. v. dunne , and j. javanainen , phys .
lett . * 88 * , 180401 ( 2002 ) .
b. braunecker , g. i. japaridze , j. klinovaja , and d. loss , phys .
b * 82 * , 045127 ( 2010 ) ; s. gangadharaiah , b. braunecker , p. simon , and d. loss , phys . rev .
lett . * 107 * , 036801 ( 2011 ) ; m. kjaergaard , k. wlms , and k. flensberg , phys .
b * 85 * , 020503 ( 2012 ) ; n. sedlmayr , j. m. aguiar - hualde , and c. bena , phys . rev .
b * 91 * , 115415 ( 2015 ) .
a. altland and m. r. zirnbauer , phys .
b * 55 * , 1142 ( 1997 ) ; a. y. kitaev , aip conf . proc . * 1134 * , 22 ( 2009 ) ; a. p. schnyder , s. ryu , a. furusaki , and a. w. w. ludwig , phys . rev .
b * 78 * , 195125 ( 2008 ) ; s. ryu , a. p. schnyder , a. furusaki , and a. w. w. ludwig , new j. phys .
* 12 * , 065010 ( 2010 ) ; x .- g .
wen , phys .
b * 85 * , 085103 ( 2012 ) .
t. l. hughes , e. prodan , and b. a. bernevig , phys .
b * 83 * , 245132 ( 2011 ) ; c .- k chiu , h. yao , and s. ryu , phys .
b * 88 * , 075142 ( 2013 ) ; ch .- k . chiu , j. c. y. teo , a. p. schnyder , sh .
ryu , arxiv:1505.03535 ; k. shiozaki and m. sato , phys .
b * 90 * , 165114 ( 2014 ) ; y. x. zhao , a. p. schnyder , and z. d. wang , arxiv:1601.07440v1 .
ryu and y. hatsugai , phys .
89 * , 077002 ( 2002 ) ; x. chen , z .- c .
gu , and x .-
wen , phys . rev .
b * 83 * , 035107 ( 2011 ) ; _ ibid _ * 84 * , 235128 ( 2011 ) ; x. chen , z .- c .
gu , z .- x .
liu , and x
wen , science * 338 * , 1604 ( 2012 ) ; l. fidkowski and a. kitaev , phys .
b * 81 * , 134509 ( 2010 ) ; a. m. turner , f. pollmann , and e. berg , phys . rev .
b * 83 * , 075102 ( 2011 ) . | we study theoretically the interplay effect of zeeman field and modulated spin - orbit coupling on topological properties of a one - dimensional dimerized lattice , known as su - schrieffer - heeger model .
we find that in the weak ( strong ) modulated spin - orbit coupling regime , trivial regions or nontrivial ones with two pairs of zero - energy states can be turned into nontrivial regions by applying a uniform ( staggered ) perpendicular zeeman field through a topological phase transition .
furthermore , the resulting nontrivial phase hosting a pair of zero - energy boundary states can survive within a certain range of the perpendicular zeeman field magnitude . due to the effective time - reversal , particle - hole , chiral , and inversion symmetries , in the presence of either uniform or staggered perpendicular zeeman field , the topological class of the system is bdi which can be characterized by @xmath0 index .
we also examine the robustness of the nontrivial phase by breaking the underlying symmetries giving rise that inversion symmetry plays an important role . |
the next generation of neutrino oscillation experiments has the potential to not only determine the remaining unknowns in the pmns matrix but also to measure its parameters with unprecedented precision .
this will mark the beginning of a period of high - precision neutrino physics , where the standard paradigms describing the neutrino sector will be put to proof and theoretical ideas about the origins of neutrino mass and leptonic flavour can be confronted with data .
one of the more popular beyond the standard model ideas applied to the neutrino sector is the introduction of a discrete flavour symmetry .
models based on this principle have been shown to be able to derive the observed structure of the pmns matrix from a small set of assumptions .
these models generally propose a discrete symmetry ( _ e.g. _ @xmath2 or @xmath3 ) which is broken spontaneously , leaving residual symmetries amongst the leptonic mass terms .
these symmetries reduce the degrees of freedom amongst the mixing parameters , generating a pattern of falsifiable predictions . by
hypothesizing which symmetries of the leptonic mass terms are residual , this idea can be used to reconstruct the flavour group independently of many model specific assumptions . in ref .
@xcite , we have shown that a quite general construction of this type ( first presented in ref .
@xcite ) leads to only @xmath4 viable models in light of the current global oscillation data .
these models fix a column of the pmns matrix which , under the assumption of unitarity , can be expressed in terms of two constraints on the pmns parameters : the _ atmospheric sum rule _ , relating @xmath5 to @xmath6 and @xmath0 , and the _ solar prediction _ , an expression for @xmath7 in terms of @xmath6 alone . in this contribution
, we shall discuss the parameter correlations of these models , and how they can be constrained by the next generation of high - precision oscillation experiments .
we shall employ the notation @xmath8 , @xmath9 and @xmath10 @xcite throughout .
.[tab : solar_pred]the solar predictions for the @xmath4 viable models identified in ref .
the model label denotes the flavour group and the pattern of breaking ; for details , see ref .
[ cols="^,^,^",options="header " , ]
the atmospheric sum rule can be written in a linearized form by @xmath11 where @xmath12 and @xmath13 are constants expressible in terms of the group theoretic parameters of each model . to test these relations , we require a strong precision on the parameter @xmath0 , which necessitates the consideration of the next generation of long - baseline experiments . these proposals seek to make accurate measurements of the appearance channels @xmath14 and @xmath15 , which are sensitive to the value of @xmath0 at a subdominant level .
although it remains a challenging measurement , two leading designs have been shown to offer significant sensitivity to @xmath0 : superbeams and neutrino factories .
such facilities would be able to constrain the atmospheric sum rules over a significant fraction of the available parameter space .
for example , an on - axis superbeam with a detector mass of @xmath16 kton ( @xmath17 kton ) and a baseline of @xmath18 km would be capable of excluding models with @xmath19 and @xmath20 for over @xmath21@xmath22 ( @xmath23@xmath24 ) of the parameter space , depending on the true value of @xmath5 @xcite .
in this section , we shall consider a circa @xmath1 km reactor experiment based on the jiangmen underground neutrino observatory ( juno ) @xcite and reactor experiment for neutrino oscillations ( reno-50 ) designs @xcite .
these facilities will be capable of high precision measurements of the @xmath25 disappearance probability .
the main goal of such experiments is to observe the subdominant oscillations whose phase depends upon the mass hierarchy .
however , they will also significantly increase the precision on the oscillation parameters @xmath7 , @xmath26 and @xmath27 , reducing their uncertainty to the sub - percent level .
the @xmath4 viable models identified in ref .
@xcite make @xmath28 distinct solar predictions , which are shown in table [ tab : solar_pred ] .
such precision will have a significant impact on the viability of the correlations predicted by flavour symmetric models . to understand the impact of these high precision measurements , we have performed a simulation based on the juno design to determine its ability to test the correlations shown in table [ tab : solar_pred ] . in our simulation
, we assume a @xmath29 kton liquid scintillator detector with a linear energy uncertainty of @xmath30 .
the juno facility will detect neutrinos from @xmath31 nearby reactors ; however , we model this by a single source at a baseline distance given by the power weighted average of @xmath32 km and a reactor power of @xmath33 gw @xcite .
we have normalised our spectrum to produce @xmath34 events , including a @xmath35 normalisation uncertainty . in fig .
[ fig : juno_sr ] , we show the allowed regions at @xmath36 significance for the models shown in table [ tab : solar_pred ] . we see that only two of the @xmath36 intervals overlap , which allows for a strong model discrimination .
the ability for juno to exclude these models independently of their atmospheric sum rules provides a great complementarity between the reactor and long - baseline programmes .
furthermore , the two indistinguishable models for juno predict very different atmospheric sum rules , @xmath37 where @xmath38 is the golden ratio , and we expect these to be distinguishable with a superbeam for most of the parameter space @xcite . the @xmath36 allowed regions for the solar predictions shown in table [ tab : solar_pred ] after @xmath28 years of data taking by juno.,width=359 ]
the next generation of neutrino oscillation experiments , with their focus on precision measurements of the underlying parameters , will allow certain classes of models with discrete flavour symmetries to be thoroughly tested . in ref .
@xcite , the role of a long - baseline superbeam experiment ( modelled after lbno or lbne ) has been shown to be able to exclude these correlations for a large fraction of parameter space . in this contribution
, we have highlighted the potential for experimental exclusion of these models at a circa @xmath1 km reactor experiment based on the juno facility . by testing the solar predictions to high accuracy
, such a facility will be able to independently distinguish between almost all models under consideration .
the complementarity between reactor and long - baseline experiments will provide a stringent test of the idea that residual symmetries are responsible for the structure of the pmns matrix . | models of leptonic flavour with discrete symmetries can provide an attractive explanation of the pattern of elements found in the leptonic mixing matrix .
the next generation of neutrino oscillation experiments will allow the mixing parameters to be tested to a new level of precision , crucially measuring the cp violating phase @xmath0 for the first time . in this contribution , we present results of a systematic survey of the predictions of a class of models based on residual discrete symmetries and the prospects for excluding such models at medium- and long - term oscillation experiments .
we place particular emphasis on the complementary role that a future circa @xmath1 km reactor experiment , _
e.g. _ juno , can play in constraining these models . |
the rotation curves of spiral galaxies provide one of the strongest lines of evidence establishing the need for dark matter in the universe ( e.g. , rubin , thonnard & ford 1980 ; bosma 1981 ) . an elaborate paradigm for dark matter
has subsequently developed : @xmath5cdm .
more recent work emphasizes larger scale constraints ( e.g. , tegmark et al .
2004 ; spergel et al .
an important question is whether this cosmic paradigm provides a satisfactory description of spiral galaxies .
a complete explanation of spiral galaxies in @xmath5cdm requires a comprehensive theory of galaxy formation .
this remains lacking .
indeed , there are a number of lingering problems on small ( galaxy ) scales .
these include the cusp - core problem ( e.g. , de blok et al .
2001a ) , the missing satellite problem ( e.g. , moore et al .
1999a ) , the dynamical friction time scale problem ( goerdt et al . 2006
; sanchez - salcedo , reyes - iturbide , & hernandez 2006 ) , and a whole suite of other incongruities and apparent contradictions that arise on the scale of individual galaxies ( mcgaugh & de blok 1998 ; sellwood & kosowsky 2001 ) . in this paper , we will concentrate on the amplitude of the rotation velocity that is observed for spiral galaxy dark matter halos and that predicted by @xmath5cdm .
considerable attention has focused on the density profile inferred for the dark matter halos of dwarf and low surface brightness ( lsb ) galaxies at small radii ( moore 1994 ; flores & primack 1994 ; ct et al .
2000 ; swaters et al . 2000 ; de blok et al .
2001a , b ; borriello & salucci 2001 ; salucci 2001 ; bolatto et al . 2002 ; marchesini et al . 2002 ; salucci et al . 2003 ; swaters et al .
2003a , b ; de blok et al .
2003 ; donato et al . 2004 ; simon et al .
2003 , 2005 ; gentile et al . 2004 , 2005 ; spekkens , giovanelli , & haynes 2005 ; kuzio de naray et al . 2006 ; zackrisson et al .
there are good reasons for this : lsb galaxies are dark matter dominated down to small radii , making the inference of their dark matter properties much less sensitive to the stellar mass than in higher surface brightness spirals . in the most straightforward analyses ,
the data appear to contradict a basic prediction of the @xmath5cdm structure formation scenario , that dark matter halos should have cusps at small radii ( a density distribution @xmath7 with @xmath8 : dubinski 1994 ; cole & lacey 1996 ; navarro , frenk , & white 1997 ( hereafter nfw ) ; tormen , bouchet , & white 1997 ; moore et al.1998 , 1999b ; tissera & dominguez - tenreiro 1998 ; nusser & sheth 1998 ; syer & white 1998 ; avila - reese , firmiani , & hernandez 1998 ; salvador - ole , solanes , & manrique 1998 ; jing 2000 ; jing & suto 2000 ; kull 1999 ; klypin et al .
2001 ; ricotti 2003 ; power et al . 2003 ; diemand et al .
the data are apparently more consistent with a nearly constant density core ( @xmath9 ) .
many different solutions to this problem have been proposed ( e.g. , el - zant , shlosman , & hoffman 2001 ; weinberg & katz 2002 ; merritt et al . 2004 ; ma & boylan - kolchin 2004 ; mashchenko , couchman , & wadsley 2006 ; tonini , lapi , & salucci 2006 ) .
interesting constraints also stem from the rotation curve data for high surface brightness ( hsb ) galaxies .
it is frequently possible to obtain quite high quality data for these bright objects , but the interpretation depends critically on the mass ascribed to baryons .
the stellar disks in these objects can be massive ( e.g. , debatista & sellwood 2000 ; salucci & burkert 2000 ; weiner , sellwood , & williams 2001 ; zavala et al .
2003 ; kassin , de jong , & weiner 2006 ) or even maximal ( e.g. , van albada & sancisi 1986 ; kent 1987 ; palunas & williams 2000 ; battaglia et al . 2006 ) .
such dynamically dominant disks leave little room for halos with steep inner density profiles .
for example , binney & evans ( 2001 ) find that cuspy halos are inconsistent with a number of constraints for the milky way ( cf .
klypin , zhao , & somerville 2002 ) . on the other hand
, it can be argued that the lack of scale length residuals in the tully - fisher relation implies that typically hsb disks are sub - maximal ( e.g. , courteau & rix 1999 ; cf .
sellwood 1999 ) .
if dark matter is dynamically significant in the inner regions of hsb galaxies , then it may be quite possible for the halos of these galaxies to have a steep cusp .
the critical issue is the stellar mass .
constraining the precise inner slope of the dark matter density profile of an individual galaxy is observationally challenging .
even if the stellar mass can be properly treated , a variety of systematic effects might affect the result ( swaters et al .
blok et al . 2003 ; simon et al . 2003 , 2005 ; spekkens et al .
there are several distinct problems here : observational uncertainties , non - circular motions that might cause misestimation of the potential , and degeneracy between fitting parameters .
observational uncertainties of concern here are those that might lead us to underestimate the steepness of the rise of the rotation curve .
this might result , for example , from beam - smearing of 21 cm observations ( e.g. , van den bosch et al .
2000 ) or mis - positioning of the slit during optical observations ( e.g. , swaters et al .
such errors are not a great concern when rotation curves are observed to be steeply rising , as is typically the case for hsb galaxies in this sample .
when the observed rotation curve rises more gradually , there is more potential for concern as to whether this behavior is truly intrinsic ( de blok , mcgaugh , & van der hulst 1997 ; de blok & mcgaugh 1997 ) .
this no longer appears to be a significant worry for lsb galaxies , as high resolution two dimensional velocity fields ( kuzio de naray et al .
2006 ) largely confirm pervious results ( e.g. , mcgaugh , rubin , & de blok 2001 ; de blok & bosma 2002 ; see also zackrisson et al .
2006 ) .
non - circular motions can be significant at small radii , in the sense that they can represent a substantial fraction of the total gravitational potential .
the contributions of circular and noncircular motions sum in quadrature . from this
, it immediately follows that non - circular motions ( @xmath10 ) are only important when comparable to the rotational motion ( @xmath11 ) .
simon et al .
( 2005 ) show some cases for which this is the case ( @xmath12 ) , albeit at quite small radii .
for the sample considered here , which all have two dimensional velocity fields , @xmath10 is typically small ( a few to tens of @xmath13 ) compared to @xmath11 at the same radius ( tens to hundreds of @xmath13 ) . we consider only intermediate and large radii where @xmath14 and the contribution of non - circular motions to the potential is not significant .
galaxies that do not satisfy this criterion are not in the sample .
a more fundamental difficulty is the degeneracy between halo fitting parameters , even when the potential is perfectly well known .
the degeneracy between stellar mass and the core radius of a pseudo - isothermal halo has long been known ( e.g. , kent 1987 ) . for nfw halos , there is substantial covariance between @xmath15 and @xmath16 even when the stellar mass is fixed ( de blok et al .
this degeneracy stems from the similarity of the inner shapes of nfw halo velocity profiles over a range of ( @xmath15,@xmath16 ) .
more generally , degeneracy is unavoidable because of a surplus of parameters .
mass models require the specification of a minimum of three parameters per galaxy ( one for the disk and two for the halo ) while empirically a single parameter suffices to describe the data ( e.g. , brada & milgrom 1999 ; mcgaugh 2004 ) . in this paper
, we adopt a model independent approach .
we do not fit specific mass models to individual galaxies .
the halo parameters derived from such models inevitably suffer from parameter degeneracies .
rather , we simply ask what range of halo parameters ( @xmath15 , @xmath16 ) are consistent with the halo velocities implied by observation . we determine the velocity attributable to the dark matter for a broad range of stellar mass estimators .
we find that the data are consistent with a nearly universal dark matter halo profile .
there is some scatter about this relation , and we would not claim that there is in fact a single halo profile appropriate to all galaxies , but the data are not wildly at odds with such a situation . the slope we find is consistent with previous determinations ( e.g. , barnes , sellwood , & kosowsky 2004 ) . if halos have the nfw form , their concentrations and/or masses ( @xmath16 ) are generally smaller than might nominally be expected in @xmath5cdm .
this is largely consistent with various previous results ( e.g. , debatista & sellwood 2000 ; weiner et al .
2001 ; borriello & salucci 2001 ; zentner & bullock 2002 ; alam , bullock , & weinberg 2002 ; mcgaugh et al .
2003 ; bell et al . 2003a ; gentile et al . 2004
; zackrison et al . 2006 ; portinari & sommer - larsen 2006 ) that suggest that real dark matter halos are less dense than predicted by @xmath5cdm .
the data and stellar mass estimators are described in 2 . in
3 we determine the halo velocity @xmath17 implied by the data for a range of stellar mass estimates .
the predictions of @xmath5cdm models and their relation to the data are described in 4 .
the implications of the result and possible ways to reconcile the apparent discord between theory and observation are discussed in 5 .
a summary of our conclusions is given in 6 .
we employ the rotation curve data and mass models compiled by sanders & mcgaugh ( 2002 ) , as trimmed for quality by mcgaugh ( 2005 ) .
these data originate from many independent observers , and represent the accumulated effort of a substantial community of astronomers .
sample galaxies span a wide range of morphological types , from sab through all spiral types to irr / im .
recent work by noordermeer et al .
( 2005 ) focussing on the earliest ( s0/sa ) type disk galaxies appears contiguous with the data used here .
more important than morphological type is the broad range of physical parameter space that our sample spans .
it includes galaxies with circular velocities @xmath18 , baryonic masses @xmath19 , scale sizes @xmath20 kpc , central surface brightnesses @xmath21 , and gas fractions @xmath22 .
no restriction is made on surface brightness or any other physical parameter .
the sample represents the dynamics of disks in the broadest sense possible with available data .
all sample galaxies have high quality , extended rotation curves derived from 21 cm data cubes .
the high precision sample of mcgaugh ( 2005 ) consists of 60 galaxies with @xmath23 independent , resolved rotation velocity measurements , each with formal accuracy of 5% or better .
we choose to exclude data with @xmath24 kpc where the controversy over cusp or core occurs ( de blok 2004 ) .
the point here is to compare reliable data with the expectations of @xmath5cdm models , free from concern over systematic uncertainties , either observational ( such as beam smearing ) or physical ( e.g. , non - circular motions ) .
these may be serious concerns for constraining the innermost slope of the density profile , but do not significantly affect the data at intermediate radii . here
the data are well resolved , and the observed circular velocity greatly exceeds the velocity dispersion . indeed ,
strong constraints exist on the ellipticities of the potentials for some sample galaxies ( schoenmakers , franx , & de zeeuw 1997 ) .
triaxiality of the halo ( e.g. , hyashi & navarro 2006 ) is not relevant to the bulk of the data here .
we exclude the possibility of non - circular motions biasing the results at small radii by truncating the inner data .
our result is not sensitive to the choice of the inner truncation radius .
the total radial range sampled is from 1 to 74 kpc , with a median radius of 9 kpc .
only a few galaxies have measured rotation extending to very large radii .
the bulk of the data ( 68% ) fall in the range @xmath25 kpc ( fig .
[ rhisto ] ) .
data in this range are not only numerous , but fall in a range where the systematic concerns discussed above have negligible impact . the same result follows if we restrict the data to this very conservative range .
for many galaxies in the sample , the contribution of the stars to the total potential can not be ignored .
this dominates all other effects by far .
it is necessary to subtract off the baryonic contribution to reveal the velocity due to the halo .
all galaxies have detailed surface photometry from which baryonic mass models are derived .
for the present work , we utilize @xmath26-band surface photometry , as this is the only band that all 60 galaxies have in common .
we are working to place the sample on a uniform @xmath27-band scale , but at present lack sufficiently deep observations of some of the fainter galaxies . using the @xmath27-band data already in hand
does not alter the basic result ( see also verheijen 2001 ; verheijen & sancisi 2001 ; kassin et al .
2006 ) .
we make no assumption about the light profile ( e.g. , exponential disk ) .
the observed light distribution is converted to a baryonic rotation curve numerically ( via gipsy : van der hulst et al .
this includes the bulge ( when present ) , stellar disk , and hi gas .
the molecular gas distribution is not available in most cases , but this is likely to be a small percentage of the baryonic mass in the majority of late type galaxies ( young & knezek 1989 ) .
moreover , the molecular gas generally follows the stellar distribution ( helfer et al .
2003 ) , so is effectively subsumed into the stellar mass - to - light ratio .
it is the estimation of stellar mass that is most crucial .
the observed rotation curves provide a good map of the total gravitational potential that is due to both baryonic and dark mass . while the total potential is well constrained , dividing it into baryonic and dark mass components is a degenerate process .
additional information is needed to break this degeneracy .
we explore three methods for estimating the stellar mass - to - light ratio @xmath28 : maximum disk , @xmath29 ; stellar population synthesis models , @xmath30 ; and that from the mass discrepancy acceleration relation , @xmath31 .
this latter in effect treats mond ( milgrom 1983 ) fits ( sanders & mcgaugh 2002 ) as an empirical phenomenology . that this phenomenology is present in the data
is all that is required , regardless of the underlying physical reason for it ( mcgaugh 2004 ) .
following mcgaugh ( 2005 ) , we can choose any arbitrary stellar mass estimate by scaling each of these prescriptions by a constant factor @xmath32 , @xmath33 , or @xmath34 : @xmath35 this scheme allows us to explore the full range of possible stellar masses ( @xmath36 ) with three independent estimators for each galaxy .
the population synthesis mass - to - light ratio @xmath30 is estimated from the observed @xmath37 color ( as tabulated by mcgaugh 2005 ) and the models of bell et al .
( 2003b ) : @xmath38 very similar results follow from other models ( e.g. , portinari et al .
the chief uncertainty of this method is the imf .
@xmath39 corresponds to the nominal `` scaled salpeter '' imf adopted by bell et al .
( 2003b ) , which for our purposes is nearly indistinguishable from the kroupa ( 2002 ) imf adopted by portinari et al .
( 2004 ; see also portinari & sommer - larsen 2006 ) . to consider the effects of a heavier or lighter imf , we simply scale by @xmath33 ( see discussion in bell & de jong 2001 ) .
we estimate the circular velocity provided by dark matter by subtracting the baryonic component from the observed rotation curves for each choice of @xmath28 : @xmath40 here @xmath41 is the rotation velocity attributed to the dark matter , @xmath42 is the observed rotation velocity , and @xmath43 is the rotation velocity attributed to baryons .
this depends on @xmath28 through @xmath44 , where @xmath45 is the rotation velocity for @xmath46 .
the hi gas component , corrected to account for helium and metals , gives @xmath47 .
molecular gas generally traces the stellar distribution ( helfer et al .
2003 ) and is subsumed into @xmath28 .
an example rotation curve is shown in fig .
[ n2403log ] .
as we increase ( decrease ) @xmath28 by varying @xmath32 , @xmath33 , and @xmath34 , the stellar contribution to the total rotation increases ( decreases ) and the corresponding contribution from the dark matter decreases ( increases ) .
though the details of each realization depend on the @xmath28 estimator and the relative importance of gas ( which does not vary ) , it is a tolerable approximation to imagine the dark and baryonic lines in fig .
[ n2403log ] sliding up and down see - saw fashion . in order to restrict our attention to the best data
, we have excised individual points with uncertainty in excess of 5% . in many cases ,
the formal precision is rather greater than this limit .
while this leaves plenty of data to analyze , various astrophysical considerations make it unrealistic to consider those data with small error bars as accurate estimates of the circular velocity of the potential ( see discussions in mcgaugh et al .
2001 ; de blok et al .
we therefore give uniform weight to the data which pass the 5% criterion . a more serious uncertainty than that in the data
is that in @xmath28 .
the estimate of stellar mass is the single dominant source of uncertainty in this analysis .
the systematic nature of this uncertainty makes it impossible to propagate it in a meaningful way .
hence we simply examine the results of adopting each @xmath28 estimator , leaving it to the reader s judgement which case might seem best . fortunately , the basic result is fairly insensitive to this choice . in fig .
[ vhqboth ] we show @xmath41 for all 60 galaxies plotted together for the case of @xmath48 .
remarkably , the inferred dark matter halos of all these galaxies fall more or less on top of each other .
this is not to say that the dark matter halos of all galaxies are the same .
quite the contrary , it is possible to observationally distinguish between the dark matter halos of individual galaxies ( sancisi 2004 ; mcgaugh 2004 ; barnes et al . 2004 ;
simon et al . 2005 ) . nonetheless , when plotted together in this fashion , the apparent variation is remarkably small .
taken together , the data suggest a relation @xmath49 fitting the data with @xmath48 gives @xmath50 and @xmath51 . while we would not claim that this is a universal halo that all galaxies share , it does give a tolerable approximation of the velocity attributable to dark matter .
we can extend this approach to other mass - to - light values by varying @xmath32 , @xmath33 , and @xmath34 .
the cases @xmath52 , @xmath39 , @xmath48 and @xmath53 ( the latter as suggested , @xmath54 corresponds to @xmath53 . ] by bottema 1993 and courteau & rix 1999 ) are illustrated in fig .
[ dmscale ] .
this figure is like fig .
[ vhqboth ] for the various mass estimators , but with shading to illustrate the density of data .
the darker the shading , the more sharply defined the relation between @xmath17 and radius .
there is some covariance between the slope @xmath55 and intercept @xmath56 .
our purpose here is to constrain the amplitude of the velocity that can be attributed to dark matter . in order not to obscure the basic result with uncertainties introduced by covariance
, we fix the slope to @xmath57 , consistent with our data ( see figs . [ dmhistog1 ] [ dmhistog04 ] ) and those of barnes et al . (
moreover , the slope does not vary appreciably when left free ; the important information is contained in the amplitude .
fixing @xmath57 changes the @xmath48 best fit intercept to @xmath58 .
this number encodes , in the simplest possible way , the amplitude of the dark matter velocity .
[ scatterplot ] show how this amplitude varies as we change @xmath28 . for @xmath39 ,
the amplitude barely changes : @xmath59 . for the full range of possibilities , from minimum to maximum disk , we find @xmath60 and @xmath61 . in other words
, the amplitude of the velocity provided by the dark matter falls in the range @xmath1 for @xmath57 .
one can also see in fig .
[ scatterplot ] variation in the scatter about the mean .
remarkably , the scatter is minimized for @xmath48 , the mond mass - to - light ratios .
this is also the choice of @xmath28 which minimizes the scatter in the baryonic tully - fisher relation ( mcgaugh 2005 ) and , by construction , that in the mass discrepancy acceleration relation ( mcgaugh 2004 ) .
irrespective of the physics that causes it , this implies that mond is an important empirical organizing relation that is relevant even to dark matter .
the scatter is larger for other mass estimators , but the basic result encapsulated by @xmath56 is the same .
cosmological simulations lead to mass distributions for dark matter halos that have gravitational potentials which can be analytically approximated as @xmath62}\right ] , \label{nfwformula}\ ] ] where @xmath63 and @xmath64 ( nfw ) .
the radius @xmath65 encloses a density 200 times the critical density of the universe ; @xmath16 is the circular velocity of the potential at @xmath65 .
the radius @xmath65 is , crudely speaking , the virial radius in a critical density universe . for current @xmath5cdm parameters ,
the virial radius is more nearly at @xmath66 ( e.g. , eke , navarro , & steinmetz 2001 ) . more recently , navarro et al .
( 2004 ) have suggested that a density profile with a continuously running slope provides a better fit to simulated halos than the original nfw formula . from an observational perspective , these forms are virtually indistinguishable .
we therefore retain the nfw formulation for its familiarity and its well quantified relation to cosmological parameters ( nfw ; mcgaugh , barker , & de blok 2003 ) .
one interesting aspect of the data is that there is no clear tendency for the dark halo rotation curves to flatten out .
the combination of disk plus halo is often close to flat , and as the disk contribution is declining , it is natural to associate the flat rotation velocity with the halo .
this is , however , an oversimplification . in most cases , the inferred dark matter contribution to the velocity is still rising at the last measured point , and it is not clear that the observed flat velocity is truly representative of the halo . in the context of simulated halos , the lack of a flat halo rotation velocity may help the models .
nfw halos do not possess a significant range of radii where @xmath67 , as needed to produce flat rotation curves .
presumably , the observed range suffices only to cover the rising portion of the halo , with a suggestion of a turn - over in only a few sufficiently extended cases .
it remains an open question why the disk and halo conspire to produce nearly flat rotation curves over the range observed .
it is suggestive that the fitted slope @xmath68 is that expected for the inner portion of nfw halos .
it is tempting to conclude that halos are indeed nfw , and we will use this to compare the observed and predicted amplitudes @xmath56 .
an important caveat is that it is not straightforward to match the radial scale of simulated halos to that of real galaxies .
it is for this reason that the amplitude @xmath56 provides the most robust comparison : it uses all the data to set the velocity scale , in effect providing an integrated measure that is not sensitive to systematic effects in the way the innermost slope may be .
the slope of the data is , if anything , too close to a constant @xmath57 .
nfw profiles have this slope at small radii , though the exact slope the simulations predict is still debated ( e.g. , ricotti 2003 ; power et al . 2003 ; navarro et al . 2004 ; diemand et al . 2005 ) .
irrespective of this detail , all simulations agree that the inner slope gradually transitions to a steeper fall off ( becoming @xmath69 at large radii ) .
the consequence is that there should be some curvature apparent in @xmath41 over the radial range sampled by the data .
this is not obvious in the observations ( fig .
[ vhqboth ] ) .
this difference between simulations and data is sufficiently large that the difference between the various proposed analytic formulas is a rather small concern .
the use of individually simulated halos ( e.g. , hayashi et al . 2004 ; cf .
de blok 2005 ) in preference to the analytic approximations should not make a difference to the radii of interest here , where the deviations between the various realizations of the simulation results are small .
in addition to the data , the upper panel of fig . [ vhqboth ] shows the predicted shape and amplitude of nfw rotation curves for a range of @xmath16 . presuming galaxy halos span a range of mass comparable to that suggested by their luminosities
, one would not only expect more curvature than observed , but also a wider spread in @xmath17 ( compare , for example , the halos with @xmath70 and @xmath71 ) .
that the data concentrate in this plane more than anticipated might suggest that we may err in presuming that halo masses span the same range of mass as the observed luminosity , or that the simulated @xmath15-@xmath16 relation built into the lines in the upper panel of fig .
[ vhqboth ] may not be precisely correct .
one must bear in mind , of course , that the data only extend over a finite range of radii , and do not generally show the inevitable maximum and decline in @xmath17 . though it is difficult to assign a specific halo to a particular galaxy , taken together the data are suggestive of an envelope composed of halos of a range of masses .
if we had complete knowledge of the potentials out to arbitrary radii , we might see individual halos peeling away from this envelope as illustrated by the lines in the bottom panel of fig .
[ vhqboth ] . for comparison with the data , we can ask what we would expect to find from a @xmath5cdm model observed in the same way as the data .
we sample a series of halos with @xmath72 ( e.g. , the lines in the top panel of fig .
[ vhqboth ] ) .
model data points are allowed to fill the space available to them within the observed range of radii .
the model points are uniformly sampled and weighted since we have no knowledge of the underlying halo mass distribution , other than that it gives rise to the broad observed range of luminosities and rotation velocities .
since no turn - over ( @xmath73 of the halo ) is observed in the data , the maximum radius to which halos are sampled is truncated along the lower edge of the observed envelope , as defined by a line with slope 1/2 and intercept 1.33 .
the result of fitting the model halos is @xmath74 and @xmath75 for the vanilla @xmath5cdm parameters of tegmark et al .
( 2004 ) . for the third year
wmap parameters ( spergel et al .
2006 ) , @xmath56 decreases to 1.61 .
this normalization is too large , being consistent with the observed value only in the limit of zero stellar mass ( @xmath76 ) .
the @xmath15-@xmath16 relation can be adjusted in order to match the nfw model to the observed amplitude of the data .
this is illustrated in the lower panel of fig .
[ vhqboth ] by lines representing nfw halos with parameters chosen to match the observed normalization @xmath56 . in effect
, this provides an empirical estimate of the @xmath15-@xmath16 relation .
we can do this for the value of @xmath56 implied by any choice of @xmath28 . to get the zero point amplitude ( @xmath56 ) for nfw ,
let us rewrite equation ( [ nfwformula ] ) in log - log space .
in the limit might seem like a dubious approximation since we have excluded the data at @xmath24 kpc .
however , it is in fact quite a good approximation both theoretically ( since @xmath77 and we are concerned with data at a small fraction of the virial radius ) and empirically ( since the observed slope is close to limiting case of 1/2 ) . ] @xmath78 we have : @xmath79+\log(h/2)\right ] + \onehalf \log r,\ ] ] where @xmath80 and the factor @xmath81 appears because @xmath82 when @xmath16 is in @xmath13 and @xmath65 is in kpc ( nfw ) . written this way
, we recognize @xmath0 with @xmath57
. we can thus relate nfw parameters to the data through the observed @xmath56 : @xmath83 - \log(h/2)\ ] ] ( the data assume @xmath84 ) . in this fashion
, we can generate the range of ( @xmath15 , @xmath16 ) that are acceptable to the data ( fig .
[ nfwcv ] ) .
note that we do not specify the particular values of these parameters for any given galaxy .
rather , we constrain the range of @xmath15-@xmath16 parameter space from which the observed halos may be drawn . the parameters of nfw halos in simulations are not arbitrary .
there is a correlation between @xmath15 and @xmath16 ( nfw ) with fairly modest @xmath85 wide . ]
scatter ( jing 2000 ; bullock et al .
2001 ; coln et al .
2004 ) . moreover , the typical concentration depends on the cosmology ( nfw ) . for a @xmath86 halo ,
mcgaugh et al .
( 2003 ) found that the mean @xmath87 where @xmath88 nfw found that the density of halos depends on the density of the universe at the time of collapse .
this depends on @xmath89 and the amplitude of the power spectrum at the appropriate scale , as encapsulated by the above formula .
the concentration varies slowly with scale @xmath16 ( nfw ) . for illustration
, we adopt @xmath5cdm parameters from the tegmark et al .
( 2004 ) and spergel et al .
we use the 6-parameter ` vanilla ' fit of tegmark et al .
( 2004 ) , and the results of the 3 year wmap - only fit from spergel et al .
( 2006 ) . while superficially these may appear similar , the concentrations of dark matter halos are quite sensitive to cosmic parameters , so there is a noticeable difference . the observed and predicted @xmath15-@xmath16 bands are illustrated in fig .
[ nfwcv ] . for the scatter in @xmath5cdm concentrations
, we adopt @xmath90 ( bullock et al .
similarly , we illustrate the @xmath91 range of observed halos ( i.e. , the band that contains 68% of the halos from the histograms in figs .
[ dmhistop1 ] and [ dmhistoq1 ] ) .
the predicted and observed bands do not match up well , having only a modest range of overlap at surprisingly small @xmath16 .
the cusp - core problem in dwarf and lsb galaxies is well known .
inspection of fig .
[ nfwcv ] reveals a more general problem with the densities of the dark matter halos of all spiral galaxies .
the data are consistent with a swath of @xmath15-@xmath16 space that does not parallel the predictions of @xmath5cdm .
the acceptable regions of halo parameter space diverge as mass increases .
this result can not be attributed to inadequate resolution or slit positioning errors .
all of the data are derived from resolved two dimensional velocity fields .
we have intentionally excluded data from the innermost region where the resolution might be questioned .
significant non - circular motions could be seen directly in the velocity fields .
the result does not change if we exclude where nfw halos provide their maximum acceleration . ]
even more of the inner data ( see fig . [ rhisto ] ; imagine truncating fig .
[ vhqboth ] at 2 or 3 kpc ) .
it is the velocity attributable to dark matter at well resolved intermediate radii that is lower than expected in @xmath5cdm .
let us examine the implications of this result , and possible ways out , starting with the theoretical expectations .
two fiducial @xmath5cdm models are illustrated in fig .
[ nfwcv ] .
the model of tegmark et al .
( 2004 ) is a good representation of standard vanilla @xmath5cdm , predicting very nearly the same swath of @xmath15-@xmath16 parameter space as other plausible choices for cosmological parameters , going back to the early work of nfw .
the wmap three year results ( spergel et al . 2006 ) noticeably lower the predicted halo concentrations .
this is due to slight changes in all of the relevant parameters .
the dominant effect is from the lower power spectrum normalization @xmath92 , a result anticipated both by cluster ( e.g. , rosati , borgani , & norman , 2002 ; diego et al .
2003 ) and galaxy work ( zentner & bullock 2002 ; mcgaugh et al . 2003
; van den bosch , mo , & yang 2003 ) .
the tilt ( @xmath93 ) and the lower matter density ( @xmath94 ) also contribute .
as we add in large scale constraints beside wmap , the parameters tend to drift back towards those of tegmark et al .
( 2004 ) .
the lower concentrations implied by the more recent wmap results do help . ignoring the trend with @xmath16 for the moment , equation
( [ concosmo ] ) implies @xmath95 for the cosmic parameters of tegmark et al .
( 2004 ) , whereas the new wmap data ( spergel et al .
2006 ) imply @xmath96 .
this difference is quite pronounced in rotation curves .
the revised wmap results at least bring cosmic data into the right ballpark : mcgaugh et al .
( 2003 ) found @xmath97 for dark matter dominated galaxies , though it should be noted that this typical value ignores problematic cases with @xmath98 or for which nfw simply does not fit .
the problem here is not so much the mean concentration , but the run of concentration with halo mass .
concentrations are acceptable for low mass halos ( those with @xmath99 ) .
they become increasingly discrepant with increasing mass . by the time we reach @xmath100 ( @xmath86 ) , the data imply @xmath101 .
this is nominally the mass scale of the milky way ( e.g. , wilkinson & evans 1999 ; sakamoto , chiba , & beers 2003 ) , and the low inferred concentration is consistent with both milky way constraints ( binney & evans 2001 ; flynn et al .
2006 ) and those for comparable galaxies ( e.g , weiner et al .
2001 ) . to obtain such a low concentration from initial conditions would require absurd changes to cosmology : @xmath102 or @xmath103 .
this basic result is virtually independent of the choice of mass - to - light ratio estimator .
the distribution of @xmath17 is not as sharply peaked for @xmath39 as it is for @xmath48 , but the mean is nearly the same , so the same result is obtained .
worse , changing @xmath33 ( or @xmath34 or @xmath32 ) simply shifts the bands left ( for @xmath104 ) or right ( for @xmath105 ) in fig . [ nfwcv ] .
it does little to alter the differing slope of the observed and predicted @xmath15-@xmath16 bands . in order to fix the mismatched slopes with @xmath28
, we would have to make the stellar mass indicator a systematic function of @xmath16 .
specifically , @xmath28 would have to get progressively smaller as halo mass increases .
this is quite contrived .
there is no evidence to suggest that the stellar populations of massive spirals are systematically lighter than those of dwarfs .
indeed , such a contention can likely already be excluded by the need for a substantial stellar contribution to the mass in at least some high surface brightness galaxies ( e.g. , debattista & sellwood 2000 ; weiner et al .
2001 ; kassin et al . 2006 ; battaglia et al .
2006 ) . in contradiction to this , jimenez ,
verde , & oh ( 2003 ) argue that rotation curve fits tolerate the expected concentrations for nfw halos .
however , this comes at the expense of unrealistic @xmath28 .
the @xmath106-band @xmath28 distribution of jimenez et al .
( 2003 ) is bimodal ( their fig . 7 ) , with roughly half of the galaxies centered on @xmath107 .
so they in fact find the same result , that mass - to - light ratios need to be quite low in order to be reconciled with the expected nfw concentrations .
a velocity - variable imf might achieve this , but it would introduce an unreasonable amount of scatter into the baryonic tully - fisher relation ( mcgaugh 2005 ) . the mismatch between the slopes of the observational and theoretical bands in fig .
[ nfwcv ] is caused by the near universality of the empirical dark matter halo found in 3 .
the data can be described , albeit crudely , by a single simple form ( equation [ singlehalo ] ) . according to nfw
, this should not be the case halos should spread out ( e.g. , top panel of fig .
[ vhqboth ] ) , not clump together as observed . in this case , we would expect to see more data at high @xmath17 at small @xmath108 .
while there are choices of @xmath28 that have some data in this region , the bulk of the data do not reside there ( fig .
[ dmscale ] ) . the slope of the observed bands in fig .
[ nfwcv ] is the result of adjusting the nfw halos concentrations to match the data with @xmath109 .
it is worth checking how this changes if we allow the slope to be free .
a @xmath28 choice that leads to a shallower slope and higher amplitude is needed to reconcile the data with the model . leaving both the slope and the intercept free in the fit , we obtain @xmath110 and @xmath111 for @xmath39 .
this steeper slope goes in the wrong direction , as it would make the @xmath39 band in fig .
[ nfwcv ] cut somewhat _ more _ steeply across the predicted band . using the two dozen galaxies for which we have @xmath112-band data and assuming a constant @xmath113 ( the @xmath39 value for the mean color )
, we obtain @xmath114 and @xmath115 , consistent with @xmath48 . increasing the stellar mass
makes things worse , maximum disk being the limiting case ( @xmath116 and @xmath117 for @xmath52 ) : there is too little velocity attributable to dark matter at small radii . lowering the stellar mass within reasonable bounds
does not help either .
for @xmath118 , @xmath119 and @xmath115 . this a band with precisely the same slope as illustrated ; the change in @xmath56 only shifts the band a bit to the right .
the case of @xmath53 ( bottema 1993 ) gives results indistinguishable from @xmath118 .
to push the stellar mass low enough to get into the right ballpark , we need @xmath120 ( @xmath121 and @xmath122 for @xmath123 ) .
this corresponds to an imf that is a factor of 5 less massive than salpeter .
this is below the practical minimum in stellar mass as determined from direct integration of the observed imf ( kroupa 2002 ) .
it is also grossly inconsistent with the coherence of the mass discrepancy acceleration relation ( mcgaugh 2004 ) and baryonic tully - fisher relation ( mcgaugh 2005 ) , the presence of baryonic features observed in rotation curves ( palunas & williams 2000 ; sancisi 2004 ; kassin et al .
2006 ; battaglia et al .
2006 ) , observed velocity dispersions ( bottema 1997 ; kregel , van der kruit & freeman 2005 ) , and hydrodynamical flows ( weiner et al .
2001 ; kranz , slyz & rix 2003 ) .
there appears to be no plausible choice of @xmath28 that is consistent with cosmological expectations and nfw , at least for halos with @xmath16 comparable to the observed circular velocity .
we have so far implicitly assumed that halo mass scales linearly with baryonic mass . this need not be the case .
one possibility to consider it that the observed luminosity function of galaxies descends from parent dark matter halos spanning a smaller range of mass .
this would require a systematic variation in the fraction of baryons that form the luminous disk ( mcgaugh & de blok 1998 ; okamoto et al .
2005 ) , but this may already be required to match the observed tully - fisher relation ( van den bosch 2000 ; mo & mao 2004 ; see also navarro & steinmetz 2000 ; eke et al .
2001 ; portinari & sommer - larsen 2006 ; dutton et al .
2006 ) .
if we drop the assumption that total mass scales with baryonic mass , it is tempting to conclude that galaxies may be born from halos in the region allowed in fig .
[ nfwcv ] .
this would mean that even massive galaxies with observed rotation velocities in excess of 200 @xmath13 ( ranging all the way up to 300 @xmath13 ) would have their origins in relatively modest halos . from fig .
[ nfwcv ] , we need @xmath124 for vanilla @xmath5cdm , and @xmath125 for wmap third year parameters .
galaxies would presumably gain their larger observed velocities during the collapse of their massive disks ( see below ) .
examination of figs .
[ vhqboth ] and [ dmscale ] reveals that halos with @xmath126 are about right to explain much of the data .
unfortunately , such low mass halos do not provide a complete explanation .
for one thing , there need to be enough baryons available in the halo to form the observed galaxy . assuming a universal baryon fraction of 0.17 ( spergel et al .
2006 ) , and the most favorable case that all baryons within @xmath65 collapse to form the visible galaxy , a halo with @xmath127 has enough baryons to make a galaxy with @xmath128 .
this assumes @xmath48 mass - to - light ratios and the baryonic tully - fisher relation of mcgaugh ( 2005 ) .
there is some room to play with the choice of @xmath28 , but this does little more than assuage the extreme assumption that all baryonscdm , the virial radius is closer to @xmath66 than @xmath65 , further reducing the number of available baryons . ] collapse to form the observed galaxy .
it is therefore rather difficult to ascribe the known spiral galaxies with @xmath129 to nfw halos of low mass but normal concentration , at least for vanilla @xmath5cdm .
this constraint is relaxed for wmap 3 year parameters , but even here a halo with @xmath130 can give at most enough baryons for a galaxy with @xmath131 ( again taking @xmath48 ) .
looking in detail , nfw halos obeying the expected @xmath15-@xmath16 relation have more curvature than observed , predicting too high a velocity at small @xmath108 and too small a velocity at large @xmath108 relative to the data .
even if we overlook the over - prediction at @xmath132 kpc , the under - prediction at @xmath133 kpc is quite serious .
the compressive formation of a massive disk can raise @xmath17 , but this process is much more effective at small radii than at large radii . in order to explain the high observed dark halo velocity at large radii
, it seems necessary to have initial halos with at least @xmath134 .
such halos grossly over - predict @xmath17 at small @xmath108 if they obey the vanilla @xmath15-@xmath16 relation .
it may be tempting to simply ignore the data at large radii , as there are relatively few with @xmath135 kpc
. however , there do exist galaxies outside this sample ( e.g. , noordermeer et al . 2005 ; spekkens & giovanelli 2006 ) with outer rotation velocities approaching 300 @xmath13 .
these galaxies show no sign of a turn down in their outer velocities , and obey the same scaling relations as those in the present sample , so there is no reason to expect them to be any different .
the need for large @xmath17 at large radii is thus a serious problem .
so far , we have compared the data to the @xmath5cdm predictions for primordial dark matter halos .
it is possible that the process of galaxy formation alters the halos .
if so , the difference between the predictions and observations might be interpreted to show what changes occur in this process .
one should bear in mind that cuspy mass distributions seem to be a generic result of cold collapse ( e.g. , huss et al .
1999 ) , and that , once established , it is very hard to unbind the mass in the deep potential wells of cuspy halos .
achieving the desired effect is thus a tall order from the outset ( sommer - larsen & dolgov 2001 ; eke et al .
2001 ) .
the primary physical effect which should occur during galaxy formation is the adiabatic contraction of the dark matter halo in response to the collapse of the baryonic disk ( e.g. , blumenthal et al .
1986 ; barnes & white 1984 ; gnedin et al . 2004 ; sellwood & mcgaugh 2005 ; choi et al .
the effect of adiabatic contraction is to further concentrate the dark matter .
this goes in the wrong direction to help with the problems discussed here . however , there are two distinct aspects of adiabatic contraction to consider .
the slope @xmath136 will steepen from its initial value ; this goes in the wrong direction .
the velocity provided by the dark matter also increases .
while this might also seem to act in the wrong direction , it means that a galaxy of a given rotation velocity may form in a halo with lower primordial @xmath16 . as discussed above
, this can help , at least in principle .
unfortunately , it seems unlikely that the effect can be large enough to help at large radii without simultaneously doing far too much at small radii .
disks heavy enough to have a substantial impact on outer velocities grossly over - predict inner velocities if the starting point is an nfw halo ( abadi et al . 2003 ; sellwood & mcgaugh 2005 ) . to make matters worse , dutton et al .
( 2006 ) argue that a large difference between the observed and initial halo velocity scales would make reconciling the normalization problem between tully - fisher and the luminosity function practically impossible .
low mass halos might also be inconsistent with constraints from weak lensing ( smith et al . 2001 ; mandelbaum et al .
2006 ) .
after adiabatic contraction , there may be some subsequent process that further alters the halo .
perhaps the most common example is feedback from star formation .
quite generically , the effects of feedback are believed to be stronger in lower mass galaxies owing to their shallower potential wells ( e.g. , thacker & couchman 2000 ) .
such an effect is unlikely to be sufficiently strong ( mac low & ferrara 1999 ; gnedin & zhao 2002 ) and acts preferentially on the wrong end of the mass function .
it is the concentrations of high mass galaxies ( if not necessarily the slope @xmath136 ) that are most deviant from the predictions of @xmath5cdm .
if the solution to this problem lies in the processes of galaxy formation , it likely resides in a rather different direction ( see , for example , el - zant et al .
2001 ; weinberg & katz 2002 ; merritt et al . 2004 ; ma & boylan - kolchin 2004 ; mashchenko , couchman , & wadsley 2006 ; tonini , lapi , & salucci 2006 ; see also sellwood 2006 ) .
another approach would be to alter the initial conditions .
a turndown in the power spectrum on galaxy scales can reduce halo concentrations .
however , such a solution would require a rather strange dip in the power spectrum .
concentrations at small scales are not unreasonable , and also make sense on larger ( cluster ) scales ( e.g. , vikhlinin et al .
it is the intermediate scale of massive spirals , giant ellipticals ( romanowsky et al .
2003 ; pierce et al .
2006 ) , and groups ( karachentsev 2005 ) where the halo concentration prediction the shape rather than the amplitude continues to suffer most clearly in lsb systems ( simon et al . 2005 ; kuzio de naray et al . 2006 ;
goerdt et al . 2006 ; sanchez - salcedo et al .
2006 ) . ] of @xmath5cdm
simulations fares worst .
another possibility is to modify the nature of dark matter .
the cusp - core problem has already motivated consideration of various possibilities , such as warm ( bode , ostriker , & turok 2001 ) , fuzzy ( hu , barkana , & gruzinov 2000 ) , self - interacting ( spergel & steinhardt 2000 ) , and meta - cold ( strigari et al . 2006 ) dark matter , among others ( e.g. , piazza & marinoni 2003 ; lee & lee 2004 ; blanchet 2006 ) .
it is not immediately clear how any of these proposals fare in the current context .
however , many proposals seem to be meant to produce dark matter halos with a core .
this in itself is not a complete solution .
one must get the density right , and do so on all relevant scales .
we have examined the rotation curves of a large sample of spiral galaxies in order to constrain the velocity that can be attributed to dark matter . taken together ,
the dark halo velocity can be approximated by @xmath137 with @xmath68 and @xmath1 .
the value of @xmath56 depends on the choice of stellar mass - to - light ratio , with the uncertainty corresponding to the full range of possibilities from minimum to maximum disk .
while the dark matter halos of individual galaxies must differ , this relation gives a tolerable approximation over the observed range . in the context of @xmath5cdm ,
the densities of halos depends on cosmology . for reasonable cosmological parameters ,
we expect @xmath6 .
this is only consistent with the data in the limit @xmath138 , which is excluded by a wide variety of independent lines of evidence .
the problem is not just that rotation curve data prefer halo models with a core over those with a cusp .
the amplitude of the velocity provided by dark matter is too low at all observed radii .
we have emphasized here the importance of spiral galaxy dark matter halo densities being substantially lower than anticipated by @xmath5cdm . for halos of @xmath139 ,
the density discrepancy is nearly an order of magnitude ( see also sellwood 2006 ) .
getting the halo density in the right ballpark would be an important step towards a successful theory .
ultimately , one must explain in detail the observed coupling between dark and baryonic components ( e.g. , sancisi 2004 ; mcgaugh 2004 ) .
we thank the referee for comments which made this manuscript more clear and led to substantial improvements in the figures .
the work of ssm is supported in part by nasa grant nag513108 .
rkdn and jhk are supported by nsf grant ast0505956 .
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 .
abadi , m. g. , navarro , j. f. , steinmetz , m. , & eke , v. r. 2003 , , 591 , 499 alam , s. m. k. , bullock , j. s. , & weinberg , d. h. 2002 , , 572 , 34 avila - reese , v. , firmiani , c. , & hernandez , x. 1998 , , 505 , 37 barnes , e.i .
, sellwood , j.a . , & kosowsky , a. 2004 , , 128 , 2724 barnes , j. , & white , s. d. m. 1984 , , 211 , 753 battaglia , g. , fraternali , f. , oosterloo , t. , & sancisi , r. 2006 , , 447 , 49 bell , e. f. & de jong , r. s. 2001 , , 550 , 212 bell , e. f. , baugh , c. m. , cole , s. , frenk , c. s. , & lacey , c. g. 2003a , , 343 , 367 bell , e. f. , mcintosh , d. h. , katz , n. , & weinberg , m. d. 2003b , , 149 , 289 binney , j. j. , & evans , n. w. 2001 , , 327 , l27 blais - ouellette s. , amram p. , carignan c. , swaters r. 2004 , , 420 , 147 blanchet , l. 2006 , gr - qc/0609121 blumenthal , g. r. , faber , s. m. , flores , r. , & primack , j. r. 1986 , , 301 , 27 bode , p. , ostriker , j. p. , & turok , n. 2001 , , 556 , 93 borriello , a. , & salucci , p. 2001
, , 323 , 285 bosma , a. 1981 , , 86 , 1825 bottema , r. 1993 , , 275 , 16 bottema , r. 1997 , , 328 , 517 brada , r. , & milgrom , m. 1999 , , 512 , l17 bullock , j. s. , kolatt , t. s. , sigad , y. , somerville , r. s. , kravtsov , a. v. , klypin , a. a. , primack , j. r. , & dekel , a. 2001 , , 321 , 559 choi , j .- h . , lu , y. , mo , h. j. , weinberg , m. d. 2006 , , 1138 cole , s. & lacey , c. 1996 , , 281 , 716 coln , p. , klypin , a. , valenzuela , o. , & gottlber , s. 2004 , , 612 , 50 ct , s. , carignan , c. , & freeman , k. c. 2000 , , 120 , 3027 courteau , s. & rix , h .- w . 1999 , , 513 , 561 daigle , o. , carignan , c. , amram , p. , hernandez , o. , chemin , l. , balkowski , c. , kennicutt , r. 2006 , , 367 , 469 de blok , w. j. g. 2004 , iau symposium , 220 , 69 de blok , w. j. g. 2005 , , 634 , 227 de blok , w. j. g. , & bosma , a. 2002 , , 385 , 816 de blok , w.j.g . , bosma , a. , & mcgaugh , s.s .
2003 , mnras , 340 , 657 de blok , w.j.g .
, & mcgaugh , s.s .
1996 , , 469 , l89 de blok , w.j.g . , & mcgaugh , s.s .
1997 , , 290 , 533 de blok , w.j.g . , &
mcgaugh , s.s .
1998 , , 508 , 132 de blok , w.j.g . ,
mcgaugh , s.s . , van der hulst , j.m .
1996 , , 283 , 18 de blok , w.j.g . ,
mcgaugh , s.s . ,
bosma , a. , & rubin , v.c .
2001a , , 552 , l23 de blok , w.j.g . ,
mcgaugh , s.s . , & rubin , v.c .
2001b , , 122 , 2396 debattista , v. p. , & sellwood , j. a. 2000 , , 543 , 704 diego , j. m. , sliwa , w. , silk , j. , & barcons , x. 2003 , , 344 , 951 diemand , j. , zemp , m. , moore , b. , stadel , j. , & carollo , m. 2005 , , 364 , 665 donato , f. , gentile , g. , & salucci , p. 2004
, , 353 , l17 dubinski , j. 1994 , , 431 , 617 dutton , a. a. , van den bosch , f. c. , dekel , a. , & courteau , s. 2006 ( astro - ph/0604553 ) eke , v. r. , navarro , j. f. , & steinmetz , m. 2001 , , 554 , 114 el - zant , a. , shlosman , i. , & hoffman , y. 2001 , , 560 , 636 flores , r. a. , & primack , j. r. 1994 , , 427 , l1 flynn , c. , holmberg , j. , portinari , l. , fuchs , b. , & jahrei , h. 2006 , , 372 , 1149 fraternali , f. , van moorsel , g. , sancisi , r. , & oosterloo , t. 2002 , , 123 , 3124 gentile , g. , burkert , a. , salucci , p. , klein , u. , & walter , f. 2005 , , 634 , l145 gentile , g. , salucci , p. , klein , u. , vergani , d. , & kalberla , p. 2004
, , 351 , 903 gnedin , o. y. , kravtsov , a. v. , klypin , a. a. , & nagai , d. 2004 , , 616 , 16 gnedin , o. y. , & zhao , h. 2002 , , 333 , 299 goerdt , t. , moore , b. , read , j. i. , stadel , j. , & zemp , m. 2006 , , 368 , 1073 helfer , t. t. , thornley , m. d. , regan , m. w. , wong , t. , sheth , k. , vogel , s. n. , blitz , l. , & bock , d. c .- j .
2003 , , 145 , 259 hu , w. , barkana , r. , & gruzinov , a. 2000 , , 85 , 1158 huss , a. , jain , b. , & steinmetz , m. 1999 , , 517 , 64 hayashi , e. , & navarro , j. f. 2006 , astro - ph/0608376 hayashi , e. , et al .
2004 , , 355 , 794 jimenez , r. , verde , l. , & oh , s. p. 2003
, , 339 , 243 jing .
y. , 2000 , , 535 , 30 jing , y.p . , &
suto , y. 2000 , , 529 , l69 karachentsev , i. d. 2005 , , 129 , 178 kassin , s. a. , de jong , r. s. , & weiner , b. j. 2006 , , 643 , 804 kent , s.m .
1987 , , 94 , 306 klypin , a. , kravtsov , a. v. , bullock , j. s. , & primack , j. r. 2001 , , 554 , 903 klypin , a. , zhao , h. , & somerville , r. s. 2002 , , 573 , 597 kranz , t. , slyz , a. , & rix , h .- w .
2003 , , 586 , 143 kregel , m. , van der kruit , p. c. , & freeman , k. c. 2005 , , 358 , 503 kroupa , p. 2002 , science , 295 , 82 kull , a. 1999 , , 516 , l5 kuzio de naray , r. , mcgaugh , s.s .
, de blok , w.j.g .
& bosma , a. 2006 , , 165 , 461 lee , t.h . , & lee , b.j .
2004 , 69 , 127502 ma , c .- p . , & boylan - kolchin , m. 2004 , , 93 , 021301 mac low , m .- m . , &
ferrara , a. 1999 , , 513 , 142 mandelbaum , r. , seljak , u. , kauffmann , g. , hirata , c. m. , & brinkmann , j. 2006 , , 368 , 715 marchesini , d. , donghia , e. , chincarini , g. , firmani , c. , conconi , p. , molinari , e. , & zacchei , a. 2002 , , 575 , 801 mashchenko , s. , couchman , h.m.p . , & wadsley , j. 2006 , , 442 , 539 mcgaugh , s.s .
2004 , , 609 , 652 mcgaugh , s.s .
2005 , , 632 , 859 mcgaugh , s.s . , barker , m.k .
, & de blok , w.j.g . , 2003 , , 584 , 566 mcgaugh , s.s .
, & de blok , w.j.g .
1998 , , 499 , 41 mcgaugh , s.s . , rubin , v.c .
, & de blok , w.j.g .
2001 , , 122 , 2381 mcgaugh , s. s. , schombert , j. m. , bothun , g. d. , & de blok , w. j. g. 2000 , , 533 , l99 merritt , d. , milosavljevic , m. , favata , m. , hughes , s.a . , & holz , d.e .
2004 , , 607 , l9 milgrom , m. 1983 , , 270 , 371 mo , h. j. , & mao , s. 2004 , , 353 , 829 moore , b. 1994 , , 370 , 629 moore , b. , ghigna , s. , governato , f. , lake , g. , quinn , t. , stadel , j. , & tozzi , p. 1999a , , 524 , l19 moore , b. , governato , f. , quinn , t. , stadel , j. , & lake , g. 1998 , , 499 , l5 moore , b. , quinn , t. , governato , f. , stadel , j. , & lake , g. 1999b , , 310 , 1147 navarro , j. f. , et al .
2004 , , 349 , 1039 navarro , j. f. , frenk , c. s. , & white , s. d. m. 1997 , , 490 , 493 ( nfw ) navarro , j. f. & steinmetz , m. 2000 , , 528 , 607 noordermeer , e. , van der hulst , j. m. , sancisi , r. , swaters , r. a. , & van albada , t. s. 2005 , , 442 , 137 nusser , a. , & sheth , r. 1998 , , 303 , 685 okamoto , t. , eke , v. r. , frenk , c. s. , & jenkins , a. 2005 , , 363 , 1299 palunas , p. & williams , t. b. 2000 , , 120 , 2884 persic , m. , & salucci , p. 1988
, , 234 , 131 piazza , f. , & marinoni , c. 2003 , , 91 , 141301 pierce , m. , et al . 2006 , , 366 , 1253 portinari , l. , & sommer - larsen , j. 2006 ( astro - ph/0606531 ) portinari , l. , sommer - larsen , j. , & tantalo , r. 2004 , , 347 , 691 power , c. , navarro , j. f. , jenkins , a. , frenk , c. s. , white , s. d. m. , springel , v. , stadel , j. , & quinn , t. 2003 , , 338 , 14 ricotti , m. 2003 , , 344 , 1237 romanowsky , a. j. , douglas , n. g. , arnaboldi , m. , kuijken , k. , merrifield , m. r. , napolitano , n. r. , capaccioli , m. , & freeman , k. c. 2003 , science , 301 , 1696 rosati , p. , borgani , s. , & norman , c. 2002 , , 40 , 539 rubin , v. c. , thonnard , n. , & ford , w. k. , jr .
1980 , , 238 , 471 sakamoto , t. , chiba , m. , & beers , t. c. 2003 , , 397 , 899 salucci , p. 2001
, , 320 , l1 salucci , p. , & burkert , a. 2000 , , 537 , l9 salucci , p. , walter , f. , & borriello , a. 2003 , , 409 , 53 salvador - ole , e. , solanes , j .- m . , & manrique , a. 1998 , , 499 , 542 sanchez - salcedo , f.j . , reyes - iturbide , j , & hernandez , x. 2006 , 370 , 1829 sanders , r.h . , & mcgaugh , s.s .
2002 , , 40 , 263 schoenmakers , r. h. m. , franx , m. , & de zeeuw , p. t. , 1997 , , 292 , 349 sellwood , j. a. 1999 , in galaxy dynamics , asp 182 , 351 sellwood , j. a. 2006 , astro - ph/0610468 sellwood , j. a. , & kosowsky , a. 2001 , asp conf .
ser . 240 : gas and galaxy evolution , 240 , 311 sellwood , j.a . , & mcgaugh , s.s .
2005 , apj , 634 , 70 simon , j. d. , bolatto , a. d. , leroy , a. , & blitz , l. 2003 , , 596 , 957 simon , j. d. , bolatto , a. d. , leroy , a. , blitz , l. , & gates , e. l. 2005 , , 621 , 757 smith , d. r. , bernstein , g. m. , fischer , p. , & jarvis , m. 2001 , , 551 , 643 sommer - larsen , j. , & dolgov , a. 2001 , , 551 , 608 spekkens , k. , giovanelli , r. , & haynes , m. p. 2005
, , 129 , 2119 spekkens , k. , & giovanelli , r. 2006 , , 132 , 1426 spergel , d. n. , et al .
2006 ( astro - ph/0603449 ) spergel , d. n. , & steinhardt , p. j. 2000 , , 84 , 3760 strigari , l. e. , kaplinghat , m. , & bullock , j. s. 2006 ( astro - ph/0606281 ) swaters , r.a . , madore , b.f . , & trewhella , m. 2000 , , 2000 , , 531 , l107 swaters , r. a. , madore , b. f. , van den bosch , f. c. , & balcells , m. 2003a , , 583 , 732 swaters , r. a. , verheijen , m. a. w. , bershady , m. a. , & andersen , d. r. 2003b , , 587 , l19 syer , d. & white , s.d.m .
1998 , , 293 , 337 thacker , r. j. , & couchman , h. m. p. 2000 , , 545 , 728 tissera , p. & dominguez - tenreiro , r. 1998 , , 297 , 177 tonini , c. , lapi , a. , & salucci , p. 2006
, , 649 , 591 tormen , g. , bouchet , f.r . , & white , s.d.m .
1997 , , 286 , 865 van albada , t. s. & sancisi , r. 1986 , royal society of london philosophical transactions series a , 320 , 447 van den bosch , f. c. 2000 , , 530 , 177 van den bosch , f. c. , mo , h. j. , & yang , x. 2003 , , 345 , 923 van den bosch , f.c . , robertson , b.e . , dalcanton , j.j .
, & de blok , w.j.g .
2000 , , 119 , 1579 van der hulst , j. m. , terlouw , j. p. , begeman , k. g. , zwitser , w. , & roelfsema , p. r. 1992 , asp conf . ser . 25 : astronomical data analysis software and systems i , 25 , 131 verheijen , m. a. w. 2001 , , 563 , 694 verheijen , m. a. w. & sancisi , r. 2001 , , 370 , 765 vikhlinin , a. , kravtsov , a. , forman , w. , jones , c. , markevitch , m. , murray , s. s. , & van speybroeck , l. 2006 , , 640 , 691 weinberg , m.d . , & katz , n. 2002 , , 580 , 627 weiner , b. j. , sellwood , j. a. , & williams , t. b. 2001 , , 546 , 931 wilkinson , m. i. , & evans , n. w. 1999 , , 310 , 645 young , j. s. , & knezek , p. m. 1989 , , 347 , l55 zavala , j. , avila - reese , v. , hernndez - toledo , h. , & firmani , c. 2003 , , 412 , 633 zackrisson , e. , bergvall , n. , marquart , t. , stlin , g. 2006 , , 452 , 857 zentner , a. r. , & bullock , j. s. 2002 , , 66 , 043003 | we examine the amplitude of the rotation velocity that can be attributed to the dark matter halos of disk galaxies , focusing on well measured intermediate radii .
the data for 60 galaxies spanning a large range of mass and hubble types , taken together , are consistent with a dark halo velocity @xmath0 with @xmath1 and @xmath2 over the range @xmath3 kpc .
the range in @xmath4 stems from different choices of the stellar mass estimator , from minimum to maximum disk . for all plausible choices of stellar mass , the implied densities of the dark halos are lower than expected from structure formation simulations in @xmath5cdm , which anticipate @xmath6 .
this problem is not specific to a particular type of galaxy or to the innermost region of the halo ( cusp or core ) ; the velocity attributable to dark matter is too low at all radii . |
the distribution - free chain ladder model ( dfcl ) of mack @xcite is a popular model for stochastic claims reserving . in this paper
we use a time series formulation of the dfcl model which allows for bootstrapping the claims reserves .
an important aspect of this model is that it can provide a justification for the classical deterministic chain ladder ( cl ) algorithm which originally was not founded on an underlying stochastic model .
moreover , it allows for the study of prediction uncertainties .
note that there are different stochastic models that lead to the cl reserves ( see for example wthrich - merz @xcite , section 3.2 ) . in the present paper
we use the dfcl formulation to reproduce the cl reserves .
the paper presents a novel methodology for estimating a bayesian dfcl model utilising a framework of approximate bayesian computation ( abc ) in a non - standard manner .
a methodology utilising markov chain monte carlo ( mcmc ) , abc and a bayesian bootstrap procedure is developed in a distribution - free setting .
the abc framework is required because we work in a distribution - free setting in which we make no parametric assumptions about the form of the likelihood .
effectively , the abc methodology allows us to overcome the fact that we can not evaluate the likelihood point - wise in the dfcl model .
typically , abc methodology circumvents likelihood evaluations by simulation from the likelihood .
however , in this case simulation from the likelihood model is not directly available because no parametric assumption is made .
we combine abc methodology with bootstrap to overcome this additional complexity that the dfcl model presents in the abc framework .
then , by using an mcmc numerical sampling algorithm combined with the novel version of abc that has the embedded bootstrap procedure , we are able to obtain samples from the intractable posterior distribution of the dfcl model parameters .
this allows us to utilise this methodology to obtain the bayesian posterior distribution of the dfcl model parameters empirically .
then we demonstrate two approaches in which we can utilise the posterior samples for the dfcl model parameters to obtain the bayesian predictive distribution of the claims .
the first approach involves using each posterior sample to numerically estimate the full predictive claims distribution given the observed claims .
the alternative approach involves using the posterior samples for the dfcl model parameters to form bayesian point estimators .
then , conditional on these point estimators , we can obtain the bayesian conditional predictive distribution for the claims .
the second approach will be relevant for comparisons with the classical and credibility approaches .
the first approach has the benefit that it integrates out of the bayesian predictive claims distribution the parameter uncertainty associated with estimation of the dfcl chain ladder parameters .
the paper then analyses the parameter estimates in the dfcl model , the associated claims reserves and the mean square errors of prediction ( msep ) from both the frequentist perspective and a contrasting bayesian view .
in doing so we analyse cl point estimators for parameters of the dfcl model , the resulting estimated reserves and the associated msep from the classical perspective .
these include non - parametric bootstrap estimated prediction errors which can be obtained via one of two possible bootstrap procedures , conditional or unconditional . in this paper
we consider the process of conditional back propagation ; see @xcite for in - depth discussion .
these classical frequentist estimators are then compared to bayesian point estimators .
the bayesian estimates considered are the maximum _ a posteriori _ ( map ) and the minimum mean square error ( mmse ) estimators . for comparison with the classical frequentist reserve estimates
, we also obtain the associated bayesian estimated reserves conditional upon the bayesian point estimators .
in addition , since in the bayesian setting we obtain samples from the posterior for the parameters we use these along with the msep obtained by the estimated bayesian point estimators to obtain associated posterior predictive intervals to be compared with the classical bootstrap procedures .
we then robustify the prediction of reserves by rao - blackwellization , that is , we integrate out the influence of the unknown variance parameters in the dfcl model .
having done this , we analyse the resultant msep .
this is again only achievable since in the bayesian setting we obtain samples from the joint posterior for the cl factors and the variances . to summarize our contribution
, the novelty within this paper involves the development and comparison of a new estimation methodology to work with the bayesian cl model for the dfcl model which makes no parametric assumptions on the form of the likelihood function ; see also gisler - wthrich @xcite .
this is unlike the works of yao @xcite and peters et al .
@xcite that assume explicit distributions in order to construct the posterior distributions in the bayesian context .
instead we demonstrate how to work directly with the intractable likelihood functions and the resulting intractable posterior distribution , using novel abc methodology . in this regard
we demonstrate that we do not need to make any parametric assumptions to perform posterior inference , avoiding potentially poor model assumptions made , as for example in the paper of yao @xcite .
* outline of this paper . * the paper begins with a presentation of the claims reserving problem and then presents the model we shall consider .
this is followed by the description of the classical cl algorithm and the construction of a bayesian model that can be used to estimate the parameters of the model .
the bayesian model is constructed in a distribution - free setting .
this is followed by a discussion on classical versus bayesian parameter estimators along with a bootstrap based procedure for the estimation of the parameter uncertainty in the classical setting .
the next section presents the methodology of abc coupled with a novel bootstrap based sampling procedure which will allow us to work directly with the distribution - free bayesian model .
we then illustrate the developed algorithm on a synthetic data set and the real data set , comparing performance to the classical results and those obtained via credibility theory .
we briefly outline the claims development triangle structure we utilise in the formulation of our models .
assume there is a run - off triangle containing claims development data with the structure given in table 1 .
[ table : tab.1 claims development triangle ] .claims development triangles .
[ cols="^ , > , > , > , > , > , > , > , > , > , > " , ] assume that @xmath0 are cumulative claims with indices @xmath1 and @xmath2 , where @xmath3 denotes the accident year and @xmath4 denotes the development year ( cumulative claims can refer to payments , claims incurred , etc ) .
we make the simplifying assumption that the number of accident years is equal to the number of observed development periods , that is , @xmath5 . at time @xmath6
, we have observations @xmath7and for claims reserving at time @xmath6 we need to predict the future claims @xmath8 moreover , we define the set @xmath9 for @xmath10 , that is , @xmath11 is the first column in table 1 . in the classical ( deterministic ) chain ladder algorithm
there is no underlying stochastic model .
it is rather a recursive algorithm that is used to estimate the claims reserves and which has proved to give good practical results .
it simply involves the following recursive steps to predict unobserved cumulative claims in @xmath12 .
set @xmath13 and for @xmath14 @xmath15since this is a deterministic algorithm it does not allow for quantification of the uncertainty associated with the predicted reserves . to analyse the associated uncertainty
there are several stochastic models that reproduce the cl reserves ; for example mack s distribution - free chain ladder model @xcite , the over - dispersed poisson model ( see england - verrall @xcite ) or the bayesian chain ladder model ( see gisler - wthrich @xcite ) .
we use a time series formulation of the bayesian chain ladder model in order to use bootstrap methods and bayesian inference .
we use an additive time series version of the bayes chain ladder model ( model assumptions 3.1 in gisler - wthrich @xcite ) .
[ model ass ] 1 .
we define the cl factors by @xmath16 and the standard deviation parameters by @xmath17 .
we assume independence between all these parameters , i.e. the prior density of @xmath18 is given by @xmath19 where @xmath20 denotes the density of @xmath21 and @xmath22 denotes the density of @xmath23 .
conditionally , given @xmath24 and @xmath25 , we have : * cumulative claims @xmath0 in different accident years @xmath3 are independent . *
cumulative claims satisfy the following time series representation @xmath26where conditionally , given @xmath27 , we have that the residuals @xmath28 are i.i.d . satisfying @xmath29 = 0~\text { and } ~\mathrm{var}\left [ \varepsilon _ { i , j}|{\cal b}_0,\bm{f } , \bm{\xi}\right ] = 1 , \label{residualassumptionsdfcl}\ ] ] and @xmath30=1 $ ] for all @xmath31 . * remark . *
note that the assumptions on the residuals are slightly involved in order to guarantee that cumulative claims @xmath0 are positive @xmath32-a.s .
[ corind ] under model assumptions [ model ass ] we have that conditionally , given @xmath33 , the random variables @xmath34 are independent .
thus , we obtain the following posterior distribution for @xmath35 , given @xmath33 , @xmath36 this result follows from theorem 3.2 in gisler - wthrich @xcite ; from prior independence of the parameters ; and the fact that @xmath37 only depends on @xmath38 , @xmath39 and @xmath0 ( markov property ) .
this has important implications for the abc sampling algorithm developed below . in order to perform the bayesian analysis we make explicit assumptions on the prior distributions of @xmath35 .
[ model ass2 ] in addition to model assumptions [ model ass ] we assume that the prior model for all parameters @xmath40 is given by : * @xmath41 , where @xmath42 is a gamma distribution with mean @xmath43 = \alpha_j \beta_j = \widehat{f}_{j}^{(cl)}$ ] ( see ) and large variance to have diffuse priors . * the variances @xmath44 , where @xmath45 is an inverse gamma distribution with mean @xmath46 = b_j/(a_j-1)=\widehat{\sigma } _ { j}^{2(cl ) } $ ] ( see below ) and large variance .
* remarks * 1 .
the likelihood model is intractable , meaning that no density can be written down analytically in the dfcl model . in formulating the bayesian model we have only made distributional assumptions on the priors for the parameters @xmath35 but not on the observable cumulative claims @xmath0 .
though we make distributional assumptions for the priors , the model is distribution - free because no distributional assumptions on the cumulative claims are made . as a result of only making assumptions on the priors , a standard bayesian analysis using analytic posterior distributions can not be performed .
one way out of this dilemma would be to re - formulate the bayesian model by making distributional assumptions ( for example , this is done in yao @xcite ) but then the model is no longer distribution - free .
another approach would be to use credibility methods ( see gisler - wthrich @xcite ) but this only gives statements for the first two moments . in the present
set up we develop abc methods that allow for a full distributional answer for the posterior distributions without making explicit distributional assumptions for the cumulative claims @xmath0 .
our priors are chosen as diffuse priors with large variances .
this again highlights the differences between specification of the prior distributions and making distributional assumptions for the actual likelihood model , these are mutually exclusive ideas .
we select the priors to ensure that we maintain several relevant aspects of the dfcl model .
in particular , it is important to utilise priors that enforce the strict positivity of the parameters @xmath47 .
we note here that the parametric bayesian model developed in yao @xcite failed in this aspect when it came to prior specification .
therefore we develop an alternative prior structure that satisfies these required properties of the dfcl model .
this section considers both classical and bayesian estimators for the chain ladder framework , including both the chain ladder factors and the variance parameters . in the classical cl method ,
the cl factors are estimated by @xmath48 given in .
the variance parameters are estimated by @xmath49 see ( 3.4 ) in wthrich - merz @xcite .
note that this estimator is only well - defined for @xmath50 .
there is a vast literature and discussion on the estimation of tail parameters .
we do not enter this discussion here but we simply choose the estimator given in mack @xcite for the last variance parameter which is defined by @xmath51 in a bayesian inference context one calculates the posterior distribution of the parameters , given @xmath33 .
as in we denote this posterior by @xmath52 . since the mcmc - abc bootstrap procedure will allow us to obtain samples from the posterior distribution of the bayesian dfcl model presented , we can now consider estimating cl point estimators using these samples .
there are two commonly used point estimators in bayesian analysis that correspond to the posterior mode ( map ) and the posterior mean ( mmse ) , respectively : @xmath53and @xmath54 , \label{mmse_f } \\
\widehat{\sigma } _ { j}^ { ( mmse ) } & = & \int ~\sigma _ { j}~\pi \left ( \sigma _ { j}|\mathcal{d}_{i}\right ) d\sigma _ { j}=e\left[\left .
\xi_j\right|\mathcal{d}_i \right ] .
\label{mmse_var}\end{aligned}\ ] ] in the case in which @xmath55 is not independent of @xmath56 , the map estimators obtained through joint maximization are optimal .
however , in practice one often works with marginal estimators for simplicity . additionally , note that for diffuse priors we find ( see corollary 5.1 in gisler - wthrich @xcite )
@xmath57 hence , using corollary [ corind ] , we obtain the approximation @xmath58 & = & e\left[\left .
c_{i , j}\right|\mathcal{d}_i,\bm{f } , \bm{\xi } \right]\right|\mathcal{d}_i \right]=c_{i , i - i}~e\left[\left .
\prod_{j = i - i}^{j-1}f_j\right|\mathcal{d}_i \right ] \nonumber\\ \label{bayespredictor } & = & c_{i , i - i}~ \prod_{j = i - i}^{j-1}e\left[\left .
f_j\right|\mathcal{d}_i \right ] ~=~c_{i , i - i}~ \prod_{j = i - i}^{j-1}\widehat{f}_{j}^ { ( mmse ) } \\\nonumber & \approx & c_{i , i - i}~ \prod_{j = i - i}^{j-1}\widehat{f}_{j}^{(cl ) } ~=~\widehat{c}_{i , j},\end{aligned}\ ] ] where on the last line we have an equality if the diffusivity of the priors @xmath20 tends to infinity .
this is exactly the argument why the bayesian cl model can be used to justify the cl predictors ; see gisler - wthrich @xcite .
in addition , the posterior samples for the dfcl model parameters , obtained via the mcmc - abc bootstrap procedure , will allow us to obtain the predictive distribution of the claims in two ways . the first is the full predictive distribution of the claims obtained after integrating out the posterior uncertainty associated with the bayesian dfcl model parameters to empirically estimate @xmath59 in practice , this numerical procedure involves taking each posterior sample for the dfcl model parameters and obtaining an estimate of the predicted claims .
the second approach involves using one of the bayesian point estimators for the parameters such as the mmse to obtain @xmath60 .
alternatively , one may consider a rao - blackwellised version of the bayesian predictive distribution of claims involving @xmath61 having numerically integrated out the bayesian posterior uncertainty associated with the dfcl variance parameters .
such methods are typically known as empirical bayesian approaches .
these results can then be applied to estimate any risk measures .
for example , if we fix a security level 95% we can calculate the var on that level , which is defined by @xmath62\bigg| \mathcal{d}_i\bigg ) = \min \left\ { x;~ p \bigg [ c_{i , j}-e\left[c_{i , j}|\mathcal{d}_i\right ] > x \bigg| \mathcal{d}_i \bigg ] \le 0.05 \right\}.\ ] ]
assume that we have calculated the bayesian predictor or the cl predictor given in .
then we would like to determine the prediction uncertainty , that is , we would like to study the deviation of @xmath63 around its predictor .
if one is only interested in second moments , the so - called conditional mean square error of prediction ( msep ) , one can often estimate the error terms analytically .
however , other uncertainty measures like value - at - risk ( var ) can only be determined numerically ; see ( [ eqns ] ) .
a popular numerical method is the bootstrap method .
the bootstrap technique was developed by efron @xcite and extended by efron - tibshirani @xcite and davison - hinkley @xcite . in the actuarial literature
the development of bootstrap procedures includes the work of taylor @xcite , taylor - mcguire @xcite , @xcite , england - verrall @xcite , @xcite and pinheiro et al .
@xcite .
this procedure allows one to obtain information regarding an aggregated distribution given a single realisation of the data .
to apply the bootstrap procedure one introduces a minimal amount of model structure such that resampling observations can be achieved using observed samples of the data . in this section
we present a bootstrap algorithm in the classical frequentist approach .
that is , we assume that the cl factors @xmath64 and the standard deviation parameters @xmath65 given in model assumptions [ model ass ] are unknown constants .
the bootstrap then generates synthetic data denoted by @xmath66 that allow for the study of the fluctuations of @xmath67 and @xmath68 ( for details see section 7.4 in wthrich - merz @xcite ) . in the presented text
we restrict ourselves to the conditional resampling approach presented in section 7.4.2 of wthrich - merz @xcite . 1 .
calculate estimated residuals @xmath69 for @xmath70 , @xmath71 , conditional on the estimators @xmath72 and @xmath73 and the observed data @xmath74 : @xmath75 2 .
these residuals @xmath76 give the empirical bootstrap distribution @xmath77 .
3 . sample i.i.d .
residuals @xmath78 for @xmath70 , @xmath71 .
4 . generate bootstrap observations ( conditional resampling ) @xmath79which defines @xmath80 .
note that for the unconditional version of bootstrap we should generate @xmath81 . for a discussion on this approach , see section 7.4.1 of @xcite .
5 . calculate bootstrapped cl parameters @xmath82and @xmath83 by @xmath84 6 .
repeat steps 3 - 5 and obtain empirical distributions from the bootstrap samples @xmath85 , @xmath82 and @xmath83 .
these are then used to quantify the parameter estimation uncertainty .
this non - parametric classical bootstrap method can be seen as a frequentist approach .
this means that we do not express our parameter uncertainty by the choice of an appropriate prior distribution .
we rather use a point estimator for the unknown parameters and then study the possible fluctuations of this point estimator .
the main difficulty now is that the non - parametric bootstrap method , as described above , underestimates the `` true '' uncertainty .
this comes from the fact that the estimated residuals @xmath86 , in general , have variance smaller than 1 ( see formula ( 7.23 ) in wthrich - merz @xcite ) .
this means that our estimated residuals are not appropriately scaled .
therefore , frequentists use several different scalings to correct this fact ( see formula ( 7.24 ) in wthrich - merz @xcite or england - verrall @xcite ) . here
, we use a different approach by introducing the novel bayesian bootstrap method embedded within an mcmc - abc algorithm to obtain empirically the posterior distribution of the bayesian dfcl model , described below . having obtained this , we can then calculate all required bayesian parameter estimates , capital reserve estimates and associated risk measures such as var . before presenting the methodology for this novel mcmc - abc algorithm
we will finalize this section with the decompositions of the msep under frequentist , bayesian and credibility approaches .
let us for the time - being concentrate on the conditional msep given by @xmath87\\ \nonumber & = & \text{var}\left(\left.c_{i , j}\right| \mathcal{d}_i\right ) + \left(e\left[\left.c_{i , j } \right| \mathcal{d}_i\right]-\widehat{c}_{i , j}\right)^2.\end{aligned}\ ] ] the first term is known as the conditional process variance and the second term as the parameter estimation uncertainty . in the frequentist approach ( i.e. for given deterministic @xmath64 and @xmath65 ) these terms
can be calculated as @xmath88 \bigr)^2~ \sum_{j = i - i}^{j-1}~ \frac{\sigma^2_j / f^2_{j}}{e\left.\left[c_{i , j}\right|c_{i , i - i}\right ] } \stackrel{def.}{=}c_{i , i - i } \gamma_{i - i},\ ] ] and @xmath89-\widehat{c}_{i , j}\right)^2 = c^2_{i , i - i}\left(\prod_{j={i - i}}^{j-1}f_j -\prod_{j={i - i}}^{j-1}\widehat{f}^{(cl)}_j\right)^2 \stackrel{def.}{=}c_{i , i - i}^2 \delta_{i - i};\ ] ] see wthrich - merz @xcite , section 3.2 .
the process variance is estimated by replacing the parameters by its estimators , @xmath90 the parameter estimation error is more involved and there we need the bootstrap algorithm .
assume that the bootstrap method gives @xmath91 bootstrap samples @xmath92 .
then the parameter estimation error is estimated by the sample variance of the product of the bootstrap observation chain ladder parameter estimates @xmath92 , which gives the estimator @xmath93 . in the bayesian setup , ( i.e. choosing prior distributions for the unknown parameters @xmath94 and @xmath95 ) we obtain a natural decomposition of the conditional msep : @xmath96\right ) & = & \text{var}\left(\left.c_{i , j}\right| \mathcal{d}_i\right)\\ \nonumber & = & e\left[\left.\text{var}\left(\left.c_{i , j } \right| \mathcal{d}_i , \bm{f},\bm{\xi}\right)\right| \mathcal{d}_i\right ] + \text{var}\left(\left.e\left[\left.c_{i , j } \right| \mathcal{d}_i,\bm{f},\bm{\xi}\right]\right| \mathcal{d}_i\right).\end{aligned}\ ] ] the average process variance is given by ( see wthrich - merz @xcite , lemma 3.6 ) @xmath97 = c_{i , i - i } \sum_{j = i - i}^{j-1 } e\left[\left
. \prod_{m = i - i}^{j-1}f_m ~\xi_j^2 \prod_{n = j+1}^{j-1 } f_n^2 \right|\mathcal{d}_i\right]\\ & & \quad = c_{i , i - i } \sum_{j = i - i}^{j-1 } \prod_{m = i - i}^{j-1}e\left[\left.f_m \right|\mathcal{d}_i\right ] e\left[\left.\xi_j^2 \right|\mathcal{d}_i\right ] \prod_{n = j+1}^{j-1 } e\left[\left.f_n^2 \right|\mathcal{d}_i\right]\stackrel{def.}{=}c_{i , i - i } \widehat{\gamma}_{i - i}^{bayes } , \nonumber\end{aligned}\ ] ] where we have used posterior independence ( [ posteriorindependence ] ) .
the parameter estimation error is given by @xmath98\right| \mathcal{d}_i\right ) = c_{i , i - i}^2 \text { var}\left ( \left.\prod_{j = i - i}^{j-1 } f_j \right| \mathcal{d}_i\right)\stackrel{def.}{=}c_{i , i - i}^2 \widehat{\delta}_{i - i}^{bayes},\ ] ] where we have used .
using ( [ posteriorindependence ] ) , we obtain for the last term @xmath99 -\prod_{j = i - i}^{j-1}e\left [ \left .
\mathcal{d}_i\right]^2 \right].\ ] ] in order to calculate these two terms given in and , we need to calculate the posterior distribution of @xmath35 , given @xmath33 .
since we do not have a full distributional model , we can not write down the likelihood function , which would allow for analytical solutions or markov chain monte carlo ( mcmc ) simulations . therefore we introduce the abc framework which allows for distribution - free simulations using appropriate bootstrap samples and a distance metric .
this will be discussed in section [ abcsection ] . as mentioned previously
, we can also consider the credibility estimates given in gisler - wthrich @xcite .
as long as we are only interested in the second moments ( i.e. conditional msep ) we can also use credibility estimators , which are minimum variance estimators that are linear in the observations . for diffuse priors we obtain the approximation given in corollary 7.2 of gisler - wthrich @xcite @xmath100\right ) = c_{i , i - i}\widehat{\gamma } _ { i - i}^{cred}+c_{i , i - i}^{2}\widehat{\delta } _ { i - i}^{cred},\]]where@xmath101 in the results section we compare the frequentist bootstrap approach , the credibility approach and the abc bootstrap approach that is described below ( see table 7 below ) .
to estimate numerically the parameters , predicted claims and associated uncertainty measures such as the msep presented in the previous sections , the bayesian approach requires the ability to sample from the posterior distribution of the dfcl model parameters . obtaining samples @xmath102 which are realisations of a random vector distributed with a posterior distribution @xmath103 in the dfcl model is difficult since the likelihood is intractable .
hence , standard numerical approaches such as markov chain monte carlo ( mcmc ) algorithms ( see gilks et al .
@xcite ) can not be directly used since they all require explicit repeated evaluation of the likelihood function at each stage of the markov chain sampling algorithm .
it is common to avoid this difficulty by making distributional assumptions for the form of the likelihood .
this then violates the dfcl model assumption but allows for relatively standard sampling procedures to be applied . in this regard ,
one possible approach involves making a specific gaussian assumption for the likelihood .
one problem with this assumption , which is evident immediately , is that it precludes skewness in the model . here
, we do not make any such assumptions and instead we work in a truly distribution - free model using abc to facilitate sampling from an intractable posterior distribution .
there is an additional complexity in the dfcl model not typically encountered when working with abc methodology .
typically , abc methodology is developed in the case in which the model likelihood can not be evaluated point - wise , but conditional on parameter values , synthetic data is easily simulated from the model ; see examples in peters - sisson @xcite and peters et al .
this is not the case in the bayesian dfcl model . under the dfcl model
the likelihood is only expressed by moment conditions , hence we can not evaluate the likelihood point - wise and also the simulation from the likelihood can not be performed directly .
this is why we introduce the novel concept of the bayesian bootstrap which is embedded within the abc methodological framework .
hence , to sample from the posterior in our dfcl model we develop a novel formulation of the abc methodology based on the bootstrap and conditional back transformation procedure , similar to that discussed in section [ section : bootstrap dfcl ] .
abc methods aim to sample from posterior distributions in the presence of computationally intractable likelihood functions . for an application in risk modelling of abc methodology , see peters - sisson @xcite . in this article
we present a novel mcmc - abc algorithm . before presenting some details of the numerical mcmc procedure
, we note that alternative numerical algorithms could be considered in the abc context .
for example , a sequential monte carlo ( smc ) based algorithms which can improve simulation efficiency can be found in del moral et al .
@xcite , sisson et al .
@xcite , peters et al .
@xcite,@xcite and marjoram et al .
@xcite . in this section
we provide a brief description of abc methodology , which describes a suite of methods developed specifically for working with models in which the likelihood is computationally intractable . here
we work with a bayesian model and consider the likelihood intractability to arise in the sense that we may not evaluate the likelihood point - wise .
the abc method we consider here embeds an intractable target posterior distribution , in our case denoted by @xmath104 , into a general augmented model @xmath105 where @xmath106 is an auxiliary vector on the same space as @xmath74 . in this augmented bayesian model ,
the weighting function @xmath107 weights the intractable posterior . in this paper
we consider the hierarchical model assumption , where we work with @xmath108 ; see reeves and pettitt @xcite . the mechanism in the abc framework which allows one to avoid the evaluation of the intractable likelihood involves replacing this evaluation with data simulation from the likelihood .
that is , given a realisation of the parameters of the model , a synthetic data set @xmath106 is generated and compared to the original data set .
this is a key aspect of the novel methodology we develop in this paper , since we utilise a bootstrap procedure to perform this simulation in the dfcl model setting .
then summary statistics @xmath109 derived from this data are compared to summary statistics of the observed data @xmath110 and a distance @xmath111 is calculated .
finally , a weight is given to these parameters according to the weighting function @xmath112 , which may give greater weight when @xmath109 and @xmath110 are close ( i.e. where @xmath111 is small ) .
for example , under the `` hard decision '' ( hd ) weighting given by @xmath113 a reward is given to summary statistics of the augmented auxiliary variables @xmath114 within an @xmath115-tolerance of the summary statistic of the actual observed data @xmath116 , as measured by distance metric @xmath117 .
hence , in the abc context , an approximation to the intractable target posterior marginal distribution @xmath104 , for which we are interested in formulating an empirical estimate , is given by @xmath118 as briefly mentioned , obtaining samples from the abc posterior can be achieved using a number of numerical procedures , in this paper we consider an mcmc approach .
the mcmc class of likelihood - free algorithm is justified on a joint space formulation , in which the stationary distribution of the markov chain is given by @xmath119 .
the corresponding target distribution for the marginal distribution @xmath120 as @xmath121 , recovering the `` true '' ( intractable ) posterior , assuming that @xmath116 are sufficient statistics and that the weighting function converges to a point mass on @xmath116 as @xmath122 ; see peters - sisson @xcite and references therein for detailed discussion .
accordingly , the tolerance @xmath115 is typically set as low as possible for a given computational budget . in this paper we focus on the class of mcmc - based sampling algorithms . the abc methodology is novel both in the statistics literature and in the actuarial literature .
it is informative to clearly provide the justification for this approach both theoretically and numerically .
the simplest understanding of abc is achieved by considering a rejection algorithm , therefore we provide a basic argument for how the abc methodology works in simple rejection sampling in appendix [ section : justification for abc ] .
the actuarial dfcl model considered in this paper requires the more sophisticated mcmc - abc methodology described below . for given observations
@xmath74 we want to sample from @xmath123 with an intractable likelihood function .
we assume that @xmath110 is either the data itself or a summary of the data such as a sufficient statistic for the model from which we assume data @xmath74 is a realisation .
we assume that , given a set of parameters values @xmath124 , we can generate from the dfcl model ( via a conditional bootstrap procedure ) a synthetic data set denoted @xmath125 .
we define a hard decision function @xmath126 for a given tolerance level @xmath127 and a distance metric @xmath128 , where @xmath129 is the indicator function which equals 1 if the event is true and 0 otherwise .
as demonstrated in appendix a , we use the approximation , - , which gives us in the bayesian dfcl model setting , @xmath130 } { e\left[g(\mathcal{d}_{i}|\mathcal{d}_{i}^{\ast})\right]}.\ ] ] in the next step the numerator of is approximated using the empirical distribution : @xmath131 \approx \pi(\bm{f},\bm{\sigma } ) \frac{1}{l } \sum_{l=1}^l g\left ( \mathcal{d}_{i}|\mathcal{d}_{i}^{\ast,(l)}(\bm{f},\bm{\sigma})\right),\ ] ] where @xmath132 . finally , we need to consider the denominator @xmath133 $ ] . in general
this has a non - trivial form that can not be calculated analytically .
however , since we use an mcmc based method the denominators cancel in the accept - reject stage of the algorithm
. therefore , the intractability of the denominator does not impede sampling from the posterior .
thus we use @xmath134\\ & \approx \pi(\bm{f } ) \pi(\bm{\sigma } ) \frac{1}{l } \sum_{l=1}^l g\left(\mathcal{d}_{i}|\mathcal{d}_{i}^{\ast,(l)}(\bm{f},\bm{\sigma})\right ) \end{aligned}\ ] ] in order to obtain samples from @xmath135 .
almost universally , @xmath136 is adopted to reduce computation but on the other hand this will slow down the rate of convergence to the stationary distribution .
note that sometimes one also uses softer decision functions for @xmath137 .
the role of the distance measure @xmath138 is evaluated by peters et al .
we further extend this analysis to the class of models considered in this paper .
we analyse several choices for the distance measure @xmath138 such as mahlanobis distance , scaled euclidean distance and the manhattan `` city block '' distance .
fan et al .
@xcite demonstrate that it is not efficient to utilise the standard euclidean distance , especially when summary statistics considered are on different scales .
additionally , using an mcmc - abc algorithm , it is important to assess convergence diagnostics . particularly when using mcmc - abc where serial correlation in the markov chain samples can be significant if the sampler is not designed carefully .
we assess autocorrelation of the simulated markov chain , the geweke @xcite time series statistic and the gelman - rubin @xcite r - statistic convergence diagnostic in an abc setting . *
concluding : * we apply three different techniques in order to treat the intractable likelihood : 1 .
abc is used to get a handle on the likelihood and therefore the intractable posterior .
2 . as a result of using abc
we need to be able to generate synthetic data samples from the dfcl model given realisations of the parameters .
these data samples come from the bootstrap algorithm .
we use a well understood mcmc based sampling algorithm that does not require calculation of the non - analytic normalizing constants for the target distribution @xmath135 .
the reason for this is that in the acceptance probability of the mcmc algorithm , the normalizing constant for the target posterior appears both in the numerator and denominator , resulting in cancellation .
the specific details of the mcmc algorithm and abc choices are provided in the appendix b.
to test the accuracy of the methodology , first we use synthetic data generated with known parameter values .
the tuning of the proposal distribution in this study is done for the simplest `` base '' distance metric , the weighted euclidean distance . to study the effect of the distance metric in a comparative fashion we shall keep the proposal distribution unchanged .
the first example we present has a claims triangle of size @xmath139 .
in this example we fix the true model parameters , denoted by @xmath140 and @xmath141 and given in table [ tab2 ] , used to generate the synthetic data set . to generate the synthetic observations for @xmath74 , we generate randomly the first column ( i.e. @xmath11 ) .
then conditional on this realisation of @xmath11 we make use of the model given in ( [ model ass ] ) to generate the remaining columns of @xmath74 , ensuring the model assumptions are satisfied .
this requires setting @xmath142 sufficiently large ( for appropriate choices of @xmath143 and @xmath144 ) and then sampling i.i.d .
realisations of @xmath145 $ ] used to obtain @xmath33 ; see the observations in table [ tab2 ] .
we perform a sensitivity analysis , studying the impact of the distance metric on the mixing of the markov chain in the case of joint estimation of the chain ladder factors and the variance parameters .
the pre - tuned coefficient of variation of the gamma proposal distribution for each parameter of the posterior was performed using the following settings ; @xmath146 , @xmath147 , @xmath148 and initial values @xmath149 for all @xmath150 .
additionally , the prior parameters for the chain ladder factors @xmath21 were set as @xmath151 and the parameters for the variance parameters @xmath152 were set as @xmath153 .
after tuning the proposal distributions during burn - in and rounding the shape parameters , we found that @xmath154 for all @xmath155 produced average acceptance probabilities for each parameter between 0.3 and 0.5 .
this is a range typically used in practice when designing mcmc sampling algorithms .
then , keeping the proposal distribution constant and using a common data set @xmath74 , we ran three versions of the mcmc - abc algorithm for 200,000 samples corresponding to : 1 . scaled euclidean distance and joint estimation of posterior for @xmath156 ; 2 .
mahlanobis distance ( modified ) and joint estimation of posterior for @xmath156 ; and 3 .
manhattan `` city block '' distance and joint estimation of posterior for @xmath156 .
we estimate the three convergence diagnostics given in appendix b. the results of this analysis are presented as a function of markov chain iteration @xmath157 post burn - in of 50,000 samples .
* autocorrelation function * : figure [ fig1 ] shows the estimated autocorrelation functions for the markov chains of the random variables @xmath158 and @xmath159 .
we analyze the marginal parameters to get a reasonable estimate of the mixing behavior of the mcmc - abc algorithm .
the results demonstrate the degree of serial correlation in the markov chains generated for these parameters as a function of lag time @xmath160 .
the higher the decay rate in the tail of the estimated acf as a function of @xmath160 , the better the mixing of the mcmc algorithm . due to the independence properties of this model
there is little difference between results obtained for scaled euclidean and mahlanobis distances . as shown in appendix c ,
the estimate of the covariance matrix is diagonal on all but the right lower @xmath161 block .
hence , we recommend using the simple scaled euclidean distance metric as it provided the best trade - off between simplicity and mixing performance .
* geweke time series diagnostic * : figure [ fig2 ] shows results for the geweke time series diagnostic .
again , we present the results for the random variables @xmath158 and @xmath159 .
note , we used the posterior mean as the sample function and a set of increasing values for @xmath162 from @xmath163 increasing in steps of 5,000 samples to @xmath91 . in each case
we split the chain in each `` window '' given by @xmath164 and @xmath165 according to recommendations from geweke et al .
we then calculate the convergence diagnostic @xmath166 which is the difference between these two means divided by the asymptotic standard error of their difference . as the chain length increases @xmath167 , the sampling distribution of @xmath168 if the chain has converged .
hence values of @xmath166 in the tails of a standard normal distribution suggest that the chain was not fully converged early on ( i.e. during the 1st window ) .
hence , we plot @xmath166 scores versus increasing @xmath162 and monitor if they lie within a @xmath169 confidence interval @xmath170 $ ] .
the results in figure [ fig2 ] clearly demonstrate the convergence properties of the distance functions differ .
again this is more material in the markov chain for the variance parameter when compared to the markov chain results for the chain ladder factor .
the main point we note is that again one would advise against use of the `` city block '' distance metric . *
gelman and rubin r statistic * : figure [ fig3 ] presents the gelman and rubin convergence diagnostic . to calculate this we ran 20 chains in parallel , each of length 10,000 samples and for each chain we discarded 250 samples as burn - in .
we then estimated the @xmath171 statistic as a function of simulation time post burn - in .
figure [ fig3 ] shows the convergence rate of the @xmath171 statistic to 1 for each distance metric on increasing blocks of 200 samples . using this summary statistic , all three distance metrics are very similar in terms of convergence rate of the r statistic to 1 .
overall , these three convergence diagnostics demonstrate that the simple scaled euclidean distance metric is the superior choice .
secondly , we see appropriate convergence of the markov chains under three convergence diagnostics which tests different aspects of the mixing of the markov chains , giving confidence in the performance of the mcmc - abc algorithm for this model . in this section we present results for the scaled euclidean distance metric , with a markov chain of length 200,000 samples discarding the first 50,000 samples as burn - in .
table [ tab3 ] shows the cl parameter estimates for the dfcl model and the associated parameter estimation error .
we define the following quantities : * @xmath172 , @xmath173 , @xmath174 and @xmath175|\sigma_{0:j-1}$ ] denote respectively the maximum a - posteriori , minimum mean square error , posterior standard deviation of the conditional distribution of chain ladder factor @xmath21 and the posterior coverage probability estimates at @xmath176 of the conditional distribution of chain ladder factor @xmath21 . each of these estimates is conditional on knowledge of the true @xmath177 .
* @xmath178 , @xmath179 , @xmath180 and @xmath175 $ ] denote the same quantities for the unconditional distribution after joint estimation of @xmath181 and @xmath182 .
* @xmath183 $ ] and @xmath184 $ ] denote the average acceptance probabilities of the markov chain .
* @xmath185 , @xmath186 , @xmath187 and @xmath175 $ ] denote the same quantities for the chain ladder variances as those defined above for chain ladder factors .
note , the estimates for @xmath178 and @xmath188 were obtained marginally . for the frequentist approach
we obtain the standard error in the estimates by using 1,000 bootstrap realisations of @xmath189 to obtain @xmath190 .
we use these bootstrap samples to calculate the standard deviation in the estimates of the parameters in the classical frequentist cl approach , given in brackets @xmath191 next to their corresponding estimators .
the standard errors in the bayesian parameter estimates are obtained by blocking the markov chain into 100 blocks of length 1,500 samples and estimating the posterior quantities on each block .
in this example we consider estimation using real claims reserving data from wthrich - merz @xcite , see table 3 .
this yearly loss data is turned into annual cumulative claims and divided by 10,000 for the analysis in this example .
we use the analysis from the previous study to justify use of the joint mcmc - abc simulation algorithm with a scaled euclidean distance metric .
we pre - tuned the coefficient of variation of the gamma proposal distribution for each parameter of the posterior .
this was performed using the following settings : @xmath146 , @xmath192 , @xmath193 and initial values @xmath149 for all @xmath194 .
here we make a strict requirement of the tolerance level to ensure we have accurate results from our abc approximation .
additionally , the prior parameters for the chain ladder factors @xmath21 were set as @xmath195 and the parameters for the variance @xmath152 priors were set as @xmath196 .
the code for this problem was written in matlab and it took approximately 10 min to simulate 200,000 samples from the mcmc - abc algorithm on intel xeon 3.4ghz processor with 2 gb ram .
after tuning the proposal distributions during burn - in we obtained rounded shape parameters + @xmath197 $ ] provided average acceptance probabilities between 0.3 and 0.5 . * estimates of @xmath143 and @xmath198 * + figures [ fig4 ] presents box - whisker plots of estimates of the distributions of the parameters @xmath181 and @xmath182 obtained from the mcmc - abc algorithm , post burn - in .
figure [ fig5 ] shows the bayesian mcmc - abc empirical distributions of the ultimate claims , @xmath63 for @xmath199 . in table
[ tab5 ] we present the predicted cumulative claims for each year along with the estimates for the chain ladder factors and chain ladder variances under both the classical approach and the bayesian model .
we see that with this fairly vague prior specified , we do indeed obtain convergence of the mcmc - abc based bayesian estimates @xmath200 to the classical estimates @xmath201 .
* dependence on tolerance @xmath115 * + figure [ fig6 ] presents a study of the histogram estimate of the marginal posterior distribution for chain ladder factor @xmath202 .
the plot was obtained by sampling from the full posterior @xmath203 for each specified tolerance value , @xmath204 .
then the samples for the particular chain ladder parameter in each plot are turned into a smoothed histogram estimate for each @xmath204 and plotted .
the results of this analysis demonstrated that when @xmath115 is large , in this model greater than around @xmath205 , the likelihood is not having an influence on the abc posterior distribution .
hence , under an mcmc - abc algorithm , this results in acceptance probabilities for the chain being artificially high , resulting in estimates of the posterior which reflect the prior distribution used ( in this case a vague prior ) . as @xmath204 is reduced , we notice that the changes in the estimate of the posterior distribution also reduces .
the aim of this study is to demonstrate that once @xmath204 reaches a small enough level , the effect of reducing it further is minimal on the posterior distribution .
we see that changing @xmath204 from @xmath206 to @xmath207 has not had a material impact on the posterior mean or variance , the change is less than @xmath208 . as a result , reducing @xmath204 past this point can not be justified relative to the significant increase in computational effort required to achieve such a further reduction in @xmath204 .
ultimately , we would like an algorithm which could work well for any @xmath204 , the smaller the better .
however , we note that with a decreasing @xmath204 in the sampler we present in this paper , one must take additional care to ensure the markov chain is still mixing and not `` stuck '' in a particular state , as is observed to be the case in all mcmc - abc algorithms . to avoid this acknowledged difficulty with mcmc - abc , one should run much longer mcmc chains or alternatively use of more sophisticated sampling algorithms such as smc samplers prc - abc based algorithms ; see sisson et al .
@xcite .
the conclusion of these findings is that a value of @xmath193 , which was used for the analysis of the data in this paper , is suitable numerically and computationally .
+ + + + + * var and msep . * + in table [ tab6 ] we present the predictive var at @xmath169 and @xmath209 levels for the ultimate predicted claims , obtained from the mcmc - abc algorithm .
these are easily obtained under the bayesian setting , using the mcmc - abc posterior samples to explicitly obtain samples from the full predictive distribution of the cumulative claims after integrating out the parameter uncertainty numerically .
in addition to this , we present the analysis of the msep under the bootstrap frequentist procedure and the bayesian mcmc - abc and credibility estimates for the total predicted cumulative claims for each accident year @xmath3 . we also present results for the sum of the total cumulative claims for each accident year , and the associated parameter uncertainty and process variance ( see section [ section : bootstrap dfcl ] for details ) .
we can make the following conclusions from these results : 1 .
the estimates of process variance for each @xmath63 demonstrate that the frequentist bootstrap and the credibility estimates are very close for all accident years @xmath3 .
the bayesian results compare favorably with the credibility results .
2 . the results for the parameter estimation error for the predicted cumulative claims @xmath63 demonstrate for small @xmath3 that the bayesian approach results in a smaller estimation error compared to the frequentist approach . for large @xmath3
, the bayesian approach produces larger estimation error relative to the credibility approach .
the total results for the process variance for @xmath210 demonstrate that the frequentist and credibility results are very close . additionally , bayesian total results are largest followed by credibility and then frequentist estimates which is in agreement with theoretical bounds . 4 . the total results for the parameter estimation error for @xmath210 demonstrate that frequentist unconditional bootstrap procedure results in the lowest total error .
the bayesian approach and credibility total parameter errors are close .
additionally , we note that the results in table 7.1 of wthrich - merz @xcite , for the total parameter estimation error under an unconditional frequentist bootstrap with unscaled residuals is also very close to the total obtained under the frequentist approach .
this paper has presented a distribution - free claims reserving model under a bayesian paradigm .
a novel advanced mcmc - abc algorithm was developed to obtain estimates from the resulting intractable posterior distribution of the chain ladder factors and chain ladder variances .
we assessed several aspects of this algorithm , including the properties of the convergence of the mcmc algorithm as a function of the distance metric approximation in the abc component .
the methodologies performance was demonstrated on a synthetic data set generated from known parameters .
next , it was applied to a real claims reserving data set .
the results we obtained for predicted cumulative ultimate claims were compared to those obtained via classical chain ladder methods and via credibility theory .
this clearly demonstrated that the algorithm is working accurately and provides us not only with the ability to obtain point estimates for the first and second moments of the ultimate cumulative claims , but also with an accurate empirical approximation of the entire distribution of the ultimate claims .
this is valuable for many reasons , including prediction of reserves which are not based on centrality measures such as the tail based var results we present .
* * * * the fist author thanks eth fim and eth risk lab for their generous financial assistance whilst completing aspects of this work at eth .
the first author also thanks the department of mathematics and statistics at the university of nsw for support through an australian postgraduate award and to csiro for support through a postgraduate research top up scholarship .
finally , this material was based upon work partially supported by the national science foundation under grant dms-0635449 to the statistical and applied mathematical sciences institute , north carolina , usa . any opinions , findings , and conclusions or recommendations expressed in this material are those of the author(s ) and do not necessarily reflect the views of the national science foundation 9 davison , a.c . ,
hinkley , d.v .
( 1997 ) . _ bootstrap methods and their application_. cambridge university press , cambridge .
del moral , p. , doucet , a. , jasra , a. ( 2006 ) .
sequential monte carlo samplers .
_ journal of the royal statistical society series .
@xmath211(3 ) , 411 - 436 . efron , b. ( 1979 ) .
bootstrap methods : another look at the jackknife . _
annals of statistics_. @xmath212(1 ) , 1 - 26 .
efron , b. , tibshirani , r.j .
( 1993 ) . _
an introduction to the bootstrap_. chapman & hall , ny . england , p.d . ,
verrall , r.j .
analytic and bootstrap estimates of prediction errors in claims reserving . _
insurance : mathematics and economics_. @xmath213(3 ) , 281 - 293 .
england , p.d . ,
verrall , r.j .
stochastic claims reserving in general insurance .
_ british actuarial journal_. @xmath214(3 ) , 443 - 518 .
england , p.d . ,
verrall , r.j .
predictive distributions of outstanding liabilities in general insurance . _
annals of actuarial science_. @xmath215(2 ) , 221 - 270 .
fan , y. , sisson , s.a . , peters , g.w .
improved efficiency in approximate bayesian computation .
_ technical report , statistics department , university of new south wales_. gelman , a. , rubin , d.b .
inference from iterative simulation using multiple sequences .
_ statistical science _
@xmath216 , 457 - 472 .
geweke , j.f .
( 1991 ) . _
evaluating the accuracy of sampling - based approaches to the calculation of posterior moments_. in j.m .
bernardo , j.o .
berger , a.p .
dawid and a.f.m .
smith ( eds . ) bayesian statistics , 4 , oxford univeristy press , oxford .
gilks , w.r .
, richardson , s. , spiegelhalter , d.j .
_ markov chain monte carlo in practice_. chapman & hall , london .
gisler , a. , wthrich , m.v .
credibility for the chain ladder reserving method .
_ astin bulletin _
@xmath217(2 ) , 483 - 526 .
gramacy , r.b . ,
samworth , r.j . , king , r. ( 2008 ) .
importance tempering .
_ preprint , arxiv : 0707.4242v5 [ stat.co]_. mack , t. ( 1993 ) . distribution - free calculation of the standard error of chain ladder reserve estimates . _ astin bulletin _ @xmath218 , 213 - 225 .
marjoram , p. , molitor , j. , plagnol , v. , tavare , s. ( 2003 ) .
markov chain monte carlo without likelihoods .
_ proceedings of the national academy of science usa _
@xmath219 , 15324 - 15328 .
peters , g.w . , sisson , s.a .
bayesian inference , monte carlo sampling and operational risk .
_ journal of operational risk _
@xmath215(3 ) , 27 - 50 .
peters , g.w . ,
fan , y. , sisson , s.a .
( 2008 ) . on sequential monte carlo , partial rejection control and approximate bayesian computation .
_ preprint , statistics department , university of new south wales_. peters , g.w . , sisson , s.a . , fan , y. ( 2008 ) . design efficiency for `` likelihood free '' sequential monte carlo samplers .
_ preprint , statistics department , university of new south wales .
_ pinheiro , p.j.r .
, andrade e silva , j.m .
and de lourdes centeno , m. ( 2003 ) .
bootstrap methodology in claim reserving . _
journal of risk insurance _
@xmath220(4 ) , 701 - 714 .
roberts , g.o . , gelman , a. , gilks , w.r .
weak convergence and optimal scaling of random walk metropolis algorithm . _
annals of applied probability _
@xmath212 , 110 - 120 .
peters , g.w . , shevchenko , p. , wthrich , m.v . , ( 2008 ) .
model uncertainty in claims reserving within tweedie s compound poisson models .
_ astin bulletin _ * 39*,1 - 33 .
peters , g.w . , nevat , i. , sisson , s.a . , fan , y. , yuan , j. ( 2009 ) .
bayesian symbol detection in wireless relay networks .
accepted to appear in _ ieee transactions on signal processing_. proakis , j.g . ,
manolakis , d.g . , ( 1996 ) .
_ digital signal processing_. upper saddle river , n.j .
prentice hall .
reeves , r.w . , pettitt , a.n .
( 2005 ) . a theoretical framework for approximate bayesian computation . presented at the international workshop for statistical modelling , sydney .
sisson , s.a . ,
fan , y. , tanaka , m. ( 2007 ) .
sequential monte carlo without likelihoods .
_ proceedings of the national academy of science usa _
@xmath221 , 1760 - 1765 .
sisson , s.a . , peters , g.w . , fan , y. , briers , m. , ( 2008 ) .
likelihood free samplers .
_ preprint , statistics department , university of new south wales .
_ taylor , g.c . ( 1987 ) .
regresion models in claims analysis i : theory .
lxxiv , 354 - 383 .
taylor , g.c . ,
mcguire , g. ( 2005 ) .
synchronous bootstrapping of seemingly unrelated regressions .
conference paper , _
36th astin colloquium , 2005 _ zurich , switzerland .
taylor , g.c . ,
mcguire , g. ( 2007 ) . a synchronous bootstrap to account for dependencies between lines of business in the estimation of loss reserve prediction error . _ north american actuarial journal _
@xmath214(1 ) , 37 - 44 .
wthrich , m.v .
, merz , m. ( 2008 ) .
_ stochastic claims reserving methods in insurance_. wiley finance .
yao , j. ( 2008 ) .
bayesian approach for prediction error in chain - ladder claims reserving .
_ conference paper presented at the astin colloqium _
, manchester uk .
the abc algorithm is typically justified in the simple rejection sampling framework .
this then extends in a straightforward manner to other sampling frameworks such as the mcmc algorithm we utilise in this paper .
we denote the posterior density from which we wish to draw samples by @xmath222 with @xmath223 , where @xmath224 denotes support of the posterior distribution and @xmath225 is the support for @xmath226 .
the abc method aims to draw from this posterior density @xmath227 without the requirement of evaluating the computationally expensive or in our setting intractable likelihood @xmath228 .
the cost of avoiding this calculation is that we obtain an `` approximation ''
. * 1st case .
* we assume that the support @xmath225 is discrete .
given an observation @xmath229 , we would like to sample from @xmath230 .
then the original rejection sampling algorithm reads as follows : * rejection sampling abc * 1 .
sample @xmath231 from prior @xmath232 ; 2 .
simulate synthetic data set of auxiliary variables @xmath233 ; 3 .
abc rejection condition : if @xmath234 then accept sample @xmath231 , else reject sample and return to step 1 .
then the chosen @xmath231 is distributed from @xmath235 .
this follows from a simple rejection argument , denote @xmath236 if @xmath231 was chosen . then , the joint density of @xmath237 conditional on @xmath238 is given by @xmath239 this implies that @xmath240 henceforth , this algorithm generates samples @xmath241 , for @xmath242
. * 2nd case .
* for more general supports @xmath225 one replaces the strict equality @xmath243 with a tolerance @xmath127 and a measure of discrepancy or a distance metric @xmath244 . in this case
the posterior distribution is given by @xmath245 .
implementing this algorithm in a rejection sampling framework gives the following : * rejection sampling abc * 1 .
sample @xmath231 from prior @xmath232 ; 2 .
simulate synthetic data set of auxiliary variables @xmath233 ; 3 .
abc rejection condition 2 : if @xmath246 then accept sample @xmath231 , else reject sample and return to step 1 . in this case the joint density of @xmath237 , conditional on @xmath247 , is given by @xmath248 note that for appropriate choices of the distance metric @xmath117 and assuming the necessary continuity properties for the densities we obtain that @xmath249 this concept was taken further with the intention of improving the simulation efficiency by reducing the number of rejected samples . to achieve this
, sufficient statistics were used to replace the comparison between the auxiliary variables ( `` synthetic data '' ) @xmath250 and the observations @xmath226 . denoting the sufficient statistics by @xmath251 and @xmath252 ,
allows one to decompose the likelihood under the fisher - neyman factorization theorem into @xmath253 for appropriate functions @xmath254 and @xmath255 . in the abc context presented above
, the consequence of this decomposition is that when @xmath256 the obtained samples are from the posterior density @xmath257 similar to .
in general , summary statistics will be used when sufficient statistics are not attainable .
we develop an mcmc - abc algorithm which has an adaptive proposal mechanism and annealing of the tolerance during burn - in of the markov chain .
having reached the final tolerance post annealing , denoted @xmath259 , we utilise the remaining burn - in samples to tune the proposal distribution to ensure an acceptance probability between the range of 0.3 and 0.5 is achieved .
the optimal acceptance probability when posterior parameters are i.i.d .
gaussian was proven to be at 0.234 ; see roberts et al . @xcite .
though our problem does not match the required conditions for this proof , it provides a practical guide . to achieve this
, we tune the coefficient of variation of the proposal , in our case it is the shape parameter of the gamma proposal distribution .
we impose an additional constraint that the minimum shape parameter value is set at @xmath260 for @xmath261 . * mcmc - abc algorithm using bootstrap samples .
* 1 . for @xmath262 initialize the parameter vector randomly , this gives @xmath263 .
initialize the proposal shape parameters @xmath264 for all @xmath155 .
2 . for @xmath265 1 . set @xmath266 .
2 . for @xmath267 1 .
sample proposal @xmath268 from a @xmath269-distribution .
we denote the gamma proposal density by @xmath270 .
this gives proposed parameter vector @xmath271 .
2 . conditional on @xmath272 , generate synthetic bootstrap data set @xmath273 using the bootstrap procedure detailed in section [ section : bootstrap dfcl ] where we replace the cl parameter estimates @xmath274 by the parameters @xmath275 .
3 . evaluate summary statistics @xmath276 and @xmath277 and corresponding decision function @xmath278 as described in section 5 .
4 . accept proposal with abc acceptance probability @xmath279 that is , simulate @xmath280 and set @xmath281 if @xmath282 .
5 . if @xmath283 and @xmath284 then check to see if tuning of the proposal is required .
define the average acceptance probability over the last 100 iterations of updates for parameter @xmath3 by @xmath285 and consider the adaption : @xmath286 then set the proposal shape parameter as @xmath287 .
the mcmc - abc algorithm presented can be enhanced by utilising an idea of gramacy et al .
@xcite in an abc setting .
this involves a combination of tempering the tolerance @xmath288 and importance sampling corrections .
we start with the choices of the abc components .
* generation of a synthetic data set * : [ synthetic data ] note that in this setting not only is the likelihood intractable but also the generation of a synthetic data set @xmath289 given the current parameter values @xmath290 is not straightforward .
the synthetic data set @xmath289 is generated using the bootstrap procedure described in section [ section : bootstrap dfcl ] .
note that both the bootstrap residual @xmath86 and the bootstrap samples @xmath291 are functions of the parameter choices ; see section 4.1 . therefore we generate for given @xmath64 and @xmath65 the bootstrap residuals @xmath292 and the bootstrap samples @xmath293 according to the non - parametric bootstrap ( see section 4.1 ) where we replace the cl parameter estimates @xmath274 by the parameters @xmath294 .
* summary statistics * : we introduce summary statistics to replace sufficient statistics when they are not attainable for a given model . then , in order to define the decision function @xmath255 , we introduce summary statistics ; see appendix [ section : justification for abc ] . for the observed data @xmath33 we define the vector @xmath295 where @xmath296 denotes the number of residuals @xmath86 . for given @xmath294 ,
we generate the bootstrap sample @xmath297 as described above .
the corresponding residuals @xmath298 should also be close to the standardized observations .
therefore , we define its empirical mean and standard deviation by @xmath299^{1/2}.\end{aligned}\ ] ] hence , the summary statistics for the synthetic data is given by @xmath300 * distance metrics * : * _ mahlanobis distance and scaled euclidean distance _ + here
we draw on the analysis of sisson et al .
@xcite that proposes the use of the mahlanobis distance metric given by@xmath301 ^{\top } ~\sigma _ { \mathcal{d}_{i}}^{-1}~ \left [ s\left ( \mathcal{d}_{i};0,1\right ) -s \left ( \mathcal{d}_{i } ^{\ast } ; \mu^\ast , s^\ast\right ) \right ] , \end{aligned}\]]where the covariance matrix @xmath302 is an appropriate scaling described in appendix [ section : mahlanobis distance metric ] .
the scaled euclidean distance is obtained when we only consider the diagonal elements of the covariance matrix @xmath303 .
+ note , the covariance matrix @xmath303 provides a weighting on each element of the vector of summary statistics to ensure they are scaled appropriately according to their influence on the abc approximation .
there are many other such weighting schemes one could conceive .
manhattan `` city block '' distance _
+ we consider the @xmath304-distance given by @xmath305 * decision function * : we work with a hard decision function given by @xmath306 * tolerance schedule * : we use the sequence @xmath307note , the use of an mcmc - abc algorithm can result in `` sticking '' of the chain for extended periods .
therefore , one should carefully monitor convergence diagnostics of the resulting markov chain for a given tolerance schedule .
there is a trade - off between the length of the markov chain required for samples approximately from the stationary distribution and the bias introduced by non zero tolerance . in this paper
we set @xmath308 via preliminary analysis of the markov chain sampler mixing rates for a transition kernel with coefficient of variation set to one .
we note that in general , practitioners will have a required precision in posterior estimates that can be directly used to determine , for a given computational budget , a suitable tolerance @xmath308 .
* convergence diagnostics : * we stress that when using an mcmc - abc algorithm , it is crucial to carefully monitor the convergence diagnostics of the markov chain .
this is more important in the abc context than in the general mcmc context due to the possibility of extended rejections where the markov chain can stick in a given state for long periods .
this can be combatted in several ways which will be discussed once the algorithm is presented .
the convergence diagnostics we consider are evaluated only on samples post annealing of the tolerance threshold and after an initial burn - in period once tolerance of @xmath308 is reached .
if the total chain has length @xmath91 , the initial burn - in stage will correspond to the first @xmath309 samples and we define @xmath310 .
we denote by @xmath311 the markov chain of the @xmath3-th parameter after burn - in .
the diagnostics we consider are given by : _ autocorrelation .
_ this convergence diagnostic will monitor serial correlation in the markov chain .
for given markov chain samples for the @xmath3-th parameter @xmath311 , we define the biased autocorrelation estimate at lag @xmath160 by @xmath312[\theta_i^{(t+\tau)}-\widehat{\mu}\left(\theta_i\right)],\ ] ] where @xmath313 and @xmath314 are the estimated mean and standard deviation of @xmath315 . _ geweke @xcite time series diagnostic . _ for parameter
@xmath315 it is calculated as follows : 1 .
split the markov chain samples into two sequences , @xmath164 and @xmath165 , such that @xmath316 , and with ratios @xmath317 and @xmath318 fixed such that @xmath319 for all @xmath162 .
2 . evaluate @xmath320 and @xmath321 corresponding to the sample means on each sub sequence .
3 . evaluate consistent spectral density estimates for each sub sequence , at frequency 0 , denoted @xmath322 and @xmath323 .
the spectral density estimator considered in this paper is the classical non - parametric periodogram or power spectral density estimator .
we use welch s method with a hanning window ; for details see appendix [ section : power spectral density estimate ] .
4 . evaluate convergence diagnostic given by @xmath324 + according to the central limit theorem , as @xmath167 one has that @xmath325 if the sequence @xmath311 is stationary .
_ gelman - rubin @xcite r - statistic diagnostic .
_ this approach to convergence analysis requires that one runs multiple parallel independent markov chains each starting at randomly selected initial starting points ( we run five chains ) . for comparison purposes we split the total computational budget of @xmath162 into @xmath326 .
the convergence diagnostic for parameter @xmath315 is calculated using the following steps : 1 .
generate five independent markov chain sequences , producing the chains for parameter @xmath315 denoted @xmath327 for @xmath328 .
2 . calculate the sample means @xmath329 for each sequence and the overall mean @xmath330 .
3 . calculate the variance of the sequence means @xmath331 4 .
calculate the within - sequence variances @xmath332 for each sequence .
calculate the average within - sequence variance , @xmath333 .
6 . estimate the target posterior variance for parameter @xmath315 by the weighted linear combination @xmath334 .
this estimate is unbiased for samples which are from the stationary distribution . in the case
in which not all sub chains have reached stationarity , this overestimates the posterior variance for a finite @xmath162 but asymptotically , @xmath335 , it converges to the posterior variance .
improve on the gaussian estimate of the target posterior given by + @xmath336 by accounting for sampling variability in the estimates of the posterior mean and variance .
this can be achieved by making a student - t approximation with location @xmath330 , scale @xmath337 and degrees of freedom @xmath338 , each given respectively by : @xmath339 and @xmath340 , where the variance is estimated as + @xmath341 note , the covariance terms are estimated empirically using the within sequence estimates of the mean and variance obtained for each sequence .
calculate the convergence diagnostic @xmath342 , where as @xmath167 one can prove that @xmath343 .
this convergence diagnostic monitors the scale factor by which the current distribution for @xmath315 may be reduced if simulations are continued for @xmath344 .
in the mahlanobis distance metric , estimation of the scaling weights is given by the covariance @xmath345 , where @xmath346 and @xmath347 are the sample mean and standard deviation of @xmath296 i.i.d .
residuals @xmath348 ( see also - ) .
next we outline the estimation of @xmath349 by a matrix @xmath350 . * starting with the elements @xmath351 with @xmath352 , we obtain from the conditional resampling bootstrap * * @xmath353 if @xmath354 or @xmath355 * * @xmath356 * considering the elements @xmath357 and also @xmath358 of the covariance matrix @xmath359 , for simplicity we set @xmath360 . * considering elements @xmath361 , we assess now @xmath362 either analytically or numerically by simulation of appropriate i.i.d . residuals .
+ * parametric approximation * * * in approximating @xmath363 and @xmath364 we assume i.i.d . samples @xmath365 . * * using the assumptions we know that : + @xmath366 , + @xmath367 = \frac{1}{(n-1)^2}[2n(1+\frac{5}{n^2})]$ ] , + @xmath368 $ ] . * * under these assumptions : + 1 .
if the distribution of @xmath348 is skewed then it is more appropriate to do a numerical approximation with the observed residuals from the bootstrap algorithm .
the precision @xmath369 from the mcmc - abc algorithm should depend on the size of the claims triangle , that is , the number of residuals @xmath296 .
this is calculated via a modified technique using welch s method ; see proakis - manolakis @xcite , 910 - 913 .
this involves performing the following steps : * split each sequence @xmath164 and @xmath165 into @xmath370 non - overlapping blocks of length @xmath371 . *
apply a hanning window function @xmath372 to the samples of the markov chain in each block .
* take the discrete fourier transform ( dft ) of each windowed block given by @xmath373 .
* estimate the spectral density ( sd ) as @xmath374 . | the intention of this paper is to estimate a bayesian distribution - free chain ladder ( dfcl ) model using approximate bayesian computation ( abc ) methodology .
we demonstrate how to estimate quantities of interest in claims reserving and compare the estimates to those obtained from classical and credibility approaches . in this context , a novel numerical procedure utilising markov chain monte carlo ( mcmc ) , abc and a bayesian bootstrap procedure was developed in a truly distribution - free setting .
the abc methodology arises because we work in a distribution - free setting in which we make no parametric assumptions , meaning we can not evaluate the likelihood point - wise or in this case simulate directly from the likelihood model .
the use of a bootstrap procedure allows us to generate samples from the intractable likelihood without the requirement of distributional assumptions , this is crucial to the abc framework .
the developed methodology is used to obtain the empirical distribution of the dfcl model parameters and the predictive distribution of the outstanding loss liabilities conditional on the observed claims
. we then estimate predictive bayesian capital estimates , the value at risk ( var ) and the mean square error of prediction ( msep ) .
the latter is compared with the classical bootstrap and credibility methods .
claims reserving , distribution - free chain ladder , mean square error of prediction , bayesian chain ladder , approximate bayesian computation , markov chain monte carlo , annealing , bootstrap |
a _ trigraph _ @xmath1 consists of a set @xmath2 of vertices , and two disjoint sets of directed edges on @xmath3 the set of _ weak edges _
@xmath4 , and the set of _ strong edges _ @xmath5 . if both edge sets @xmath6 , viewed as relations on @xmath3 , are symmetric , we have a _ symmetric _ , or _
undirected _ trigraph .
a weak , respectively strong , edge @xmath7 is called a _ weak _ , respectively _ strong , loop _ at @xmath8 .
the _ adjacency matrix _ of a trigraph @xmath1 , with respect to an enumeration @xmath9 of its vertices , is the @xmath10 matrix @xmath11 over @xmath12 , in which @xmath13 if @xmath14 is not an edge , @xmath15 if @xmath14 is a weak edge , and @xmath16 if @xmath14 is a strong edge .
note that a trigraph @xmath1 is symmetric if and only if its adjacency matrix is symmetric .
we consider the class of digraphs included in the class of trigraphs , by viewing each digraph @xmath1 as a trigraph with the same vertex set @xmath17 , and with the weak edge set @xmath18 and strong edge set @xmath19 .
conversely , if @xmath1 is a trigraph , the _ associated digraph _ of @xmath1 is the digraph with the same vertex set @xmath17 , and with the edge set @xmath20 .
moreover , the _ underlying graph _ of the trigraph @xmath1 is the underlying graph of the associated digraph , and the _ symmetric graph _ of the trigraph @xmath1 is the symmetric graph of the associated digraph . to be specific
, @xmath21 is an edge of the _ underlying graph _ of @xmath1 just if @xmath22 or @xmath23 , and @xmath21 is an edge of the _ symmetric graph _ of @xmath1 just if @xmath22 and @xmath23 .
these conventions allow us to extend the usual graph and digraph terminology to trigraphs .
we speak , for instance , of _ adjacent vertices , components , neighbours , cutpoints , or bridges _ of a trigraph @xmath1 , meaning the corresponding notions in the associated digraph of @xmath1 , or in its underlying graph ; and we speak of _ symmetric edges , symmetric neighbours , etc .
_ in a trigraph @xmath1 , meaning the edges , neighbours , etc . , in the symmetric graph of @xmath1
let @xmath24 be a digraph and @xmath1 a trigraph .
a _ homomorphism _ of @xmath24 to @xmath1 is a mapping @xmath25 such that the following two conditions are satisfied for each @xmath26 : 1 . if @xmath27 then @xmath28 2 .
if @xmath29 then @xmath30 . in other words , edges of @xmath24 must map to either weak or strong edges of @xmath1 , and non - edges of @xmath24 must map to either non - edges or weak edges of @xmath1 . if each vertex @xmath8 of the digraph
@xmath24 has a _ list _ @xmath31 , then a _ list homomorphism _ of @xmath24 to @xmath1 , with respect to the lists @xmath32 , is a homomorphism @xmath33 of @xmath24 to @xmath1 such that @xmath34 for all @xmath35 .
following standard practice @xcite , we also call a homomorphism of @xmath24 to @xmath1 an _ @xmath1-colouring of @xmath24 _ , and a list homomorphism of @xmath24 to @xmath1 ( with respect to the lists @xmath32 ) a _ list @xmath1-colouring _ of @xmath24 ( with respect to @xmath32 ) .
suppose @xmath1 is a fixed trigraph .
the _ @xmath1-colouring problem _ @xmath36 has as instances digraphs @xmath24 , and asks whether or not @xmath24 admits an @xmath1-colouring .
the _ list @xmath1-colouring problem _ @xmath36 has as instances digraphs @xmath24 with lists @xmath32 , and asks whether or not @xmath24 admits a list @xmath1-colouring with respect to @xmath32 . as noted earlier , @xmath1 could be a digraph , viewed as a trigraph ( with @xmath18 , @xmath37 ) .
digraph homomorphism and list homomorphism problems have been of much interest @xcite . if necessary , we will emphasize the distinction between _ trigraph _ list homomorphism problems and _ digraph _ list homomorphism problems , depending on whether @xmath1 is a trigraph or a digraph respectively , i.e. , whether @xmath1 has any strong edges or not .
however , note that the input @xmath24 is always a digraph . for a fixed trigraph @xmath1 , the list @xmath1-colouring problem l - hom@xmath36 concerns the existence of vertex partitions of the input digraphs @xmath24 . for instance , if @xmath1 is the undirected trigraph with @xmath38 , with a strong loop at @xmath39 and a weak edge joining @xmath40 and @xmath39 , then an @xmath1-colouring of @xmath24 is precisely a partition of @xmath41 into a clique and an independent set .
thus @xmath24 is @xmath1-colourable if and only if @xmath24 is a split graph .
many graph partition problems , especially those arising from the theory of perfect graphs , can be formulated as trigraph homomorphism ( or list homomorphism ) problems ; this is discussed in detail in @xcite .
equivalently , all these problems can be described in terms of the adjacency matrix of the trigraph , in the language of matrix partitions and list partitions ( see @xcite ) . in this paper
it will be more convenient to emphasize the trigraph ( rather than the matrix ) terminology , since we are dealing with the structure of the trigraph @xmath1 .
it is generally believed @xcite that for each digraph @xmath1 the @xmath1-colouring problem hom@xmath36 is @xmath0-complete or polynomial time solvable .
( this is equivalent to the so - called _ csp dichotomy conjecture _ of feder and vardi @xcite . )
one special case when the dichotomy conjecture is known to hold is the case of undirected graphs ( i.e. , symmetric digraphs ) . in this case , hom@xmath36 is polynomial time solvable if @xmath1 has a loop or is bipartite and is @xmath0-complete otherwise @xcite . for the list homomorphism problem ,
it is shown in @xcite that l - hom@xmath36 is polynomial time solvable if @xmath1 is a so - called _ bi - arc graph _ ( a simultaneous generalization of reflexive interval graphs and bipartite graphs whose complements are circular arc graphs ) , and is @xmath0-complete otherwise . a more general result of bulatov @xcite handles all constraint satisfaction problems , implying , in particular , that for each digraph @xmath1 , the list @xmath1-colouring problem l - hom@xmath36 is @xmath0-complete or polynomial time solvable .
by contrast , this is not known for trigraph list homomorphism problems , and in @xcite it is only proved that for each trigraph @xmath1 , the list @xmath1-colouring problem is @xmath0-complete or _ quasi - polynomial _ ( of complexity @xmath42 ) .
all list @xmath1-colouring problems l - hom@xmath36 for trigraphs @xmath1 with three or fewer vertices have been classified as @xmath0-complete or polynomial time solvable in @xcite . for symmetric trigraphs with four vertices , this has been accomplished in @xcite , with the exception of a single trigraph @xmath1 ; the corresponding problem remains open and has earned the name _ the stubborn problem _ ( figure [ fig : ex1]a ) .
the best known algorithm for this problem has complexity @xmath43 , implying it is unlikely to be @xmath0-complete @xcite .
thus ( polynomial / @xmath0-complete ) dichotomy for trigraph list homomorphism problems seems less likely than for _ digraph _ list homomorphism problems . in this paper
we prove dichotomy for the class of _ trigraph trees _
, i.e. , for trigraphs whose underlying graph is a tree .
it turns out that if @xmath1 is a trigraph tree , then the list @xmath1-colouring problem is polynomially equivalent to a list @xmath44-colouring problem where @xmath44 is a _ digraph _
obtained from @xmath1 by removing all vertices with a strong loop and removing all other strong edges @xmath21 ( and their converses @xmath45 if any ) .
we conduct the proof of this result in such a way that it in fact implies the dichotomy for a large class of trigraphs , which includes all digraphs and all trigraph trees .
we think of these trigraphs as _ tree - like _ , although it is only the structure of the strong edges that is tree - like . for trigraphs that are not tree - like ( in our definition )
, we illustrate the possible complications .
we believe that our class of tree - like trigraphs covers an important portion of the class of trigraphs @xmath1 in which the strong edges do not significantly impact the complexity of the list @xmath1-colouring problem .
let @xmath1 be a trigraph , and let @xmath24 be a digraph with list @xmath46 for each @xmath47 .
we shall denote by @xmath48 the number of vertices in @xmath24 , and by @xmath49 the number of vertices in @xmath1 .
we say that lists @xmath50 are _ a reduction of @xmath32 _ , if we have @xmath51 for each @xmath47 . in the following text
, we shall say that lists @xmath32 can be _ reduced _ to satisfy property @xmath52 to mean that for every instance consisting of a graph @xmath24 with lists @xmath32 , there exist lists @xmath50 on @xmath24 such that a. the lists @xmath50 can be found in polynomial time b. the lists @xmath50 are a reduction of @xmath32 c. the lists @xmath50 satisfy property @xmath52 , and d. @xmath24 has a list @xmath1-colouring with respect to @xmath32 if and only if @xmath24 has a list @xmath1-colouring with respect to @xmath50 .
we shall say that lists @xmath32 can be _ transformed _ to satisfy property @xmath52 to mean that for every instance @xmath24 with lists @xmath32 , there exists a set @xmath53 of lists such that a. the set @xmath54 can be constructed in polynomial time b. each @xmath55 is a reduction of @xmath32 c. each @xmath55 satisfies property @xmath52 , and d. @xmath24 has a list @xmath1-colouring with respect to @xmath32 if and only if there exists an @xmath56 such that @xmath24 has a list @xmath1-colouring with respect to @xmath55 .
note that while reducing the lists results in a single problem , transforming the lists produces a family of problems , as illustrated in figure [ fig:1 ] .
@xmath57[f]{\phantom{s}}="a1 " ; ( -1,-0.5)*[o][f]{\phantom{s}}="a2 " ; { \ar " a1";"a2 " } ; " a1"+(0,0.4)*{l } ; " a2"+(0.1,-0.4)*{l ' } ; ( 0,0)*[r]{\mbox{\em `` reduced '' } } ; ( -1,-2.5)*[o][f]{\phantom{s}}="a1 " ; ( -2,-3.5)*[o][f]{\phantom{s}}="a2 " ; ( -1.5,-3.5)*[o][f]{\phantom{s}}="a3 " ; ( -1,-3.5)*[o][f]{\phantom{s}}="a4 " ; ( -0.5,-3.5)*[o]{\ldots}="a5 " ; ( 0,-3.5)*[o][f]{\phantom{s}}="a6 " ; { \ar " a1";"a2 " } ; { \ar " a1";"a3 " } ; { \ar " a1";"a4 " } ; { \ar " a1";"a6 " } ; " a1"+(0,0.4)*{l } ; " a2"+(0.1,-0.3)*{_{l_1 } } ; " a3"+(0.1,-0.3)*{_{l_2 } } ; " a4"+(0.1,-0.3)*{_{l_3 } } ; " a6"+(0.25,-0.35)*{_{l_{p(n ) } } } ; ( 0,-2.7)*[r]{\mbox{\em `` transformed '' } } ; ( 4.3,0.5)*[o][f]{\phantom{s}}="a1 " ; ( 4.3,-0.5)*[o][f]{\phantom{s}}="a2 " ; ( 4.3,-1.5)*[o][f]{\phantom{s}}="a3 " ; ( 4.3,-1.9)*[o]{\vdots } ; ( 4.3,-2.5)*[o][f]{\phantom{s}}="a4 " ; ( 4.3,-3.5)*[o][f]{\phantom{s}}="a5 " ; { \ar " a1";"a2 " } ; { \ar " a2";"a3 " } ; { \ar " a4";"a5 " } ; " a1"+(0,0.4)*{l } ; " a5"+(0.1,-0.4)*{l^ * } ; ( 4.8,0)*[r]{arc } ; ( 4.8,-1)*[r]{sep } ; ( 4.8,-2)*[r]{arc } ; ( 4.3,-4.5)*{arc+sep } ; ( 9,1)*[o][f]{\phantom{s}}="a0 " ; ( 9,0)*[o][f]{\phantom{s}}="a1 " ; ( 8,-1)*[o][f]{\phantom{s}}="a21 " ; ( 8.5,-1)*[o][f]{\phantom{s}}="a22 " ; ( 9,-1)*[o][f]{\phantom{s}}="a23 " ; ( 9.5,-1)*[o]{\ldots}="a24 " ; ( 10,-1)*[o][f]{\phantom{s}}="a25 " ; { \ar " a0";"a1 " } ; { \ar " a1";"a21 " } ; { \ar " a1";"a22 " } ; { \ar " a1";"a23 " } ; { \ar " a1";"a25 " } ; ( 8,-2)*[o][f]{\phantom{s}}="a31 " ; ( 8.5,-2)*[o][f]{\phantom{s}}="a32 " ; ( 9,-2)*[o][f]{\phantom{s}}="a33 " ; ( 9.5,-2)*[o]{\ldots}="a34 " ; ( 10,-2)*[o][f]{\phantom{s}}="a35 " ; { \ar " a21";"a31 " } ; { \ar " a22";"a32 " } ; { \ar " a23";"a33 " } ; { \ar " a25";"a35 " } ; ( 8,-2.4)*[o]{\vdots } ; ( 8.5,-2.4)*[o]{\vdots } ; ( 9,-2.4)*[o]{\vdots } ; ( 10,-2.4)*[o]{\vdots } ; ( 8,-3)*[o][f]{\phantom{s}}="a41 " ; ( 8.5,-3)*[o][f]{\phantom{s}}="a42 " ; ( 9,-3)*[o][f]{\phantom{s}}="a43 " ; ( 9.5,-3)*[o]{\ldots}="a44 " ; ( 10,-3)*[o][f]{\phantom{s}}="a45 " ; ( 8,-4)*[o][f]{\phantom{s}}="a51 " ; ( 8.5,-4)*[o][f]{\phantom{s}}="a52 " ; ( 9,-4)*[o][f]{\phantom{s}}="a53 " ; ( 9.5,-4)*[o]{\ldots}="a54 " ; ( 10,-4)*[o][f]{\phantom{s}}="a55 " ; { \ar " a41";"a51 " } ; { \ar " a42";"a52 " } ; { \ar " a43";"a53 " } ; { \ar " a45";"a55 " } ; " a0"+(0,0.4)*{l } ; " a5"+(0.1,-0.4)*{l^ * } ; ( 10,0.8)*[r]{weak } ; ( 9.5,0.4)*[r]{domination } ; ( 10.5,-0.5)*[r]{s\mbox{-}d } ; ( 10.5,-1.5)*[r]{arc } ; ( 10.5,-2.5)*[r]{sep } ; " a51"+(0.1,-0.35)*{_{l^*_1 } } ; " a52"+(0.1,-0.35)*{_{l^*_2 } } ; " a53"+(0.1,-0.35)*{_{l^*_3 } } ; " a55"+(0.25,-0.4)*{_{l^*_{p(n ) } } } ; ( 9,-5)*{arc+sep+s\mbox{-}d } ; \endxy $ ] we say that lists @xmath32 are _ non - empty _ if the list @xmath58 for each vertex @xmath47 is not empty . clearly , if @xmath24 admits a list homomorphism to @xmath1 with respect to @xmath32 , then the lists @xmath32 are non - empty . hence ,
to avoid trivial cases in what follows , we shall always assume that lists @xmath32 are non - empty .
we say that lists @xmath32 are _ arc - consistent _ if for each @xmath59 and each @xmath60 , there exists @xmath61 such that ( i ) @xmath62 if @xmath63 , ( ii ) @xmath64 if @xmath65 , ( iii ) @xmath66 if @xmath67 , and ( iv ) @xmath68 if @xmath69 . [ lem:1 ] lists @xmath32 can be reduced to be arc - consistent .
if some @xmath59 violate the above condition for @xmath60 , no @xmath1-colouring of @xmath24 with respect to @xmath32 maps @xmath70 to @xmath71 . hence , @xmath71 can be removed from @xmath58 without changing the existence of solutions .
we repeatedly test for such violations and reduce the lists if a violation is found .
after at most @xmath72 such steps either we obtain arc - consistent lists .
( note that if one final list is empty then all are empty , by arc - consistency . )
we say that lists @xmath32 contain _ representatives _ , if there is a set @xmath73 such that a. for each @xmath74 , the list @xmath75 , and b. for each @xmath76 , there is a vertex @xmath74 with @xmath77 , [ lem:2 ] lists @xmath32 can be transformed to contain representatives .
we construct the set @xmath78 as follows .
the lists @xmath55 are obtained in performing in every possible way the following process . for each vertex @xmath79 , we either select a vertex of @xmath24 and set its list to @xmath80 , or we remove @xmath71 from all lists .
the resulting lists @xmath55 contain representatives for the set @xmath81 of vertices @xmath71 for which some list in @xmath24 was set to @xmath80 .
moreover , the vertices of @xmath82 do not appear in @xmath55 .
the _ symmetric trigraph @xmath83 associated with @xmath1 _ is the trigraph on the vertices of @xmath1 with strong edges @xmath21 iff @xmath84 , and with weak edges @xmath21 iff @xmath85 or @xmath68 but @xmath86 .
a graph @xmath24 is _ chordal _ if @xmath24 contains no induced cycle of length four or longer .
chordal completion _
@xmath87 of @xmath24 is a chordal graph on the vertices of @xmath24 with @xmath88 .
a chordal completion @xmath87 is _ minimal _ if no chordal completion @xmath89 of @xmath24 satisfies @xmath90 .
@xcite[prop : tarjan ] let @xmath87 be a minimal chordal completion of @xmath24 , and let @xmath91 be a clique of @xmath24 .
then there are no edges in @xmath87 between vertices of different components of @xmath92 .
we say that lists @xmath32 contain _ strong representatives _ , if for each strong loop @xmath71 of @xmath1 , there is a set @xmath93 such that a. @xmath94 whenever @xmath95 , b. each vertex of @xmath96 is a symmetric neighbour of @xmath71 , and c. @xmath97 is connected in the symmetric graph of @xmath1 .
[ lem:3 ] lists @xmath32 can be transformed to contain strong representatives .
let @xmath24 be a digraph with lists @xmath32 , and let @xmath33 be a list @xmath1-colouring of @xmath24 with respect to @xmath32 .
by lemmata [ lem:1 ] and [ lem:2 ] , we may assume that the lists @xmath32 are arc - consistent and contain representatives .
let @xmath71 be a strong loop of @xmath1 . since @xmath33 is a homomorphism , if @xmath98 for @xmath59 , we must have both @xmath99 and @xmath100 in @xmath101 . hence , the set @xmath102 induces a symmetric clique in @xmath24 .
let @xmath87 be the symmetric graph of @xmath24 , and @xmath83 be the symmetric trigraph associated with @xmath1 .
clearly , @xmath91 also induces a clique in @xmath87 .
moreover , it is easy to show that @xmath33 is also a list @xmath83-colouring of @xmath87 with respect to @xmath32 . now , suppose that @xmath71 appears on some list .
then , since the lists @xmath32 contain representatives , there is a vertex @xmath103 with @xmath104 .
let @xmath105 denote the subset of @xmath17 containing @xmath71 and its symmetric neighbours .
let @xmath106 denote the subset of @xmath41 containing @xmath107 and its symmetric neighbours .
let @xmath108 denote the subset of @xmath41 containing all vertices @xmath109 with @xmath60 .
since the lists @xmath32 are arc - consistent , we must have @xmath110 .
also , for each @xmath111 , we have @xmath112 .
now , since @xmath33 respects the lists @xmath32 , we have @xmath60 , for each @xmath113 .
hence , @xmath114 , and therefore , @xmath91 induces a clique in @xmath115 $ ] .
furthermore , since @xmath110 , we have @xmath112 for each @xmath116 .
hence , @xmath33 restricted to @xmath108 is a homomorphism of @xmath115 $ ] to @xmath117 $ ] .
in particular , each component of @xmath115-c$ ] maps by @xmath33 to a unique component of @xmath117-x$ ] .
now , let @xmath89 be a minimal chordal completion of @xmath115 $ ] .
let @xmath118 be the vertices of a maximal clique of @xmath89 that completely contains @xmath91 . by proposition [ prop : tarjan ] , all vertices of @xmath119 belong to one component of @xmath115-c$ ] .
hence , by the above remark , there is a unique component @xmath120 of @xmath117-x$ ] such that @xmath121 for each @xmath122 .
it now follows that @xmath24 admits a list @xmath1-colouring with respect to @xmath32 if and only if for some maximal clique @xmath118 of a minimal chordal completion @xmath89 of @xmath115 $ ] , and some component @xmath120 of @xmath117-x$ ] , the graph @xmath24 admits a list @xmath1-colouring with respect to @xmath32 such that the vertices outside of @xmath118 do not map to @xmath71 , and the vertices in @xmath118 map to @xmath71 or the vertices of @xmath120 .
hence , we can modify the lists @xmath32 by removing @xmath71 from the vertices outside @xmath118 and by reducing the lists of the vertices in @xmath118 to @xmath123 .
such lists , clearly , contain a strong representative for @xmath71 .
( take @xmath124 to be @xmath123 . )
to conclude , we remark that the graph @xmath87 and the sets @xmath108 and @xmath105 , as well as , a minimal chordal completion @xmath89 of @xmath115 $ ] by the result @xcite can be found in polynomial time . since @xmath89 is chordal , it has at most @xmath48 maximal cliques @xcite
. also , there are at most @xmath49 different components @xmath120 of @xmath117-x$ ] .
hence , we can reduce the problem to at most @xmath125 different instances with strong representatives .
the proof is now complete .
note that if lists @xmath32 contain representatives , respectively strong representatives , then any non - empty reduction @xmath50 of @xmath32 also contains representatives , respectively strong representatives .
let @xmath126 be a set of edges of @xmath1 .
let @xmath127 denote the graph obtained from @xmath24 by removing all edges @xmath63 such that @xmath128 for some @xmath129 and @xmath61 .
we say that lists @xmath32 are _ separator - consistent on @xmath126 _ , if for each component @xmath91 of @xmath127 and each component @xmath120 of @xmath130 , a. either there exists a list @xmath120-colouring of @xmath91 with respect to @xmath32 , b. or no vertex of @xmath120 appears on the list of any vertex of @xmath91 . [ lem:4 ] if l - hom@xmath131 is polynomial time solvable , then the lists @xmath32 can be reduced to be separator - consistent on @xmath126 .
we obtain a separator - consistent lists @xmath50 from @xmath32 as follows . for each component @xmath91 of @xmath127 , and each component @xmath120 of @xmath130 , we test if @xmath91 admits a list @xmath120-colouring with respect to @xmath32 . if not , then we remove all vertices of @xmath120 from the lists of the vertices of @xmath91 . since homomorphisms map connected graphs only to connected graphs , the claim follows . in particular , we have the following property .
[ lem:5 ] if l - hom@xmath131 is polynomial time solvable , then the lists @xmath32 can be reduced to be arc - consistent and separator - consistent on @xmath126 .
we apply lemmata [ lem:1 ] and [ lem:4 ] to @xmath32 until the lists no longer change .
since at each step the lists are reduced , the claim follows .
we remark that for @xmath73 , we say that the lists @xmath32 are _ separator - consistent on @xmath81 _ , if they are separator - consistent on @xmath126 , where @xmath126 is the set of edges of @xmath1 with at least one endpoint in @xmath81 .
let @xmath81 and @xmath132 be two sets of vertices of @xmath1 such that each vertex of @xmath81 has a strong loop , and no vertex of @xmath132 has a loop .
we say that lists @xmath32 are _ sparse - dense - consistent on @xmath81 and @xmath132 _ if for each @xmath74 such that @xmath133 , we have @xmath75 or @xmath134 . [ lem:6 ] if l - hom@xmath135)$ ] and l - hom@xmath136)$ ] are polynomial time solvable , then the lists @xmath32 can be transformed to be sparse - dense - consistent on @xmath81 and @xmath132 .
let @xmath137 and @xmath138 be classes of digraphs closed under taking induced subgraphs .
suppose that there is a constant @xmath139 such that each digraph in @xmath140 has at most @xmath141 vertices .
then by @xcite , for any @xmath48-vertex digraph @xmath24 , there are at most @xmath142 partitions @xmath143 such that @xmath144\in\mathcal s$ ] and @xmath145\in\mathcal d$ ] .
( we call each such partition an
_ @xmath146-partition _ of @xmath24 . )
all these partitions can be enumerated in time @xmath147 where @xmath148 is the complexity of recognizing digraphs in @xmath137 and in @xmath138 .
now , let @xmath137 be the set of all digraphs admitting an @xmath149$]-colouring and @xmath138 be the set of all digraphs admitting an @xmath150$]-colouring ; we have @xmath151 .
let @xmath152 denote the vertices @xmath70 of @xmath24 with @xmath153 .
we observe that any list @xmath1-colouring of @xmath24 with respect to @xmath32 induces an @xmath154-partition of @xmath155 $ ] .
hence , for each @xmath154-partition @xmath156 of @xmath155 $ ] , we construct lists @xmath50 on @xmath1 as follows : @xmath157 if @xmath158 , @xmath159 if @xmath116 , and @xmath160 otherwise . clearly
, @xmath50 is sparse - dense consistent on @xmath81 and @xmath132 .
now , since there are at most @xmath161 such partitions , the claim follows .
let @xmath162 be a set of vertices of @xmath1 , and let @xmath163 be two vertices of @xmath1 .
we say that @xmath164 _ weakly dominates @xmath71 on @xmath162 _ , if for each @xmath165 , we have that @xmath166 is a weak edge in @xmath1 whenever @xmath167 is a weak edge in @xmath1 , and @xmath168 is a weak edge in @xmath1 whenever @xmath169 is a weak edge in @xmath1 .
we say that _
@xmath164 weakly dominates @xmath71 _ if @xmath164 weakly dominates @xmath71 on @xmath17 .
[ lem:7 ] if @xmath164 weakly dominates @xmath71 , then the lists @xmath32 can be reduced to satisfy @xmath170 whenever @xmath171 . by lemmata [ lem:1 ] and [ lem:2 ] , we may assume that the lists @xmath32 are arc - consistent and contain representatives .
let @xmath33 be a list @xmath1-colouring of @xmath24 with respect to @xmath32 .
let @xmath172 be a mapping such that @xmath173 if @xmath174 and @xmath171 , and @xmath175 otherwise .
we show that @xmath172 is also a homomorphism and the claim will follow .
let @xmath176 be any vertex ( including @xmath71 and @xmath164 ) of @xmath1 .
suppose that @xmath177 , @xmath178 is not an edge , and @xmath176 appears on some list .
we claim that no list @xmath58 contains both @xmath71 and @xmath164 . to prove this ,
let @xmath179 be a vertex with @xmath180 .
suppose that @xmath181 .
if @xmath182 , then by arc - consistency of @xmath32 , we obtain @xmath183 .
if @xmath184 , similarly @xmath185 , a contradiction . moreover , by symmetry , if @xmath186 and @xmath187 is not an edge , or if @xmath71 and @xmath164 exchange places , we also have that no list contains both @xmath71 and @xmath164 . in addition , since @xmath164 weakly dominates @xmath71 , we obtain that either both @xmath71 and @xmath164 have strong loops , or both have no loops , or @xmath164 has a weak loop . now , it is not difficult to directly verify that @xmath172 is a homomorphism using the above observations and the fact that @xmath164 weakly dominates @xmath71 .
hence , @xmath24 has a list @xmath1-colouring with respect to @xmath32 if and only if @xmath24 has a list @xmath1-colouring with respect to the lists @xmath50 obtained from @xmath32 by removing @xmath71 from each list @xmath188 that also contains @xmath164 . that concludes the proof .
let @xmath81 , @xmath132 , and @xmath152 be three sets of vertices of @xmath1 such that each vertex of @xmath81 has a strong loop , no vertex of @xmath132 has a loop , and each vertex of @xmath152 has a weak loop .
suppose that each vertex of @xmath152 weakly dominates each vertex of @xmath132 , and that we have @xmath189 whenever @xmath190 .
[ lem:8 ] let @xmath191 be as above . if l - hom@xmath135)$ ] and are polynomial time solvable , then the lists @xmath32 can be transformed to be arc - consistent , separator - consistent on @xmath81 , and sparse - dense - consistent on @xmath81 and @xmath132 . we observe that by lemma [ lem:7 ] , we may assume that no list @xmath58 contains both a vertex of @xmath132 and a vertex of @xmath152 .
hence , either @xmath192 , or @xmath153 , or @xmath193 for each @xmath47 .
now , let @xmath55 be one of the lists we obtain from @xmath32 by applying sparse - dense consistency on @xmath81 and @xmath132 ( lemma [ lem:6 ] ) .
then we must have either @xmath194 , or @xmath195 , or @xmath196 , or @xmath197 for each @xmath47 .
in particular , any reduction of @xmath55 must also satisfy this condition .
hence , we can apply lemma [ lem:5 ] to @xmath55 , and the claim follows .
let @xmath198 be a partition of the vertices of @xmath1 .
let @xmath126 be the edges of @xmath1 that have exactly one endpoint in @xmath199 .
let @xmath81 respectively @xmath132 be the vertices of @xmath199 respectively @xmath108 with at least one incident edge in @xmath126 .
we say that @xmath24 is _ separable _ on @xmath126 , if for each vertex @xmath74 , a. @xmath188 contains at most one vertex of @xmath81 and at most one vertex of @xmath132 , b. if @xmath188 contains a vertex of @xmath81 , it contains no vertex of @xmath200 , and c. if @xmath188 contains a vertex of @xmath132 , it contains no vertex of @xmath201 .
[ lem:9 ] if l - hom(@xmath202 ) is polynomial time solvable , then l - hom(@xmath1 ) is polynomial time solvable on the class of all digraphs @xmath24 separable on @xmath126 .
first , we observe if @xmath24 with lists @xmath32 is separable on @xmath126 , then for any reduction @xmath50 of @xmath32 , the graph @xmath24 with lists @xmath50 is separable on @xmath126 .
hence , by lemmata [ lem:1 ] and [ lem:2 ] , we may assume that the lists @xmath32 are arc - consistent and contain representatives .
next , we show that l - hom@xmath203)$ ] can be solved in polynomial time on any @xmath127 given that @xmath24 is separable on @xmath126 .
indeed , consider a component @xmath91 of @xmath127 , and let @xmath50 be lists such that @xmath204 for each @xmath205 .
since @xmath24 is separable on @xmath126 , we have either @xmath206 , or @xmath207 . hence
, if @xmath108 denotes the vertices with @xmath207 , then arc - consistency of @xmath32 implies that @xmath91 has a list @xmath208$]-colouring with respect to @xmath32 if and only if @xmath209 has a list @xmath210-colouring with respect to @xmath50 .
similarly , for l - hom@xmath211)$ ] .
hence , l - hom@xmath131 can be solved in polynomial time for @xmath127 given @xmath24 is separable on @xmath126 .
it now follows by lemma [ lem:5 ] that we may assume that the lists @xmath32 are arc - consistent and separator - consistent on @xmath126 .
we also assume that lists @xmath32 are non - empty , since otherwise there is no solution .
let @xmath212 be the trigraph constructed from @xmath213 $ ] by adding vertices @xmath214 and @xmath215 with weak loops such that @xmath214 has a weak symmetric edge to each vertex of @xmath81 , and @xmath215 has a weak symmetric edge to each vertex of @xmath132 .
let @xmath216 be the lists obtained from @xmath32 by replacing by @xmath214 each @xmath217 such that @xmath218 , and replacing by @xmath215 each @xmath217 such that @xmath219 .
also , let @xmath220 and @xmath221 .
let @xmath222 be a list @xmath212-colouring of @xmath24 with respect to @xmath216 .
let @xmath33 be a list @xmath131-colouring of @xmath127 with respect to @xmath32 such that for each @xmath47 , we have @xmath223 if @xmath224 , and @xmath225 if @xmath226 .
since the lists @xmath32 are separator - consistent on @xmath126 , such a colouring can be found .
we make some observations about @xmath33 .
first , note that @xmath174 if @xmath227 where @xmath76 .
indeed , this follows since @xmath24 is separable on @xmath126 , and @xmath222 respects the lists @xmath216 .
similarly , @xmath228 if @xmath229 . moreover , @xmath230 if @xmath231 , and @xmath232 if @xmath233 .
it now follows that @xmath33 is a homomorphism of @xmath24 to @xmath1 with respect to @xmath32 . finally , we observe that the lists @xmath216 are all of size at most two .
hence , the mapping @xmath222 can be found in polynomial time by the standard reduction to @xmath234 . that concludes the proof .
in this section , we prove the dichotomy for l - hom@xmath36 for trigraph trees @xmath1 , i.e. , for trigraphs @xmath1 whose underlying graph is a tree .
partial results along these lines are included in @xcite ; specifically , the case when the underlying graph of @xmath1 is a path is solved there .
let @xmath44 be the digraph obtained from a trigraph @xmath1 by removing all edges @xmath21 such that at least one of @xmath235 is a strong edge of @xmath1 , and by removing all vertices @xmath71 such that @xmath236 is a strong loop in @xmath1 .
[ thm : trees ] if @xmath1 is a trigraph tree , then l - hom@xmath36 is polynomially equivalent to l - hom@xmath237 .
if @xmath1 is a trigraph tree , then l - hom@xmath36 is polynomial time solvable or @xmath0-complete .
suppose that l - hom@xmath36 is polynomial time solvable .
observe that , since the underlying graph of @xmath1 is a tree , each connected component of @xmath44 is an induced subgraph of @xmath1 .
it follows that l - hom@xmath237 is also polynomial time solvable .
suppose that l - hom@xmath237 is polynomial time solvable .
we prove the theorem by induction on the size of @xmath17 .
hence , we shall assume that for each vertex @xmath71 of @xmath1 , l - hom(@xmath238 ) is polynomial time solvable .
let @xmath24 with lists @xmath32 be an instance of l - hom@xmath36 . if @xmath1 contains no strong loops or strong edges , then @xmath239 and there is nothing to prove .
suppose that @xmath1 contains a strong loop at @xmath71 .
by lemmata [ lem:1 ] , [ lem:2 ] , and [ lem:3 ] we may assume that the lists @xmath32 are arc - consistent and contain representatives and strong representatives .
consider the strong representative @xmath124 of @xmath71 .
let @xmath108 denote the vertices @xmath47 with @xmath60 . if @xmath71 has no symmetric neighbours , @xmath240 .
hence , @xmath77 whenever @xmath95 .
therefore , since the lists @xmath32 are arc - consistent , @xmath24 admits a list @xmath1-colouring with respect to @xmath32 if and only if @xmath241 admits a list @xmath242-colouring with respect to @xmath32 .
since l - hom@xmath242 is polynomial time solvable , the claim follows .
now , suppose that @xmath71 has symmetric neighbours .
since the underlying graph of @xmath1 is a tree , it follows that @xmath243 where @xmath164 is a symmetric neighbour of @xmath71 .
first , suppose that @xmath164 has no loop .
we apply lemma [ lem:8 ] for @xmath244 , @xmath245 , and @xmath246 , and we see that we may assume that the lists @xmath32 are arc - consistent and sparse - dense - consistent on @xmath80 and @xmath247 .
that is , we have @xmath248 or @xmath249 or @xmath250 for each @xmath47 . in particular , because @xmath243 , we have @xmath248 whenever @xmath60 . since the lists @xmath32 are arc - consistent , the claim follows exactly as above .
next , suppose that @xmath164 has a weak loop .
let @xmath120 be the component of @xmath238 to which @xmath164 belongs .
observe that @xmath164 weakly dominates @xmath71 on @xmath251 .
we apply lemma [ lem:8 ] for @xmath244 , @xmath252 , and @xmath253 .
this implies that we may assume that the lists @xmath32 are arc - consistent and separator - consistent on @xmath81 .
moreover , @xmath241 admits a list @xmath242-colouring with respect to @xmath32 , since otherwise lhom@xmath36 has no solution for @xmath24 with lists @xmath32 .
hence , we let @xmath222 be a list @xmath242-colouring of @xmath241 with respect to @xmath32 such that @xmath222 restricted to any component @xmath91 of @xmath241 is a @xmath120-colouring as long as a vertex of @xmath120 appears on the list of some vertex of @xmath91 .
since the lists @xmath32 are separator - consistent on @xmath81 , such a mapping must exist .
in fact , @xmath222 can be constructed in polynomial time , since l - hom@xmath242 is polynomial time solvable .
we extend this mapping to @xmath24 and show that this yields a homomorphism .
let @xmath33 be a mapping defined as follows .
@xmath254 note that since @xmath243 , we have for @xmath116 either @xmath248 or @xmath255 .
hence , the mapping @xmath33 is well - defined .
moreover , @xmath33 clearly respects the lists @xmath32 .
we show that @xmath33 is also a homomorphism .
suppose that @xmath256 for some @xmath63 .
clearly , at least one of @xmath257 must belong to @xmath108 , since @xmath222 is a homomorphism .
also , if @xmath248 , or @xmath77 , or both @xmath258 , we obtain a contradiction by arc - consistency of @xmath32 .
hence , we can assume , by symmetry , that @xmath116 , @xmath259 , and @xmath228 . since the list @xmath32 are arc - consistent , there exists @xmath260 such that @xmath261 .
in particular , @xmath262 since @xmath259 , and hence , @xmath263 belongs to @xmath120 .
therefore , if @xmath91 is the component of @xmath241 to which @xmath8 belongs , @xmath222 restricted to @xmath91 is a @xmath120-colouring .
in particular , @xmath264 .
now , if @xmath265 , both @xmath71 and @xmath164 can not belong to @xmath58 by arc - consistency . if @xmath266 , then @xmath267 since @xmath164 weakly dominates @xmath71 on @xmath251 .
therefore , @xmath268 .
but , since the lists @xmath32 are arc - consistent , and @xmath255 , we obtain a contradiction .
suppose now that @xmath269 for @xmath67 . clearly , @xmath116 or @xmath270 .
again , if @xmath248 , or @xmath77 , or @xmath258 , we have a contradiction .
hence , we can assume , by symmetry , that @xmath116 , @xmath259 , and @xmath228 . also , since @xmath271 is an edge and @xmath259 , @xmath272 belongs to @xmath120 .
as above , @xmath273 can neither be a non - edge nor a weak edge by arc - consistency of @xmath32 , respectively the fact that @xmath164 weakly dominates @xmath71 on @xmath251 .
therefore , @xmath265 , and again , since the lists @xmath32 are arc - consistent and @xmath255 , we have a contradiction .
this proves that @xmath33 is indeed a list @xmath1-colouring of @xmath24 with respect to @xmath32 , and clearly , @xmath33 can be constructed in polynomial time . next ,
suppose that @xmath164 has a strong loop .
consider the strong representative @xmath274 .
if @xmath275 or @xmath276 where @xmath277 , it follows that @xmath248 whenever @xmath60 . in this case , the claim follows , as above , from arc - consistency of @xmath32 .
hence , we may assume @xmath278 .
we observe that @xmath279 contains two components , a component @xmath199 which contains @xmath71 , and a component @xmath108 which contains @xmath164 .
therefore , if we let @xmath244 and @xmath245 , then the list @xmath188 of any vertex @xmath74 either does not contain both @xmath71 and @xmath164 , or @xmath280 .
this shows that @xmath24 is separable on @xmath281 .
in addition , l - hom@xmath282 is polynomial time solvable .
therefore , by lemma [ lem:9 ] , we can solve l - hom@xmath36 for @xmath24 with the lists @xmath32 in polynomial time .
finally , suppose that @xmath1 contains a strong edge @xmath21 , and @xmath45 is an edge of @xmath1 .
let @xmath199 be the component of @xmath279 which contains @xmath71 , and @xmath108 be the component which contains @xmath164 .
let @xmath244 and @xmath245 . if @xmath71 does not appear on the list of any vertex in @xmath24 , then @xmath24 with the lists @xmath32 is an instance of which is polynomial time solvable .
similarly for @xmath164 .
hence , since the lists @xmath32 contain representatives , there exist vertices @xmath283 such that @xmath104 and @xmath284 . now , suppose that the list @xmath58 contains @xmath71 . by arc - consistency
, we must have @xmath285 . since @xmath164 is not adjacent to any vertex of @xmath200 and @xmath285
, no vertex of @xmath200 can appear on the list @xmath58 .
similarly , if @xmath171 , no vertex of @xmath201 appears on the list @xmath58 .
this shows that @xmath24 is separable on @xmath286 .
if @xmath45 is not an edge of @xmath1 , the same argument shows that @xmath24 is separable on @xmath281 . in both cases ,
the claim follows by lemma [ lem:9 ] .
that concludes the proof .
in this section , we extend the dichotomy from the previous section to a larger class of trigraphs which we refer to as _ tree - like_. in fact , we prove the dichotomy for this class using the same tools we used in the previous section .
we have stated these tools in a sufficiently general way to easily allow for this extension .
_ domination property _
we say that an ordering @xmath287 of vertices is a _ domination ordering _
, if @xmath288 weakly dominates @xmath289 whenever @xmath290 .
let @xmath1 be a trigraph , and suppose that @xmath71 is strong loop of @xmath1 .
for a component @xmath120 of @xmath238 , let @xmath291 denote the set of all symmetric neighbours of @xmath71 in @xmath120 . recall that according to our definitions , @xmath164 is a symmetric neighbour of @xmath71 just if both @xmath21 and @xmath45 are in @xmath292 .
we say that a component @xmath120 satisfies _ the domination property _ if 1 .
no vertex of @xmath291 has a strong loop , and 2 .
the vertices of @xmath291 with weak loops admit a domination ordering , each weakly dominates @xmath71 on @xmath251 , and each weakly dominates every vertex of @xmath291 without a loop . _ matching property _
let @xmath126 be the edges of @xmath1 .
we say that @xmath126 _ separates @xmath1 _ if each edge @xmath128 has its endpoints @xmath163 in different components of @xmath130 .
thus @xmath126 separates @xmath1 if and only if there is a partition of @xmath17 such that @xmath126 is the set of all edges between different parts .
let @xmath293 consist of all edges @xmath294 such that neither @xmath21 nor @xmath45 is a strong edge of @xmath1 .
we say that @xmath126 satisfies _ the matching property _ if 1 .
@xmath126 separates @xmath1 , 2 . for any @xmath295 ,
both @xmath236 and @xmath296 are strong loops , 3 .
if @xmath297 , and neither @xmath21 nor @xmath166 is a bridge of @xmath1 , then @xmath169 is not a symmetric edge , and @xmath236 or @xmath298 is a strong loop , 4 . if @xmath299 , and neither @xmath21 nor @xmath169 is a bridge of @xmath1 , then @xmath168 is not a symmetric edge , and @xmath296 or @xmath298 is a strong loop .
if @xmath295 , and @xmath169 and @xmath300 are symmetric edges with @xmath301 , then @xmath302 and @xmath303 .
if @xmath295 , and @xmath169 and @xmath300 are symmetric edges with @xmath304 , then @xmath302 and @xmath305 .
we are now in a position to define the class @xmath306 of _ tree - like trigraphs_. 1
. if @xmath1 has no strong edges ( or loops ) , then @xmath307 2 .
if @xmath1 contains a strong loop at @xmath71 such that @xmath308 and each component of @xmath238 satisfies the domination property , then @xmath307 .
3 . if @xmath1 contains a set @xmath126 of edges such that @xmath309 and the set @xmath126 satisfies the matching property , then @xmath307 .
[ thm:4 ] if @xmath1 is a tree - like trigraph , i.e. , if @xmath310 , then is polynomially equivalent to l - hom@xmath237 . in particular , is polynomial time solvable or @xmath0-complete .
we prove the theorem by structural induction on @xmath1 .
we shall assume that @xmath1 is connected , otherwise we treat each component of @xmath1 separately . if @xmath1 is a digraph , there is nothing to prove .
suppose that @xmath1 contains a strong loop at @xmath71 such that @xmath311 and each component of @xmath238 satisfies the domination property .
if l - hom@xmath36 is polynomial time solvable , then so is l - hom@xmath242 . on the other hand , since @xmath311 , we shall assume , by induction , that l - hom@xmath242 is polynomial time solvable . let @xmath24 with lists @xmath32 be an instance of l - hom@xmath36 . by lemmata [ lem:1 ] , [ lem:2 ] , and [ lem:3 ]
, we may assume that the lists @xmath32 contain representatives and strong representatives .
consider the strong representative @xmath124 .
let @xmath108 be the vertices @xmath47 with @xmath60 .
it follows that there is a unique connected component @xmath120 of @xmath238 such that @xmath312 .
let @xmath244 , let @xmath132 be the vertices of @xmath124 with no loops , and @xmath152 be the vertices of @xmath124 with weak loops . since @xmath120 satisfies the domination property , we have @xmath313 .
also , @xmath152 admits a domination ordering @xmath314 , and each vertex of @xmath152 dominates every vertex of @xmath132 . for @xmath47 ,
let @xmath315 denote the vertex @xmath316 with the smallest index @xmath317 .
now , by lemma [ lem:6 ] , we may assume that for each @xmath47 , either @xmath248 , or @xmath318 , or @xmath185 .
moreover , l - hom@xmath135)$ ] and l - hom@xmath319 are polynomial time solvable since @xmath244 , and by the inductive hypothesis , respectively .
hence , by applying lemma [ lem:8 ] to @xmath191 , we see that may assume that the lists @xmath32 are arc - consistent , separator - consistent on @xmath81 , and sparse - dense - consistent on @xmath81 and @xmath132 .
now , let @xmath222 be a list @xmath242-colouring of @xmath241 with respect to @xmath32 , such that @xmath222 reduced to any component @xmath91 of @xmath241 is a @xmath120-colouring as long as a vertex of @xmath120 appears on the list of some vertex of @xmath91 .
since the lists @xmath32 are separator - consistent on @xmath81 , such mapping must exist and can be found in polynomial time .
we extend the mapping @xmath222 to a mapping @xmath33 as follows .
@xmath320 by the above remark , the mapping @xmath33 is well - defined .
it also clearly respects the lists @xmath32 .
it remains to show that @xmath33 is a homomorphism .
the proof of this follows exactly as in the proof of theorem [ thm : trees ] , and hence , we skip further details .
now , suppose that @xmath1 contains edges @xmath126 satisfying the matching property and such that @xmath321 .
again , we assume that l - hom@xmath322 is polynomial time solvable , and that the lists @xmath32 are arc - consistent , contain representatives and strong representatives .
first , we observe that any subset @xmath323 of @xmath126 satisfies the matching property in @xmath324 .
in particular , @xmath325 .
therefore , we can proceed by induction on @xmath126 .
we denote by @xmath326 the vertices of @xmath1 with at least one incident edge from @xmath126 .
we observe that if a vertex @xmath327 does not appear on any list , we can remove from @xmath126 all edges incident to @xmath71 .
since there is at least one such edge , the claim follows by induction .
hence , we shall assume that each vertex of @xmath326 appears on some list .
let @xmath327 be a strong loop .
consider the strong representative @xmath124 .
we claim that either @xmath243 where @xmath128 , or @xmath124 belongs to a component of @xmath130 . to prove this ,
suppose that @xmath328 and @xmath128 . clearly , all of @xmath21 , @xmath329 , and @xmath330 are symmetric edges
. since @xmath126 separates @xmath1 , at least one of @xmath331 must be in @xmath126 .
but that contradicts the matching property . now , suppose that @xmath124 belongs to a component of @xmath130 .
let @xmath332 denote the union of lists of vertices @xmath47 with @xmath60 .
we have @xmath333 .
since @xmath327 , there exists @xmath164 with @xmath128 or @xmath334 . by symmetry , suppose that @xmath128 , and suppose also that @xmath21 is a weak edge .
hence , @xmath164 is a strong loop .
consider the strong representative @xmath274 .
suppose first that @xmath274 also belongs to a component of @xmath130 .
clearly , @xmath335 .
in particular , @xmath336 where @xmath337 is the union of lists @xmath58 with @xmath171 . furthermore , for @xmath338 and @xmath339 , we have by the matching property that @xmath340 are not edges of @xmath1 , and either @xmath341 or @xmath342 is not an edge of @xmath1 . in particular , if @xmath341 , then arc - consistency of @xmath32 implies @xmath343 .
hence , it follows that the only edge in @xmath1 from @xmath332 to @xmath337 is the edge @xmath21 .
this implies that for any @xmath63 with @xmath60 and @xmath61 , we have @xmath248 and @xmath344 by arc - consistency of @xmath32 .
therefore , we can remove all such edges @xmath99 , and after that , we can remove @xmath21 from @xmath126 .
the claim now follows by induction .
next , suppose that @xmath345 where @xmath346 .
let @xmath347 .
by the matching property , we have @xmath348 .
also , either @xmath349 or @xmath342 is not an edge , and either @xmath350 or @xmath351 is not an edge . if @xmath341 , then arc - consistency of @xmath32 implies @xmath352 .
if @xmath353 and @xmath342 is not an edge , then arc - consistency of @xmath32 implies @xmath354 .
again , the only edge from @xmath332 to @xmath337 is the edge @xmath21 , and the claim follows by induction .
therefore , we may assume that any strong loop @xmath327 is either incident to a strong edge of @xmath126 , or we have @xmath243 where @xmath128 .
recall that @xmath126 separates @xmath1 .
first , suppose that @xmath130 contains exactly two components .
we prove that @xmath24 is separable on @xmath126 .
suppose that we have @xmath60 where @xmath327 .
let @xmath120 be the component of @xmath130 which contains @xmath71 .
suppose that there exists @xmath164 such that @xmath128 or @xmath355 and one of @xmath356 is strong .
then , by the matching property and arc - consistency of @xmath32 , the list @xmath58 contains no vertex of @xmath120 other than @xmath71 .
if no such @xmath164 exists , then @xmath236 must be a strong loop .
hence , by the above assumption , we have @xmath243 where @xmath128 .
therefore , @xmath58 again contains no vertex of @xmath120 other than @xmath71 .
the claim now follows by lemma [ lem:9 ] .
finally , suppose that @xmath130 contains more than two components .
let @xmath323 be a smallest subset of @xmath126 such that @xmath357 is disconnected .
since @xmath1 is connected and @xmath130 is disconnected , the set @xmath323 must exist . as remarked above
, @xmath358 satisfies the matching property in @xmath357 .
also , it can be seen that @xmath323 satisfies the matching property in @xmath1 .
since @xmath323 contains at least one edge and @xmath357 contains exactly two components because of minimality , the claim follows by induction .
that concludes the proof .
the class @xmath306 admits a few natural extensions which we shall only mention tangentially .
for instance , we shall observe the following simple fact .
[ thm : simple ] if each vertex of @xmath1 has a strong loop , and the symmetric graph of @xmath1 contains no triangles , then l - hom@xmath36 is polynomial time solvable .
let @xmath24 with lists @xmath32 be an instance of l - hom@xmath36 .
by lemma [ lem:3 ] , we may assume that the lists @xmath32 contain strong representatives .
since the symmetric graph of @xmath1 has no triangles , it follows that for each @xmath359 , the strong representative @xmath124 contains at most two elements .
the problem now can be reduced to @xmath234 which is polynomial time solvable . as a consequence
, we can extend the class @xmath306 by adding another basis clause to its recursive description : 1 .
if each vertex of @xmath1 has a strong loop , and the symmetric graph of @xmath1 contains no triangles , then @xmath307 . similarly , if @xmath1 has no weak edges , i.e. , if @xmath360 , then l - hom@xmath36 is polynomially solvable @xcite .
hence , we can also add the following clause : 1 . if @xmath1 has no weak edges , then @xmath307 . since the recursive description of the class @xmath306 is complex
, we shall identify a subclass of @xmath306 which can be defined directly .
let @xmath361 denote all edges @xmath21 of @xmath1 such that either a. @xmath21 or @xmath45 is a strong edge of @xmath1 , or b. @xmath21 and @xmath45 are weak edges of @xmath1 , and @xmath71 and @xmath164 are strong loops of @xmath1 .
we say that @xmath1 is a
_ special tree - like trigraph _ if there is a set @xmath362 of edges of @xmath1 such that 1 .
@xmath363 satisfies the matching property , and 2 . for every strong loop @xmath71 of @xmath364 ,
each component @xmath120 of @xmath365 satisfies the domination property in @xmath364 .
let @xmath137 denote the class of all special tree - like trigraphs .
also , let @xmath366 denote the class of all trigraph trees .
@xmath367 .
first , let @xmath1 be in @xmath366 .
consider the set @xmath368 . since each edge in @xmath363 is a bridge of @xmath1 , conditions ( m1 ) , ( m3 ) and ( m4 )
are satisfied for @xmath363 .
also , the edges of @xmath363 do not form cycles in @xmath1 since @xmath1 is a trigraph tree , and hence , ( m5 ) and ( m6 ) are satisfied .
therefore , @xmath363 satisfies the matching property . on the other hand , for every vertex @xmath71 of @xmath1 with a strong loop
, each component of @xmath238 satisfies the domination property , since @xmath71 is adjacent to at most one vertex of this component .
it follows that @xmath369 .
now , for @xmath369 , it suffices to observe that edges @xmath21 of @xmath370 have both @xmath71 and @xmath164 strong loops .
hence , after removing @xmath363 using ( t3 ) and then removing all vertices with strong loops using ( t2 ) , we obtain precisely @xmath44 which is a digraph .
hence , by ( t1 ) , we conclude that @xmath371 . if @xmath1 is a special tree - like trigraph , i.e. , if @xmath369 , then is polynomial time solvable or @xmath0-complete .
although some simple extensions of tree - like , or special tree - like trigraphs are possible ( such as , say , theorem [ thm : simple ] ) , we shall mention in the conclusions some example trigraphs outside @xmath306 for which the first part of theorem [ thm:4 ] fails .
in this section , we describe an extension of the notion of strong representatives to strong edges of trigraphs .
the _ underlying trigraph @xmath83 of @xmath1 _ is the trigraph on the vertices of @xmath1 with strong edges @xmath21 such that @xmath372 or @xmath373 , and with weak edges @xmath21 such that @xmath374 and @xmath375 or @xmath376 .
we denote by @xmath377 the digraph on the vertices of @xmath24 with edges @xmath21 such that @xmath378 or @xmath379 for some @xmath380 .
we denote by @xmath381 the trigraph on the vertices of @xmath1 with strong edges @xmath21 such that @xmath372 or @xmath382 for some @xmath380 , and with weak edges @xmath21 such that @xmath383 , and @xmath375 or @xmath384 for some @xmath380 .
we say that an edge @xmath21 of @xmath1 is _ admissible _ , if there exist vertices @xmath59 such that @xmath60 and @xmath61 .
we say that lists @xmath32 contain _ representatives for strong edges _ ,
if for each admissible strong edge @xmath21 in @xmath1 , there is a set @xmath385 such that a. @xmath386 whenever @xmath95 or @xmath61 , b. each vertex of @xmath387 is a neighbour of @xmath71 or a neighbour of @xmath164 , c. if @xmath71 and @xmath164 have no common neighbours , then @xmath388 contains only neighbours of @xmath71 or only neighbours of @xmath164 , d. if @xmath45 is a strong edge , then each vertex of @xmath389 is a symmetric neighbour of @xmath71 or a symmetric neighbour of @xmath164 , e. if @xmath45 is a strong edge , and @xmath71 and @xmath164 have no common symmetric neighbours , then contains only symmetric neighbours of @xmath71 or only symmetric of @xmath164 .
[ lem:10 ] lists @xmath32 can be transformed to contain representatives for strong edges . by lemmata [ lem:1 ] and [ lem:2 ]
, we may assume that the lists @xmath32 are arc - consistent and contain representatives .
let @xmath33 be a list @xmath1-colouring of @xmath24 with respect to @xmath32 , and let @xmath21 be an admissible strong edge of @xmath1 .
let @xmath390 , let @xmath105 denote all vertices of @xmath17 that are neighbours of @xmath71 or @xmath164 , and let @xmath108 denote all vertices @xmath47 with @xmath60 or @xmath171 .
we have @xmath114 .
also , since the lists @xmath32 are arc - consistent , we have @xmath112 for each @xmath116 .
let @xmath87 be the underlying graph of @xmath24 and @xmath83 be the underlying trigraph of @xmath1 .
it is easy to verify that @xmath33 is a homomorphism of @xmath87 to @xmath83 . in particular
, @xmath33 is a surjective mapping from of @xmath108 to @xmath391 , where @xmath392 .
clearly , @xmath393 .
now , it is not difficult to prove that @xmath33 is a homomorphism of @xmath394)^2 $ ] to @xmath395)^2 $ ] .
( for this , we need the above remark about surjectivity . )
we observe that @xmath91 induces a clique in @xmath394)^2 $ ] .
hence , using the same argument as in the proof of lemma [ lem:3 ] , there is a maximal clique @xmath118 of a minimal chordal completion @xmath89 of @xmath394)^2 $ ] such that for each @xmath396 , we have @xmath121 where @xmath120 is a unique component of @xmath395)^2-\{x , y\}$ ] .
in particular , if @xmath71 and @xmath164 have no common neighbours , then @xmath251 either contains only neighbours of @xmath71 or it contains only neighbours of @xmath164 .
hence , we let @xmath397 . now , if @xmath45 is also a strong edge , we obtain @xmath385 by replacing @xmath24 with the symmetric graph of @xmath24 , and replacing @xmath1 with the symmetric trigraph of @xmath1 .
the remainder of the proof follows exactly as in lemma [ lem:3 ] .
a trigraph @xmath1 is a _ trigraph cycle _ if the underlying graph of @xmath1 is a cycle .
let @xmath1 be a trigraph cycle .
we say that @xmath1 is a _ good cycle _
if @xmath1 has at least one of the following : a. two strong edges , or b. three consecutive strong loops , or c. two pairs of consecutive strong loops , or d. a strong edge and a distinct pair of consecutive strong loops , or e. two strong loops joined by a nonsymmetric edge , or f. a strong loop whose neighbours have no loops , or g. a strong loop with non - symmetric edges to neighbours , or h. a strong edge with at least one endpoint having no loop . if @xmath1 is a good cycle , then the problem l - hom@xmath36 is polynomial time solvable or @xmath0-complete .
figure [ fig : cycle1 ] illustrates example trigraph cycles whose complexity is not determined by our theorem .
these are cycles @xmath1 that contain vertices @xmath163 such that either @xmath372 and @xmath398 , or all of @xmath236 , @xmath21 , @xmath45 , and @xmath296 are edges ( weak or strong ) but at least one is strong .
( only three typical cases of this are shown in the figure . )
we prove the theorem by reducing the problem to an induced subgraph of @xmath1 . unfortunately , if @xmath1 is a trigraph cycle that is not a good cycle , such reduction may not be possible at all .
the complement of the stable cutset problem ( figure [ fig : ex1]c ) is a good example illustrating this difficulty .
we assume that @xmath1 has at least five vertices , since otherwise the claim follows from @xcite .
let @xmath24 with lists @xmath32 be an instance of l - hom@xmath36 .
by lemmata [ lem:1 ] , [ lem:2 ] , [ lem:3 ] , and [ lem:10 ] , we assume that the lists @xmath32 are arc - consistent , contain representatives , strong representatives , and representatives for strong edges .
also , we assume that each vertex @xmath71 of @xmath1 appears on some list , since otherwise we can reduce the problem @xmath238 , which is a trigraph tree , and the claim follows by theorem [ thm : trees ] .
suppose that @xmath1 contains two strong edges @xmath399 , @xmath400 .
if @xmath401 satisfies the matching property , then we are done by theorem [ thm:4 ] . otherwise , we must have @xmath402 and @xmath403 .
( the case @xmath404 and @xmath405 is symmetric . )
let @xmath176 be any vertex ( including @xmath406 ) of @xmath1 other than @xmath164 .
by arc - consistency of @xmath32 , no list @xmath58 contains both @xmath164 and @xmath176 since @xmath21 and @xmath166 are strong , but at least one of @xmath407 is not an edge .
hence , we have @xmath249 whenever @xmath171 . by arc - consistency of @xmath32 , we can reduce the problem to @xmath408 and we are done . next , if @xmath1 contains two pairs of consecutive strong loops @xmath163 and @xmath409 with possibly @xmath410 , or a strong edge @xmath21 and a pair of strong loops @xmath409 where @xmath411 , then @xmath412 satisfies the matching property , and again we are done . hence , suppose that @xmath1 contains strong loops @xmath163 where @xmath21 is an edge , but @xmath45 is not . consider the strong representatives @xmath124 and @xmath274 .
clearly , @xmath124 and @xmath274 are disjoint . also , the only edge from @xmath124 to @xmath274 is the edge @xmath21 .
hence , by arc - consistency of @xmath32 , if @xmath63 with @xmath60 and @xmath61 , we have @xmath248 and @xmath344 . in particular , we can remove all such edges @xmath99 , and afterwards , we remove the edge @xmath21 from @xmath1 .
now , since @xmath413 is a trigraph tree , by theorem [ thm : trees ] , we conclude that l - hom@xmath414 is polynomially equivalent to l - hom@xmath242 .
hence , we are done .
next , suppose that @xmath1 contains a strong loop @xmath71 with neighbours @xmath415 having no loops .
consider the strong representative @xmath124 .
if @xmath243 , then we apply lemma [ lem:8 ] for @xmath244 , @xmath245 , and @xmath246 .
this yields that the lists @xmath32 are arc - consistent and sparse - dense - consistent on @xmath81 and @xmath132 . in particular , @xmath248 whenver @xmath60 .
similarly , if @xmath416 .
hence , we can reduce the problem to @xmath238 and we are done . now , suppose that @xmath1 contains a strong loop @xmath71 with non - symmetric edges between @xmath71 and its neighbours @xmath415 .
clearly , we have @xmath240 . hence , we can reduce the problem to @xmath238 and we are done .
finally , suppose that @xmath1 contains a strong edge @xmath21 whose one endpoint has no loop . by symmetry
suppose that @xmath71 has no loop .
let @xmath417 be the other neighbour of @xmath71 and let @xmath418 be the other neighbour of @xmath164 . consider the representative @xmath385 .
we have that either @xmath419 or @xmath420 . by arc - consistency of @xmath32 ,
no list @xmath58 contains both @xmath164 and @xmath418 since @xmath21 is strong , but @xmath351 is not an edge .
similarly , no list @xmath58 contains both @xmath71 and @xmath164 , or both @xmath71 and @xmath417 .
hence , if @xmath420 , we have @xmath248 whenever @xmath60 , and if @xmath419 , we have @xmath249 whenever @xmath171
. therefore , we can reduce the problem to @xmath238 or to @xmath408 , and we are done .
@xmath421[f]{\phantom{s } } ; " a4"*[o][f]{\phantom{s } } ; { \ar@{-}@/^0.1pc/ " a4";"a4"+(-0.07,0.3 ) } ; { \ar@{-}@/^0.2pc/ " a4"+(-0.07,0.3);"a4"+(0.07,0.3 ) } ; { \ar@{-}@/_0.1pc/ " a4";"a4"+(0.07,0.3 ) } ; { \ar@{-}@/^0.1pc/ " a3";"a3"+(-0.07,0.3 ) } ; { \ar@{-}@/^0.2pc/ " a3"+(-0.07,0.3);"a3"+(0.07,0.3 ) } ; { \ar@{-}@/_0.1pc/ " a3";"a3"+(0.07,0.3 ) } ; " a4"+(0,-0.2)*{x } ; " a3"+(0,-0.2)*{y } ; \endxy \quad \xy / r3pc/ : ( 1.5,0);(0,0);{\xypolygon10"a"{~>{}{\phantom{s } } } } ; " a3";"a4";{\ellipse : a(-55),^{. } } ; { \ar@/^0.15pc/ " a3";"a4 " } ; { \ar@/^0.15pc/ " a4";"a3 " } ; " a3"*[o][f]{\phantom{s } } ; " a4"*[o][f]{\phantom{s } } ; { \ar@{=}@/^0.1pc/ " a4";"a4"+(-0.07,0.3 ) } ; { \ar@{=}@/^0.2pc/ " a4"+(-0.07,0.3);"a4"+(0.07,0.3 ) } ; { \ar@{=}@/_0.1pc/ " a4";"a4"+(0.07,0.3 ) } ; { \ar@{=}@/^0.1pc/ " a3";"a3"+(-0.07,0.3 ) } ; { \ar@{=}@/^0.2pc/ " a3"+(-0.07,0.3);"a3"+(0.07,0.3 ) } ; { \ar@{=}@/_0.1pc/ " a3";"a3"+(0.07,0.3 ) } ; " a4"+(0,-0.2)*{x } ; " a3"+(0,-0.2)*{y } ; \endxy \quad \xy / r3pc/ : ( 1.5,0);(0,0);{\xypolygon10"a"{~>{}{\phantom{s } } } } ; " a3";"a4";{\ellipse : a(-55),^{. } } ; { \ar@/^0.15pc/ " a3";"a4 " } ; { \ar@/^0.15pc/ " a4";"a3 " } ; " a3"*[o][f]{\phantom{s } } ; " a4"*[o][f]{\phantom{s } } ; { \ar@{=}@/^0.1pc/ " a4";"a4"+(-0.07,0.3 ) } ; { \ar@{=}@/^0.2pc/ " a4"+(-0.07,0.3);"a4"+(0.07,0.3 ) } ; { \ar@{=}@/_0.1pc/ " a4";"a4"+(0.07,0.3 ) } ; { \ar@{-}@/^0.1pc/ " a3";"a3"+(-0.07,0.3 ) } ; { \ar@{-}@/^0.2pc/ " a3"+(-0.07,0.3);"a3"+(0.07,0.3 ) } ; { \ar@{-}@/_0.1pc/ " a3";"a3"+(0.07,0.3 ) } ; " a4"+(0,-0.2)*{x } ; " a3"+(0,-0.2)*{y } ; \endxy \quad \xy / r3pc/ : ( 1.5,0);(0,0);{\xypolygon10"a"{~>{}{\phantom{s } } } } ; " a3";"a4";{\ellipse : a(-55),^{. } } ; { \ar@{->}@/^0.2pc/ " a3";"a4 " } ;
{ \ar@{=>}@/^0.2pc/ " a4";"a3 " } ; " a3"*[o][f]{\phantom{s } } ; " a4"*[o][f]{\phantom{s } } ; { \ar@{=}@/^0.1pc/ " a4";"a4"+(-0.07,0.3 ) } ; { \ar@{=}@/^0.2pc/ " a4"+(-0.07,0.3);"a4"+(0.07,0.3 ) } ; { \ar@{=}@/_0.1pc/ " a4";"a4"+(0.07,0.3 ) } ; { \ar@{-}@/^0.1pc/ " a3";"a3"+(-0.07,0.3 ) } ; { \ar@{-}@/^0.2pc/ " a3"+(-0.07,0.3);"a3"+(0.07,0.3 ) } ; { \ar@{-}@/_0.1pc/ " a3";"a3"+(0.07,0.3 ) } ; " a4"+(0,-0.2)*{x } ; " a3"+(0,-0.2)*{y } ; \endxy $ ] finally , we describe how we can use our results to classify the complexity of finding surjective list homomorphisms for some trigraphs .
we say that a homomorphism @xmath33 of @xmath24 to @xmath1 is a _ a ( vertex ) surjective homomorphism _ if @xmath33 is a surjective mapping of @xmath41 onto @xmath17 .
the _ surjective list @xmath1-colouring problem _ sl - hom@xmath36 takes as input a digraph @xmath24 with lists @xmath32 , and asks whether or not @xmath24 admits a surjective list homomorphism to @xmath1 with respect to @xmath32 .
let @xmath422 be the digraph obtained from a trigraph @xmath1 by removing all vertices @xmath71 with a strong loop at @xmath71 or a strong edge @xmath21 or @xmath45 for some @xmath164 .
[ thm:5 ] if @xmath1 is a special tree - like trigraph , i.e. , if @xmath369 , then is polynomially equivalent to sl - hom@xmath423 .
in particular , sl - hom@xmath36 is polynomial time solvable or @xmath0-complete .
let @xmath1 be any trigraph ( not necessarily a special tree - like ) .
first , we observe that sl - hom@xmath36 is polynomially reducible to l - hom@xmath36 , since sl - hom@xmath36 is a special case of l - hom@xmath36 . now
, suppose that @xmath1 contains a digraph @xmath212 as an induced subgraph .
let @xmath424 with lists @xmath216 be an instance of l - hom@xmath425 .
let @xmath83 be the digraph associated with @xmath1 , and let @xmath24 be the disjoint union of @xmath424 and @xmath83 .
define @xmath426 for @xmath427 , and @xmath428 for @xmath429 .
it now follows that @xmath424 admits a list @xmath212-colouring with respect to @xmath216 if and only if @xmath24 admits a surjective list @xmath1-colouring with respect to @xmath32 .
this yields that l - hom@xmath423 is polynomially equivalent to sl - hom@xmath423 .
in fact , we can also conclude that sl - hom@xmath423 is polynomially reducible to sl - hom@xmath36 .
now , let @xmath1 be a special tree - like digraph .
let @xmath362 be the set of edges from the definition of @xmath1 .
let @xmath24 with lists @xmath32 be an instance of sl - hom@xmath36 .
we can assume that each vertex of @xmath1 appears on some list , since otherwise there is no solution .
now , following the proof of theorem [ thm:4 ] , we conclude ( depending on @xmath32 ) that there is a set @xmath430 such that the instance @xmath431 of l - hom@xmath36 is polynomially reducible to an instance of l - hom@xmath432 ) .
in fact , the proof implies the instance is reducible to an instance of l - hom@xmath433 where @xmath434 are the vertices incident to the edges of @xmath435 .
moreover , since each vertex of @xmath1 appears on some list , the edges of @xmath436 are only between strong loops .
hence , @xmath434 contains all vertices of @xmath1 incident to strong edges .
therefore , @xmath437 .
also , by the definition of @xmath1 , for each strong loop @xmath71 of @xmath438 , each connected component of @xmath439 satisfies the domination property .
hence , by theorem [ thm:4 ] , @xmath433 is polynomially equivalent to @xmath423 . also , as remarked earlier , @xmath423 is polynomially equivalent to l - hom@xmath423 .
hence , @xmath36 is polynomially reducible to sl - hom@xmath423 .
that concludes the proof .
the proof of this theorem in particular implies the following
. if @xmath1 is a digraph , then l - hom@xmath36 and sl - hom@xmath36 are polynomially equivalent .
in fact , similarly , one can prove a stronger statement .
if for no vertices @xmath440 of @xmath1 we have @xmath372 and @xmath441 , then l - hom@xmath36 and sl - hom@xmath36 are polynomially equivalent .
note that if we have vertices @xmath440 as described above , then an instance @xmath24 in which @xmath71 occurs on some list allows us to use arc - consistency to make sure that no vertex has both @xmath164 and @xmath417 on its list ; while if @xmath71 appears on no list , this no longer happens .
we investigated the list ( and surjective list ) homomorphism problems for trigraphs @xmath1 . when @xmath1 is a digraph , we know that each such problem is polynomial time solvable or @xmath0-complete @xcite
however , there are small trigraphs @xmath1 for which such dichotomy is not known , for instance , the trigraph in figure [ fig : ex1]a ( corresponding to the so - called `` stubborn '' problem from @xcite ) .
hence , we have tried to identify properties of trigraphs @xmath1 which allow us to prove dichotomy . with this in mind , we have defined the class @xmath443 of tree - like trigraphs , and we proved that these trigraphs enjoy dichotomy . slightly easier to define
is the class @xmath137 of special tree - like trigraphs , included in @xmath443 . in particular , the class of trigraphs whose underlying graph is a tree , is included in @xmath137 ( and hence in @xmath444 ) .
we now offer some tangential evidence that our class @xmath443 carves out a reasonable portion of trigraphs where l - hom(@xmath1 ) is polynomially equivalent to l - hom(@xmath44 ) .
for instance , one way to violate the matching property is by having a vertex with a strong loop adjacent to two other vertices with strong loops joined by a symmetric edge .
the trigraph in figure [ fig : ex1]b illustrates this possibility , and also illustrates that a polynomial time solvable ( in this case trivial ) problem l - hom(@xmath44 ) can arise from an @xmath0-complete problem l - hom(@xmath1 ) .
( in this case , @xmath24 admits a homomorphism to @xmath1 if and only if the complement of @xmath24 is 3-colourable . )
the trigraph in figure [ fig : ex1]c illustrates another way to have this take place . here
, the two other vertices have weak loops and are joined by a strong symmetric edge . in this case
l - hom(@xmath1 ) is also np - complete , since it corresponds ( in the complement ) to the stable cutset problem @xcite .
thus we again have an easy with a hard l - hom(@xmath1 ) . of course , it is possible that some class of trigraphs enjoys dichotomy for reasons different from polynomial equivalence of l - hom(@xmath1 ) and l - hom(@xmath44 ) .
however , the stubborn problem ( figure [ fig : ex1]a ) illustrates the fact that there are trigraphs @xmath445 where even the dichotomy is not clear .
we note that the trigraph @xmath1 for the stubborn problem ( the trigraph in figure [ fig : ex1]a ) violates the domination property , since the statement ( d2 ) does not hold for the strong loop at @xmath71 .
finally , we have also introduced the surjective list homomorphism problem sl - hom(@xmath1 ) as an interesting variant of l - hom(@xmath1 ) .
we were able to completely classify the complexity of sl - hom(@xmath1 ) for trigraphs @xmath1 in the class @xmath446 by proving polynomial equivalence with sl - hom(@xmath447 .
this result implies , in particular , that the two problems , l - hom(@xmath1 ) and sl - hom(@xmath1 ) , may not necessarily have the same complexity for all trigraphs @xmath1 .
this difference was not noted before , since the two problems are polynomially equivalent for all digraphs @xmath1 .
in fact , we have given a general condition for trigraphs @xmath1 under which the two problems sl - hom(@xmath1 ) and l - hom(@xmath1 ) are polynomially equivalent .
nonetheless , the trigraph @xmath1 in figure [ fig : ex1]d illustrates a case where sl - hom(@xmath1 ) is polynomial time solvable , because using representatives and arc - consistency reduces all lists to size at most two , while l - hom(@xmath1 ) is @xmath0-complete , because @xmath238 corresponds to the stable cutset problem @xcite . | trigraph list homomorphism problems ( also known as list matrix partition problems ) have generated recent interest , partly because there are concrete problems that are not known to be polynomial time solvable or @xmath0-complete .
thus while _ digraph _ list homomorphism problems enjoy _ dichotomy _ ( each problem is @xmath0-complete or polynomial time solvable ) , such dichotomy is not necessarily expected for trigraph list homomorphism problems .
however , in this paper , we identify a large class of trigraphs for which list homomorphism problems do exhibit a dichotomy .
they consist of trigraphs with a tree - like structure , and , in particular , include all trigraphs whose underlying graphs are trees .
in fact , we show that for these tree - like trigraphs , the trigraph list homomorphism problem is polynomially equivalent to a related _ digraph _ list homomorphism problem .
we also describe a few examples illustrating that our conditions defining tree - like trigraphs are not unnatural , as relaxing them may lead to harder problems .
trigraph , list homomorphism , matrix partition , trigraph homomorphism , surjective list homomorphism , dichotomy , trigraph tree |
in recent years ultrashort , high - harmonic generated vacuum and extreme ultraviolet ( vuv / xuv ) laser pulses have become available .
they open the way towards time - resolved observation of electronic dynamics in plasmas , atoms and condensed matter , for an overview see e.g. @xcite . electronic motion in atoms and relaxation processes can now be studied experimentally on sub - femtosecond time scales , e.g. @xcite , for a recent overview see @xcite .
this creates the need for complementary theoretical investigations .
interaction of laser radiation with matter has been actively studied in recent years using a quantum kinetic theory which was derived using nonequilibrium greens functions ( negf ) methods , e.g. @xcite .
in contrast , application of negf to single atoms has only recently been attempted @xcite . while the ground state properties were very well reproduced , time - dependent calculations describing the ionization dynamics of atoms during and after a laser pulse turned out to be computationally very expensive @xcite . in this paper
we give a brief overview on the application of negf to time - resolved laser - atom interaction and outline the current problems .
furthermore , we discuss an alternative approach which is based on the @xmath0electron wave function the multiconfiguration time - dependent hartree - fock method ( mctdhf ) .
the latter is quite efficient in application to few - electron atoms and small molecules .
this is demonstrated by numerical results for a four - electron one - dimensional model atom subject to a short laser pulse .
the unperturbed hamiltonian of @xmath1 electrons in the atom is given by ( we use atomic units ) @xmath2 where the potential of the nucleus , @xmath3 , as well as the two - particle coulomb interaction @xmath4 are assumed to be spin - independent .
for times @xmath5 , the atom is disturbed by a time - dependent external field , and the hamiltonian is modified : @xmath6 where the electromagnetic field is described in dipole approximation , and the field envelope is assumed to have a gaussian shape with a pulse duration @xmath7 and @xmath8 @xmath9 \exp \left[- \frac{(t - t_{\text{0}})^2}{2\tau^2}\right ] \,.\end{aligned}\ ] ] to treat the interaction of an ensemble of atoms with the electromagnetic field a statistical theory is required .
it can be based on density operators @xcite or nonequilibrium greens functions .
here we discuss the latter concept .
starting point are the fermionic field operators of the electrons ( the nuclei are assumed to be fixed ) @xmath10 und @xmath11 , which guarantee that the spin statistics theorem is obeyed , @xmath12 where @xmath13 and we denoted @xmath14 [ spin indices will be suppressed in the following ] .
the nonequilibrium properties of the electrons are completely determined by the two - time correlation functions ( averaging is over the @xmath1-particle density operator of the system ) @xmath15 the two independent functions @xmath16 and @xmath17 contain the complete statistical information , in particular the nonequilibrium one - body density matrix , according to @xmath18 the equations of motion of @xmath16 and @xmath17 are the keldysh / kadanoff - baym equations ( kbe ) , e.g. @xcite which are the most general starting point for developing a quantum statistical theory of matter in a strong electromagnetic field . in recent years
the kbe have been successfully solved numerically for spatially homogeneous plasmas , e.g. @xcite and for a weakly inhomogeneous electron plasma @xcite .
yet these solutions require to propagate the correlation functions in the two - time plain which is computationally very expensive .
therefore , approximate methods have been developed .
one possibility consists in expressing the correlation functions by the simpler single - particle distribution function @xmath19 [ or density matrix @xmath20 , eq .
( [ rho - def ] ) ] via the so - called generalized kadanoff - baym ansatz @xcite . for plasmas interacting with a strong field
this ansatz has been generalized in refs .
@xcite . the emerging quantum kinetic equation for the wigner distribution which includes nonlinear field effects , inverse bremsstrahlung
, collisional absorption and harmonic generation has been solved in ref .
the second direction of recent research in the kinetic theory of laser - plasma interation was devoted to include neutral atoms and their excitation and ionization .
starting from the kbe atomic bloch equations were derived @xcite which allow to compute the absorption spectrum of an ensemble of atoms in thermodynamic equilibrium .
this approach is very useful for weak excitation near the linear response regime .
however , in the case of strong field ionization of atoms we expect essential nonequilibrium dynamics effects related to ionization of several electrons and electron - electron correlations developing rapdily on femtosecond and even sub - femtosecond time scales .
these phenomena which have already been observed in experiments require a fully nonequilibrium approach .
starting again from the kbe we have in recent years developed a quantum kinetic approach for the electron dynamics in a single atom .
the main difference to the previous applications to plasmas is that now the system of electrons under study is spatially strongly inhomogeneous due to the confinement by the potential @xmath21 of the atomic nucleus . to solve this problem the two - time correlation functions ( [ ggtls_def ] )
were expanded into a basis of single - particle states or hartree - fock orbitals .
this approach was successfully tested for `` artificial atoms '' , i.e. electrons in quantum dots in refs .
more recently it was applied by the present authors to electrons in small atoms @xcite such as a one - dimensional beryllium model atom following earlier work by van leeuwen and co - workers @xcite . in these works
it was found that the ground state of few - electron atoms is well reproduced with simple approximations for the selfenergy such as the second born approximation .
the situation is more complex for time - dependent calculations . while lowest order calculations using a time - dependent hartree - fock selfenergy are well feasible @xcite with the inclusion of correlations , simulations could only be carried out for very short times .
this is a particular problem of ionization dynamics where one has to simultaneously treat bound and free electron states which requires large basis sets . on the other hand ,
the interaction of atoms with a strong laser field can , in many cases , be assumed to be a coherent process .
further , at sufficiently low density , the interaction between atoms can be neglected .
also , at room temperature conditions the density matrix of the atom is dominated by the contribution of the ground state wave function , and one can instead of quantum statistics apply a pure state description . at the conditions described above the initial state @xmath22 of an atom
is given by the ground state wave function or a coherent superposition of atomic eigenfunctions .
then the dynamics following an external exciation is described by the time - dependent @xmath1-electron schrdinger equation ( tdse ) with the hamiltonian ( [ h ] ) @xmath23 electrons , depending on the dimensionality .
therefore , approximate methods have been developed .
a particularly successful approach is the _ multiconfiguration time - dependent hartree - fock _ ( mctdhf ) method due to cederbaum @xcite , see also @xcite and references therein .
we have , therefore , implemented the mctdhf method for photoionzation of atoms in a strong field and briefly outline the main idea .
starting point is the following ansatz for the many - electron wave function as linear combination of time - dependent slater determinants which are denoted in terms of the occupation numbers @xmath24 of the single - particle orbitals @xmath25 the sum runs over an integer number @xmath26 of determinants . in the limit
@xmath27 the method yields the exact wave function , whereas for @xmath28 one recovers the time - dependent hartree - fock approximation ( tdhf ) .
thus , by increasing @xmath29 one can systematically improve the treatment of electronic correlations .
the equations of motion for the coefficients @xmath30 and the single - particle orbitals @xmath31 are obtained by a variational principle and read , with a matrix notation ( with respect to slater determinants ) @xmath32\ , , \end{aligned}\ ] ] where @xmath33 and @xmath34 denote the single - particle and two - particle density matrix , and @xmath35 is the single - particle hamiltonian .
further , @xmath36 is the operator of the mean - field potential which , in coordinate representation , is given by @xmath37 . eqs .
( [ eq_wavefunction],[eq_orbitals ] ) show that the expansion coefficients @xmath38 obey a time - dependent schrdinger equation , whereas the single - particle orbitals @xmath39 are the solutions of a more complicated non - linear integro - differential equation . [ be_ion0 ]
we now present numerical results for photoionization of a four - electron one - dimensional model atom ( beryllium , @xmath40 ) for which we presented nonequilibrium greens functions results in ref .
@xcite . for the @xmath41 simulations we use regularized coulomb potentials , @xmath42 , and @xmath43^{-1/2 }
\ $ ] where the screening parameters are introduced to avoid divergencies .
this model has been used successfully in many studies of atom - laser interaction where also the influence of @xmath44 has been investigated .
following refs .
@xcite we use @xmath45 .
a sketch of the atomic core potential and the ( square of the ) occupied hf basis functions ( shifted by the orbital energies ) is shown in fig .
[ be_ion0 ] together with the total electron density .
there are two bound states at the energies @xmath46 and @xmath47 , each occupied by two electrons .
the first task is to obtain accurate results for the ground state .
the results of our mctdhf calculations for the ground state energy are presented in the table .
convergence with respect to the correlation energy is reached for @xmath48 . in comparison
, the ground state energy for the same system from negf calculations in second born approximation is found to be equal to @xmath49 , i.e. on the level of mctdhf with @xmath50 .
.ground state energy of the 1d beryllium atom with @xmath45 for different numbers of determinants @xmath29 .
we used a basis with 50 eigenfunctions of the single - particle hamiltonian which assured convergence with respect to the basis size . [ cols="^,^,^,^,^,^,^,^,^,^",options="header " , ] as a next step , we use the ground state wave function as the initial state and expose the atom to a time - dependent laser field of the form ( [ eq : gauss ] ) with a fixed number of cycles choosing the pulse width according to @xmath51 and an amplitude of @xmath52 .
we consider two frequencies , @xmath53 and @xmath54 which corresponds to a keldysh - parameter @xmath55 , [ @xmath56 is the ionization potential , and @xmath57 the ponderomotive potential , @xmath58 of @xmath59 and @xmath60 . in the first case
we expect that direct photoionization is not possible and the dominant process will be tunnel ionization .
in contrast , in the second case multiphoton ionization should be expected .
the time - dependent numerical results of our mctdhf calculations are shown in fig .
[ be_ion ] . in the lower row
we show the ionization yield [ time - dependent expectation value of the number of ionized electrons @xmath61 .
interestingly , @xmath62 is much larger for the smaller photon energy ( case 1 ) which confirms our previous results @xcite .
this is due to the much larger electron excursion amplitude in the low frequency laser field [ recall that , for a free electron , the amplitude is @xmath63 . for the larger photon frequency @xmath64
, @xmath65 is strongly reduced .
nevertheless , the two electrons in the upper state with energy @xmath66 are driven substantially away from the nucleus during one half - cycle of the field but they are driven back during the following half - cycle . as a consequence , the ionization probability is substantially reduced . in this second case , mctdhf calculations with different numbers of determinants @xmath29 are very close to each other , and even a tdhf calculation yields the qualitatively correct result .
more interesting is the behavior for the smaller frequency . here
two electrons are driven far away from the nucleus and they occupy continuum states which leads to strong correlation effects . as a result , tdhf calculations while overall giving the correct picture become inaccurate already after about one laser cycle ( @xmath67 ) .
[ be_ion ] summarizing , we have presented mctdhf calculations for a four electron model atom exposed to strong laser pulses with two different frequencies .
these calculations are rather efficient and allow to follow the time - resolved ionization dynamics and the history of electronic correlations in single atoms .
they show that , overall , tdhf gives the correct tendencies .
however , it fails to accurately describe the dynamics of electrons at longer times , in particular in cases of strong ionization . in future work
we will concentrate on correlation dynamics and on the inneratomic relaxation as well as on further comparisons with nonequilibrium greens functions results .
this work is supported by the deutsche forschungsgemeinschaft via sfb - tr 24 and by the u.s .
department of energy award de - fg02 - 07er54946 . | theoretical approaches to the photoionization of few - electron atoms are discussed .
these include nonequilibrium greens functions and wave function based approaches . in particular , the multiconfiguration time - dependent hartree - fock method is discussed and applied to a model one - dimensional atom with four electrons .
we compute ground state energies and the time - dependent photoionization by the field a strong laser pulse with two different frequencies in the ultraviolet . |
finite order invariants of immersions of a closed orientable surface into @xmath0 have been introduced in o , where all order 1 invariants have been classified .
explicit formulae for the majority of order 1 invariants have been given in a , f . all higher order invariants have been classified in h , and the analogue of all the above for non - orientable surfaces has appeared in n. a first step in the study of finite order invariants of immersions of surfaces into general 3-manifolds appears in q where one specific order 1 invariant has been studied . in the present work we study finite order invariants of null - homotopic immersions of a closed orientable surface @xmath1 into an aspherical orientable 3-manifold @xmath2 .
the immersions being null - homotopic will enable us to lift the immersions to the universal covering @xmath3 of @xmath2 , and @xmath2 being aspherical will guarantee a simple form for @xmath4 for discrete subset @xmath5 .
we develop the foundations for the study of finite order invariants in our setting , as has been done in o for the case @xmath6 ( and analogous to chord diagrams and the 1-term and 4-term relations in knot theory ) .
we then classify all order 1 invariants .
revisions to this paper will appear at www.math.biu.ac.il/@xmath7tahl/publications.html
let @xmath2 be an aspherical orientable 3-manifold , i.e. @xmath8 for all @xmath9 ( a sufficient condition being that @xmath10 ) .
let @xmath11 be the universal covering , so @xmath3 is contractible . for @xmath12
denote @xmath13 . if @xmath14 is a path in @xmath2 with initial point @xmath15 and if @xmath16 then we denote the lift of @xmath14 to @xmath3 with initial point @xmath17 by @xmath18 .
if @xmath19 is another point and @xmath14 is a path from @xmath15 to @xmath20 then we define a bijection @xmath21 by @xmath22 , where @xmath23 denotes concatenation from left to right .
we choose @xmath12 , @xmath16 and for each @xmath19 we choose a path @xmath24 from @xmath15 to @xmath20 .
we then label the points in each @xmath25 by elements of @xmath26 via the bijection @xmath27 .
we now check in what way this labeling depends on our choices .
so let @xmath28 , @xmath29 and for each @xmath19 , @xmath30 a path from @xmath31 to @xmath20 , be another such choice .
in order to identify the elements of @xmath26 with those of @xmath32 we need to make one additional choice , the dependence on which will be apparent as well , namely we choose a path @xmath33 from @xmath15 to @xmath31 and identify @xmath26 with @xmath32 by @xmath34 where @xmath35 is the path inverse to @xmath33 ( i.e. @xmath36 ) .
finally , let @xmath37 be the unique ( up to homotopy ) path in @xmath2 from @xmath15 to @xmath31 such that @xmath38 .
now for each @xmath19 : @xmath39 .
so given the identification between @xmath26 and @xmath32 determined by @xmath33 , the labeling of the points in @xmath25 have changed by left multiplication by @xmath40 and right multiplication by @xmath41 . note that @xmath40 depends only on @xmath42 and so we get left multiplication by the _ same _
element for all points @xmath20 .
we summarize this in the following : [ dep ] the labeling of the points of @xmath25 by elements of @xmath26 is well defined up to left multiplication by one common element in @xmath26 for all @xmath20 , and right multiplication by an element in @xmath26 which may depend on @xmath20 .
once the dependence on choices has been established , we fix @xmath43 once and for all , and so the labeling of the points in @xmath25 for each @xmath19 is from now on fixed .
note that this labeling is necessarily _ not _ locally constant
. there will be two families of bijections that we will have occasion to use , and we will now see how they are expressed in terms of the labeling .
first let @xmath44 be a deck transformation , then the restriction of @xmath45 to each @xmath25 is a bijection onto itself .
let @xmath46 be the unique element such that @xmath47 .
then for any @xmath48 and any @xmath19 , @xmath49 , that is , in terms of the labeling , @xmath45 is given by common left multiplication by @xmath50 .
( by _ common _ we mean as above , that it is the _ same _
@xmath50 for all @xmath20 . )
secondly , given @xmath51 let @xmath52 be a path from @xmath20 to @xmath53 , then @xmath52 defines a bijection @xmath54 given by @xmath55 .
we have for any @xmath48 , @xmath56 , that is , in terms of the labeling , @xmath57 is given by right multiplication by @xmath58 .
it is also clear from the above calculations that any common left multiplying element , and any right multiplying element , can be realized by appropriately choosing @xmath45 and @xmath52 respectively .
we summarize this in the following : [ dk ] 1 .
the effect of a deck transformation is a common left multiplication , and any such common left multiplication can be realized by a deck transformation .
the effect of @xmath57 is right multiplication , and any such right multiplication can be realized by some @xmath52 from @xmath20 to @xmath53 .
we have noted that our labeling is not locally constant . by proposition dk(2 ) we see that any such discontinuity is always given by right multiplication .
denote by @xmath59}}$ ] the group ring of @xmath26 with coefficients in @xmath60 .
we will in fact not use the full ring structure , but only the left and right actions of @xmath26 on @xmath59}}$ ] . for each @xmath19
, @xmath61 is a free abelian group with basis in one - to - one correspondence with @xmath25 .
this is true by a mayer - vietoris sequence , since @xmath3 is contractible .
so via our labeling , for each @xmath19 we may identify @xmath61 with @xmath59}}$ ] .
also note @xmath62 .
any deck transformation induces an automorphism of @xmath61 for each @xmath20 .
it follows from proposition dk(1 ) that in terms of the identification of @xmath61 with @xmath59}}$ ] , this automorphism ( of @xmath59}}$ ] as an abelian group ) is given by left multiplication by some common @xmath46 .
similarly , given @xmath51 , a path @xmath52 from @xmath20 to @xmath53 naturally induces an isomorphism from @xmath61 to @xmath63 , and it follows from proposition dk(2 ) that in terms of the identifications with @xmath59}}$ ] , this isomorphism is given by right multiplication by some element of @xmath26 .
let @xmath2 be an oriented aspherical 3-manifold , let @xmath1 be a closed oriented surface , and let @xmath64 be a regular homotopy class of immersions @xmath65 which are null - homotopic .
a ce point of an immersion @xmath65 is a point of self intersection of @xmath66 for which the local stratum in @xmath67 corresponding to the self intersection , has codimension one .
we distinguish four types of ces which we name @xmath68 . in the notation of hk
they are respectively @xmath69 , @xmath70 , @xmath71 , @xmath72 .
the four types may be demonstrated by the following local models , where letting @xmath73 vary , we obtain a 1-parameter family of immersions which is transverse to the given codim 1 stratum , intersecting it at @xmath74 .
@xmath75 : @xmath76 , @xmath77 .
@xmath78 : @xmath76 , @xmath79 .
@xmath80 : @xmath76 , @xmath81 , @xmath82 .
@xmath83 : @xmath76 , @xmath81 , @xmath84 , @xmath85 .
see figure hoi , which corresponds to some small @xmath86 .
if @xmath87 is one of the above four ce types , then we denote by @xmath88 the number of sheets involved in the given configuration , that is , @xmath89 , @xmath90 , @xmath91 .
let @xmath92 be the space of all immersions with precisely @xmath93 ces , in particular , @xmath94 is the space of all stable immersions . a choice of one of the two sides of the local codim 1 stratum at a given point of the stratum , is represented by the choice of @xmath95 or @xmath96 in the formulae above .
we will refer to such a choice as a _ co - orientation _ for the configuration of the self intersection . for types @xmath75 and @xmath80 , the configuration of the self intersection at the two sides of the stratum is distinct , namely , for @xmath95 there is an additional 2-sphere in the image of the immersion , and we permanently choose this side ( @xmath95 ) as our positive side for the co - orientation . for types @xmath78 and @xmath83
, the configuration of the self intersection on the two sides of the strata is the same .
we will now further discuss the symmetries of our four configurations .
a symmetry of a ce configuration is an orientation preserving diffeomorphism from a neighborhood of the ce in @xmath2 to itself , which maps sheets onto sheets .
a symmetry @xmath97 will be said to _ preserve _ the co - orientation of the ce if the image under @xmath97 of motion into some side of the stratum , is again motion into the same side , ( after identifying each sheet with its image ) .
the comment of the previous paragraph may now be stated as follows : all symmetries of @xmath98 configurations preserve co - orientation , whereas configurations @xmath99 admit co - orientation reversing symmetries .
we further notice , that configurations @xmath100 admit co - orientation preserving symmetries that realize any given permutation of the sheets , whereas for configuration @xmath83 , any co - orientation preserving symmetry induces an even permutation of the sheets , and any co - orientation reversing symmetry induces an odd permutation of the sheets .
these facts may be seen by slightly resolving the ce into one of the sides , and attempting to construct the diffeomorphism @xmath97 for the resolved configurations . given an immersion @xmath101 , an _ @xmath93-co - orientation _ for @xmath66 is a choice of co - orientation at each of the @xmath93 ce points @xmath102 of @xmath66 .
( this is called a temporary co - orientation in o , h , n . )
a _ proper _ @xmath93-co - orientation for @xmath103 is an @xmath93-co - orientation where at each ce of type @xmath75 and @xmath80 , the co - orientation chosen is the permanent one mentioned above .
( this definition is slightly weaker than that in h. ) for immersion @xmath103 , and proper @xmath93-co - orientation @xmath104 for @xmath66 , let @xmath19 be one of its @xmath93 ce points and we define @xmath105 as the symbol @xmath106 which is composed of : \(1 ) a symbol @xmath107 which is the configuration of the given ce .
\(2 ) a sequence of elements @xmath108 which is determined as follows : let @xmath109 be a lift of @xmath66 .
there are @xmath88 sheets of @xmath110 passing through points of @xmath25 which in turn are labeled by elements @xmath108 ( section lb ) .
if the ce is of type @xmath75,@xmath78 or @xmath80 then the order of the @xmath111s is arbitrarily chosen . as to @xmath83 configuration , we choose once and for all a rule by which an orientation on a simplex determines an ordering on its four faces , up to an even permutation .
now , we resolve the ce of type @xmath83 positively with respect to @xmath104 , i.e. slightly deform @xmath66 ( not @xmath112 ) near @xmath20 , to obtain a stable immersion in a neighborhood of @xmath20 , and write the elements @xmath113 in order determined by the orientation of @xmath2 restricted to the small simplex created by this resolution .
( this ordering convention is related to the symmetry properties discussed above ) .
\(3 ) to each @xmath114 , there corresponds a sign @xmath115 which is determined as follows : we look at @xmath116 ( not the lift ) , and resolve the ce at @xmath20 according to @xmath104 . for types
@xmath117 this creates a little sphere in the image of @xmath66 , which bounds a little 3-cell @xmath118 in @xmath2 ( for @xmath83 this is a simplex and for @xmath98 recall that @xmath104 is proper ) .
each @xmath111 we have , corresponds to a sheet involved in the ce .
this sheet may have @xmath118 on its non - preferred side , ( determined by the orientations of @xmath1 and @xmath2 ) , in which case we set @xmath119 , or otherwise @xmath118 is on its preferred side , in which case @xmath120 . for type
@xmath78 the region @xmath118 is not bounded by the local configuration , but may still be defined , e.g. for @xmath96 in the formula for @xmath78 above , @xmath118 will be a region consisting of points close to the origin and satisfying @xmath121 .
now the signs @xmath122 are determined by the _ opposite _
convention than above , namely , a sheet having @xmath118 on its _ preferred _ side will have @xmath119 , and if @xmath118 is on its _ non - preferred _ side we set @xmath120 .
the reason for the opposite convention is to obtain slightly nicer formulae in the end .
( it is consistent with the conventions in o , chosen there in a similar way for the same reason ) .
\(4 ) the subscript @xmath123}}$ ] which will be called the _ degree _ of the ce , is determined as follows : let @xmath109 be the same lift as used in ( 2 ) .
wherever @xmath110 passes through a point of @xmath25 , we push it slightly into the preferred side of @xmath110 in @xmath3 , obtaining a map @xmath124 .
this map represents an element of @xmath61 which we identify with an element @xmath125}}$ ] ( section lb ) .
if @xmath126 are the @xmath93 ces of @xmath103 then we define @xmath127 to be the @xmath93-tuple @xmath128 , so @xmath129 where @xmath130 is the set of all @xmath93-tuples of symbols of the form @xmath106 .
in addition to the permanent choices of section lb , @xmath131 also depends on the choice of lift @xmath112 , and various ordering choices .
we thus define an equivalence relation on @xmath130 to be the equivalence relation generated by the following operations : 1 .
any permutation of the @xmath93 symbols . 2 .
for @xmath132 , any permutation of the @xmath88 elements @xmath133 ( each pair @xmath134 goes together ) .
3 . for @xmath135 , any _ even _ permutation of the four elements @xmath136 .
4 . given @xmath137 replace each one of the @xmath93 symbols @xmath106 of the @xmath93-tuple , by @xmath138 ( the same @xmath97 for all @xmath93 symbols ) . 5 . given @xmath137 ,
replace _ one _ of the @xmath93 symbols @xmath106 by @xmath139 .
( it is of course the same @xmath97 multiplying @xmath111 and @xmath140 on the right within one symbol , @xmath97 being allowed to vary only between the @xmath93 symbols ) .
denote by @xmath141 the set of equivalence classes of @xmath130 under this equivalence relation .
note that because of the _ common _ left action in item ( 4 ) , @xmath141 is _ not _ simply the set of unordered @xmath93-tuples of elements of @xmath142 .
we now define @xmath143 to be the equivalence class of @xmath127 .
so @xmath144 is well defined , independent of all choices , including the choice of lift @xmath112 ( proposition dk(1 ) ) .
denote by @xmath145 the set of all pairs @xmath146 where @xmath103 and @xmath104 is a proper @xmath93-co - orientation for @xmath66 .
then @xmath147 and we claim : [ srj ] the map @xmath147 is surjective .
we are given an @xmath93-tuple @xmath17 of symbols of the form @xmath106 .
begin with any immersion @xmath148 and lift it to @xmath109 .
choose points @xmath149 as locations for the @xmath93 ces we will construct . deform @xmath112 by regular homotopy so that pieces of @xmath1 will pass the right points of @xmath3 above each @xmath150 , to produce the correct labelings appearing as superscripts in the symbols in @xmath17 .
continue the deformation near each such point so that for the projection back to @xmath2 the right configurations will be created with the right signs , and inducing the right ordering in case of ce of type @xmath83 . finally , move some other pieces of @xmath1 across the right points of @xmath151 to obtain the correct degree @xmath123}}$ ] .
the projection back to @xmath2 gives the desired immersion .
we also note that for types @xmath99 , the relation between the symbols obtained for a given co - orientation of a ce , and that obtained for the opposite co - orientation , is as follows : for @xmath78 , all remains the same except for the signs @xmath122 which are all reversed . for @xmath83 ,
all signs are reversed , and in addition an odd permutation is performed on @xmath136 .
( the degree @xmath140 remains unchanged since its definition does not involve the co - orientation . )
we call the symbol obtained from a symbol of type @xmath78 or @xmath83 in this way , the _ reversed _ symbol .
( recall that @xmath105 is only defined for _ proper _ @xmath104 and so for @xmath98 the co - orientation may not be reversed . )
we recall the definition of an ab equivalence , appearing in o : a regular homotopy between two immersions @xmath152 is called an ab equivalence if it is alternatingly of type a and b , where 1 .
@xmath153 ( @xmath154 ) is of type a if it is of the form @xmath155 where @xmath156 is an immersion and @xmath157 , @xmath158 are isotopies .
2 . @xmath153 ( @xmath154 ) is of type b if @xmath159 and there are little balls @xmath160 centered at the @xmath93 ce points of @xmath161 such that @xmath162 fixes @xmath163 and moves @xmath164 within @xmath165
. given two immersions @xmath152 and @xmath93-co - orientations @xmath166 for @xmath167 respectively , we now define _ co - oriented _ ab equivalence , or cab equivalence , between @xmath146 and @xmath168 to be an ab equivalence between @xmath66 and @xmath169 which respects @xmath166 , i.e. if we carry @xmath104 at each ce of @xmath66 continuously along the ab equivalence , then we arrive at @xmath169 with @xmath93-co - orientation @xmath170 .
we now prove : [ cab ] let @xmath171 and @xmath166 @xmath93-co - orientations for @xmath167 respectively
. then @xmath146 and @xmath168 are cab equivalent iff @xmath172 .
if @xmath146 and @xmath168 are cab equivalent , let @xmath162 ( @xmath173 ) be the given cab equivalence , and @xmath174 ( @xmath173 ) , the @xmath93-co - orientation carried continuously along ( so @xmath175 ) .
if a lift @xmath176 of @xmath177 is used for defining @xmath178 a each @xmath179 , and all choices of orderings are also carried along continuously , then by proposition dk(2 ) , @xmath178 may only change along the way by right multiplication .
we may still have that @xmath180 is different than the lift of @xmath169 used to define @xmath181 , in which case we will gain a common left multiplication .
the choices of ordering may also be different , but all together we get @xmath172 .
for the converse , assume @xmath172 .
since our allowed ordering of superscripts , and equivalence of ordering , all correspond to the possible symmetries of the configurations , there is an ambient isotopy @xmath182 which brings the ces of @xmath66 onto the ces of @xmath169 , such that the @xmath93-co - orientation @xmath104 carried continuously along , coincides with @xmath170 , such that the orientations of the corresponding sheets match , ( since the signs @xmath122 coincide , ) and such that all labelings in @xmath26 and degrees in @xmath59}}$ ] coincide up to right multiplication by an element in @xmath26 and common left multiplication by an element in @xmath26 .
since the orientations of corresponding sheets match , we may continue with a regular homotopy of the form @xmath183 for isotopy @xmath184 , until we have the same discs in @xmath1 participating in the ces of @xmath66 and @xmath169 , and the restriction of @xmath66 and @xmath169 to those discs coincides precisely , and we still have matching labeling and degrees up to right and common left multiplication . by proposition
dk any right multiplication can be realized by dragging a ce around some loop in @xmath2 and back to its place to match @xmath169 , and any common left multiplication can be realized by changing the choice of lift .
so we perform such ambient isotopies and change of lift until we have that all labelings and degrees coincide precisely .
this means that now , not only @xmath167 but also their chosen lifts @xmath185 coincide on the discs participating in the ces , and that if we slightly deform @xmath66 and @xmath169 by pushing each sheet of the ces slightly into its preferred side , then @xmath186 represent the same element in @xmath187 .
( note that @xmath188 , where the projections are induced by inclusion . )
form this point on we may proceed exactly as in the proof of o proposition 3.4 , where instead of working in @xmath0 with set of designated points @xmath189 , we work in @xmath3 with set of designated points @xmath151 .
we then compose the obtained regular homotopy with @xmath190 obtaining the desired cab equivalence in @xmath2 .
as already seen in o , a ce of type @xmath78 or @xmath83 may be cab equivalent to itself with the opposite co - orientation , which means that the stratum corresponding to this ce in @xmath67 is one sided . from proposition cab
we see that this happens iff the corresponding symbol is equivalent to its reversed symbol . for this to happen clearly
the number of @xmath191 and @xmath192 signs should be equal , as this number is preserved under equivalence .
so a one sided stratum may occur only for symbols of the form @xmath193 and @xmath194 .
if all @xmath111 are equal then we clearly do get a symbol equivalent to its reversed , and so a one sided stratum . in o where @xmath195 is trivial ( @xmath6 ) , it is thus clear which are the one sided strata .
but for general aspherical manifold this is intricately related to the structure of @xmath196 . in the following two examples we show a case where the stratum for @xmath193 is one sided though @xmath197 , and then a case of an @xmath2 where the stratum is one sided only when @xmath198 , though @xmath196 is nontrivial .
let @xmath199 be the connect sum of four projective planes , with @xmath200 the standard four loops for which @xmath201 .
let @xmath2 be the 3-manifold obtained from @xmath202 $ ] by attaching a 2-handle to the loop
@xmath203 in @xmath204
. then @xmath205 and one can show that @xmath2 is indeed aspherical .
we look at the symbol @xmath206 ( the subscript is the zero element of @xmath207 $ ] ) .
we note that @xmath208 , e.g. by looking at the quotient obtained by adding the relations @xmath209 .
the reversed symbol is @xmath210 .
left multiplication by @xmath15 and right multiplication by @xmath211 gives @xmath212 ( since @xmath213 ) , which is equivalent back to @xmath214 by permuting the superscripts .
and so the @xmath214 stratum is one sided . on the other hand , for @xmath215 , if @xmath17 denotes a generator of @xmath196 and say @xmath216 then @xmath217 is not equivalent to the reversed symbol @xmath218 since the property of having the @xmath191 sign attached to the smaller power of @xmath17 will be preserved under any left or right multiplication by element @xmath219 .
given an @xmath93-co - orientation @xmath104 for @xmath103 and a subset @xmath220 , we define @xmath221 to be the immersion obtained from @xmath66 by resolving all ces of @xmath66 at points of @xmath222 into the positive side with respect to @xmath104 , and all ces not in @xmath222 into the negative side .
now let @xmath223 be any abelian group and let @xmath224 be an invariant , i.e. a function which is constant on each connected component of @xmath94 .
given @xmath101 and an @xmath93-co - orientation @xmath104 for @xmath66 , @xmath225 is defined as follows : @xmath226 where @xmath227 is the number of elements in @xmath222 . if @xmath166 are two @xmath93-co - orientations for the same immersion @xmath66 then @xmath228 and so having @xmath229 is independent of the @xmath93-co - orientation @xmath104 .
an invariant @xmath224 is called _ of finite order _ if there is an @xmath93 such that @xmath229 for all @xmath230 .
the minimal such @xmath93 is called the _ order _ of @xmath231 .
the group of all invariants on @xmath94 of order at most @xmath93 is denoted @xmath232 .
. if @xmath152 and @xmath166 are @xmath93-co - orientations for @xmath167 respectively , such that @xmath146 and @xmath168 are cab equivalent , then @xmath234 , the argument being the same as in o proposition 3.8 .
and so we have by lemma srj and proposition cab , that each @xmath233 induces a well defined function @xmath235 .
the map @xmath236 induces an injection @xmath237 where @xmath238 is the group of all functions from @xmath141 to @xmath223 . if @xmath103 and @xmath166 are two @xmath93-co - orientations for @xmath66 which differ at precisely one ce , then for any @xmath233 , @xmath239 .
if the ce where we have reversed the co - orientation is of type @xmath78 or @xmath83 , and located at @xmath19 , then we have already noticed the relation between @xmath105 and @xmath240 , and so we now get an equation that must be satisfied by any @xmath241 in order for it to lie in the image of @xmath242 , namely , if @xmath243 are any @xmath244 symbols then for any symbol @xmath245 , @xmath246 must satisfy @xmath247 where @xmath248 denotes the sign opposite to @xmath122 .
we will write such an equation in the short form @xmath249 . for @xmath83 configuration
we similarly get @xmath250 ( note the transposition @xmath251 ) . also note that in the above two equations , the operation which produces the @xmath93-tuple of symbols on the right from the @xmath93-tuple of symbols on the left , is indeed well defined on the classes in @xmath141 .
let @xmath148 be an immersion with a self intersection of local codim 2 at @xmath20 and @xmath244 additional self - intersections of local codim 1 ( i.e. ces ) at @xmath252 .
we look at a 2-parameter family of immersions which moves @xmath1 only in a neighborhood of @xmath20 , such that the immersion @xmath66 corresponds to parameters @xmath253 and such that this 2-parameter family is transverse to the local codim 2 stratum at @xmath66 . in this 2-parameter family of immersions
we look at a loop which encircles the point of intersection with the codim 2 strata , i.e. a circle around the origin in the parameter plane .
this circle crosses the local codim 1 strata some @xmath190 times . between each two intersections we have an immersion in @xmath254 with the same @xmath244 ces , at @xmath252 . at each intersection with the local codim 1 strata ,
an @xmath93th ce is added , obtaining an immersion in @xmath255 .
let @xmath256 be the @xmath190 immersions in @xmath255 so obtained . for each @xmath257
, choose a proper @xmath93-co - orientation @xmath258 for @xmath259 , such that the co - orientation chosen for each @xmath260 , @xmath261 is the same in all @xmath262 .
let @xmath263 , @xmath264 be @xmath265 or @xmath266 according to whether we are passing the @xmath93th ce of @xmath259 in the direction of its co - orientation determined by @xmath258 , or in the opposite direction , respectively . for an invariant @xmath231 ,
it is easy to show that the following equation holds : @xmath267 . looking at @xmath237 we thus obtain additional equations that must be satisfied by a function in @xmath238 in order for it to lie in the image of @xmath242 .
we will now find all equations on @xmath238 obtained in this way .
as above , the equations will be written in short form as equations on the symbols .
we may assume ( by moving the codim 2 singular point @xmath20 if necessary ) , that there is a neighborhood @xmath268 of @xmath20 such that the labeling by elements of @xmath26 is locally constant in @xmath269 , and such that all motion involved in our 2-parameter family of immersions takes place inside @xmath268 .
and so we will be able to follow the labelings during our loop of immersions .
the local codim 2 strata may be divided into six types which we name after the types of ces appearing in a 2-parameter family of immersions , transverse to the given stratum : @xmath270 , @xmath271 , @xmath272 , @xmath273 , @xmath274 , @xmath275 . in the notation of hk
they are respectively : @xmath276 , @xmath277 , @xmath278 , @xmath279 , @xmath280 , @xmath281 .
formula and sketch for local model for such strata , the bifurcation diagrams , and the equations obtained , are as follows .
the symbol @xmath282 represents a string of @xmath134s of the appropriate length ( the same string for all appearances of @xmath282 within the same diagram or equation ) .
@xmath270 : @xmath76 , @xmath283 .
@xmath284 @xmath271 : @xmath76 , @xmath81 , @xmath285 .
@xmath286 @xmath272 : @xmath76 , @xmath84 , @xmath287 .
@xmath288 @xmath273 : @xmath76 , @xmath84 , @xmath289 .
@xmath290 @xmath274 : @xmath76 , @xmath81 , @xmath84 , @xmath291 .
@xmath292 @xmath275 : this configuration is a quintuple point , i.e. five sheets passing through a point , any three of which are in general position .
we model this by five planes passing through a point in @xmath0 .
a 2-parameter family may be constructed by fixing three of the planes , and allowing the other two planes to move parallel to themselves .
a loop in the shape of a rectangle around the origin of the parameter plane may be constructed as follows : the three fixed planes @xmath293 intersect in a triple point @xmath17 .
start with the plane @xmath294 very close to @xmath17 , and @xmath295 somewhat further away .
for the first edge of the rectangle , move @xmath294 across @xmath17 passing a first quadruple point . in this process
a little simplex @xmath296 has vanished and a new little simplex @xmath297 has been created from the same four planes @xmath298 . for the second edge of the rectangle ,
move @xmath295 across the whole of @xmath297 , passing four quadruple points on the way , when crossing the four triple points of @xmath297 . for the third edge move @xmath294 back to its original place , thus crossing @xmath17 again , in the opposite direction , recreating @xmath296 in place of @xmath297 , this being the sixth quadruple point .
finally move @xmath295 back to place , across @xmath296 , adding four more quadruple points , and so completing a total of ten quadruple points .
we co - orient each of these ten quadruple points according to the direction we are passing it along this loop , and so each of the ten terms in the equation that we are producing , will have a @xmath191 sign in front of it .
we will call the first and third edge of the rectangle described above `` short edges '' ( since they pass only one quadruple point ) and the second and fourth edges `` long edges '' .
we will first find the relation between the symbols of two consecutive quadruple points along a long edge , and then see that the same relation holds when passing to a short edge , and back .
so let @xmath299 be two vertices of the little simplex , which @xmath295 passes consecutively .
let @xmath300 be the edge of the simplex connecting @xmath299 .
by affine transformation of @xmath0 we may assume that @xmath301 , that the two planes intersecting to create the edge @xmath300 are the planes @xmath302 and @xmath303 , and that the plane @xmath295 which is in motion , crossing @xmath20 and then @xmath53 , is the plane @xmath304 where @xmath179 increases in the range @xmath305 for some small @xmath190 . for concreteness
we look at the time @xmath306 , and so @xmath307 . as to the remaining two planes
, one must pass @xmath20 , and the other must pass @xmath53 .
they can not be vertical planes ( since we already have two vertical planes ) , and so they are given by @xmath308 and @xmath309 .
the coefficients @xmath310 must both be nonzero since if say @xmath311 then we get three planes @xmath312 , @xmath313 , @xmath302 which are not in general position ( when parallelly moved to have a common point ) .
similarly @xmath314 are nonzero . by rotating the configuration around the @xmath315 axis we may assume @xmath316 .
we claim that we must then get also @xmath317 .
indeed , in order that @xmath295 will pass @xmath20 and @xmath53 consecutively , it must be that the other two vertices of the simplex lie outside the region @xmath318 of @xmath0 .
these two points are @xmath319 and @xmath320 .
the @xmath315 component of @xmath321 is @xmath322 , and this is outside the interval @xmath323 $ ] iff @xmath324 .
( if @xmath325 then there is no intersection , which implies that the three planes are not in general position . ) in the same way @xmath326 , using @xmath327 . and so the simplex @xmath328 created from the passage of @xmath295 across @xmath20 ( bounded by the planes @xmath329 ) , lies in the octant @xmath330 , whereas the simplex @xmath331 that is about to vanish when @xmath295 arrives at @xmath53 ( bounded by the planes @xmath332 ) , lies in the octant @xmath333 which is the exact opposite octant . and
so the sign @xmath122 with which each of the common planes @xmath334 appears in @xmath328 and @xmath331 is the opposite sign .
but since we have co - oriented all quadruple points according to our direction of motion , the simplex @xmath331 is not the one with which we determine the signs for the symbol @xmath335 of the quadruple point at @xmath53 , but rather the simplex @xmath336 which will be created _
after _ we cross @xmath53 .
as we have already noticed , the signs of the faces for @xmath336 are all opposite to the corresponding signs for @xmath331 , and so finally the common planes with @xmath328 will have the same sign as in @xmath328 .
we apply a similar two step argument to determine the relation between the orderings of the faces in the two symbols @xmath337 .
we note that if some ordering of the faces of @xmath328 is consistent with the orientation of @xmath328 ( restricted from @xmath0 ) then the ordering for the faces of @xmath331 obtained by simply replacing the plane @xmath308 with the plane @xmath309 is _
_ in__consistent with the orientation of @xmath331 . and so this same ordering of these same four planes , now considered as the faces of @xmath336 , _ is _ consistent with the orientation of @xmath336 . and
so for the superscripts of two consecutive symbols we finally get , that the @xmath134 corresponding to the plane which appears only in the first quadruple point and not the second ( the plane @xmath308 in our case ) should be dropped , and exactly in its place should be written the @xmath134 corresponding to the new plane which participates in the second quadruple point but not in the first ( the plane @xmath309 in our case ) .
the three @xmath134s corresponding to the three common planes ( the planes @xmath338 in our case ) must remain in their place , and with the signs @xmath122 unchanged .
we now show that when passing from the last quadruple point of a long edge to the quadruple point of the next short edge , or from that to the first quadruple point of the next long edge , then the same relation holds .
indeed , looking say at the first and second quadruple point in our rectangle described above , let @xmath339 be the triple point of @xmath297 which @xmath295 passes first along the long edge , for the occurrence of the second quadruple point .
let @xmath340 be one of the planes @xmath298 such that if @xmath340 is moved in the direction such that @xmath297 increases in size , then the point @xmath339 will move toward @xmath295 and eventually cross it .
so we can start from the same initial position as before , but instead of moving @xmath294 and then @xmath295 , we move only plane @xmath340 , such that @xmath296 will vanish , passing the first quadruple point , @xmath297 is created and increases in size until @xmath339 crosses @xmath295 thus passing the second quadruple point .
so we have obtained the same two quadruple points , where now four of the planes are fixed and only the plane @xmath340 is in motion . but
this is precisely the motion analyzed above to obtain the relation between two consecutive quadruple points along a long edge .
finally it remains to determine the degrees @xmath123}}$ ] of the ten symbols .
we do this again by analyzing the relation between the degrees of two consecutive quadruple points . following the intersection point between @xmath341 as it continuously moves up from @xmath20 to @xmath53 , we see that there is no change in the degree due to these three planes .
( recall that the labeling is locally constant above the region of interest . )
there may , on the other hand , be a change due to the planes @xmath342 and @xmath343 .
if the preferred side of @xmath344 is facing upward , which happens iff the corresponding sign @xmath345 appearing in the symbol @xmath346 for the first quadruple point is @xmath192 , then when computing the degree @xmath347 for the first quadruple point , @xmath344 is slight pushed upward so as to pass above the point @xmath20 . but as @xmath295 moves upward to @xmath53 , @xmath344 does not participate in the second quadruple point , and remains below the point @xmath53 which is where the degree is now computed . so in this case there is a change of @xmath348 to the degree due to the plane @xmath344 .
if on the other hand the plane @xmath344 is facing downward , which happens iff the sign @xmath345 appearing in @xmath346 is @xmath191 , then the plane @xmath344 is pushed _ downward _ when @xmath347 is computed , and again it will remain below the point @xmath53 when the degree of the second quadruple is computed , and so there will be no change in the degree due to @xmath344 .
the two cases may be written in one formula as follows . for sign @xmath349 ,
define @xmath350 then the change in degree between the first and second quadruple point , due to the plane @xmath344 is @xmath351 . a similar analysis as to the effect of the plane @xmath352 on the degree
will give @xmath353 .
together we get that if @xmath352 is the plane missing from the first quadruple point , and @xmath344 appears there with sign @xmath345 , and @xmath344 is missing from the second quadruple point , and @xmath352 appears there with sign @xmath122 , then the change in the degree between the first and second quadruple point is @xmath354 .
looking at the circle ( or rectangle ) of ten quadruple points , we see that the same 4-tuple of planes participates in a quadruple point twice , and with opposite co - orientation , and so the ordering of the corresponding @xmath355 at the two times differ by an odd permutation .
they occur at the precise opposite timing along the circle , i.e. five places apart , and so the ordering of quadruple points along the circle is that all five possible 4-tuples appear one after the other , and then they appear again in the same order , to complete the cycle of ten quadruple points .
so given a quadruple point , either it or the matching quadruple point on the opposite side of the circle , have the property , that the ordering of superscripts may be chosen as @xmath136 so that in the next quadruple point the plane corresponding to @xmath356 will be missing , in the following quadruple point the plane corresponding to @xmath357 will be missing , and so on . finally , since the two matching quadruple points ( i.e. those with the same 4-tuple of planes ) occur with opposite co - orientation , the sign attached to each @xmath358 will be opposite ( i.e. @xmath355 is replaced by @xmath359 ) .
since we have seen that if a plane is common to two consecutive quadruple points , it appears in their superscripts with the same sign , and since a given plane will be missing exactly once between its appearance in a given quadruple point , and the occurrence of the matching quadruple point , it follows that whenever a plane is missing from a quadruple point , then it appears in the following quadruple point with opposite sign than it does in the previous one . to write our final formula , let us choose the first quadruple point and ordering @xmath136 as mentioned above , and
we may write its degree as @xmath360 where @xmath361 corresponds to the plane missing from the first quadruple point , and @xmath362 is the sign with which it will appear in the second quadruple point .
the first symbol is thus @xmath363 , and the combination of all analysis above produces the following equation : @xmath364 now let @xmath365 be the subgroup consisting of all functions @xmath241 which satisfy the above equations ( 1 ) ( 6 ) together with our first two equations @xmath249 and @xmath250 .
then we have obtained that the image of the injection @xmath366 is contained in @xmath367 .
finding the precise image of @xmath242 for all @xmath93 ( as has been done for the case @xmath6 in o , h ) , would give a full classification of all finite order invariants .
in the next section we will show that the image of @xmath368 is all @xmath369 , by this classifying all order one invariants .
we define a `` universal '' abelian group @xmath370 , by the abelian group presentation which takes as generators the elements of @xmath142 , and as relations , the set of all equations used above to define @xmath367 ( and in particular @xmath369 ) .
note that for defining @xmath369 , an expression such as @xmath249 was merely a short form for writing @xmath371 , this being a condition on an element @xmath372 for being included in the subgroup @xmath369 , whereas , in the definition of @xmath370 , @xmath249 is an actual relation in the presentation of @xmath370 by generators and relations .
we define the universal element @xmath373 , to be the function that assigns to each element in @xmath142 , the generator corresponding to it in @xmath370 . by definition of @xmath370 , indeed @xmath373 .
we will now establish the existence of an order 1 invariant @xmath374 satisfying @xmath375 .
this will prove our desired result , that for any abelian group @xmath223 , @xmath376 is surjective .
indeed , given @xmath377 , there is a ( unique ) homomorphism @xmath378 such that @xmath379 , and we get @xmath380 .
let @xmath148 be some immersion and @xmath383 a null - homotopy from @xmath66 to a constant map @xmath384 with @xmath385 .
using the parallelizibility of @xmath2 we extend @xmath78 to a homotopy of bundle monomorphisms , from @xmath386 to a bundle monomorphism @xmath31 which covers @xmath387 . by the smale - hirsch theorem applied to @xmath388 , there is an immersion @xmath389 with @xmath390 homotopic to @xmath31 as bundle monomorphisms .
now by the smale - hirsch theorem applied to @xmath391 , @xmath66 is regularly homotopic to @xmath392 , and so @xmath381 .
fix an immersion @xmath393 with image contained in a small ball @xmath382 , as provided by lemma b. for any @xmath394 take a generic regular homotopy @xmath162 from @xmath392 to @xmath66 , and to each ce which occurs along @xmath162 choose a proper co - orientation .
let @xmath395 be the sum of the elements in @xmath396 corresponding to the co - oriented ces occurring along @xmath162 , each taken with a sign @xmath397 according to whether @xmath162 passes it in the direction of its co - orientation .
an opposite choice of co - orientation for a ce along @xmath162 ( relevant for types @xmath99 ) , will produce the negative element in @xmath370 , but this element will appear with opposite sign in the sum and so this sum is well defined , for given @xmath162 .
we would like to show that in fact @xmath398 is independent of the regular homotopy @xmath162 , and so @xmath398 is a well defined invariant .
this is equivalent to showing that this sum is 0 along any closed @xmath162 , i.e. @xmath162 such that @xmath399 .
once we know @xmath398 is well defined , it is clear that it is of order 1 , and @xmath375 .
the fact that the sum of values is 0 along any null - homotopic loop follows from the definition of @xmath369 and @xmath370 as demonstrated in o , and so the sum along loops induces a well defined homomorphism @xmath400
. we must show @xmath401 .
fix a disc @xmath402 , and let @xmath403 be given by restriction , @xmath404 . then @xmath199 is a fibration with fiber @xmath405 .
so we get an exact sequence @xmath406 let @xmath407 be the subgroup consisting of loops obtained by composing @xmath408 with a motion of @xmath409 in @xmath2 through embeddings , beginning and ending with the inclusion of @xmath409 in @xmath2 .
it is clear that @xmath410 maps @xmath411 onto @xmath412 , and so the elements of @xmath411 and those coming from @xmath413 generate @xmath414 .
now @xmath415 is 0 on elements of @xmath411 , since no ces occur along such loops , and so it remains to show that @xmath415 is 0 on @xmath413 .
it is shown in the proof of q theorem 3.4 , that @xmath413 is generated by a loop which only moves a small disc @xmath416 which is disjoint from @xmath45 , and @xmath268 moves only within a small ball @xmath417 . in our case
we may take @xmath418 , and so we finally need only to check the value of @xmath415 on loops moving @xmath1 in @xmath409 .
if we take our labeling to be locally constant above @xmath409 , then we can have all lifts used to define our symbols to be contained in the same lift of @xmath409 , constantly labeled by some fixed element @xmath137 . and
so all symbols will be of the form @xmath419 . by a permutation of the superscript , which is even in case @xmath87
is @xmath83 , we may assume that all @xmath191 signs appear first , and so the superscript is characterized only by the _ number _ of @xmath191 signs , which is precisely the way the superscript of a symbol is defined in o for immersions @xmath420 .
the subscript is characterized by the coefficient @xmath421 of @xmath422 which is also simply the degree as defined in o for immersions @xmath423 . and
so @xmath415 being 0 on all loops moving @xmath1 in @xmath409 , follows from the fact that @xmath398 appearing in o for the case @xmath6 , is well defined .
indeed , for @xmath398 of o , the sum of symbols along a loop of immersions in @xmath409 , is 0 in the group @xmath424 of o. we have noticed that our symbols of the form @xmath419 correspond to the symbols of o , but furthermore , any equation on the symbols appearing in o , has a corresponding equation on our symbols of the form @xmath419 , by considering the same codim 2 configuration in @xmath409 . and
so the sum in our present @xmath424 must also be 0 .
this completes the proof of : let @xmath1 be a closed orientable surface , @xmath2 an aspherical orientable 3-manifold , and @xmath425 a regular homotopy class of null - homotopic immersions of @xmath1 into @xmath2 .
then for any abelian group @xmath223 , the injection @xmath426 is surjective .
we conclude with the following remark .
our classification is in terms of the group @xmath427 which as we have seen is the same as the group @xmath428 . in o
where @xmath6 and so @xmath429 , the structure of @xmath370 is completely understood and accordingly the group of order 1 invariants . in the general case presented here , the structure of @xmath370 depends on the structure of @xmath196 , most notably through the equivalence relation which produces @xmath142 from @xmath430 .
the complexity of this dependence may be seen in the discussion following proposition cab , where the question is whether a given symbol is equivalent to its reversed symbol . and
so a closed explicit classification as in o ( not to mention explicit formulae for the invariants , as in f ) , will require extensive analysis of @xmath196 for any given @xmath2 . | we study finite order invariants of null - homotopic immersions of a closed orientable surface into an aspherical orientable 3-manifold .
we give the foundational constructions , and classify all order one invariants . [ section ] [ thm]lemma [ thm]proposition [ thm]corollary [ thm]definition [ thm]remark |
following the pioneering discovery of the feigenbaum - coullet - tresser period doubling universality in unimodal maps @xcite , @xcite , @xcite universality has been demonstrated to be a rather generic phenomenon in dynamics .
such renormalization approach to universality has been very successful in one - dimensional dynamics , and has led to explanation of universality in unimodal maps @xcite , @xcite,@xcite , critical circle maps @xcite , @xcite , @xcite and holomorphic maps with a siegel disks @xcite , @xcite , @xcite . universality has been abundantly observed in higher dimensions , in particular , in two and more dimensional dissipative systems ( cf .
@xcite , @xcite ) , in area - preserving maps , both as the period - doubling universality @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , and as the universality associated with the break - up of invariant surfaces @xcite ,
@xcite , @xcite , @xcite , and in hamiltonian flows @xcite,@xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite .
it has been established that the universal behavior in dissipative and conservative higher dimensional systems is fundamentally different .
the case of of the dissipative systems is often reducible to the one - dimensional feigenbaum - coullet - tresser universality ( @xcite , @xcite ) .
the latter case is very different , and at present there is no deep understanding of universality in conservative systems , other than in the `` trivial '' case of the universality for systems `` near integrability '' @xcite , @xcite , @xcite , @xcite , @xcite .
the study of the interesting cases of universality for maps and flows far from linear , is at present confined to numerics ( for instance , @xcite,@xcite , @xcite ) ) , or computer - assisted proofs @xcite , @xcite , @xcite .
the latter approach turned to be quite powerful , albeit rather specialized and time and effort consuming .
very little analytic machinery exists for higher - dimensional maps and flows far from linear .
an infinite period - doubling cascade in families of area - preserving maps was observed by several authors in early 80 s @xcite , @xcite , @xcite , @xcite , @xcite .
a typical period - doubling scenario can be illustrated with the area - preserving h ' enon family ( cf .
@xcite ) : @xmath5 maps in this family posses a fixed point @xmath6 which is stable for @xmath7 .
when @xmath8 this fixed point becomes unstable , at the same time an orbit of period two is born with @xmath9 , @xmath10 .
this orbit , in turn , becomes unstable at @xmath11 , giving birth to a period @xmath12 stable orbit .
generally , there exists a sequence of parameter values @xmath13 , at which the orbit of period @xmath14 turns unstable , while at the same time a stable orbit of period @xmath15 is born .
the parameter values @xmath13 accumulate on some @xmath16 .
the crucial observation is that the accumulation rate @xmath17 is universal for a large class of families , not necessarily hnon .
d. f. escande , f. doveil , renormalization method for computing the threshold of the large scale stochastic instability in two degree of freedom hamiltonian systems , _
* 26 * ( 1981 ) 257284 .
m. j. feigenbaum , quantitative universality for a class of nonlinear transformations , _
_ * 19 * ( 1978 ) 2552 .
m. j. feigenbaum , universal metric properties of non - linear transformations , _
* 21 * ( 1979 ) 669706 .
h. koch , on the renormalization of hamiltonian flows , and critical invariant tori , _ discrete contin .
* 8 * ( 2002 ) , 633646
. h. koch , a renormalization group fixed point associated with the breakup of golden invariant tori , _ discrete contin .
* 11 * ( 2004 ) , no .
4 , 881909 .
h. koch , _ existence of critical invariant tori _
, preprint mp_arc 04210 ( 2004 ) s. koci , renormalization of hamiltonians for diophantine frequency vectors and kam tori,_nonlinearity _ * 18 * ( 2005 ) 25132544 .
r. s. mackay , renormalization approach to invariant circles in area - preserving maps , _ physica _ * d7 * ( 1983 ) 283300 .
r. s. mackay , renormalisation in area preserving maps , thesis , princeton ( 1982 ) .
world scientific , london ( 1993 ) .
a. mehr and d.f .
escande , _ physica _ * d13 * ( 1984 ) 302 .
d. sullivan , bounds , quadratic differentials and renormalization conjectures , in : mathematics into the twenty - first century , ams centennial publications , vol .
ii , amer . math .
soc . , providence , r.i .
( 1992 ) 417 - 466 . | it has been observed that the famous feigenbaum - coullet - tresser period doubling universality has a counterpart for area - preserving maps of @xmath0 .
a renormalization approach has been used in a computer - assisted proof of existence of an area - preserving map with orbits of all binary periods in @xcite and @xcite . as it is the case with all non - trivial universality problems in non - dissipative systems in dimensions more than one , no analytic proof of this period doubling universality exists to date .
we argue that the period doubling renormalization fixed point for area - preserving maps is almost one dimensional , in the sense that it is close to the following hnon - like map : @xmath1 where @xmath2 solves @xmath3 we then give a `` proof '' of existence of solutions of small analytic perturbations of this one dimensional problem , and describe some of the properties of this solution .
the `` proof '' consists of an analytic argument for factorized inverse branches of @xmath2 together with verification of several inequalities and inclusions of subsets of @xmath4 numerically . finally , we suggest an analytic approach to the full period doubling problem for area - preserving maps based on its proximity to the one dimensional . in this respect ,
the paper is an exploration of a possible analytic machinery for a non - trivial renormalization problem in a conservative two - dimensional system . |
qcd is usually quantized perturbatively .
the lagrangean @xmath1 is split into the sum of two terms @xmath2 where @xmath3 describes free quarks and gluons , and @xmath4 their interaction .
then @xmath3 is quantized : the hilbert space is the fock space of free quarks and gluons and the interaction describes scattering between them .
any observable @xmath5 is computed as a power series expansion in the renormalized coupling constant @xmath6 @xmath7 where @xmath8 are finite amplitudes .
the expansion eq(2 ) is not convergent , not even as an asymptotic series@xcite .
nevertheless it is an empirical fact that , as @xmath9 or for @xmath10 @xmath11 and a few terms of the expansion describe physics correctly .
the lack of convergence of the expansion eq(2 ) reflects the instability of fock vacuum .
quarks and gluons are confined and therefore free quarks and gluons are not a good zeroth order approximation to physics .
a non perturbative quantization of the theory is needed to identify the true vacuum state .
this is provided by the feynman path integral formulation .
the key quantity is the partition function @xmath12 , in terms of which all field correlators can be computed .
@xmath13[d\psi][d\bar\psi]\exp{-\left[s(a_\mu,\bar\psi,\psi)\right]}\ ] ] @xmath12 in eq(4 ) is a functional integral and is defined by approximating euclidean space time by a discrete set of points in a finite volume , so that the number of integration variables is finite and the integral is an ordinary well defined integral . a sequence of such integrals is then computed on sets of points which tend to cover the volume densely .
the limit of infinite volume is then performed .
if @xmath12 is finite and well defined after these limiting procedures one says that the theory exsists as a field theory .
the procedure also ensures the existence of the analytic continuation to minkowskian space time . for qcd
asymptotic freedom insures that the first of these limits exists .
the existence of a mass gap in the theory also insures that the infinite volume limit exists .
a fully rigorous mathematical proof of these statements still does not exist , but it is physically reasonable to say that qcd exists as a field theory .
lattice formulation of qcd is nothing but an approximant of feynman integral in the sequence which defines it : at sufficiently low value of the unrenormalized coupling and at sufficiently large volume it will provide the physical amplitudes with any required precision from first principles ( modulo technical computational difficulties ) .
non perturbative phenomenology is based on wilson s operator product expansion ( ope ) , @xmath14 where @xmath15 s describe short distances and are usually computed in perturbation theory and the operators @xmath8 describe large distances.the sum in eq(4 ) is ordered by incresing order of the dimensions in mass of the local operators @xmath8 .
vacuum expectation values ( vev ) of eq(4 ) involve in the rhs vev s of composite operators ( condensates ) .
they enter in the svz sum rules@xcite .
the gluon condensate @xmath16 , and the quark condensate @xmath17 are examples of them . a condensate like @xmath18 has dimension 4 in mass , and therefore , by renormalization group arguments is related to the running coupling constant at the scale @xmath19 as @xmath20 which is non analytic in g and undefined in perturbation theory .
matrix elements of eq(4 ) between quark states are relevant to weak interaction matrix elements at distances @xmath21 , or to structure functions of deep inelastic scattering . in this talk
i will concentrate on the vacuum state , namely on condensates ( sect 2 ) , topology ( sect3 ) , and confinement of color ( sect4 )
the ope of the product of two conserved currents @xmath22 taken on the vacuum reads @xmath23 where i is the identity operator ( @xmath24),and at short distances @xmath25 modulo logs,@xmath26 and @xmath27 @xmath28 modulo logs .
eq(6 ) is a theorem in perturbation theory , an assumption in the presence of non perturbative effects .
the coefficients @xmath29 describe short distances , the condensates describe large distance physics .
after fourier transform @xmath30 with @xmath31 a dispersive representation for the l.h.s . of eq(7 ) , if @xmath32 is the electromagnetic current is @xmath33 where @xmath34 can be taken from experiment .
an appropriate average on @xmath35 of both members of eq(10 ) gives the svz@xcite sum rules , in which non perturbative effects are parametrized in terms of the condensates @xmath36 and @xmath37 .
the result is a good phenomenology and a determination of the condensates @xmath38 however the perturbative expansion of @xmath39 which provides the coefficient @xmath40 , when resummed at higher orders , is ambiguous by terms @xmath41 @xcite,@xcite which mimic the terms with @xmath18 and @xmath42 .
the definition of the condensates is then intrinsically ambiguous .
they could be defined as the coefficients of the terms @xmath43 in the expansion eq(7 ) , but a priori , due to the ambiguity , they could be process dependent , i.e. dependent on the currents under consideration . an ope of the levels of bound @xmath44 energy levels @xcite@xcite can be also performed .
this amounts to compute the quadratic stark effect produced on the levels by the fluctuating vacuum chromoelectric field .
if the finite correlation length of the field strength correlators is taken into account@xcite@xcite the relevant quantity is the gauge invariant correlator@xcite @xmath45 with @xmath46 the parallel transport from 0 to @xmath47 along the path c , or @xmath48 .
a general parametrization of such correlators is@xcite@xcite @xmath49|0\rangle=\nonumber\\ & & \hskip-20pt(g_{\mu\rho}g_{\nu\sigma } - g_{\mu\sigma}g_{\nu\rho})[d(x^2)+d_1(x^2 ) ] + \\ & & \hskip-20pt+\left[x_\mu x_\rho g_{\nu\sigma}-x_\mu x_\sigma g_{\nu\rho}+x_\nu x_\sigma
g_{\mu\rho}-x_\nu x_\rho g_{\mu\sigma}\right ] \frac{\partial d_1(x^2)}{\partial x^2}\nonumber\end{aligned}\ ] ] in this language @xmath50 .
the correlators @xmath51 can be computed on the lattice @xcite@xcite and with them their ope . .
the line is the best fit to eq(15).,title="fig : " ] -0.5 in .
the line is the best fit to eq(16).,title="fig : " ] -0.5 in their typical behaviour is shown in figs 1 and 2 .
a convenient parametrization of the lattice data @xmath52 is @xcite@xcite @xmath53 @xmath54 is a split point regulator of @xmath36 .
a best square fit to the lattice data gives , for pure gauge su(3)@xcite ( quenched , no quarks ) @xmath55 .
in full qcd @xcite the same quantities can be computed .
an extrapolation to realistic quark masses gives @xmath56 and @xmath57 .
the phenomenological value for @xmath36 is @xmath58@xcite from quenched to full qcd the gluon condensate decreases by almost an order of magnitude , and the correlation length increases by 50% .
lattice vacuum correlators are the input of the stochastic approach to qcd @xcite@xcite . in a similar way the fermion correlators can be computed@xcite @xmath59 @xmath60
can be viewed as a split point regulator of @xmath61 .
lattice data @xcite give @xmath62 consistent with the value obtained by use of svz sum rules and a correlation length @xmath63
the @xmath0 problem was a problem of the free quark model of gellmann .
the singlet axial current is conserved in the model , but in nature the corresponding symmetry is neither realized a la wigner ( no parity doublets exist in the hadron spectrum ) , nor a la goldstone : indeed a goldstone - broken symmetry would imply @xcite @xmath64 , which is badly violated by the observed value @xmath65 . hence the idea of gellmann that the symmetry of hadrons could be abstracted from the free quark model was not correct for the axial singlet .
qcd solves the puzzle .
the singlet axial current is anomalous in qcd and @xmath0 is not a symmetry , being broken at the quantum level .
@xmath66 @xmath67 ( @xmath68 ) is named topological charge density . on continuous configurations the topological charge
@xmath69 is an integer ( second chern number ) . by use of different lattice regulators and of different values of the unrenormalized coupling constant .
the physical value of @xmath70 is independent of these choices within statistical errors . ] for quantized fields @xmath71 is an operator .
the topological susceptibility @xmath70 is defined as the response of the vacuum to q(x ) @xmath72 an argument based on @xmath73 shows that @xmath74 is related to the topological susceptibility of the @xmath75 vacuum.@xcite @xmath76 as @xmath73 dynamical quarks can be neglected and with the same approximation @xmath77 can be taken equal to the physical value 3 , so that @xmath70 is the topological susceptibility of the quenched su(3 ) vacuum
. an improved version of the argument including mixing to the octet gives@xcite @xmath78 the prediction is in both cases @xmath79 .
computing @xmath70 from first principles can provide a cross check of qcd and of the 1/@xmath77 expansion at the same time .
this problem has a long story : tha main difficulty has been for a long time to understand how topology , which is based on continuity , can exist on the lattice which has a discrete structure .
the way out is basic field theory@xcite : any lattice version of the operator @xmath71 , @xmath80 , is related to the continuum counterpart , say in the @xmath81 scheme , by a multiplicative renormalization @xmath82 the product q(x)q(0 ) is singular as @xmath83.the ope gives @xmath84 in computing @xmath70 an additive renormalization is also required.@xcite in conclusion@xcite @xmath85 where @xmath86 is the lattice topological susceptibility of the @xmath87 sector of the vacuum .
a technique developed to compute @xmath88 and @xmath86 is known as `` heating''@xcite : it is based on the fact that topology has a much longer autocorrelation time @xmath89 in montecarlo upgrading than the autocorrelation time @xmath90 of the local quantum fluctuations which dominate the renormalization constants .
therefore @xmath91 can be measured on the configurations obtained at times t such that @xmath92 from a zero field configuration , which has topological charge zero .
similarily @xmath93 can be measured on configurations prepared from an initial configuration consisting of a single instanton , which has q=1 . at times
long enough with respect to @xmath90 but very short compared to @xmath89 @xmath93 will be equal to z .
this method is known as `` field - theoretical '' .
the result of this procedure is @xmath94@xcite recently an independent determination has been made , using the divergence of the singlet axial current operator , i.e .
the lhs of eq ( 17 ) .
the advantage is in the use of a formulation of the fermions on the lattice which preserves chiral symmetry , and hence has multiplicative renormalization @xmath95 the result is@xcite @xmath96 in complete agreement with the field theoretical determination .
also determinations based on cooling techniques @xcite give consistent estimates . in conclusion
the @xmath0 problem is solved in qcd , and additional legitimation is added to the ideas of @xmath73 . an important issue , as we shall see in the next section , is the behaviour of @xmath70 at the deconfining phase transition .
quarks and gluons are visible at short distances , but have never been observed as free particles .
this fact has lead to the conjecture that colored particles never appear in asymptotic states , a property known as confinement of color .
confinement should be derivable from the qcd lagrangean .
quarks have been searched in nature since they were introduced by gellman as fundamental constituents of hadrons@xcite , with negative results : only upper limits to their abundance and to their production rate have been established@xcite . the ratio of quark abundance in nature , @xmath97 to that of protons @xmath98 has been given the upper limit @xmath99 to be compared to the expectation based on the standard cosmological model @xmath100 .
as for production in particle reactions the best limit is provided by the inclusive cross section @xmath101 namely @xmath102 to be compared with the expectation , in the absence of confinement , @xmath103 in both cases the ratio is depressed by a factor @xmath104 .
the only natural explanation is that these ratios are exactly zero , i.e. that confinement is an absolute property , and as such based on some symmetry .
experiments have been and are being performed to detect a deconfining transition at high energy density from hadronic matter to a state of free quarks and gluons ( quark gluon plasma ) , in high energy collisions of heavy ions@xcite .
the major difficulty with them is to identify an observable quantity by which deconfinement could be detected .
the partition function of a field theory at temperature t is the euclidean feynman path integral extending in euclidean time from @xmath105 to @xmath106 , with periodic boundary conditions for boson fields , antiperiodic for fermions . for qcd this
is computed on a lattice of spatial extension @xmath107 and temporal extension @xmath108 with @xmath109 , and the temperature is @xmath110 with @xmath111the lattice spacing in physical units , which can be tuned by varying the unrenormalized coupling constant @xmath112 . in quenched formulation ( pure gauge , no dynamical quarks )
the criterion consists in looking at the large distance behaviour of the static @xmath113 potential , @xmath114 .
@xmath115 is related to the correlator of polyakov lines @xmath116
@xmath117 by the relation @xmath118 where @xmath119 is the extension of the lattice in the time direction . by definition
the polyakov line is the trace of the parallel transport across the lattice in the time direction . @xmath120\ ] ] by cluster property @xmath121 if @xmath122 it follows @xmath123 if @xmath124 @xmath125 a transition is observed on the lattice at @xmath126 from a region @xmath127 where @xmath128=0 to a region @xmath129 where @xmath124 .
a finite size scaling analysis allows to determine the order of the transition , which for su(3 ) is first@xcite@xcite .
@xmath128 can be then assumed as an order parameter for confinement , and the corresponding symmetry is @xmath130 in the presence of dynamical quarks ( full qcd ) @xmath130 is not a symmetry anymore and @xmath128 is not an order parameter .
string breaking is expected to take place : when pulling the static @xmath113 pair apart from each other energy is converted into light @xmath132 pairs , and the potential is not linear any more , even if there is confinement . at @xmath133 another symmetry is present , chiral symmetry , which is spontaneously broken at @xmath134 , and , as lattice data show , is restored at some @xmath135 .
chiral symmetry , however , is explicitely broken by quark masses , and can not be the symmetry responsible for confinement .
a phase dyagram is usually drawn for this system as in fig(4 ) . the line which joins the first order phase transition at @xmath138 ( quenched ) to the chiral transition at @xmath139 corresponds to the values @xmath140 at which the susceptibilities @xmath141 ( specific heat ) and @xmath142 are maximum@xcite@xcite .
an analysis can be made of the chiral phase transition at @xmath139 , assuming that the pions and sigmas are the relevant degrees of freedom@xcite .
chiral symmetry fixes then the form of the free energy and renormalization group arguments allow to predict the order of the transition and the corresponding universality class .
for @xmath143 the transition is first order .
for @xmath144 the order depends on the relative weight of the chiral o(4 ) part of the action and the term describing the @xmath0 anomaly .
if the anomaly disappears below @xmath135 the transition is first order , and such is the transition along the line at @xmath145 .
if instead the anomaly persists up to @xmath135 the transition is second order the universality class is @xmath146 and the line at @xmath145 is a crossover . a reliable numerical analysis of this issue is not yet available , but for some reason the second option is more popular
. the order and the universality class can be investigated by studying the behaviour of the critical specific heat as the volume goes to infinity by a finite size scaling analysis .
a good order parameter should have a behaviour in that limit consistent with that of @xmath141 .the order of the transition and the universality class should be identified by the symmetry responsible for confinement .
a valid candidate symmetry is dual superconductivity of the vacuum@xcite@xcite@xcite .
the basic idea is that the chromoelectric field acting between colored particles is channeled into dual abrikosov flux tubes , whose energy is proportional to the length .
here dual means interchange of electric with magnetic with respect to ordinary superconductors .
a disorder parameter @xmath147 has been defined which detects dual superconductivity : it is the vev of an operator @xmath19 carrying magnetic charge which has zero vev in a phase in which vacuum has a definite magnetic charge , and can have a non zero vev if monopoles condense i.e. the vacuum is a superposition of states with different magnetic charge . in principle @xmath148 or@xmath149 i.e. superconductivity or not depends on the procedure used to identify magnetic charges ( abelian projection ) .
however it can be shown that monopole condensation ( or non condensation ) is an abelian projection independent statement@xcite@xcite . for quenched theory the result of numericall simulatios
is that this disorder parameter is consistent with @xmath150@xcite@xcite : it is non zero for @xmath127 and strictly zero for @xmath129 . by finite size
scaling analysis of the corresponding susceptibility @xmath151 the same critical indices are obtained as with @xmath152 .
the operator @xmath19 is perfectly well defined also in the presence of dynamical quarks , and can be used to investigate the phase diagram of fig(4 ) .
the result is that the region below the line is really confined , the one above it is deconfined@xcite . @xmath153 as a function of @xmath154 at fixed @xmath155 has a peak at the same value as the other susceptibilities , i.e. on the line of fig(4 ) . @xmath147
thus provides a good criterion for confinement . the finite size scaling analysis works as follows . a priori @xmath156 depends on the coupling constant @xmath157 i.e. on the lattice spacing @xmath158 , on the mass @xmath155 and on the lattice size @xmath159
.the dependence on @xmath157 can be traded with the dependence on the correlation length @xmath160 so that , for dimensional reasons @xmath161 as @xmath154 approaches @xmath135 from below , or @xmath162 , @xmath160 diverges as @xmath163 @xmath164 and @xmath165 are critical indices .
if the transition is second order or weak first order @xmath160 goes large as @xmath135 is approached and @xmath166 can be put equal to zero .
because of eq(37 ) @xmath167 can be traded with @xmath168 .
the scaling law follows @xmath169 in the quenched case the second dependence does not exist.the scaling can be tested on the lattice data and the critical index @xmath164 can be extracted@xcite@xcite .
in particular for the peak one finds @xmath170 for the case of full qcd the problem has two scales and the analysis is more complicated .
preliminary results@xcite exclude the critical indices of @xmath146 and seem consistent with a first order phase transition . if this will be confirmed by the additional data in preparation either the anomaly disappears below @xmath135 , or the relevant degrees of freedom which dominate are not the pion and the sigma@xcite .
the first possibility can be checked on the lattice .
the dependence of the topological susceptibility at the deconfining transition has been [email protected](5 ) shows the existing data .
a more precise analysis is needed to settle the problem . in conclusion
there is strong evidence that dual superconductivity is the mechanism of confinement . a candidate disorder parameter exists , which is well defined and seems to work .
the corresponding symmetry is dual superconductivity , which is however not clearly understood formally , being a common feature of all the abelian projections : consequently the effective action is not known .
a.h.mueller , nucl .
phys.b250 , 327(1985 ) m.a.shifman , a.i.veinshtein , v.i .
zakharov , nucl .
b147 , 385 , 448 , 519 ( 1979 ) h.g.dosch in @xmath173(s.i.f .
varenna 1995 , a.di giacomo and p.dyakonov eds ) h.leutwyler , phys.lett.b98 , 447 , 1981 m.b.voloshin , nuclphys.b154 , 365 , 1979 d.gromes , phys.lett.b 116 , 179 , 1972 m.campostrini , a.di giacomo , s.olejnik,zeit .
phys.c31 , 577 , 1986 m.campostrini,a.di giacomo , g.mussardo , zeit.phys.c25 , 173 , 1984 h.g.dosch , phys.lett.b190 , 177 , 1987 h.g.dosch,y.simonov , phys.lett.b205 , 339 , 1988 a.di giacomo , h.panagopoulos , phys.lett.b285 , 133 , 1992 a.di giacomo , e.meggiolaro , h.panagopoulos,nucl.phys.b483 , 371 , 1997 s.narison , phys.lett.b387 , 162 , 1996 a.di giacomo , h.g.dosch , v.i.shevchenko,yu.a.simonovphys.repts.372,319,2002 m.delia , a.di giacomo , e.meggiolaro , phys.lettb , 408 , 315 , 1997 m.delia , a.di giacomo , e.meggiolaro , phys.rev .
d59 , 054503 , 1999 s.weinberg , phys.rev.d11 , 3583 , 1975 e.witten , nucl.phys.b156 , 269 , 1979 g.veneziano , nucl.phys.b159 , 213 , 1979 m.campostrini,a.di giacomo , h.panagopoulos , phys.lett b212 , 206 , 1988 p.di vecchia , k.fabricius , gc.rossi , g.veneziano , phys.lett.b108 , 323 , 1982 m. campostrini , a. di giacomo , h. panagopoulos , e. vicari , nucl .
b329 , 683 , 1990 a.di giacomo , e.vicari,phys.lett.b275 , 429 , 1992 b.alles , m.delia , a.di giacomo , nucl.phys.b434 , 481 , 1997 l.giusti,gc.rossi,m.testa,g.veneziano , nucl.phys.b628 , 234 , 2002 m.teper , phys.lett.b171 , 86 , 1986 m.gellmann , phys.rev . 125 , 1067 , 1962 review of particle properties , e.p.j.15 , 2000 see e.g.the proceedings of quark matter 2002 , north holland , h.gutbrod , j.aichelin , and k.werner eds m. fukugita , m. okawa , a. ukawa , phys .
63 , 1768 , 1989 f.r.brown et al . ,
phys.rev.lett.61 , 2058 , 1988 f. karsch and e. laermann , phys .
d 50 , 6954 ( 1994 ) s. aoki et al .
( jlqcd collaboration ) , phys.rev .
d 57 , 3910 , 1998 .
r. pisarski and r. wilczek , phys .
d 29 , 338 ( 1984 ) . g.thooft in high energy physics eps conference palermo 1975 , a.zichichi ed . s.mandelstam
phys.rep.23c,245,1976 g.thooft nucl.phys.b190 455 1981 a. di giacomo acta physica polonica b25 227 1994 l. del debbio , a. di giacomo , g. paffuti , phys
b 349 , 513 , 1995 l. del debbio , a. di giacomo , g. paffuti , p.pieri phys . lett .
b355 , 255 , 1955 a. di giacomo , b. lucini , l. montesi , g. paffuti , phys .
d 61 , 034503 , 2000 a. di giacomo , b. lucini , l. montesi , g. paffuti , phys . rev .
d 61 , 034504 , 2000 a. di giacomo , g.paffuti , nucl.phys .
proc.suppl.1006 , 604 , 2002 a. di giacomo , independence of dual superconductivity on the abelian projection hep - lat/0206018 .
a. di giacomo , g. paffuti , the abelian projection revisited ( lattice 2003 ) hep - lat/0309019 , to appear in the proceedings .
carmona , m. delia , l. del debbio , a. di giacomo , b. lucini , g. paffuti , phys .
d 66 , 011503 ( 2002 ) j.m.carmona , m. delia , l.del debbio , a. di giacomo , b. lucini , g.paffuti , c.pica ( lattice 2003 ) b.alles,m.delia , a.di giacomo , phys.lett.b483 , 139 , 2000 | non perturbative results from lattice qcd will be discussed , namely : vacuum condensates and qcd sum rules ; @xmath0 and topology ; confinement of color . |
although fluxoid quantization in superconductors was first demonstrated experimentally over 40 years ago,@xcite there has recently been a resurgence of interest in fluxoid dynamics in a ring geometry . for example
, it has been proposed that the interacting dipole moments in an array of superconducting rings can provide a model experimental system for studying magnetism in ising antiferromagnets .
@xcite this possibility has become particularly attractive with the development of @xmath2-rings : superconducting rings with an intrinsic quantum mechanical phase change of @xmath2 upon circling the ring , in the absence of supercurrents or externally applied fields .
such @xmath2 phase changes can be produced either by the momentum dependence of an unconventional superconducting order parameter , @xcite or by magnetic interactions in the tunneling region of a josephson weak link in the ring.@xcite @xmath2-rings are an ideal model system for the ising antiferromagnet , since they have a degenerate , time - reversed ground state in the absence of an externally applied magnetic field .
recent progress @xcite has made it possible to reliably make very large closely packed arrays of @xmath2-rings , which show strong antiferromagnetic correlations in their spin " orientations upon cooling in zero magnetic field.@xcite superconducting ring experiments have also been proposed@xcite and performed @xcite to test for the presence of visons " , vortex - like topological excitations which may result from electron fractionalization in the high - t@xmath3 cuprate superconductors .
in addition , superconducting rings can provide a model system for the early evolution of the universe , through the study of quenched fluctuations in superconducting rings .
@xcite each of these studies depend critically on an understanding of the fluxoid dynamics during the cooldown process .
finally , there has been a resurgence of interest lately in macroscopic quantum coherence effects in superconducting rings , in conjunction with applications to quantum computation.@xcite in these experiments noise due to thermally activated motion of fluxoids must be considered , at least at elevated temperatures .
therefore the study of fluxoid dynamics in rings is one of much current topical interest .
thermally activated vortex dynamics in superconductors have been studied extensively,@xcite including in ring geometries.@xcite magnetic noise in high temperature cuprate superconductors has been shown to arise from thermally activated hopping of vortices between pinning sites.@xcite fluxoid dynamics have been studied in superconducting rings interrupted by josephson weak links,@xcite and in mesoscopic rings , with the coherence length comparable to the ring dimensions.@xcite however , the system we have chosen to study , relatively large rings with long magnetic penetration depths ( because they are strongly underdoped ) , is in many respects a particularly simple one .
first , our rings have no intentional josephson weak links . many quantitative details specific to phenomena related to flux quantization can be treated within the london approach , which is not bound by the rigid temperature restriction of the ginzburg - landau theory . the smallness of the ring with respect to the pearl length @xcite @xmath4 ( in the thin film limit , the london penetration depth @xmath5 , the film thickness ) simplifies considerably the problem of a ring in an applied magnetic field
. studies of the transitions between quantum states in thin film rings are relevant for understanding the dynamics of multiply connected mesoscopic superconducting devices in general and of the telegraph noise in these devices in particular .
( a ) scanning squid microscope image of an 80 @xmath6 m diameter ring cooled in a magnetic induction of 6.9 mg , resulting in a fluxoid number of n = 10 , and imaged in zero field at t = 6 k. ( b ) cross - sections through the center of the ring in ( a ) , cooled in an induction of 0.7 mg , resulting in a fluxoid number of n = 1 , and imaged in zero field at temperatures of 6 k and 30 k ( dots ) , and modelling as described in the text ( line ) . the data and fit for t = 6 k have been offset vertically by 0.05 @xmath7 for clarity .
the insets at the bottom of ( b ) show schematics of the ring and squid pickup loop geometries .
@xmath8 is the flux through the squid pickup loop .
, width=240 ] our measurements were made on 300 nm thick films of the high - temperature superconductor bi@xmath0sr@xmath0cacu@xmath0o@xmath1 ( bscco ) , epitaxially grown on ( 100 ) srtio@xmath9 substrates using magnetron sputtering .
the oxygen concentration in these films was varied by annealing in oxygen or argon at 400 - 450 @xmath10c .
the films were photolithographically patterned into circular rings using ion etching .
the rings had outside diameters of 40 , 60 , and 80 @xmath6 m , with inside diameters half the outside diameters .
the film for the current measurements had a broad resistive transition ( 90% of the extrapolated normal state resistance at t = 79 k , 10% at t = 46
k ) with a zero - resistance @xmath11 of 36 k before patterning .
such broad resistive transitions are characteristic of both single crystals @xcite and thin films @xcite of bscco , and may be indicative of oxygen inhomogeneity . in this paper
we will treat the rings as homogeneous and cylindrically symmetric .
this view is supported by two facts : 1 ) the squid images are homogeneous , at least within the spatial resolution set by the 17.8 @xmath6 m pickup loop size , at all temperatures ( see e.g. fig . [
fig : bigcrsvt ] ) ; and 2 ) the pearl penetration depth is quite long , of order 100 @xmath6 m , at the temperatures of interest .
this long penetration depth might be expected to average out spatial inhomogeneities . nevertheless , it is possible that the fluxoid transitions in our samples are dominated by paths with relatively low barrier heights.@xcite it is therefore remarkable that the simple model described in this paper qualitatively describes our results in the presence of this inhomogeneity .
the critical temperatures of the rings were slightly lower after patterning than the blanket coverage film , presumably due to additional oxygen removal in the ion etching step .
the critical temperature of the individual ring being measured was determined by squid inductive measurements , as described below .
the rings were magnetically imaged using a variable sample temperature scanning superconducting quantum interference device ( squid ) microscope,@xcite which scans a sample relative to a squid with a small , well shielded , integrated pickup loop ( a square loop 17.8 @xmath6 m on a side for these measurements ) , the sample temperature being varied while the low-@xmath11 squid remains superconducting . ( a ) difference @xmath12 in squid signal directly above the 80 @xmath6 m diameter ring minus that with the squid far from the ring , with the ring in the n = 1 fluxoid state ( solid circles ) ; meissner screening signal @xmath12 with an applied induction of @xmath13 0.2 mg ( diamonds ) ; and amplitude of the telegraph noise due to switching between fluxoid states at @xmath14 ( squares ) , all as a function of temperature .
( b ) expanded view of the data close to the ring superconducting temperature @xmath11 .
, width=240 ] figure [ fig : bigcrsvt](a ) shows a scanning squid microscope image of an 80 @xmath6 m outside diameter ring , cooled in an induction of 6.9 mg , which results in a vortex number n = 10 in the ring , and imaged in zero field at low temperature . for consistency ,
all of the measurements presented in this paper were made on this ring .
measurements on a number of rings of all three sizes were also performed , with quite comparable results .
the dots in figure [ fig : bigcrsvt](b ) are cross - sections through the center of the 80 @xmath6 m ring with a ring fluxoid number of n = 1 in zero field at two temperatures .
the solid lines are fits to the ssm images , taking into account the detailed current distributions in the rings .
such fits are used in this paper to determine the temperature - dependent pearl penetration length @xmath15 , which is an important parameter in modelling the fluxoid dynamics .
( a ) calculated dependence of the squid difference signal @xmath12 above minus away from the ring , for the ring and pickup loop geometry used in this paper , as a function of the pearl length @xmath15 ( solid line ) .
the dashed line shows that @xmath12 is calculated to be nearly inversely proportional to @xmath15 .
( b ) calculated dependence of the pearl length @xmath15 on temperature for the 80 @xmath6 m ring , assuming the linear dependence of @xmath12 on temperature indicated by the dashed line in fig .
[ fig : bigphivt ] ( solid line ) .
the dashed line is proportional to @xmath16 , the ideal temperature dependence for @xmath15.,width=240 ] fig .
[ fig : bigphivt ] shows the results of a number of such measurements as a function of temperature on this ring .
the solid circles in fig . [ fig : bigphivt]a are the difference @xmath12 between the squid signal with the pickup loop centered over the ring and that with the pickup loop far from the ring , with the ring in the n = 1 fluxoid state .
such measurements can not be made closer than about 1k from t@xmath3 because the ring switches to the n = 0 state .
the solid diamonds are measurements @xmath12 with a small applied induction @xmath17 = 0.2 mg with the ring in the n = 0 state .
the squares are the change in the flux through the squid when the ring spontaneously changes fluxoid number in the telegraph noise described below .
the dashed line is a linear extrapolation of the @xmath12 ( circles and squares ) data ; the zero crossing of this line provides an estimate of the critical temperature t@xmath3 for this ring , t@xmath3 = [email protected] k. histogram of the number of switches observed , as a function of the externally applied flux , for a squid pickup loop positioned directly above the 80 @xmath6 m outer diameter underdoped bscco ring at t = 30.9 k. the sweep rate was 200@xmath7/s , with the data stored in 512 bins.,width=240 ] our modelling of the supercurrent distributions in these rings is as follows : consider a thin film ring of thickness @xmath19 with radii @xmath20 in the plane @xmath21 .
the london equations for the film interior read @xmath22 where @xmath23 is the supercurrent density , @xmath24 is the superconducting flux quantum , @xmath25 is the order parameter phase , and @xmath26 is the vector potential . since the current in the ring must be single valued , @xmath27 , where @xmath28 is the azimuth and the integer @xmath29 is the winding number ( vorticity ) of the state .
integrating @xmath23 over the film thickness @xmath30 , we obtain : @xmath31 where @xmath32 is the sheet current density directed along the azimuth @xmath28 .
the vector potential @xmath33 can be written as @xmath34 where the last term represents a uniform applied field @xmath35 in the @xmath36 direction and @xmath37 is the vector potential of the field created by a circular unit current of radius @xmath38:@xcite @xmath39,\nonumber\\ k^2 & = & \frac{4\rho r}{(\rho+r)^2+z^2}\,.\label{eq : a}\end{aligned}\ ] ] here , @xmath40 and @xmath41 are the complete elliptic integrals in the notation of ref . .
fluxoid transition rates @xmath42 for the transition @xmath43 vs. the externally applied flux @xmath44 ( swept towards positive @xmath44 ) for a bscco ring of 80 @xmath6 m outer diameter , with t@xmath3 = 32.5 k , at various temperatures .
the solid symbols are experiment .
the solid and dashed lines are the predictions of the model described in the text.,width=230 ] substituting eq .
( [ eq : int ] ) and ( [ eq : a ] ) into ( [ eq : current ] ) , we obtain an integral equation for @xmath32 : @xmath45,\label{int.eq}\end{aligned}\ ] ] where @xmath46 .
this equation is solved by iteration for a given integer @xmath29 and field @xmath35 to produce current distributions which we label as @xmath47 .
after @xmath47 is found , the field outside the ring can be calculated using eq .
( [ eq : a ] ) : @xmath48+h\,.\label{h(r , z)}\end{aligned}\ ] ] the flux through the squid is obtained numerically by integrating eq .
( [ eq : fieldz ] ) over the pickup - loop area .
the lines in fig . [ fig : bigcrsvt]b are two - parameter fits of this integration of eq .
( [ eq : fieldz ] ) to the data , resulting in @xmath36 = 3.5 @xmath6 m , and @xmath15 = 7 @xmath6 m ( corresponding to @xmath49 = 1 @xmath6 m ) at t = 6 k , and @xmath15 = 54 @xmath6 m ( @xmath49 = 2.8 @xmath6 m ) at t = 30 k. this value ( @xmath49 = 1@xmath6 m ) for the low - temperature in - plane penetration depth is at first surprising , given the observed values of @xmath50 0.2 @xmath6 m for bscco near optimal doping.@xcite however , one might expect the penetration depth to be larger for our underdoped films because of their lower t@xmath3 , following the uemura relation @xmath51 t@xmath3 .
@xcite further , these films have large normal - state resistivities @xmath52 1200 @xmath53-cm , meaning that they are in the dirty limit , and close to the metal - insulator transition.@xcite the zero - temperature penetration depth of a dirty - limit superconductor is given by @xmath54 .
taking the bcs value @xmath55 , with t@xmath3 = 30 k gives @xmath56 = 0.7 @xmath6 m .
it is expected that fluctuations in the superfluid density could further increase the penetration depth in these layered superconductors.@xcite to model the fluxoid dynamics data presented in this paper , it is necessary to estimate the temperature - dependent pearl length @xmath15 and the energy associated with supercurrent flow in our rings .
we can infer the temperature dependence of the pearl length from the temperature dependence of @xmath12 as follows .
numerical integration of eq .
( [ eq : fieldz ] ) for our ring and squid pickup loop geometry as a function of the pearl length @xmath15 gives the solid line in fig .
[ fig : biglamvt]a .
the calculated @xmath12 is nearly inversely proportional to @xmath15 , as shown by the dashed line in fig .
[ fig : biglamvt]a . the linear dependence of @xmath12 on temperature indicated by the dashed line in fig . [ fig : bigphivt ] results in a temperature dependence of the pearl length @xmath57 for this ring indicated by the solid line in fig .
[ fig : biglamvt ] . since the london penetration depth @xmath58 ( @xmath59 = t / t@xmath3 )
, @xcite @xmath4 should be approximately proportional to @xmath16 as indicated by the dashed line in fig .
[ fig : biglamvt]b .
the fluxoid number @xmath29 of a ring can be changed by varying the externally applied flux @xmath60 , where @xmath61 is the effective area , @xcite and can be monitored by positioning the squid pickup loop directly over it .
@xcite in the limit @xmath62 , the current around the loop can be found by integrating eq .
( 2 ) to obtain @xmath63 .
this result also can be obtained from more detailed calculations of the system energy e(n , h ) in equilibrium .
@xcite we assign an experimental value for the effective ring area using the telegraph noise data of fig . [
fig : teltoyft ] , and assuming the peaks are spaced by @xmath7 .
this gives a value of 2895 @xmath6m@xmath64 , in comparison with the calculated value of 2719 @xmath6m@xmath64 .
this discrepancy of about 6% could be due to variations in the photolithography of the rings , or errors in the calibration of the helmholtz coils which apply the magnetic fields .
we used the experimental value for @xmath61 in determining the flux scales in figs .
[ fig : histgram ] , [ fig : ratefig ] , and [ fig : teltoyft ] . at temperatures sufficiently close to t@xmath3 the fluxoid number changes by one fluxoid at a time , as determined by the agreement ( to within 10% ) with our calculations for @xmath65 of the measured spacing in applied flux between vortex switching events .
switching distributions @xmath66 were obtained by repeatedly sweeping the applied field , in analogy with measurements of josephson junctions switching into the voltage state.@xcite an example is shown in fig .
[ fig : histgram ] the transition rates @xmath42 of the fluxoid states were determined from such data using@xcite @xmath67 where @xmath68 labels the largest @xmath69 in a given switching histogram peak , and @xmath70 is the flux interval between data points .
the dots in figure [ fig : ratefig ] show the results from such experiments from the 80 @xmath6 m ring at several reduced temperatures .
the assignment of the starting fluxoid number @xmath29 in this data was made by following the transitions as they evolved with temperature from the two - state telegraph noise ( see fig .
[ fig : teltoyft ] ) described below . at temperatures sufficiently close to t@xmath3 and applied fluxes close to a half - integer multiple of @xmath7
, two - state telegraph noise was observed in the squid pickup loop signal when the loop was placed directly above a ring .
an example is shown in fig .
[ fig : teltime ] . telegraph noise signal vs. time for the ring of figure [ fig : histgram ] at t = 31.4 k , @xmath14 .
, width=240 ] the frequency of this telegraph noise oscillates with the applied flux , with period @xmath7 , and peaks at @xmath71 , @xmath29 an integer , as shown in fig .
[ fig : teltoyft ] .
( a)telegraph noise signal vs. applied flux @xmath44 for the 80 @xmath6 m ring at t = 31.6 k. the dots are the data , the solid line is a fit to the model described in the text .
( b ) replot of the data of ( a ) on a log - linear scale .
the dots are the data , the solid lines are fits to an exponential dependence on @xmath44 , in segments @xmath72 and @xmath73 , @xmath29 an integer . , width=240 ]
several general observations can be made about the fluxoid dynamics observed in our experiments .
first , the dynamics are nearly periodic in the applied field , with a period given by the applied field times the effective ring area @xmath63 ( see figs . [
fig : histgram ] , [ fig : ratefig ] , and [ fig : teltoyft ] ) .
this scaling with the effective ring area has been confirmed for three different ring sizes .
second , the fluxoid transition rates depend exponentially on the applied flux , both for the fluxoid escape measurements of fig .
[ fig : ratefig ] , and in the telegraph noise data of fig .
[ fig : teltoyft ] .
the latter becomes clear when this data is plotted on a log - linear scale , as in fig .
[ fig : teltoyft](b ) .
third , at a particular applied field , both the fluxoid escape rates and the telegraph noise frequencies depend exponentially on temperature .
an example for the 80 @xmath6 m ring is shown in fig .
[ fig : ratevst2 ] . plots of the telegraph noise frequency and the fluxoid transition rate , for the n = 0 to n = 1 transition , at an applied flux of @xmath14 , as a function of temperature .
the symbols are the data ; the solid line is the model described in the text .
the fluxoid transition rate data was obtained by extrapolating straight line fits to the data of fig .
[ fig : ratefig].,width=240 ] we consider the mechanism for transitions between fluxoid states as a thermally activated nucleation of a vortex in , and transport of this vortex across , the ring wall .
the relevant energies in the proposed process are 1 ) the energy required to nucleate a vortex , and 2 ) the kinetic and magnetic energies associated with supercurrents in the ring . the maximum vortex energy in a straight thin film superconducting strip of width @xmath74 ( carrying no transport supercurrent):@xcite @xmath75 where @xmath76 is the vortex core size .
the energy of the ring in a state with the winding number @xmath29 is@xcite @xmath77 clearly , the prefactor @xmath78 coincides with the ring energy in the state @xmath79 in zero applied field : @xmath80 see appendix a. we inferred the temperature dependence of the pearl length from our squid microscope measurements above ( see fig .
[ fig : biglamvt ] ) .
once the pearl length is known , it is possible to calculate the temperature dependence of the energy of our ring .
this is done by setting @xmath79 and @xmath81 , and integrating the solution of eq .
( [ eq : giter ] ) to obtain the total supercurrent , and the total energy in the ring from eq .
( [ eq10 ] ) .
figure [ fig : bigengvt]a shows the results of such a calculation for @xmath82 as a function of @xmath15 .
figure [ fig : bigengvt]b plots @xmath82 as a function of @xmath83 for the @xmath84 m ring .
( a ) calculated dependence of the total energy of our 80 @xmath6 m outside diameter ring for @xmath79 , @xmath81 on the pearl length @xmath15 .
( b ) calculated temperature dependence of the total energy for @xmath79 , @xmath81 , using the calculated temperature dependence ( solid line ) of @xmath15 from fig .
[ fig : biglamvt ] . , width=240 ] figure [ fig : ringedia ] shows a simplified schematic of the energies involved in the thermally activated process @xmath85 which is accomplished by a vortex ( or an antivortex ) crossing the ring .
the ring has an initial ring energy @xmath86 , and a final ring energy @xmath87 . within this simple scheme ,
the energy barrier for the process is @xmath88 here , we have used eq .
( [ eq9 ] ) ; the upper ( lower ) sign is for a vortex ( antivortex ) since the corresponding energy is @xmath89 .
it should be noted that the model we consider here is by no means exact .
it disregards an intricate interplay between vortex currents and those flowing in the ring in a certain quantum state @xmath29 ( the currents in the ring are not a simple superposition of vortex currents and those in the absence of a vortex - even within a linear london approach - because a vortex causes the vorticity @xmath29 to depend on the vortex position ) .
for this reason , there is no point - within our model - to calculate
exactly " the magnetic moment ; instead , we consider @xmath6 as a fitting parameter .
we further simplify the model by considering only transitions between the ground state and the first excited state as in the case of a two - level system .
for the two - level system , the rtn rate @xmath90 , where @xmath91 are probabilities to find the system in the states 1,2 and @xmath92 are the lifetimes . since @xmath93 , we readily get @xmath94,@xcite and @xmath95 if the system is in the ground state @xmath29 , the closest state of a higher energy depends on the applied field .
( [ eq9 ] ) it easy to verify that for @xmath96 , the closest state is @xmath97 , whereas for @xmath98 , the first excited state is @xmath99 .
we begin with the latter possibility .
the rate of the transition @xmath85 is @xmath100 since the transition can be accomplished by both vortices and antivortices .
here , @xmath101 is anattempt frequency " , @xmath102 and @xmath103 denote corresponding barriers divided by @xmath104 for vortices and antivortices ( for brevity the argument @xmath105 of the @xmath106 s is omitted ) .
this expression can be easily factorized with the help of an identity @xmath107\,\exp[(x+y)/2]$ ] : @xmath108/t\ } } { 2\nu_0\,\cosh(\mu h/ t)}\ , , \label{t1}\ ] ] where eq .
( [ eq11 ] ) has been used and for brevity we set @xmath109 .
similarly we obtain : @xmath110/t\ } } { 2\nu_0\,\cosh(\mu h/ t)}\ , .
\label{t2}\ ] ] now eq .
( [ nu ] ) yields @xmath111 } \ , .
\label{nu1}\ ] ] the same calculation for the applied field @xmath96 gives @xmath112 } \ , .
\label{nu2}\ ] ] the factors @xmath113 $ ] oscillate with the period @xmath114 because in the ground state the number @xmath29 is the closest integer to the value of @xmath105 . due to these factors , @xmath115 has maxima at @xmath116 .
clearly , the peaks of @xmath115 become sharper when the parameter @xmath117 increases .
the numerator @xmath118 provides an increase of the maxima with increasing applied field .
physically , this happens because the vortex magnetic moment reduces the energy barrier by @xmath119 .
if @xmath120 , the maxima increase quadratically with field : @xmath121
. this is , in fact , the case for our data .
one does not expect @xmath115 to increase without a limit : at a certain applied field , the barrier for the vortex entry splits in two and the vortex can stay in a metastable equilibrium at the ring .
our model does not hold for such fields . the solid line in fig .
[ fig : teltoyft ] shows a fit of eqs .
( [ nu1 ] ) and ( [ nu2 ] ) to the experimental data .
the best fit parameters were @xmath101 = 1.1@xmath12210@xmath123 s@xmath124 , @xmath125erg , @xmath126erg , and @xmath127erg / g . from fig .
[ fig : biglamvt]b we read @xmath15(t = 31.6 k ) = 240 @xmath6 m .
taking @xmath128 nm,@xcite and @xmath129 m , we calculate @xmath130 erg , a factor of 2.4 larger than the value extracted from the fit .
as discussed above , sample inhomogeneities or surface defects@xcite could reduce the barrier to entry of vortices in type - ii superconductors . from fig .
[ fig : bigengvt]b we read @xmath131erg , smaller than the value obtained from the fit by a factor of 2.3 .
our value for the attempt frequency is within the range ( 10@xmath132 - 10@xmath133 hz ) suggested for attempt frequencies from experiments on thermally activated flux jumps in high temperature superconductors .
@xcite therefore our model provides a good description of the magnetic field dependence of the telegraph noise at a fixed temperature , using values for the vortex nucleation energy @xmath134 and the ring supercurrent energy coefficient @xmath82 that are within approximately factors of 2 of values calculated from experimental measurements on the same ring .
note that our estimate of the attempt frequency is very sensitive to the value of the coherence length @xmath76 .
indeed , the factor @xmath135 combined with @xmath136 of eq . ( [ eq8 ] ) yields @xmath137 with a large exponent @xmath138 .
the same model provides good agreement with the temperature dependence of the fluxoid transition rates and telegraph noise frequencies for the @xmath139 transition at @xmath14 shown in fig .
[ fig : ratevst2 ] . the solid line in fig .
[ fig : ratevst2 ] is the prediction of eq .
( [ n , n+1 ] ) , scaling the value for @xmath140 at t = 31.6 k from the fit of fig .
[ fig : teltoyft ] by @xmath141 .
the solid lines in fig . [ fig : ratefig ] show the predictions of eq .
( [ t1 ] ) , using the model outlined above , with the fit values from the telegraph noise data of fig .
[ fig : teltoyft ] , with @xmath134 and @xmath82 scaled in temperature according to the calculated curves in figs .
[ fig : biglamvt ] and [ fig : bigengvt ] respectively .
the predictions of the model diverge from experiment for lower temperatures and fluxoid numbers .
in particular , the model predicts that the slope of the fluxoid transition rates with applied flux should _ increase _ as the temperature is reduced .
however , as can be seen from fig .
[ fig : ratefig ] , although these slopes are relatively insensitive to temperature , if anything they _ decrease _ with decreasing temperature .
somewhat better agreement with experiment ( the dashed lines in fig .
[ fig : ratefig ] ) is obtained if @xmath82 is taken to have the temperature independent value obtained from the fit to telegraph noise data of fig .
[ fig : teltoyft ] , with @xmath142 as before .
we can speculate on some of the sources of the differences between the predictions of our model and experiment .
first , the model does not take into account interactions between the bulk vortex and the supercurrents .
as discussed above , we have also implicitly assumed that the rings are spatially homogeneous , with a sharp superconducting transition temperature .
the resistive transitions are in fact quite broad .
this broadening could be a source of the apparently reduced temperature dependence of @xmath82 and reduced vortex nucleation energy that we observe .
finally , spatial inhomogeneities could reduce the effective width of the rings . in summary
then , we have measured single fluxoid transitions and two - state telegraph noise in superconducting thin film rings as a function of applied magnetic field and temperature at temperatures close to t@xmath3 . the long penetration depths in the underdoped cuprate films used allowed measurements over a relatively broad temperature range .
the measurements are generally consistent with a model in which the fluxoid transitions are mediated by thermally activated nucleation of a bulk vortex in , and transport of the vortex across , the ring wall .
we presented a model which qualitatively explains some of the features of the data , but other features remain puzzling .
we would like to thank t. senthil and m.p.a .
fisher for suggesting ring experiments to us , and d. bonn , w.a .
hardy , r. koch , k.a .
moler , d.j .
scalapino , f. tafuri , and s. woods for useful conversations .
we would like to thank r. mints for detailed discussions and calculations of the fluxoid transition process .
this manuscript has been authored in part by iowa state university of science and technology under contract no .
w-7405-eng-82 with the u.s .
department of energy .
the work of vgk was partially supported by the binational us - israel science foundation .
the magnetic part of the energy for the state @xmath29 in zero applied field is @xmath143 .
substitute here the vector potential from eq .
( 2 ) to obtain @xmath144 . since the kinetic part is the integral over the volume of the quantity @xmath145 , the first term in @xmath146 is @xmath147 .
further , @xmath148 and we have @xmath149 . m.b ketchen , w.j .
gallagher , a.w .
kleinsasser , s. murphy , and j.r .
squid 85 : superconducting quantum interference devices and their applications _ , ed .
hahlbohm and h. lbbig , walter de gruyter , berlin , 1985 , p. 865 . | we have measured the dynamics of individual magnetic fluxoids entering and leaving photolithographically patterned thin film rings of the underdoped high - temperature superconductor bi@xmath0sr@xmath0cacu@xmath0o@xmath1 , using a variable sample temperature scanning squid microscope .
these results can be qualitatively described using a model in which the fluxoid number changes by thermally activated nucleation of a pearl vortex in , and transport of the pearl vortex across , the ring wall . |
the simplest of many - electron systems are those with two electrons like ions of the helium isoelectronicc series or a harmonium atom . since the schr@xmath9dinger equation for these systems can not be solved analytically , accurate solutions are obtained by using the variational principle for the energy .
thus an approximate parametrized form of the wavefunction is chosen incorporating the effect of electron - electron interactionon in the wavefunction and the expectation value of the hamiltonian ( atomic units are used ) @xmath10 is minimized with respect to the parameters in the wavefunction .
this leads to an optimized approximate wavefunction and the corresponding energy .
the various forms used for the wavefunctions are those by hylleraas@xcite , chandrasekhar@xcite and kinoshita@xcite . as an example of how these forms incorporate the effect of electron - electron interaction
, we consider the chandrasekhar wavefunction for two - electron atom given as @xmath11 where @xmath4 and @xmath5 are the parameters to be optimized and @xmath12 is the normalization constant .
the idea behind writing the wavefunction in the form above is as follows . in the ground state of a two electron atom ,
each electron moves essentially in a hydrogen - atom like @xmath13 orbital ( of the form @xmath14)but because of electron closer to nucleus , the effective nuclear charge is different for the outer electron .
hence the parameters for the orbitals of the two electrons are different .
symmetrization of the resulting product wavefunction then leads to the form given in equation @xmath15 .
another way of obtaining accurate wavefunctions has been given by hylleraas@xcite where the wavefunction is expressed as @xmath16 where @xmath17 , @xmath18 , @xmath19 , and @xmath20 and @xmath21 are variational parameters .
the most accurate calculation using this form of the wavefunction has been that of frankowski and pekeris@xcite , who used a @xmath22-term wavefunction to get highly accurate energies for the helium isoelctronic series . employing a @xmath23-term wavefunction freund et al .
@xcite got essentially the same results as those of @xcite . using another modification to hylleraas type wavefunction ,
koga @xcite constructed a @xmath8-term wavefunction that gives energy within @xmath24 nano - hartree of the energy given in @xcite .
in addition , usefulness of the work of @xcite lies in the fact that they expressed the electronic density of the @xmath8-parameter wavefunction in a simple analytical form with @xmath25 parameters . along similar lines , in @xmath26 umrigar and gonze@xcite
constructed very accurate densities for he isoelectronic series by using @xmath27-term wave function having the same form of the basis set as used by freund .
the resulting densities have been used in the past to construct exchange - correlation potential@xcite and to calculate@xcite polarizability of two electron atoms and also to study derivative discontinuity of the exchange - correlation@xcite of potential of density functional theory .
examples of references @xcite are sufficient to show that the availability of accurate densities facilitate studies in fundamental aspects of density functional theory@xcite . as such methods that can give accurate densities with relative ease are desirable to facilitates such studies . with this in mind
le sech@xcite introduced a semianalytical wavefunction for the ground state of two electron atoms and ions that gives energy within parts per million ( ppm ) of the answers of references @xcite .
the wavefunction proposed by le sech is motivated by the intuitive aspects of the chandrashekhar wavefunction and correlation factor given by earlier studies @xcite .
the wavefunction is given as @xmath28 where @xmath12 is the normalization constant and @xmath29 is the atomic number of the atomic system .
the factor @xmath30 is chosen so that cusp condition@xcite is satisfied exactly by the wavefunction .
the parameters @xmath4 and @xmath5 are obtained variationally by minimizing the expectation value of the hamiltonian with respect to them .
the energies so obtained for h^-^ , he , li^+^ , be^2+^ , b^3+^ are given in along with the exact energies obtained in @xcite .
the differences between the two energies are also shown .
we mention that the values of parameters @xmath4 and @xmath5 and the corresponding energies obtained by us are slightly different from those reported in ref @xcite .
it is clear from that the wavefunction @xmath31 of leads to accurate energies with the difference from those of @xcite becoming smaller with increasing z. the difference for h^-^ is @xmath32 parts per million and goes down to @xmath33for @xmath34 .
we note that the wavefunction of can be improved further with the inclusion of one more parameter and leads to a closer agreement with the exact energies .
+ the question we now ask is how accurate are the densities given by @xmath31 ?
this is important from the point of view of having an accurate and easily accesible density of two electron systems if we were to employ them to investigate the foundational aspects of density functional theory of many electron systems in general and two electron systems in particular . to this end
we present in the values of @xmath35 , @xmath36 and in table [ tb3 ] and [ tb4 ] various moments of the densities obtained from the le sech wavefunction .
a comparision of the results is made with the accurate values of those obtained from the densities given in @xcite .
it is seen that the densities given by @xmath31 are smaller than the exact values of the densities at @xmath37 . since the kato - cusp condition is satisfied by both @xmath31 and @xmath38 , the derivative @xmath39 are also small in comparision with the exact density .
this implies that the le sech densities may be little more spread out than the exact densities .
next in and , we present various moments of the density obtained from @xmath40 of and compare them with their exact counterparts .
it is seen that @xmath41 and @xmath42 for @xmath40 are less than those obtained from@xcite the exact density .
this is consistent with the density for @xmath40 being smaller than the exact one near the nucleus . for @xmath43 , @xmath44 and @xmath45 on the hand
, it is observed that their values are larger than the exact moments for most of the cases ( exception is for h^-^ ) .
this again shows that for @xmath40 , the density is more spread out in general .
we also test the density in terms of it satisfying the ionization potential ( @xmath46 ) theorem in kohn - sham density functional theory .
furthermore , we also study the accuracy of the corresponding kohn - sham exchange - correlation potential . according to the @xmath46 theorem
@xcite the highest occupied orbital eigenenergy is equal to negative of the ionization potential of a many electron system .
accordingly for a kohn - sham system for a given ground - state density of two electron system , its eigenvalue should be equal to @xmath47 where @xmath48 is the energy of the two - electron system . to test the accuracy of the le sech densities ,
we construct the corresponding kohn - sham system using the zhao - parr@xcite method , which is described briefly in the following .
+ for the orbitals of noninteracting system of electrons the kinetic energy @xmath49 and ground state density @xmath50 are given by + @xmath51 for a given ground state density , the kohn - sham system is formed by minimizing @xmath52 with respect to @xmath53 s with the constraint . @xmath54[\rho(r')-\rho_{0}(r')]}{|\mathbf{r}-\mathbf{r'}|}}\mathrm{d}\mathbf{r}\mathrm{d}\mathbf{r'}=0
\nonumber\ ] ] which is equivalent to @xmath55 , and @xmath56 .
this leads to the equation @xmath57 where @xmath58}{|\mathbf{r}-\mathbf{r'}|}}\mathrm{d}\mathbf{r ' } \label{pzp}\nonumber\ ] ] where @xmath59 and @xmath60 are the lagrangian multipliers to enforce the two constraints above .
is solved self - consistently using a herman - skillman@xcite program modified suitably . since at the solution point @xmath61
, the value of @xmath59 should go to infinity such that @xmath62 remains finite . in this paper
we have performed all the calculations for @xmath63 .
we point out that for two electron systems , the kohn - sham systems can also be found directly from the laplacian of the density .
we however , chose zhao - parr method because it is more general and more accurate . shown in
are the values of @xmath64 and @xmath65 for @xmath31 for the ions studied in this paper .
it is seen for @xmath40 that the values of @xmath66 and @xmath65 are comparable to each other . however , they are not close enough to satisfy the @xmath46 theorem . furthermore , the values of @xmath65 are close to the experimental ionization potential , which shows that the total energies @xmath48 for @xmath40 is accurate . on the other hand @xmath64
is not accurate for @xmath40 .
this is particularily evident for the densities of the @xmath67 ion and he atom where the difference between the two is @xmath68 and @xmath69 , percent respectively .
since the eigenvalue @xmath66 is determined by the asymptotic decay of the electron density , it is evident that asymptotically , the density given by le sech wavefunction is not accurate .
this is further evidenced by the corresponding kohn - sham exchange - correlation shown in figure @xmath70 where we have plotted the exchange - correlation potential of he for both le sech density as well as the density given in@xcite and another accurate density of umrigar and gonze@xcite .
it is clear that le sech density leads to smaller magnitude of the exchange - correlation potential near the nucleus , which in turn gives densities which are more spread out .
+ comparision of the exchange - correlation potential calculated for given le sech density ( @xmath71 ) , modified le sech density ( @xmath72 ) and for the density given in@xcite ( @xmath73 ) and for the accurate density of umrigar and gonze@xcite(@xmath74).,title="fig:",scaledwidth=90.0% ] [ figure ] it is evident from the results presented so for that the le sech wavefunction needs to be improved if it were to give accurate densities .
the rest of the paper is then devoted to modifying the le sech wavefunction so that it leads to densities that can be employed to study exact properties in density functional theory .
we wish to point out that after completing our work , we discovered that the above equation was solved@xcite in @xmath89 to obtain ground state energy and polarizability of the he atom .
however the scope of the present paper is much wider . in this paper
we investigate the wavefunction not only for the energy but also for the density it gives rise to .
we present the results obtained by solving for the two electron atomic systems studied so for and show the accuracy of the energy as well as the density that is achieved by self - consistently determined @xmath80 . from the results presented below ,
it becomes clear that provides a computationally straightforward method of obtaining highly accurate wavefunctions for two - electron systems .
the method can also be applied with equal ease to other two - electrons systems such as electrons in a confined environment or harmonic oscillator potential .
+ given in are the results for the total energies of the modified le sech wavefunction @xmath80 of . as is evident , for the modified wavefunction
, the total energies are better than those for the le sech functions .
the modified wavefuction reduces the error of energies from le sech wavefunction by roughly a factor of @xmath90 .
this is because the wavefunctions has been made more flexible .
in addition to giving improved energies , @xmath80 also gives much better densities than @xmath40 .
we present these results in the following .
+ in , we also present the density and its derivative at @xmath37 obtained from @xmath80 of .
we see that the values of modified le sech density and its derivative are very close the values obtained using exact expression given in @xcite . in and we present the various moments of the density obtained by @xmath80 . these moments are compared with the moments given in @xcite .
we get all the moments calculated by the modified le sech density close to the moments given in @xcite .
this implies that the modified le sech density is very close to the exact density .
+ in , @xmath46 theorem is tested for both le sech and modified le sech density , where the @xmath66 are calculated by solving equation ( 7 ) .
we see that the modified le sech density satifies the @xmath46 theorem almost exactly .
finallly in we show the exchange - correlation potential by using le sech density , modified le sech density , density given in@xcite and for the density obtained by umrigar and gonze@xcite .
we see that the exchange - correlation potential given by modified le sech densities is very close to the exact potential .
le sech wavefunction@xcite , which gives very accurate energies for he - like systems , has been modified so that both the energies and densities obtained from the modified wavefunction are highly accurate .
the densities so obtained can be used with confidence to perform fundamental density functional theory investigations . along this line of investigation
we are now employing the method proposed to study adiabatic connection in density functional theory .
we are grateful to prof .
j. umrigar for providing exact densities for two electron atoms .
50 _ quantum chemistry _ translated and edited by hinne hettema ( world scientific series in @xmath91 century chemistry ) : vol @xmath69 .
this book has the english translation of all the hylleraas papers .
e. a. hylleraas @xmath92 _ z. phys .
_ @xmath93 e. a. hylleraas @xmath94 _ z. phys .
_ @xmath95 e. a. hylleraas @xmath96 _ naturwissenschaften _
@xmath97 e. a. hylleraas @xmath96 _ z. phys .
_ @xmath98 chandrasekhar s @xmath99 _ astrophys .
@xmath100 t. kinoshita @xmath101 _ phys .
_ @xmath102 t. kinoshita @xmath103 _ phys .
rev . _ @xmath104 k. frankowski and c. l. pekeris @xmath105 _ phys .
_ @xmath106 david e. fruend @xmath107 _ phys .
@xmath108 t. koga et al .
_ @xmath110 c. j .
umrigar and xavier gonze @xmath26 _ phys .
@xmath111 manoj k. harbola @xmath26 _ chem .
_ @xmath112 manoj k. harbola @xmath113 _ phys .
@xmath114 r. g. parr and w. yang @xmath115 _ density - functional theory of atoms and molecules _
( oxford science publications ) c. le sech @xmath116 _ j. phys .
b : atom . mol .
_ @xmath117 j. o. hirschfelder @xmath118 _ j. chem .
_ @xmath119 t. kato @xmath101 _ commun .
pure appl .
_ @xmath120 j. p. perdew et al .
@xmath121 _ phys .
@xmath122 m. levy , j. p. perdew and v. sahni @xmath107 _ phys .
a _ @xmath123 q. zhao and r. g. parr @xmath109 _ j. chem .
_ @xmath124 f. herman and s. skillman @xmath118 _ atomic structure calculations _ ( prentice - hall publications ) david r. lide @xmath125 _ crc handbook of chemistry and physics _
t. d. h. baber and h. r. hasse @xmath89 _ math . proc . of the cambridge phil . | it is shown for two electron atoms that ground - state wavefunctions of the form @xmath0 where @xmath1 and @xmath2 are the coordinates of two electrons and @xmath3 , can be made highly accurate by optimizing @xmath4 , @xmath5 and @xmath6 .
this is done by solving a variationally derived equation for @xmath6 for a given @xmath4 and @xmath5 and finding @xmath4 and @xmath5 so that the expectation value of the hamiltonian is minimum . for the set @xmath7 the values for various quantities obtained from the above wavefunction
are compared with those given by @xmath8-parameter wavefunction of koga @xciteand are found to be matching quite accurately(within ppm ) with them . |
this section has two purposes .
firstly , we review some well - known phenomena in order to motivate subsequent developments . after that
, we provide a background of the phenomena with some notation , terminology and notions .
_ * volatility smiles . * _ in a nutshell , vanilla options with different maturities and strikes have different volatilities implied by the well - known formula of @xcite . implied volatility is quoted as the market expectation about the average future volatility of the underlying asset over the remaining life of the option . thus compared to historical volatility it is the forward looking approach . for many years
, practitioners and academics have tried to analyse the volatility smile phenomenon and understand its implications for derivatives pricing and risk management . in @xcite , their link between the real - world and risk - neutral processes of the underlying would be complete by non - traded sources of risk .
@xcite found that the dynamics of the risk premium , when volatility is stochastic , is not a traded security . a number of models and extensions of , or alternatives to , the black - scholes model , have been proposed in the literature : the local volatility models of @xcite , @xcite ; a jump - diffusion model of @xcite ; stochastic volatility models of @xcite , @xcite and others ; mixed stochastic jump - diffusion models of @xcite and others ; universal volatility models of @xcite , @xcite , @xcite , @xcite and others ; regime switching models , etc . from a hedging perspective
, traders who use the black - scholes model must continuously change the volatility assumption in order to match market prices .
their hedge ratios change accordingly in an uncontrolled way : the models listed above bring some order into this chaos . in the course of time , the general consensus , as advocated by practitioners and academics , is to _ choose _ a model that produces hedging strategies for both vanilla and exotic options resulting in profit and loss distributions that are sharply peaked at zero .
we argue that such a model , if recovered ( or implied ) from option prices , by no means nearly explains this phenomenon , but is a means only to describe the implied volatility surface . _
* convexity adjustments . *
_ one of many well - known adjustments is the convexity adjustment - the implied yield of a futures and the equivalent forward rate agreement contracts are different .
this phenomenon implies that market participants need to be paid more ( or less ) premium .
the common approach , as used by most practitioners and academics , is to adjust futures quotes such that they can be used as forward rates .
naturally , this approach depends on an model that is used for this purpose . for the extended vasicek known as @xcite and @xcite model ,
explicit formulae can be derived .
the situation is different for models whose continuous description gives the short rate a log - normal distribution such as the @xcite and @xcite models : for these , in their analytical form of continuous evolution , futures prices can be shown to be positively infinite @xcite and @xcite . in subsequent developments ,
we shall offer a different approach to this phenomenon .
_ * parallel derivative markets .
* _ in an economic system , a financial market consists of a risk - free money account , _ primary _ and _ parallel _ markets .
examples of primary markets are stocks and bonds , and examples of parallel markets are derivatives such as forward , futures , vanilla options , credits which are derived from the same primary asset .
market makers can trade and make prices for derivatives in a parallel market without references to another .
the framework is as follows : a complete probability space @xmath0 with a filtration @xmath1 satisfying the usual conditions of right - continuity and completeness .
@xmath2 denotes a fixed and finite time horizon ; furthermore , we assume that @xmath3 is trivial and that @xmath4 .
let @xmath5 be a continuous semimartingale representing the price process of a risky asset .
the absence of arbitrage opportunities implies the existence of an probability measure @xmath6 equivalent to the probability measure @xmath7 ( the real world probability ) , such that @xmath8 is a @xmath6-martingale .
denote by @xmath9 the set of coexistent equivalent measures @xmath6 .
a financial market is considered such that @xmath10 .
uniqueness of the equivalent probability measure @xmath6 implies the market is _
complete_. the fundamental theorem of asset pricing establishes the relationship between the absence of arbitrage opportunities and the existence of an equivalent martingale measure and in a basic framework was proved by @xcite , @xcite .
the modern version of this theorem , established by @xcite , states that the absence of arbitrage opportunities is essentially " equivalent to the existence of an equivalent martingale measure under which the discounted ( primary asset ) price process is a martingale . for simplicity
, we consider only one horizon of uncertainty @xmath11 $ ] .
a _ contingent claim _ , or a _ derivative _
, @xmath12 is a payoff at time @xmath13 , contingent on the scenario @xmath14 .
the derivative has the special form @xmath15 for some function @xmath16 . here ,
@xmath8 is referred to as the primary ( or the ` underlying ' ) .
more generally , @xmath17 depends on the whole evolution of @xmath8 up to time @xmath13 and is a random variable @xmath18 in financial terms , every contingent claim can be replicated by means of a _ trading strategy _ ( or interchangeably known as _ hedging strategy _ or a _ replication portfolio _ ) which is a portfolio consisting of the primary asset @xmath8 and a risk - free money account @xmath19 .
let @xmath20 and @xmath21 be a predictable process and an adapted process , respectively . @xmath22 and @xmath23 are the amounts of asset and money account , respectively , held at time @xmath24 . in this section , for ease of exposition , we assume that @xmath25 for all @xmath26 .
the _ value _ of the portfolio at time @xmath24 is given by @xmath27 for @xmath26 .
it can be shown that the trading strategy @xmath28 is _ admissible _ such that the value process @xmath29 is square - integrable and have right - continuous paths and is defined by @xmath30 for @xmath26 . for @xmath6-almost surely , every contingent claim @xmath17 is attainable and admits the following representation @xmath31 where @xmath32 $ ] . moreover , the strategy is _ self - financing _ , that is the cost of the portfolio ( also known as derivative price ) is a constant @xmath33 @xmath34 the constant value @xmath33 represents a perfect replication or a _
perfect hedge_. thus far , we have presented the well - known mathematical construction of a hedging strategy in a complete market where every contingent claim is attainable . in a complete market ,
derivative prices are unique - no arbitrage opportunities exist .
derivatives can not be valuated in a parallel market at any price other than @xmath33 . from financial and economic point of view
, the phenomena imply that the market is _ incomplete _ , arbitrage opportunities exist and may not be at all eliminated .
a derivative can be valued at different prices and hedged by mutually exclusively trading in risky assets ( or derivatives ) in parallel markets where market makers engage in market activities : investments , speculative trading , hedging , arbitrage and risk management .
in addition , market makers expose themselves to market conditions such as liquidity , see for instance @xcite .
we argue that exposure to the variability of market activities , market conditions and generally to uncertain future events constitutes a basis of arbitrage opportunity which we shall call _ intrinsic risk_. in general , market incompleteness is a principle under which every contingent claim bears intrinsic risks . let us postulate an assumption as a basis for subsequent reasonings and discussions .
+ * assumption .
* _ the market is incomplete and there exist intrinsic risks embedded in every contingent claim . _
+ while the assumption is theoretical , it is rather realistically a proposition with the phenomena as proof .
in a mathematical context , let @xmath35 be the set of all intrinsic risks , that is the set of all real valued functions on @xmath36 .
denote by @xmath37 the measure to the intrinsic risk @xmath38 on the scenario @xmath14 . as a measure of intrinsic risk
, @xmath39 is a mapping from @xmath35 into @xmath40 . as a basic object of our study
, @xmath39 shall therefore be the random variable on the set of states of nature at a future date @xmath13 .
generally , @xmath39 depends on the evolution of the primary asset up to time @xmath13 and may also depend on the contingent claim : @xmath41 the superscript indicates the dependence of a particular contingent claim @xmath17 .
this leads to a new representation of @xmath17 @xmath42 we now introduce the kunita - watanabe decomposition @xmath43 where @xmath44 is a square - integrable martingale orthogonal to @xmath8 .
thus , we have @xmath45 where @xmath46 and @xmath47 . this representation of @xmath17
have been extensively dealt with , see for example @xcite . by incompleteness , the derivative value @xmath48 represents a perfect hedge , which manifests an initial intrinsic value of risk @xmath49 . in relation to the hedging strategy
( [ hgrepresenttation ] ) , the measure of intrinsic risk shall be considered as the value of all possible future capital which , required to control the risk incurred by the market maker ( such as hedger ) and invested in the primary asset , makes not only the contingent claim acceptable , but its valuation fair . from a mathematical point of view
, market incompleteness implies that there exists in the set @xmath9 an equivalent measure , not necessarily a martingale and/or unique measure , that is assigned to a parallel market .
thus , intrinsic risk may depend on the derivative and is not necessarily unique , as such its measure takes many forms some of which we shall consider for applications . in the remaining of this paper , we shall not discuss further on the abstract representations ( [ hgrepresenttation ] ) and ( [ hrep ] ) , but present them in a more descriptive ( down to earth ) framework - the continuous time framework .
in this section we propose a continuous time financial market consisting of a primary price process @xmath8 and a risk - free money account @xmath50 .
we shall define a measure of intrinsic risk and show that perfect hedging strategies can be constructed .
we also show that the existence of intrinsic risk provides an internal consistency in pricing and hedging a contingent claim .
let @xmath51 be a brownian motion on the complete probability space @xmath0 .
the underlying price process of @xmath8 satisfies the sde @xmath52 where @xmath53 and @xmath54 are lipschitz continuous functions so that a solution exists .
@xmath55 and @xmath56 can be functions of @xmath8 .
the price process of @xmath50 is given by @xmath57 where @xmath58 is a lipschitz continuous function .
we expand the portfolio value process ( [ portvalue ] ) as follows : @xmath59\nonumber \\ & = & \nu(t)v(t)dt + \alpha(t)\sigma(t)x(t ) dw(t),\nonumber\end{aligned}\ ] ] where @xmath60 is a @xmath6-brownian motion and is defined by @xmath61 and @xmath62 here , @xmath6 is some martingale measure .
indeed , the theory of the girsanov change of measure , see for example @xcite , shows that there exists such a martingale measure @xmath6 equivalent to @xmath7 and which excludes arbitrage opportunities .
more precisely , there exists a probability measure @xmath63 such that @xmath64 and @xmath8 is a @xmath6-martingale .
such a martingale measure @xmath6 is determined by the right - continuous square - integrable martingale @xmath65\ ] ] for @xmath26 .
and explicitly @xmath66 and @xmath67 satisfies novikov s condition
@xmath68 < \infty.\ ] ] it is not hard to see that the price process @xmath8 under @xmath6 is given by @xmath69 note that the martingale measure @xmath6 and @xmath67 are , _ if unique _ , theoretically and practically well - known as the _ risk - neutral measure _ and the _ market price of risk _ , respectively .
the risk - neutral valuation formula is given by @xmath70 = d(t)e_{\mathbb{q}}\left[\left .
\frac{1}{d(t)}h(x(t ) ) \right| \mathcal{f}(t ) \right].\ ] ] the expectation is taken under the measure @xmath6 .
it is important to note that in the risk - neutral world the essential theoretical assumptions are : ( 1 ) the _ true _ price process ( [ xpsde ] ) is correctly specified and ( 2 ) prices of derivatives @xmath17 are drawn from this price process , that is derivative prices are uniquely determined by formula ( [ qvalformula ] ) .
these assumptions , if not violated , lead to a complete market and the trading strategy ( [ qhedgestrat ] ) and the measure @xmath6 are unique .
however , in practice as we argued earlier , these assumptions are strongly violated ; as a result market completeness and uniqueness of derivative prices are no longer valid .
that is @xmath6 is no longer risk - neutral , but only an equivalent measure in the set @xmath9 .
we now consider the representation ( [ hgrepresenttation ] ) in a continuous time framework , the measure of intrinsic risk can be defined , without loss of generality , in terms of changes in values in a future time interval @xmath71 $ ] as follows . * definition . *
_ the measure of intrinsic risk in a time interval @xmath72 is defined by @xmath73 , where @xmath74 is a continuous adapted process representing a rate of intrinsic risk . _ as was represented earlier in ( [ hgrepresenttation ] ) , the evolution of a trading strategy shall be adaptable to adjust for the measure of intrinsic risk which can be considered an additional / less capital required in a time interval @xmath72 , that is @xmath75\nonumber \\ & = & \nu(t)v(t)dt + \alpha(t)\sigma(t)x(t)dz(t),\nonumber\end{aligned}\ ] ] where @xmath76 is a @xmath77-brownian motion and is given by @xmath78 and @xmath77 is a measure equivalent to @xmath7 .
thus , @xmath79 .
analogously , @xmath80 is defined as an _ intrinsic price of risk_. under @xmath77 measure , the price process of @xmath8 under @xmath77 is given by @xmath81 consequently the fair value of a contingent claim is given by the formula @xmath82
= d(t)e_{\mathbb{s}}\left[\left .
\frac{1}{d(t)}h(x(t ) ) \right| \mathcal{f}(t ) \right].\ ] ] from a pragmatic standpoint , what is needed in determining prices of derivatives and managing their risks is to allow sources of uncertainty that are epistemic ( or subjective ) rather than aleatory in nature . in theory , the value of a derivative can be perfectly replicated by a combination of other derivatives provided that these derivatives are uniquely determined by the formula ( [ qvalformula ] ) .
in practice , prices of derivatives ( such as futures , vanilla options ) on the same primary asset are not determined by ( [ qvalformula ] ) from statistically or econometrically observed model ( [ xpsde ] ) , but made by individual market makers who , with little , if not at all , knowledge of the _ true _ price process , have used their personal perception of the future .
we argue further on this point as follows .
if we let @xmath83 be the price process of a derivative in a derivative market ( such as futures in particular , @xmath84 $ ] , since its contract is not necessarily connected with a physical primary asset ) , @xmath85 must have an abstract dynamics and is assumed to satisfy a sde @xmath86 where @xmath13 denotes a fixed time horizon larger than or equal to the maturity of any contingent claim , @xmath87 is a lipschitz continuous function so that a solution exists .
we now show that @xmath88 is a @xmath77-brownian motion - the source of randomness that drives the derivative price process @xmath85 .
we introduce a change of time , see for example @xcite .
let @xmath89 be a positive function such that @xmath90 which is finite for finite time @xmath91 and increases almost surely .
define @xmath92 , let @xmath85 be a replacement of @xmath8 , i.e. @xmath93 whose solution is given by @xmath94 with @xmath95 . rearranging the drift term leads to @xmath96 where @xmath97 here , we see the concurrence of the sdes ( [ xosde ] ) and ( [ yosde ] ) , the source of randomness @xmath88 is the very @xmath77-brownian motion ( [ zobrownian ] ) .
we have just shown that the measure @xmath77 is subjective in the sense that the valuation of a contingent claim is not only subjected to the dynamics of the primary asset price , but also subject to an exogenous measure of risk @xmath98 .
we shall call the measure @xmath77 the _ risk - subjective measure_. the connection between the risk - subjective measure and the risk - neutral measure described by ( [ zobrownian ] ) is far more precise than that found in @xcite . an important note here is that the trading strategy ( [ oportmod ] ) is equivalent to the risk - free money account , that is the growth of portfolio value ( [ portvalue ] ) is at the risk - free rate @xmath99 . in terms of pricing and hedging , the presence of intrinsic risk
imposes an internal consistency and implies that possible arbitrage exists in the market ( the primary market and its associated derivative markets ) .
in this section , we shall first discuss some problems related to asset models in parallel markets so as to provide some background for subsequent applications . in the light of intrinsic risk ,
the sde ( [ yqsde ] ) in reality may represent a risky asset price process in parallel markets such as : ( 1 ) futures price process , or ( 2 ) an implied price process recovered from option prices where @xmath87 is the implied volatility .
attempts of recovering the implied price process were pioneered , for examples , by @xcite , @xcite , @xcite , @xcite and references therein .
market makers indeed have dispensed with the correct specification ( [ xpsde ] ) and directly use an implied price process as a tool to prescribe the dynamics of the implied volatility surface .
a practice of recovering an implied price process from observed derivative prices ( such as vanilla option prices ) and use it to price derivatives is known as _
instrumental approach _
, described in @xcite . a practical point that is more pertinent to the instrumental approach
is that the prices of exotic derivatives are given by the price dynamics that can take into account or recover the volatility smile . with reference to intrinsic risk ,
an implied price process is a mis - specification for the primary asset , this was discussed in @xcite and was shown that successful hedging depends entirely on the relationship between the mis - specified volatility @xmath87 and the true local volatility @xmath56 , and the total hedging error is given by , assuming zero risk - free rate , @xmath100 note that this hedge error resembles the term ( [ zetatimechge ] ) .
clearly , the hedging error is an intrinsic price of risk presented as traded asset in the hedging strategy ( [ oportmod ] ) , but not in ( [ qhedgestrat ] ) .
before we illustrate a number of applications for pricing and hedging with specific form of the measure of intrinsic risk , let us state a general result for derivative valuation .
we have established the risk - subjective valuation formula ( [ ovalformula ] ) where the risk - subjective price process is given by ( [ xosde ] ) .
the risk - subjective value @xmath101 of a contingent claim @xmath15 given by @xmath102\ ] ] is a unique solution to @xmath103 with @xmath104 and @xmath105 .
the result is obtained by directly applying the feynman - kac formula .
we have shown that the trading strategy ( [ oportmod ] ) yields the risk - free rate of return on the value of a derivative , and also the intrinsic risk is perfectly hedged by delta - hedging representation ( [ hrep ] ) . as unpredictable as a market , prices in a parallel market ( such as futures and corresponding vanilla options )
may not be driven by the same source of randomness that drives the primary asset ( such as stock and bond ) .
motivated by results ( [ zetatimechge ] ) and ( [ hedgeerror ] ) , in the present framework it makes sense to formulate @xmath98 by an abstract form @xmath106 where @xmath56 is the volatility of the underlying asset , @xmath87 the volatility of a risky asset in a parallel market .
we propose that @xmath98 takes a general form of an exponential family @xmath107 the parameter @xmath108 and @xmath104 . as a result ,
( [ speczeta ] ) is a special case .
+ _ remark .
_ while the diffusion term @xmath56 accounts for the distributional property of the primary asset price , the exogenous term @xmath98 accounts for a phenomenon such as volatility smile .
the existence of intrinsic risk appears to undermine the true probability distribution of the underlying , however it emphasises its important role in determining the values of derivatives .
it ensures maximal consistency in pricing and hedging contingent claims that are path - dependent / independent and particularly derivatives on volatility ( such as variance swap , volatility swap ) .
it insists on a realistic dynamics for the underlying asset as far as delta - hedge is concerned . in practice , forward contracts
are necessarily associated with the primary asset ( such as stock and bond ) and therefore their prices are determined by ( [ qvalformula ] ) and hedged by ( [ qhedgestrat ] ) . as was illustrated in the previous section , can be determined by ( [ ovalformula ] ) which includes a measure of intrinsic risk , @xmath98 , as a convexity adjustment .
hedgers holding the primary asset in their hedging portfolio would receive dividends which are assumed to be a continuous stream of payments , whereas hedgers holding other hedge instruments ( such as futures , vanilla options ) do not receive dividends . in this case
, @xmath98 can be considered as dividend yield and @xmath109 is the amount of dividend received in a time interval @xmath72 . @xmath98 may also be a non - negative function representing the hazard rate of default in a time interval @xmath72 , this well - known approach was proposed in @xcite and references therein .
suppose that @xmath110 is the risk - free rate of return of a foreign money account and @xmath111 the measure of risk that accounts for volatility smile , ( [ risksubjectivepde ] ) is then a direct application to foreign market derivatives where @xmath112 .
this is indeed the simplest application of risk - subjective valuation . as an exogenous variable to the risk - subjective price process
( [ xosde ] ) , @xmath98 of a particular form would become a mean of reversion .
this is a desirable feature in a number of well - known interest rate models such as extended model of @xcite , @xcite model .
with reference to the _ liquidity preference theory _ or the _ preferred habitat theory _ of @xcite , a term premium for a bond can be represented as a measure of intrinsic risk .
it is well - known among both academic and practitioners that the standard complete market framework often failed , see for example @xcite
. incomplete market framework becomes crucial in understanding and explaining well - known market anomalies . in this article
we have introduced the notion of intrinsic risk and derived the risk - subjective measure @xmath77 equivalent to the real - world measure @xmath7 , where @xmath113 . at a conceptual level
, the theory of girsanov change of measure allows us to recognise that the crucial role of @xmath77 , rather than the expectation @xmath114 $ ] , is assigned to the price of a derivative ( such as futures , vanilla option ) .
in addition , the intrinsic risk as a structure is what needed to be imposed on the mutual movements of the primary and derivative markets so that , at least , the pricing and hedging derivatives ( such as swaps and caplets ) can be undertaken on a consistent basis .
apart from such conceptual aspect , the measure @xmath77 does not undermine the role of the measure @xmath7 in that a lot of knowledge about the primary market is known at any given time @xmath24 .
more precisely , the market s expectation ( predictions ) in terms of a measure @xmath77 at time @xmath24 is given by the conditional probability distribution @xmath115 \mbox { on } \mathcal{f}_{\mathbb{s}}(t)\ ] ] where @xmath116 is the information available given by the primary market at time @xmath24 , and @xmath117 is the information generated by derivatives ( such as vanilla options ) with maturities @xmath118 . a final remark : in view of the last financial crises , the market has evolved and there is an apparent need , both among practitioners and in academia , to comprehend the problems caused by an excessive dependence on a specific asset modeling approach , by ambiguous specification of risks and/or by confusions between risks and uncertainties ( volatilities ) . as a result , we presented a continuous time framework that , we believe , brings unity , simplicity and consistency to two important aspects : pricing with correctly specified model for a primary risky asset , and hedging risks that can be correctly understood and specified .
in addition , the framework proposed in this article is rigorous in the sense that the true meanings of properties and relationships of intrinsic risk and volatility are self - consistent such that their values are not arbitrarily assigned nor should their properties be misused by ignorance .
blacher , g. ( 2001 ) a new approach for designing and calibrating stochastic volatility models for optimal delta - vega hedging of exotic options conference presentation at global derivatives , juan - les - pins .
follmer , h. & schweizer , m. ( 1991 ) hedging of contingent claims under incomplete information . in _
applied stochastic analysis _ ( m. h. a. davis and r. j. elliott , eds . ) , stochastics monographs , * 5 * , gordon and breach , london / new york , 38 .
sandmann , k. & sondermann , d. ( 1994 ) ont the stability of log - normal interest rate models and the pricing of eurodollar futures , discussion paper , dept . of statistics ,
faculty of economics , sfb 303 , universit at bonn . | we review the nature of some well - known phenomena such as volatility smiles , convexity adjustments and parallel derivative markets .
we propose that the market is incomplete and postulate the existence of an intrinsic risk in every contingent claim as a basis for understanding these phenomena . in a continuous time framework
, we bring together the notion of intrinsic risk and the theory of change of measures to derive a probability measure , namely risk - subjective measure , for evaluating contingent claims .
this paper is a modest attempt to prove that measure of intrinsic risk is a crucial ingredient for explaining these phenomena , and in consequence proposes a new approach to pricing and hedging financial derivatives . by adapting theoretical knowledge to practical applications ,
we show that our approach is consistent and robust , compared with the standard risk - neutral approach . implied volatility , convexity adjustment , primary and parallel markets , incomplete markets , intrinsic risk , risk - neutral measure , risk - subjective measure , fair valuation , delta - hedging . |
supermassive black holes ( smbh ) , with masses between @xmath4 and @xmath5 , are found at the center of nearly every galaxy where there have been sensitive searches .
while we suspect that the seeds for these smbh might all have a common origin , their formation mechanism is not well understood .
many galaxies at redshifts @xmath6 contain quasars , i.e. , smbh in the midst of luminous accretion .
the soltan argument @xcite suggests black hole masses are largely accounted for via growth due to luminous accretion .
there are far fewer quasars at low redshift @xcite , implying that at some point , smbh cease their luminous accretion .
the quasar _ turnoff _
mechanism is not well understood ( thacker et al .
finally , the black hole mass - stellar velocity ( @xmath7 ) relation @xcite suggests that smbh are in some way co - evolving with their host galaxies , but the @xmath7 relation merely describes an end state and is not a theoretical explanation . as we know of no rapid process by which smbh can lose mass , the evolutionary tracks for smbh consist of stages at progressively higher masses .
these stages must , in order , involve ( 1 ) a formation mechanism , ( 2 ) a period of luminous growth ( the ` quasar phase ' ) , perhaps along with a period of nonluminous growth ( ` turnoff ' ) , and ( 3 ) a period in which smbh lie at the centers of galaxies without substantial growth , as we observe them today .
much current theoretical research concerns the origin and turnoff phases of quasar evolution , while the ` quasar phase ' appears to be relatively well understood . in this paper
we find a new feature of smbh evolution during their quasar phase .
we examine the evolution of the quasar locus in mass - luminosity space as a function of redshift .
the @xmath8 locus is traditionally shown with all quasars on the same plot ( as in figure [ fig : allml ] ) .
the large size of the sloan digital sky survey ( sdss ) dr5 quasar catalogue @xcite allows a subdivision into several redshift bins , each containing thousands of quasars .
the eddington limit produces an absolute upper bound on the luminosity of quasars which is proportional to the black hole mass , @xmath9 @xcite . while strictly applicable only for a spherical accretion flow of ionized gas , models for more realistic accretion configurations with rotation are still limited by a luminosity of this order ( except under special circumstances , e.g. begelman 2002 ) . using a smaller data set ( @xmath10 ) than sdss , kollmeier et al .
( 2006 ) appeared to confirm the applicability of @xmath11 to quasars , using an @xmath8 locus ( similar to figure [ fig : allml ] ) to show that the most luminous quasars reach but do not exceed @xmath12 .
for the lowest black - hole masses at every redshift , we confirm the conclusion of @xcite , but we show that the quasars with the highest smbh mass at every redshift fall well short of @xmath11 . in [ sec : methods ] , we review the methods used to estimate masses and bolometric luminosities .
while we have added nothing original to this methodology , the remainder of our results are entirely dependent upon its accuracy . in [
sec : lm ] , we subdivide the sdss dr5 quasar catalogue by redshift and show that the quasar mass - luminosity distribution does not match what we should expect given our current theoretical understanding of quasar accretion .
in particular , we show that instead a sub - eddington boundary ( seb ) is present in each redshift bin .
we also use our limited statistics to make a first estimate for how the seb evolves with redshift . in [ sec : tests ] , we consider several alternative explanations for the discrepancies between the observed quasar locus and the eddington luminosity , focusing on potential sources of measurement uncertainty or bias . in [ sec : discussion ] , we comment on the potential implications of our new results .
the remainder of this work relies upon an ability to accurately estimate the bolometric luminosities and central black hole masses of quasars at cosmological redshifts .
the primary results in this work are drawn from the shen et al .
( 2008 ) virial mass catalogue for sdss dr5 @xcite quasars .
the luminosity determination is fairly straightforward and uses the relatively settled techniques discussed in richards et al .
the mass determination , on the other hand , uses relatively new techniques , and has a larger uncertainty .
therefore , we review here the basic assumptions in virial mass estimation .
potential sources of error or bias in the shen et al .
virial masses are discussed in more detail in
[ sec : tests ] the determination of black hole masses from spectral emission lines relies upon two basic assumptions : ( 1 ) that the orbital velocity of gas in the broad - line region ( blr ) is dominated by the virial velocity due to the central black hole and ( 2 ) that there is a scaling relationship between luminosity and radius . as a result of ( 1 ) , we can calculate the mass @xmath13 of the black hole from emission lines in the blr as @xmath14 where @xmath15 is the radius and @xmath16 the velocity of gas emitting the blr spectral lines .
marconi et al . ( 2009 ) suggest corrections to the virial approximation for radiation pressure might be needed at high luminosity , particularly when using civ lines to determine @xmath16 .
this first scaling relationship is based upon black hole masses determined using reverberation mapping , which uses the time delay between variability in the continuum and emission lines to determine @xmath15 ( cf . peterson & horne 2004 ) .
virial mass estimates also require assumption ( 2 ) , an empirical scaling relationship very close to the @xmath17 that we would expect for a black body @xcite , although thermal processes in the accretion disc are likely to be substantially more complex .
this second assumption transforms ( [ eq : virial ] ) into a scaling relation of the form @xmath18 \nonumber,\end{aligned}\ ] ] where fwhm is the full width at half - maximum of the corresponding line profile , @xmath19 is the luminosity per unit wavelength at rest - frame wavelength @xmath20 , and @xmath21 and @xmath22 are constants .
vestergaard & peterson ( 2006 ) determined these constants in black hole mass scaling relations based on h@xmath0 and civ emission lines ( vp06 ) , while mclure & jarvis ( 2002 ) and mclure & dunlop ( 2004 ) developed a mass relation ( md04 ) by scaling mgii - based estimates against h@xmath0-based estimates .
the scaling relations used in this work are summarized in table [ table : vmesummary ] .
multiple scaling relations are required because sdss spectra only cover the range 39009100 , so that none of these three blr emission lines are accessible over the entire redshift range of the sdss catalogue .
the additive constants @xmath23 are determined by calibrating virial mass scaling relationships against other methods for black hole mass estimation as part of a ` black hole mass ladder'@xcite .
direct estimates based upon local stellar and gas kinematics ( cf .
ferrarese & ford 2005 provide our best estimates for nearby black holes .
reverberation mapping is calibrated against these estimates , and the virial mass scaling relations are in turn calibrated against reverberation mapping .
these calibrations use very few quasars compared to the 62,185 in the shen et al .
( 2008 ) sample : only 28 quasars are used to calibrate the h@xmath0 and 27 to calibrate the civ scaling relations@xcite .
.summary of the virial mass estimates used [ cols="^,^,^,^,^,^",options="header " , ] [ table : tests ] a large , unknown systematic error in virial @xmath24 measurements is the most plausible explanation by comparison for how the seb might be produced without additional underlying physics , but requires that virial mass estimates are @xmath25 dex incorrect , reverberation mapping is @xmath25 dex incorrect , and that quasars at low redshift lie within a spread of just 1.3 decades of central black hole mass .
it would appear likely that the explanation for these new limits lies in new theory rather than observational errors .
quasar catalogues such as the sloan digital sky survey are now large enough to yield useful statistical conclusions after being subdivided . in this work ,
we have explored just a few of the many ways in which the catalogue might be divided .
it is immediately apparent that when the quasar distribution is considered at any given redshift , the most massive central black holes do not accrete at their eddington luminosity , but rather all fall well short of @xmath26 . while there are many measurements that contribute to these quasar distributions , each with its own set of assumptions , a close examination reveals no evidence that any these are responsible for massive quasars remaining sub - eddington .
rather , it appears that this ` sub - eddington boundary ' is a new physical limit for quasars .
thus , quasar accretion must be more complicated than had been previously thought .
the sub - eddington boundary helps to recast the problem in a two - dimensional form .
this form emphasizes the luminosity as being a function of the mass and accretion rate , fundamental properties of the quasar , in a non - trivial way .
quasar luminosity functions and quasar mass functions are both one - dimensional projections of the mass - luminosity plane .
the full two - dimensional quasar distribution contains complexities difficult to discern from either of these projections .
the sub - eddington boundary presents a theoretical puzzle .
not only must an improved theory explain the sub - eddington boundary , but it must also explain its detailed shape and explain the evolution of that shape with redshift . while the slope of the sub - eddington boundary may not vary greatly with redshift ,
the mass scale at which the boundary becomes relevant is larger at higher redshift .
the results in this paper can be used to develop a series of tests which should be capable of discriminating between models . given the complications introduced by the sub - eddington boundary , the case could be made that we really understand surprisingly little about the supermassive black holes that appear to be at the center of nearly every galaxy . not only is their seeding mechanism unknown , but as shown in this paper , the growth mechanism is quite poorly understood , contrary to expectations .
the eddington limit is relevant at low black hole masses , but is only part of the story .
the sub - eddington boundary developed in this work is the latest addition to a growing collection of puzzles regarding every phase of the evolution of galactic nuclei and surrounding regions .
the authors would like to thank mihail amarie , forrest collman , doug finkbeiner , margaret geller , lars hernquist , gillian knapp , avi loeb , ramesh narayan , jerry ostriker , and michael strauss for valuable comments .
this work was supported in part by chandra grant number go7 - 8136a ( chandra x - ray center ) . | we use 62,185 quasars from the sloan digital sky survey dr5 sample to explore the relationship between black hole mass and luminosity .
black hole masses were estimated based on the widths of their h@xmath0 , mgii , and civ lines and adjacent continuum luminosities using standard virial mass estimate scaling laws .
we find that , over the range @xmath1 , the most luminous low - mass quasars are at their eddington luminosity , but the most luminous high - mass quasars in each redshift bin fall short of their eddington luminosities , with the shortfall of order ten or more at @xmath2 .
we examine several potential sources of measurement uncertainty or bias and show that none of them can account for this effect .
we also show the statistical uncertainty in virial mass estimation to have an upper bound of @xmath3 dex , smaller than the 0.4 dex previously reported .
we also examine the highest - mass quasars in every redshift bin in an effort to learn more about quasars that are about to cease their luminous accretion .
we conclude that the quasar mass - luminosity locus contains a number of new puzzles that must be explained theoretically .
[ firstpage ] black hole physics galaxies : evolution galaxies : nuclei quasars : general accretion , accretion discs |
the purpose of this section is to review the construction of @xmath1-cobordisms in @xcite , chapter 6 .
we include it here for convenience of the reader , and also to fix the terminology , since this will be used throughout the paper .
recall that this construction is used in @xcite to prove that a @xmath8-null immersion of a union of transverse pairs , with algebraically trivial intersections , is @xmath1-cobordant to an embedding .
the starting point for the construction is a transverse pair of framed spheres @xmath12 in a @xmath0-manifold @xmath5 .
typically in the applications the transverse pairs have algebraically trivial intersections .
this means that @xmath3 , @xmath6 have a `` preferred '' intersection point , and all other intersection and self - intersection points are paired up with whitney disks .
let @xmath13 denote a ball in @xmath5 around the preferred intersection point between @xmath3 and @xmath6 , and consider the @xmath0-manifold @xmath14 .
the intersection of @xmath3 and @xmath6 with @xmath15 is the hopf link @xmath16 .
consider parallel copies @xmath17 , @xmath18 of the spheres , and let @xmath19 denote the circles of their intersection with @xmath15 .
@xmath20 , @xmath21 are parallel copies of the components of @xmath16
. attach @xmath2-dimensional @xmath4-handles to @xmath22 along @xmath23 , @xmath24 thus surgering @xmath5 along these circles .
figure [ cobord ] shows the effect of surgery on @xmath25 .
the cores of the handles ( viewed in surgered @xmath25 ) are capped off with the disks bounded by @xmath20 , @xmath21 in @xmath15 , slightly pushed out of the ball @xmath13 .
@xmath26 this gives two spheres @xmath27 , @xmath28 , transverse to @xmath3 , @xmath6 respectively . they are embedded , but intersect in a single point , figure [ cobord ] . the strategy for using them in @xcite is the following .
consider the clifford linking tori , one for each pair of `` extra '' double points of @xmath3 and @xmath6 .
these tori are geometrically dual to the whitney disks for @xmath29 , and they have caps provided by @xmath30 .
when @xmath31 is assumed to be @xmath8-null in @xmath5 , the resulting capped surfaces are used embedded whitney disks , thus solving the embedding problem for @xmath29 .
the crucial observation is that the cores of the @xmath4-handles attached to @xmath5 may be capped off by disks parallel to @xmath32 , which are now embedded .
the resulting embedded @xmath4-spheres intersect the respective co - cores in a single point , and are used to attach @xmath33-handles .
the union of @xmath22 and @xmath4- and @xmath33-handles is an @xmath1-cobordism since the boundary map computing the relative homology with @xmath34-coefficients is given by the identity matrix .
we refer the reader to @xcite for the definition and basic properties of capped gropes . the ( iterated ) splitting operation from @xcite will be used later in the paper , and is briefly reviewed in this section .
a disk - like grope is _ dyadic _ if all component surfaces are either disks or punctured tori .
this means each non - cap surface has exactly one pair of dual subgropes attached to it .
a grope has _ dyadic branches _ if all of the subgropes above the base level are dyadic . in a grope with dyadic branches ,
the caps are in one - to - one correspondence with the _ dyadic labels _ : for each dual pair of surfaces in the grope , label one by @xmath35 and the other by @xmath7 .
a cap gets a label by reading off the sequence of @xmath35s or @xmath7s encountered in a path going from the base surface to the cap .
the splitting operation splits a surface into two pieces , at the cost of doubling the dual subgrope .
it can be used to decompose branches into dyadic branches , and can separate intersection points distinguished by properties unaffected by the dual doubling . in particular , it is used to separate double points that have different dyadic labels , or that represent different group elements in @xmath36 .
suppose @xmath3 is a component surface of a grope , not part of the base .
let @xmath6 be the surface it is attached to , and @xmath16 the dual subgrope
. now suppose @xmath37 is an embedded arc on @xmath3 , with endpoints on the boundary , and disjoint from attaching circles of higher stages . in the 3-dimensional model , sum @xmath6 with itself by a tube about @xmath37 ( the normal @xmath38 bundle ) , and discard the part of @xmath3 that lies inside the tube .
this splits @xmath3 into two components .
@xmath16 is a dual for one component ; obtain a dual for the other by taking a parallel copy of @xmath16 .
the following lemma is proved by an inductive application of this idea , descending from the caps downward to the first stage surface .
( splitting ) [ just splitting ] _ any grope can be transformed by iterated splitting to one with dyadic branches , and such that each cap satisfies : _ 1 .
there are no self - intersections ; 2 .
all caps intersecting the given one have the same label ; and 3 .
the fundamental group classes of the loops through the intersection points are the same .
further , the subset of @xmath39 occuring as double point loops is the same as that of the original grope .
we will need an iterated version of this lemma , where the intersections are uniformized to distance @xmath40 .
this is made precise by the notion of an @xmath40-_type_. we refer to @xcite for the formal definition of an @xmath40-type , and of a _ collision _ at distance @xmath40 .
[ n splitting ] ( splitting to distance @xmath40 ) _ given a positive integer @xmath40 and a capped grope , there is a splitting so that _ 1 .
every branch has an @xmath40-type , 2 .
there are no collisions at distance @xmath41 . given an integer @xmath40 and a capped grope , its splitting to distance @xmath40 ( outcome of lemma [ n splitting ] )
has the following algebraic - combinatorial description .
consider any cap of the split grope .
its intersections going out through chains of @xmath40 branches are encoded by a tree , whose valence equals the class of the grope , and whose edges are labelled by a finite collection of group elements represented by the intersections of the original grope .
the paths in this tree are mimicked by geometric moves on the caps : the resulting double point loop represents the group element equal to the product of the labels along the path in the tree .
however , this combinatorial graph is not , in general , mirrored geometrically as a tree of gropes , due to the existence of cycles .
this is the subject of section [ cycles ] .
this is a preliminary step showing that there is an analogue of splitting of capped surfaces ( lemma [ just splitting ] above ) for transverse pairs of spheres , up to @xmath1-cobordism .
the @xmath1-cobordism is completed when eventually the immersed spheres are homotoped to embeddings .
an extension to transverse pairs with higher whitney disk data is discussed in section [ splitting towers ] .
let @xmath29 be a framed transverse pair in @xmath42 , and suppose @xmath3 has two intersection points @xmath43 , @xmath44 that need to be separated .
consider two parallel copies @xmath18 , @xmath45 of @xmath6 , and let @xmath46 be the distinguished intersections between @xmath18 and @xmath3 , and between @xmath45 and @xmath3 , respectively .
let @xmath37 be a simple closed curve in @xmath3 , dividing it into two disks @xmath47 , @xmath48 , so that @xmath49 and @xmath50 .
attach a @xmath4-handle to @xmath51 along @xmath52 , thus surgering the @xmath0-manifold along the curve @xmath37 .
now both disks @xmath47 , @xmath48 are capped with copies of the core of the attached handle , giving immersed spheres @xmath17 , @xmath53 .
this is similar to the argument in section 6.3 of @xcite , which converts a disk - embedding problem into a sphere problem .
@xmath54 this creates two transverse pairs @xmath55 and @xmath56 , where @xmath17 and @xmath53 each inherits one of the intersections @xmath43 , @xmath44 .
the original intersection point @xmath57 is doubled , so both @xmath18 , @xmath45 have an intersection .
if one of the spheres @xmath17 , @xmath53 is homotopic to an embedding , then we will attach to it a @xmath33-handle , which together with the @xmath4-handle attached earlier yields an @xmath1-cobordism . to complete this @xmath1-cobordism , and also to solve the original surgery problem , one needs to work with both pairs of transverse spheres @xmath55 , @xmath56 . as in lemma
[ n splitting ] , given any @xmath58 , there is an iterated splitting of @xmath59 so that each sphere has an @xmath40-type , and there are no collisions at distance @xmath41 .
this is the analogue of splitting of capped surfaces .
here we show that there is an extension of the arguments above from transverse pairs of spheres ( dually : capped surfaces ) to transverse pairs together with layers of whitney disks ( dually : capped gropes ) .
let @xmath12 be a transverse pair with algebraically trivial intersections , so all extra intersection points are paired up by whitney disks .
given @xmath60 , it may be arranged that : there are @xmath61 layers of whitney disks for the intersections of the spheres , and for each @xmath62 the interiors of whitney disks at height @xmath63 are disjoint from all surfaces below that height ( i.e. we have a transverse pair of whitney towers of height @xmath61 . ) this may be done , for example , by converting @xmath59 into a transverse pair of capped gropes of height @xmath61 , and then contracting some of the surface stages . @xmath64 the splitting of transverse pairs described in section [ splitting spheres ] will be applied to _
pairs _ of intersections , rather than individual intersections .
thus at all times the intersections will stay paired up , and whenever a parallel copy of a surface ( sphere or whitney disk ) is taken , it acquires also a parallel copy of the whole corresponding higher whitney data .
the remaining ingredient is the splitting of whitney disks .
suppose two intersections of some surfaces @xmath20 , @xmath21 are paired up with a whitney disk @xmath65 , and the interior of @xmath65 contains two collections @xmath43 , @xmath44 of points that need to be separated .
consider a curve @xmath37 in @xmath65 with one endpoint in @xmath20 and the other one in @xmath21 , separating @xmath43 and @xmath44 .
perform a finger move on one of the surfaces , say @xmath20 , along @xmath37 , so that two new @xmath20
@xmath21 intersections are created .
the disk @xmath65 is replaced by two whitney disks , @xmath66 and @xmath67 , containing @xmath43 , @xmath44 respectively .
this pushes the splitting problem to a lower stage .
whitney towers are split , using an inductive application of this step together with the splitting of spheres described in section [ splitting spheres ] .
before introducing the general construction for capped gropes in section [ grope scobordism ] , we illustrate the problem , and our approach to it , in the setting of transverse pairs of spheres .
start with a transverse pair of spheres with algebraically trivial intersections , and split it to distance @xmath40 . in this setup
, the splitting of transverse pairs described above will be applied to _
pairs _ of intersections , rather than individual intersections .
thus at all times the intersections will stay paired up , and whenever a parallel copy of a sphere is taken , its intersections acquire also a parallel copy of the corresponding whitney disks .
@xmath68 perform the @xmath1-cobordism construction ( attach a pair of @xmath4-handles ) in a neighborhood of the distinguished intersection point of each transverse pair of spheres , as described in section [ construction ] .
now each whitney disk has a dual capped torus .
we are going to focus on this collection @xmath69 of capped tori , in the complement of the original spheres .
algebraically we have a line , subdivided into intervals labelled by group elements , encoding the intersections of the spheres , as in @xcite .
( note that this line also encodes , algebraically , the intersections among the @xmath69 . )
one can not assume however that this picture holds geometrically as well , i.e. that there is a different capped torus for each vertex of the line .
this is because of the presence of cycles . by definition ,
a _ cycle _ is a path , embedded in the graph ( in this context , in the line ) whose endpoints correspond to the same capped torus .
for example , suppose two spheres @xmath3 , @xmath6 forming a transverse pair intersect each other , as in figure [ cycle ] , and let @xmath70 be the group element corresponding to the intersections .
no splitting is necessary : both spheres have an @xmath40-type for each @xmath40 , and the corresponding graph is a line , with each edge labelled by the same group element @xmath71 .
however , there is no line of tori geometrically : the caps of the dual capped torus intersect each other , this is a cycle of length @xmath7 .
the following construction is used to unravel cycles , without changing the uniform algebraic pattern of intersections . before considering the general problem , here
is the description for the cycle in figure [ cycle ] .
consider @xmath40 copies of one of the spheres , say @xmath6 , denote them @xmath72 .
now there are @xmath40 whitney disks , @xmath73 , @xmath74 , and let @xmath75 denote the corresponding clifford torus , dual to @xmath73 .
@xmath76 perform the construction of section [ construction ] ( attach a pair of @xmath4-handles ) in a neighborhood of the distinguished interesection point of each pair @xmath77 , @xmath74 .
if all spheres @xmath78 are eventually shown to be homotopic to embedding , their parallel copies will be used as attaching maps for @xmath33-handles .
these , together with the @xmath4-handles attached earlier , form an @xmath1-cobordism , since the boundary map @xmath79 is given by an upper triangular matrix .
continue the construction of the capped tori : there is a unique choice for the cap of @xmath80 , intersecting @xmath81 : it has to go over the dual sphere to @xmath81 , provided by the corresponding @xmath4-handle .
there are , however , @xmath40 choices for the other cap of each @xmath80 , intersecting @xmath3 .
we pick a cyclic shift , and it is this choice that is at the core of the construction . for
each @xmath82 let the cap of @xmath80 go over the @xmath4-sphere , dual to @xmath3 near the intersection with @xmath83 , and finally the cap of @xmath84 goes over the sphere dual to @xmath3 near the intersection with @xmath85 .
all caps are disjoint from each other , except for the unavoidable intersections that are inherited from the dual spheres in the @xmath1-cobordism construction .
figure [ unravel ] illustrates the case @xmath86 . the outcome is a cyclic chain of capped surfaces @xmath87 , with the caps of @xmath80 intersecting @xmath88 and @xmath89 the @xmath40-cyclic cover of the capped torus in figure [ cycle ]
this is a geometric realization , with injectivity radius @xmath90 , of the algebraic graph encoding the intersections of the spheres .
we now continue the construction in the general case , for an arbitrary transverse pair of spheres , split to distance @xmath40 .
the solution here is going to be an implant of a segment of length @xmath40 of the universal cover of figure [ unravel ] . more precisely ,
pick one of the transverse pairs , @xmath91 , and look at the algebraic graph of intersections up to radius @xmath40 .
it corresponds to a sequence of pairs of spheres , @xmath92 ( actually , multiple parallel copies due to splitting ) .
we assign the labels @xmath93 to the spheres respecting their linear order .
the edges of the graph correspond to the intersections between @xmath94 and @xmath95 .
moreover , some of the pairs coincide ( precisely in the presence of cycles . )
@xmath96 recall that just like in the capped surface case , the spheres in a pair are assigned dyadic labels .
the notation may be misleading , as @xmath3 , @xmath6 do not correspond to such labels : @xmath97 , @xmath98 may have the same dyadic label .
replace @xmath97 with @xmath40 copies : @xmath99 .
since @xmath97 may occur elsewhere in the chain , say @xmath100 for some @xmath63 , all such @xmath94 are replaced with @xmath40 copies as well .
note that @xmath94 can never coincide with @xmath101 for any @xmath63 , @xmath102 , as this would imply a collision at distance @xmath103 , and would contradict the splitting assumption .
if @xmath104 has not been affected , replace it with @xmath40 copies .
continue this procedure until each @xmath105 is replaced with @xmath40 copies .
now we have a segment of the infinite cyclic cover of figure [ unravel ] , and we follow the same construction as in that example : assign caps to the capped tori with a cyclic shift .
the claim is that the result is a sequence of length @xmath40 of capped tori , _ without cycles _ , exactly mimicing the algebraic line of intersections of the spheres .
if there were no cycles to begin with , our construction just replicates the chain of capped surfaces , without altering the algebraic intersections .
if there were cycles , they are eliminated .
this is the analogue of the @xmath1-cobordism construction for transverse pairs of spheres ( desrcibed in section [ construction ] . )
the new feature of the construction for capped gropes is that it can be applied at every surface stage . at the end of this section
we give the proof of theorem [ geometric tree ] .
recall that every surface ( including the caps ) , above the first stage , in a capped grope has a transverse grope .
it is constructed using two copies of the dual surface , cf section 1.4 in @xcite . to fix notations , suppose @xmath106 is a surface in a capped grope , and let @xmath3 , @xmath6 be surfaces attached to a symplectic pair of circles @xmath107 in @xmath106 .
then the transverse capped grope @xmath108 for @xmath3 is built of two parallel copies of @xmath6 ( and everything attached to @xmath6 ) , and analogously the transverse grope @xmath109 for @xmath6 is made of two copies for @xmath3 .
the transverse gropes @xmath110 are used to resolve intersections of surfaces with @xmath3 and @xmath6 ( and with other surface stages and caps attached to @xmath3 , @xmath6 ) respectively . the base surfaces
@xmath111 , @xmath112 of @xmath108 , @xmath109 intersect in two points near the intersection point @xmath113 of the circles @xmath107 in @xmath106 .
consider a @xmath0-ball @xmath13 around the intersection point @xmath113 in @xmath106 .
the ball @xmath13 is chosen so that @xmath114 and @xmath115 are disks , and @xmath13 contains the two intersection points @xmath116 . consider the circles @xmath117 , @xmath118 of intersection of @xmath111 , @xmath112 with @xmath15 .
as in the construction of section [ construction ] , attach two @xmath4-handles to @xmath22 along @xmath119 , @xmath120 .
the cores of the @xmath4-handles ( viewed in the surgered @xmath0-manifold ) are capped with the disks @xmath121 , @xmath122 .
the result is a pair of spheres , geometrically dual to the surfaces @xmath111 , @xmath112 .
each of these two spheres is embedded , but they intersect in two points . consequently , the spheres may be used to resolve intersections with @xmath3 , @xmath6 and with other surfaces attached to @xmath3 , @xmath6 , but the price is the presence of new intersections between the surfaces pushed off @xmath3 and @xmath6 .
these spheres are useful for finding embedded caps for the grope .
the crucial observation is that once the caps of the grope are improved to embeddings , then the @xmath33-handles can be attached to complete the @xmath1-cobordism .
the attaching @xmath4-spheres are constructed as follows : a parallel copy of the core of each @xmath4-handle is capped off by an embedded disk built of two parallel copies of the dual capped grope ( an embedded disk may be found now , since the caps are assumed to be embedded . ) attaching the @xmath33-handles completes the construction of an @xmath1-cobordism , since the boundary map @xmath123 computing the relative homology with @xmath34-coefficients is the identity matrix .
we conclude with the proof of theorem [ geometric tree ] . starting with a transverse pair of spheres ,
convert one of them into a capped grope
a capped grope of height @xmath4 suffices , and we assume for simplicity that this is the case . let @xmath40 be a positive integer .
split the grope to the distance @xmath40 ( lemma [ n splitting ] . )
fix a cap @xmath124 of the split grope ; as explained in section [ grope splitting ] , the intersections among the caps of the split grope are encoded algebraically by a tree @xmath125 , up to the distance @xmath40 from @xmath124 .
the goal is to convert this into a geometric picture where the capped gropes correspond to the vertices of the tree , and the intersections between their caps are encoded by the edges .
consider all genus @xmath7 pieces of the base surface ( branches ) that appear in the tree @xmath125 .
the gropes have two surface stages and caps .
the split grope has dyadic branches ; fix a genus one piece of the base surface .
it branches into two second stage surfaces , and then into four caps .
adapting to the setting of gropes the cyclic construction of section [ cycles ] , consider @xmath40 parallel copies of one of the second stage surfaces , and then @xmath40 parallel copies of one in each dual pair of caps .
the @xmath1-cobordism construction described above will be applied at each intersection point of the attaching circles of all surfaces ( second stage surfaces and caps ) present in the picture .
this is the model that will be applied at each branch of the split capped grope .
finally , the construction for resolving cycles in section [ cycles ] is generalized from the the context of transverse pairs of spheres ( intersections encoded by a line ) to capped gropes ( intersections encoded by a tree . )
the intersections between the caps are pushed down and off the grope using the transverse spheres provided by the @xmath1-cobordism construction .
this is done with cyclic shift with respect to the numbering of the parallel copies of the caps , and of the dual spheres .
the stage , to which the intersection is pushed down to , is determined based on the dyadic labels of the caps .
the proof that the cycles are resolved is analogous to the sphere case , in section [ cycles ] . | the disk embedding lemma is a technique underlying the topological classification results in @xmath0-manifold topology for good fundamental groups .
the purpose of this paper is to develop new tools for disk embedding that work up to @xmath1-cobordism , without restrictions on the fundamental group . as an application
we show that a surgery problem gives rise to a collection of capped gropes that fit in the framework of control theory .
[ thm]lemma [ thm]proposition [ thm]corollary [ thm]conjecture [ section ] the disk embedding lemma is a technique underlying the topological classification results in @xmath0-manifold topology .
it is conjectured to fail , in general , when the fundamental group of the @xmath0-manifold is free .
a weaker lemma , disk embedding up to @xmath1-cobordism , would imply @xmath0-dimensional surgery but not the @xmath2-dimensional @xmath1-cobordism conjecture .
recall its statement : * conjecture*. _ suppose @xmath3 is an immersion of a @xmath4-sphere into a @xmath0-manifold @xmath5 , and there is a framed immersed @xmath4-sphere @xmath6 with trivial algebraic self - intersection , and algebraic intersection @xmath7 with @xmath3 .
then there is an @xmath1-cobordism of the immersion @xmath3 to a topological embedding .
_ the advantage of this formulation is the extra flexibility in changing the ambient @xmath0-manifold , which could make it applicable in contexts where the disk embedding lemma in unavailable .
we refer the reader to @xcite for a discussion of the consequences of this conjecture .
it has been established @xcite under an additional assumption that the transverse pair of spheres is @xmath8-null in @xmath5 ( in other words , the inclusion @xmath9 induces the trivial map on fundamental groups . ) the purpose of this paper is to develop new techniques for disk embedding that work up to @xmath1-cobordism , for any fundamental group . after reviewing the foundational material
the construction of @xmath1-cobordisms and splitting of gropes the following constructions are presented here .
sections [ splitting spheres ] , [ splitting towers ] : the grope splitting technique @xcite , @xcite is extended , up to @xmath1-cobordism , to transverse pairs of spheres , and more generally to pairs of whitney towers .
section [ grope scobordism ] : the @xmath1-cobordism construction for transverse pairs of spheres @xcite is adapted to the setting of capped gropes , so it can be applied at every surface stage .
section [ cycles ] : a technique for eliminating cycles , which translates the algebraic - combinatorial data into a geometric configuration of gropes .
in particular , the following statement is proved .
[ geometric tree ] _ given a surgery problem , there is a procedure for converting it , up to an @xmath1-cobordism , into a large collection of capped gropes whose intersections ( measured in @xmath10 ) are encoded by a tree , up to any given scale . _
this is a refinement of the outcome of grope splitting in @xcite .
in addition to the algebraic manipulations of the group elements represented by the double point loops of the gropes ( which were sufficient for the proof of the disk embedding lemma in the subexponential growth case in @xcite ) , this enables one to use the geometric techniques of control theory ( cf @xcite , chapter 5.4 . ) _ acknowledgements .
_ i would like to thanks frank quinn for many discussions . |
the x - ray emission from galaxy clusters has enabled the study of what is assumed to be the largest virialized systems in the universe .
the x - ray electron temperature , @xmath8 , measures the depth of the galaxy cluster potential , while the x - ray luminosity , @xmath1 , emitted as thermal bremsstrahlung by the intracluster plasma , measures primarily the baryonic number density within this potential .
when combined with the fact that the electron temperature is a robust estimator of the galaxy cluster total mass , @xmath6 , the x - ray emission observations can be used in cosmological studies to derive constraints on the power spectrum of the initial density perturbations and on cosmological parameters ( e.g. , eke _ et al .
_ 1998 ; oukbir & blanchard 1997 ; bahcall , fan & cen 1997 ) . however , most of the important conclusions from such studies are dependent on the accuracy of used scaling relations between the cluster properties , such as the cluster x - ray luminosity , total mass , gas mass , and the electron temperature .
independent of the large scale distribution , baryonic component of the individual clusters has been used to constrain the mass density of the universe , @xmath9 ( e.g. , briel et al .
1992 ; white et al .
1993 ; white & fabian 1995 ; david , jones & forman 1995 ; evrard 1997 ) , but such studies can be subject to variations in physical properties from one cluster to another , e.g. , due to cluster cooling flows ( fabian et al .
1994 ; white , jones & forman 1997 ; allen & fabian 1998 ; markevitch 1998 ) .
apart from x - ray emission observations , another well known probe of the intracluster gas distribution is the sunyaev - zeldovich ( sz ) effect ( sunyaev & zeldovich 1970 , 1980 ) .
the sz effect is the scattering of the cosmic microwave background ( cmb ) radiation via hot electrons of the x - ray emitting gas through an inverse - compton process , producing a distortion in the cmb spectrum . in recent years , increasingly sensitive observations of galaxy clusters , first with single - dish telescopes and now with interferometers
, have produced accurate maps of the cmb temperature change resulting from the sz effect .
unlike the x - ray emission from galaxy clusters , the magnitude of the cmb temperature decrement due to the sz effect is independent of the redshift , and allows a direct probe to the distant universe .
it is likely that the potential use of the sz effect as a cosmological probe to distant universe is yet to be fully realized , and more work , both together and independent of the x - ray properties of galaxy clusters , is required .
the combination of x - ray emission and sz effect towards a given cluster can be used to determine the distance to that cluster , from which the hubble constant can be derived . also , the sz effect , which measures that gas mass within galaxy clusters , can be used to constrain the total mass density of the universe ( see , e.g. , myers et al .
other than cosmological uses , the sz effect and the x - ray emission from clusters can be used to study cluster gas physics , since these two observables depends differently on the electron number density and temperature distributions . given
the ever increasing galaxy cluster sz data , we have initiated a study to investigate the ways in which the sz effect can be used both as a cosmological tool , as well as a tool to understand the distribution and physical properties of baryonic content within clusters . here in this paper
, we present the observed relation between the cmb temperature change due to the sz effect , @xmath0 , and the x - ray luminosity , @xmath1 , of a sample of galaxy clusters . in section 2
, we present a brief introduction to the sz effect and the x - ray emission , and formulate the expected relation between @xmath0 and @xmath1 . in section 3
, we present the used cluster sample in which sz effect has been observed and derive the @xmath10 relation , while in section 4 , we discuss the observed relation in terms of the galaxy cluster properties . in the same section
, we use this relation as an application to study the cluster contribution to arcminute scale cosmic microwave background anisotropies .
[ cols="<,^,^,^,^ , < " , ] notes : column ( 6 ) indicates references for the x - ray luminosity , temperature and sz measurement , respectively ( for clusters with multiple sz measurements , the second column under the same cluster refers only to the sz data ) : 1 .
arnaud & evrard ( 1998 ) ; 2 .
myers et al .
( 1997 ) ; 3 .
herbig et al .
( 1995 ) ; 4 .
allen & fabian ( 1998 ) ; 5 .
grainge et al .
( 1996 ) ; 6 .
uyaniker et al .
( 1997 ) ; 7 .
jones ( 1995 ) ; 8 .
holzapfel et al .
( 1997 ) ; 9 .
birkinshaw et al .
( 1991 ) ; 10 .
grainge et al .
( 1993 ) ; 11 .
carlstrom et al .
( 1996 ) ; 12 .
neumann & bhringer ( 1997 ) ; 13 .
tsuru et al .
( 1996 ) ; 14 .
hughes & birkinshaw ( 1998 ) ; 15 .
holzapfel ( 1996 ) ; briefly , the sz effect is a distortion of the cosmic microwave background ( cmb ) radiation by inverse - compton scattering of thermal electrons within the hot intracluster medium ( sunyaev & zeldovich 1970 , 1980 ) .
the change in the cmb brightness temperature observed is : @xmath11 where @xmath12\ ] ] is the frequency dependence with @xmath13 , @xmath14 ( fixsen _ et al .
_ 1996 ) and @xmath15 , @xmath16 and @xmath17 are the electron density , electron temperature and cross section for the thomson scattering .
the integral is performed along the line of sight through the cluster .
we refer the reader to a recent review by birkinshaw ( 1998 ) on the sz effect and its observation . at the rayleigh - jeans ( rj ) part of the frequency spectrum , @xmath18 .
the other important observable of the hot intracluster gas is the x - ray emission , whose surface brightness @xmath19 can be written as : @xmath20 where @xmath21 is the redshift and @xmath22 is the x - ray spectral emissivity of the cluster gas due to thermal bremsstrahlung emission within a certain energy band @xmath23 ( @xmath24 ) . by combining the sz intensity change and the x - ray emission observations of a given cluster , the angular diameter distance to the cluster
can be derived due to the different dependence of the x - ray emission and the sz effect on the electron number density , @xmath15 ( e.g. , cavaliere _ et al . _ 1977 ) .
combining the distance measurement with redshift allows a determination of the hubble constant , h@xmath25 , as a function of certain cosmological parameters ( e.g. , see , hughes & birkinshaw 1998 ) .
other than the derivation of the hubble constant , sz effect and x - ray emission can in principle be used to constrain the cosmological parameters based on distance measurements of a large sample of galaxy clusters with a wide range of redshift .
the present paper discusses the relation between @xmath0 and @xmath1 , the x - ray luminosity , for a sample of galaxy clusters .
we can estimate the expected relation based on @xmath26 and @xmath27 dependences between the sz effect and the x - ray emission .
the sz effect is due to the pressure integrated along the line of sight , i.e. @xmath28 , while the x - ray emission is due to the thermal bremsstrahlung with @xmath29 . here , we igonore the contribution from x - ray line emission . by removing the @xmath30 dependence between @xmath0 and @xmath1
, one gets @xmath31 .
the relation between @xmath1 and @xmath8 has been well studied based on numerical simulations ( e.g. , cavaliere _ et al . _
1997 ) and observed data ( e.g. , mushotzky & scharf 1997 ; allen & fabian 1998 ; arnaud & evrard 1998 ) . assuming that @xmath32 , @xmath33 based on numerical simulations , cavaliere _
( 1997 ) predicts that @xmath34 at the scale of groups and flattens to @xmath35 for rich clusters , and saturates to @xmath36 for high temperature clusters .
currently , the sz effect has been detected towards rich clusters with moderate to high electron temperatures , resulting in @xmath37 to @xmath38 , when @xmath39 varies from 5 to 2 .
the observed data currently suggest that @xmath40 , less than previous studies which suggested values for @xmath41 to 3.3 ( e.g. , david et al . 1993 ) . also , since @xmath0 is a measurement of the pressure along the line of sight to the cluster , the integral of @xmath0 through a cylindrical cut of the cluster is directly proportional to the gas mass within that cylinder .
this assumes that the cluster gas is isobaric .
if we expect the gas mass to scale with luminosity according to a relation of the form @xmath42 , then , @xmath43 the second estimate for the @xmath2 relation should be more accurate since the measurement of gas mass accounts for the integrated flux along the line of sight , while the former estimate relies slightly on the assumption of spherical geometry for galaxy clusters and that no cluster gas clumping is present .
in order to derive the observed relation between @xmath0 and @xmath1 , we have compiled a sample of galaxy clusters for which sz data are available from literature , and for which accurate x - ray results are also available .
we list these clusters , x - ray luminosities in the 2 to 10 kev band , electron temperature , sz temperature decrement and references to each of the x - ray luminosity , temperature and sz experiments in table 1 . given that the tabulated sz observations were made at different frequencies , we have scaled the sz temperature decrement to the rj part of the spectrum so that the comparison between the sz effect and the x - ray luminosity of galaxy clusters is uniform . also , some of the sz experiments ignored the higher order corrections to the inverse - compton scattering in evaluating the sz temperature decrement . using the relativistic corrections presented by itoh _
( 1997 ) , we have corrected for the sz temperature decrement , which are mainly important for clusters with high electron temperatures ( @xmath44 10 kev ) .
such corrections only change the sz temperature decrement by about 3% to 5% , but the relation between @xmath0 and @xmath1 presented here is not heavily dependent on such corrections .
another uncertainty associated with the tabulated values is the accuracy of x - ray luminosities . except for cl0016
, we have used the published luminosity values from allen & fabian ( 1998 ) and arnaud & evrard ( 1998 ) .
these studies have taken into account the variations in the x - ray luminosities due to cluster cooling flows , so the values tabulated here should be unbiased from cooling flow effects . however , we note that these luminosities may have both statistical and systematic uncertainties of the order 15% , resulting from poor calibration to uncertainties associated with removal of individual cluster cooling flow regions in calculating the luminosities
. our sz cluster sample is primarily made of interferometric observations of the sz effect ( carlstrom et al .
1996 ; grainge et al .
1993 ) , suzie observations at the caltech sub - mm observatory ( holzapfel et al .
1997 ) , and ovro single - dish observations of the low redshift clusters ( myers et al .
1997 ) .
the published sz temperature decrement values in myers et al .
( 1997 ) are not corrected for the beam dilution and switching .
however , based on modeling of the cluster gas distribution for individual clusters , the authors have calculated various efficiencies in this process ( see , table 8 in myers et al .
we used these values to convert the observed sz temperature decrements to beam - corrected values , which are presented in table 1 . in fig . 1
, we show the observed sz temperature decrement vs. the x - ray luminosity in the 2 to 10 kev band , to which we have fitted a relation of the form @xmath45 . for clusters with multiple sz measurements ,
we have used the mean value of the reported temperature decrement but have appropriately scaled the uncertainty so that it covers the whole range suggested by the two separate measurements . using a maximum - likelihood minimization ,
the data are best explained by : @xmath46 where the uncertainty is the 1@xmath47 statistical error . in order to test the expected @xmath0 and @xmath1 relation based on the x - ray emission data through measurements of the cluster gas mass
, we complied a list of gas mass estimates from published data . in fig . 2 , we show the gas mass within 0.5 mpc of the cluster center as derived by white et al .
( 1997 ) for a sample of @xmath48 150 clusters .
the derived relation between gas mass , @xmath49 , and the cluster luminosity , @xmath1 , is : @xmath50 which is very similar to the relation found by shimasaku ( 1997 ) for 40 clusters studied in jones & forman ( 1984 ) and arnaud _ et al . _
for the sz observational data in table 1 , we derived @xmath51 .
the slope is consistent with the numerically simulated values for @xmath39 ranging from 5 to 3 , but slightly inconsistent with a @xmath52 relation of the form @xmath53 , which is expected for clusters under the self - similar model .
the current observational data suggest that @xmath54 ( markevitch 1998 ) to @xmath55(arnaud & evrard 1998 ) , which is consistent with the present estimate of the @xmath2 slope .
the @xmath4 relation , as derived based on the observed mass data for a large sample of clusters in white et al . ( 1997 ) suggests a slope of @xmath56 to the @xmath2 relation , which is again consistent with the observed value . since the sz temperature decrement is a direct estimate on the cluster gas mass , and assuming that the measured gas mass is essentially similar to the gas mass the sz experiments would observed based on the geometry , then @xmath57 .
however , effects due to cluster projections and other associated systematic errors ( see , below ) may produce a slightly different observed relation that one expected based on these simple arguments .
we note that the observed @xmath4 relation is based on the gas masses within inner 0.5 mpc from the cluster centers . depending on the redshift and the instrumental properties ( beam size , resolution etc )
, individual sz experiments may be sensitive to a different radius from the cluster center .
the @xmath0 values are derived by modeling the observed flux over a larger radius than 0.5 mpc .
thus , comparsion of the two relations may be slightly problematic , however , since the total gas mass out to a larger radius would scale in a self - similar manner , we do not expect the slope to be affected .
however , comparsion of the normalizations are not currently possible .
also , we have used two different estimates of luminosities luminosities in 2 to 10 kev for @xmath2 relation and bolometric luminosities for @xmath4 relation but here again , we do nt expect the slope to have been affected systematically by these differences .
other than statistical uncertainties associated with the measurement of sz decrements and x - ray luminosities , it is likely that there are additional effects contributing to the observed scatter in the @xmath2 relation .
such contributions may come from projection effects of galaxy clusters , which are mostly modeled using spherical geometries , variations in gas mass fraction from one cluster to another , and effects due to gas clumping and nonisothermality .
for example , if a cluster is prolate and elongated along the line of sight , then the observed sz temperature decrement would be larger than what is expected based on the x - ray properties of that cluster ( see , cooray 1998 ) .
in such a case , one may also find a substantially lower value for the hubble constant .
since we are considering the galaxy cluster sample in a statistical manner , the derived relations such as @xmath2 , should be unbiased of effects that may arise due to cluster projections .
however , this requires that the clusters in the sample are distributed randomly , so that the whole sample is unbiased .
this may not likely to be true with the current sample , as most of the current sz targets are clusters which are likely to have been previously studied due to certain properties .
such clusters , which may have been selected due to high x - ray luminosities or gravitational lensing effects , are likely to be prolate clusters elongated along the line of sight such that the observed decrement may be slightly higher than the value expected for the luminosity of that cluster .
this is more likely to be the case with strong gravitational lensing clusters ; the fact that both a1689 and a2163 have similar temperature decrements but different x - ray luminosities clearly suggests this possibility . also , high luminosity clusters are likely to be strong cooling flows , but the luminosity values used here , except for cl0016 , have been corrected for such cooling flow contributions . for the present sample , if biased effects exist either at the low or high end of the luminosities , then the derived relation may have been affected . from the present sz sample , a2218 , a773 , and cl0016 have been observed by multiple observational programs .
the difference between these separate measurements are of the order 15% to 25% .
other than physical systematics effects discussed above , it is likely that the derived relation has an additional intrinsic dispersion as high as 25% .
such differences are likely to arise when clusters are modeled using different parameters and that the observational effects , such as the beam dilution produced by the instrument , are not properly taken into accounted .
it is likely that the @xmath2 relation will be properly studied using carefully selected sample of galaxy clusters that are planned to be observed with ground - based interferometers , and space - based observatories such as _
planck_. the current cluster sample used to derive the @xmath2 relation ranges in redshift from 0.02 to 0.54 , with only one cluster beyond a redshift of 0.2 .
therefore , we are unable to study any redshift evolution of this relation
. however , we note that when using this relation to study clusters at high redshifts , it may be important to consider the possible evolutionary effects . however , if such effects exist in the @xmath10 relation , then it would be primarily due to the evolution of the x - ray luminosity function , but other effects , such as a systematic change in the cluster baryonic gas mass fraction with redshift , can also produce deviations in the @xmath2 relation . in order to test the accuracy of the derived relation between sz effect and x - ray luminosity of galaxy clusters ,
we now consider additional observable cluster properties .
since the x - ray luminosity and the gas mass of a cluster is a measurement of the baryonic content , the presented @xmath10 is primarily a probe of the baryonic mass distribution .
when the @xmath10 relation is combined with scaling relations such as @xmath3 and @xmath58 ( hjorth , oukbir & kampen 1998 ) , one can constrain both the baryonic and dark matter distribution separately from each other . in fig . 3
, we show the observed temperature decrements as a function of the cluster electron temperatures in table 1 . even though the involved uncertainties are high , we find that the @xmath59 is : @xmath60 the slope of this relationship is again consistent with simple scaling - law arguments .
the luminosity - gas mass relation observed from the x - ray measurements in fig .
2 , suggests @xmath61 , which is in good agreement with the relation derived from the sz observational data . as suggested earlier , this relation is more likely to be accurate since the integrated pressure along the line of sight is simply proportional to the gas mass within the line of sight , which is the gas mass within the cylinder defined by geometry .
a direct probe of the total cluster mass along the line of sight is the mass derived based on gravitational lensing observations .
smail et al .
( 1997 ) studied lensing properties of a sample of galaxy clusters observed with the _ hubble space telescope _ and measured both weak and strong lensing properties of the clusters .
they derived a relationship between cluster luminosity , @xmath1 , and mean shear , @xmath62 , of the form : @xmath63 here , @xmath64 is a measurement of the average tangential shear strength of galaxy clusters , and is directly proportional to the cluster total mass responsible for gravitationally lensing the background galaxies towards the foreground cluster .
thus , total mass can be written as @xmath65 .
since @xmath66 , we find that the ratio , @xmath67 , is @xmath68
. this ratio measures the cluster gas mass fraction , which has been used in literature to constrain the total mass density of the universe based on baryonic - mass density as derived based on nucleosynthesis arguments ( see , e.g. , evrard 1997 ) .
it is likely that this ratio is independent of the cluster luminosity , suggesting that the gas mass fraction within clusters is constant from one cluster to another .
recently , arnaud & evrard ( 1998 ) studied the changes in cluster gas mass fraction from one cluster to another and suggested that these changes are likely be due to heating processes within clusters , such as due to winds .
if such effects exist , then the @xmath2 relation would also be affected contributing to the observed dispersion .
we use the x - ray luminosity function ( xlf ) of clusters of galaxies from the _ rosat _ brightest cluster sample ( ebeling et al .
1997 ) , which is an x - ray selected , flux limited sample of 172 clusters compiled from the _ rosat _ all - sky survey data , to study the local cluster contribution to arc - minute scake cosmic microwave background ( cmb ) anisotropies .
in order to compare with the current limits of the compton @xmath7-parameter based on _ firas _ observations , we convert our @xmath2 relation to a @xmath69 relation using : @xmath70 the xlf is represented by : @xmath71 where @xmath72 , @xmath73 and @xmath74 . here , @xmath1 is the luminosity in the 2 to 10 kev band . using the luminosity function and the @xmath75 relation , we can write the average @xmath76 parameter due to galaxy clusters as : @xmath77 where @xmath78 and @xmath79 are the lower and upper limits of xlf , respectively . in order to describe the cluster volume
@xmath80 we adopt a @xmath81-model for the cluster gas distribution with a core radius @xmath82 .
we assume a @xmath81 of 2/3 to describe all clusters .
we allow for the total cluster scale radius @xmath83 to vary with luminosity of the form , @xmath84 , where @xmath85 is a free - parameter . according to mohr & evrard ( 1997 )
, the x - ray size scales with temperature as @xmath86 . with the @xmath52 relation ,
the x - ray size is expected to scale with luminosity as @xmath87 to @xmath88 . using the tabulated data in white et al .
( 1997 ) , we obtained a relation of the form , @xmath89 between the tabulated core radii for clusters , @xmath90 , and their luminosities , @xmath1 .
we note that the core - radii from white et al .
( 1997 ) do not necessarily represent the true underlying sacle - size of the x - ray emission .
in their analysis , each cluster core - radius was treated as a parameter which was varied to obtain a flat temperature profile , under their assumption for the form of the gravitational potential . for the purpose of the present calculation , however , we are only interested in core - radii as a representation for the relative distribution of size - scales . since we also vary most of the parameters , such as the slope in the @xmath91 relation , and the fact that our final results our presented as a ratio between the cluster radius and the core - radius , the use of a @xmath91 relation from data in white et al .
( 1997 ) should not affect the final conclusions .
for the rest of the paper we assume that @xmath85 ranges from 0 to 1 . in order to test the accuracy of derived @xmath2 relation
, we also vary its slope @xmath64 , between 0 and 2.5 . in fig .
4 we show the derived @xmath76 values as a function of @xmath92 . it is likely that cluster sizes , in general , are of the order 15 to 25 core radii .
each of the individual plots correspond to the different @xmath91 relation at steps of 0.25 between 0 and 1 , while each of the curves represent the different @xmath64 values at steps of 0.5 between 0 and 2.5 . for our preferred values of @xmath85 and @xmath64 , @xmath93 to 0.5 and @xmath94 to 1.0 , we find that the cluster contribution to @xmath7-parameter through the sz effect is at least two orders of magnitude less than the current upper limit based on _ firas _ observations of @xmath95 ( mather et al .
if all clusters have a constant size , independent of the x - ray luminosity , then the @xmath76 can be as high as @xmath96 , but this possibility is excluded by the observational data , which suggests that cluster size varies with the luminosity .
in general , we find that the cluster contribution to @xmath7-parameter is @xmath97 , which is consistent with previous estimates ( see , e.g. , ceballos & barcons 1994 ) .
the xlf used here has been calculated for clusters out to redshifts of 0.3 , so that the derived results may be valid to clusters out to the same redshift .
however , evidence for no evolution in the xlf out to redshifts of 0.8 has recently appeared on literature ( rosati et al . 1998 ) , and thus , our results may be valid to a much higher redshift . as studied in barbosa et al .
( 1996 ) , in a low @xmath9 universe , galaxy cluster contribution to compton @xmath7 parameter may be as high as @xmath98 , with most of the contribution coming from clusters at @xmath99 .
other than using the @xmath2 relation as a probe of cluster physics , the observed @xmath2 relation can also be used in the study of high redshift clusters , where sz temperature decrements have been observed but no x - ray emission have been detected .
for example , given that we now know the @xmath2 relation , it is reasonably possible to constrain the expected x - ray luminosity of such a detection , and then , perhaps , based on the observed x - ray flux threshold , put a lower limit on the redshift .
detection of such high redshift clusters has strong implications for cosmological world models , and based on the expected redshift and the mass of such clusters , one can constrain cosmological parameters with high accuracy ( e.g , bartlett , blanchard & barbosa 1998 ) .
the @xmath2 and @xmath4 allows one to directly relate the observed temperature decrements to gas mass , and under the assumption of a constant baryonic fraction , to relate to the total mass of the cluster .
it should then be possible to apply the press - schechter ( ps ) formalism to number density of such cluster masses , and constrain the cosmological parameters , with little dependence on the x - ray observations .
however , to perform such an analysis one requires data from a complete sample of galaxy clusters , and such sz samples are currently not available ; observations of a luminosity - selected unbiased sample of galaxy clusters would be useful in the future to constrain the cosmological parameters , using scaling relations such as the one presented here .
based on observations of the sunyaev - zeldovich effect in galaxy clusters and the x - ray luminosity , we have derived a relation between the two observables .
we have studied this relation in terms of other cluster properties , and have found it to agree with the @xmath52 , @xmath4 and @xmath5 relations . using the observed @xmath2 relation and the x - ray luminosity function for galaxy clusters , we have derived the local cluster contribution to the compton @xmath7-parameter through the sz effect .
these values are at least two orders of magnitudes lower than the current upper limit based on _ firas _ observations . given the planned observations of sz effect in large sample of galaxy clusters , such as with _ planck _ and other ground based interferometers , it is likely that a relation such as @xmath2 would be useful both in making predictions and deriving important cosmological parameters .
we leave such studied to be carried out in future , using complete and unbiased samples of galaxy clusters .
i would like to thank david a. white , the referee , for constructive comments on the manuscript , john carlstrom and bill holzapfel for a helpful suggestions on an early draft of the paper and dan reichart for useful discussions .
tsuru , t. , koyama , k. , hughes , j. p. , arimoto , n. , kii , t. , hattori , m. 1996 , in : uv and x - ray spectroscopy of astrophysical and labatory plasmas , yamashita , k. , watanabe , t. ( eds . ) , universal academic press , tokyo , p. 375 . | we present the observed relation between @xmath0 , the cosmic microwave background ( cmb ) temperature decrement due to the sunyaev - zeldovich ( sz ) effect , and @xmath1 , the x - ray luminosity of galaxy clusters .
we discuss this relation in terms of the cluster properties , and show that the slope of the observed @xmath2 relation is in agreement with both the @xmath3 relation based on numerical simulations and x - ray emission observations , and the @xmath4 relation based on observation .
the slope of the @xmath2 relation is also consistent with the @xmath5 relation , where @xmath6 is the cluster total mass based on gravitational lensing observations .
this agreement may be taken to imply a constant gas mass fraction within galaxy clusters , however , there are large uncertainties , dominated by observational errors , associated with these relations . using the @xmath2 relation and the cluster x - ray luminosity function , we evaluate the local cluster contribution to arcminute scale cosmic microwave background anisotropies .
the compton distortion @xmath7-parameter produced by galaxy clusters through sz effect is roughly two orders of magnitude lower than the current upper limit based on _ firas _ observations .
cosmology : observations cosmology : theory galaxies : clusters : general cosmic microwave background |
the year 2003 marks the 50th anniversary of the discovery of hypernucleus . in these 50 years
, the physics of hypernucleus and strange hadrons has played a key role in understanding both the strong and weak interactions of quarks ( and gluons ) .
starting from the eight - fold way , or the flavor su(3 ) symmetry , various symmetries of hadron spectroscopy and interactions , such as current algebra , were born from the strangeness physics . in 1990
s , we have seen a marvelous development both in experimental and theoretical hypernuclear physics . it has accomplished fairly good description of the strong , electromagnetic and weak interactions of hyperons in nuclear medium .
especially a series of new experimental data on the weak decays have lead us to a nearly full understanding of the mechanisms of the weak decays of the hyperon in nuclei .
this is the subject of this report . in section 2 ,
i summarize the current status of various theories and models of the nonmesonic weak decay . in section 3 ,
the model calculations are compared with recent experimental data . in section 4 ,
i concentrate on the @xmath2 rule and argue that the rule may not be valid in the nmwd . in section 5 , conclusions are given .
the weak decay of @xmath3 in hypernuclei is dominated by the nonmesonic decay modes , @xmath4 , because the mesonic decay , @xmath5 is suppressed .
this is easily understood by realizing that the mesonic decay emits a nucleon of momentum , @xmath6 mev / c , and thus is suppressed by the pauli exclusion in the final states . in contrast , the nonmesonic weak decay ( nmwd ) may emit nucleons with momentum @xmath7 mev / c .
this momentum corresponds to the region of the @xmath8 interaction where the short - range repulsion is important .
it indicates that the nmwd of @xmath3 hypernuclei probes the short distance part of the baryonic weak interaction .
the nonmesonic decay of @xmath3 is most simply realized by two kinds of the two - body decay modes , @xmath9 ( proton induced ) and @xmath10 ( neutron induced ) , while more complicated processes , like @xmath11 , may contribute .
we start with the mechanism of the two - body decay modes .
study of the hypernuclear weak decays showed steady but significant development in 1990 s .
the progress was mostly lead by the new generation of experiments , carried out with high intensity secondary beams of mesons combined with high resolution spectrometers .
new developments include measurement of the decay neutron spectrum and also back - to - back coincidence measurement . before such progress ,
our knowledge on nmwd was very limited .
the simplest view was proposed just after the discovery of the hypernucleus by ruderman and karplus@xcite , where the authors considered one - pion exchange process with a @xmath12 weak vertex for the nmwd and calculated the ratio of the mesonic to nonmesonic decay rates in nuclear matter .
they concluded that the then - available emulsion data are consistent only if the spin of @xmath3 is @xmath13 or @xmath14 .
later , block and dalitz@xcite compared the one - pion exchange ( ope ) process with experiment carefully .
the weak coupling constants at the @xmath12 vertex are fixed by the decay rate and the asymmetry of the @xmath5 decay .
it was realized that although the total decay rate is roughly reproduced the ratio of the neutron induced decay rate , @xmath15 , and the proton induced decay rate , @xmath16 , is in trouble .
in fact , the large momentum transfer enhances the contribution of the tensor part of ope , which induces a strong @xmath17 transitions .
because of the isospin symmetry in the final state , no tensor transition is allowed for the @xmath10 , which leads to the relation @xmath18 .
unfortunately , the experimental data do not show this disparity between the two modes . as the transferred momentum is high between two baryons
, we expect to have shorter range interactions contribute to the nmwd .
it is natural to follow the description of the strong interaction of the baryons , in which the various heavy meson exchanges are coherently added up .
the @xmath19 exchange for the weak decay was first introduced by mckellar and gibson@xcite .
takeuchi et al.@xcite applied their model to @xmath20 and 5 hypernuclei .
these two pioneering works considered only the tensor part of the @xmath19 exchange , where the factorization " ansatz was used to determine the weak @xmath21 coupling constants . later , dubach et al.@xcite proposed to employ the su(3)/su(6 ) symmetry to evaluate the weak coupling constants of the pseudoscalar and vector mesons to the octet baryons .
the parity violating ( pv ) couplings of the pseudoscalar mesons are determined by the su(3 ) symmetry from the @xmath22 couplings , while the su(6 ) symmetry is required for the vector mesons and furthermore an extra parameter is introduced .
the parity conserving ( pc ) part of the weak coupling is assumed to come from the baryon pole diagrams .
similar assumptions have been employed in the calculation of nonleptonic decay of the hyperons .
it is known , however , that they do not reproduce both the pc and pv decay rates satisfactorily .
thus it is important to evaluate the weak coupling constants directly from qcd .
dubach et al.@xcite analyzed the nmwd of @xmath3 in nuclear matter by employing the above prescription for the weak coupling constants .
all the members of the pseudoscalar and vector nonet mesons are taken into account .
the full vector meson exchanges including the central part of the interaction were considered also by parreno et al.@xcite for the calculation of the decay rates of @xmath23 .
it is well known that exchanges of two pions correlated into the scalar - isoscalar meson , `` @xmath24 '' , is crucial in nuclear force to make nuclear bound states . in the weak decay , shmatikov@xcite and itonaga et al.@xcite considered @xmath25 in the scalar @xmath24 channel and @xmath19 channel .
the advantage of this approach is that the weak coupling constant can be estimated from the weak @xmath5 and @xmath26 vertices .
the contributions are found to be sizable in the @xmath27 transitions and also in the tensor transition , although the results from the two groups do not necessarily agree with each other .
as the short range baryonic interaction plays important roles , the nucleon substructure must be taken into account for the complete understanding of nmwd .
the @xmath28 dependence of the hypernuclear life - times also indicates that short - range interaction plays an important role in nmwd.@xcite it is known that the short range @xmath8 repulsion comes from the quark exchange interaction.@xcite similarly , the direct quark ( dq ) weak process was introduced in order to explain the short - range part of the nmwd of hypernuclei . in the dq calculation ,
an effective hamiltonian is introduced for the weak interactions of quarks , which consists of various four - quark vertices , @xmath29 and @xmath30 . in the pioneering work done by cheung et al.@xcite
, these operators were not correctly taken into account .
maltman and shumatikov@xcite considered the @xmath31 and @xmath32 exchange with dq , and pointed out that the @xmath2 may be violated by the dq .
inoue et al.@xcite showed that the dq @xmath33 exchange gives a large @xmath0 ratio , which is one of the difficulties of ope .
they also pointed out that a strong @xmath34 transition is realized in the @xmath27 amplitudes .
later , sasaki et al.@xcite showed that the @xmath0 ratio becomes as large as @xmath35 when we take account of @xmath31 , @xmath32 and dq exchanges .
another notable approach to the baryonic weak interaction is the effective interaction approach , which was first considered by block and dalitz@xcite , and later elaborated by jun et al.@xcite and parreno et al.@xcite they are mostly phenomenological but the last one is based on the chiral effective theory .
the latter can reproduce the basic features of the experimental data employing sufficient number of parameters .
nuclear effects on the nmwd are considered in various aspects .
the medium effects on ope and one - boson exchange ( obe ) have been studied by many groups , where the short - range correlation and final state interaction are two important effects.@xcite contribution of two - nucleon induced decay , @xmath36 , especially in extracting the @xmath0 ratio was studied by spanish groups.@xcite their results seem to agree rather well with recent experimental spectrum . contributions .
( b ) shows a coherent process.,height=226 ] the roles of virtual @xmath37 excitation is another interesting contribution to the nmwd .
it was first considered by bando et al.@xcite , who introduced the @xmath38 decay process .
recently , sasaki et al.@xcite pointed out the importance of coherent @xmath37 mixing in the @xmath20 ( @xmath39 ) hypernuclei , and found that it gives 10% effects .
it may be closely related to the `` overbinding '' problem of the s shell hypernuclei@xcite and is an interesting subject to pursue further .
main observables in nmwd are the nonmesonic decay rates , @xmath40 , the partial rates @xmath16 , @xmath15 and the proton asymmetry parameter @xmath41 . @xmath42&@xmath40&@xmath0&@xmath41&@xmath43&0.52&0.70&@xmath44&sasaki et al.@xciteobe ( all)&0.32&0.46&@xmath44 & parreno et al.@xcite @xmath45&0.42&0.39&@xmath46 or 0.12&itonaga et al.@xciteexp&@xmath47&@xmath48&@xmath49&kek e462/e508@xcite [ table1 ] the theoretical prediction of the @xmath50 ratio , @xmath0 , has a long history . in 1963 , block and dalitz showed that the ope underestimates the ratio by a factor 10 . until 1997 ,
the obe with all mesons showed no improvement giving about @xmath51 .
the calculation in the @xmath52 model performed in 1998 gave a larger value , 0.49 , but adding the @xmath32 exchange at that time reduced the value to 0.20 for @xmath42 .
in 2000 , we found an error in the sign in the @xmath32 exchange amplitudes commonly used .
the revised values of @xmath53 and @xmath54 , which denote the @xmath55 ( @xmath56 ) transition , were shown to give a larger value of the @xmath50 ratio , 0.45 for @xmath57 and 0.70 for @xmath43 for @xmath42.@xcite the meson exchange calculations have also agreed with this conclusion .
now the kek e462 and e508 experiments@xcite have successfully singled out the back - to - back decay events , which are not contaminated by the final state rescattering , and have obtained the relatively lower value of the @xmath50 ratio .
finally , the theories and experiments agree with each other and the @xmath50 puzzle has been solved .
the nucleon spectrum emitted from the nmwd has also been measured and compared with the theoretical one .
agreement is not complete yet and it seems that we still need more realistic final state description .
the remaining problem is in the proton asymmetry parameter , @xmath58 , which describes the decay proton asymmetry with respect to the polarization of @xmath3.@xcite as is shown in table 1 , theoretical calculations mostly give a large negative value , while recent experiments indicate small or positive asymmetry .
future studies of nmwd may concentrate on the @xmath59 decay amplitudes , which have not been determined well experimentally nor theoretically .
they are the key to understand the proton asymmetry @xmath58 and @xmath34 in nmwd .
more observables are being measured in the future .
one of the interesting experiments is the reversed process , i.e. , the weak production of @xmath3 in @xmath60 scattering .
it was shown that the @xmath61 amplitudes may dominate the process.@xcite although the expected cross section is tiny , of the order @xmath62 , an experiment is planned at rcnp@xcite .
the decay of double hypernuclei is a new and interesting subject , as this decay has a new decay modes , @xmath63 or @xmath64 , which are supposed to be dominated by the @xmath59 amplitudes.@xcite
it is well known that the strangeness changing weak decay strongly favors the @xmath2 transition .
the experimental ratios of the @xmath2 and @xmath34 amount to 20 in the decay amplitudes of the kaon and the hyperon . yet , the mechanism for the @xmath2 dominance is not completely understood .
indeed , the standard theory of the electroweak interaction predicts the amplitudes of the same order , which can be seen from the fundamental vertex , @xmath65 , @xmath66 ( @xmath67 ) .
the final @xmath68 system has either isospin 1/2 or 3/2 with the ratio 2 to 1 , which comes from the clebsch - gordan coefficients .
thus naive expectation contradicts with data .
nonleptonic weak interaction.,width=566 ] the difficulty is partially solved by considering the perturbative qcd corrections on the weak transition process .
the one - loop qcd corrections associated with the renormalization group ( rg ) improvement are shown to enhance ( suppress ) the @xmath2 ( @xmath34 ) amplitudes.@xcite the mechanism can be understood by considering the color - flavor structure of the gluon exchange interactions between quarks .
the color - magnetic attraction in the scalar diquark , @xmath56 , @xmath69 , @xmath70 @xmath71 , in the final state enhances the @xmath2 . also the so - called penguin diagram , which is purely @xmath2 and mixes the right - handed flavor - singlet current ,
was shown to contribute significantly.@xcite the low energy effective interaction for the nonleptonic @xmath72 decay is given in terms of various four quark operators , whose coefficients are given as the solution of the rg equation at a renormalization scale of order 1 gev .
the rg equation takes into account the leading log summation and also the operator mixings .
the coefficients show clearly the enhancement of the @xmath2 and the suppression of the @xmath34 vertices .
unfortunately , the perturbative correction seems not sufficient in explaining the observed @xmath2 to @xmath34 ratio .
the consensus is that extra 3 - 5 times enhancement is acquired from `` nonperturbative '' qcd corrections .
this is formally given by the corrections to the matrix elements of the operators , which are renormalized at a scale @xmath73 gev , as all the physics above @xmath74 should be included in the coefficients .
the matrix elements are to be evaluated by , for instance , the lattice qcd , effective theories , and models . as for the @xmath2 enhancement in the @xmath32 decay , morozumi et al.@xcite proposed @xmath75 mechanism , in which @xmath24 is a scalar @xmath76 meson of mass @xmath77 mev .
this process enhances the @xmath2 amplitude as the @xmath24 mass is close to the kaon mass . a model calculation based on the njl model
was performed and showed that the experimental ratio can be reproduced fairly well in this mechanism.@xcite when the hyperon decays are concerned , the @xmath2 enhancement may be understood by combination of the soft - pion relation and the color symmetry argument .
miura - minamikawa@xcite and pati - woo@xcite ( mmpw ) showed that no @xmath34 amplitude comes from the weak vertex diagram in which either the initial two quarks or the final two quarks belong to the same baryon .
this theorem is a consequence of the color symmetry of the constituent quarks of the baryon . for the pionic decays of hyperons ,
the pv transition matrix elements can be reduced into the baryon - baryon matrix elements by the soft - pion technique , for example , @xmath78| \lambda\rangle = -{i\over 2f_{\pi } } \langle n| h^{\rm pc } | \lambda\rangle\ ] ] where we use the following relations satisfied by the weak effective lagrangian , @xmath79 = 0 \qquad \ , [ q_5 ^a , h^{\rm pv } ] = - [ i^a , h^{\rm pc } ] .\ ] ] then the baryon - baryon matrix elements , ex .
@xmath80 , evaluated in the constituent quark model contains only @xmath2 transitions due to the mmpw theorem . as for the pc amplitudes , contributions of the pole diagrams again guarantee the @xmath2 dominance .
are the the baryon - baryon weak matrix elements large enough to explain baryon decays ?
the matrix elements are sensitive to the baryon size . in the dq calculation ,
the nonrelativistic quark model wave function with @xmath81 fm is used .
this gives a large enough matrix element for the @xmath5 decay .
the mit bag model tends to underestimate the matrix elements .
the next interesting question then is whether the @xmath2 is dominant in nmwd of the hyperons . in the two - baryon weak interactions
, we do not see any enhancement mechanism for @xmath2 , nor suppression of @xmath34 .
indeed , it has been pointed out that the dq mechanism gives @xmath34 amplitudes comparable to @xmath2.@xcite maltman and shmatikov also suggested that the @xmath82 couplings may contain significant @xmath34.@xcite it is therefore important to check whether the @xmath34 transition is significant in the hypernuclear decays , for which a few proposal were made for the decays of the @xmath83 -shell hypernuclei @xmath20 and 5.@xcite the @xmath84 of the weak transition will be clearly seen in the @xmath27 decay amplitudes .
we define the ratio , @xmath85 , where @xmath86 ( @xmath87 ) denotes the proton induced decay rate for the total angular momentum @xmath88 (= 0 or 1 ) . then @xmath89 is a clear indicator of @xmath84 : @xmath90 the partial decay rates of @xmath91 , @xmath92 and @xmath42 , are the key quantities .
we define two ratios of the observables : @xmath93 then the following theorem can be proved if one assumes that the nonmesonic decay is dominated by the two - body processes : + the current experimental data are not conclusive : @xmath94 but it indicates @xmath95 and therefore possibility of violation of the @xmath2 rule .
a related topic is the @xmath96 decay of hypernuclei , which is not induced by the @xmath5 decay , but needs to involve another proton .
i applied the soft pion technique again to the @xmath96 emission and showed that the decay through @xmath97 is strongly hindered in the soft pion ( @xmath83 wave ) limit.@xcite instead , the @xmath34 transition gives nonvanishing matrix element in the soft pion limit , and therefore the @xmath96 emission rate is directly connected to the @xmath34 amplitudes of nmwd .
thus a large @xmath96 emission rate may indicate the violation of the @xmath2 rule in nmwd .
after 50 years of study , the nmwd of hypernuclei is fairy well understood in terms of the models of meson exchanges and also by taking account of the quark substructure of the baryon .
the long standing problem of the @xmath0 discrepancy has been solved in cooperation of the theoretical and experimental efforts .
the decay nucleon spectrum indicates strong final state interactions .
yet , we still need some refinement to reproduce the proton asymmetry parameter . the predicted @xmath2 violation should be tested by experiment .
it is crucial to measure @xmath98 precisely
. the @xmath96 decay rate may also be useful in determining the role of @xmath34 component in nmwd .
the roles of virtual @xmath37 ( and also @xmath99 in the double hypernuclei ) are left for the future study .
after all , our goal is to understand the structure and dynamics of hypernuclei as well as to study the qcd corrections on the nonleptonic weak interaction in nuclei . in 2007
, the j - parc is expected to start new series of hypernuclear experiments . combined studies of spectroscopy and weak decay
may provide high quality data so that high precision analyses become possible .
j. cohen , progress in part .
physics , * * ( ) .
h. bando , t. motoba and j. zofka , perspectives of meson science , chap . 20 , eds . t. yamazaki et al . , ( elsevier science , 1992 ) , m. ruderman and r. karplus , phys . rev . * * ( ) . m.m .
block and r.h .
dalitz , phys . rev
. lett . * * ( ) .
mckellar and b.f .
gibson , phys .
c * * ( ) .
k. takeuchi , h. takaki and h. bando , prog .
phys . * * ( ) .
a. parreno , a. ramos and c. bennhold , phys .
c * * ( ) .
dubach , g.b .
feldman , b.r .
holstein and l. de la torre , * * ( ) .
a. parreno , a. ramos and c. bennhold , phys .
c * * ( ) .
m. shmatikov , nucl . phys . * * ( ) .
k. itonaga , t. ueda and t. motoba , nucl .
phys . * * ( ) ; phys .
c * * ( ) .
h. bhang et al . , `` strangeness nuclear physics '' , ed . by il - t .
cheon , ( 1999 ) 303 ; in this proceedings . m. oka and k. yazaki , phys
. lett .
b * * ( ) ; m. oka , k. shimizu and k. yazaki , prog .
phys . * * ( ) .
cheung , d.p .
heddle and l.s .
kisslinger , phys .
c * * ( ) .
k. maltman and m. shmatikov , phys .
lett . b * * ( ) .
t. inoue , s. takeuchi and m. oka , nucl .
phys . * * ( ) .
t. inoue , s. takeuchi and m. oka , nucl .
phys . * * ( ) .
k. sasaki , t. inoue and m. oka , nucl .
phys . * * ( ) ; nucl .
phys . * * ( ) ( e ) . j .-
jun , phys .
c * * ( ) .
a. parreno , c. bennhold and b.r .
holstein , nucl - th/0308074 ( 2003 ) .
e. oset and l.l .
salcedo , nucl .
phys . * * ( ) ; a. parreno , a. ramos and e. oset , phys .
c * * ( ) ; a. parreno and a. ramos , phys .
c * * ( ) .
a. ramos , e. oset and l.l .
salcedo , phys .
c * * ( ) ; w.m .
alberico , a. de pace , g. garbarino and a. ramos , phys .
c * * ( ) .
h. bando , y. shono , h. takaki , int .
j. mod .
phys . * * ( ) .
k. sasaki , t. inoue and m. oka , nucl .
phys . * * ( ) .
h. nemura , y. akaishi and y. suzuki , phys .
lett . * * ( ) .
h. nabetani , t. ogaito , t .sato and t .kishimoto , phys .
c * * ( ) .
j. haidenbauer et al . , phys .
c * * ( ) ; t. inoue , k. sasaki , m. oka , nucl .
phys . * * ( ) ; t. inoue , k. sasaki , s. takeuchi and m. oka , prog .
phys . * * ( ) .
t. kishimoto , nucl .
phys . * * ( ) ; s. minami et al .
, in this proceedings .
m.k . gaillard and b.w .
lee , phys .
lett . * * ( ) .
g. altarelli and l. maiani , phys .
lett . * * ( ) .
vainshtein , v.i . zakharov and m.a .
shifman , * * ( ) .
gilman and m.b .
wise , phys .
d * * ( ) .
paschos , t. schneider and y.i .
wu , nucl .
phys . * * ( ) .
t. morozumi , c.s .
lim and a.i .
sanda , phys .
lett . * * ( ) .
t. inoue , m. takizawa , m. oka , prog .
phys . * * ( ) .
k. miura and t. minamikawa , prog .
phys . * * ( ) .
pati and c.h .
woo , phys .
d * * ( ) .
k. maltman and m. shmatikov , phys .
c * * ( ) ; a. parreno , a. ramos , c. bennhold and k. maltman , phys .
b * * ( ) . c. dover ,
xith european conference on few - body phys . , ( 1987 ) ; j. cohen , phys .
c * * ( ) ; r.a .
schumacher , nucl .
phys . * * ( ) .
m. oka , nucl .
phys . * * ( ) . | i review the current status of understanding the mechanism of the nonmesonic weak decays ( nmwd ) of hypernuclei
. long standing problem on the @xmath0 ratio has been solved by considering short - range weak interactions properly .
this leaves a few remaining problems , such as the asymmetry of the emitted proton with respect to the @xmath1 polarization and validity of the @xmath2 in the nonmesonic weak decay . |
the harmonic oscillator is one of the most appealing models due to its well known solutions in classical mechanics and in quantum mechanics , and due to a wide variety of applications in various branches of physics @xcite . in classical mechanics
its solution is unique , and so it is in the canonical non - relativistic treatment in quantum mechanics @xcite . in this quantum approach ,
the position and momentum operators ( @xmath6 and @xmath7 , respectively ) satisfy the canonical commutation relation @xmath8 = - i\hbar $ ] , and the model is described by its hamiltonian @xmath9 and the heisenberg equations .
this algebraic description can be translated to a description by means of wave functions . in the position representation , for example
, the operator @xmath7 corresponds to multiplying functions by the variable @xmath10 and the heisenberg operator equations correspond to schrdinger s equation , the solutions of which are expressed through hermite polynomials .
these solutions or wave functions lead to important position probability functions .
to study the applicability of a system modelled by a harmonic oscillator , one often turns to expressions for the joint quasi - probability function of momentum and position , i.e. to a quantum analogue of phase space . the wigner distribution function ( d.f . )
@xcite is the main theoretical tool here , as it is considered as the closest quantum analogue of a classical distribution function over the phase space .
the quantum harmonic oscillator is one of the first systems for which the wigner d.f .
was computed analytically . nowadays , there are also advanced experimental techniques that enable one to measure phase space densities for certain quantum systems @xcite , and thus compare theoretical results with experimental results @xcite .
it is in this context that certain deformations or generalizations of the quantum harmonic oscillator are of relevance . in particular , deformations that are still exactly solvable and for which the wigner d.f .
can be computed analytically are highly interesting .
one such model is the @xmath10-deformed quantum harmonic oscillator , being exactly solvable @xcite and having found applications in various physical models @xcite .
its wigner d.f . has been computed in @xcite .
a deformation or generalization of the canonical quantum harmonic oscillator that is more fundamental in nature was proposed by wigner @xcite .
it is known as the one - dimensional wigner quantum oscillator or as the parabose oscillator . in wigner
s approach , the canonical commutation relations are not assumed . instead of that , the compatibility of hamilton s equations and the heisenberg equations is required .
this approach has been the fundamental starting point of so - called wigner quantum systems @xcite .
in such an approach , one typically obtains several solutions ( several algebraic representations ) , of which only one corresponds to the canonical solution .
it is thus a natural and fundamental extension of canonical quantum mechanics .
it is also of growing importance as deviations from the canonical commutation relations ( in particular of commuting position operators in the context of non - commutative quantum mechanics ) are supported by investigations from several field @xcite .
the one - dimensional wigner quantum oscillator itself was already solved by wigner @xcite , leading to an oscillator energy spectrum similar to the canonical one but with shifted ground state energy equal to some arbitrary positive number @xmath11 . in a sense , this number @xmath11 can be considered as the deformation parameter , and the solutions of the wigner quantum oscillator reduce to the canonical one when @xmath12 .
wigner s formulation was also the inspiration for the introduction of parabosons ( and parafermions ) , as the common boson commutator relation of the canonical case is , in wigner s approach , replaced by a triple relation that is also the defining relation of a paraboson @xcite .
later on , one realized that lie superalgebras are the natural framework for this algebraic description @xcite , and in particular that the representations of the wigner quantum oscillator are in one - to - one correspondence with the unitary irreducible lowest weight representations of the lie superalgebra @xmath13 .
so the wigner quantum oscillator , being the generalization of the canonical one , could play an important role to explain various phenomena more precisely , and its importance is undeniable .
however , so far its application has been restricted in comparison with the canonical harmonic oscillator due to the fact that its behaviour in phase space has not been studied .
in other words , there is no expression of the joint quasi - probability distribution function ( wigner d.f . ) of position and momentum available for the wigner quantum oscillator .
the main problem is that the treatment a wigner d.f
. usually starts from the assumption that the commutator of @xmath6 and @xmath7 is a constant , which is no longer the case for the non - canonical wigner quantum oscillator .
it is precisely this problem which is being tackled and solved here . in this paper
, we first recall some basic facts of the one - dimensional wigner quantum oscillator or parabose oscillator .
the emphasis is on the algebraic description , in terms of the lie superalgebra @xmath13 and representations characterized by a positive real number @xmath11 ( section 2 ) . in section 3 we discuss the main definition of the wigner d.f . in the current non - canonical setting .
the main part of the paper is devoted to computing the wigner d.f.@xmath14 analytically , for the parabose oscillator in the @xmath15th excited pure state @xmath16 . for @xmath17 ( and @xmath18 ) , the calculation is somewhat simpler than in the general case , and this is treated first in section 4 .
the calculation involves various computational tools . in some steps ,
series are term wise integrated , so one should pay attention to convergence . here
, we shall often assume that the parameter @xmath11 is of the form @xmath19 , with @xmath20 a nonnegative integer ; in that case all series appearing are finite and there are no convergence problems .
the final expression of @xmath21 is essentially a series in terms of laguerre polynomials , depending only upon @xmath22 , or alternatively a confluent hypergeometric series in @xmath22 , see .
section 5 then deals with the general case : the computation of @xmath14 . here , the analysis is quite involved , and some intermediate calculations are performed in an appendix b. we obtain two alternative expressions for @xmath14 , and , both as series of laguerre polynomials in @xmath22 .
section 6 deals with some plots of the newly obtained wigner distribution functions , discusses some properties and summarizes some conclusions . to end the introduction ,
let us introduce some of the notation and classical results we are going to use .
many of the formulas encountered in this article will involve binomial coefficients , factorials etc .
it is therefor convenient to use the notation for ( generalized ) hypergeometric series @xcite .
let @xmath23 and @xmath24 be nonnegative integers , a hypergeometric series with @xmath23 numerator and @xmath24 denominator parameters is then defined as @xmath25 where @xmath26 is the rising factorial or pochhammer symbol . when one of the numerator parameters @xmath27 is a negative integer , the series is terminating and in fact a polynomial in @xmath28 .
of course , if one of the denominator parameters is also a negative integer , it has to be more negative such that the numerator vanishes before the denominator does .
all classical orthogonal polynomials can be written using this hypergeometric notation .
in particular one has for the laguerre polynomials @xcite : @xmath29 here , we also introduced an alternative notation for the confluent hypergeometric series @xmath30 . if the parameter @xmath31 is @xmath32 , we will simply write @xmath33 instead of @xmath34 . often , the @xmath33 are called laguerre polynomials , while the @xmath35 are called the generalized laguerre polynomials . for the classical orthogonality of the laguerre polynomials to hold @xmath31 needs to be bigger than @xmath36 . in particular , if @xmath31 is a negative integer , the definition is no longer valid since one has a negative integer appearing as a denominator parameter .
the laguerre polynomials can be written in an alternative way such that they are defined for all values of @xmath31 : @xmath37 many well known summation formulas can be very neatly expressed using the hypergeometric notation .
most notable are the binomial summation theorem : @xmath38 with @xmath39 if @xmath11 is not a negative integer , and the chu - vandermonde summation theorem for a terminating gauss hypergeometric series : @xmath40 with @xmath15 a nonnegative integer .
there exists also a very extensive theory of hypergeometric transformations , a very important transformation being kummer s transformation for confluent hypergeometric series : @xmath41
in this paragraph we briefly describe the treatment of a harmonic oscillator as a wigner quantum system .
although first introduced by wigner almost 60 years ago , and described in some articles and book chapters , it may still be useful to recall the main results . consider the hamiltonian for a one - dimensional harmonic oscillator ( in units with mass and frequency both equal to 1 ) : @xmath42 where @xmath43 and @xmath44 denote respectively the momentum and position operator of the system .
wigner @xcite already noted that there are other solutions besides the canonical one if one only requires the compatibility between the hamilton and the heisenberg equations .
thus , one drops the canonical commutation relation @xmath45 = -i$ ] ( we have taken @xmath46 ) , and instead one imposes the equivalence of hamilton s equations @xmath47 and the heisenberg equations @xmath48,\quad \dot { \hat q}= i[{\hat h } , { \hat q}].\ ] ] expressing this equivalence leads to the following compatibility conditions : @xmath49 = i{\hat q},\quad [ { \hat h } , { \hat q } ] = -i{\hat p}.\ ] ] so , one has to find operators @xmath43 and @xmath44 , acting in some hilbert space , such that the compatibility conditions hold , with @xmath50 given by .
also , in this hilbert space @xmath43 and @xmath44 have to be self - adjoint .
the solutions to can be found by introducing two new ( yet unknown ) operators @xmath51 and @xmath52 : @xmath53 or equivalently @xmath54 it is then easily checked that @xmath55 and that the compatibility conditions are equivalent with @xmath56 = \pm 2 { \hat b}^\pm .
\label{ccs2}\ ] ] the relations are in fact the defining relations of one pair of parabose operators @xmath57 @xcite .
moreover , from the self - adjointness of the position and momentum operators it follows that @xmath58 it is known that the lie superalgebra generated by two odd elements @xmath57 subject to the restriction is the lie superalgebra @xmath13 @xcite .
so , the solutions to our problem are given by the star representations of the lie superalgebra @xmath13 .
these are known , and are characterized by a positive real number @xmath11 and a vacuum vector @xmath59 , such that @xmath60 the representation space can then be shown to be the hilbert space @xmath61 with orthonormal basis vectors @xmath16 ( @xmath62 ) and with the following actions : @xmath63 from which it immediately follows that @xmath64 using the actions , one can now compute the action of the commutator @xmath45 $ ] : @xmath65|2n\rangle = -2ai|2n\rangle , \quad [ { \hat p},{\hat q}]|2n+1\rangle = -2(1-a)i|2n+1\rangle .
\label{action - comm - pq}\ ] ] so the canonical commutation relations are only satisfied for @xmath12 and any representation with @xmath66 thus leads to a non - canonical solution .
it is also important to note that for such non - canonical solutions , the commutator @xmath45 $ ] is not constant .
this last fact has a serious impact on the computation of exponential operators involving linear combinations of @xmath43 and @xmath44 since the baker - campbell - hausdorff theorem does not give rise to a simple formula .
the energy spectrum is now very easy to determine as well using and : @xmath67 so one has an equidistant energy spectrum with ground level given by @xmath11 .
this once again confirms that only @xmath12 yields the canonical solution .
also the wave functions for the parabose oscillator in the position and momentum representation are known .
working in the position representation , the operator @xmath44 is still represented by multiplication by @xmath10 , the operator @xmath43 however is no longer given by its canonical realization @xmath68 , but instead has a different realization with an extra term ( depending on @xmath11 ) coming in .
using this realization the time - independent schrdinger equation can then be solved yielding the following expressions for the orthonormal wave functions in terms of laguerre polynomials ( * ? ? ?
* chapter 23 ) and @xcite : @xmath69 again , for @xmath12 , one recovers the very well known expression for the wave functions for a harmonic oscillator in terms of hermite polynomials , since there are formulas relating @xmath70 to @xmath71 and @xmath72 respectively .
a major advantage of the phase - space formulation of quantum mechanics is the fact than one is no longer dealing with operators , but instead with constant - number equations .
more in particular , the expectation value of an arbitrary operator @xmath73 can be calculated using the distribution function @xmath74 as follows : @xmath75 with @xmath76 the scalar function obtained by replacing the operators @xmath43 and @xmath44 in the expression for @xmath73 by scalar variables @xmath77 and @xmath10 .
herein , @xmath78 is the density operator , representing the state of the system .
also , both integrations run from @xmath79 to @xmath80 ; this will be the case for all integrations in this article .
it is also known that the distribution function @xmath81 depends on the so - called correspondence rule used , this is , on the way one associates a function of non - commuting operators ( thus an operator ) to a given function of scalar variables .
a very popular association rule is the weyl correspondence , yielding as distribution function the so - called wigner distribution function . very often , the definition for the wigner d.f .
is written as follows : @xmath82 this definition , however , is completely unsuitable for the case of non - canonical quantum mechanics , since it relies heavily on the canonical commutation relations .
indeed , it was shown by tatarskii @xcite ( see also section 2.1 of lee s review paper @xcite ) that the wigner d.f .
defined by first principles ( see equation here ) gives rise to expression only if @xmath43 and @xmath44 satisfy @xmath45 = -i$ ] .
so if the canonical commutation relations are not satisfied , one should go back to use this first principles approach , which is described in the rest of this section .
another approach is the following @xcite : is it clear that in order to define an association rule between functions of scalar variables @xmath83 and operators @xmath84 it suffices to define this correspondence for the function @xmath85 , with @xmath86 and @xmath87 real variables .
indeed , an arbitrary analytical function @xmath83 can be obtained from @xmath88 in the following way : @xmath89 note that this expression is well - defined since the partial derivate operators commute .
the operator corresponding to the function @xmath83 under the chosen association rule , i.e. for the particular choice of @xmath90 , is then given by @xmath91 once a choice for @xmath90 is made , the corresponding distribution function @xmath92 may be calculated as follows : @xmath93 where @xmath94 denotes the state of the system .
this definition is independent of the canonical commutation relations and still allows the computation of averages by use of .
we now make the following choice for @xmath90 @xcite : @xmath95 where we have introduced the operator @xmath96 : @xmath97 since this is the only correspondence rule we are using in this article , the subscript @xmath98 will be dropped from the notation . on the other hand , as we will be concentrating on calculating explicit expressions for the distribution function @xmath99 for the pure states @xmath16 , we will introduce the subscript @xmath15 to indicate this : @xmath100 since in this article our goal will be to compute an explicit expression for , it is convenient to recall the expression for the wigner d.f .
@xmath101 in the case of canonical commutation relations : @xmath102 where @xmath103 are the laguerre polynomials .
so , whatever result we obtain for @xmath104 , replacing the parameter @xmath11 by @xmath105 should yield .
in this section , we compute an explicit expression for @xmath21 .
this will already show the techniques involved in the computation of the wigner d.f . for arbitrary states @xmath16 , but the resulting calculations are not quite so lengthy .
to reiterate , the goal is to compute @xmath106 thus , obviously , in a first step we are going to determine the matrix elements @xmath107 , and in a second step the integration will be performed . the operator @xmath108 being defined as @xmath109 we first try to find @xmath110 . to this end , we write @xmath111 in terms of @xmath57 : @xmath112 it is then immediately clear that @xmath113 since an odd number of applications of @xmath51 and @xmath52 on @xmath59 does not yield a term corresponding to @xmath59 in the resulting expansion , as can be seen from . on the other hand ,
using , one finds easily that @xmath114 this suggests , as will be proved later , that one has @xmath115 using and , one thus finds that @xmath116 where we have used the fact that @xmath117 .
this means that we now have : @xmath118 in order to compute this integral , it is useful to introduce a quadratic decaying exponential into the integrand by applying kummer s transformation on the confluent hypergeometric series .
one then has : @xmath119 where we have swapped summation and integration .
we assume that the parameter @xmath11 is such that this is possible . in any case ,
when the parameter @xmath11 is of the form @xmath120 this can certainly be done since one is then only dealing with finite summations . in appendix
a , it is shown that @xmath121 such that @xmath122 it is important to remark that , apart from the exponential factor , the expression is a polynomial in @xmath22 whenever the parameter @xmath11 is of the form . in this case
, it is possible to simplify the expression even further .
expand the laguerre polynomials as single sums and switch the two ( finite ) summations . on the resulting inner sum vandermonde s summation theorem can be applied .
after some further easy manipulations one then finds : @xmath123 clearly , in the canonical case ( i.e. @xmath124 ) the resulting expression reduces correctly to @xmath125 . to compute @xmath126
, we first have to find an expression for the matrix elements @xmath127 .
again , one finds the matrix elements with odd powers of @xmath96 to be zero . for the even powers
, one finds using and : @xmath128 indicating that @xmath129 since these matrix elements are the only place where the parameter @xmath11 appears , the expression for @xmath126 is also found by replacing @xmath11 with @xmath130 in the expression for @xmath21 : @xmath131
we now would like to find an explicit expression for the following integral : @xmath132 as is clear from the previous section , the key in calculating this integral is finding an expression for the matrix elements @xmath133 .
as before , matrix elements with odd powers of @xmath96 are easily seen to be zero . it will also be the case that matrix elements for a state @xmath134 will be determined from those with state @xmath135 simply by replacing @xmath11 with @xmath130 .
we will now give two different expressions for the matrix elements @xmath136 .
we will show that they are equivalent and will compute the wigner d.f . using each of them .
the proof that the proposed expressions are in fact the correct matrix elements is relegated to an appendix , despite the importance of the result to the calculations .
the proof , although in itself not difficult , requires rather lengthy calculations as one also needs the off - diagonal matrix elements .
basically , it involves checking that a certain expression satisfies the recurrence relation for the matrix elements ( as well as the correct boundary conditions ) .
we have the following expressions for the matrix elements @xmath136 : @xmath137 or alternatively @xmath138 note that both expressions reduce to for @xmath17 . to see that both expressions are indeed equivalent , it is easiest to write them in hypergeometric notation . for formula
we have @xmath139 whilst formula is @xmath140 so , we would like a transformation for hypergeometric series that relates a series of type @xmath141 with one of type @xmath142 . in krattenthaler s list
@xcite , transformation t3231 does just that and is indeed applicable to the @xmath141-series at hand .
it will turn out that we will need to apply yet another transformation , namely t4301 from the list , to the resulting series .
finally , a doubling formula for rising factorials , see e.g. ( * ? ? ?
* i.25 ) , has to be applied .
we thus have : @xmath143 this shows that formulas and are indeed equivalent .
the next step is to compute expressions for @xmath144 , by using the series expansion for the operator @xmath108 .
we first use : @xmath145 this simply follows from swapping the summation order and writing the inner sum again as a hypergeometric series .
the same procedure is applicable to formula , but here one gets : @xmath146 basically , all that is left to do is to integrate these expressions against the complex exponential @xmath147 . for the expression , we proceed as in the case of the ground state . we apply kummer s transformation to each of the confluent hypergeometric series .
one will then get a double summation over integrals of the type .
( here , we thus again assume that the order of summation and integration may be switched , which is certainly the case when the parameter @xmath11 is of the form . )
application of for each of the integrals in the series , will yield a triple sum expression for @xmath148 , with one summation hiddenin the laguerre polynomial .
explicitly , we have the following result : @xmath149 as a sanity check , we would like to see that for @xmath12 formula reduces to @xmath150 .
this means that the summation part of should equal @xmath151 .
this is not immediately clear , so we substitute @xmath12 and write the laguerre polynomial as an explicit summation and the triple sum then becomes : @xmath152 where the summation over @xmath153 is now performed first .
the gauss hypergeometric series , which terminates because either @xmath154 or @xmath155 is a non positive integer , can be summed in a unified way by applying transformation ( * ? ? ?
* t2140 ) first : @xmath156 apply this summation , and one gets a double summation .
the summation over the index @xmath157 can then be summed using the binomial theorem , and the result follows , showing that for @xmath12 expression reduces correctly to the known expression .
can we also do the integration starting from ?
clearly , we need a kummer - like transformation for @xmath158-series that introduces a quadratic decaying exponential . fortunately ,
such a transformation was published quite recently by paris @xcite .
his main transformation expresses one @xmath158-series as an exponential function multiplied by an infinite series containing @xmath158-series .
we , however , only need a simpler version where a finite sum of @xmath158-series is involved : @xmath159 as is stands , expression is a double sum . after applying
, it will be a triple sum , and since integrating using introduces an extra summation , the resulting expression will be a four - fold summation . since this computation is quite similar to the ones we have already done , we just state the result : @xmath160 there now exist quite a number of formulas relating sums of laguerre polynomials to another laguerre polynomial , one of them being the following @xcite : @xmath161 this identity is also valid for @xmath162 , provided one defines the laguerre polynomials as in .
application of this identity then leads immediately to : @xmath163 also in this case , reduction to the known result for @xmath12 can be verified , and is easier than in the previous case , since the identity needed is recorded as @xcite .
we thus have found closed expressions for @xmath148 namely and . from this
we now immediately find the corresponding expressions @xmath164 .
we state the following : @xmath165 this follows immediately from the fact , shown in appendix b , that @xmath166 together with the observation that this is the only place in which the parameter @xmath11 plays a role in the computation of the wigner distribution function .
parastatistics is one of the alternative approaches in particle statistics proposing that there are particles obeying statistics different from bose - einstein or fermi - dirac @xcite .
it involves an additional parameter ( often referred to as the order of statistics ) , and for one particular value of this parameter one recovers the common bose - einstein or fermi - dirac statistics . in the current case , where we just considered one parabose oscillator ,
this parameter is @xmath11 , and for @xmath12 one is reduced to the canonical quantum oscillator .
the simple case treated here can also be viewed as a deviation from canonical quantum mechanics .
the one - dimensional quantum oscillator is treated as a wigner quantum system , and allows several representations of @xmath13 labelled by a positive number @xmath11 . in some sense , @xmath11 ( or @xmath167 )
can be viewed as a quantity measuring the deviation from canonical quantum mechanics , see .
we have now obtained , for the first time , explicit expressions of the wigner d.f . for the parabose oscillator stationary states .
this allows one to consider `` phase space densities '' for a `` parastatistics quantum system '' or for a `` non - canonical quantum system '' .
let us now examine some plots of the wigner distribution functions .
first of all , note that the wigner d.f .
@xmath14 is , just as in the canonical case , a function of @xmath22 only .
so the 3d - plots of @xmath14 in @xmath168-space are rotational invariant .
this is why it is more convenient to give only two - dimensional plots of @xmath14 as a function of @xmath23 , where @xmath169 in figure 1 we have plotted the ground state wigner d.f .
@xmath21 of the one - dimensional parabose oscillator for values of the parameter @xmath11 equal to @xmath170 and @xmath171 ( or @xmath172 ) .
one can see that for @xmath12 ( @xmath124 , the well - known canonical case ) one finds the usual gaussian behaviour of the distribution function . as @xmath11 ( or @xmath20 ) increases , the shape changes quite significantly . for values of @xmath11
larger than @xmath105 , the shape of @xmath21 is rather similar to the shape of the wigner d.f .
for the first excited state of the canonical harmonic oscillator .
in fact , it follows from that @xmath21 for @xmath173 is equal to @xmath126 for @xmath12 , so in that specific case it is identical to the canonical wigner d.f.of the first excited state . as @xmath11 increases , the maximum value of @xmath21 moves further away from the origin .
the ground state of the parabose oscillator behaves as the first excited state of a canonical quantum oscillator but with increasing energy as @xmath11 increases . in figure 2
we plot the distribution functions @xmath14 for a fixed @xmath11-value ( namely @xmath173 ) , and for a number of stationary states @xmath174 ) .
note the separation of even and odd states near the origin .
the properties of the wigner distribution functions are in agreement with the `` wave - mechanical representation '' for the parabose oscillator , as discussed in ( * ? ? ?
* chapter 23 ) .
the wave functions in the position representations have also been given here in .
note that these wave functions are also the solutions of a wave equation of the form @xmath175 the even solutions are the functions @xmath176 with @xmath177 and @xmath178 , and the odd solutions are the functions @xmath179 with @xmath180 and @xmath178 . equation is often referred to as that of the non - relativistic singular harmonic oscillator , or of the calogero - sutherland model @xcite .
the singular behaviour of the last term of the hamiltonian in is once more in agreement with the negative values of the wigner distribution function near the origin . to summarize
, we have in this paper constructed the wigner distribution function for the one - dimensional parabose oscillator .
this is a quantum system for which the hamiltonian is given by the standard expression , and for which the position and momentum operators @xmath6 and @xmath7 do not satisfy the canonical commutation relations , but instead the more general compatibility conditions .
this non - canonical situation makes the definition ( and computation ) of the wigner function more difficult .
the approach of section 3 allowed us to define the right wigner function , satisfying all necessary properties .
the computation of the wigner function @xmath14 for stationary states is still quite involved , and uses some techniques from hypergeometric series .
the main results of the computation are given in , and .
on the basis of some plots , the behaviour of the wigner distribution function is further discussed in the context of parastatistics or in the context of the wave - mechanical representation of the parabose oscillator .
( @xmath181).,scaledwidth=90.0% ] , for fixed value of @xmath173 ( @xmath182 ) and @xmath183 .
, scaledwidth=90.0% ]
one of the authors ( e.j . ) would like to acknowledge that this work is performed in the framework of the fellowship 05 - 113 - 5406 under the intas - azerbaijan ys collaborative call 2005 .
another author ( s.l . ) was supported by project p6/02 of the interuniversity attraction poles programme ( belgian state belgian science policy ) .
we have to show that @xmath184 for @xmath185 and @xmath77 and @xmath10 real .
let us call the left hand side of the expression @xmath186 , and apply the binomial theorem to the factor @xmath187 : @xmath188 each of the two simple integrals above has a known expression in terms of hermite polynomials @xcite : @xmath189 yielding @xmath190 finally , there is a known summation formula for hermite polynomials @xcite : @xmath191 application of this summation formula immediately yields the desired result .
in this appendix , it is our aim to prove formula .
we will do so by given formulas for the matrix elements @xmath192 , with @xmath193 , which will be seen to reduce to formula when @xmath194 .
the proof relies on the fact that the formulas given below satisfy the same recurrence and the same boundary conditions as the matrix elements themselves . with @xmath96 as
before , one sees that @xmath195 and hence one gets , using and : @xmath196 so , if we now denote @xmath192 as @xmath197 , the recurrence that @xmath197 satisfies is : @xmath198 from the fact that we are dealing with orthonormal states @xmath16 , it follows immediately that @xmath199 and also from the actions and it is clear that @xmath200 these two boundary conditions determine the unique solution of the recurrence . we will now give explicit formulas that satisfy the boundary conditions and , and we will verify that these formulas also satisfy . the verification of that last fact will boil down to checking that two polynomials are equal . as a polynomial basis
, we will however not choose the classical basis @xmath201 , but rather we will work with the polynomials @xmath202 .
we now state the formulas , @xmath203 and @xmath204 where , in both cases , @xmath205 .
it is immediately clear that formulas and satisfy the boundary conditions and .
also , the reduction to for @xmath206 is readily seen . since the explicit formulas for the matrix elements are different depending on @xmath207 being positive or not , checking that and indeed satisfy will involve three cases : @xmath208 , @xmath194 and @xmath209 .
since all three are very similar , we will only deal with the case @xmath210 here .
we will abbreviate the summation in as @xmath211 , then substituting the expression in allows simplification of the powers of @xmath212 and @xmath213 as well as the square roots , and leads to the following that needs to be checked : @xmath214 this is a polynomial identity in the variable @xmath215 .
let us denote @xmath215 as @xmath216 .
the coefficients of the polynomials @xmath217 are immediately clear for the lhs . for the right hand side
, one can use the following easy identities : @xmath218 this allows to extract the coefficient of @xmath217 in the right hand side .
this coefficient is a rational function in the variables @xmath11 , @xmath157 , @xmath153 and @xmath207 .
after clearing the denominator and simplifying the remaining rising factorials , one ends up with a polynomial of a low degree that needs to be identically zero , and that is indeed the case .
so , although the checking is rather tedious , it is at the same time also simple and indeed shows that our formulas for the matrix elements are correct .
we end this appendix by noting that one indeed has : @xmath219 indeed , it is easily checked that the recurrence satisfied by @xmath220 is identical to but with @xmath11 replaced by @xmath130 .
also , the boundary conditions for @xmath221 are identical to and and as these do not involve @xmath11 the statement is now proved .
in this appendix , we are briefly going to discuss some aspects of the convergence of the main expressions and .
the main result is that expressions and converge absolutely when @xmath222 for all @xmath77 and @xmath10 . in the origin ,
i.e. @xmath223 , the condition @xmath222 is also necessary for convergence . for the odd states
, we thus have absolute convergence for the expressions derived from and everywhere for @xmath224 .
the only remaining question is thus what happens for the even states when @xmath225 in the case @xmath226 ( and @xmath66 ) . first ,
let us recall the well known fact that if one has two series of positive terms @xmath227 and @xmath228 such that @xmath229 , then convergence of @xmath228 implies that of @xmath227 .
secondly , there are some well known tests for convergence of positive series , one of them is raabe s test which can be used if the ratio test fails : let @xmath227 be a series of positive terms and let : @xmath230 one then has that @xmath231 converges if @xmath232 and diverges whenever @xmath233 . when @xmath234 , the test is inconclusive .
let us abbreviate @xmath22 as @xmath235 .
the convergence of then reduces to the convergence of @xmath236 next , we are going to use an inequality for the laguerre polynomials @xmath237 which gives an upper bound for @xmath238 not depending on @xmath157 @xcite ) : @xmath239 this gives us : @xmath240 the ratio test fails on the series @xmath241 , but raabe s test
gives : @xmath242 so clearly , the series @xmath241 converges if @xmath222 , and hence the series is absolutely convergent in that case .
note that the fact that @xmath241 diverges for @xmath243 does not imply anything about the convergence of .
the issue about the convergence of thus remains open for @xmath244 .
when @xmath245 , however , something more can be said : @xmath246 a gauss hypergeometric series @xmath247 converges if and only if @xmath248 @xcite , which in this case amounts to @xmath222 . for the convergence of we follow the same strategy .
first , it is clear that the convergence of @xmath249 is a sufficient condition for the convergence of since the summation over @xmath153 is finite .
then , we again use , and in this case we get @xmath250 raabe s test yields in this case that the series @xmath241 is convergent if @xmath251 . since @xmath252 ( see the summation bounds in ) , a sufficient ( and necessary ) condition for this to be true is @xmath222 . for @xmath245 ,
use of the convergence of the gauss hypergeometric series for @xmath253 again yields convergence if and only if @xmath222 .
finally , for the functions @xmath164 , we will thus have ( absolute ) convergence everywhere for @xmath224 . | in the beginning of the 1950 s , wigner introduced a fundamental deformation from the canonical quantum mechanical harmonic oscillator , which is nowadays sometimes called a wigner quantum oscillator or a parabose oscillator .
also , in quantum mechanics the so - called wigner distribution is considered to be the closest quantum analogue of the classical probability distribution over the phase space . in this article , we consider which definition for such distribution function could be used in the case of non - canonical quantum mechanics .
we then explicitly compute two different expressions for this distribution function for the case of the parabose oscillator .
both expressions turn out to be multiple sums involving ( generalized ) laguerre polynomials .
plots then show that the wigner distribution function for the ground state of the parabose oscillator is similar in behaviour to the wigner distribution function of the first excited state of the canonical quantum oscillator .
* the wigner distribution function for the one - dimensional parabose oscillator * + e. jafarov@xmath0 , s. lievens@xmath1 and j. van der jeugt@xmath2 + @xmath3 department of applied mathematics and computer science , + ghent university , krijgslaan 281-s9 , b-9000 gent , belgium + [email protected] [email protected] , [email protected] + @xmath4 institute of physics , azerbaijan national academy of sciences , + javid avenue 33 , az1143 , baku , azerbaijan + [email protected] + @xmath5 institute of mathematics , statistics and actuarial science + university of kent , cornwallis building , kent ct2 7nf , united kingdom + [email protected] _ keywords : _ wigner distribution function , non - canonical quantum mechanics , phase space , parabose oscillator + _ running head : _ wigner d.f . for the parabose oscillator |
the linear variation of mode frequencies with the changing solar magnetic activity is well established @xcite .
however , detailed studies based on high quality uniform data indicate the complexity of the relationship between mode frequencies and solar activity .
for example , @xcite have shown that there is a latitudinal variation in frequencies and splitting coefficients .
in addition , @xcite found a year - wise distribution in linear - regression slopes ( i e. the change in frequency per unit change in activity ) and the degree of correlation .
thus , with the availability of continuous data for a complete solar cycle , it is important to study correlations of oscillation frequencies with different measures of the solar activity in order to understand the source of the variability .
the analysis presented here uses oscillation data sets obtained from the global oscillation network group ( gong ) and michelson doppler imager ( mdi ) _ onboard solar and helipspheric observatory ( soho ) _ and solar activity data .
it covers a period of about 13 years , i.e. a complete solar cycle including both minimum phases at the beginning and end of solar cycle 23 .
the 128 108-day gong data sets spanning over the period from 1995 may 7 to 2008 feb 27 are continuous while the 61 72-day mdi data sets have two large gaps in 1998 - 1999 .
the mdi data cover the period from 1996 may 1 to 2008 sept 30 .
the activity data used here are : the integrated emission from the solar disk at 10.7 cm wavelength ( @xmath0 ) , the line - of sight magnetic field strength from kitt peak observatory ( kpmi ) , the international sunspot number ( @xmath1 ) , the mt .
wilson sunspot index ( mwsi ) , the magnetic plage strength index ( mpsi ) from mt .
wilson and the total solar irradiance ( tsi ) .
details of the various data sets are given in @xcite .
the number of _
p- _ modes analysed here are 479 for gong and 876 for mdi data sets in the frequency range 1.5 @xmath2 4.0 mhz .
these modes are observed in all data sets of gong or mdi .
we also analyze 76 _ f_-modes observed in all mdi data sets .
the frequency shifts ( @xmath3 ) are calculated with respect to the reference frequency which is determined by taking an average of the frequencies of a particular multiplet ( @xmath4 ) .
the temporal variation of gong frequency shifts ( @xmath3 ) with various measures of solar activity ( @xmath5 ) are shown in figure 1 .
it is evident that the frequency shifts follow the general trend of the solar activity .
the correlation coefficients between @xmath3 and @xmath5 obtained in all cases are comparable .
however , we find significant different correlation coefficients when we divide the activity cycle into four phases as shown in figure 2 ; the periods of minimum activity at the beginning and end of the solar cycle ( _ phase i _ ) , rising activity ( _ phase ii _ ) , high - acivity ( _ phase iii _ ) , and declining activity ( _ phase iv _ ) . in the right panel of figure 2 , we compare the phase - wise pearson s linear correlation coefficients ( @xmath6 ) for both gong and mdi data sets with those obtained for the complete cycle .
it is interesting to note that the correlation between @xmath3 and solar activity changes significantly from phase to phase ; the rising and declining phases are better correlated than the low- and high - activity phases .
the frequencies during _ phase i _ do not correlate well with any of the proxies . during _ phase iii _ , we obtain significant correlations for @xmath0 and kpmi while a substantial decrease is noticed for @xmath1 , mwsi and tsi .
figure 3 ( left panel ) shows the temporal variation of _
f_-mode frequencies .
we notice two distinct features in frequency shifts ; the persisting strong 1-year periodicity as discussed by several authors @xcite and the frequencies at current solar minimum ( 2007 - 2008 ) are lower than those at the previous minimum ( 1996 ) .
the correlation coefficients obtained with the original and smoothed frequency shifts are also shown in the right panel .
the smoothed frequency shifts are the running mean of five points to minimize the effect of 1-year periodicity .
it is seen that smoothing enhances the correlation in all cases .
the variation in correlation coefficients for _ f_-modes at different phases are consistent with those for
_ p_-modes . in both cases ,
mdi data sets have good correlation with tsi at the low - activity phase that requires a detailed investigation .
in summary , the improved and continuous measurements of intermediate - degree mode frequencies for a complete solar cycle demonstrate that , while the frequencies vary in phase with the solar activity , the degree of correlation between frequencies and activity indices differs from one activity phase to another .
although there is a strong correlation during rising and declining activity phases , we find a significant decrease in correlation at the low- and high - activity phases .
antia , h. m. , basu , s. , pintar , j. , & schou , j. 2001 , in esa sp-464 , : proceedings of soho 10/gong 2000 workshop : helio- and asteroseismology at the dawn of the millennium , ed .
a. wilson & p. l. palle ( noordwijk : esa publications ) 27 | we analyze intermediate degree _ p_- and _ f_-mode eigenfrequencies measured by gong and mdi / soho for a complete solar cycle to study their correlation with solar activity .
we demonstrate that the frequencies do vary linearly with the activity , however the degree of correlation differs from phase to phase of the cycle . during the rising and the declining phases ,
the mode frequencies are strongly correlated with the activity proxies whereas during the low- and high - activity periods , the frequencies have significantly lower correlation with all the activity proxies considered here . |
reliable simulations of biomolecules are one of nowadays grand challenges of computational science . in particular the problem of protein folding thermodynamics starting purely from an aminino - acid sequence has received major attention . until recently the prevailing view has been that it is elusive to search for the native states with present day simulational techniques due to limitations of time scale and force field accuracy .
now a barrier appears to be broken , as it was reported @xcite that large scale distributed computing allows to achieve folding of the 23-residue design mini - protein bba5 , which is relatively insensitive to inaccuracies of the force field .
the relaxation dynamics of the computer simulation is found to be in good agreement with the experimentally observed folding times and equilibrium constants .
molecular dynamics ( md ) technics , for a review see @xcite , tend to be the method of first choice for simulations of biomolecules .
one of the attractive features of md is that it allows to follow the physical time evolution of the system under investigation .
nevertheless , there has also been activity based on the metropolis method @xcite , which allows only for limited dynamical insights and has its strength in the generation of configurations which are in thermodynamical equilibrium .
a major advantage of metropolis simulations of biomolecules is that they allows for updates which are large moves when one has to follow dynamical trajectories .
such updates may help to overcome the kinetic trapping problem , which is due to a large number of local minima in the free energy space of typical biomolecules .
already metropolis simulations of the canonical gibbs - boltzmann ensemble may jump certain barriers . using _ generalized ensembles _ , which enlarge the gibbs - boltzmann ensemble and/or replace the canonical weights by other weighting factors , further progress has been made .
for reviews see @xcite . to my knowledge _ umbrella sampling _
@xcite was the first generalized ensemble method of the literature .
its potential for applications to a wide range of interesting physical problems remained for a long time dormant .
apparently , one reason was that the computational scientists in various areas shied away from performing simulations with an a - priori unknown weighting factor .
as li and scheraga put it @xcite : `` the difficulty of finding such a weighting factor has prevented wide applications of the `` umbrella sampling '' method to many physical systems . ''
this changed with the introduction of the multicanonical approach @xcite to complex systems @xcite .
besides having the luck that the controversy surrounding its first application got quickly resolved by analytical results @xcite , the multicanonical approach addressed aggressively the problem of finding reliable weights . nowadays
a starter kit of fortran programs is available on the web @xcite .
the replica exchange or parallel tempering ( pt ) method @xcite uses a generalized ensemble which enlarges the canonical configuration space by allowing for the exchange of temperatures within a set of canonical ensembles . in this and other generalized ensembles
the metropolis dynamics at low temperatures can be accelerated by excursions to disordered configurations at higher temperatures .
note that , in contrast to multicanonical simulations , the probabilities of canonically rare configurations are not enhanced by pt . for the purpose of simulating biomolecules
pt was , e.g. , studied in ref.@xcite . in this article
we discuss a biased updating scheme @xcite , which enhances already the dynamics of canonical metropolis simulations and is easily integrated into generalized ensemble simulations too .
the latter point is illustrated for parallel tempering .
the biased updating scheme is inspired by the funnel picture of protein folding @xcite .
at relatively high temperatures probability densities of the dynamical variables are estimated and used at lower temperatures to enhance the chance of update proposals to lie in the statistically relevant regions of the configuration space . in some sense
this is nothing else but an elaboration of the original importance sampling concept of metropolis et al.@xcite .
this article is organized as follows .
all - atom protein models , the funnel picture and the rugged metropolis updating are introduced in the next section . in the subsequent section numerical results
are presented for the brain peptide met - enkephalin .
a short summary and conclusions are given in the final section .
proteins are linear polymers of the 20 naturally occurring amino acids .
small proteins are called peptides .
the problem of protein folding is to predict ( at room temperature and in solvents ) the 3d conformations ( native structures ) from the sequence of amino acids .
a conformational energy function models the interactions between the atoms ( units kcal / mol ) : @xmath2 the contributions to the energy function are ( charges are in units of the elementary charge ) : @xmath3\ . \end{aligned}\ ] ] the solvent accessible surface method allows for an approximative inclusion of solvent interactions : @xmath4 where @xmath5 is the solvation parameter for atom @xmath6 and @xmath7 the conformation dependent solvent accessible surface area .
the @xmath8 are the distances between the atoms and the @xmath9 are the torsion angles for the chemical bonds . the parameters @xmath10 and @xmath11 are determined from crystal structures of amino acids and a number of thus obtained force fields are given in the literature .
bond length and angles fluctuate little and are normally set constant .
the important degrees of freedom are the dihedral angles @xmath12 and @xmath13 , see fig.[fig_dia ] . ] ] the funnel picture of bryngelson and wolyness @xcite gives qualitative insight into the process of protein folding , see fig.[fig_funnel ] , where a schematic sketch of the free energy versus a suitable _ reaction coordinate _ is given
. however , there is no generical definition of a good reaction coordinate , as the funnel lives in the high - dimensional configuration space .
here we give a parameter free funnel description @xcite from higher to lower temperatures , which suggests a method for designing the a - priori metropolis weights for simulations of biomolecules .
to be definite , we use the all - atom energy function @xcite ecepp/2 ( empirical conformational energy program for peptides ) .
our dynamical variables @xmath14 are then the dihedral angles , each chosen to be in the range @xmath15 , so that the volume of the configuration space is @xmath16 .
let us define the _ support _ of a pd of the dihedral angles .
the support of a pd is the region of configuration space where the protein wants to be .
mathematically , we define @xmath17 to be the smallest sub - volume of the configuration space for which @xmath18 holds . here
@xmath19 is a probability , which ought to be chosen close to one , e.g. , @xmath20 .
the free energy landscape at temperature @xmath21 is called _ rugged _ , if the support of the pd consists of many disconnected parts ( this depends of course a bit on the adapted values for @xmath22 and `` many '' ) .
that a protein folds at room temperature , say @xmath23 , into a unique native structure @xmath24 means that its pd @xmath25 describes small fluctuation around this structure .
we are now ready to formulate the funnel picture in terms of pds @xmath26 which are ordered by the temperatures @xmath27 , namely @xmath28 the sequence ( [ rho_r ] ) constitutes a protein _ funnel _ when , for a reasonable choice of the probability @xmath22 and the temperatures ( [ t_order ] ) , the following holds : 1 .
the pds are rugged .
the support of a pd at lower temperature is contained in the support of a pd at higher temperature @xmath29 e.g. for @xmath20 , @xmath30 and @xmath31 .
3 . with decreasing temperatures @xmath27 the support @xmath32 shrinks towards small fluctuations around the native structure .
properties 2 and 3 are fulfilled for many systems of statistical physics , when some groundstate stands in for the native structure .
the remarkable point is that they may still hold for complex systems with a rugged free energy landscape , i.e. , with property 1 added . in such systems
one finds typically local free energy minima , which are of negligible statistical importance at low temperatures , while populated at higher temperatures . in simulations at low temperatures the problem of the canonical ensemble approach
is that the updating tends to get stuck in those local minima .
this prevents convergence towards the native structure on realistic simulation time scales .
on the other hand , the simulations move quite freely at higher temperatures , where the native structure is of negligible statistical weight .
nevertheless , the support of a protein pd may already be severely restricted , as we shall illustrate .
the idea is to use a relatively easily calculable pd at a higher temperature to improve the performance of the simulation at a lower temperature .
the metropolis importance sampling would be perfected , if we could propose new configurations @xmath33 with their canonical pd @xmath34 . due to the funnel property
2 we expect that an _ estimate _
@xmath35 from some sufficiently close - by higher temperature @xmath36 will feed useful information into the simulation at temperature @xmath21 . the potential for computational gains is large because of the funnel property 3 .
the suggested scheme for the metropolis updating at temperature @xmath27 is to propose new configurations @xmath33 with the pd @xmath37 and to accept them with the probability @xmath38\ .\ ] ] this equation biases the a - priori probability of each dihedral angle with an estimate of its pd from a higher temperature . in previous literature
@xcite such a biased updating has been used for the @xmath39 theory , where it is efficient to propose @xmath40 at each lattice size @xmath6 with its single - site probability .
for our temperatures @xmath27 the ordering ( [ t_order ] ) is assumed .
with the definition @xmath41 the simulation at the highest temperature , @xmath42 , is performed with the usual metropolis algorithm .
we have thus a recursive scheme , called rugged metropolis ( rm ) in the following .
when @xmath43 is always a useful approximation of @xmath44 , the scheme zooms in on the native structure , because the pd at @xmath45 governs its fluctuations . to get things working , we need to construct an estimator @xmath46 from the numerical data of the rm simulation at temperature @xmath27 .
although this is neither simple nor straightforward , a variety of approaches offer themselves to define and refine the desired estimators . in the following we work with the approximation @xmath47 where the @xmath48 are estimators of reduced one - variable pds defined by @xmath49 the implementation of the resulting algorithm , called rm@xmath50 , is straightforward , as estimators of the one - variable reduced pds are easily obtained from the time series of a simulation .
the cpu time consumption of rm@xmath50 is practically identical with the one of the conventional metropolis algorithm .
to illustrate the developed ideas , we rely on the brain peptide met - enkephalin , which is a numerically well - studied @xcite .
met - enkephalin is determined by the amino - acid sequence tyr
gly gly phe met or , in the short notation , y g g f m , where @xmath51 our simulations are performed with a variant of smmp @xcite ( simple molecular dynamics for protein ) using fully variable @xmath52 torsion angles .
for the data analysis we keep a times series by writing out configurations every 32 sweeps . ]
fig.[fig_dia13 ] show the probability densities of the dihedral angle @xmath53 ( for the notation see table [ tab_vac_auto ] ) at 400@xmath54k and 300@xmath54k .
this angle is chosen because it illustrates the possibility of large moves in the metropolis updating , which jump barriers of a molecular dynamics simulation .
namely , a single update may take us directly from each of the three populated regions to each other , whereas one encounters barriers when one has to move in small increments @xmath55 . to evaluate the relative performance of different algorithms
, we measure the integrated autocorrelation times @xmath56 for the energy and each dihedral angle .
the integrated autocorrelation times are directly proportional to the computer run times needed to achieve the same statistical accuracy . for an observable @xmath57
the autocorrelations are @xmath58 where @xmath59 labels the computer time . defining @xmath60 , the time - dependent integrated autocorrelation time is given by @xmath61 formally the integrated autocorrelation time
@xmath56 is defined by @xmath62 .
numerically this limit can not be reached as the noise of the estimator increases faster than the signal .
nevertheless , one can calculate reliable estimates by reaching a window of @xmath59 values for which @xmath63 becomes flat , while its error bars are still reasonably small . .
integrated autocorrelation times .
[ tab_vac_auto ] [ cols="^,^,^,^,^,^,^,^,^",options="header " , ] columns 5 and 6 of table [ tab_vac_auto ] collect our estimates of the integrated autocorrelation times from conventional , canonical metropolis simulations at 400@xmath54k and 300@xmath54k .
we find a remarkable slowing down .
for the energy , which is characteristic for the over - all behavior , @xmath56 increases by a factor of about ten . for certain angles , e.g. @xmath64 , @xmath56 increases
even by factors larger than twenty . at 400@xmath54k conventional metropolis simulations allow to calculate observables with relative ease , while this is no longer the case at 300@xmath54k .
our aim is to use information from the 400@xmath54k simulation to improve on the performance at 300@xmath54k .
k and 400@xmath54k .
[ fig_energy ] ] one finds that the energy histograms of the 400@xmath54k and the 300@xmath54k simulation have a considerable overlap , see fig.[fig_energy ] .
therefore , the two temperatures are well - suited to be combined into a pt @xcite simulation .
we abstain here from introducing additional temperatures , because our aim is to get a clear understanding of the improvement of the 300@xmath54k simulation due to pt input from 400@xmath54k .
k. [ fig_auto ] ] in fig.[fig_auto ] integrated autocorrelations times for the energy variable are compared at 300@xmath54k .
the order of the curves agrees with the order in the figure legend . from up to down : conventional canonical metropolis simulation , rm@xmath50 improved canonical metropolis simulation with input from 400@xmath54k , pt simulation coupled to 400@xmath54k and the rm@xmath50 improved pt simulation .
the variable @xmath59 of equation ( [ tau_int_t ] ) is given in units of 32 sweeps due to the way our data are recorded . in the range shown , i.e. up to @xmath65 sweeps , a window exist for the pt and the rm@xmath50 improved pt simulations , which allows to estimate the integrated autocorrelation times of these simulations . for the other two simulations
one has to go to even larger @xmath59 values , but it remain possible to estimate @xmath56 . for the energy and all 24 dihedral angles the @xmath56 estimates are collected in table [ tab_vac_auto ] . for the energy we find a decrease from @xmath66 to @xmath67 due to the rm@xmath50 improvement of the canonical metropolis simulation , i.e. approximately a factor of 2 .
the pt improvement of this simulation is even larger , namely from @xmath66 to @xmath68 , i.e. approximately a factor of 2.5 . in both cases
the cpu time spent at 400@xmath54k is not part of the equation .
this is justified as one should anyway understand the system first at temperatures where one does not suffer from long autocorrelation times .
for pt the factor 2.5 is the improvement in real time when two identical pc nodes with some parallel software like mpi ( message passing interface ) are available .
most remarkably , the two improvements multiply : for the rm@xmath50 improved pt simulation the autocorrelation time is down to @xmath69 , only about one more factor of two away from the @xmath56 value at 400@xmath54k .
inspection of the @xmath56 values for the 24 dihedral angles shows large differences . for the conventional canonical simulation
the range is from means that our resolution of measurements every 32 sweeps is too crude to show the autocorrelations , which one may still expect on the scale of a few sweeps . ]
@xmath70 for @xmath71 to @xmath72 for @xmath73 .
for the pt - rm@xmath50 simulation it becomes reduced to a range from @xmath70 to @xmath68 .
this suggests that it is not efficient to simulate by sweeps , where each dihedral angle is updated once per sweep . instead
, an algorithm ( systematic or random ) where the number of updates per angle is proportional to its integrated autocorrelation time is expected to be more efficient .
obviously , this can be implemented without major changes of the existing code , but tests of this idea have not yet been completed .
after all these improvements the 300@xmath54k simulation is expected to be no more autocorrelated than the conventional , canonical simulation at 400@xmath54k .
all - atom metropolis simulations of small biomolecules deserve a place on their own in the arsenal of computational biophysics .
a good understanding of smaller pieces of a larger protein is one of the ingredients , which can pave the way towards the understanding of large systems like proteins .
major algorithmic problems remain to be solved before metropolis simulations of all - atom models may become directly suitable for applications like the folding of small to medium sized proteins .
one algorithmic problem is that of correlated moves of two or more dihedral angles .
such moves promise to overcome ( jump ) essential free energy barriers in the case of larger molecules . within conventional canonical metropolis simulations
the acceptance rates for simultaneous moves of two and more angles are prohibitively small .
the rm concept promises major inroads , but details have not yet been tested out .
an even more challenging problem is the inclusion of solvent effects . here
the large moves of metropolis updates create the _ cavity problem_. we need a cavity in the surrounding water to accommodate the move and updates of the dihedral angles do not create one .
this appears to be one of the major reasons why large scale simulations of proteins with a surrounding solvent are to an overwhelming extent done within the md framework , where the cavity problem is avoided due to joint , small moves of all degrees of freedoms . within the metropolis approach effective solvent models
may provide a solution .
they are , e.g. , defined by a solvent contribution like @xmath74 in eq.([e_sol ] ) .
unfortunately , no reliable determination of the parameters entering this equation exists presently in the literature , as was demonstrated by recent simulations @xcite .
however , it appears to be within the reach of metropolis simulations to lead the way towards determining reliable parameterizations .
* the rm@xmath50 approximation to the rugged metropolis ( rm ) method @xcite leads already to considerable improvements over conventional metropolis simulations of met - enkephalin at 300@xmath54k . as rm@xmath50 is easily implemented and needs no additional computer time , it should be used whenever metropolis simulations of suitable systems are done . * for larger systems one - variable moves alone will not work due to correlations between the dihedral angles .
the rm approach promises sufficiently large acceptance rates for multi - variable moves . *
ultimately , each biomolecule of interest may need its own , specifically designed , metropolis algorithm .
this task includes to determine reliable parameters for a solvent model like the one of eq.([e_sol ] ) .
i am indebted to yuko okamoto for many useful discussion and to robert swendsen for kindly informing me during this meeting about a dynamically optimized monte carlo method @xcite , which is tailored for the simulation of biomolecules .
further , i would like to thank james gubernatis and the other organizers of the los alamos metropolis workshop for their kind hospitality .
this work was partially supported by the u.s .
department of energy under contract no .
de - fg02 - 97er41022 .
bouzida , d. , kumar , s. , and swendsen , r. , _ phys .
a _ , * 45 * , 88948901 ( 1992 ) ; _ proceedings of the twenty - sixth annual hawaii international conference on system sciences _ , ed .
mudge , t.n . , milutinovic , v. , and hunter , l. ( ieee computer society press ) , pp.736742 ( 1993 ) . | metropolis simulations of all - atom models of peptides ( i.e. small proteins ) are considered .
inspired by the funnel picture of bryngelson and wolyness , a transformation of the updating probabilities of the dihedral angles is defined , which uses probability densities from a higher temperature to improve the algorithmic performance at a lower temperature . the method is suitable for canonical as well as for generalized ensemble simulations .
a simple approximation to the full transformation is tested at room temperature for met - enkephalin in vacuum .
integrated autocorrelation times are found to be reduced by factors close to two and a similar improvement due to generalized ensemble methods enters multiplicatively .
address=@xmath0department of physics , florida state university , tallahassee , fl 32306 , usa + @xmath1school of computational science and information technology + florida state university , fl 32306 , usa |
distributions of matter in form of shells have been a useful tool in astrophysics and general relativity , often providing simplified but analytically tractable models in cosmology , gravitational collapse and supernovae . in general relativity ,
the pioneering work on constructing solutions of the einstein field equations representing shells was done by israel @xcite . in his formalism , the shells are constructed by joining two manifolds across a three - dimensional time - like surface , and the matter representing the shell is interpreted as a discontinuity of the extrinsic curvature on the surface .
this formalism has been used to study several models of relativistic shells . @xcite
give a brief review of some of these models and also discuss shells which admit closed equations of state , including non - linear polytropic models .
an exact solution of a static shell surrounding a black hole was obtained by @xcite , and the stability of such a configuration against radial perturbations was studied by @xcite , while non - radial perturbations were analysed by @xcite .
non - radial perturbations of self - gravitating static fluid shells in the newtonian gravity were studied by @xcite . in this work ,
we propose a newtonian analytical potential - density pair that represents a family of spherical shells with varying thickness .
for this potential , we calculate the rotation curves of test particles and make a first study of stability of the shells by considering the stability of circular orbits of the test particles .
this is done in section [ sec_ns ] . in section [ sec_gr ]
, we consider a general relativistic version of the newtonian shells . for this task , we employ a particular isotropic form of a metric in spherical coordinates .
we calculate the components of the energy - momentum tensor of the shell s matter and also analyse the geodesic circular motion of test particles and discuss their stability .
finally , in section [ sec_dis ] we summarize our results .
in this section , we present a simple potential - density pair that describes a family of spherical shells of matter
. we shall consider the following potential : @xmath0 where @xmath1 is a parameter with dimensions of length and @xmath2 is a non - negative integer . by using the poisson equation in spherical coordinates ,
@xmath3 one obtains a mass density given by @xmath4 for @xmath5 , the pair ( [ eq_phi_n ] ) and ( [ eq_rho_n ] ) is reduced to the plummer model @xcite that has a monotone decreasing density profile .
however , for @xmath6 the density vanishes on @xmath7 and thus we have a shell - like distribution of matter . fig .
[ fig1](a ) shows some curves of the dimensionless mass density @xmath8 as function of @xmath9 for @xmath10 , @xmath11 and @xmath12 . for higher values of @xmath2 ,
the mass density becomes more concentrated ( thinner shell ) , with maximum at @xmath13 for large values of @xmath14 , the density ( [ eq_phi_n ] ) decays as @xmath15 , thus very fast for large @xmath2 , as can be seen in fig .
[ fig1](a ) .
( equation [ eq_rho_n ] ) , as function of @xmath9 for @xmath10 , @xmath11 and @xmath12 .
( b ) the circular velocity @xmath16 ( equation [ eq_vc_n ] ) , and ( c ) the angular momentum @xmath17 ( equation [ eq_h_n ] ) as functions of @xmath9 for @xmath10 , @xmath11 and @xmath12 . ] the circular velocity @xmath18 and angular momentum per unit mass @xmath19 of test particles moving in circular orbits are given by @xmath20 curves of the circular velocity @xmath16 and of the angular momentum @xmath17 for @xmath10 , @xmath11 and @xmath12 are depicted in figs [ fig1](b ) and ( c ) , respectively .
the stability of circular orbits can be determined using the rayleigh criterion of stability for a rotating fluid ( @xcite ; see also @xcite ) , and that can be adapted to study circular orbits of test particles in self - gravitating systems ( see e. g. @xcite for an application of the rayleigh criterion of stability in general relativity ) .
the condition for a stable orbit is @xmath21 using equation ( [ eq_h_n ] ) , it is straigthforward to check that the orbits are always stable .
it is instructive to consider the limit @xmath22 . in this case
, we have an infinitesimal thin shell on @xmath23 , and the limits of the potential ( [ eq_phi_n ] ) and circular velocity ( [ eq_vc_n ] ) are @xmath24 thus , for @xmath25 the potential is constant and a test particle is at rest inside the shell , and if @xmath26 the shell exerts on the test particle the same gravitational force as a point mass located on @xmath7 , in agreement with newton s shell theorem .
now , we study a general relativistic version of the newtonian potential - density pair discussed in the previous section . for this , it is convenient to choose a particular metric in isotropic form in spherical coordinates @xmath27 , which we write as @xmath28 where @xmath29 . in this particular form of metric
, the schwarzschild solution is given by @xmath30 .
a similar form of the metric ( [ eq_metric ] ) in cylindrical coordinates was used by @xcite to construct general relativistic versions of newtonian models for distributions of matter in galaxies . for metric ( [ eq_metric ] ) ,
the einstein equations @xmath31 yield the following expressions for the non - zero components of the energy - momentum tensor @xmath32 : @xmath33 \mbox{. } \label{eq_tthth}\end{gathered}\ ] ] because @xmath32 has only diagonal non - zero components , the energy density is directly given by @xmath34 and the pressures or tensions along a direction @xmath35 read @xmath36 .
the ` effective newtonian density ' @xmath37 reads @xmath38 the relation between the function @xmath39 and the newtonian potential @xmath40 is obtained by comparing equation ( [ eq_rho_n1 ] ) or ( [ eq_ttt ] ) with equation ( [ eq_poisson ] ) in the non - relativistic limit when @xmath41 , @xmath42 now , using equation ( [ eq_f_phi ] ) with the potential ( [ eq_phi_n ] ) , equations ( [ eq_ttt])([eq_rho_n1 ] ) can be cast as @xmath43 where the dimensionless variables and parameters are as follows : @xmath44 , @xmath45 , @xmath46 , @xmath47 , @xmath48 , and where we defined @xmath49 and @xmath50 .
we assume again that @xmath6 . for @xmath5 , equations ( [ eq_eps])([eq_rho_n2 ] ) reduce to the perfect fluid sphere derived by @xcite , which has as equation of state similar to the classical polytrope of index 5 .
note that the relation between @xmath51 and @xmath52 is independent of the radius and is simply @xmath53 . in order to be physically meaningful
, the components of the energy - momentum tensor should satisfy the energy conditions .
the strong energy condition states that @xmath54 , whereas the weak energy condition imposes the condition @xmath55 .
the dominant energy condition requires @xmath56 .
equations ( [ eq_eps])([eq_rho_n2 ] ) show that the weak energy condition is always satisfied , and the strong condition holds if @xmath57 .
this condition also ensures that we have pressures everywhere .
the inequality @xmath58 is satisfied when @xmath59 , and we have @xmath60 if @xmath61 .
thus , there exists an interval of the parameter @xmath62 where all energy conditions are satisfied .
( equation [ eq_eps ] ) , ( b ) the effective newtonian density @xmath47 ( equation [ eq_rho_n2 ] ) , ( c ) the radial pressure @xmath63 ( equation [ eq_p_r ] ) and ( d ) the azimuthal / polar pressure @xmath64 ( equation [ eq_p_phi ] ) as functions of @xmath44 for @xmath10 , and parameter @xmath65 , @xmath66 and @xmath67 . ] ( equation [ eq_eps ] ) , ( b ) the effective newtonian density @xmath47 ( equation [ eq_rho_n2 ] ) , ( c ) the radial pressure @xmath63 ( equation [ eq_p_r ] ) and ( d ) the azimuthal / polar pressure @xmath64 ( equation [ eq_p_phi ] ) as functions of @xmath44 for @xmath11 , and parameter @xmath68 , @xmath67 and @xmath69 . ] figs [ fig2](a)(d ) display , respectively , some curves of the energy density @xmath70 , the effective newtonian density @xmath71 , the radial pressure @xmath51 and the azimuthal / polar pressure @xmath52 as functions of the radial coordinate for the shell with @xmath10 and parameter values @xmath72 , @xmath66 and @xmath67 . in figs [ fig3](a)(d ) , the same quantities are displayed for the shell with @xmath11 and parameter values @xmath73 , @xmath67 and @xmath69 .
all quantities have the profile of a shell - like distribution of matter . as the value of the parameter @xmath62
is increased , the densities and pressures are smoothed out .
the densities and pressures become more concentrated when @xmath2 is increased . as we did in the newtonian model
, we also study the circular orbits and stability of test particles . without loss of generatity , if the geodesic motion is confined on the @xmath74 plane , the expressions for the circular velocity and for the angular momentum are formally the same as in cylindrical coordinates , and we have @xcite @xmath75 } } \mbox{. } \label{eq_h_r1}\end{gathered}\ ] ] for the shell potential ( [ eq_phi_n ] ) , we get @xmath76 where @xmath77 , @xmath78 and @xmath50 .
( equation [ eq_vc_r2 ] ) , as a function of @xmath44 for the relativistic shell model with ( a ) @xmath10 and ( b ) @xmath11 .
( c)(d ) the angular momentum @xmath78 ( equation [ eq_h_r2 ] ) , as a function of @xmath44 for the relativistic shell model with ( c ) @xmath10 and ( d ) @xmath11 . ]
some curves of the circular velocity @xmath77 and angular momentum @xmath78 for relativistic shells are shown in figs [ fig4](a)(d ) . in figs [ fig4](a)(b ) , the curves of rotation are displayed for the shells with @xmath10 and @xmath11 , respectively , and the corresponding curves of angular momentum are plotted in figs [ fig4](c)(d ) . as the values of the parameter @xmath45 is increased , the velocities becomes less relativistic .
the minimum value of @xmath62 such that the rotation curve is sublumial has to be found numerically .
table [ tab1 ] shows some approximate minimum values for @xmath62 for shells with @xmath10 to @xmath12 .
note that these minimum values are always larger than those required by the energy conditions .
.approximate minimum values of the parameter @xmath45 that ensures sublumial rotation curves for some values of @xmath2 .
[ cols="<,^,^,^,^,^,^,^ " , ] as functions of @xmath79 and @xmath62 for circular geodesic orbits in shells with @xmath10 to @xmath11 . ]
the curves of angular momentum in figs [ fig4](c)(d ) show regions of instability at high relativistic velocities . in the newtonian limit ,
the orbits are always stable , thus this instability is a purely relativistic effect . using the rayleigh criterion of stability , curves of @xmath80 as functions of @xmath79 and @xmath62 are shown in fig .
[ fig5 ] for models of shells with @xmath2 from @xmath10 to @xmath11 . in each graph
, the minimum value of @xmath62 is that given in table [ tab1 ] . in the inner part of the curves ,
orbits are unstable .
the regions of instability become larger as the parameter @xmath2 is increased .
we proposed a very simple potential - density pair in newtonian gravity that describes a family of spherical shells whose thickness is function of a parameter @xmath2 . in the limit @xmath22 ,
one obtains an infinitesimal thin shell .
this family of shells is found to be stable by the rayleigh criterion of stability .
we also studied a general relativistic version of the newtonian model of shells by using a particular metric in isotropic form in spherical coordinates .
these relativistic shells have radial pressures that are different from the equal azimuthal and polar pressures , but are proportional to for a fixed value of @xmath2 .
also , parameters can be chosen such that the matter in the shells satisfies all the energy conditions .
the rayleigh criterion of stability shows that the relativistic shells have regions of unstable circular orbits of test particles , and this instability is a pure relativistic effect .
d. v. thanks fapesp for financial support , and p. s. l. thanks fapesp and cnpq for financial support .
99 israel w. , 1966 , nuovo cimento b , 44 , 1 nuovo cimento b , 48 , 463 ( erratum ) .
kijowski j. , magli g. , malafarina d. , 2006 , gen .
gravitation , 38 , 1697 .
frauendiener j. , hoenselaers c. , konrad w. , 1990 , classical quantum gravity , 7 , 585 .
brady p.r . , louko j. , poisson e. , 1991 , phys . rev .
d , 44 , 1891 .
schmidt b.g . , 1998 ,
d , 59 , 024005 .
bik j. , schmidt b.g .
, 1999 , apj , 521 , 708 .
plummer h.c . , 1911 , mnras , 71 , 460 .
binney j. , tremaine s. , 2008 , galactic dynamics , 2nd edn .
princeton univ . press ,
princeton , nj . rayleigh j.w.s . , 1917 ,
london a , 93 , 148 .
landau l.d .
, lifshitz e.m . , fluid mechanics , 2nd edn .
pergamon press , oxford .
letelier p.s . , 2003 ,
d , 68 , 104002 .
vogt d. , letelier p.s . , 2005 , mnras , 363 , 268 .
buchdahl h.a . , 1964 ,
apj , 140 , 1512 . | a family of spherical shells with varying thickness is derived by using a simple newtonian potential - density pair .
then , a particular isotropic form of a metric in spherical coordinates is used to construct a general relativistic version of the newtonian family of shells .
the matter of these relativistic shells presents equal azimuthal and polar pressures , while the radial pressure is a constant times the tangential pressure .
we also make a first study of stability of both the newtonian and relativistic families of shells .
_ keywords _ : gravitation . |
cerium , the first of the lanthanide series ( rare earths ) with an inner 4f - electron exhibits a broad range of unusual electronic , magnetic and structural properties in the solid state @xcite .
most intriguing is the transformation under pressure @xcite between two face - centered - cubic ( fcc ) phases called @xmath6-ce and @xmath7-ce . while @xmath6-ce is stable at room temperature ( @xmath8 ) and ambient pressure ( @xmath9 ) , it transforms to @xmath7-ce at @xmath10 gpa .
this apparent isostructural phase transition @xmath11 is accompanied by a @xmath1215% volume reduction and a change in magnetism . while in the @xmath6-phase the magnetic susceptibility @xmath13 follows a curie weiss law which is attributed to the magnetic moment of the 4f - electron , it is a quasi @xmath8-independent in the @xmath7-phase and reminiscent of pauli paramagnetism @xcite .
the isostructural nature of the @xmath11 transition is at variance with the landau theory of phase transitions @xcite which would require a change of space group symmetry . since the establishing of the fcc structure for both phases @xcite various microscopic theories have been proposed . nowadays
two theoretical models are well represented in the literature .
a mott - transition scenario for 4f - states is based on the concept of a localized non bonding state of the 4f - electron in @xmath6-ce and an extended metallic bonding state in the @xmath7-phase @xcite .
the increased binding capability due to the metallic f state is considered then as the reason for the marked volume reduction in the @xmath7-phase @xcite . in
the following this model has been supported by band structure calculations @xcite . on the other hand
more recent inelastic neutron scattering experiments on @xmath7-ce found that the magnetic form factor is quite different @xcite from the one calculated @xcite under the assumption that the 4f states are itinerant .
the absence of extended 4f states in the solid is also supported by many - electron calculations on the ce dimer @xcite where it is found that the 4f - electrons do not participate in the chemical bond .
the second widely recognized theoretical model is based on spin fluctuations .
these are the driving mechanism in the kondo - volume - collapse ( kvc ) model of the @xmath11 transition @xcite . in first approximation
the 4f - electron stays localized in both phases .
however in the @xmath7-phase the hybridization between the 4f - electrons and the conduction electrons ( anderson impurity hamiltonian , see e.g. ref . )
is much more intense than in the @xmath6-phase .
this interaction leads to a screening of the 4f local moments in the energy ground state .
the volume dependence of the hybridization coupling or equivalently of the kondo - temperature @xmath14 is used to obtain the equation of state .
the interpretation of the @xmath11 transition in ce within the kvc scenario @xcite has found strong support by the analysis of electron spectroscopy data @xcite .
notwithstanding much effort , both the mott transition and the kvc model remain under debate @xcite and there is no consensus on the driving mechanism of the phase transition @xcite . a common feature is the absence of any symmetry breaking at the transition as would be required by landau theory @xcite . on the basis of thermodynamical data it has been suggested @xcite that the @xmath11 transformation is in fact a first order phase transition which becomes of second order beyond a tricritical point .
therefore , @xmath7-ce should have lower symmetry than @xmath6-ce .
however a distorted lattice has been discarded by x - ray diffraction experiments @xcite .
a mechanism of symmetry lowering without lattice distortion ( positions of the ce nuclei fcc in the @xmath6 and @xmath7 phase ) has been suggested by the present authors @xcite .
the main concepts are the support of quadrupolar electronic charge - density fluctuations by the 4f - electrons in the @xmath6-phase on a compressible fcc lattice and the collective orientational ordering of the quadrupolar densities in the @xmath7 phase @xcite on four simple cubic ( sc ) sublattices .
the space group symmetry lowering at the transition @xmath11 is @xmath0 , the local density symmetry is @xmath15 .
it is accompanied by a uniform lattice contraction so that the fcc structure is conserved .
although this latter aspect is isostructural " , the symmetry lowering of the electronic structure is fully consistent with landau theory . in reciprocal space
the active electronic density mode condenses at the @xmath16 point of the brillouin zone .
a phase transition @xmath0 occurs in solid c@xmath5 ( fullerite ) @xcite . here
the symmetry lowering is due to the orientational ordering of the c@xmath5 molecules . at the transition
the cubic lattice constant contracts discontinuously @xcite , the center of mass points of the molecules still occupy an fcc lattice . from the latter point of view , the phase transition would be isostructural " .
however , the molecular order on four sc sublattices entails an order of the constituent c atoms and hence the @xmath2 structure leads to characteristic reflections in x - ray @xcite and neutron scattering @xcite experiments .
the problem of measuring the @xmath2 structure directly in ce is complicated by the fact that the ce nuclei remain on an fcc lattice .
although the quadrupolar charge density ordering could be detectable by synchrotron radiation , the intensity of the additional @xmath2 reflections is likely to be too weak and has not yet been uncovered @xcite . on the other hand
there have been two recent experimental results which are very specific and point to the relevance of the quadrupolar ordering at the @xmath11 transition .
time - differential perturbed angular correlation ( tdpac ) experiments in solid ce in a pressure range up to 8 gpa detect an appreciable electric field gradient ( efg ) in @xmath7-ce which is almost four times larger than in the cubic @xmath6 phase and close to values in the noncubic phases @xmath17 and @xmath18 @xcite .
this finding rules out the @xmath19 symmetry in @xmath7-ce and evidences in support of the antiferroquadrupolar order suggested by theory @xcite .
the other experimental support of the antiferroquadrupolar ordering is more indirect but very specific too .
phonon dispersion measurements by inelastic x - ray scattering on elemental ce across the @xmath11 transition reveal strong changes in the dispersion shape @xcite .
in particular a pronounced softening of certain phonon branches is found in the @xmath7-phase toward the @xmath16 point of the brillouin zone .
given the above mentioned fact that the discussion on the validity of the mott transition or kvc scenario remains open , there have been in the last decade increased experimental efforts in investigating the lattice dynamics at the phase transition @xmath11 @xcite . from these studies
it results that the lattice vibrations play an important role at the @xmath11 phase transition and that the inclusion of lattice dynamics in the theoretical description is of paramount importance . in this respect
we recall that anomalous elastic behavior at the @xmath11 transition was originally discovered in ultrasound experiments @xcite . from inelastic neutron scattering experiments
@xcite performed on single crystal of @xmath6-ce it was concluded that premonitory effect of the @xmath20 transition are present in the phonon dispersion curves of @xmath6-ce at room temperature .
however , until recently the relevance of these results has not been sufficiently appreciated , while all research efforts were concentrated on the electronic properties .
the anomalies of the elastic properties at the @xmath11 transition in ce [ 34,21,33 ] have their counterpart at the fcc@xmath21sc transition in c@xmath5 fullerite @xcite .
in particular there is a striking resemblance of the pressure dependence of the bulk modulus in ce with corresponding experimental results in c@xmath5 @xcite .
the content of the paper is as follows . in sec .
[ sec : model ] we recall the main features of the model of interacting 4f - electron quadrupolar densities on a compressible fcc lattice .
the hamiltonian and the resulting condensation scheme from the disordered @xmath6-phase to the quadrupolar ordered @xmath7-phase are described .
the coupling to the lattice is included .
next ( sec .
[ sec : phase_trans ] ) we extend the model by adding an external hydrostatic pressure . various quantities that characterize the phase transition ( order parameter , susceptibility and correlation length ) are derived as functions of temperature and pressure . in sec .
[ sec : el_res ] we calculate the dynamic displacement displacement correlation function , taking into account the coupling of the quadrupolar electron densities to the crystal lattice . in the static limit the inversion of the displacement
displacement correlation function tensor leads to the elastic constants c@xmath22 and @xmath4 .
while in c@xmath22 the coupling to quadrupolar electron density fluctuations leads to remarkable anomalies at the @xmath1 phase transition , the coupling is absent in c@xmath23 .
the results of the theory are compared with experiments .
conclusions of the paper are presented in sec .
[ sec : con ] .
here we will remind the model hamiltonian which describes electronic quadrupolar charge density fluctuations on a compressible fcc lattice . in previous work
we have first treated quadrupolar charge fluctuations due to solely 4f - electrons @xcite . later on we have taken into account conduction electrons @xcite as well as f- and d - electron intra site correlations @xcite .
since these extensions can be cast into an effective hamiltonian with essentially the same structure as the one originally studied @xcite , we will restrict ourselves here to 4f - electrons as a generic case .
we consider @xmath24 ce atoms located on a non rigid fcc lattice with nuclear positions @xmath25 , @xmath26 , 2 , ... , @xmath24 . here
@xmath27 denote the equilibrium positions and @xmath28 the displacements due to lattice vibrations .
we assume that core and valence electrons follow adiabatically the nuclear displacements .
the hamiltonian comprises three parts : @xmath29 where @xmath30 stands for the 4f - electrons , @xmath31 for the lattice and @xmath32 for the coupling of both .
explicitly one has @xmath33 where @xmath34 is the single particle potential and @xmath35 the quadrupole - quadrupole interaction on the rigid lattice .
we assume that the 4f - electrons have coordinates @xmath36 and are localized on spheres centered at @xmath27 and with radii @xmath37 , @xmath381.378 a.u .
the single particle potential is due to the cubic crystal field in presence of spin - orbit coupling .
one obtains @xmath39 where @xmath40 and @xmath41 are the eigenvalues and eigenstates , @xmath42 , 2 , ... , 14 .
the angular part of the wave function of the 4f - electron ( angular momentum quantum number @xmath43 ) at site @xmath44 is described by the @xmath45 spin orbitals @xmath46 , @xmath47 , where @xmath48 .
for details see refs . .
the electronic quadrupolar interaction is given by @xmath49 where @xmath50 , @xmath511 , 2 , 3 are quadrupolar density operators and @xmath52 interaction matrix elements between nearest neighbors on the fcc lattice .
the density operators are defined by @xmath53 with @xmath54 here we restrict ourselves to the three quadrupolar functions @xmath55 , @xmath56 , 2 , 3 , which transform as the irreducible representation @xmath57 of the cubic point group @xmath58 ( @xmath59 ) .
these symmetry adapted functions ( saf s ) are linear combinations of spherical harmonics belonging to the manifold @xmath60 and are tabulated in ref . .
they transform as the cartesian components @xmath61 , @xmath62 , @xmath63 for @xmath56 , 2 , 3 , respectively . defining fourier transforms
@xmath64 @xmath65 we obtain @xmath66 with @xcite @xmath67 } \nonumber \\
\label{m10}\end{aligned}\ ] ] here @xmath6 , @xmath68 and @xmath7 are quadrupole - quadrupole interaction coefficients on the fcc lattice .
the quantities @xmath69 , and @xmath70 , @xmath71 , @xmath72 is the cubic lattice constant , account for the fcc lattice structure . at the @xmath16 point of the brillouin zone for @xmath73 ,
the interaction matrix @xmath74 becomes diagonal .
the star of @xmath75 has the arms @xmath76 , @xmath77 , @xmath78 . for @xmath79 one
has @xmath80 , and similarly for @xmath81 and @xmath82 with permutation of indices @xmath83 .
hence the quadrupolar interaction becomes attractive with the largest twofold degenerate eigenvalue @xmath84 at each arm of @xmath85 .
as we have shown previously @xcite this interaction is compatible with a symmetry lowering @xmath0 , characterized by an orientational order of the quadrupolar densities on four different sc sublattices ( see fig . [ fig0 ] ) .
the corresponding condensation scheme reads : @xmath86 here the superscript @xmath87 stands for a thermal expectation value and @xmath88 is the order parameter amplitude .
since the three arms of @xmath85 are involved one speaks of a triple-@xmath89 antiferroquadrupolar order ( afq ) .
notice that a condensation scheme similar to eqs .
( [ m11a ] ) , ( [ m11b ] ) also holds for the phase transitions with the symmetry @xmath0 in nao@xmath90 @xcite and in c@xmath5-fullerite @xcite .
although icosahedral symmetry of the c@xmath5 molecule implies that not quadrupoles but higher order multipoles ( @xmath91 , 10 , ... ) determine the orientational interactions , the corresponding saf s transform as irreducible representations of symmetry @xmath57 of the cubic point group . a compelling mathematical reason to consider the symmetry lowering @xmath0 as a candidate for the isostructural phase transition in ce is the fact that it leads to a lattice contraction while the center of mass points still occupy an fcc lattice . in addition
the transition is of first order . indeed
the number of symmetry elements is thereby reduced by a factor 3 ( from 48 to 16 ) , which implies the existence of a third order cubic invariant in free energy @xcite .
the lattice dynamics is described by the phonon hamiltonian @xmath92 where @xmath93 is the kinetic energy and @xmath94 the potential energy in harmonic approximation . in fourier space
one has @xmath95 @xmath96 .
the displacements @xmath97 and the conjugate moments @xmath98 are related to the variables in real space by @xmath99 @xmath100 where @xmath101 is the atomic mass .
one has the usual commutation rules
@xmath102=[p , p]=0 $ ] and @xmath103 = i \hbar \delta_{\vec{q } \vec{k } } \delta_{ij } .
\label{m15}\end{aligned}\ ] ] in the long wavelength limit the dynamical matrix @xmath104 is given by @xmath105 } \nonumber \\
\label{m16}\end{aligned}\ ] ] where @xmath106 are the bare elastic constants in absence of coupling to the quadrupolar electronic fluctuations .
the coupling between quadrupolar charge density fluctuations and lattice dynamics has been derived previously @xcite . in the long wavelength regime for the lattice displacements and for quadrupolar fluctuations near the @xmath16 point of bz
we have @xmath107 here the prime on the sum over @xmath7 indicates the following restrictions : for @xmath108 , @xmath109 ; @xmath110 , @xmath111 ; @xmath112 , @xmath113 .
the coupling matrix is given by @xmath114 here @xmath115 , @xmath116 refers to the 12 nearest neighbors of @xmath44 on the fcc lattice , @xmath117 is the first order derivative with respect to the lattice displacement component @xmath118 of the quadrupole - quadrupole interaction @xmath119 .
the structure of the fcc lattice implies that @xmath120 , where for @xmath108 or @xmath121 , @xmath122 ; for @xmath123 or @xmath124 , @xmath125 ; for @xmath112 or @xmath121 , @xmath126 . within the quadrupolar model we obtain that @xmath127 . carrying out the summation over @xmath128 and exploiting the symmetry of the lattice , we rewrite eq .
( [ m17 ] ) as @xmath129 under the proviso of the summation restriction over @xmath7 we define for @xmath130 or @xmath131 @xmath132 and hence @xmath133 only longitudinal lattice displacements or equivalently longitudinal lattice strains
@xmath134 occur on the right hand side of eq .
( [ m19 ] ) . since in addition @xmath135 for @xmath130 and
@xmath131 occur on the same footing , the coupling @xmath32 leads to a striction of the lattice with conservation of cubic symmetry @xcite .
notice that there is no coupling to transverse lattice displacement waves or equivalently to shear strains @xmath136 , @xmath137 , in @xmath32 . in the next section
we will investigate the influence of the quadrupolar density fluctuations on the low frequency lattice dynamics .
in order to study the tripple-@xmath89 antiferroquadrupolar phase transition on a compressible lattice as a function of temperature and pressure , we first recall some concepts of the underlying free energy .
the phase transition is of first order .
we calculate the order parameter as a function of temperature and pressure .
we show that the transition temperature increases linearly with pressure .
finally we study the order parameter susceptibility in the disordered and the ordered phase . taking into account the quadrupole - quadrupole interaction on a rigid fcc lattice we have written the helmholtz free energy as a landau expansion in terms of the order parameter amplitude @xmath88 @xcite : @xmath138 here
@xmath139 is the free energy of the disordered phase which has to be calculated with the cubic crystal field @xmath34 , eq .
( [ m3 ] ) .
the coefficients of the order parameter terms are @xmath140 , \label{fe23 } \\ b & = & -t\,x_{123}^{(3)}\,/[x^{(2)}]^3 , \label{fe24 } \\
c & = & { \frac{t}{8(x^{(2)})^4}}\ , \left [ \,9(x^{(2)})^2-x_{1111}^{(4)}-6x_{1122}^{(4)}+ { \frac{24(x_{123}^{(3)})^2}{x^{(2 ) } } } \right ] , \nonumber \\ \label{fe25}\end{aligned}\ ] ] where @xmath141 is the quadrupolar interaction and @xmath8 the temperature in unergy units ( @xmath142 ) .
the quantities @xmath143 , @xmath144 and @xmath145 are single particle thermal expectation values @xcite , which are calculated by means of the single particle potential .
numerical values of relevant parameters are given in table 1 . .
parameters @xmath143 , @xmath144 , @xmath6 , @xmath146 , @xmath147 , @xmath148 , @xmath149 , @xmath150 calculated from ref . .
@xmath151 and @xmath152 estimated from experiment @xcite .
[ table1 ] [ cols="^,^ " , ] in case of a compressible lattice @xcite the hamiltonian @xmath32 , eq .
( [ m21 ] ) , gives the free energy term @xmath153 , \label{fe28}\end{aligned}\ ] ] where @xmath154 etc .
are the lattice strains .
since @xmath127 , this term favors @xmath155 , i.e. a volume contraction .
the lattice hamiltonian @xmath31 leads to the elastic contribution @xmath156 , \label{fe29}\end{aligned}\ ] ] where @xmath157 is the volume per atom .
in presence of an applied external pressure @xmath9 , we consider the gibbs free energy @xcite @xmath158 minimizing @xmath159 with respect to @xmath134 for @xmath130 and @xmath131 , we get an isostructural contraction of the cubic lattice @xmath160 \kappa_l .
% ( 29 ) \label{fee29}\end{aligned}\ ] ] here @xmath161 is the compressibility .
isostructural contraction of the lattice means that there is no symmetry breaking associated with the change of the center of mass positions of the ce atoms at the @xmath11 transition .
we completely agree with conclusions from high - pressure and high - temperature x - ray diffraction experiments that the structure remains fcc across the @xmath11 transformation and retains crystallographic orientation during the transformation @xcite . within the present theory
a symmetry change occurs solely in the quadrupolar electronic charge densities whereby the @xmath2 ordering of the latter is compatible with the fcc structure of the lattice . eliminating @xmath134 from eq .
( [ fe30 ] ) and retaining only linear terms in @xmath9 we rewrite @xmath159 as @xmath162/n = f_0/n + a ' \rho^2 + b \rho^3 + c'\rho^4 , \nonumber \\ % ( 30 ) \label{fe31}\end{aligned}\ ] ] where @xmath163 , % ( 31 ) \label{fe32a}\end{aligned}\ ] ] with @xmath164 and where @xmath165 with @xmath166 .
the occurrence of the cubic invariant in the free energy implies that the phase transition is of first order . the coexistence condition @xmath167 establishes a relation between the transition temperature @xmath168 and the corresponding pressure @xmath151 : @xmath169 .
% ( 34 ) \label{fe33}\end{aligned}\ ] ] in first approximation @xmath168 increases linearly with pressure , in agreement with the experimental @xmath170 phase diagram of ce @xcite .
notice that the analogue of eq .
( [ fe33 ] ) also holds for c@xmath5 fullerite @xcite , where experiments @xcite show a linear pressure dependence of @xmath168 . with the parameters of table [ table1 ] we obtain @xmath171 k. in accordance with the experimental situation
@xcite we consider a fixed temperature @xmath172 and a variable pressure .
then @xmath173 , @xmath174 , corresponds to the disordered phase ( in casu @xmath6-ce ) , while @xmath175 , @xmath176 corresponds to the ordered phase ( @xmath7-ce ) .
minimization of @xmath159 at @xmath168 and @xmath177 with respect to @xmath88 leads to the order parameter amplitude @xmath178 since @xmath179 , the order parameter @xmath88 has to be negative and hence the @xmath180 sign has to be chosen in front of the square root in eq .
( [ fe33b ] ) .
the discontinuity of the order parameter on the transition line , eq .
( [ fe33 ] ) , is given by @xmath181 where we remind that @xmath168 is an implicit function of @xmath151 .
with the parameters of table [ table1 ] we obtain @xmath182 . here and in the following
we assume that the quantities @xmath183 , @xmath184 and @xmath185 , all which are single expectation values , are not affected by moderate variations of pressure near @xmath151 . in the ordered phase we obtain by means of eqs .
( [ fe33b ] ) and ( [ fe33 ] ) : @xmath186 where @xmath187 . since @xmath188 , the order parameter amplitude increases in absolute value with the increment of pressure above @xmath151 .
[ fig1 ] we have calculated the order parameter amplitude by means of eq .
( [ fee37 ] ) . here and below the cubic lattice constant @xmath72 stands for @xmath189 in the ordered phase and @xmath190 in the disordered phase . in the following we will need the wave number dependent order parameter susceptibilities @xmath191 in the disordered phase and in the ordered phase .
the susceptibility is defined in terms of the fluctuations @xmath192 of the local order parameter from its average value : @xmath193 where @xmath71 for @xmath194 , respectively .
hence , @xmath195 we first consider the disordered phase . in case of a rigid lattice
we obtain within molecular field theory @xmath196^{-1}_{\alpha \alpha } .
\label{fe34}\end{aligned}\ ] ] we recall that @xmath197 is the quadrupolar interaction matrix eq .
( [ m10 ] ) while @xmath198 is the @xmath199 unit matrix . the single particle expectation value @xmath200 has the same value for @xmath113 , and 3 .
since we consider order parameter fluctuations near the phase transition , only wave vectors @xmath201 near the star of @xmath202 are relevant . in accordance with the condensation scheme ( [ m11a ] ) , ( [ m11b ] )
the matrix @xmath197 becomes diagonal at @xmath203 , where @xmath204 .
expansion about @xmath205 gives @xmath206 similar expressions with cyclic permutation of indices are obtained for @xmath207 and @xmath208 near @xmath209 and @xmath210 , respectively . in case of a non - rigid lattice
we see from eq .
( [ fe32b ] ) that the applied pressure acts as addition to the quadrupolar interaction at @xmath202 . replacing in eq .
( [ fe35 ] ) @xmath141 by @xmath211 , we rewrite for the order parameter susceptibility in the disordered phase at temperature @xmath168 and pressure @xmath212 : @xmath213 } .
% ( 42 ) \label{fee40}\end{aligned}\ ] ] here the correlation length @xmath214 is defined by @xmath215^{1/2 } .
% ( 43 ) \label{fee41}\end{aligned}\ ] ] expressions similar to eq .
( [ fee40 ] ) are obtained for @xmath216 and @xmath217 near @xmath218 and @xmath210 , respectively , and with @xmath219 replaced by @xmath220 and @xmath221 , respectively .
making use of eqs .
( [ fe33 ] ) and ( [ fe32b ] ) , we obtain the correlation length @xmath222 } \right]^{1/2 } , % ( 44 ) \nonumber \\
\label{fee42}\end{aligned}\ ] ] where @xmath223 . with increasing pressure @xmath224 in the disordered phase
, the correlation length reaches its maximum value @xmath225^{1/2 } .
( 45 ) \label{fee43}\end{aligned}\ ] ] toward the onset of the phase transition .
likewise the susceptibility becomes maximum at @xmath226 and @xmath227 : @xmath228 in the ordered phase the order parameter susceptibility is given by @xmath229 .
\label{fe40}\end{aligned}\ ] ] applying mean - field theory @xcite we obtain for the quadrupolar model on a rigid lattice : @xmath230^{-1}_{\alpha \alpha } , % ( 48 ) \nonumber \\
\label{fe41}\end{aligned}\ ] ] where @xmath231 .
\quad \label{fe42}\end{aligned}\ ] ] numerical calculation show that @xmath232 . in expression ( [ fe40 ] ) @xmath88 is the order parameter amplitude .
proceeding now as before we consider a nonrigid lattice and an external pressure . using eqs .
( [ fe32b ] ) and ( [ fe35 ] ) we define the correlation length in the ordered phase at temperature @xmath168 and pressure @xmath233 : @xmath234^{1/2 } .
% ( 50 ) \nonumber \\
\label{fe43}\end{aligned}\ ] ] here the subscript @xmath235 stands for ordered quadrupolar phase . the corresponding order parameter susceptibility for the component @xmath236 is then given by @xmath237 } .
% ( 51 ) \label{fe44}\end{aligned}\ ] ] notice that the discontinuity of the order parameter at the first order phase transition leads to a drop of the correlation length and hence of the order parameter susceptibility . with @xmath238 given by eq .
( [ fe33a ] ) , we obtain @xmath239 } \right]^{1/2 } , % eq .
( 52 ) \nonumber \\ \label{fee49}\end{aligned}\ ] ] and since @xmath240 , @xmath241 .
the correlation length in the ordered phase is calculated by means of eqs .
( [ fe43 ] ) , ( [ fee37 ] ) and ( [ fe33 ] ) .
the result reads @xmath242^{-1/2 } \label{fee50}\end{aligned}\ ] ] where @xmath243 numerical evaluation of the quadrupolar model shows that @xmath244 .
hence the correlation length decreases with increasing pressure in the ordered phase . using eq .
( [ fee42 ] ) for @xmath245 and eq .
( [ fe43 ] ) for @xmath246 we have calculated the squared correlation length @xmath247 as a function of pressure . the plot is shown in fig .
[ fig2 ] . from eqs .
( [ fee40 ] ) and ( [ fe44 ] ) for @xmath248 we see that fig .
[ fig2 ] also describes the pressure dependence of the order parameter susceptibility at @xmath226 .
the expressions of the order parameter susceptibility will be used in the next section where we study the temperature and pressure dependence of the elastic response near the first order phase transition .
we will derive the static displacement displacement response function matrix @xcite by using the well known green s functions techniques @xcite
. inversion of the response function in the long wavelength limit leads to the elastic constants .
the fourier transform of the retarded green s function of two operators @xmath249 and @xmath147 is defined by @xmath250 \rangle , \label{m22}\end{aligned}\ ] ] with frequency @xmath251 , @xmath252 .
the skew brackets @xmath253 stand for a thermal average with the system s hamiltonian @xmath254 . in heisenberg representation
the time dependence reads @xmath255 , @xmath256 .
we quote the equation of motion @xcite : @xmath257 \rangle + \langle \langle [ a,{\cal h } ] ; b \rangle \rangle_z , \label{m23}\end{aligned}\ ] ] with the identity @xmath258 ; b \rangle \rangle_z = - \langle \langle a ; [ b,{\cal h } ] \rangle \rangle_z .
\label{m24}\end{aligned}\ ] ] the dynamic displacement displacement green s function is defined by @xmath259 with lattice displacements and conjugate momenta specified in eqs .
( [ m14a ] ) , ( [ m14b ] ) .
we recall that the hamiltonian is given by eq .
( [ m1 ] ) , where the parts have been specified subsequently in sect . [ sec : model ] . in the following
we treat @xmath260 as a dynamic variable .
thereby we retain only the fluctuation contribution of the order parameter variable , writing @xmath261 applying twice the equation of motion ( [ m23 ] ) to @xmath262 we use the commutation rules eq .
( [ m15 ] ) as well as the fact that the electronic variables @xmath263 and @xmath264 commute with the lattice variables @xmath265 and @xmath266 .
the result reads @xmath267 here the summation is understood over the repeated index @xmath268 .
we recall that @xmath260 has been defined by eq .
( [ m20 ] ) . in a similar way we obtain by using eq .
( [ m24 ] ) @xmath269 taking the static limit @xmath270 in eqs .
( [ m26 ] ) and ( [ m27 ] ) and combining the results we find @xmath271 .
\nonumber \\ \label{m28}\end{aligned}\ ] ] here the subscript @xmath272 stands for @xmath270 .
we introduce the static susceptibilities @xcite @xmath273 and rewrite eq .
( [ m28 ] ) as @xmath274 .
\nonumber \\ \label{m31}\end{aligned}\ ] ] we remind that in the long wavelength limit the static displacement susceptibility is related to the elastic constants @xcite @xmath275 by @xmath276 where @xmath277 is the mass density .
hence eq . ( [ m31 ] ) allows us to calculate the elastic constants in presence of the coupling @xmath32 between electronic and lattice degrees of freedom .
we recall that in absence of the coupling the bare elastic constants are given by @xmath278 [ see eq .
( [ m16 ] ) ] . for cubic crystals we have in voigt
s notation the elastic constants @xmath279 , @xmath280 , @xmath281 . with @xmath282 and @xmath283 ,
( [ m31 ] ) reduces in the long wavelength limit to @xmath284^{-1 } .
% ( 67 ) \label{m33}\end{aligned}\ ] ] here we have made use of eqs .
( [ m32 ] ) and ( [ m16 ] ) . on the basis of the hamiltonian , eq .
( [ m1 ] ) , the obtained expressions for @xmath285 and @xmath3 are rigorous results . since @xmath286 , the quadrupolar - elastic coupling , eq .
( [ m19 ] ) , leads to a reduction of the elastic constant @xmath3 in comparison with the bare quantity @xmath287 .
this reduction as a consequence of the @xmath32 coupling is responsible for the relative softening of the corresponding @xmath288[001 ] phonon branch in @xmath6-ce @xcite . in contradistinction to @xmath3 , the shear elastic constant @xmath4 is not affected by the coupling of the lattice to quadrupolar electronic density fluctuations . as we have shown in sec .
ii , there is no coupling to shear strains @xmath289 in @xmath32 , eq .
( [ m19 ] ) . hence taking @xmath290 in eq .
( [ m31 ] ) , we obtain for @xmath291 and @xmath292 : @xmath293 or equivalently by means of eqs .
( [ m16 ] ) and ( [ m32 ] ) @xmath294 .
pressure experiments exhibit a discontinuity of the shear modulus @xcite at the phase transition . however , this effect is solely due to the lattice contraction and a concomitant change of interatomic forces .
due to the absence of a direct coupling between lattice shears and orientational density fluctuations there are no precursor effects .
we want to study the anomalous behavior of the elastic constant @xmath3 near the quadrupolar ordering transition . from eq .
( [ m33 ] ) it follows that the important quantity is the four - point ( four factors @xmath88 ) susceptibility @xmath295 . at high temperature and in the long wavelength regime
such that the length scale of order parameter fluctuations is large in comparison with the interatomic spacing we use classical statistical mechanics .
we then have @xmath296 where @xmath297 in appendix we show that in the limit @xmath298 @xmath299 with @xmath236 , 3 . here
@xmath191 is the order parameter susceptibility for the component @xmath7 .
cubic symmetry implies that @xmath300 we calculate the right hand side of eq .
( [ m36 ] ) by using @xmath301 where @xmath302 is the volume of the crystal and @xmath24 the number of atoms .
so far the considerations of the present section are general and hold for the disordered as well as for the ordered phase . since the expressions of the order parameter susceptibility in the disordered and in the ordered phase , eqs .
( [ fee40 ] ) and ( [ fe44 ] ) , respectively , exhibit the same wave vector dependence , we can treat both cases simultaneously . since the integrand in eq .
( [ m45 ] ) does not depend on @xmath303 , the integration over @xmath304 in the interval @xmath305 yields @xmath306 . on the other hand the integrand vanishes for large values of @xmath307 .
hence we integrate over a circle and extend the radius to @xmath308 . taking into account eq .
( [ m38 ] ) , we obtain @xmath309 where @xmath310 stands for @xmath311 in the disordered phase and for @xmath312 in the ordered quadrupolar phase .
substitution of the result into eq .
( [ m33 ] ) gives the longitudinal elastic constant @xmath313^{-1 } , % ( 75 ) \label{m58}\end{aligned}\ ] ] where we have defined @xmath314 with @xmath315 , eq .
( [ fee42 ] ) , in the disordered phase and @xmath316 , eq .
( [ fee50 ] ) , in the ordered phase . near the phase transition in the disordered phase we have @xmath317^{-1 } .
% ( 77 ) \label{m73}\end{aligned}\ ] ] the increase of @xmath245 or equivalently of @xmath318 with @xmath223 leads to a decrease of @xmath3 which reaches its minimum value @xmath319^{-1 } % ( 78 ) \label{m74}\end{aligned}\ ] ] at @xmath320 . on the other hand at the onset of the phase transition one has @xmath321^{-1 } .
% ( 79 ) \label{m75}\end{aligned}\ ] ] the discontinuity of the order parameter and the concomitant drop of the correlation length @xmath241 results in a positive jump of @xmath3 at the first order phase transition : @xmath322}{[1 + \xi_{d}(t_1,p_1 ) ] [ 1 + \xi_{q}(t_1,p_1 ) ] } .
% ( 80 ) \label{m80n}\end{aligned}\ ] ] here we have approximated @xmath72 in the prefactor by @xmath323 . in the ordred phase the decrease of the correlation length @xmath246 with increasing pressure leads to an increase of @xmath324^{-1 } .
% ( 81 ) \label{m76}\end{aligned}\ ] ] the scenario described by eqs .
( [ m73])([m76 ] ) is in full agreement with experiment . indeed
the pioneering experiments on elastic properties of ce under pressure by voronov et al .
@xcite as well as recent high resolution ultrasonic measurements @xcite show that the propagation velocity of longitudinal ultrasonic waves decreases when the phase transition toward the @xmath7 phase is approached with increasing pressure in the @xmath6 phase . at the first order phase transition the longitudinal sound velocity exhibits a stepwise increase .
with increasing pressure in the @xmath7 phase the longitudinal sound velocity increases continuously .
a corresponding behavior of the bulk modulus as a function of pressure was deduced from high - resolution neutron and synchrotron x - ray powder diffraction @xcite experiments .
it was argued that the softening of the bulk modulus in the @xmath6 phase with increasing pressure @xmath325 is a direct consequence of the softening of @xmath3 .
we close this section by observing that in c@xmath5 fullerite there occurs a marked lowering of the bulk modulus @xmath147 if at fixed @xmath8 the fcc@xmath21sc phase transition is approached with increasing pressure @xcite .
the phenomenon is attributed to orientational reordering of the c@xmath5 molecules .
given the similarities of the lattice related phenomena at the @xmath11 phase transition in ce and the orientational phase transition in c@xmath5 fullerite , we conclude that these transitions are isomorphic . from the mathematical point of view @xcite all essential aspects ( @xmath0 ) are the same , while the constituents ( electronic quadrupolar densities versus icosahedral molecules ) are different .
the elastic properties of a model of interacting 4f - electron quadrupolar densities on a compressible fcc lattice have been investigated as a function of temperature and pressure . on the basis of previous theoretical work @xcite this model , supported by recent nuclear spectroscopy experiments @xcite ,
is taken as representative for the @xmath1 isostructural " phase transition in ce .
in particular , we have studied by analytical theory the pressure dependent anomalies of the elastic constant @xmath3 at the phase transition from the quadrupolar orientationally disoredered phase which we identify with the @xmath6-phase ( space group @xmath19 ) to the quadrupolar orientationally ordered phase which we identify with the @xmath7-phase ( space group @xmath2 ) . as a result
we find that @xmath3 ( equivalently the longitudinal sound velocity ) decreases by approaching the phase transition from the disoredered phase with increasing pressure @xmath224 , at the first order phase transition @xmath3 exhibits a positive jump , in the ordered phase @xmath3 increases continuously with pressure @xmath326 .
on the other hand we find that the elastic constant @xmath4 ( equivalently the shear sound velocity ) exhibits no precursor effects near the transition .
these theoretical results are in full qualitative agreement with experiments [ 34,21,22,33 ] on elastic anomalies at the @xmath1 isostructural " phase transition in solid ce . we notice that the hamiltonian @xmath32 , eq . ( [ m17 ] ) , which is quadratic in the electronic quadrupolar order parameter variables and linear in the lattice displacements , accounts as well for the elastic anomalies as for the isostructural " lattice contraction .
the hamiltonian is reminiscent from the compressible ising model @xcite which is quadratic in the spin variables and linear in the lattice displacements .
however , in the present case symmetry properties are more subtle and account for the interplay between antiferroquadrupolar order in the electronic densities and isostructural " lattice contraction . here
a remark on symmetry reduction at the phase transition is in order .
the quadrupolar interaction has its origin in the repulsive coulomb interaction between 4f - electrons on neighboring atoms .
the ordering of the electronic quadrupoles on four sc sublattices reduces the repulsion and acts as an effective attraction . in reciprocal space the quadrupolar interaction matrix @xmath74 , eq .
( [ m10 ] ) , has negative eigenvalues at the @xmath16-point of the brillouin zone of the fcc lattice .
phonon dispersions of ce measured by synchrotron radiation show pressure dependent anomalies related to the @xmath1 transition at the @xmath16-point of the brillouin zone @xcite .
the explanation of these experiments is an outstanding challenge for theoretical work .
a further problem is the quantitative improvement of the theory which should lead to increase the magnitude of the coupling coefficient @xmath146 in eq .
( [ m19 ] ) for @xmath32 . at present
we believe that the theory should be extended by including the intrasite coupling between 4f - electrons and conduction electrons ( say 5d ) and a possible influence of this coupling on the interatomic bonding .
an indication of the relevance of such a mechanism is provided by recent work on the ce dimer @xcite .
likely such an extension of the theory would also contribute to elucidate the concept of a mott transition @xcite versus a kondo scenario @xcite at the @xmath1 transition .
the authors acknowledge useful discussions with k. parlinski , m. krisch , f. decremps , r.m .
pick , g. roth , g. gntherodt , a.v .
tsvyashchenko and e.v .
. financial support has been provided by the research group theory of condensed matter , university of antwerp and by the institut fr kristallographie , rwth aachen . here
we express @xmath295 as a function of the order parameter susceptibility @xmath191 .
starting from eqs .
( [ m35 ] ) and ( [ m70n ] ) we first notice that @xmath327 where @xmath328 we rewrite eq .
( [ m35 ] ) as @xmath329 .
\nonumber \\ \label{a3}\end{aligned}\ ] ] we then approximate the four - point function @xmath330 by the factorization scheme @xmath331 or equivalently @xmath332 substituting this result into eq .
( [ a3 ] ) and using eq .
( [ fe39n ] ) we obtain @xmath333 taking the limit @xmath334 leads to the result of eq .
( [ m36 ] ) .
d.c . koskenmaki and k.a .
gschneidner , jr . , _
handbook on the physics and chemistry of rare earths _
gschneidner , jr . , and l. eyring ( amsterdam : north - holland , 1978 ) , p. 337 .
p. w. bridgman , proc .
arts sci . *
76 * , 55 ( 1948 ) .
macpherson , g.e .
everett , d. wohlleben , m.b .
maple , phys .
* 26 * , 20 ( 1971 ) .
l. d. landau and e. m. lifshitz . : _ statistical physics _ , ( pergamon , bristol , 1995 ) , vol . 5 .
a. w. lawson , t .- y .
tang , phys . rev . *
76 * , 301 ( 1949 ) .
b. johansson , phil . mag . * 30 * , 469 ( 1974 ) .
b. johansson , i.a .
abrikosov , m. aldn , a.v .
ruban , and h.l .
skriver , phys .
lett . * 74 * , 2335 ( 1995 ) .
murani , s.j .
levett , and j.w .
taylor , phys .
lett . * 95 * , 256403 ( 2005 ) .
a. hjelm , j. trygg , o. eriksson , b. johansson , and j. wills , phys .
b * 50 * , 4332 ( 1994 ) .
x. cao and m. dolg , mol .
* 101 * , 1967 ( 2003 ) .
roos , r. lindh , p .- a .
malmqvist , v. veryazov , and p .- o .
widmark , j. phys .
a * 112 * , 11431 ( 2008 ) .
nikolaev , phys .
rev . a * 84 * , 012512 ( 2011 ) .
allen , r.m .
martin , phys .
* 49 * , 1106 ( 1982 ) ; j.w .
allen and l.z .
liu , phys .
b * 46 * , 5047 ( 1992 ) .
m. lavagna , c. lacroix , and m. cyrot , phys .
a * 90 * , 210 ( 1982 ) .
p. fulde , _ electron correlations in molecules and solids _
( springer heidelberg , 1995 ) .
liu , j.w .
allen , o. gunnarsson , n.e .
christensen , and o.k .
andersen , phys .
b * 45 * , 8934 ( 1991 ) .
b. johansson , a. v. ruban , and i. a. abrikosov , phys .
. lett . * 102 * , 189601 ( 2009 ) .
b. amadon , s. biermann , a. georges , and f. aryasetiawan , phys .
lett . * 96 * , 066402 ( 2006 ) . j .-
rueff , j .-
iti , m. taguchi , c.f .
hague , j .-
mariot , r. delaunay , j .-
kappber , and n. jaonen , phys .
lett . * 96 * , 237403 ( 2006 ) .
g.eliashberg , h.capellmann , jetp lett . *
67 * , 125 ( 1998 ) .
jeong , t.w . darling , m. j. graf , th .
proffen , r. h. heffner , y. lee , t. vogt , j. d. jorgensen , phys .
lett . * 92 * , 105702 ( 2004 ) .
lipp , d. jackson , h. cynn , c. aracne , w. j. evans , and a. k. mcmahan , phys .
* 101 * , 165703 ( 2008 ) .
a.v . nikolaev and k.h .
michel , eur .
j. b * 9 * , 619 ( 1999 ) .
a.v . nikolaev and k.h .
michel , eur .
j. b * 17 * , 15 ( 2000 ) .
a.v . nikolaev and k.h .
michel , phys .
b * 66 * , 054103 ( 2002 ) .
r. sachidanandam and a.b .
harris , phys .
67 * , 1467 ( 1991 ) ; p.a .
heiney , j.e .
fischer , a.r .
mcghie , w.j .
romanow , a.m. denenstein , j.p .
mccauly jr . , a.b .
smith iii , and d.e .
cox , ibid . * 67 * , 1468 ( 1991 ) .
david , r.m .
ibberson , j.c .
matthewman , k. prassides , t.j.s .
dennis , j.p .
hare , h.w .
kroto , d.r.m .
walton , nature , * 353 * , 147 ( 1991 ) .
david , r.m .
ibberson , t.j.s .
dennis , j.p .
hare , k. prassides , europhys .
lett . * 18 * , 219 ( 1992 ) .
heiney , g.b.m .
vanghan , j.e .
fischer , n. coustel , d.e .
cox , j.r.d .
copley , d.a .
neumann , w.a .
kamitakahara , k.m .
creegan , d.m .
cox , j.p .
mccauley , jr . , and
smith iii , phys .
b * 45 * , 4544 ( 1992 ) .
f. decremps , l. belhadi , d.l .
farber , k.t .
moore , f. occelli , m. ganthier , a. polian , d. antonangeli , c. m. aracne - ruddle , and b. amadon , phys .
lett . * 106 * , 065701 ( 2011 ) .
tsvyashchenko , a.v .
nikolaev , a.i .
velichkov , a.v .
salamatin , l.n .
fomicheva , g.k .
ryasny , a.a .
sorokin , o.i .
kochetov , m. budzynski , and k.h .
michel , phys .
b * 82 * , 092102 ( 2010 ) ; a.v .
tsvyashchenko , a.v .
nikolaev , a.i .
velichkov , a.v .
salamatin , l.n .
fomicheva , g.k .
ryasny , a.a .
sorokin , o.i .
kochetov and m. budzynski , jetp * 111 * , 627 ( 2010 ) .
m. krisch , d.l .
farber , r. xu , d. antonangeli , c. m. aracne , a. beraud , t .- c .
chiang , j. zarestky , d. y. kim , e. i. isaev , r. ahuja , and b. johansson , pnas ( usa ) * 108 * , 9343 ( 2011 ) .
f. decremps , d. antonangeli , b. amadon , and g. schmerber , phys .
b * 80 * , 132103 ( 2009 ) .
f.f.voronov , l.f.vereshchagin and v.a.goncharova , doklad .
nauk sssr , * 135 * , 1104 ( 1960 ) , engl
soviet phys .
doklady , * 135 * , 1280 ( 1960 ) . c. stassis , t. gould , o.d .
mcmasters , and k.a .
gschneidner , jr . , and r.m .
nicklow , phys .
b * 19 * , 5746 ( 1979 ) .
w. schranz , a. fuith , p. dolinar , h. warhanek , m. haluska , and h. kuzmany , phys .
lett . * 71 * , 1561 ( 1993 ) .
a. lundin and b. sundqvist , phys .
b * 53 * , 8329 ( 1996 )
bradley and a.p .
cracknell , _ the mathematical theory of symmetry in solids _ , ( clarendon , oxford , 1972 ) .
p. zielinski and k. parlinski , j. phys .
c : solid state phys . *
17 * , 3287 ( 1984 ) .
michel , j.r.d .
copley , d.a .
neumann , phys .
68 * , 2929 ( 1992 ) .
d. lamoen , k.h .
michel , phys .
b * 48 * , 807 ( 1993 ) .
moore , l. belhadi , f. decremps , d.l .
faber , j.a .
bradley , f. occelli , m. gauthier , a. polian , and c.m .
aracne - ruddle , acta materialia * 59 * , 6007 ( 2011 ) .
samara , j.e .
schirber , b. morosin , l.v .
hansen , d. loy , and a.p .
sylwester , phys .
lett . * 67 * , 3136 ( 1991 ) . g. kriza , j .- c .
ameline , d. jrome , a. dworkin , h. swarc , c. fabre , d. schtz , a. rassat , p. bbernier , and a. zahab , j. phys . i ( france ) * 1 * , 1361 ( 1991 ) .
p.m. chaikin and t.c .
lubensky , _ principles of condensed matter physics _ , cambridge university press , cambridge ( 1995 ) .
w. gtze and k.h .
michel , in _ dynamical properties of solids _ ,
horton and a.a .
maradudin eds .
, north - holland publ . comp .
1974 , ch . 9 ,
zubarev , soviet physics uspekhi * 3 * , 320 ( 1960 ) . , t. gowers , editor , princeton university press ( 2008 ) , p. 27 . h. wagner and j. swift , z. phys .
* 239 * , 182 ( 1970 ) .
d.j . bergman and b.i .
halperin , phys .
b * 13 * , 2145 ( 1976 ) .
k. gofryk , e.k.h .
salje , d.j .
safarik , j.l .
smith , x. ding , a.c .
lawson , p.s .
riseborough , b. mihaila , and j.c .
lashley ( submitted ) . | starting from a model of 4f - electron generated quadrupolar densities on a compressible fcc lattice , the elastic anomalies at the @xmath0 phase transition are studied by means of analytical theory .
the model is taken as representative for the @xmath1 phase transition in ce .
the coupling of the ( linear ) lattice displacements to the square of the quadrupolar orientational density fluctuations renormalizes the elastic constants .
the condensation of the quadrupolar densities into the orientationally ordered @xmath2 structure is studied as function of temperature and pressure .
precursor effects of the transition lead to an anomalous softening of the elastic constant @xmath3 while @xmath4 exhibits no such softening .
the theoretical results are in excellent qualitative agreement with pressure experiments on the elastic constants ( equivalently on sound velocities ) at the @xmath1 transition in ce .
lattice dynamical analogies in theory and striking similarities in experimental results with the @xmath0 transition in c@xmath5 fullerite are discussed . |
the direct search for physics beyond the standard model(sm ) has been unsuccessful thus far .
there have been anomalies in some of the observables in the flavour sector , with deviations from the sm predictions at the level of few sigma @xcite .
in fact , the presence of new physics(np ) at possibly high scales may very well be deduced only from precision measurements of some of the rare meson decays .
the absence of flavour changing neutral currents ( fcncs ) at the tree level allows the possibility of virtual new physics particles to be present in the loop level diagrams that contribute to these processes .
detailed study of rare charm decays was performed in ref .
predictions for these decays in various extensions of the sm including extensions of the higgs , gauge and fermion sectors were obtained .
rare charm decays were also recently studied in refs .
. the focus of this study will be the radiative decays of charmed mesons . while both the inclusive and exclusive radiative @xmath2 meson decays have been extensively discussed in the literature ,
less attention has been paid to the @xmath3 meson radiative decays as their branching ratios are expected to be much smaller due to the almost complete gim suppression .
moreover , charm radiative decays will be dominated by long distance contributions , which can hide the presence of new physics particles that may appear in the loop of the short distance penguin contributions .
nevertheless , in ref .
@xcite it was pointed out that a measurement of the difference in the rates of the exclusive modes , @xmath4 and @xmath5 in which the long distance effects are expected to cancel , would indicate short distance new physics if the data reveals a difference of rates which is more than @xmath6 .
but in general , due to the large uncertainties in the long distance contributions , any definite conclusion regarding np will not be feasible from a measurement of the radiative decay rates for the inclusive @xmath7 case nor for any individual exclusive channel , unless the np short distance contribution is larger than that from the long distance effects .
in fact , the possibility of enhancement above the otherwise dominant long distance effects , in presence of a fourth generation model with large mixing angles of the @xmath8 quark , @xmath9 had been pointed out in ref . @xcite .
fourth generation models are now inconsistent with the lhc data , however , models with vector - like charge @xmath10 quarks , for which the authors of ref .
@xcite claimed that their results were also applicable , are still viable .
in fact , in the last couple of years many detailed studies of the phenomenology of vector - like quarks and constraints from the flavour sector have been performed @xcite .
apart from the enhancement in the decay rate , which will be subject to the relative size of the short distance and long distance effects , np could also be searched through a measurement of the polarization of the photon produced in the decay .
the sm has a robust prediction regarding the photon polarization in @xmath7 decays and hence , a measurement of the photon polarization can pin down the presence of np .
this had been earlier pointed out for the case of @xmath2 radiative decays in refs .
@xcite . in the sm ,
the photons from the short distance ( sd ) penguin contribution in the @xmath7 decays will be mostly left handed up to corrections of @xmath11(@xmath12 ) .
this dominance of left handed polarization can get masked in the presence of long distance ( ld ) effects .
however , the fraction of the right polarized photons will vary in different models and may possibly even allow one to distinguish between different models of np .
we explore the effects of the presence of a down - type isosinglet vector - like quark model on the @xmath7 decay rate , as well as on the photon polarization for this model with an additional left - right symmetry .
the decay rate evaluation requires an estimation of both the sd as well as ld components , which are described in the next section . in sec .
[ vlq ] some details of the down type isosinglet vector - like quark model are discussed , including the modifications to the wilson coefficients in its presence .
[ lrsm ] contains a short discussion on lrsm and the results for the bare level sd contributions to the amplitudes for the emission of the left and right handed photons in this model .
[ iva ] gives our results for the branching ratios ( br s ) in the sm and in the different np models . in sec .
[ ivb ] we present our analysis of the polarization function in the lrsm and the lrsm with vector - like quark . finally in sec .
[ v ] , we conclude .
the long distance contributions being non - perturbative in nature are hard to estimate .
the long distance contributions can come either from pole diagrams or vector meson dominance ( vmd ) diagrams . at the quark level
, the pole contribution corresponds to the annihilation diagrams @xmath13 with a photon attached to any of the four quark lines .
they are actually a subset of a more general class of long distance contributions , which include two - particle intermediate states and extends up to all higher n - particle intermediate states .
phenomenologically however , the single - particle or pole terms are the most accessible .
the underlying quark processes in the vmd contributions are @xmath14 , followed by @xmath15 .
all these long distance effects are rather hard to calculate from first principles but can be estimated in models .
hence , it is important that the observables chosen for uncovering np be independent of the long distance effects .
however , we provide an updated estimate for the long distance amplitudes and branching ratios for charm decays following the methods of ref .
@xcite in appendix [ ldbr ] .
( -5,0 ) ( -1.5,0 ) ; ( -4,0 ) arc ( 180:.5:.8 ) ; ( -2.8,0.8)(-1.7,1.8 ) ; at ( -4.1,0.8 ) w ; at ( -2.3,0.8 ) w ; at ( -5.2,0 ) c ; at ( -1.3,0 ) u ; at ( -2.1,1.8 ) @xmath16 ; at ( -3.2,-0.2 ) d , s , b ; ( -5,0 ) ( -1.5,0 ) ; ( -4,0 ) arc ( -180:.5:.8 ) ; ( -3.2,0)(-1.7,1 ) ; at ( -3.2,-1.2 ) w ; at ( -5.2,0 ) c ; at ( -1.3,0 ) u ; at ( -2.1,1.1 ) @xmath16 ; at ( -3.2,-0.2 ) d , s , b ; the amplitude for the flavour changing radiative transitions were first evaluated by inami and lim @xcite . as pointed out by ref .
@xcite , those formulas need to be appropriately modified for the case of @xmath7 decay .
the sm lagrangian for the processes @xmath7 , which arises at the loop level as shown in fig.[ctougamma fig ] is given by , @xmath17 where the mass of the final quark @xmath18 has been neglected and @xmath19 . the coefficient @xmath20 is a function of the internal quark masses and the ( qcd uncorrected ) contribution to the amplitude @xmath20 is given by , @xmath21\nonumber\\ & = \sum_{p=1,2}q_p\sum_{q = d , s , b}v_{ci}^{*}v_{ui}g_p(r_{q } ) , \end{aligned}\ ] ] where @xmath22 with @xmath23 ( @xmath24 , @xmath25 , @xmath26 ) being the masses of the down - type quarks running in the fermion loop in the penguin diagrams .
the functions @xmath27 defined in @xcite are given in appendix [ g ] .
@xmath28 and @xmath29 are the charges of the @xmath30 boson emitted from the initial quark in the loop diagram , and that of the internal quark running in the loop , respectively .
the inclusive decay rate for a @xmath0 process within the sm is given by , @xmath31 this results in the following inclusive br for the @xmath0 process , @xmath32 which is normalized with respect to the inclusive semi - leptonic br for @xmath33 decays , to get rid of the uncertainty in the charm quark mass .
the function @xmath34 is the phase space suppression factor and is given by , @xmath35 in the case of @xmath36 decay , up - type quarks flow in the loop and the heavy top quark contribution dominates and induces penguin operators already at the electroweak scale .
in contrast , all the down type quarks are massless compared to the electroweak scale , resulting in no penguin contribution at this scale within the sm for the case of @xmath7 .
however , the presence of a heavy down type vector - like quark , will result in a non - vanishing penguin contribution at the electroweak scale . within the sm ,
the enhancement of the radiative decay rates in presence of qcd corrections was pointed out in ref .
while the enhancement was by a factor of two in the case of @xmath36 , it was expected to be more dramatic in the case of charm radiative decays @xcite .
it is hence important to write down the weak effective hamiltonian with all the dimension-6 operators and calculate the corresponding wilson coefficients within the renormalization - group improved perturbation theory which are discussed below .
the rg evolution of the wilson coefficients for charm decays in context of the sm to the next - to - leading order ( nlo ) in qcd corrections is performed .
the calculation for the complete set of operators relevant for charm decays had previously been done up to nlo in the ndr scheme @xcite and to the next - to - next - to - leading order ( nnlo ) in the @xmath38 scheme @xcite . in this article , we work in the @xmath38 scheme since the anomalous dimension matrices at the leading order ( lo ) ( @xmath39 ) and at nlo ( @xmath40 ) are readily available in ref .
the short distance evolution of the wilson coefficients has to be divided into two steps .
the first task is to integrate out the weak gauge bosons at a scale @xmath41 .
this is done by calculating the @xmath42 s at the scale @xmath41 by matching the effective theory with five active flavors @xmath43 , @xmath44 , @xmath25 , @xmath45 , @xmath26 onto the full theory .
as mentioned earlier , no penguin operators are generated at this point , since all the down - type quarks ( @xmath44 , @xmath25 and @xmath26 ) are to be treated as massless @xcite and the gim mechanism is in full effect .
the effective hamiltonian for the scale @xmath46 is then given by , @xmath47.\ ] ] here , @xmath48 the effective anomalous dimension matrix @xmath49 is calculated in the effective theory with five flavours .
using this matrix , the @xmath50 s are evolved down to the scale @xmath51 , and the @xmath52 s are obtained .
the next step is to integrate out the @xmath26 quark as an effective degree of freedom at the scale @xmath51 .
this is accomplished by matching the effective five flavour theory onto the effective theory for four flavours .
this generates the penguin operators with the wilson coefficients depending upon @xmath53 solely through the coefficients @xmath54 .
the effective hamiltonian at the scale @xmath55 is then given by @xmath56\ ] ] where @xmath57 in all of the above , @xmath58 and @xmath59 are the chirality projection operators .
the colour indices are suppressed in currents of the form @xmath60 .
the @xmath61 are the generators of @xmath62 .
the @xmath42 s are the wilson coefficients which contain the complete short distance ( perturbative qcd ) corrections . for the case of radiative
charm decays under consideration here , the operators @xmath63 and @xmath64 are not relevant and therefore not shown in the above list .
we will hence consider only the set of wilson coefficients @xmath65 which are evolved down from the @xmath66 scale to the @xmath67 scale using the @xmath49 matrix now evaluated in the effective theory with four flavours to obtain the @xmath68 s .
hence , at each order @xmath69 , the vector of the wilson coefficients @xmath42 at the scale @xmath70 may be schematically written as @xmath71 where @xmath72 is the number of active flavours at the corresponding scale , @xmath73 is the matching matrix between the effective five flavour theory above the scale @xmath74 to the effective four flavour theory below the scale @xmath74 , the index @xmath75\{lo , nlo } specifies the order in qcd corrections at which the corresponding quantities are being calculated .
the @xmath76 s are the evolution matrices related to the effective anomalous dimension matrix @xmath77 and are discussed in detail below in the context of the lo and the nlo calculations along with a few other important features .
we closely follow ref .
@xcite in the following discussion .
let us start with the full @xmath78 effective anomalous dimension matrix at the leading order ( @xmath79 ) which can be assimilated in parts from @xcite .
it is given in eqn .
( [ g0 ] ) in appendix [ lonloanomalous ] with the full dependence on the number of active flavours(@xmath72 ) and charges(@xmath80 ) of the internal quark and the decaying quark .
now , let @xmath81 be the matrix that diagonalizes @xmath82 , so that @xmath83_{diag}.\ ] ] the lo evolution matrix @xmath84 for evolving the @xmath42 s down from the scale @xmath85 to @xmath86 is then given by @xmath87_{diag}v^{-1}\ ] ] where @xmath88 is the strong coupling constant .
a few comments are in order at this point .
it was specified previously that the only operators relevant for the case of charm decays within the sm , above the scale @xmath74 are @xmath89 and @xmath90 .
hence , the matrix @xmath91 is essentially a @xmath92 matrix .
the lo values of @xmath93 , which are basically the initial conditions are well known and are given by : @xmath94 hence we have , for the scale @xmath46 @xmath95 at this point , all the other wilson coefficients ( @xmath96 to @xmath97 ) are zero .
they get their values from the matching at the scale @xmath66 . however , the matching matrix @xmath98 to lo and hence , for the lo evolution , the coefficients @xmath96 to @xmath97 remain vanishing even after the matching procedure . the resulting @xmath99 column vector @xmath100 is then multiplied with the @xmath78 evolution matrix @xmath101 to obtain the values of the @xmath42 s at the charm scale . the renormalization scheme independent wilson coefficient @xmath102 relevant for radiative charm decays is then obtained at lo using the relation @xmath103 where @xmath104 @xcite . the nlo expression for the evolution matrix is given by @xmath105 where @xmath106 @xmath81 was defined previously in eqn .
( [ vlo ] ) and the matrix @xmath107 is defined by @xmath108 with @xmath109 the expression for the @xmath78 @xmath110 matrix with the complete effective flavour and charge dependence can again be collected in parts from @xcite . due to its large size , we provide the matrix in two separate @xmath111 and @xmath112 blocks in appendix [ lonloanomalous ] ( see eqns .
[ g11 ] and [ g12 ] ) .
it is easy to see that one encounters a term of the order of @xmath113 on expanding the expression for @xmath114 ( eqn .
( [ nlo ] ) ) .
however , a calculation of the nlo contribution necessarily requires that all terms higher than the first order in @xmath88 be discarded and hence , special care should be taken in using eqn.([nlo ] ) for the nlo evolution .
similar to the case of the lo evolution , the only relevant coefficients above the @xmath66 scale are @xmath115 and @xmath116 , calculated up to the nlo order this time .
the expressions can be found in @xcite and in the @xmath38 scheme are given by @xmath117 the coefficients @xmath118 however are non - vanishing after the matching procedure at nlo , since the matching matrix @xmath119 is now defined by @xmath120 the non - zero elements of the matrix @xmath121 for charm decays being @xcite @xmath122 the full set of nlo coefficients @xmath123 for the case of charm decays in the sm is then given by @xmath124 where @xmath125 is an @xmath99 column vector whose first two elements are @xmath126 and @xmath127 and the rest are zero . once the values at the charm scale are obtained , the corresponding value for @xmath102 can be obtained from eqn .
( [ c7eff ] ) .
the sm contains three generation of quarks , however , the number of generations is not predicted by the theory . a simple extension of the sm would be to have a chiral fourth generation of quarks and leptons .
presence of a fourth generation would have a significant effect on the higgs sector of the sm and is now ruled out by the higgs production and decay processes data at the lhc .
however , the so called vector - like quarks , which do not receive their masses from yukawa couplings to a higgs doublet , are consistent with the present higgs data .
they are distinguished from the sm quarks by their vector coupling to gauge bosons , i.e. , both the left handed , @xmath128 and right handed , @xmath129 chiralities of these fermions transform the same way under the sm gauge groups @xmath130 .
these exotic fermions occur for example in the grand unified theory based on @xmath131 @xcite . in general these fermions
could either be singlets or doublets or triplets under @xmath132 . here
we consider the case of a down type isosinglet quark . in the sm ,
the quark mixing matrix is a @xmath133 unitary matrix which is specified in terms of three angles ( @xmath134 ) and a @xmath135-violating phase , @xmath136 .
a @xmath137 unitary quark mixing matrix is parametrized in terms of 3 additional angles ( @xmath138 ) and two more @xmath135 violating phases , @xmath139 . in ref .
@xcite , a chi - squared fit to many flavour observables was performed to obtain the preferred central values , along with the errors of all the elements of the measurable @xmath140 quark mixing matrix . to evaluate the sd contribution of the radiative decay rate in the presence of the vector - like isosinglet quark , the central values of the mass and mixing
angles are obtained from the results of the fit in ref .
@xcite are used .
having discussed the evolution of the wilson coefficients for the sm in full detail we will now simply specify how the contribution of the vector - like quark model modifies the sm coefficients .
the down - type vector - like quark induces a @xmath141-mediated fcnc in the down - type quark sector .
however for the @xmath142 transitions that are of interest to us , no new set of operators are introduced and hence the anomalous dimension matrices along with the coefficients @xmath115 and @xmath116 , remain exactly the same as that in the sm , up to nlo . which arise as a result of integrating out the top quark as a heavier degree of freedom at the electroweak scale . since the @xmath8 is also heavier than the @xmath30 boson , one needs to integrate it out too at this scale .
hence , at the nnlo level , the expressions for @xmath115 and @xmath116 change for this model as compared to sm . ]
the fundamental difference in this model is that at the electroweak scale , the coefficients @xmath143 will not be zero . while the down - type quarks running in the penguin loop in the sm can be treated as massless and hence do not contribute , the vector - like @xmath144 quark , which couples with all the up - type sm quarks being heavier than @xmath53 will generate a value for the coefficients @xmath145 and @xmath97 at the electroweak scale itself .
the values are @xmath146 where the functions @xmath147 defined in @xcite are given in appendix [ g ] .
we have calculated these coefficients in this model for two benchmark values for the mass of the @xmath144 quark in accordance with @xcite .
our results are displayed in table [ tab1 ] .
our values for the coefficients in the sm match exactly with ref .
@xcite if we use their values for the parameters @xmath148 , @xmath66 , @xmath53 and @xmath149 .
we find there is more than an order enhancement in the values of the coefficients @xmath102 and @xmath150 at the nlo level in the case of this vector - like quark model compared to the sm .
however , we should mention here that our nlo results for the np model are not exact in the sense that we have not calculated the expressions for these coefficients at the nlo level at the @xmath30 scale .
the lo results are exact . [
cols="^,^,^,^,^,^,^ " , ] in table [ br ] we display the results for the calculation of the long - distance @xmath151 amplitudes .
the individual numbers for type - i pole and vmd contributions are shown separately along with the branching ratios .
these results are essentially an update of the results in ref . @xcite using the same techniques .
the major updates are : * inclusion of new modes like @xmath152 , @xmath153 , @xmath154 , @xmath155 , @xmath156 in the type - i pole amplitudes . * updated form factors taken from ref .
@xcite used in calculating the vmd amplitudes . * updated @xmath157 ( vector - pseudoscalar ) and @xmath158 ( tensor - pseudoscalar ) decay widths from ref .
@xcite used for the evaluation of the couplings @xmath159 and @xmath160 respectively for the type - i pole amplitudes . * inclusion of @xmath161 mixing in calculating the type - i @xmath162 ,
@xmath163 and @xmath164 amplitudes .
the corresponding mixing angles and decay constants have been obtained from ref .
inclusion of the parity - violating ( pv ) part for the @xmath165 type - i pole amplitude .
the decay constants for the corresponding scalars involved have been taken from ref .
@xcite for @xmath166 and @xcite for @xmath167 respectively . *
the decay constants are taken from ref .
@xcite for the light vector mesons and from ref .
@xcite for the light pseudoscalar mesons .
for the decay constants of the @xmath168 and @xmath169 mesons , we use ref . @xcite . we have only calculated the type - ii pole contribution to the mode @xmath170 .
this is because : * the corresponding decay widths for @xmath171 and @xmath172 essential for calculating the type - ii pole contributions to the @xmath173 and @xmath174 decay modes respectively are only given as limits in ref .
* for decay modes of @xmath33 other than @xmath175 ( for eg .
@xmath176 , @xmath177 etc . ) , the corresponding decay constants for the final state particles are not available .
we provide the effective anomalous dimension matrix in the @xmath38 scheme in this appendix . for the case of charm
decays we have @xmath178 , @xmath179 and @xmath180 .
while evolving the @xmath42 s down from the @xmath53 to the @xmath66 scale , one has to make the assignments @xmath181 and @xmath182 . for the corresponding evolution from the @xmath66 to the @xmath67 scale ,
the values are @xmath183 and @xmath184 . at the lo level , it is given by @xmath185 at the nlo level , due to its large size , we present the matrix in @xmath111 and @xmath112 blocks . it reads @xmath186 @xmath187
in this appendix , we provide the form of the @xmath189 functions mentioned in sections [ vector_like ] taken from ref .
@xmath190 where @xmath191 and @xmath192 in the above , @xmath193 , @xmath194 and @xmath195 with @xmath67 , @xmath196 and @xmath53 denoting the masses of the charm quark , up quark and the w - boson respectively .
the @xmath197 stands for the masses of the down - type quarks running in the fermionic penguin loop for a @xmath0 transition .
hence , for the sm @xmath198 and for the vector - like quark model @xmath199 with @xmath200 , @xmath201 , @xmath66 and @xmath202 standing for the masses of the @xmath44 , @xmath25 , @xmath26 and @xmath144 quarks respectively .
44ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) , link:\doibase 10.1103/physrevd.66.014009 [ * * , ( ) ] , in http://inspirehep.net/record/1392013/files/arxiv:1509.01997.pdf[__ ] ( ) @noop * * , ( ) , link:\doibase 10.1016/s0370 - 2693(00)00731 - 0 [ * * , ( ) ] , link:\doibase 10.1016/0370 - 2693(88)90994-x [ * * , ( ) ] link:\doibase 10.1007/jhep03(2012)070 [ * * , ( ) ] , link:\doibase 10.1088/1742 - 6596/447/1/012061 [ * * , ( ) ] link:\doibase 10.1007/jhep10(2015)027 [ * * , ( ) ] , @noop ( ) , link:\doibase 10.1103/physrevd.92.013002 [ * * , ( ) ] @noop ( ) ,
link:\doibase 10.1103/physrevd.66.054008 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.52.6383 [ * * , ( ) ] , link:\doibase 10.1143/ptp.65.297 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.61.013008 [ * * , ( ) ] , link:\doibase 10.1016/0370 - 2693(96)00694 - 6 [ * * , ( ) ] , link:\doibase 10.1140/epjc / s2002 - 01090 - 5 [ * * , ( ) ] , link:\doibase 10.1007/jhep08(2016)091 [ * * , ( ) ] , link:\doibase 10.1016/s0550 - 3213(01)00366 - 2 [ * * , ( ) ] , link:\doibase 10.1016/s0550 - 3213(98)00131-x [ * * , ( ) ] , link:\doibase 10.1016/j.nuclphysb.2005.01.047 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.95.102004 [ * * , ( ) ] , link:\doibase 10.1016/s0370 - 2693(97)00324 - 9 [ * * , ( ) ] , , link:\doibase 10.1103/physrevd.35.2194 [ * * , ( ) ] link:\doibase 10.1103/physrevd.10.275 , 10.1103/physrevd.11.703.2 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.11.2558 [ * * , ( ) ] link:\doibase 10.1103/physrevd.12.1502 [ * * , ( ) ] link:\doibase
10.1140/epjc / s10052 - 012 - 2241 - 5 [ * * , ( ) ] , link:\doibase 10.1007/jhep08(2012)023 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.49.5894 [ * * , ( ) ] , link:\doibase 10.1007/jhep12(2013)102 [ * * , ( ) ] , under preparation ( s.mandal , a.biswas , n.sinha ) .
link:\doibase 10.1103/physrevd.40.1569 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.118.051801 [ ( ) , 10.1103/physrevlett.118.051801 ] , , link:\doibase 10.1103/physrevlett.112.161801 [ * * , ( ) ] , link:\doibase 10.1088/1126 - 6708/2000/06/029 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.72.034029 [ * * , ( ) ] link:\doibase 10.1088/1674 - 1137/40/10/100001 [ * * , ( ) ] link:\doibase 10.1103/physrevd.85.057501 [ * * , ( ) ] , \doibase http://dx.doi.org/10.1016/j.physletb.2003.10.102 [ * * , ( ) ] link:\doibase 10.1142/s0217732306019955 [ * * , ( ) ] , link:\doibase 10.1016/j.physletb.2006.02.042 [ * * , ( ) ] , link:\doibase 10.1016/j.nuclphysbps.2016.02.030 [ * * , ( ) ] , | we show that for a heavy vector - like quark model with a down type isosinglet , branching ratio for @xmath0 decay is enhanced by more than @xmath1 as compared to that in the standard model when qcd corrections to next - to - leading order are incorporated . in a left - right symmetric model ( lrsm ) along with a heavy vector - like fermion , enhancement of this order can be achieved at the bare ( qcd uncorrected ) level itself .
we propose that a measurement of the photon polarization could be used to signal the presence of such new physics inspite of the large long distance effects .
we find that there is a large region within the allowed parameter space of the lrsm , as well as in the model with vector - like quark with additional left - right symmetry , where the photon polarization can be dominantly right - handed . |
the single - impurity anderson model ( siam ) has been a paradigm of a strongly correlated electron system .
@xcite this seemingly simple model gives rise to dynamical screening of the local spin by the electrons in the fermi sea , leading to a crossover from a weak coupling system at high temperatures ( @xmath0 ) to a strongly coupled one at low @xmath0 , with the relevant temperature scale given by the kondo temperature , @xmath1 .
the recent observation of the kondo effect in quantum dots ( qd s ) , whose parameters can be tuned continuously , and which allow probing of various properties , @xcite has yielded strong theoretical efforts in this direction .
@xcite recent experiments on a qd embedded on one branch of the aharonov - bohm interferometer @xcite created additional interest in kondo effects in _
complex networks_. while the high-@xmath0 behavior of the siam can be adequately described by perturbation theory or poor - man scaling , @xcite and the low-@xmath0 behavior is described by fermi liquid theory , @xcite there is no _ simple _ theory that describes the model s dynamical properties correctly in the whole temperature range , including _ both _ the high- and the low-@xmath0 limits .
this crossover has been described by the computationally demanding numerical renormalization group ( nrg ) approach .
@xcite other methods to describe the crossover are the conserving t - matrix approximation " @xcite ( which overestimates the unitarity sum - rule ) , or the more limited non - crossing approximation " , @xcite and to some extent , quantum monte carlo calculations . @xcite
some of these methods have also been applied to the qd in the aharonov - bohm interferometer .
@xcite however , none of these methods has the flexibility to follow _ analytically _ the effects of various network parameters ( e.g. a magnetic flux ) on experimentally measurable quantities , in particular dynamical ones . in this paper
we discuss the siam for a qd which is embedded in a _
general complex network_. many of the interesting physical properties of the system can be expressed explicitly in terms of the single electron ( retarded ) green function ( gf ) on the dot , @xmath2 , for electrons with spin component @xmath3 and energy @xmath4 ( measured relative to the fermi energy ) . here , @xmath5_+\rangle.\end{aligned}\ ] ] ( for the dot gf , @xmath6 , where @xmath7 destroys an electron with spin @xmath3 on the dot ) .
one example concerns the local density of states ( ldos ) on the dot , @xmath8/\pi.\label{rho}\end{aligned}\ ] ] another example is the conductance @xmath9 between two leads which are connected to sites on the network . for a general mesoscopic structure ,
this ( linear response ) conductance has the form @xcite @xmath10 where @xmath11 is the fermi distribution function ( we set the fermi energy at @xmath12 ) , @xmath13 and @xmath14 is often given in terms of the retarded @xmath15 , the advanced @xmath16 and non - interacting parameters characterizing the network .
for the simple siam ( dot and leads ) , @xmath14 is proportional to @xmath17 .
@xcite here we use the equations of motion method to derive a simple analytic approximate expression for @xmath15 , for a qd on a general network .
our approximate formula exhibits the correct behavior both at high and at very low temperatures . for simplicity
, we consider a qd with a single level ( of energy @xmath18 ) .
electron - electron interactions are assumed to exist only on the qd , and we take their energy @xmath19 to be infinite . in our scheme , one first solves for the gf on the dot _ in the absence of the electron - electron interactions _ ,
i.e. @xmath20 .
this ( spin - independent ) gf can be written in the form @xmath21 with the energy dependent ( complex ) self - energy ( se ) @xmath22 this is easily done exactly , for any finite network ; it only involves the solution of a finite set of linear equations . our approximate explicit expression for @xmath2
is then given in a single equation [ see eq .
( [ ggg ] ) below ] .
somewhat surprisingly , this equation depends on the network parameters _ only _ via @xmath23 .
although approximate , this expression allows for detailed systematic investigations of the kondo effect as function of the system parameters , on a variety of complex networks . even if sometimes only qualitatively correct , such systematic studies help to investigate new physical phenomena on a broad variety of mesoscopic systems
furthermore , the dynamical mean field theory @xcite ( which was developed to address the physics of the periodic anderson and hubbard models ) iterates the local gf of the siam , which is calculated at each stage in terms of the effective se created by the rest of the lattice . since our gf
is easily calculated in terms of @xmath24 , it is an ideal candidate for such calculations .
we are not aware of alternative simple analytic expressions which obey the necessary requirements at both high and low @xmath0 . in sec .
[ ii ] we follow refs . , and derive the equation - of - motion ( eom ) for the dot gf . extending these references , this eom is derived here for a _ generalized case _ , in which the dot sits " within an _ arbitrary complex network_. the eom for the dot gf involves higher - order gf s ( consisting of more operators ) , whose eom s introduce in turn more gf s .
one then terminates this hierarchy by decoupling out averages of operators .
the latter are then found using the fluctuation - dissipation relationship , from the relevant gf s . in this manner ,
the treatment becomes self - consistent .
the successful decoupling must keep as much of the electronic correlations as possible .
@xcite for example , a widely used earlier approximation @xcite neglected some dot - lead correlations and therefore gave reasonable results _ only _ at @xmath25 . for this reason ,
gerland _ et al . _
@xcite had to combine the eom method at high @xmath0 with the nrg at low @xmath0 .
after we correct these earlier calculations , and include all the necessary correlations , we obtain good results for _ all _ @xmath0 .
an earlier application of the eom method solved the integral equations numerically , on the simplest interferometer geometry . @xcite the above decoupling scheme produces an integral equation for @xmath2 , which generalizes that found in refs . and .
a new approximate analytical solution for this equation is found in sec .
the low temperature limit , the fermi liquid conditions , and the kondo behavior are discussed in sec .
[ fermi ] .
unlike earlier papers ( spanning over four decades ) , our solution gives good qualitative results for the _ whole parameter range _ : it has the correct high-@xmath0 behavior and fulfills several low-@xmath0 fermi liquid relations . for the simplest siam , the network "
is represented by a band with a density of states @xmath26 . in the broad band limit
, the self - energy on the qd can be approximated by its value at the fermi energy , @xmath27 , and @xmath28 represents the width of the impurity state in the absence of interactions ( @xmath29 represents the approximately energy - independent dot - band coupling ) . in this case
, our equations reduce to those discussed in the earlier literature .
we emphasize again that even the solution for this simple case is new and non - trivial , since earlier work either did not have a close analytical formula or missed some correlations , leading to wrong results at low @xmath0 .
as stated , our result for @xmath2 depends on the details of the general network only via the non - interacting self - energy , @xmath30 . to demonstrate the power of our result , we present here a few simple examples . section [ iv ] contains explicit results for the simple case of the siam , when the qd is coupled to two one - dimensional leads . for this case , we easily calculate various features of the kondo behavior , including the peak in @xmath31 at the fermi energy , the plateau in the conductance in the so - called unitary limit and the plateau in the transmission phase " at @xmath32 . a second example , of the so called t " network , is solved in sec .
[ t ] . in this case
, one has interference between the wave function on the dot and that on the intersection point .
our simple approximate formula exhibits the fano vanishing of the transmission and the associated so - called anti - kondo effect , as seen experimentally .
@xcite here we also extend earlier theoretical work , which used complicated techniques .
@xcite finally , sec .
[ v ] contains our summary .
our hamiltonian , @xmath33 contains the dot part @xmath34 with a single energy level @xmath35 and an electron - electron interaction @xmath19 ( @xmath36 ) .
all the other parts of @xmath37 can be arbitrary , but non - interacting : @xmath38 describes the network without the dot , and @xmath39 describes the coupling of the dot to the network .
here , @xmath40 creates an electron with spin @xmath3 and energy @xmath41 on the site @xmath42 of the network .
the coefficients @xmath43 and @xmath44 represent tight - binding hopping matrix elements . in the absence of a magnetic flux , these coefficients can be chosen to be real .
the flux then turns them complex , with phases which relate to the aharonov - bohm effect and with @xmath45 .
@xcite the main part of this paper concerns the above general hamiltonian , which assumes no specific details on the structure of the network .
however , for some specific applications , it is convenient to attach a finite mesoscopic network to several semi - infinite leads , which connect to electron reservoirs .
we thus divide @xmath46 into two parts , @xmath47 where @xmath48 has the same form as @xmath49 , except that the sum is restricted to sites which belong to the finite mesoscopic part of the net ( excluding the dot and the leads ) , and @xmath50,\nonumber\\ { \cal h}_l&=&-\sum_{rs } j_{rs } c^{(l)\dagger}_{r\sigma } c^{(l)}_{s\sigma}+h.c .. \label{leads}\end{aligned}\ ] ] here @xmath51 are the indices of the points on the mesoscopic structure which are connected to lead number @xmath52 , while @xmath53 destroys an electron of spin @xmath3 on the site @xmath54 of the lead .
usually , the lead @xmath52 has a continuous spectrum , with eigenstates @xmath55 and eigenenergies @xmath56 , in a band of width @xmath57 .
equation ( [ leads ] ) can then be written as @xmath58,\nonumber\\ { \cal h}_l&=&\sum_{k}\epsilon_k c^{(l)\dagger}_{k\sigma } c^{(l)}_{k\sigma},\label{leads1}\end{aligned}\ ] ] and also @xmath59 with obvious definitions .
when the network consists only of the dot and the leads ( i.e. @xmath60 ) then this has exactly the form of the standard siam model , discussed broadly in the literature .
if all the leads have similar bands , then the electrons in the whole system also have a continuous spectrum in the range @xmath61 .
all integrals over @xmath4 below will be thus over this energy band , and it is due to resonances in such integrals that we need to calculate the retarded gf at @xmath62 .
from now on we return to the general hamiltonian , eqs .
( [ h1]-[h4 ] ) .
the eom for the retarded gf on the qd reads @xmath63 where @xmath64 from now on , we shall use the shorthand @xmath4 for @xmath62 , except where it matters .
specifically , @xmath65 we next define the matrix @xmath66 which represents the system without the dot . in the following , we shall need the inverse matrix , @xmath67^{-1 } , \end{aligned}\ ] ] which is the green function for the system without the dot . with this definition
one finds @xmath68 where @xmath69 we can now calculate the second term on the rhs of eq .
( [ eomg0 ] ) , and show that @xmath70 with @xmath71 at @xmath20 , eq .
( [ eomg0 ] ) thus reduces to eq .
( [ g0 ] ) .
clearly , @xmath24 can be calculated for the _ non - interacting _ case .
all the other gf s can similarly be expressed in terms of @xmath15 .
for example , @xmath72 where @xmath73 .
another way to represent @xmath74 is to start from @xmath75 , and then take the complex conjugate only of the hopping coefficients @xmath76 ( namely reverse all their phases ) , without touching @xmath62 .
similarly , @xmath77 since @xmath78 is an infinite matrix , its inversion may not be trivial .
formally , we can denote the eigenenergies and eigenstates of @xmath46 by @xmath79 and @xmath80 , and then write @xmath81 however , in the examples it is useful first to eliminate the parts which involve the sites on the leads .
examples for this procedure are presented in appendix [ a ] .
the eom for @xmath82 ( with only on - site hubbard interaction on the dot ) does not involve gf s of more arguments : @xmath83 with three new gf s , @xmath84 the inhomogeneous term in eq .
( [ eomgamma ] ) , @xmath85 , represents the average number of electrons with spin @xmath3 on the dot .
this number needs to be determined self - consistently , via the fluctuation - dissipation theorem : @xmath86 we shall return to this condition below .
we now write the eom s for the new gf s , @xmath87 , and use several approximations for their solution .
firstly , it is easy to see from the eom for @xmath88 that this gf is of order @xmath89
. therefore , it contributes to @xmath90 only at order @xmath91 , and can be ignored for @xmath92 [ we need @xmath90 only to order @xmath89 , see eq .
( [ eomg0 ] ) ] .
secondly , we introduce a decoupling scheme for the new gf s that appear in the eom s of the @xmath87 s , @xmath93 we omit thermal averages of the form @xmath94 , which include two destruction ( or creation ) operators , and @xmath95 with @xmath96 , relevant only for states with a net magnetic moment .
the latter assumption means that we restrict the discussion only to symmetric states , with @xmath97 .
using these approximations , the eom for @xmath98 becomes @xmath99 the derivation of this equation also required an additional term , @xmath100 $ ] .
however , this term vanishes . here and below
, we calculate equilibrium thermal averages by using the fluctuation - dissipation theorem , @xmath101 equation ( [ g1 ] ) is practically the same as eq .
( [ gnd ] ) , which was used to derive @xmath24 in terms of the @xmath43 s .
applying the same algebra to eq .
( [ g1 ] ) then yields the analog to eq .
( [ sigext ] ) , @xmath102 using the same approximations , the eom of @xmath103 is @xmath104 in which @xmath105 although the first two terms on the rhs of eq .
( [ gamma2 ] ) are similar to those in eq .
( [ g1 ] ) , we now need to calculate the other two terms . since these terms require thermal averages , which we express using the fluctuation - dissipation eq .
( [ fd ] ) , they end up with integrals which involve @xmath106 , leading finally to our integral equation for the dot gf .
the details of these calculations are presented in appendix [ app2 ] . inserting the results for the necessary thermal averages into the rhs of eq .
( [ gamma2 ] ) , the results are used in eq .
( [ xxx ] ) for the combination @xmath107 which is needed in the rhs of eq .
( [ eomgamma ] ) . adding also the corresponding combination for @xmath108 [ eq . ( [ g1sof ] ) ] , yields our final result for @xmath90 .
it remains to insert it back into the eom ( [ eomg0 ] ) for the dot green function @xmath15 .
this yields an integral equation for this function , @xmath109\nonumber\\ & + & ( 1+\sigma_0 g_{dd\sigma})\int\bigl ( -\frac{d\omega ' } { 2\pi i}\bigr ) \frac{f(\omega ' ) } { \omega -\omega ' + i\eta}\bigl [ g'_{dd-\sigma}(\sigma_0-\sigma ' _ 0)-(\omega ' \rightarrow\omega ' ^{\ast})\bigr ] , \ \ \
\eta\rightarrow 0^{+}.\label{gsof1}\end{aligned}\ ] ] here and below , @xmath15 and @xmath24 are understood to be functions of @xmath110 , and the primes denote a dependence on @xmath111 , with @xmath112 .
two comments are in place here .
first , for the simplest siam , when @xmath113 is independent of @xmath4 , we have @xmath114 , so that only the parts with @xmath115 survive .
furthermore , in this case one can factorize @xmath28 out of the integrals .
this reproduces the integral equation of lacroix , @xcite @xmath116^\ast)\nonumber\\ & -&(1-i\delta_b g_{dd\sigma})\int\bigl ( -\frac{d\omega ' } { 2\pi i}\bigr ) \frac{f(\omega ' ) } { \omega -\omega ' + i\eta}[g'_{dd-\sigma}]^\ast\bigr ] .\end{aligned}\ ] ] 2 however , even for this simple case there has not been an analytic solution that covers the whole parameter range .
second , discarding @xcite correlations between the dot and other sites on the net , e.g. @xmath117 , amounts to neglecting the terms containing @xmath118 in eq .
( [ gsof1 ] ) .
this ends up with a breakdown of the fermi liquid conditions at @xmath119 , and with a bad approximation for @xmath120 .
we now restrict the discussion to non - magnetic " states , and replace @xmath121 . equation ( [ gsof1 ] ) can then be written as @xmath122 where @xmath123 and @xmath124 and @xmath125 are functions of @xmath4 , given by @xmath126,\nonumber\\ & & { \cal b}(\omega ) = \int\bigl ( -\frac{d\omega ' } { 2\pi i}\bigr ) \frac{f(\omega ' ) } { \omega -\omega ' + i\eta}\nonumber\\ & & \times\bigl [ g'_{dd}(\sigma_{0}'-\sigma_{0})-(\omega ' + i\eta \rightarrow\omega'-i\eta)\bigr ] .\label{ab}\end{aligned}\ ] ] for reasons that will be explained later , it is now convenient to replace @xmath127 inside the integrals by @xmath128 + 1/2 $ ] . the part related to the constant 1/2 can then be calculated using the kramers - kronig relations , @xmath129 and thus we find @xmath130,\nonumber\\ & & { \cal b}(\omega ) = \int\bigl ( -\frac{d\omega ' } { 2\pi i}\bigr ) \frac{f(\omega ' ) -1/2}{\omega -\omega ' + i\eta}\nonumber\\ & & \times\bigl [ g'_{dd}(\sigma_{0}'-\sigma_{0})-(\omega ' + i\eta \rightarrow\omega'-i\eta)\bigr ] .\label{abnew}\end{aligned}\ ] ] we next follow lacroix , @xcite and assume that the integrals are dominated by the region @xmath131 , namely that @xmath132 and thus @xmath133 where @xmath134 and @xcite @xmath135,\end{aligned}\ ] ] ( @xmath136 is the digamma function ) . a similar procedure yields @xmath137 . \label{aapprox}\end{aligned}\ ] ] the reason for using the above transformation is that when one analyzes the equivalent equations for finite @xmath19 , then one requires particle hole symmetry , namely the relation @xmath138 where the ` tilde ' denotes particle - hole transformed quantities . in that case , this relation is equivalent to the relations @xmath139 these relations should hold also for any approximate solution .
@xcite it is easy to check that these relations are obeyed by our approximate expressions ( [ bapprox ] ) and ( [ aapprox ] ) , but would not hold if we made the approximation ( [ lacapp ] ) on the original equations , before shifting @xmath127 by 1/2 . @xcite taking @xmath140 as a parameter , and defining the scaled variable @xmath141 the solution of eq .
( [ gsof ] ) can be written as @xmath142 with @xmath143^{1/2},\label{qq}\\ { \rm s}&=&z^2+\frac{9}{16}-z\re x+(\delta n-\frac{3}{4})\im x , % \frac{1}{4\delta^{2}|g|^{2}}-\frac{\re [ { \rm % x}^{\ast}/g]}{2\delta } + \delta n\im{\rm x}. \label{ss}\end{aligned}\ ] ] and we have omitted the explicit dependence on @xmath4 for brevity . for some purposes it is more convenient to write eq .
( [ apprg ] ) as @xmath144 where the real functions @xmath145 and @xmath146 , which result from the strong interactions on the dot , depend on @xmath0 and on @xmath23 and are given by @xmath147 equation ( [ ggg ] ) forms our general result .
we emphasize again that all we need to know is the non - interacting self - energy , @xmath30 . as seen explicitly from the above equations , our analytic expression for @xmath148 depends on the parameter @xmath149 , which needs to be determined self - consistently , via eq .
( [ selfcon ] ) .
we discuss this condition in a specific example below .
at @xmath119 , one easily sees that @xmath150 diverging at the fermi energy @xmath151 . at @xmath152 we thus have @xmath153 and @xmath154 , with @xmath155}{[z+\sqrt{z^2+\delta n(1.5-\delta n)}]^2+\delta n^2},\end{aligned}\ ] ] and @xmath156/[2\delta(0)]$ ] . therefore , @xmath157+i\bigr ) } .\label{gt0}\end{aligned}\ ] ] for a very deep level on the dot , @xmath158 , @xmath159 approaches @xmath160 , and @xmath148 approaches @xmath161 , ending up with high plateaus in the ldos and in the phase of the complex @xmath162 sticking to @xmath32 .
all of these characteristics , which are hallmarks of the kondo behavior , @xcite are observed in the examples below . as @xmath35 moves up from large negative values , eq .
( [ gt0 ] ) shows a gradual crossover away from this unitary " limit , and @xmath148 becomes mainly real and small as @xmath163 .
( [ gt0 ] ) also ensures that the gf obeys the fermi - liquid relations @xcite at @xmath119 : ( i ) the imaginary part of the se on the fermi level is the same as in the absence of interactions , i.e. @xmath164 , confirming the unitarity limit ; @xcite ( ii ) for @xmath165 , the ldos there [ i.e. , @xmath166 approaches the finite value @xmath167 $ ] .
we emphasize that these results apply for _ any _ complex network ; the only input is @xmath30 .
the kondo effect concerns the behavior of the density of states near the fermi energy at low temperature . in that regime , @xmath148 is dominated by its imaginary part , and thus @xmath168.\end{aligned}\ ] ] for @xmath169 , rne has @xmath170 , @xmath171 and the leading @xmath4-dependence of @xmath172 is @xmath173 we thus end up with a logarithmic cusp in @xmath174 at @xmath12 .
the related narrow peak reaches one half of its peak value ( equal to @xmath167 $ ] ) when @xmath175 .
one definition of the kondo energy @xmath1 is to identify @xmath1 with the half - width of this peak . solving @xmath176 and using eq .
( [ asym ] ) thus yields @xmath177 , i.e. @xmath178 .
this result , which agrees with that of lacroix , @xcite is qualitatively similar but quantitatively different from the presumably exact @xmath1 as given by haldane , @xcite which has @xmath179 .
however , @xmath1 only represents some crossover energy scale , and we expect the above solution to follow the qualitative variations of @xmath1 with the system parameters , which are contained in @xmath180 and in @xmath181 .
it should be emphasized that @xmath1 depends on _ both _
@xmath181 _ and _ @xmath180 ( some authors ignore the part coming from the real part of the se , @xmath180 ) . for example , for the aharonov - bohm interferometer
@xmath1 oscillates with the flux .
@xcite at the fermi energy , @xmath12 , and for fixed @xmath182 , one expects @xmath183 to decrease with increasing @xmath0 . in this limit , @xmath184 , with @xmath185 . for @xmath186
we then find the asymptotic relation @xmath187 this density of states decreases to one half of its maximum @xmath119 value at a temperature @xmath188 , with @xmath1 given in eq .
( [ tkondo ] ) .
although the result ( [ rhot ] ) is qualitatively correct , and is also consistent with our @xmath1 , its explicit low - temperature dependence disagrees with exact expectations : we obtain a logarithmic behavior , whereas fermi liquid theory predicts a @xmath189 dependence at low @xmath0 .
@xcite this is an artifact of our approximation .
however , at high @xmath0 we do recover the usual logarithmic variation . in the opposite limit , of @xmath190 ,
one finds that @xmath191 , and eq .
( [ apprg ] ) implies that @xmath192 approaches @xmath193 .
the appearance of @xmath140 in the numerator , in place of 1 , results from the fact that some weight of the spectral function is pushed to infinite @xmath19 . @xcite
another interesting point concerns the factor @xmath194 in eqs .
( [ gg ] ) and ( [ ggg ] ) .
this factor may be explained heuristically @xcite by the fact that while both spin directions are accessible for tunnelling into the dot , only a single one can tunnel out of it .
in the following two examples , we assume simple semi - infinite one - dimensional leads , with identical nearest - neighbor hopping matrix elements , @xmath195 , and with a lattice constant @xmath196 .
the eigenenergies of each @xmath197 are therefore @xmath198 , with eigenfunctions @xmath199 ( @xmath200 is the length of the lead ) , and @xmath201 .
we also assume that the leads @xmath52 and @xmath202 are attached to the mesoscopic network only at one site , with hopping matrix elements @xmath203 and @xmath204 .
for the simple siam , when the two leads are directly connected to the dot , we show in appendix [ a ] that the leads generate a self - energy @xmath205 on the dot , with @xmath206 determined by @xmath207 and with @xmath208 here we have used the notation @xmath209 with @xmath210 being the density of the band states .
the width of the non - interacting resonance at @xmath35 is thus equal to @xmath211 . in the figures presented below we use the symmetric case , @xmath212 . substituting @xmath213 into eqs .
( [ apprg])-([ss ] ) then yields @xmath148 .
we start by discussing the self - consistency condition for @xmath214 .
figure [ nd ] shows an example of our self - consistent solutions for @xmath214 , at @xmath119 : we start from an initial guess for @xmath140 , then calculate @xmath214 from eq .
( [ selfcon ] ) , and iterate ; the procedure converges after a few iterations .
also shown is @xmath215 , which represents the total change in the electron occupation in the system due to the dot , as determined from the friedel phase , defined via @xcite @xmath216 \equiv \re[g_{dd}(0)]/\im[g_{dd}(0 ) ] .
\label{friedel}\end{aligned}\ ] ] for @xmath217 , @xmath215 approaches 1 , while @xmath214 remains slightly smaller .
this small difference could reflect an additional occupation of other sites in the leads .
it could also result from the inaccuracy of our approximation for @xmath218 , which becomes worse as @xmath4 moves away from the fermi energy ; the integral in eq .
( [ selfcon ] ) contains contributions from all @xmath4 .
however , this small difference has only a small effect on the other calculations presented below .
= 6.truecm although it is easy to solve for @xmath214 for each set of parameters and use the resulting @xmath140 for other calculations , the qualitative results are only weakly affected if one uses an arbitrary smooth variation of @xmath214 from 0 to 1 as @xmath35 varies from @xmath219 to @xmath220 . in the calculations below
, we used such a smooth variation for @xmath140 , with a width of order @xmath28 .
an alternative approximate expression for @xmath140 follows from eq .
( [ apprg ] ) .
it turns out that for most of the integration range in eq .
( [ selfcon ] ) , @xmath162 is dominated by the first term there . neglecting the second term
, one has @xmath221 , with @xmath222/\pi$ ] depending only on the non - interacting parameters .
thus , @xmath223 . for @xmath28 between 0.1 and 0.01 and for @xmath224 , we find this estimate to be within @xmath225 from the full self - consistent value .
figure [ dost ] presents the ldos [ eq .
( [ rho ] ) ] versus @xmath4 , for the simple siam , with the parameters as indicated .
plots at lower @xmath0 are indistinguishable from the one shown at @xmath226 .
our approximation reproduces the kondo peak at low @xmath0 , in addition to the much broader peak at @xmath35 .
figure [ dosed ] shows @xmath227 at the fermi energy ( @xmath12 ) versus @xmath35 .
note the very slow ( logarithmic ) increase of @xmath228 with decreasing @xmath0 , in agreement with eq .
( [ rhot ] ) .
= 6.truecm = 6.truecm we now turn to the conductance , @xmath9 . for the simple siam
, one has @xcite @xmath229 equation ( [ ggg ] ) implies that @xmath230 for @xmath119 , when @xmath231 , we also have @xmath232 . at finite temperatures , these two expressions for the conductance exhibit the same qualitative behavior as @xmath227 ( fig .
[ dosed ] ) .
the quantitative difference between the two expressions , which increases with @xmath0 and with large negative @xmath35 , represents a breakdown of unitarity which may be an artifact of our approximation . at low @xmath0
, @xmath233 is practically a delta function , and eq .
( [ cond ] ) yields @xmath234 , exhibiting the qualitative behavior shown in fig .
[ dosed ] . at
higher @xmath0 , the peak in @xmath9 slightly below @xmath235 becomes lower and broader than that of @xmath236 . at fixed @xmath237 , increasing @xmath0 results with a decreasing @xmath9 , but with an interesting superimposed peak at @xmath238 ) ( when the peak at @xmath35 starts to contribute ) .
some experiments @xcite place the qd on one branch of an open aharonov - bohm interferometer , and attempt to extract the _ transmission phase _ for scattering through the dot . the transmission phase is usually related to the phase of @xmath148.@xcite without getting into the question of what is really measured in the interferometer , @xcite it is still of interest to study the latter phase .
figure [ phase ] thus shows the transmission " , represented by @xmath239 , and the transmission phase " , represented by [ see eq .
( [ friedel ] ) ] @xmath240 as function of @xmath35 for several temperatures .
interestingly , at high @xmath0 this phase simply grows smoothly from zero to @xmath241 through the resonance , similarly to the non - interacting case or to the coulomb blockade case .
however , as @xmath0 decreases , the peak in the transmission " broadens towards negative @xmath35 , eventually reaching a plateau for @xmath119 ( see also fig .
[ dosed ] ) . at the same time , the phase develops an intermediate plateau at @xmath32 .
@xcite this plateau begins at an energy @xmath35 which is roughly given by @xmath242 , as defined in eq .
( [ tkondo ] ) . studying
the energy where this phase grows from zero to @xmath32 thus suggests another way to define the crossover temperature @xmath1 .
our equations also become very simple for the t " network , when the dot sits on a side branch . such a network has recently attracted both theoretical @xcite and experimental @xcite interest , as the simplest realization of the fano - kondo effect . the mesoscopic network now consists of a single site 0 " .
this site has one bond connected to the dot , with hopping @xmath243 , and two bonds connected to two leads , with hopping @xmath203 and @xmath204 . in this case , the matrix @xmath244 in appendix [ a ] is of order @xmath245 , and eqs .
( [ dd ] ) and ( [ aaa ] ) yield @xmath246 $ ] . using @xmath28 from eq .
( [ delta ] ) , we have @xmath247 in this case , @xmath248 of eq .
( [ cond ] ) is given by @xcite @xmath249 the calculation of @xmath250 follows directly from eqs .
( [ gdm ] ) and ( [ glm ] ) , @xmath251 therefore , @xmath252 is a linear combination of @xmath253 and of @xmath254 , with coefficients which depend only on the non - interacting parameters .
these equations reproduce those found e.g. in refs . and .
however , at this point those authors use complicated numerical schemes to obtain @xmath148 , which do not allow for systematic studies of the dependence on the various parameters .
in contrast , we can easily obtain the approximate @xmath148 analytically , with the same ease as for the previous example .
all we need to do is to substitute eq .
( [ sig0 t ] ) into eqs .
( [ apprg])-([ss ] ) . for the non - interacting case ,
@xmath148 is simply given by eq .
( [ g0 ] ) .
figure [ tcond ] then compares the results for the conductance ( in units of @xmath255 ) , with and without interactions . without interactions ( thin lines ) , for @xmath256 the upper panel in fig .
[ tcond ] shows a symmetric fano vanishing of the conductance at @xmath235 . as @xmath257 becomes more negative
( lower panel there ) , the conductance assumes a typical non - symmetric fano shape , with smaller conductances at large @xmath258 .
the interactions ( thick lines ) have a negligible effect at @xmath259 , when the dot is not occupied .
however , at @xmath182 and @xmath119 ( thickest lines ) the interactions cause strong changes . for @xmath256 and large negative @xmath35 ,
the occupation on the dot causes the so - called anti - kondo effect , @xcite where the formation of the kondo singlet causes destructive interference which yields zero conductance . for @xmath260 , the fano zero
is shifted to lower @xmath35 , the peak disappears and the asymptotic conductance increases .
as @xmath0 increases ( medium thick lines ) , the behavior at negative @xmath35 gradually returns to that of the non - interacting case , and the fano zero disappears : the conductance is always non - zero .
again , all of these phenomena are qualitatively similar to those found in refs . , and .
= 6truecm a slightly more complex t - network is obtained by placing another non - interacting site , 1 " , between the site 0 and the dot , with energy @xmath261 and with hopping elements @xmath243 to both site 0 and the dot .
our mesoscopic system now contains two sites , 0 and 1 , and the methods of appendix [ a ] yield @xmath262 @xmath263 with @xmath264 figure [ t11 ] shows a few examples on how the interplay between @xmath257 and @xmath261 can change the dependence of the conductance on @xmath35 for the fully interacting case at @xmath119 . for @xmath265 , the graph looks exactly like that for the simple siam , discussed in sec .
v. clearly , the presence of the intermediate point turned the destructive interference into a constructive one .
changing @xmath257 to positive ( negative ) values then simply shifts the whole curve to the left ( right ) .
changing @xmath261 to non - zero values generates either a `` fano '' zero in the conductance ( when @xmath260 ) , similar to the results in fig . [ tcond ] , or a fano " resonance ( when @xmath266 ) . at @xmath256 one observes a change from a zero to a resonance as @xmath261 changes from positive to negative values .
all of these parameters can be easily changed using the various gate voltages , e.g in the setup of ref . .
without going into much further discussion , it is clear that the intermediate point on the side branch is very effective in changing the interference pattern between fano - kondo resonances and anti - resonances .
data from such experiments can then be used to obtain information on @xmath148 .
we have derived an approximate analytic expression for a complex mesoscopic network , which contains a quantum dot with electron - electron interactions , and which may connect to several leads .
our formulae correct and generalize earlier expressions , and give a good qualitative interpolation between the fermi liquid behavior at very low temperatures and the simpler high temperature one .
although our formulae reproduce many features required by the fermi liquid theory , they are still approximate , and should thus only be used for discussing the qualitative variation of various quantities on the parameters characterizing the network .
however , the simplicity of our expressions allows for relatively easy comparisons with experiments and with more complicated numerical work .
we have demonstrated the use of our formulae for the two simple cases of a single quantum dot attached to two leads and of the
t " network , where the dot is connected to another site which couples to the two leads .
indeed , we have reproduced and extended all the expected phenomena for these two cases . in an upcoming publication @xcite we shall apply this scheme to the interesting case of the aharonov - bohm interferometer .
* acknowledgements * we acknowledge helpful discussions with a. schiller , and support to ym from the israel science foundation ( isf ) .
this project was carried out in a center of excellence supported by the isf ( oe and aa ) . 2
we need to invert the matrix @xmath268 .
we start by dividing our hilbert space into two parts , called 1 " and 2 " . then
, we write @xmath269 and the inverse matrix @xmath270 as @xmath271 it is now easy to obtain the dyson - like equation , @xmath272^{-1}m_{21}\bigr ) ^{-1}.\label{dd}\end{aligned}\ ] ] in our calculations we only need @xmath273 for @xmath274 within the finite mesoscopic part of the network ( possibly including the dot ) . identifying this part with the subspace
1 " above , we thus only need the matrix elements of the finite matrix @xmath244 . to obtain these via eq .
( [ dd ] ) , we need the green function for the leads , @xmath275^{-1}$ ] . for our eq .
( [ leads ] ) , the matrix @xmath276 separates for the different leads , and we end up with @xmath277^{-1},\label{minv1}\end{aligned}\ ] ] with the self - energy matrices @xmath278 here , the matrices @xmath276 and @xmath279 are all of order @xmath280 , where @xmath281 is the number of sites in the mesoscopic part of the network .
also , @xmath282 is the green function for the disconnected lead @xmath52 . for the one - dimensional leads , we use the identity @xmath283 with @xmath284 , to obtain the self - energy due to the lead @xmath52 , which is attached at the single point @xmath51
, @xmath285 the same result also applies when the mesoscopic part is empty , and the leads are connected directly to the dot .
following the same logic that led to eqs .
( [ sigext ] ) and ( [ g1sof ] ) , the formal solution to eq .
( [ gamma2 ] ) gives @xmath287.\label{gg2}\end{aligned}\ ] ] calculating the thermal average @xmath288 via the fluctuation - dissipation eq .
( [ fd ] ) , and using eqs .
( [ gmd ] ) and ( [ gdm ] ) , we end up with an integral which requires the sum @xmath289 for this and for similar sums , it is helpful to use the identity @xmath290 which follows from eq .
( [ mm ] ) , together with a careful use of @xmath291 . using eq . ( [ sig0 ] ) , the sum in eq .
( [ sum1 ] ) then becomes @xmath292.\end{aligned}\ ] ] @xmath294\nonumber\\ & -&(1+\sigma_0g_{dd\sigma})\int\bigl ( -\frac{d\omega ' } { 2\pi i}\bigr ) \frac{f(\omega ' ) } { \omega -\omega ' + i\eta}\bigl [ g'_{dd-\sigma}(\sigma_0-\sigma ' _ 0)-(\omega ' \rightarrow\omega ' ^{\ast})\bigr ] .\label{xxx}\end{aligned}\ ] ] for a review , see e.g. p. fulde , j. keller and g. zwicknagel , solid state phys .
* 41 * , 1 ( 1988 ) .
d. goldhaber - gordon , h. shtrikman , d. mahalu , d. abusch - magder , u. meirav and m. a. kastner , nature * 391 * , 156 ( 1998 ) . for a review , see e.g. n. s. wingreen , materials science and engineering b * 84 * , 22 ( 2001 ) .
w. g. van der wiel , s. de franceschi , t. fujisawa , j. m. elzerman , s. tarucha , and l. p. kouwenhoven , science * 289 * , 2105 ( 2000 ) ; y. ji , m. heiblum , d. sprinzak , d. mahalu and h. shtrikman , science * 290 * , 779 ( 2000 ) ; y. ji , m. heiblum and h. shtrikman , phys . rev . lett . * 88 * , 076601 ( 2002 ) .
see , e.g. , n. sivan and n. s. wingreen , phys .
b * 54 * , 11622 ( 1996 ) and references therein .
p. nozires , j .
* 17 * , 31 ( 1974 ) ; k. yamada , prog .
. phys . * 53 * , 970 ( 1975 ) ; * 54 * , 316 ( 1975 ) .
t. a. costi and a. c. hewson , j. phys .
matter * 5 * , l361 ( 1993 ) .
s. kirchner , j. kroha , and p. wlfle , phys .
b * 70 * , 165102 ( 2004 ) . n. e. bickers , d. l. cox and j. w. wilkins , phys . rev .
b * 36 * , 2036 ( 1987 ) ; q. qin and h. keiter , z. phys .
b * 65 * , 89 ( 1991 ) . j. e. hirsch and r. m. fye , phys .
* 56 * , 2521 ( 1986 ) ; r. n. silver , j. e. gubernatis , d. s. sivia and m. jarrell , phys . rev . lett . *
65 * , 496 ( 1990 ) .
w. hofstetter , j. knig , and h. schoeller , phys .
lett . * 87 * , 156803 ( 2001 ) ; j. knig and y. gefen , phys .
lett . * 86 * , 3855 ( 2001 ) ; phys .
b * 65 * , 045316 ( 2002 ) ; c. h. lewenkopf and w. a. weidenmller , cond - mat/0401523
. for a recent review , see e.g. a. george , cond - mat/0403123 , l. dworin , phys . rev . * 164 * , 818 ( 1967 ) .
this paper shows that all self - consistent methods , to any order of truncation , have difficulties near the particle - hole symmetry point .
such difficulties should not show up in the limit @xmath92 discussed here .
j. a. appelbaum and d. r. penn , phys .
b * 3 * , 942 ( 1971 ) . c. lacroix , j. phys .
f * 11 * , 2389 ( 1981 ) .
luo , z - j .
ying , and s - j .
wang , phys .
b * 59 * , 9710 ( 1999 ) .
u. gerland , j. von delft , t. a. costi and y. oreg , phys .
lett . * 84 * , 3710 ( 2000 ) .
b. r. bulka and p. stefaski , phys .
* 86 * , 5128 ( 2001 ) .
d. r. hamann , phys . rev . *
158 * , 570 ( 1967 ) ; p. e. bloomfield and d. r. hamann , phys . rev . * 164 * , 856 ( 1967 ) .
k. kobayashi , h. aikawa , a. sano , s. katsumoto and y. iye , phys .
b * 70 * , 035319 ( 2004 ) .
k. kang , s. y. cho , j. j. kim and s. c. shin , phys .
b * 63 * , 113304 ( 2001 ) ; m. e. torio , k. hallberg , a. h. ceccatto and c. r. proetto , phys . rev .
b * 65 * , 085302 ( 2002 ) ; a. a. aligia and c. r. proetto , phys .
b * 65 * , 165305 ( 2002 ) .
a. aharony , o. entin - wohlman and y. imry , phys .
lett . * 90 * , 156802 ( 2003 ) and references therein . n. e. bickers , rev .
mod . phys . * 69 * , 845 ( 1987 ) , table ix .
d. c. langreth , phys . rev . * 150 * , 712 ( 1966 ) . | we study the kondo effect of a quantum dot placed in a complex mesoscopic structure . assuming that electronic interactions are taking place solely on the dot , and focusing on the infinite hubbard interaction limit , we use a decoupling scheme to obtain an explicit analytic approximate expression for the dot green function , which fulfills certain fermi - liquid relations at zero temperature .
the details of the complex structure enter into this expression only via the self - energy for the non - interacting case .
the effectiveness of the expression is demonstrated for the single impurity anderson model and for the t - shaped network . 2 |
during the recent upsurge of activity in the field of bose - einstein condensation ( bec ) , some older questions concerning the weakly interacting bose gas have been revived .
one such un - resolved question is the effect of a weak repulsion on the critical temperature @xmath0 .
calculations on this question has produced widely dissimilar results : increases and decreases of @xmath0 proportional to @xmath1 , @xmath2 , @xmath3 , etc .
have been reported by various authors .
( here @xmath1 is the scattering length . )
it is only during the past 2 - 3 years that the community has been converging toward a consensus . considering the diverse approaches to the @xmath0 problem , and the variety of issues that have arisen in this connection
, it seems appropriate at this point to attempt to draw some connections between the different approaches , and point out some misconceptions and fine points .
the purpose in this article is to review and comment on existing work , rather than to present new calculations .
the model is described by the hamiltonian @xmath4 where @xmath5 is the free - gas spectrum , and @xmath6 , @xmath7 are bosonic operators . in principle
, one is interested in any possible interaction function @xmath8 , but the simplest possible form is a delta function in real space , so that @xmath9 is momentum - independent .
to first order @xmath10 is related to the @xmath11-wave scattering length @xmath1 by @xmath12 .
a dimensionless measure of the interaction is the quantity @xmath13 , where @xmath14 is the density .
the program is to examine the shift in the critical temperature @xmath0 , as compared to a noninteracting gas of the same density , as a function of @xmath13 , for small @xmath13 . in section [ sect_background ] ,
the background is set by summarizing different calculations and basic results .
section [ sect_heart ] contains the comments on various approaches to the @xmath0 problem . in the concluding section [ sect_further ] ,
i suggest some directions for additional calculations .
early predictions for the shift in transition temperature appear in the work of lee , yang and huang , around forty years ago , as summarized in @xcite .
recently , huang has published @xcite a prediction for an increase in @xmath0 due to the inter - particle repulsion , proportional to the square root of the interaction strength , @xmath15 .
schakel in a short note @xcite , and baym _ et al _ in @xcite , have both discussed the problem with this work ; in particular @xcite contains an analysis showing how huang s result is generated spuriously from a truncation of the virial expansion used at mean field level .
a similar spurious result appearing in the literature is that of toyoda @xcite , who in 1982 performed mean - field - like calculations deriving a landau - ginzburg type thermodynamic function .
his prediction was a _
decrease _ of the transition temperature , @xmath16 .
again , this is believed to be an artifact of mean field approximations .
the present understanding is that there is no mean field effect on the transition temperature , and that higher - order correlations are required to find the leading shift in @xmath0 . a more recent work that also predicts a negative shift in @xmath0 is the canonical - ensemble work of wilkens _ et al _ @xcite .
this work has been critiqued in @xcite .
i discuss some related issues in section [ sect_canonical ] .
in addition , in view of the continuing appearance of negative - shift predictions , i give in section [ sect_direction+ir ] a simple argument for the _ increase _ of the transition temperature due to the addition of a repulsive interaction .
this argument makes use of eq , which is derived independent of any particular method of treating interaction effects . during the past several years
, a large body of work has been published by baym , lalo and collaborators @xcite , and a number of authors have done followup work @xcite .
these workers predict a positive shift in @xmath0 that to leading order is _ linear _ in the scattering length ; @xmath17 .
attempts to calculate the coefficient @xmath18 , however , have continued to give fluctuating results , and this has prompted the baym group to predict @xcite a logarithmic contribution at the _ next _ order , i.e. , a @xmath19 contribution .
the presence of such a non - analyticity might explain the difficulty in any numerical estimate of @xmath18 .
stoof s renormalization - group analysis @xcite of the three - dimensional bose gas has led to a prediction @xcite of an increase of the transition temperature proportional to @xmath20 , i.e. , a non - analyticity at _ leading _ order .
a leading non - analytic behavior would actually explain even better the uncertain results obtained in attempts to calculate the coefficient @xmath18 , based on the assumption of linear shift @xmath21 .
an older attack on the @xmath0 problem is that of kanno @xcite . in a series of papers ,
kanno introduced a `` quasilinear '' canonical transformation approach to the many - body problem , which he used to calculate ( among other things ) the free energy and hence transition temperature of the bose gas .
since kanno s work has mostly been ignored by the current @xmath0 community , i give a description of his work in section [ sect_kanno ] .
i also show how some of his derivations can be performed in a more transparent way
. schakel s work on developing an effective theory @xcite of the weakly non - ideal bose gas has led to the prediction @xcite of a shift in the transition temperature that is strictly linear in the interaction .
the connection of this calculation with other approaches has not yet been explored in detail .
some issues are pointed out in section [ sect_schakel ] .
our recent report @xcite addresses the question of @xmath0 within a quasiparticle framework .
other than working out details of a quasiparticle description ( ultraviolet divergence , effect of quasiparticle lifetime ) , we also provide a non - selfconsistent method of regularizing the infrared divergences appearing in the description of the transition point .
this results in a shift of @xmath0 approximately @xmath22 @xmath23 .
the procedure is uncontrolled ; however , this result points strongly to a non - analytic contribution in the dependence of @xmath0 on the interaction .
other relevant calcultaions include kleinert s `` five - loop '' determination @xcite of the coefficient @xmath18 mentioned above , assuming a puely linear leading shift , and the `` linear @xmath24-expansion '' calculations @xcite of de souza cruz _ et al_. kleinert has questioned @xcite the @xmath24-expansion method , but a detailed analysis of the relationship of this approach to others seems to be lacking in the literature . for a momentum - independent potential , the two mean field contributions to the self - energy ( hartree and fock ) are equal , @xmath25 .
since the self - energy is momentum - independent , it can simply be absorbed into a redefinition of the chemical potential .
@xmath26 here @xmath27 .
the transition now simply corresponds to @xmath28 instead of @xmath29 , and @xmath0 remains unchanged .
the lesson here is that a momentum - dependent self - energy is required for a shift in @xmath0 . in second - order calculations ,
( e.g. in @xcite , ) it is usual to have the 1st - order self - energy already absorbed in the chemical potential , i.e. , to use @xmath30 as the chemical potential .
this allows one to exclude those second - order diagrams which can be obtained by 1st - order self - energy insertions into 1st - order diagrams .
_ spatially extended interaction_. in a real system , of course , the description of the interaction in terms of a single parameter ( @xmath31 or @xmath1 ) is valid only for very weak interactions or at very low temperatures / densities ; if these conditions are not satisfied one has to take into account the detailed shape of the potential @xmath32 which is not a delta function in real space ( not a constant in momentum space ) any more . in such a case
we get a momentum - dependent fock term , and hence a contribution to @xmath0 from the first - order self - energy . assuming a gaussian @xmath33 of width @xmath1 and height @xmath34 , one gets a correction to the spectrum that can be incorporated as an effective mass @xmath35^{-1}$ ] , resulting in a negative shift @xcite : @xmath36 the negative shift at higher densities is present in liquid helium and presumably in other bose fluids
. however , for weak interactions , this shift is less important than the leading @xmath37 effect , which is the current theoretical challenge . as a highly correlated liquid
, the properties of liquid helium are not expected to be described well by the dilute weakly - interacting bose gas model .
however , the deviation of the lambda transition temperature ( 2.2k ) , from the bec temperature of an ideal gas with the same density ( 3.1k ) , can be mostly explained using a simple effective mass description . in other words ,
the experimental @xmath0 is reproduced quite closely by replacing the helium atomic mass @xmath38 by a quasiparticle effective mass @xmath38 * in the ideal gas expression @xmath39^{2/3}\rba n^{2/3}$ ] .
the effective mass in the liquid exceeds the bare mass , because of the inertia of the medium which has to make way for any atom to move .
this results in a lowering of @xmath0 .
the effect is essentially equivalent to that described in the last section for a momentum - dependent interaction .
feynman @xcite calculated a partition function for liquid helium that takes into account the effective mass enhancement described above . using this partition function ,
chapline calculated @xcite a @xmath0 quite close to the actual @xmath40-point of helium .
this is equivalent to using the effective mass in the ideal gas expression for @xmath41 . according to current understanding ,
therefore , one expects the following variation of @xmath0 with @xmath42 : an initial linear - like increase , followed by a maximum and then a decrease due to the effective - mass mechanism described above , until eventually @xmath43 becomes negative .
this kind of dependence has been seen in the monte - carlo calculation of grter , ceperley , lalo @xcite , and in reppy s experiments @xcite described below .
one could think of studying the @xmath0 problem experimentally , e.g. , in the context of trapped atomic bose gases .
unfortunately , trap effects dominate in the dependence of @xmath0 on the interaction for these systems .
turning on a repulsion for a trapped bose gas causes an expansion of the gas , reducing the density and hence decreasing the transition temperature @xmath44 @xcite .
in contrast to the intrinsic effect under review here , trapping effects on @xmath0 are found at mean - field level already .
a different experimental realization of the weakly non - ideal bose gas is the @xmath45he - vycor system explored by reppy _ et al _ @xcite . here
the small-@xmath13 regime is reached by enormously decreasing the @xmath45he density .
reppy s investigation of the transition temperature shows a linear - like increase of @xmath0 at low @xmath13 , followed by a maximum and a decrease to values lower than @xmath41 .
the general behavior is satisfying , but for the initial increase , the data is not yet good enough to choose between @xmath46 , @xmath47 , or @xmath48-like behavior .
the work of baym , lalo and collaborators @xcite establishes definitely that a repulsive interaction _ increases _
@xmath0 , and that the dependence of @xmath43 on the interaction is of linear order , as opposed to quadratic or @xmath49 .
one should note , however , that the baym group s work relies heavily on power - counting and dimensional arguments .
this kind of argument typically does not catch logarithmic corrections .
the possibility of leading non - analytic dependence of @xmath43 on the interaction , therefore , needs to be further examined .
it has been suggested @xcite that the large-@xmath50 calculation @xcite is a reliable indicator that the leading shift is purely linear , because of intuition gained from other systems where large-@xmath50 calculations capture all logarithms .
i would like to point out , however , that the @xmath51 models belong to a different universality class , and have different excitation spectra as compared to the usual ( @xmath52 ) bose gas .
more care is therefore required in interpreting results from the large-@xmath50 approach .
the baym - lalo collaboration seems to have been the first to relate the single - particle spectrum at @xmath0 to the shift @xmath43 .
another noteworthy aspect is the use of ursell operators @xcite , and a corresponding diagram technique , to describe particle correlations .
( as far as i can tell , for this problem , the ursell operator technique does not give any extra information compared to the usual green function techniques . ) once the shift in @xmath0 is related to the modification of the single - particle spectrum , the direction of the shift can be argued using well - known facts about the spectrum of the bose gas .
we will use the following expression for the shift , as appearing in @xcite : @xmath53 here @xmath54 is the bose function at the critical point ; @xmath55 ; and @xmath56 is the quasiparticle spectrum at @xmath0 , i.e. , @xmath57 . the expression used by baym _
et al _ @xcite , @xmath58 $ ] , is the high - temperature ( zero matsubara frequency ) version of . noting that the major contribution to comes from the infrared
, one can write approximately @xmath59 \ , \ , .\ ] ] this equation might introduce or leave out logarithmic corrections ; however the following argument is unchanged for . if the quasiparticle spectrum @xmath56 is `` harder '' than the bare spectrum @xmath60 , i.e. , if @xmath56 is sub - quadratic in @xmath61 or quadratic with an effective mass @xmath62 , then the @xmath63 integral dominates and the shift is positive . on the other hand
if the spectrum at the transition point is `` softened '' by the interaction , then eq [ or ] leads to a decrease in the transition temperature .
since the weakly interacting bose gas is known to have a linear ( bogoliubov ) infrared spectrum below the critical temperature , and a quadratic spectrum above @xmath0 , it is extremely reasonable to assume , even without any rg or perturbation - theoretic calculations , that the spectrum at the critical point should be something intermediate between linear and quadratic .
as pointed out above , this means a _ positive _ shift in the transition temperature .
the long - wavelength spectrum at the transition point should be @xmath64 , where @xmath65 is the anomalous dimension .
the present system falls in the same universality class as the @xmath66 model , or the @xmath52 quantum rotor model . for this universality class , large-@xmath50 and @xmath67-expansion calculations
give respectively @xmath68 and @xmath69 , and a recent monte carlo calculation @xcite has produced the value @xmath70 .
second - order treatments of the self - energy ( e.g. , the baym group s @xcite and our @xcite work ) over - estimates the modification of the spectrum , giving a @xmath71 infrared behavior ( @xmath72 ) .
on the other hand , some other treatments ( such as stoof s @xcite and schakel s @xcite ) neglect the modification altogether ( @xmath73 ) .
the rg flow equations derived by bijlsma & stoof @xcite allow one to numerically locate the transition point .
under rg flow , the physical effective chemical potential flows to positive and negative values , if one starts out respectively in the bose - condensed and the symmetry - unbroken phase .
the critical point is therefore identified by finding parameters for which the chemical potential flows to zero .
stoof finds @xcite that the dependence of @xmath43 on @xmath74 has more structure than purely linear ; @xmath75 . the differences between stoof s and the baym group s results merit further study .
stoof has suggested @xcite that the source of the difference is not the infrared but the _ ultraviolet_. the correction to the spectrum , @xmath76 , has an ultraviolet @xmath77 behavior in the baym group s treatment @xcite . in the rg framework ,
the corresponding quantity is the running coupling constant @xmath78 , which shows no ultraviolet logarithms in the calcultation of ref @xcite .
this leads to a logarithmic difference in the two predictions for @xmath0 .
stoof also suggests @xcite that an improved rg calculation , including the momentum dependence of the effective two - body interaction at the transition point , would shed light on the difference between the two approaches . in treating his effective - action theory near the transition point
, schakel @xcite is also faced with infrared ( ir ) divergences .
one of the remarkable features of this work is an innovative analytic continuation procedure used to regulate ir problems .
baym _ et al _ have remarked @xcite that schakel s prescription ( @xmath79-function regularization ) must be incorrect because it gives an incorrect result for the compressibility of the ideal gas .
this argument , unfortunately , is derived from an equation that schakel derived explicitly for the nontrivial broken - symmetry ground state @xcite of an interacting gas ( @xmath80 ) .
the same equations are not applicable to the noninteracting case ( @xmath81 ) for which the structure of the vacuum is different ( more trivial ) .
the situation is therefore more complicated than indicated in baym _
et al _ s note . in response , schakel has provided a more detailed explanation and justification @xcite for the @xmath79-function continuation procedure .
it seems to me that the regularization procedure itself is quite possibly sound .
my concern with schakel s derivation is different just below the transition point ( broken - symmetry phase ) , schakel uses a quadratic spectrum .
a more accurate treatment would use the phonon - like spectrum @xmath82 .
the effect such a spectrum would have on the thermodynamic potential or on @xmath0 is not clear .
( redoing schakel s @xmath0 calculation with a phonon - like spectrum seems quite difficult . ) in wilkens _
et al _ s paper @xcite , calculations have been performed using a canonical ensemble of @xmath50 particles , and the shift @xmath43 obtained is linear and _ negative _ in the scattering length .
it has been pointed out @xcite that the error lies in evaluating quantities ( derivatives ) at @xmath84 , instead of evaluating them at the expectation value of @xmath83 .
et al _ @xcite have also used finite - size scaling ideas in combination with a first - order calculation in the canonical ensemble , and obtain a positive linear shift .
the idea that the zero - momentum state occupancy @xmath85 becomes `` microscopic '' at and above the transition needs to be clarified .
one finds @xmath86 for @xmath87 , where @xmath50 is the total number of particles , but not necessarily @xmath88 .
in fact , for both the ideal and the interacting bose gas , @xmath89 , which is large compared to @xmath90 though small compared to @xmath50 . in view of the continuing appearance of the @xmath91 misconception ,
i give here a simple derivation of @xmath92 , together with an expression for the coefficient @xmath93 , for the ideal - gas case .
let us consider the ideal - gas relationship @xmath94 , where @xmath95 , and @xmath96 refers to a bose - einstein integral function @xcite . since @xmath97 is small near the transition , one can use the robinson expansion @xcite for @xmath98 . also , inverting @xmath99 one gets @xmath100 . combining these yields , near the transition , @xmath101 since @xmath102 at the transition
, the coefficient of @xmath103 needs to vanish .
therefore @xmath104 the @xmath105 occupancy @xmath85 can actually be calculated for all @xmath106 , using an extra lagrange multiplier @xmath107 conjugate to @xmath85 , in addition to usual chemical potential @xmath108 coupling to @xmath50 .
using this `` two chemical potentials '' formalism @xcite , one can show that for the ideal gas @xmath109 occurs at a temperature @xmath110 , well above the transition point . for the interacting - gas case , an equation corresponding to is difficult to derive , but finite - size scaling arguments @xcite show that @xmath111 continues to hold , even though the nonideal bose gas belongs to a different universality class . the definition of the transition point used in ref @xcite is in terms of the probability @xmath112 for having a condensate occupancy @xmath113 .
the claim is that @xmath114 decreases monotonically with @xmath115 for all temperatures above the transition , and has a peak at nonzero @xmath115 for in the condensed phase , and therefore the transition should be defined as the thermodynamic limit of the zero of @xmath116 .
the correspondence between the canonical particle - counting statistics , and the grand - canonical ideas of average occupancies , is not trivial . one important point to note
is that a monotonically decreasing @xmath117 does not imply a zero expectation value for @xmath85 , even though @xmath117 has a maximum at @xmath118 .
( for example , using @xmath119 gives @xmath120 $ ] . ) the stipulation of a monotonically decreasing @xmath117 therefore does not imply a zero or @xmath121 value of @xmath85 at the transition .
kanno developed an unusual approach @xcite to study many - body systems , and also used this approach to calculate bose gas properties @xcite .
_ the formalism_. + kanno s calculation of thermodynamic quantities involves a `` quasilinear '' canonical transformation @xcite of the free - particle operators @xmath6 ( @xmath7 ) : @xmath122 resulting in a hamiltonian having a particular ( diagonal - like ) form .
the formalism involves examining the properties of the coefficients @xmath123 , and the coefficients appearing in the quasi - diagonal hamiltonian . applying this formalism to the weakly repulsive bose gas @xcite , one can derive the bogoliubov spectrum , entropy and pressure above and below the transition , and many other quantities .
kanno does not provide direct connections to the usual language of perturbative many - body theory .
i would like to point out that many of his physical results , if not all , are equivalent to results obtained in more mainstream approaches .
for example , the quantities @xmath124 and @xmath125 , appearing in the second paper of the series @xcite , are equivalent to the normal and anomalous self - energies that appear in modern descriptions . _
the shift in @xmath0_. + kanno s defines @xcite a modified thermodynamic potential @xmath126 with an unusual lagrange multiplier @xmath30 conjugate to the number of non - condensate particles , @xmath127 , as opposed to the usual chemical potential which is conjugate to the total particle number @xmath50 .
it follows that @xmath128 once @xmath129 is calculated in some approximation , comparing this relation to @xmath130 for the free bose gas at the transition point yields the shift @xmath43 , in terms of the quantity @xmath131 .
the quantity @xmath30 itself is obtained from a consideration of the free energy @xmath132 .
the actual value of @xmath85 chosen by the system is the one that minimizes @xmath133 , i.e. , @xmath134 this equation allows one to express @xmath131 in terms of the interaction strength . _
computing @xmath129 in more usual language_. + as advertised in section [ sect_background ] , i show here how the modified thermodynamic potential @xmath129 ( which is the quantity relevant to the @xmath0 problem ) can be obtained from a diagrammatic second - order perturbative calculation .
the feynman rules for a diagrammatic perturbative calculation of the modified therodynamic potential @xmath135 are very similar to those for the usual @xmath136 .
one has to use a momentum - dependent chemical potential @xmath137 , and separate out @xmath105 from the momentum - sums at each stage .
two first order and two second - order diagrams need to be calculated : @xmath138 other second - order diagrams are excluded by considering the chemical potential to be renormalized upto first order , as discussed in section [ sect_meanfield ] .
the zeroth and first order terms can be evaluated exactly .
uv divergences appearing at second order can be removed by eliminating the bare potential @xmath10 in favor of the 2-body @xmath139-marix , as done for the self - energy in @xcite .
the result is @xmath140 the expression for @xmath129 obtained in ref @xcite is essentially the same as eq .
_ comparison & remarks_. + the parameter @xmath131 corresponds to the infrared cutoff used in the the non - selfconsistent regulariztion of @xcite .
eq gives the same result for the parameter @xmath131 as that obtained in @xcite for the chemical potential @xmath108 using @xmath141 .
the reason for this is that @xmath142 represents the energy of a zero - momentum quasiparticle with respect to the chemical potential , i.e. , @xmath143 .
therefore eq is equivalent to the usual condition @xmath144 for the bose - condensation point .
the shift @xmath43 obtained using eqs , , is the same as that obtained by our non - selfconsistent regularization scheme in @xcite .
the distinctive idea in kanno s work is the introduction of a modified thermodynamic function @xmath129 such that @xmath145 replaces the right side in the criticality condition @xmath146 .
this bypasses the issue of self - consistency and the modification of the spectrum at @xmath0 , and the use of the chemical potential as a cutoff folllows naturally .
to summarise , this article reviews aspects of the calculation of the @xmath0 shift due to a small repulsive interaction in a bose gas .
this question has generated a number of calculations involving diverse techniques ; the present discussion is a first step toward developing an organized overview of the different approaches .
the issue that needs to be resolved urgently , in my opinion , is the possibility of a logarithm - like correction to the leading shift in @xmath0 .
presumably , a better understanding of the spectrum at the critical point ( both at the infrared and the ultraviolet ) should help settle this issue .
two directions need to pursued in this regard .
first , as suggested by stoof , an improved rg calculation , taking into account the momentum dependence of the effective interaction , would improve the ultraviolet behavior of the spectrum in his formalism .
also , it might be worth retaining the wavefunction renormalization ( critical index @xmath65 ) , which changes the infrared spectrum at @xmath0 from @xmath147 to @xmath148 .
this is neglected in @xcite on the grounds that @xmath65 is small .
a second approach suggests itself when we note that the quasiparticle treatments @xcite have all calculated self - energies in terms of the two - body @xmath139-matrix @xmath149 .
this quantity describes all possible collisions between _ two _ particles , without taking into account the fact that the surrounding medium has an effect on these collisions .
many - body corrections arising from the surrounding gas can be treated by using the many - body @xmath106-matrix , which is the sum of ladder diagrams .
the two - body @xmath139-matrix has no temperature - dependence , while the temperature - dependence of the effective potential ( vanishing at @xmath0 ) is captured by the many - body @xmath106-matrix @xcite .
stoof suggests @xcite that this might explain why the quasiparticle analysis ( perturbative treatment of spectrum modification ) over - adjusts the infrared spectrum .
determining the self - energies and quasiparticle spectra in terms of the many - body @xmath106-matrix has been attempted @xcite , using the relations provided by shi & griffin @xcite between the bare potential @xmath10 , the @xmath139-matrix , and the many - body @xmath106-matrix .
our attempt was unsuccessful due to the difficulty of disentangling multiple frequency and momentum inter - dependences at second - order .
* one possibility is to use finite - size scaling ideas , in conjunction with second - order perturbation theory in the grand - canonical ensemble .
this would be a complementary to the calculation of mueller & baym @xcite , which uses finite - size scaling with first - order perturbation theory in the canonical ensemble .
* it would be nice to see schakel s calculation of the thermodynamic potential , just below @xmath0 , done with the spectrum treated more properly , either with a bogoliubov - like spectrum or with the @xmath150 spectrum .
* as mentioned in section [ sect_background ] , a detailed comparison is required between the @xmath24-expansion work @xcite and other approaches to @xmath0 .
informative discussions with h.t.c .
stoof and f. lalo were made possible by the generosity of ect * , trento ( italy ) .
adriaan m.j .
schakel answered many e - mail questions .
my phd advisor , a. e. ruckenstein , introduced me to the problem , and the calculations in @xcite were performed under his guidance .
m. holzmann , g. baym , j .-
blaizot , and f. lalo , phys . rev .
lett . * 87 * , 120403 ( 2001 ) .
e. mueller , g. baym , and m. holzmann , j. phys .
b , * 34 * , 4561 ( 2001 ) ; also cond - mat/0105359 .
p. arnold and b. tomsik , phys .
* a62 * , 063604 ( 2000 ) .
kashurnikov , n.v .
prokofev , and b.v .
svistunov , phys . rev
* 87 * , 120402 ( 2001 ) .
p. arnold and g. moore , phys .
. lett . * 87 * , 120401 ( 2001 ) ; phys . rev .
* e64 * , 066113 ( 2001 ) ; + p. arnold , g. moore , and b. tomsik , phys . rev . * a 65 * , 013606 ( 2002 ) .
stoof , phys . rev . * a 45 * , 8398 ( 1992 ) . m. bijlsma and h.t.c .
stoof , phys . rev . *
a54 * , 5085 ( 1996 ) . h.t.c .
stoof , private communication .
m. bijlsma and h.t.c .
stoof , phys .
* a55 * , 498 ( 1997 ) .
s. kanno , part i in prog .
theor . phys . * 41 * , 966 ( 1969 ) ; and part ii in _ ibid . _ * 41 * , 1145 ( 1969 ) . s. kanno , prog .
phys . * 41 * , 949 ( 1969 ) .
m. wilkens , f. illuminati and m. krmer , j. phys .
b * 33 * , l779 ( 2000 ) . a.m.j .
schakel , comment on los alamos preprint server , cond - mat/0004142 a.m.j .
schakel , int .
b * 8 * , 2021 ( 1994 ) ; " boulevard of broken symmetries , habilitationschrift freie universitt berlin ( 1998 ) .
also available as pre - print : cond - mat/9805152 .
schakel , private communication .
schakel , preprint cond - mat/0301050 .
k. huang , phys .
lett . * 83 * , 3770 ( 1999 ) .
t. toyoda , ann .
, n.y . , * 141 * , 154 ( 1982 ) .
f. de souza cruz , m.b .
pinto , and r.o .
ramos , phys .
b 64 * , 014515 ( 2001 ) ; + f. de souza cruz , m.b .
pinto , r.o .
ramos , and p. sena , phys .
rev . * a 65 * , 053613 ( 2002 ) ; + j .-
kneur , m.b .
pinto , and r.o .
ramos , phys .
lett . * 89 * , 210403 ( 2002 ) ; + e. braaten & e. radescu , preprint cond - mat/0206186 for properties of the bose - einstein integral functions , see for example , k. huang , _ statistical mechanics _ , 2nd edition , wiley 1988 , and r.k .
pathria , _ statistical mechanics _ , pergamon 1972 . some properties ( expansions ) were first derived in j.e .
robinson , phys . rev .
* 83 * , 678 ( 1951 ) . | a number of calculations have appeared for the bose - einstein condensation temperature ( @xmath0 ) of a weakly repulsive dilute bose gas . after a short survey of previous work ,
i point out several issues related to these approaches , and outline future calculations . |
multi - wavelength light curves of blazar flares show complex and diversified features . while in some cases there is a time lag between gamma - ray and x - ray / optical flares ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , in other cases an orphan flare in a certain wavelength was detected ( e.g. * ? ? ?
* ; * ? ? ?
even if a time - dependent model is adopted , such a variety of behaviors may be difficult to reproduce with by a one - zone model @xcite . while spatial gradients of the physical parameters in the emission regions @xcite may explain some fraction of the lags , some flares have spectral evolutions thatare too complex to be modeled , even with time - dependent multi - zone radiative transfer simulations @xcite .
this may imply that inhomogeneous emission regions evolve with a longer timescale than a dynamical one . such nontrivial properties in a blazar flare
make it difficult to probe physical processes such as electron acceleration or cooling . in 2013 december , the _ fermi_-large area telescope ( lat ) detected one of the most intense flares in the gamma - ray band from flat spectrum radio quasar ( fsrq ) 3c 279 , reaching @xmath0 for the integrated flux above 100mev ( * ? ? ?
* hereafter h15 ) .
the flux level is comparable to the historical maximum of this source observed at the gamma - ray band @xcite .
the gamma - ray flare showed a very rapid variability with an asymmetric time profile with a shorter rising time of @xmath1hr and a longer falling time of @xmath2hr .
we can expect that this extraordinary flare was emitted from a sufficiently compact region that can be regarded as homogeneous , which is different from other usual flares . in this case
, the decaying timescale may directly correspond to the cooling timescale , and the flare is an ideal target for discussing the physical processes .
another important property of the flare event of 3c 279 , a very hard photon index of @xmath3 , was observed in the @xmath4mev band by _
such a hard photon index has been rarely observed in fsrqs , whose luminosity peak from inverse - compton ( ic ) scattering is usually located below 100mev . while the mean of the @xmath5 distribution in fsrqs corresponds to about 2.4 @xcite , hard photon indices @xmath6
only have been occasionally observed in some bright fsrqs during rapid flaring events @xcite . in order to reproduce the hard photon index by ic scattering in the fast cooling regime
, the index of parent electrons should be much harder than two , which can hardly be generated in a normal shock acceleration process .
in addition , the flare event of 3c 279 indicates a high compton dominance parameter @xmath7 , leading to extremely low jet magnetization with @xmath8 ( h15 ) . to explain the flare event of 3c 279 , rather than assuming prompt electron injection by the shock acceleration
, we propose the stochastic acceleration ( sa ) model , which is phenomenologically equivalent to the second - order fermi acceleration ( fermi - ii ; e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* and references therein ) .
the sa may be driven by magnetic reconnection @xcite .
otherwise , hydrodynamical turbulences that drive the acceleration are possibly induced via the kelvin
helmholtz instability as an axial mode ( e.g. * ? ? ?
* ) , or the rayleigh taylor and richtmyer meshkov instabilities as radial modes @xcite .
broadband spectra of blazars in the steady state have been successfully fitted with recent sa models @xcite .
the flare state should be also tested with such models to show the wide - range applicability of the sa .
this is the first attempt to apply a fermi - ii model to explain both broadband spectra and light curves of fsrqs simultaneously . in this letter
, we perform time - dependent simulations of the emissions from 3c 279 with the sa . starting from modeling a steady emission ,
the gamma - ray flare is reproduced by decreasing the magnetic field for the steady model .
we demonstrate that the sa model agrees with the observed spectrum and light curve .
we adopt the numerical simulation code used in @xcite . in this model , a conical outflow with an opening angle @xmath9 , where @xmath10 is its bulk lorentz factor , is ejected at radius @xmath11 from the central engine .
the evolutions of electron and photon energy distributions are calculated in the comoving frame taking into account the sa , synchrotron emission , ic scattering with the klein
nishina effect , adiabatic cooling , @xmath12 pair production , synchrotron self - absorption , and photon escape .
the sa is characterized by the energy diffusion coefficient , @xmath13 .
@xcite conservatively assumed the kolmogorov - like diffusion as @xmath14 . here , we adopt , however , @xmath15 , which corresponds to the hard - sphere scattering .
this choice leads to reasonable spectra of 3c 279 without complicated assumptions such as nontrivial temporal evolution of the diffusion coefficient or electron injection rate .
if the cascade of the turbulence stops at a certain length scale larger than the gyro - radius of the highest - energy electrons , the mean free path of electrons becomes comparable to this scale independently of electron energies .
in this case , the energy diffusion can be approximated as the hard - sphere scattering ( e.g. * ? ? ? * ) .
the recent magnetohydrodynamical simulations accompanying the inverse cascade shows the @xmath16 spectrum in turbulences @xcite , which also support the hard - sphere approximation .
the volume we consider is a conical shell with a constant width of @xmath17 , then the isotropically equivalent volume @xmath18 ( the actual volume is @xmath19 ) .
hereafter , the values in the shell frame are denoted with prime characters . during the dynamical timescale @xmath20 in the plasma frame ,
electrons are injected at a constant rate @xmath21 in the volume @xmath22 monoenergetically ( @xmath23 ) and accelerated with a constant coefficient @xmath24 . as done in @xcite
, we can consider the temporal evolutions of the injection rate and the diffusion coefficient .
however , this simple model with constant @xmath21 and @xmath24 is sufficient to reproduce the photon spectrum of 3c 279 . the average magnetic field in the comoving frame
is assumed to behave as @xmath25 .
the evolution of the photon spectrum for observers is computed taking into account the relativistic motion and curvature of the jet surface .
as a reference case , we consider one of the most active periods in the gamma - ray band during the first two years of the _ fermi_-lat observations . in @xcite ,
this period is denoted as period `` d '' in 2009 .
although this period corresponds to an event with a prominent flare , the broadband spectrum in the paper is averaged over five days .
if the emission zone is inside the broad emission region as suggested by the short variability timescale reported in h15 , the flare state is significantly longer than the variability timescale .
therefore , we adopt a steady emission model for period d in 2009 . by assuming continuous steady ejection of the shells from @xmath11 , we model the steady photon spectrum , though the plasma and its emission evolves with @xmath26 in the shell frame .
the model parameters are @xmath27 pc , @xmath28 , @xmath29 ( @xmath30 ) , @xmath31 ( @xmath32 ) , and @xmath33 g. we adopt the same model as that of @xcite for the external radiation of the broad emission lines with the photon temperature @xmath34 ev and the energy density @xmath35 in the shell frame . the cooling timescale in this external radiation field
is written as @xmath36 as shown in figure [ fig : ele1 ] ( a ) , electrons are continuously accelerated between @xmath11 and @xmath37 , then they are rapidly cooled via ic scattering after the shutdown of the acceleration .
the electron spectrum at @xmath37 in the low - energy region is consistent with the assumed power - law index of @xmath38 in the broken power - law model of @xcite . in the highest - energy region , though the klein
nishina effect suppresses the ic cooling effect , the synchrotron cooling ( @xmath39 ) prevent the acceleration above @xmath40 mev .
the resultant photon spectrum well reproduces the observed spectrum from far - infrared to gamma - ray bands ( see figure [ fig : ele1 ] ( b ) ) .
the low - energy cutoff at @xmath41 ev is due to the synchrotron self - absorption .
the x - ray flux is originated from the synchrotron self - compton ( ssc ) emission .
those x - ray data strongly constrain the emission radius @xmath42 . thus , our sa model can naturally produce a hard electron spectrum , and the steady photon spectrum agrees with the observed one in 2009 .
based on this result , we will probe the intensive flare in 2013 in the next section .
the most intensive flare denoted as period `` b '' ( on mjd 56646 ) in h15 shows a very hard spectrum with @xmath43 and a short variability with an hourly scale in the gamma - ray band observed with _
during a short time interval of the gamma - ray flare period ( 0.2 days ) , there were simultaneous optical observations , whose results did not show any correlated variability with the gamma - ray flare as presented in h15 .
the x - ray observations in the period are available from _ swift_-bat transient monitor results by the _
swift_-bat team @xcite .
the data provided an upper limit in the 15 - 50 kev band .
we adopt the same values for @xmath42 , @xmath10 , @xmath44 , and @xmath45 as those in the previous section . by changing @xmath24 , @xmath21 , and @xmath46
, we attempt to fit the spectrum of the flaring period b in 2013 .
we consider one shell that contributes to the flare emission . the energy diffusion coefficient and injection rate
are slightly increased from the values in the steady model to @xmath47 ( @xmath48 ) , and @xmath49 ( @xmath50 ) , respectively .
hereafter , we call this the `` fiducial '' flare model .
no significant concurrent flare in the optical bands implies that the optical photons are emitted from other steady components . as discussed in section 4.3 in h15
, the lack of overall correlation between the optical and gamma - ray bands in 2013 - 2014 also suggests the different origin of the optical component .
the synchrotron flux of the flare should be below the observed flux level so that an upper limit for the magnetic field in the flare zone will be given . here
, we adopt a very low value of @xmath51 g. figure [ fig : ele2 ] ( a ) shows the evolution of the electron energy distribution in this flare model .
electrons are accelerated to higher energies compared to the case in the steady model .
the power - law distributions above @xmath52 ev are due to not only the larger @xmath24 but also the inefficiency of the synchrotron cooling owing to the low magnetic field .
the secondary bumps at @xmath53 ev for @xmath54-@xmath55 are attributed to the generation of secondary electron
positron pairs via internal @xmath12 absorption .
the resultant gamma - ray spectra shown in figure [ fig : ele2 ] ( b ) agree well with the observed gamma - ray data . here , we add an underlying component ( that overlaps the solid gray line in the figure ) to explain the gamma - ray flux before the flare and the optical data .
the flare spectrum has a higher synchrotron peak energy than the model in h15 , since our flare model shows a drastic growth of the maximum energy of electrons compared to the steady model .
as remarked above , the weak magnetic field strikingly increases the maximum energy of electrons .
even for this low magnetic field , the optical flux is slightly enhanced during the flare .
the steady behavior of the optical light curve may prefer a weaker magnetic field , but we regard this as a conservative upper limit . the sharp cutoff at @xmath56 ev in the photon spectrum is due to the @xmath12 absorption inside the emission region .
some fraction of photons above @xmath57 ev escape from the shell .
the model flux at the @xmath40 gev band is still higher than the detection limit for erenkov telescopes .
however , it should be noted that we have neglected the @xmath12 absorption after the escape from the shell .
the absorption by the broad emission lines during propagation may greatly suppress the flux around @xmath40 gev .
even for the same values of @xmath42 and @xmath10 as in the steady model , the light curve is well reproduced as shown in figure [ fig : lc ] .
thus , the emission zones of the intense flare in 2013 and the active period in 2009 may be located at similar distances from the central engine .
the observed asymmetric profile in the light curve is favorable for our simple one - shell emission - zone model .
the strong cooling due to the external ic yields the rapid decay of the light curve .
the evolutions of energy densities in figure [ fig : ened ] clearly show the energy input by the sa and rapid cooling just after the end of the acceleration . at @xmath58 ,
the energy density ratio of the magnetic field to electrons is quite low as @xmath59 .
the observational constraints , of course , do not determine the model parameter uniquely .
however , the essential parameter for determining the gamma - ray spectral shape is only @xmath24 in our model ( the role of @xmath21 is just normalizing the flux level , and the value of @xmath46 does not affect the gamma - ray spectral shape ) . in figure
[ fig : comp ] , we compare several photon spectral models derived with different parameter sets .
when we reduce @xmath24 by a factor of two from the fiducial model ( `` low-@xmath60 '' model : @xmath61 , @xmath62 , the others are the same ) , the peak photon energy does not reach 10 gev .
conversely , we increase the diffusion coefficient as shown in the `` high-@xmath60 '' model ( @xmath63 , @xmath64 , @xmath65 , the others are the same ) . in this case
, we need an even weaker magnetic field .
the peak time of the light curve is delayed due to the lower @xmath21 compared to the fiducial model . in figure
[ fig : lc ] , we shift the light curve by 1.5 hr earlier . other physical parameters ( @xmath10 , etc . ) of the jet were derived from the steady model . however , as h15 supposed , we also try to increase @xmath10 in our model .
the initial radius should be increased as @xmath66 to keep the variability timescale .
such an example ( `` high-@xmath10 '' model : @xmath67 , @xmath68 , @xmath69 , @xmath70 , @xmath24 is the same ) is shown in figure [ fig : comp ] .
due to the relatively short @xmath71 ( @xmath72 ) , the maximum electron energy grows as high as @xmath73 ev .
a very low magnetic field ( @xmath74 ) is necessary again to suppress the synchrotron flux in the x - ray band .
the strong cooling due to the higher @xmath75 makes the gev spectrum too soft compared to the observed gamma - ray spectrum .
the simple sa model can reasonably explain the very hard spectrum and short variability in the intensive flare in 2013 .
the turbulence driving the particle acceleration may be generated by the hydrodynamical instability or the magnetic reconnection .
compared to the steady model for the active period in 2009 , the drastic alteration we need is the decrease of the magnetic field .
the other parameters have almost similar values .
the absence of the optical flare implies the weak magnetic field ( @xmath76 g ) .
the requirement of the magnetic field decrease at gamma - ray flare stages was suggested by @xcite as well .
the required low magnetic field seems irrelevant to the energy source for the particle acceleration .
therefore , a hydrodynamical instability is responsible for driving the sa .
when @xmath28 , the variability timescale is consistent with @xmath77 pc as shown in figure [ fig : lc ] .
this distance from the engine also agrees with the constraint by the x - ray ssc component in the active period in 2009 .
the size of the central engine may be @xmath78 pc for the black hole mass of @xmath79 .
if we adopt the simplest model for the jet acceleration due to the magnetic energy dissipation @xcite , the bulk lorentz factor at @xmath11 should be @xmath80 , which is inconsistent with the postulated value of @xmath10 .
given the variability timescale @xmath81 , the initial radius can be scaled as @xmath82 when we change @xmath10 . however , even in this case , the maximum lorentz factor at @xmath42 inferred from the magnetic dissipation model increases by a factor of only @xmath83 . for the jet acceleration model by the poynting flux dissipation , not only the low magnetic field but also the short variability timescale are problematic .
this problem is also raised for the very short gamma - ray flare ( a few hundreds seconds ) of bl lac objects like pks 2155304 @xcite . for such ssc - dominant objects , however , the distance from the engine is not well constrained compared to fsrqs .
the tiny change of the diffusion coefficient @xmath24 , in spite of the drastic decrease of the magnetic field , seems enigmatic .
the assumed value of @xmath15 may be favorable for this invariant behavior of @xmath24 . in this low magnetic field case
, the average energy gain per scattering may be proportional to @xmath84 , where @xmath85 is the average turbulence velocity , rather than the alfvn velocity .
the pitch angle diffusion approximation @xcite and power - law magnetic turbulence of @xmath86 , where @xmath87 is the wavenumber , leads to @xmath88 .
the last factor of @xmath89 implies that electrons interact with higher ( lower ) amplitude turbulence at longer ( shorter ) wavelengths for a weaker ( stronger ) magnetic field . if @xmath90 , @xmath15 results in @xmath91 , which is independent of @xmath92 .
alternatively , magnetic bottles as `` hard spheres '' @xcite may be formed in turbulence independently of the strength of the magnetic field .
such requirements for the turbulence property motivate us to probe the hydrodynamical instabilities in blazar jets .
the authors thank the anonymous referee for the useful comments .
we also thank f. takahara , m. kusunose , k. toma , j. kakuwa , k. nalewajko and g. m. madejski for useful discussion .
this study is partially supported by grants - in - aid for scientific research no.80399279 from the ministry of education , culture , sports , science and technology ( mext ) of japan .
abdo , a. a. , ackermann , m. , ajello , m. , et al .
2010 , , 710 , 810 abdo , a. a. , ackermann , m. , ajello , m. , et al .
2010 , , 463,919 ackermann , m. , ajello , m. , atwood , w. , et al .
2015 , arxiv:1501.06054 aharonian , f. , akhperjanian , a. g. , bazer - bachi , a. r. , et al .
2007 , , 664 , l71 asano , k. , takahara , f. , kusunose , m. , toma , k. , & kakuwa , j. 2014 , , 780 , 64 beresnyak , a. , yan , h. , & lazarian , a. 2011 , , 728 , 60 blandford , r. , & eichler , d. 1987 , phr , 154 , 1 baejowski , m. , blaylock , g. , bond , i. h. , et al .
2005 , , 630 , 130 brandenburg , a. , kahniashvili , t. , & tevzadze , a. g. 2015 , , 114 , 075001 chen , x. , fossati , g. , liang , e. p. , & bttcher , m. 2011 , , 416 , 2368 diltz , c. , & bttcher , m. 2014 , j. high ene .
astrop . , 1 , 63 drenkhahn , g. 2002 , , 387 , 714 fossati , g. , buckley , j. h. , bond , i. h. , et al .
2008 , , 677 , 906 hayashida , m. , madejski , g. m. , nalewajko , k. , et al .
2012 , , 754 , 114 hayashida , m. , nalewajko , k. , madejski , g. m. , et al .
2015 , , 807 , 79 janiak , m. , sikora , m. , nalewajko , k. , et al .
2012 , , 760 , 129 kakuwa , j. , toma , k. , asano , k. , kusunose , m. , & takahara , f. 2015 , , 449 , 551 katarzyski , k. , ghisellini , g. , mastichiadis , a. , tavecchio , f. , & maraschi , l. 2006 , , 453 , 47 krawczynski , h. , coppi , p. s. , & aharonian , f. 2002 , , 336 , 721 krawczynski , h. , hughes , s. b. , horan , d. , et al .
2004 , , 601 , 151 krimm , h. a. , holland , s. t. , corbet , r. h. d. , et al .
2013 , , 209 , 14 kusunose , m. , takahara , f. , & li , h. 2000 , , 536 , 299 lazarian , a. , vlahos , l. , kowal , g. , yan , h. , beresnyak , a. , de gouveia dal pino , e. m. 2012 , space sci .
, 173 , 557 lefa , e. , rieger , f. m. , & aharonian , f. 2011 , , 740 , 64 matsumoto , j. , & masada , y. 2013 , , 772 , l1 mizuno , y. , hardee , p. e. , & nishikawa , k. 2007 , , 662 , 835 pacciani , l. , tavecchio , f. , donnarumma , i. , et al . 2014 , , 790 , 45 stawarz , .
, & petrosian , v. 2008 , , 681 , 1725 wehrle , a. e. , et al .
1998 , , 497 , 178 zrake , j. 2014 , , 794 , 26 | the very short and bright flare of 3c 279 detected with _ fermi_-lat in 2013 december is tested by a model with stochastic electron acceleration by turbulences .
our time - dependent simulation shows that the very hard spectrum and asymmetric light curve are successfully reproduced by changing only the magnetic field from the value in the steady period .
the maximum energy of electrons drastically grows with the decrease of the magnetic field , which yields a hard photon spectrum as observed .
rapid cooling due to the inverse - compton scattering with the external photons reproduces the decaying feature of the light curve .
the inferred energy density of the magnetic field is much less than the electron and photon energy densities .
the low magnetic field and short variability timescale are unfavorable for the jet acceleration model from the gradual poynting flux dissipation . |
the first theoretical constructions related with directed transport have been formulated in an early work by feynman @xcite .
this opened a field whose relevance and activity has been increasing since then .
the motivation is twofold , in the first place several fundamental questions have originated from these ideas and have been only partially answered @xcite . as a consequence ,
the subject grew into a major new field of statistical physics . on the other hand ,
the wealth of possible applications has determined the emergence of a very broad area of research in physics .
ratchets , generically defined as periodic systems having a dissipative dynamics associated with thermal noise and unbiased perturbations ( driving them out of equilibrium ) , can be used to model a wide range of different phenomena . in order to give just a few examples we can mention molecular motors in biology @xcite , nanodevices ( rectifiers , pumps , particle separators , molecular switches and transistors ) @xcite , and coupled josephson junctions @xcite . on the other hand , there is a great interest in the theoretical description and experimental implementations of cold atoms subjected to time - dependent standing waves of light .
they play a central role in many studies of the quantum dynamics of nonlinear systems ( dynamical localization , decoherence , quantum resonance , etc . )
in particular , the so - called optical ratchets , i.e. directed transport of laser cooled atoms , have been successfully implemented in this sort of experiments @xcite .
the appearance of a net current can be classically explained by the necessary condition of breaking all spatiotemporal symmetries leading to momentum inversion @xcite .
this , along with curie s principle indicates that transport should be present , given that it is not forbidden . as an example of this situation we can mention hamiltonian systems ( with necessarily mixed phase spaces ) where a chaotic layer should be asymmetric @xcite . in the more general dissipative case , chaotic attractors having this property are necessary @xcite .
it is usual that the same principle translates almost directly into the quantum domain @xcite , and very similar behaviors arise .
but sometimes quantum mechanics introduces new effects @xcite and the results depart from the classical ones .
for example , the efficiency of a forced thermal quantum ratchet has been calculated in @xcite . in that work
the authors find that the quantum current is higher in comparison with the classical one at the lowest values of the temperature .
as this parameter increases the discrepancies diminish and finally they become negligibly small .
recently , a quantum chaotic dissipative ratchet has been introduced in @xcite . in this example
directed transport appears for particles under the influence of a pulsed asymmetric potential in the presence of a dissipative environment at zero temperature .
the asymmetry of the quantum strange attractor is at the origin of the quantum current , in close analogy with what happens at the classical level .
indeed , this work provides with the parameters needed for a possible implementation using cold atoms in an optical lattice . for a somewhat similar dynamics , the case of weak
coupling and low temperature has been studied in @xcite . in the present work
we extend the study of @xcite for a wide range of couplings with the environment and different temperature values .
we have verified that there is a strong dependence of the current behavior on the coupling strength .
if we compare this with the results found in @xcite , for instance , we could not find a generic enhancement of the quantum current for finite temperatures . instead of that
, we could identify a close quantum - to - classical correspondence when considering thermal effects only at the classical level .
in fact , we have found that the finite @xmath0 effects on the quantum current are analogous to the influence of the classical thermal fluctuations on the classical transport . in the following
we describe the organization of this paper . in section
ii we present our model for the system and for the environment , explaining the methods we have used to investigate the current behavior . in section iii
we show the results where the roles of @xmath0 , the coupling strength and the temperature are analyzed in detail .
finally , in section iv we summarize and point out our conclusions .
in this section we describe the approach used to model the system plus the environment .
we study the motion of a particle in a periodic kicked asymmetric potential given by @xmath1 \sum_{m=-\infty}^{+ \infty } \delta ( t - m \tau)\ ] ] where @xmath2 is the kicking period , @xmath3 is the strength of the kick , and @xmath4 and @xmath5 are parameters that allow to introduce a spatial asymmetry @xcite .
the effects of the environment are taken into account by means of a velocity dependent damping and thermal fluctuations . at the classical level ,
these ingredients are incorporated in the following map : @xmath6 + \xi \\ \overline{x}= x + \tau \overline{n } \end{array } \right.\ ] ] in these expressions , @xmath7 is the momentum variable conjugated to @xmath8 and @xmath9 is the dissipation parameter , with @xmath10 .
the thermal noise @xmath11 is related to @xmath9 , according to @xmath12 , where @xmath13 is the boltzmann constant and @xmath14 is the temperature , making the formulation consistent with the fluctuation - dissipation relationship . by performing the change of variables
@xmath15 , @xmath16 , and @xmath17 ( where @xmath18 and @xmath19 ) , we can eliminate the period from the classical expressions , and define the new map @xmath20 + \tilde \xi \\ \overline{x}= x + \overline{p } \end{array } \right . \label{cl_map}\ ] ] in the quantum version of the model the system hamiltonian is given by @xmath21 where the quantization has been performed in such a way @xcite that @xmath22 , @xmath23 and @xmath24 .
this amounts to saying that , being @xmath25=i\tau$ ] , there is an effective planck constant given by @xmath26 .
then , the classical limit corresponds to @xmath27 , while keeping @xmath28 constant . in order to incorporate dissipation and thermalization to the quantum map we consider the coupling of the system to a bath of non interacting oscillators in thermal equilibrium .
the degrees of freedom of the bath are eliminated introducing the usual weak coupling , markov and rotating wave approximations @xcite .
this leads to a lindblad equation in action representation @xmath29 \ , + \ , \\ \!\!\ ! & + & \ ! g \ , \sqrt{n^+_{th } ( \omega_n , t ) n^+_{th}(\omega_{n'},t ) } \ , \ { [ \hat{l}_n , \hat{\rho } \hat{l}_{n'}^{\dagger } ] + [ \hat{l}_n \hat{\rho},\hat{l}_{n'}^{\dagger}]\ } \\ \!\!\ ! & + & \ ! g \ , \sqrt{n^-_{th } ( \omega_n , t ) n^-_{th}(\omega_{n'},t ) } \ { [ \hat{l}_n^{\dagger } , \hat{\rho } \hat{l}_{n ' } ] + [ \hat{l}_n^{\dagger } \hat{\rho},\hat{l}_{n'}]\ } \end{array } \label{qdiff}\ ] ] where the frequencies @xmath30 are the energy differences between two neighboring levels of the rotator . the population densities of the bath found in eq .
[ qdiff ] are given by @xmath31 the system operators @xmath32 describe transitions towards the ground state of the free rotator . requiring quantum to classical correspondence at short times
we fix the coupling constant @xmath33 . for @xmath34
we recover the master equation used in @xcite for the pure dissipative case .
for finite temperature , the last term in eq .
[ qdiff ] describes the thermal excitation of the rotator through absorption of heat bath energy .
finally , eq . [ qdiff ] will be integrated numerically without further approximations .
since we are interested in chaotic transport , throughout the following calculations we will use the set of parameters given by @xmath35 , @xmath36 , and @xmath37 . in the hamiltonian limit , this case shows no visible stability islands in phase space .
we have first studied some classical aspects of our system at zero temperature , beginning with the bifurcation diagrams in terms of @xmath38 as a function of the parameter @xmath39 $ ] ( see fig .
[ fig : bifdiag](_a _ ) ) .
it should be mentioned that the chaotic attractors set in very fast for small @xmath9 . but this is not the case for larger values where the transient times can be very long .
for this reason , we have calculated these diagrams with the last @xmath40 iterations of the map , after the first @xmath41 have been discarded .
we have randomly taken @xmath42 initial conditions inside the unit cell ( @xmath43 , @xmath44 $ ] ) .
several regular and chaotic windows alternate .
the former are characterized by simple attractors ( stable fixed points of the dissipative map ) and the latter are dominated by chaotic attracting sets .
the width in @xmath38 of these sets grows as dissipation weakens ( @xmath45 ) .
the main quantity characterizing transport is the current @xmath46 , where @xmath47 stands for the average taken on the initial conditions , and @xmath48 is the moment after the @xmath49th iteration of the map . since bifurcations that change the shape of strange attractors also play a role in determining the values of the asymptotic current we restrict our analysis to the chaotic window approximately located at @xmath50 $ ] indicated by an arrow .
the inset of fig .
[ fig : bifdiag](a ) displays the asymptotic current @xmath51 as a function of @xmath9 ( circles stand for @xmath34 ) . at this @xmath9 range ,
100 kicks are enough to reach the stationary behavior , independently of the initial distributions .
as pointed out in @xcite , dissipation induces an asymmetry of the strange attractor which is responsible for the directed transport . on the other hand this dissipation mechanism contracts phase space and makes the higher energies inaccessible for the system .
the final value of @xmath52 results from the interplay between both effects .
in fact , increasing dissipation does not increase the transport and the largest values of @xmath52 are obtained for the lowest values of dissipation , i.e. @xmath53 ( for example @xmath54 for @xmath55 ) .
the minimum current in this window is reached for an intermediate value of dissipation ( @xmath56 ) .
we then consider the case of finite temperature .
the bifurcation diagram corresponding to @xmath57 is shown in fig .
[ fig : bifdiag](b ) .
it is clear that the effect of temperature consists of smoothing the finer structure of the chaotic attractors that is present for the smallest values of @xmath9 .
even for this extremely low value of @xmath14 the detailed features have almost completely disappeared , with the exception of the black lines corresponding to the highest values of the density distributions .
the other very interesting effect is that temperature erases the regular windows allowing for a continuously chaotic behavior .
this could be of much relevance in obtaining large ratchet currents without the need for an extremely fine tuned , weak dissipation @xcite ( this will addressed in future studies @xcite ) .
as shown in the inset , low temperatures ( diamonds correspond to @xmath57 ) lead to a noticeable enhancement of the asymptotic current @xmath58 ( around @xmath59 for @xmath56 ) .
we now turn to compare the classical and quantum behaviors .
firstly , we analyze the currents ( which in the quantum case is given by @xmath60 ) . in fig .
[ fig : allasympcurrentsvst ] we display @xmath58 as a function of @xmath14 , for three different values of @xmath9 and @xmath61 . at the classical level
low temperatures lead to an enhancement of the current for intermediate values of the dissipation ( see fig .
[ fig : allasympcurrentsvst ] upper and middle panels corresponding to @xmath9= 0.7 , 0.75 ) . in the case of weak dissipation ( @xmath62 in the lower panel of fig . [ fig : allasympcurrentsvst ] ) , which displays larger values of @xmath58 , the effect of thermal noise is negligible . for higher temperatures , the thermal effects reduce @xmath52 in all cases .
this can be interpreted as follows : thermal noise reduces the energy loss caused by dissipation ( with no kicks , the system would attain a boltzmann distribution ) , so higher energies can be reached , in comparison with a pure dissipative process .
but since this diffusion also tends to blur the asymmetry of the strange attractor , the two effects compete and transport has a maximum for low values of t , and then decreases . at the quantum level we can clearly see that the previously described thermal enhancement is not generally present , at least for the @xmath61 values we have considered .
it is important to note that these values are consistent with experimental implementations using for example , cold sodium atoms in a laser field having a wave length @xmath63 ( for more details see @xcite ) .
however , we observe a very slight growth of the current for the @xmath64 case with @xmath62 ( see the blue triangles in the lower panel of fig . [
fig : allasympcurrentsvst])indicating that the temperature dependence of the current is very sensitive to the particular dynamics of the system . for a different example ,
an enhancement of the quantum transport has been observed @xcite , hence it is difficult to extract universal behaviors . for @xmath65
the quantum current is lower than the classical one for all temperatures , as already pointed out in @xcite , but only for @xmath34 .
the same happens in the case of weak damping @xmath62 ( nevertheless , for @xmath64 and @xmath66 both currents coincide ) .
the case @xmath67 shows a different behavior . at @xmath34
the quantum currents ( for any of the @xmath61 values we have considered ) are larger than the classical ones , that is , there is an enhancement due to the finite size of @xmath61 .
also , the maximal quantum current corresponds to @xmath68 , and this is valid for all the temperatures shown ( for @xmath69 it is still greater than the classical one ) . for larger quantum coarse - graining ( see @xmath70 ) quantum currents decrease . in this sense
, it seems that the effect of quantum fluctuations on the quantum directed currents is analogous to the effect of those of thermal origin on the classical ones .
small fluctuations of thermal or quantum mechanical origin assist directed transport while large fluctuations ( corresponding to high temperatures or to large values of @xmath61 respectively ) blur the asymmetry of the attractor and thus kill the net current .
it is interesting to note that in the case @xmath67 the thermal diffusion associated with the temperature which gives the maximal value of the current ( @xmath71 for @xmath72 ) is of the order of the quantum coarse - graining @xmath61 corresponding to the strongest quantum current .
for @xmath65 the classical current attains its maximum value at @xmath69 ( @xmath73 ) , which corresponds to a value of @xmath61 that we were not able to consider in our numerical calculations .
the analogy between thermal noise and quantum coarse - graining can also be appreciated when looking at the asymptotic poincare sections and husimi distributions ( displayed in fig .
[ fig : phasespace ] for @xmath74 , @xmath75 ) .
as expected , at zero temperature the quantum husimi function reproduces well the main patterns of the classical attractor but shows less fine structure ( see the upper panels ) .
if a small temperature is introduced the fine details of the classical distribution are smoothed out and both distributions look more alike ( see the lower panels corresponding to @xmath76 ) . on the other hand the quantum distributions at zero and finite temperatures
are practically indistinguishable , indicating that the quantum coarse - graining is at least of the order of the thermal one for these values of @xmath14 .
we finally study @xmath77 as a function of @xmath49 ( i.e. , the number of iterations of the map ) .
results for @xmath67 are shown in fig .
[ fig : currentvstg075 ] , where different temperatures and @xmath61 values have been considered .
besides the mentioned fact that the asymptotic value is reached very rapidly , we notice that the transient behavior shows a very close quantum - to - classical correspondence .
the classical current peak observed at @xmath78 for low temperatures ( @xmath76 ) is also present in the quantum current when @xmath64 and @xmath34 .
this peak disappears from the classical current at larger temperatures ( @xmath79 ) , and so does the quantum one at @xmath70 and @xmath34 .
so the analogy seems to hold at all times .
in this work we have analyzed the behavior of a chaotic dissipative system that shows directed transport under the influence of a thermal bath , both in its classical and quantum versions .
we have varied the strength of the coupling with the environment and also the temperature .
we have found that the transport enhancement effect due to a finite temperature is highly dependent on the system specific properties .
in fact , it depends on the coupling strength of the system with the environment and also on the @xmath0 size .
moreover , we could find an analogy between the effects caused by thermal and quantum fluctuations .
these results open the possibility for many further studies that include finding ways of obtaining large ratchet currents in experimentally realistic situations in kicked becs and cold atoms experiments
. these are one of the best candidates to test our results since even becs show an unavoidable fraction of noncondensed atoms when kicked .
if kicks become strong , thermal excitations will be of much relevance rather than a negligible effect . with the sort of calculations presented in this paper the effects of this fraction on the transport properties of the system
could be estimated .
f. l. moore _
et al . _ ,
lett . * 75 * , 4598 ( 1995 ) ; j. ringot _
et al . _ ,
85 * , 2741 ( 2000 ) ; h. ammann _
lett . * 80 * , 4111 ( 1998 ) ; m. b. darcy _ et al . _ ,
lett . * 87 * , 074102 ( 2001 ) . | we study a chaotic ratchet system under the influence of a thermal environment . by direct integration of the lindblad equation
we are able to analyze its behavior for a wide range of couplings with the environment , and for different finite temperatures .
we observe that the enhancement of the classical and quantum currents due to temperature depend strongly on the specific properties of the system .
this makes difficult to extract universal behaviors .
we have also found that there is an analogy between the effects of the classical thermal noise and those of the finite @xmath0 size .
these results open many possibilities for their testing and implementation in kicked becs and cold atoms experiments . |
many phenomenological studies on susy particle searches have been performed in the minimal supersymmetric standard model ( mssm ) with real susy parameters . in general , however , some of the susy parameters may be complex , in particular the higgsino mass parameter @xmath9 , the gaugino mass parameters @xmath10 and the trilinear scalar coupling parameters @xmath11 of the sfermions @xmath12 .
the su(2 ) gaugino mass parameter @xmath13 can be chosen real after an appropriate redefinition of the fields .
not only the cp - violating observables ( such as fermion edms ) but also the cp - conserving observables ( such as cross sections and decay branching ratios ) depend on the phases of the complex parameters , because in general the mass - eigenvalues and the couplings of the susy particles involved are functions of the underlying complex parameters .
for example , the decay branching ratios of the staus @xmath14 and @xmath15-sneutrino @xmath16 can be quite sensitive to the complex phases of the stau and gaugino - higgsino sectors @xcite .
therefore , in a complete phenomenological analysis of production and decays of the susy particles one has to take into account that @xmath11 , @xmath9 and @xmath17 can be complex . in this article based on @xcite
we study the effects of the phases of @xmath18 , @xmath19 , @xmath9 and @xmath3 on the decay branching ratios of the stops @xmath20 and sbottoms @xmath21 with @xmath22 ( @xmath23 ) being the lighter ( heavier ) squark .
we take into account the explicit cp violation in the higgs sector .
in the mssm the squark sector is specified by the mass matrix in the basis @xmath24 with @xmath25 or @xmath26 @xmath27 with @xmath28 a_q m_q & = & \left\ { \begin{array}{l } ( a_t - \mu^*\cot\beta ) \ ; m_t~~(\sq=\st)\\ ( a_b - \mu^*\tan\beta ) \ ;
m_b~~(\sq=\sb ) \end{array } \right .
\label{eq : d } \\ & = & \,\ , |a_q m_q| \ , e^{i\varphi_{\sq } } \,\ , ( -\pi < \varphi_{\sq } \leq \pi ) .
\label{eq : e } \ ] ] here @xmath29 is the third component of the weak isospin and @xmath30 the electric charge of the quark @xmath31 .
@xmath32 and @xmath33 are soft susy breaking parameters and @xmath34 with @xmath35 @xmath36 being the vacuum expectation value of the higgs field @xmath37 @xmath38 .
we take @xmath39 ( @xmath40 ) and @xmath9 as complex parameters : @xmath41 and @xmath42 with @xmath43 . diagonalizing the matrix ( [ eq : a ] ) one
gets the mass eigenstates @xmath44 and @xmath45 @xmath46 with the masses @xmath47 and @xmath48 ( @xmath49 ) , and the mixing angle @xmath50 @xmath51 the @xmath52 mixing is large if @xmath53 , which may be the case in the @xmath54 sector due to the large @xmath55 and in the @xmath26 sector for large @xmath56 and @xmath57 .
we assume that the gluino mass @xmath58 is real .
we write the u(1 ) gaugino mass @xmath3 as @xmath59 .
inspired by the gaugino mass unification we take @xmath60 and @xmath61 . in the mssm higgs sector with explicit cp violation
the neutral higgs mass eigenstates @xmath62 and @xmath63 @xmath64 are mixtures of cp - even and cp - odd states . for the radiatively corrected masses and mixings of the higgs bosons we use the formulae of ref.@xcite .
possible important decay modes of @xmath4 are : _ 1 & & t , t _
i , b _ j , _ 1 w^+ , _ 1 h^+
[ eq : i ] + _ 2 & & t , t _
i , b _ j , _ 1 z^0 , _ 1,2 w^+ , _ 1 h_k^0 , _ 1,2 h^+ . [ eq : j ] those of @xmath5 are analogous .
the cp phase dependences of the decay widths stems from those of the involved mass - eigenvalues , mixings and couplings among the interaction - eigenfields . in @xcite
it is found that the masses , mixings and couplings of the involved susy particles ( stops , sbottoms , charginos @xmath65 ( @xmath66 ) , neutralinos @xmath67 ( @xmath68 ) , and higgs bosons ) can be very sensitive to the cp phases in a large region of the mssm parameter space .
for the @xmath69 ( @xmath70 ) sectors we find : 1 .
the mass - eigenvalues @xmath71 are sensitive to the phases @xmath72 via @xmath73 if and only if @xmath74 _ and _ @xmath75 ( with @xmath76 and @xmath77 ) ( see eqs .
( [ eq : d]),([eq : g ] ) ) .
the @xmath78-mixing angle @xmath50 ( given by @xmath79 ) is sensitive to @xmath72 via @xmath73 if and only if @xmath80 _ and _ @xmath75 ( see eqs .
( [ eq : d]),([eq : g ] ) , ( [ eq : h ] ) ) .
3 . the @xmath78-mixing phase @xmath81 in eq.([eq : e ] ) is sensitive to ( @xmath82 , @xmath83 ) , @xmath82 and @xmath83 if @xmath75 , @xmath84 and @xmath85 , respectively . for large squark mixing
the term @xmath86 can result in a large phase dependence of the decay widths @xcite ( see eq.([eq : f ] ) ) .
therefore , we expect that the widths ( and hence the branching ratios ) of the @xmath4 and @xmath5 decays are sensitive to the phases ( @xmath87 , @xmath83 , @xmath88 ) in a large region of the mssm parameter space .
in order to improve the convergence of the perturbative expansion @xcite we calculate the tree - level widths by using the corresponding tree - level couplings defined in terms of effective " mssm running quark masses @xmath89 ( i.e. those defined in terms of the effective running yukawa couplings @xmath90 @xmath91 ) . for the kinematics , e.g. , for the phase space factor we use the on - shell masses obtained by using the on - shell ( pole ) quark masses @xmath92 .
we take @xmath93 gev , @xmath94 gev , @xmath95 gev , and @xmath96 gev .
we fix @xmath97 and @xmath98 gev , i.e. @xmath99 gev . in our numerical study
we take @xmath56 , @xmath100 , @xmath101 , @xmath102 , @xmath103 , @xmath57 , @xmath104 , @xmath105 , @xmath83 , @xmath88 and @xmath106 as input parameters , where @xmath107 and @xmath102 are the on - shell squark masses .
note that for a given set of the input parameters we obtain two solutions for ( @xmath108 , @xmath109 ) corresponding to the two cases @xmath110 and @xmath111 from eqs .
( [ eq : a])-([eq : d ] ) and ( [ eq : g ] ) with @xmath55 replaced by @xmath112 . in the plots we impose the conditions as described in @xcite in order to respect experimental and theoretical constraints , such as the lep limits and @xmath113 data .
the experimental limits on the edms of electron , neutron and @xmath114hg strongly constrain the susy cp phases . here
we adopt the scenario of @xcite where the first two generations of the sfermions are very heavy and hence @xmath87 , @xmath83 and @xmath88 are practically unconstrained .
we have also checked that the electron and neutron edm constraints at two - loop level @xcite are fulfilled in the numerical examples studied in this article . in fig.1
we show the @xmath104 dependence of the @xmath115 decay branching ratios for @xmath116 and @xmath117 with @xmath56=8 , ( @xmath100,@xmath101,@xmath102)=(400,700,200 ) gev , @xmath118 gev , @xmath119 gev , @xmath105=@xmath88=0 , and @xmath120 gev in the case @xmath110 .
the case @xmath111 leads to similar results .
we see that the @xmath115 decay branching ratios are very sensitive to @xmath104 and depend significantly on @xmath83 . for large @xmath121
we have obtained results similar to those for @xmath56=8 @xcite . [ 0.8 ] @xmath104 dependence of the @xmath115 decay branching ratios for @xmath116 ( a ) and @xmath117 ( b ) with @xmath56=8 , ( @xmath100,@xmath101,@xmath102)=(400,700,200 ) gev , @xmath118 gev , @xmath119 gev , @xmath105=@xmath88=0 , and @xmath120 gev in the case @xmath110 .
note that the @xmath122 and @xmath123 modes are kinematically forbidden here.,title="fig : " ] [ 0.8 ] @xmath104 dependence of the @xmath115 decay branching ratios for @xmath116 ( a ) and @xmath117 ( b ) with @xmath56=8 , ( @xmath100,@xmath101,@xmath102)=(400,700,200 ) gev , @xmath118 gev , @xmath119 gev , @xmath105=@xmath88=0 , and @xmath120 gev in the case @xmath110 .
note that the @xmath122 and @xmath123 modes are kinematically forbidden here.,title="fig : " ] for @xmath124 and @xmath5 decays we have obtained results similar to those for the @xmath115 decays @xcite .
we have also found that for small @xmath125 the @xmath124 decay can be fairly sensitive to @xmath88 @xcite .
we now study to which extent one can extract the underlying mssm parameters ( such as @xmath56 , @xmath33 etc . ) from measured observables ( such as masses , cross sections and branching ratios ) .
the observables are functions of the mssm parameters .
therefore , in case the number of measured observables is larger than that of the mssm parameters , the whole data set of the observables over - constrains the mssm parameters . in this case
the mssm parameters and their errors can be determined by a @xmath126 fit to the experimental data of the observables . in the fit
the errors of the observables are propagated to those of the mssm parameters .
the assumptions for the expected experimental errors of the observables measured at tesla , clic and lhc are described in @xcite .
our strategy for the parameter determination is as follows : 1 .
take a specific set of values of the underlying mssm parameters .
calculate the masses of @xmath127 , @xmath128 , @xmath129 , @xmath130 , @xmath131 , the production cross sections for @xmath132 , and @xmath133 , and the branching ratios of the @xmath127 and @xmath128 decays , and estimate the expected experimental errors of these observables .
3 . regard these calculated values as real experimental data with definite errors .
determine the underlying mssm parameters and their errors from the `` experimental data '' by a @xmath126 fit .
we consider two scenarios , one with small @xmath134 and one with large @xmath134 .
the small @xmath134 scenario is characterized by the following underlying mssm parameters : @xmath135 169.6 gev , @xmath136 408.8 gev , @xmath137 623.0 gev , @xmath138 800 gev , @xmath139 , @xmath140 , @xmath141 300 gev , @xmath142 gev , @xmath143 , @xmath144 1000 gev , and @xmath145 900 gev .
( here we do not assume the unification relation between @xmath146 and @xmath13 . )
the large @xmath134 scenario is specified by : @xmath135 360.0 gev , @xmath136 198.2 gev , @xmath137 691.9 gev , @xmath147 600 gev , @xmath148 , @xmath149 1000 gev , @xmath150 , @xmath151 , @xmath141 200 gev , @xmath142 gev , @xmath152 30 , @xmath144 1000 gev , and @xmath145 350 gev .
the resulting values of the observables and their expected experimental errors are shown for the two scenarios in @xcite .
we regard these calculated values as real experimental data with definite errors .
we determine the underlying mssm parameters and their errors from the `` experimental data '' on these observables by a @xmath126 fit .
the results obtained are shown in table 2 of @xcite .
as one can see , all parameters except @xmath19 can be determined rather precisely .
@xmath134 can be determined with an error of about 3% in both scenarios .
the relative error of the squark mass parameters squared @xmath153 is in the range of 1% to 2% .
@xmath154 and @xmath155 can be measured within an error of 2 3% independently of @xmath134 .
the situation for @xmath19 is considerably worse : in case of small @xmath134 one gets only an order of magnitude estimate .
the reason is that both the bottom squark mixing angle and the bottom squark couplings depend only weakly on @xmath19 for small @xmath134 . in case of large @xmath134
the situation improves somewhat in particular for the imaginary part of @xmath19 .
a @xmath126 fit using only real mssm parameters would result in a totally wrong parameter determination with a significantly larger value of @xmath126 .
we have found that the analogous fit procedure using only real mssm parameters gives a much larger value for @xmath126 : @xmath156 with dof=61 for the scenario with @xmath157 and @xmath158 with dof=61 for the scenario with @xmath159 , where dof is degree of freedom.
we have studied the decays of @xmath0 and @xmath1 in the mssm with complex parameters @xmath160 and @xmath3 .
we have shown that including the cp phases strongly affects the branching ratios of @xmath0 and @xmath1 decays in a large domain of the mssm parameter space .
this could have an important impact on the search for @xmath4 and @xmath5 and the determination of the underlying mssm parameters at future colliders . | we study the decays of top squarks ( @xmath0 ) and bottom squarks ( @xmath1 ) in the minimal supersymmetric standard model ( mssm ) with complex parameters @xmath2 and @xmath3 .
we show that including the corresponding phases strongly affects the branching ratios of @xmath0 and @xmath1 decays in a large domain of the mssm parameter space .
this could have an important impact on the search for @xmath4 and @xmath5 and the determination of the underlying mssm parameters at future colliders .
tgu-34 + uwthph-2004 - 26 + hep - ph/0409347 + a. bartl@xmath6 , s. hesselbach@xmath6 , @xmath7 , t. kernreiter@xmath6 and w. porod@xmath8 [ cols= " < " , ] |
[ [ synthesis . ] ] synthesis. + + + + + + + + + + + polycrystalline samples of @xmath0iro@xmath1 ( @xmath2 na , li ) and li@xmath4sno@xmath1 were synthesized by solid state synthesis .
high purity starting materials @xmath0co@xmath1 ( @xmath2 na , li ) ( 99.995% alfa aesar ) and ir metal powder ( @xmath98 alfa aesar ) or sno@xmath4 ( 99.995% alfa aesar ) were mixed in the ratio 1.05 : 1 and placed in an alumina crucible with a lid and given heat treatments between 750 @xmath99c and 950 @xmath99c in steps of 50 @xmath99c with intermediate grindings and pelletizing after each step .
the li@xmath4sno@xmath1 sample was given a further heat treatment at 1000 @xmath99c . in an attempt to grow single crystals
the li@xmath4iro@xmath1 sample was dissolved in excess licl flux at 850 @xmath99c for 6 hrs and then cooled to 590 @xmath99c at 3 @xmath99c / hr . after washing majority of the licl flux with de - ionized water we obtained a fine black powder which turned out to be highly ordered li@xmath4iro@xmath1 polycrystalline samples .
this powder was pelletized and sin tered at 900 @xmath99c for 48 hrs to get a hard pellet for resistivity and heat capacity measurements .
[ [ structure . ] ] structure. + + + + + + + + + + + powder x - ray diffraction ( xrd ) scans of polycrystalline @xmath0iro@xmath1 ( @xmath60 na , li ) and li@xmath4sno@xmath1 were found to be single phase and could be indexed to the monoclinic _ c2/c _
( no . 15 ) structure .
rietveld refinements,@xcite of the x - ray patterns gave the unit cell parameters @xmath100 = 5.4187(2 ) , @xmath101 = 9.3689(3 ) @xmath87 = 10.7731(5 ) and @xmath102 for na@xmath4iro@xmath1 and @xmath100 = 5.1678(4 ) , @xmath101 = 8.9347(7 ) @xmath87 = 9.7825(4 ) and @xmath103 for li@xmath4iro@xmath1 . a representative xrd scan for li@xmath4iro@xmath1 obtained by licl flux method
is shown in fig .
[ fig - xrd ] along with the rietveld refinement shown as the curve through the data .
regions where small peaks of licl appeared have been excluded in the fits .
the inset of fig .
[ fig - xrd ] shows the xrd data and rietveld refinement at low angles on an expanded scale .
the highly ordered nature of the sample can be seen from the separation of the lines between @xmath104 and @xmath105 which in a disordered sample would appear merged together and would also show a highly anisotropic shape with a tail extending to the higher angle side of the peak.@xcite
the resistivity @xmath106 versus temperature @xmath38 for li@xmath4iro@xmath1 is shown in fig .
[ fig - res ] .
the data clearly point to insulating behavior with a room temperature value @xmath107 cm and an activation gap @xmath108 k. the insulating resistivity coupled with the local - moment magnetism indicates that li@xmath4iro@xmath1 is a mott insulator . versus temperature @xmath38 data for polycrystalline li@xmath4iro@xmath1 shown on a semi - log plot .
[ fig - res ] ]
we applied the wien2k @xcite implementation of the all electron linearized augmented plane wave method in the lda+u+so approximation , within the second variational treatment using u=3ev and j=0.6ev . for the lattice structure the empirical monoclinic lattice structure
c2/c @xcite was used .
the moments directions were constrained along the three crystallographic axes .
the moments connected by half a unit cell in the c direction , were aligned antiferromagnetically .
the basis size was determined by rkmax=8.0 and the brillouin zone was sampled with a 5x5x2-mesh . for the zig - zag , nel and stripy configuration , the moment direction with the lowest total energy
is found to be ` a ' , ` c ' and ` a ' respectively .
we note that the constrained moment directions investigated in this study are not meant to represent the moment direction of the theoretical ground state , which in general could be away from the crystallographic axes and non - collinear .
we also note that another closely related space group c2/m has been suggested for li@xmath4iro@xmath1 in the literature @xcite .
since our x - ray diffraction patterns could be fit equally well with either space group , we have used c2/c in our dft calculations for li@xmath4iro@xmath1 to be consistent with what was previously used in calculations for na@xmath4iro@xmath1 @xcite .
to capture the experimental data we consider an extended heisenberg - kitaev model which is augmented by second and third nearest neighbor heisenberg couplings @xmath30 and @xmath31 , respectively .
explicitly , the hamiltonian is given by where nearest - neighbor exchange is indicated by the @xmath110 brackets , and second and third nearest neighbor exchange by @xmath111 and @xmath112 , respectively .
@xmath30 and @xmath31 are given in units of the nearest neighbor heisenberg coupling .
using a pseudo - fermion functional renormalization group ( frg ) calculations we were able to ( i ) determine the phase diagram , ( ii ) characterize the magnetic ordering of the different phases , and ( iii ) calculate thermodynamic properties such as the curie - weiss scale , the ordering scale , and the frustration parameter .
quickly summarized , the main outcome of the frg calculation is the magnetic spin susceptibility as the linear response to a small external magnetic field , calculated via kubos formula .
note that the magnetic field always points along one of the cubic axes .
since the frg naturally computes the spin susceptibility as a function of a frequency cutoff parameter @xmath113 , we are able to extract thermodynamic properties of the model system via an identification of precisely this frequency cutoff parameter @xmath113 with a temperature @xmath38 .
our calculations use cluster sizes of the hexagonal lattice of 112 lattice sites . for further technical details about the pseudo - fermion frg method and the extraction of the curie - weiss scales and ordering scales from the frg data we refer to refs . .
first , we turn to the overall phase diagram of model . this phase diagram has recently been calculated @xcite on a classical level , which revealed four phases with different types of long - range order , i.e. , antiferromagnetic nel order , stripy order , spiral order , and zig - zag order ( for an illustration of the ordering patterns , see e.g. ref . ) . in our approach
, the inclusion of quantum fluctuations in the frg flow enables us to treat the system far beyond the classical limit .
however , despite the quantum nature of our method , we clearly see that in a wide parameter regime the phase boundaries remain roughly the same as compared the to classical case . in order to exemplify this result , fig .
[ fig - phases ] shows the phase diagram for @xmath114 with @xmath115 .
this cut through the parameter space exhibits stripy , spiral and zigzag ordered phases with boundaries which almost agree with the classical one s. only in the vicinity of the limits @xmath116 or @xmath117 quantum fluctuations have a more significant effect on the phase diagram and may even destroy the long - range order , see fig . 3 of the main text .
the different types of order in the phase diagram of fig .
[ fig - phases ] also reveal themselves through their different peak structures of the spin susceptibility in @xmath118-space .
[ fig - order ] displays these susceptibility profiles in the entire extended ( second ) brillouin zone for selected parameter points . in the case of stripy order
, large response peaks can be seen at an @xmath119-point position ( due to the external magnetic field along one of the cubic axes the susceptibility profile loses its six fold rotation symmetry ) . upon entering the spiral regime
, the ordering peak leaves the commensurate @xmath119 point and moves continuously towards the @xmath120 point . as one approaches the phase boundaries to the zig - zag phase , the ordering peaks broaden in @xmath121 direction and finally split into two separate peaks which mark the @xmath118-space positions of the commensurate zig - zag order .
these peaks are located in the middle between the @xmath120 point and the corners of the extended brillouin zone .
recall that there are also three distinct zig - zag ordering configurations corresponding to three differently oriented zig - zag paths in the hexagonal lattice .
each of these configurations relates to two ordering peaks at @xmath118 and @xmath122 . as can be seen in the right plot of fig .
[ fig - order ] the symmetry - breaking property of the external magnetic field only allows for two zig - zag ordering configurations , which corresponds to four peaks in the brillouin zone .
finally , we turn to thermodynamic properties of model , in particular the suppression of magnetic ordering with regard to the curie - weiss temperature , which is often considered a measure of frustration and quantified by the frustration parameter @xmath123 , i.e. the ratio of the curie - weiss temperature @xmath7 and the ordering temperature @xmath8 . in fig .
[ fig - f ] , we again consider the case @xmath114 .
the qualitative behavior of @xmath35 shown there is characteristic also for smaller and larger values of @xmath20 .
the stripy ordered phase can clearly be identified as the region with the smallest frustration , which is in agreement with the fact that this order is classically saturated at @xmath114 , @xmath124 . as the boundary to the spiral phase
is crossed , the frustration parameter undergoes a pronounced increase . in general
, @xmath35 grows with increasing @xmath30 and @xmath31 ( at least if @xmath31 is not too large ) and the largest amount of frustration i s obtained near the transition line between spiral order and zig - zag order .
hence , by adding @xmath30 and @xmath31 interactions , the frustration can be enhanced by a factor of three as compared to the bare heisenberg - kitaev case . as shown in fig .
3 of the main text , @xmath35 decreases monotonically with increasing @xmath20 , while the general profile of frustration in the @xmath30-@xmath31 plane remains similar to fig .
[ fig - f ] .
however , depending on @xmath30 and @xmath31 , for small @xmath20 a non - magnetic phase may appear , which causes a sharp rise of the frustration parameter in the nearby ordered phases .
hence , in regimes of small @xmath20 , the capability of @xmath30 and @xmath31 interactions to increase the frustration is most pronounced . | combining thermodynamic measurements with theoretical density functional and thermodynamic calculations we demonstrate that the honeycomb lattice iridates @xmath0iro@xmath1 ( @xmath2na , li ) are magnetically ordered mott insulators where the magnetism of the effective spin - orbital s=1/2 moments can be captured by a heisenberg - kitaev ( hk ) model with heisenberg interactions beyond nearest - neighbor exchange .
experimentally , we observe an increase of the curie - weiss temperature from @xmath3 k for na@xmath4iro@xmath1 to @xmath5 k for li@xmath4iro@xmath1 , while the antiferromagnetic ordering temperature remains roughly the same @xmath6 k for both materials .
using finite - temperature functional renormalization group calculations we show that this evolution of @xmath7 , @xmath8 , the frustration parameter @xmath9 , and the zig - zag magnetic ordering structure suggested for both materials by density functional theory can be captured within this extended hk model . combining our experimental and theoretical results
, we estimate that na@xmath4iro@xmath1 is deep in the magnetically ordered regime of the hk model ( @xmath10 ) , while li@xmath4iro@xmath1 appears to be close to a spin - liquid regime ( @xmath11 ) .
[ [ introduction . ] ] introduction. + + + + + + + + + + + + + + the fundamental importance of the kitaev model , which describes the highly anisotropic exchange of su(2 ) spin-1/2 moments on the honeycomb lattice , has quickly been appreciated for its rare combination of microscopic simplicity and an exact analytical solution @xcite .
it has also become an archetypal example of a microscopic model that
depending on the spatial anisotropy of its couplings harbors three of the currently most sought - after collective states in condensed matter physics @xcite : a gapless spin liquid with emergent majorana fermion excitations , a gapped @xmath12 spin liquid , and a topologically ordered phase with non - abelian quasiparticle statistics ( in the presence of a magnetic field perpendicular to the honeycomb lattice ) @xcite .
especially , physical realizations of topological states of the latter form which support majorana fermion zero modes @xcite are intensely searched for in various candidate systems including certain fractional quantum hall systems @xcite , unconventional superconductors @xcite , as well as heterostructures of topological insulators , semi - metals , or semiconductors with conventional @xmath13-wave superconductors @xcite , not only because of their fundamentally new character but also due to their possible application in topological quantum computation proposals @xcite .
a direct realization of the kitaev model could provide yet another alternative path to this goal .
first proposals to engineer implementations of the kitaev model were discussed in the context of optical lattices @xcite and superconducting circuits @xcite .
more recently , it has been put forward that strong spin - orbit coupling in certain mott insulating transition metal oxides @xcite could inherently give rise to kitaev - type couplings of effective spin - orbital degrees of freedom . among the best candidate materials are layered iridates of the form @xmath0iro@xmath1 , which exhibit mott insulating ground states and where the ir@xmath14 form effective @xmath15 moments as it was recently observed for na@xmath4iro@xmath1 @xcite . on a microscopic level
, it has been argued that the strong spin - orbit coupling in these 5d transition metal systems leads to orbital dependent anisotropic in - plane exchange that precisely mimics the kitaev couplings @xcite . for real materials , however , further interactions will inevitably be present and in particular one might expect that the original spin exchange is not completely damped and isotropic heisenberg interactions will compete with the anisotropic kitaev couplings @xcite .
such a heisenberg - kitaev ( hk ) model can be written down in its simplest form as @xmath16 where the @xmath17 are pauli matrices for the effective @xmath18 and @xmath19 labels the three different links for each spin of the honeycomb lattice .
it has been shown @xcite that the isotropic heisenberg exchange in the first term of model enters as antiferromagnetic coupling , while the anisotropic kitaev exchange ( in the second term ) is ferromagnetic .
varying the relative coupling @xmath20 of the two exchange terms a sequence of three different phases has been found @xcite : a conventional nel antiferromagnet for @xmath21 , a so - called a stripy antiferromagnet for @xmath22 , and a spin - liquid state for @xmath23 .
but while the @xmath0iro@xmath1 materials have been appreciated from a theory perspective as possible candidate materials to look for kitaev - like and hk - like physics @xcite , there has so far been very limited experimental data available for these layered iridates . for li@xmath4iro@xmath1
there have been two conflicting reports @xcite , with one report @xcite suggesting paramagnetic behavior between @xmath24 k and 300 k without any sign of magnetic order , while the second report @xcite indicated an anomaly in the magnetic susceptibility below @xmath25 k , which was also accompanied by a hysteresis between zero - field - cooled and field - cooled data , suggesting glassy behavior .
no heat capacity data has so far been available for li@xmath4iro@xmath1 . for single crystal na@xmath4iro@xmath1
some of us have earlier shown @xcite mott insulating behavior with antiferromagnetic ordering below @xmath26 k. subsequent resonant magnetic x - ray scattering measurements @xcite on single crystals were consistent with either stripy or zig - zag magnetic order , with supplementary dft calculations indicating that zig - zag order was the more likely magnetic ground state for na@xmath4iro@xmath1 .
this gave rise to another theoretical puzzle , since the original hk - model with nearest - neighbor exchange , i.e. model , allows for stripy magnetic ordering but not zig - zag order .
finite - temperature calculations @xcite for model pointed to another discrepancy with experimental observations , since the theoretical calculations indicated that the competition of the heisenberg and kitaev exchanges in model does not lead to a substantial suppression of the magnetic ordering transition with regard to the curie - weiss scale and the frustration parameter @xmath27 was found to never exceed @xmath28 @xcite , while for na@xmath4iro@xmath1 experiments indicate @xmath29 @xcite . pieces of this puzzle were recently solved when it was shown that taking into account heisenberg interactions beyond the nearest neighbor exchange can indeed stabilize the zig - zag ordering pattern @xcite . for antiferromagnetic exchanges , the latter
are also expected to introduce geometric frustration . in the following
we will expand this discussion of the role of further neighbor heisenberg exchange and by providing a detailed comparison of theoretical and experimental results we will establish a microscopic description of the layered iridates @xmath0iro@xmath1 in terms of such an extended heisenberg - kitaev model . quickly summarizing our main results we report magnetic and heat capacity measurements on high quality polycrystalline samples of @xmath0iro@xmath1 ( @xmath2 na , li ) .
magnetic measurements show local moment behavior with effective spin @xmath18 moments . both magnetic and heat capacity measurements show sharp anomalies at @xmath26 k for both materials indicating bulk antiferromagnetic ordering . for both materials
our dft calculations indicate that the most likely magnetic order is of zig - zag type .
finite - temperature functional renormalization group ( frg ) calculations for an extended hk model including next - nearest ( @xmath30 ) and next - to - next - nearest ( @xmath31 ) neighbor antiferromagnetic heisenberg interactions are then used to confirm the type of magnetic order , and study the evolution of the curie - weiss temperature scale @xmath7 , the ordering scale @xmath32 , and the frustration parameter @xmath33 as @xmath20 and the various @xmath34 s are varied .
it must be emphasized that in contrast to the classical phase diagram of the extended heisenberg - kitaev model discussed earlier@xcite our frg calculations are performed on the quantum level .
we show that the experimentally observed evolution of @xmath7 , @xmath8 , and @xmath35 and the observed magnetic order can all be very well captured within this extended hk model .
comparison of experiments with calculations suggest that while the kitaev term is small for na@xmath4iro@xmath1 , the li@xmath4iro@xmath1 system with @xmath11 sits quite close to the spin - liquid state in the kitaev limit @xmath36 .
[ [ magnetic - susceptibility . ] ] magnetic susceptibility. + + + + + + + + + + + + + + + + + + + + + + + + + the magnetic susceptibility @xmath37 versus @xmath38 data for @xmath0iro@xmath1 ( @xmath2 na , li ) are shown in fig .
[ fig - chi ] .
the @xmath39 data between @xmath40 k and 300 k were fit by the curie - weiss expression @xmath41 with @xmath42 , @xmath43 , and @xmath7 as fitting parameters .
the fit , shown in fig .
[ fig - chi ] as the solid curve through the data , gives the values @xmath44
@xmath45/mol , @xmath46 @xmath45 k / mol , and @xmath47 k , for na@xmath4iro@xmath1 and , @xmath48 @xmath45/mol , @xmath49 @xmath45 k / mol , and @xmath50 k , for li@xmath4iro@xmath1 , respectively .
assuming a @xmath51-factor @xmath52 the above values of @xmath43 correspond to an effective moment of @xmath53 and @xmath54 , for na@xmath4iro@xmath1 and li@xmath4iro@xmath1 , respectively .
these values of @xmath55 are close to the value 1.74 @xmath56 expected for spin = 1/2 moments .
this local - moment formation along with the insulating resistivity ( see auxiliary material @xcite ) indicates that like its sister compound li@xmath4iro@xmath1 is indeed a mott insulator .
the value of the weiss temperature @xmath57 k for li@xmath4iro@xmath1 further suggests that the effective antiferromagnetic exchange interactions have weakened when compared to the na@xmath4iro@xmath1 system .
however , the @xmath39 data for li@xmath4iro@xmath1 also show an anomaly at @xmath58 k suggesting that an antiferromagnetic transition occurs at roughly the same temperature as for na@xmath4iro@xmath1 .
this is further supported by our heat capacity results presented below .
the insets ( a ) and ( b ) in fig .
[ fig - chi ] show the @xmath39 data at low temperatures to highlight the anomaly seen at the onset of the antiferromagnetic transition below @xmath26 k in both materials .
[ [ heat - capacity . ] ] heat capacity. + + + + + + + + + + + + + + + in fig .
[ fig - heatcap ] we show heat capacity data divided by temperature @xmath59 versus temperature @xmath38 for @xmath0iro@xmath1 ( @xmath60 na , li ) , and for the non - magnetic analog li@xmath4sno@xmath1 .
the anomaly at @xmath26 k in the data for both @xmath0iro@xmath1 ( @xmath60 na , li ) materials confirms bulk magnetic ordering .
a small bump is also observed around @xmath24 k in the @xmath61 for li@xmath4iro@xmath1 .
this most likely arises due to a small amount ( @xmath62 ) of disorder in the sample.@xcite the magnetic contribution @xmath63 for li@xmath4iro@xmath1 shown in the inset of fig .
[ fig - heatcap ] was obtained by subtracting the @xmath61 data of li@xmath4sno@xmath1 from the data of li@xmath4iro@xmath1 .
the latter reveals a clearly more visible lambda - like anomaly at @xmath26 k. a slight depression of @xmath32 in an applied magnetic field of @xmath64 t was observed ( not shown ) which points to the antiferromagnetic nature of the magnetic ordering in li@xmath4iro@xmath1 .
the entropy @xmath65 obtained by integrating the @xmath66 versus @xmath38 data is also shown in fig .
[ fig - heatcap ] inset . just above @xmath32
the entropy is only about 15%rln2 .
such a reduced entropy at the transition was also observed earlier for single crystalline na@xmath4iro@xmath1.@xcite the small entropy points to the reduced ordered moment and the possible proximity to a non - magnetic ground state . [
[ magnetic - ordering . ] ] magnetic ordering. + + + + + + + + + + + + + + + + + + + from the similarities in the anomalies seen in @xmath67 and @xmath43 data for both the na and li systems , it would seem likely that the kind of magnetic order would also be similar for the two . to resolve the magnetic structure for li@xmath4iro@xmath1 , spin
density function calculations within the lda+u+so approximation were performed for the nel , stripy , and zig - zag configurations with the moments constrained along the crystallographic axes @xcite .
the results are summarized in table [ tab : tab1 ] .
we find that as for na@xmath4iro@xmath1 @xcite , the zig - zag configuration has the lowest energy and is hence the most likely magnetic structure for li@xmath4iro@xmath1 . .
total energy e@xmath68 per ir for li@xmath4iro@xmath1 for three antiferromagnetic configurations , obtained from collinearly constrained lda+u+so simulations .
[ cols="^,^,^,^,^",options="header " , ] we now turn to the evolution of magnetic properties as we go from the na to the li compound . from our @xmath39 data above we find that the curie - weiss temperature @xmath7 decreases from @xmath69 k to @xmath70 k on going from na@xmath4iro@xmath1 to li@xmath4iro@xmath1 possibly indicating that the effective magnetic interactions are weaker for li@xmath4iro@xmath1 . surprisingly however ,
both @xmath39 and @xmath61 show that both materials order magnetically at roughly the same temperature @xmath6 k. the frustration parameter @xmath9 , however , reduces from @xmath71 for na@xmath4iro@xmath1 to @xmath72 for li@xmath4iro@xmath1 . in previous theoretical calculations @xcite for the thermodynamics of the hk model ,
the ordering temperature @xmath8 was found to be largely insensitive to variations of @xmath20 whereas @xmath7 was found to decrease monotonically with @xmath20 within the stripy magnetic phase .
the increase of the weiss temperature scale on increasing @xmath20 is a direct consequence of the fact that the two coupling terms in the hk model enter with opposite coupling signs . increasing the relative strength of the ferromagnetic kitaev term thus leads to an increase of the weiss temperature scale .
these theoretical trends thus seem to match well with what is observed in our experiments .
there are , however , two issues where experiments differ from predictions for the hk model .
first , the zig - zag magnetic order obtained from dft for both na@xmath4iro@xmath1 and li@xmath4iro@xmath1 is not one of the three phases of the hk model @xcite .
secondly , a maximum frustration parameter @xmath28 was found in calculations for the hk model@xcite , which is much smaller than the experimentally observed value @xmath29 for na@xmath4iro@xmath1@xcite . to resolve these discrepancies
it has recently been argued @xcite that further nearest neighbor antiferromagnetic heisenberg exchange interactions should be added to the original hk model , which can indeed stabilize the zig - zag magnetic order .
it was further demonstrated @xcite that the experimental magnetic susceptibility data for @xmath0iro@xmath1 ( @xmath60 na , li ) materials can only be fit when including the kitaev term in this expanded microscopic model . the inclusion of further than nearest - neighbor antiferromagnetic interactions is further expected to introduce geometric frustration in addition to the frustration arising from the competition between the heisenberg and kitaev couplings of the original model .
we have therefore expanded our frg calculations @xcite to such an extended hk-@xmath30,@xmath31 model and determine its thermodynamic properties by extracting the high - temperature cw behavior ( from the rg flow ) , the onset of magnetic ordering ( from the breakdown of the rg flow ) , and the nature of the various ground states by calculating momentum - resolved magnetic susceptibility profiles as further detailed in the auxiliary material @xcite .
we focused our calculations on the parameter regime @xmath73 and @xmath74 .
a representative plot of the ordering scale @xmath75 as a function of @xmath20 for fixed @xmath76 is shown in figure [ fig - t - vs - alpha ] with the inset showing the evolution of the frustration parameter @xmath35 .
our calculations indicate that zig - zag order is stabilized for an extended range @xmath77 in agreement with recent calculations for the hk-@xmath78 model @xcite . around the kitaev limit for @xmath36
we find an extended non - magnetic spin - liquid phase , which connects directly to the one of the original hk model .
for @xmath79 we obtain another non - magnetic ground state , evidently arising from the further nearest - neighbor exchange @xmath78 frustrating the nearest - neighbor @xmath34 and suppressing the nel state in favor of a valence bond dimer crystal @xcite .
our results summarized in fig . [ fig - t - vs - alpha ] further indicate that the potentially counter - intuitive experimental observation of @xmath8 staying roughly the same in going from na@xmath4iro@xmath1 to li@xmath4iro@xmath1 even though @xmath7 decreases by a factor of @xmath80 in fact agrees well with our calculations showing that the ordering scale stays more or less constant for the zig - zag ordered ground state in the regime @xmath77 . finally , we note that our calculations also indicate that the frustration parameter @xmath9 decreases monotonically with @xmath20 in the region where magnetic order is found ( see the inset in fig .
[ fig - t - vs - alpha ] ) , which is a direct consequence of the curie - weiss scale @xmath7 decreasing monotonically in this region .
interestingly , for small @xmath20 the geometric frustration induced by the further nearest - neighbor exchange becomes more evident and the parameter @xmath35 reaches values much larger than obtained for the original hk model @xcite and in fact becomes comparable in size to what is observed experimentally for na@xmath4iro@xmath1 where @xmath29 . to place the @xmath0iro@xmath1 materials on the diagram in fig .
[ fig - t - vs - alpha ] we note that the zig - zag ordered ground state indicated in dft calculations for both materials is found only in the parameter range @xmath77 in the presence of significant second and third neighbor exchange .
additionally , an enhanced frustration parameter @xmath35 is found only for small @xmath81 before the system transitions to a non - magnetic ground state ( for @xmath79 ) and the ordering temperature starts to drop drastically .
we therefore place na@xmath4iro@xmath1 at @xmath82 . in contrast , li@xmath4iro@xmath1 with zig - zag order at @xmath83 k and @xmath28 can be placed at @xmath84 putting it considerably closer to the spin - liquid regime @xcite around the kitaev limit beyond @xmath36 .
finally , we note that in going from the na to the li system the @xmath85 lattice parameters are reduced by @xmath86 while the @xmath87 parameter is reduced by @xmath88 .
thus , substituting na by li is equivalent to preferentially applying chemical pressure along the @xmath87 axis ( @xmath89 to the honeycomb planes ) .
this most likely leads to the iro@xmath90 octahedra becoming more symmetrical within the @xmath91-plane which in turn enhances the parameters @xmath92 ( in the notation of ref . )
leading to an increased kitaev coupling .
this is consistent with the value of @xmath93 estimated above for li@xmath4iro@xmath1 which puts its closer to the kitaev limit . in summary
, we have shown that magnetic properties of the mott insulating iridates @xmath0iro@xmath1 , in particular the evolution of thermodynamic observables , i.e. the curie - weiss temperature @xmath7 and ordering temperature @xmath8 , as well as their low - temperature magnetic order can be captured within an extended heisenberg - kitaev model .
our detailed comparison of experiment and theory in particular suggests that while na@xmath4iro@xmath1 is located deep in a magnetically ordered regime , li@xmath4iro@xmath1 lies close to the spin - liquid regime around the kitaev limit ( @xmath36 ) .
future experiments will further investigate whether the application of @xmath87-axis pressure can push these systems deeper into the orbitally dominated regime , and in particular whether li@xmath4iro@xmath1 can be pushed into the spin - liquid phase .
[ [ acknowledgments . ] ] acknowledgments. + + + + + + + + + + + + + + + + + ys acknowledges support from alexander von humboldt foundation and from dst , india .
sm acknowledges support from the erasmus mundus eurindia project .
jr is supported by the deutsche akademie der naturforscher leopoldina through grant lpds 2011 - 14 .
rt is supported by an sitp fellowship by stanford university .
wk and tb acknowledge support from doe # de - ac02 - 98ch10886 a. kitaev , ann . phys .
( n.y . ) * 321 * , 2 ( 2006 ) .
l. balents , nature * 464 * , 199 ( 2010 ) .
h. jiang _
et al . _
phys .
rev .
b * 83 * , 245104 ( 2011 ) .
a. stern , nature * 464 * , 187 ( 2010 ) .
g. moore and n. read , nucl .
phys .
b * 360 * , 362 ( 1991 ) .
n. read and d. green , phys .
rev .
b * 61 * , 10267 ( 2000 ) .
l. fu and c. l. kane , phys .
rev .
lett . * 100 * , 096407 ( 2008 ) .
j. d. sau _ et al .
_ , phys .
rev .
lett .
* 104 * , 040502 ( 2010 ) .
j. alicea , phys .
rev .
b * 81 * , 125318 ( 2010 ) . c. nayak _ et al . _ ,
rev .
mod .
phys . * 80 * , 1083 ( 2008 ) .
l .- m .
duan _
et al . _ ,
phys .
rev .
lett . * 91 * , 090402 ( 2003 ) .
j. q. you _ et al .
_ , phys .
rev .
b * 81 * , 014505 ( 2010 ) . g. jackeli and g. khaliullin , phys
. rev .
lett . * 102 * , 017205 ( 2009 ) .
j. chaloupka _
et al . _ ,
phys .
rev .
lett . * 105 * , 027204 ( 2010 ) . y. singh and p. gegenwart , phys .
rev .
b * 82 * , 064412 ( 2010 ) .
j. reuther _
et al . _ ,
phys .
rev .
b * 84 * , 100406(r ) ( 2011 ) . i. felner and i. m. bradaric , physica b * 311 * , 195 ( 2002 ) .
h. kobayashi _
et al . _ ,
j. mater .
chem .
* 13 * , 957 ( 2003 ) .
x. liu _
et al . _ ,
phys .
rev .
b * 83 * , 220403(r ) ( 2011 ) . i. kimchi and y - z . you , phys .
rev .
b * 84 * , 180407(r ) ( 2011 ) .
j. reuther , d. a. abanin , and r. thomale , phys .
rev .
b * 84 * , 014417 ( 2011 ) .
s. manni _
et al . _ ,
unpublished .
we note that such a relative placement of the two sister compounds can also be obtained from their relative magnetic susceptibility , which has been studied @xcite for the original hk model .
it has been predicted that as @xmath20 varies within the stripy phase , @xmath94 .
thus , the higher zero - temperature susceptibility of the li system , see fig .
[ fig - chi ] , indeed points to a higher value of @xmath20 with regard to the na system . a more quantitative analysis of the relative zero - temperature susceptibilities , yields @xmath95 .
conservatively placing the na system at @xmath96 , one obtains @xmath97 consistent with the above estimates from the extended hk model .
f. trousselet _
et al . _ , phys .
rev .
b * 84 * , 054409 ( 2011 ) . see the auxiliary material available in the appendix . * auxiliary material * |
with the advent of the x - ray free - electron lasers @xcite there are new possibilities to explore matter on atomic and single molecule levels . on these length scales , of the order of a few ngstrm , quantum effects play an important role in the dynamics of the electrons .
quantum effects have been measured experimentally both in the degenerate electron gas in metals and in warm dense matters @xcite .
it has also been found that quantum mechanical effects must be taken into account in intense laser - solid density plasma interaction experiments @xcite .
the interaction of large amplitude electromagnetic waves with the plasma can lead to various parametric instabilities @xcite . at laser intensities around @xmath0 and above
, the nonlinearity associated with relativistic electron mass increase in short laser pulses plays a significant role .
furthermore , the relativistic ponderomotive force @xcite of intense laser pulses produces density modifications .
thus , in a classical plasma , nonlinear effects associated with relativistic electron mass increase and relativistic ponderomotive force very important , since they provide the possibility of the modulational instability @xcite followed by a compression and localization of intense electromagnetic waves .
in addition to the modulational instability , there are relativistic raman forward and backward scattering instabilities @xcite and the two - plasmon decay @xcite instability that lead to strong collisionless heating of the plasma in the relativistic regime .
the parametric instabilities of intense electromagnetic waves in magnetized plasmas have also been investigated @xcite .
however , for intense electromagnetic waves interacting with the plasma in the x - ray and @xmath1-ray regimes , both relativistic and quantum effects must me taken into account on equal footing . accordingly , in this paper , we present a simple nonlinear model , based on the klein - gordon ( kg ) equation coupled with the maxwell equations that are capable of treating both the relativistic and quantum effects .
our work has applications in laboratories @xcite , in quantum free electron laser systems @xcite , as well as in astrophysical settings @xcite where white dwarf cores @xcite and neutron stars @xcite are strong sources of x - rays and @xmath1-rays .
the manuscript is organized as follows . in sec .
ii , we present our mathematical model for the coupled kg and maxwell equations , exhibiting nonlinear interactions between relativistic electrons and electromagnetic fields .
linear properties of the electrostatic and electromagnetic waves are discussed in sec .
section iv shows hoe our governing equations lead to the wave equation that reveals the phenomena of relativistic self - focusing and relativistic self - induced transparency of electromagnetic waves .
section v is concerned with the theoretical and numerical investigations of the relativistic parametric instabilities in the quantum regime .
section vi deals with relativistic optical solitary waves .
the nonlinear dynamics of interacting intense localized electromagnetic pulses , as well as the new phenomena of the formation of nonlinear bernstein - greene - krushkal ( bgk)-like modes and associated electron acceleration are described in sec .
section viii contains a brief summary and conclusions .
historically , the klein - gordon equation ( kge ) for an electron is obtained from the relativistic relation between the energy @xmath2 and the momentum @xmath3 , viz .
@xmath4 where @xmath5 is the speed of light in vacuum and @xmath6 the electron mass . by the substitution @xmath7 and @xmath8 in ( 1 ) , where @xmath9 is the planck constant divided by @xmath10
, we obtain the kge for a free electron as @xmath11 where @xmath12 is the electron wave function .
the free - particle kge fulfills the continuity equation @xmath13 where @xmath14 and @xmath15 we have multiplied the right - hand sides of eqs .
( [ eq4 ] ) and ( [ eq5 ] ) by the electron charge @xmath16 , so that @xmath17 can be interpreted as the electric charge density and @xmath18 as the electric current density . since @xmath17 is neither positive or negative definite , it can not be interpreted as a probability density , however , it can be interpreted as a charge density which need not has a definite sign .
we now wish to use the charge and current densities as sources for the self - consistent electromagnetic scalar and vector potentials @xmath19 and @xmath20 for a quantum plasma .
we , therefore , let @xmath12 represent an ensemble of the electrons . introducing the electromagnetic potentials into the kge
, we make the usual substitutions @xmath21 and @xmath22 , obtaining @xmath23 where we have defined the energy and momentum operators as @xmath24 and @xmath25 respectively .
the electric charge and current densities are now obtained as @xmath26 , \label{rho_e}\ ] ] and @xmath27 , \label{j_e}\ ] ] respectively .
we note that the charge and current densities obey the continuity equation @xmath28 the self - consistent vector and scalar potentials are obtained from the electromagnetic wave equations @xmath29 and @xmath30 where @xmath31 is the magnetic vacuum permeability and @xmath32 is the electric permittivity in vacuum , and @xmath33 is the neutralizing positive charge density due to the ions . for immobile , singly charged ions
, one can assume that @xmath34 , where @xmath35 is the equilibrium ion number density .
using the coulomb gauge @xmath36 , we obtain from eqs .
( [ vector_pot ] ) and ( [ scalar_pot ] ) @xmath37 and @xmath38 respectively .
taking the divergences of both sides of eq .
( [ wave2 ] ) , we have @xmath39 so that eq .
( [ wave2 ] ) can be written as @xmath40 equations ( [ kg ] ) , ( [ poisson ] ) and ( [ wave3 ] ) are our desired system that describes intense laser - plasma interactions in the quantum regime .
the non - relativistic limit is obtained from eq .
( [ kg ] ) by substituting @xmath41 , and by using the condition @xmath42 , together with the normalization of @xmath43 such that @xmath44 is the electron number density at the equilibrium . in this limit ,
( [ kg ] ) , yields the schrdinger equation @xmath45 here , and in what follows , we have used a simplified model and neglected the electron degeneracy pressure .
the latter is important in dense matters where the electron degeneracy pressure appears due to the pauli exclusion principle . for a non - relativistic plasma ,
the quantum statistical pressure has been introduced in a nonlinear schrdinger model @xcite , but this has to be investigated for relativistic quantum plasmas .
in the absence of the electromagnetic field ( viz .
@xmath46 ) , we still have electrostatic waves due to the charge separation between the electrons and ions . at short wavelengths ,
the quantum effects become important and give rise to dispersive effects in the electrostatic wave . at these wavelengths
, there is an interplay between collective electron oscillations and free electron motion . when the wavelength is comparable to the compton wavelength , the electrons become relativistic , and there are relativistic corrections to the dispersion relation for the electrostatic wave . in the derivation of the dispersion relation for relativistic electrons ,
it is convenient to first make the transformation @xmath47 , where the wave function @xmath48 obeys the wave equation @xmath49 and the electron charge density is @xmath50 we next linearize the system ( 19 ) by setting @xmath51 and @xmath52 , where @xmath53 + complex conjugate , @xmath54 , and where @xmath55 . separating different fourier modes , we obtain from ( 19 ) the dispersion relation for the electrostatic oscillations as @xmath56 , where @xmath2 is the dielectric constant and the electron susceptibility is @xmath57 } { \hbar^2(\omega^2-c^2 k^2)^2 - 4 m_e^2 c^4\omega^2 } , \label{chi_e}\ ] ] where @xmath58 is the electron plasma frequency .
we note that in the classical limit @xmath59 , we have @xmath60 , while in the non - relativistic limit @xmath61 , we have @xmath62 . after some reordering of terms , the dispersion relation can be written as @xmath63 in the classical limit @xmath59 we obtain the langmuir oscillations @xmath64 , while in the limit @xmath61 , we retain the non - relativistic result @xmath65 on the other hand , in the limit @xmath66 , eq .
( [ es_disp ] ) yields two possibilities , one of which is the langmuir oscillations at the plasma frequency , @xmath64 and the other one is oscillations with the frequency @xmath67 .
the latter corresponds to a negative energy state , which can be interpreted as positronic state .
we note that there is a non - dimensional quantum parameter @xmath68 in eq .
( [ es_disp ] ) that determines the relative importance of the quantum effect .
typical values are @xmath69 for the electron number density @xmath70 in solid density laser - plasma experiments and @xmath71 may be representable of modern laser - compressed matter experiments with @xmath72 .
this corresponds to @xmath73 and @xmath74 for @xmath69 , and @xmath75 and @xmath76 for @xmath71 , where @xmath77 is the electron skin depth .
the non - relativistic result ( [ es_disp2 ] ) is valid for electrostatic waves with wave numbers in the range @xmath78 . for @xmath79 ,
the quantum corrections to @xmath80 are different from ( [ es_disp2 ] ) and turns the wave frequency slightly lower than @xmath81 . however , this effect is negligible for small values of @xmath82 , and may be smaller than the degeneracy electron pressure effect , which is neglected here .
on the other hand , the limit @xmath83 corresponds to relativistic particles with @xmath84 , where @xmath85 is the reduced compton wavelength . for @xmath86
, we obtain the relativistic free particle dispersion relation @xmath87 where the upper sign ( - ) corresponds to the motion of a free electron and the lower sign ( + ) can be interpreted as the motion of a free positron . vs. @xmath88 ) for the electrostatic oscillations for @xmath69 ( solid curves ) , @xmath71 ( dashed curves ) , where @xmath89 . for @xmath83 ,
the particle motion turns from weakly relativistic to ultra - relativistic.,width=321 ] in fig . 1
, we have plotted the solutions of the dispersion relation ( [ es_disp ] ) for @xmath69 and @xmath71 .
both the electron plasma oscillations and the positronic states are shown .
the electron plasma oscillations have a cutoff frequency @xmath90 when @xmath91 , while the positronic states have a cutoff frequency @xmath92 , corresponding to @xmath93 at @xmath66 in fig .
1 . for the electron plasma oscillations ,
the increase in the wave frequency due to the quantum effect becomes noticeable approximately where @xmath94 , or @xmath95 where @xmath96 is the bohr radius .
this corresponds to a wavelength of @xmath97 for @xmath69 and @xmath98 for @xmath71 .
the positronic states are associated with , for example , the zitterbewegung effect @xcite , in which the interference between the positive and negative energy states are predicted to give oscillations on compton wavelength scales in space .
the zitterbewegung effect is still debated and has not yet been observed in experiments .
it is well - known @xcite that a large amplitude electromagnetic wave propagating in a classical plasma changes the dispersive properties of the plasma due to the relativistic mass increase of the electrons .
we show here that the same effect occurs in our klein - gordon - maxwell system .
we consider for simplicity the propagation of a right - hand circularly polarized electromagnetic ( cpem ) wave of the form @xmath99 $ ] , where @xmath100 is the wave frequency and @xmath101 the wavenumber . due to the circular polarization ,
the oscillatory parts in the nonlinear term proportional to @xmath102 in the klein - gordon equation vanish . assuming that @xmath12 depends only on time and not on space , and that @xmath103 , we obtain from eq . ( [ kg ] ) @xmath104 where @xmath105 can be interpreted as the relativistic gamma factor due to the electron mass increase in the cpem wave field .
equation ( [ kg_time ] ) has the solution @xmath106 where the constant @xmath107 is determined by assuming the constant density @xmath108 in eq .
( [ rho_e ] ) . inserting ( [ psi ] ) into ( [ rho_e ] ) with @xmath108
, we obtain @xmath109 on the other hand , inserting ( [ psi ] ) into ( [ j_e ] ) we have @xmath110 which can be inserted into ( [ wave2 ] ) to obtain @xmath111 equation ( [ plane_wave ] ) admits the nonlinear dispersion relation @xmath112 which predicts a relativistic downshift of the cpem wave frequency due to the relativistic electron mass increase in the cpem wave field . since the effective plasma frequency is decreased by a factor @xmath113 , the model predicts the well - known self - induced transparency where the cpem wave can propagate at frequencies below the electron plasma frequency .
this is identical to the case of classical plasmas @xcite .
we now consider the instability of an intense cpem wave in the quantum regime . in the presence of intense electromagnetic waves
, we have the relativistic down - shift in the wave frequency given in ( [ omega0 ] ) , as well as the possibility of exciting electrostatic oscillations via the parametric instabilities . as an example
, we will here consider stimulated raman scattering instability , in which an intense electromagnetic wave decays into a daughter em wave and an electron plasma wave .
the two - plasmon decay instability , in which the cpem wave decays into two electrostatic waves , will be treated elsewhere .
it is convenient to first introduce the transformation @xmath114 , where @xmath105 and @xmath115 is the amplitude of the em carrier wave @xmath116 .
the wavefunction @xmath48 obeys the modified klein - gordon equation @xmath117 and the electron charge density is given by @xmath118 now , we linearize our system by introducing @xmath119 ( where @xmath120 is assumed to be constant ) , @xmath121 , and @xmath122 . using @xmath34 into eq . ( [ poisson ] ) , we note that the equilibrium quasi - neutrality requires that @xmath120 is normalized such that @xmath123 . using that @xmath116 fulfills the plane wave equation ( [ plane_wave ] ) , the linearized kge ( [ kg5 ] ) , poisson s equation ( [ poisson ] ) and the em wave equation ( [ wave3 ] ) then become @xmath124 @xmath125 and @xmath126\ } , \end{split}\ ] ] respectively .
we note that the term proportional to @xmath127 in eq .
( [ kg_linear ] ) gives rise to the two - plasmon decay , which we , however , do not consider here .
we now introduce the fourier representations @xmath128 , @xmath129 + c.c .
, @xmath130 + c.c . , and @xmath131 $ ] + c.c . , where we introduced @xmath132 and @xmath133 , and c.c
. stands for complex conjugate . in one of the steps , we take the scalar product of both sides of the em wave equation by @xmath134 and use the fact that @xmath135 = ( { \bf k}_\pm\times\widehat{\bf a}_0)\cdot ( \widehat{\bf a}_0^\ast \times{\bf k}_\pm)=-|{\bf k}_{\pm}\times \widehat { \bf a}_0|^2 $ ] . separating different fourier modes and eliminating the fourier coefficients , we find the nonlinear dispersion relation @xmath136 } \bigg[\frac{c^2 e^2|{\bf k}_+\times \widehat{\bf a}_0|^2}{k_{+}^2 d_a(\omega_{+},{\bf k}_{+ } ) } + \frac{c^2 e^2|{\bf k}_{-}\times \widehat{\bf a}_0|^2}{k_{-}^2 d_a(\omega_{-},{\bf k}_{-})}\bigg ] , \end{split } \label{nonlin_disp2}\ ] ] where the electromagnetic sidebands are governed by @xmath137 . the electric susceptibility in the presence of the laser field
is given by @xmath138 } { \gamma_a[\hbar^2(\omega^2-c^2k^2)^2 - 4\gamma_a^2 m_e^2 c^4 \omega^2]}.\ ] ] after reordering of terms , the nonlinear dispersion relation ( [ nonlin_disp2 ] ) can be written as @xmath139=0 , \end{split } \label{nonlin_disp}\ ] ] where the electron plasma oscillations in the presence of the laser field are represented by @xmath140 we note that @xmath141 gives the dispersion relation for pure electrostatic oscillations in the presence of a large amplitude electromagnetic wave . in the classical limit @xmath142 , we have @xmath143 , and the nonlinear dispersion relation takes the form @xmath144 , \end{split}\ ] ] which can be written in a more familiar form as @xmath145=0 , \end{split}\ ] ] with @xmath146 .
these results can be compared with , for example , the dispersion relations obtained in refs .
@xcite for the relativistic case and in @xcite for the non - relativistic case . to proceed with the numerical evaluation of the nonlinear dispersion relation , we choose a coordinate system such that the cpem takes the form @xmath147 and @xmath148 , and , without loss of generality , we choose @xmath149
then , we have @xmath150 , @xmath151 , @xmath152|\widehat{a}_0|^2 $ ] , and @xmath153
. we also use that the carrier wave @xmath116 obeys the nonlinear dispersion relation @xmath154 . vs. @xmath155 and @xmath156 ) for stimulated raman scattering in the presence of a large amplitude cpem wave , for the amplitudes @xmath157 , @xmath158 , and @xmath159 ( left to right panels ) where @xmath160 for @xmath69 ( top panels ) and @xmath71 ( bottom panels ) . , width=321
] we now assume that the wave frequency is complex valued , @xmath161 , where @xmath162 is the real frequency and @xmath163 the growth rate , and solve numerically the dispersion relation ( [ nonlin_disp ] ) for @xmath164 . in fig .
2 , we have plotted the growth rate for stimulated raman scattering instability as a function of the wavenumbers @xmath155 and @xmath156 , for a few values of @xmath165 and @xmath166 . for all cases in fig .
2 , we used @xmath167 , which corresponds to a wavelength of @xmath168 for @xmath69 and to @xmath169 for @xmath71 , which is in the x - ray regime .
we observe that for @xmath69 , there is a broad spectrum of unstable waves , in particular for @xmath158 and @xmath170 . for @xmath71 ,
we observe a reduction in the spectrum of unstable waves and the growth rate ( relative to the electron plasma frequency ) is slightly reduced .
this is due to the fact that the wavelength of the unstable electrostatic oscillation approaches the critical wavelength , where quantum dispersive effects become important compared to the plasma frequency oscillations . for @xmath69 ,
this wavelength is @xmath171 , corresponding to a critical wavenumber @xmath172 , and for @xmath71 , we have @xmath173 , corresponding to @xmath174 . hence , for @xmath71 we have @xmath175 , which leads to the reduction of the growth rate due to the quantum dispersion effect . furthermore , it should be mentioned that we do not find raman - type instabilities involving the positronic states in fig .
this is consistent from the point of view of the conservation of charges , since the production of positrons would violate the conservation of electric charges . vs. @xmath155 and @xmath156 ) for the modulational instability in the presence of a large amplitude cpem wave , for the amplitudes @xmath157 ( left panel ) , @xmath158 ( middle panel ) and @xmath159 ( right panel ) where @xmath160 .
we used @xmath71 for all cases . , width=321 ] in addition to stimulated raman scattering instabilities , we have also the modulational instability that dominates for the pump frequencies @xmath176 , and the corresponding wavenumbers @xmath177 .
the modulational instability usually occurs for small modulation wavenumbers , and saturates nonlinearly by the formation of relatively small localized structures / solitary waves . in the past ,
such nonlinear structures have been studied for classical plasmas in 1d @xcite and 3d @xcite .
we have investigated the modulational instability for the cpem dipole field with @xmath178 , and have plotted the results in fig . 3 for different amplitudes @xmath157 , @xmath158 and @xmath159 .
we find that the growth rate is relatively insensitive to the quantum parameter @xmath82 .
we have used @xmath71 in fig . 3 , but @xmath179 gives almost identical results .
this is understandable since the modulational instability takes place on relatively large scales and the quantum effect is thus negligible .
however , we will investigate the quantum effect on the relatively small scale nonlinear structures below .
here we illustrate the existence of large amplitude localized cpem wave excitations at the quantum scale in our system .
we restrict our investigation to one - space dimension , which has also been studied for classical plasmas @xcite . far away from the local excitation
, one can assume that the dynamics of the plasma is non - relativistic . to shorten the algebraic steps , it is convenient first to introduce a new wave - function @xmath180 and the potential @xmath181 via the transformations @xmath182 and @xmath183 , and which satisfy the kge @xmath184 where @xmath185 . in this gauge ,
the wave function @xmath43 is non - oscillatory in time , and the new potential takes the value @xmath186 , far away from the solitary wave where the plasma is at rest .
the electron charge density is expressed as @xmath187 .
\label{e - charge}\ ] ] we now study quasi - steady state structures propagating with a constant speed @xmath188 , so that @xmath189 and @xmath190 , where @xmath191 and @xmath192 .
the cpem wave vector potential is of the form @xmath193 $ ] .
it is convenient to introduce the eikonal representation @xmath194 $ ] , where @xmath195 and @xmath196 are real - valued .
then , the kge ( [ kg3 ] ) takes the form @xmath197
-i \hbar e v_0 \frac{d\phi } { d\xi } p \\ & -2i\hbar e \phi v_0 \left(\frac{dp}{d\xi}+
i p\frac{d\theta}{d\xi}\right ) + ( e^2\phi^2-m_e^2 c^4\gamma_a^2)p=0 .
\end{split } \label{kg4}\ ] ] setting the imaginary part of eq .
( [ kg4 ] ) to zero , we obtain @xmath198 where @xmath199 the solution of eq .
( [ pu ] ) is @xmath200 constant . using the boundary conditions
@xmath201 , @xmath103 ( hence @xmath202 ) and @xmath203 at @xmath204 , we have @xmath205 .
hence , we obtain @xmath206 where we have denoted @xmath207 the electron charge density ( [ e - charge ] ) now takes the form @xmath208 which , by using eq .
( [ theta ] ) , can be written as @xmath209 and hence poisson s equation ( [ poisson ] ) , with @xmath34 , becomes @xmath210 on the other hand , by setting the real part of eq .
( [ kg4 ] ) to zero , we have @xmath211 + 2\hbar e\phi v_0 p\frac{d\theta}{d\xi}+ ( e^2 \phi^2-m_e^2 c^4\gamma_a^2)p=0 , \end{split}\ ] ] which , by using eq .
( [ theta ] ) , can be written as @xmath212p=0 .
\label{p}\ ] ] finally , inserting the ansatz @xmath193 $ ] for the vector potential , together with the current @xmath213 into eq . ( [ wave3 ] ) , we obtain the em wave equation @xmath214a=0 , \label{a}\ ] ] where we used @xmath215 , and where @xmath216 is a nonlinear eigenvalue of the system that determines the wave frequency @xmath100 .
the coupled system ( [ poisson2 ] ) , ( [ p ] ) and ( [ a ] ) describes the profile of electromagnetic solitary waves in a quantum plasma .
it has the conserved quantity @xmath217 , where @xmath218 the conserved quantity @xmath219 can be used to check that the numerical scheme used to solve the nonlinear system ( [ poisson2 ] ) , ( [ p ] ) and ( [ a ] ) produces correct results .
( left panels ) and @xmath220 ( right panels ) for standing solitary waves ( @xmath221 ) , and for @xmath222 .
, width=321 ] , @xmath223 , and @xmath222 . , width=170 ] in fig .
[ fig : comparison3 ] , we have compared the present model with our previous results in ref .
@xcite where we used a simplified model to describe the nonlinear interaction interaction of intense cpem wave with a quantum plasma .
we used the same parameters as in fig . 2 , of ref .
@xcite to produce the profiles of the cpem wave potential , the electron charge density and the electrostatic potential .
we observe that the present results are almost identical to our previous work @xcite . for our sets of parameters ,
the quantum effect on the profiles of the solitary waves are small , and there is only a slight difference in the profiles of the electron density for the two values @xmath220 and @xmath71 . for standing solitary waves , such as the ones in fig .
[ fig : comparison3 ] , the solutions are localized with exponentially decaying tails . by linearizing the system ( [ poisson2 ] ) , ( [ p ] ) and ( [ a ] )
one can show that far away from the soliton , @xmath224 decays as @xmath225 , while @xmath195 and @xmath19 are proportional to @xmath226 , where @xmath88 is found from the dispersion relation @xmath227 for @xmath221 ( and @xmath228 ) , we see immediately that there exist only complex - valued @xmath88 , which means that the quasi - stationary wave solutions decay exponentially far away from the solitary wave .
however , in the classical limit @xmath59 , we instead have the plasma wake oscillations given by real valued @xmath229 .
hence , an electromagnetic pulse will create an oscillatory wake that extends far away from the em pulse . in one - space dimension
, there also exist special classes of propagating localized em envelope solutions @xcite .
in addition to the wake oscillations , we also have quantum oscillations in quantum plasmas . in fig .
5 , we show an example of a slowly moving envelope soliton , where small - scale oscillations in the charge density are clearly visible .
we note that the cold fluid results can be retained in the classical limit @xmath59 . then , eq . ( [ p ] ) can be written as @xmath230 by setting @xmath231 , where @xmath232 is the electron number density , in eq .
( [ rho_e2 ] ) and solving for @xmath195 , we obtain @xmath233\frac{m_e c^2}{e\phi},\ ] ] which can be inserted into ( [ p2 ] ) to obtain @xmath234/\bigg [ \frac{n_e^2}{n_0 ^ 2}-\frac{v_0 ^ 2}{c^2}\bigg(\frac{n_e}{n_0}-1\bigg)^2 \bigg]^{1/2 } = \frac{e\phi}{\gamma_a m_e c^2 } , \label{fluid}\ ] ] which relates @xmath232 to @xmath181 and @xmath235 at a given speed @xmath188 . the relation ( [ fluid ] )
can also be obtained from the cold electron fluid model @xcite and hence confirms the classical limit of the quantum model used here . if , furthermore , we assume standing waves such that @xmath221 , then we have from ( [ fluid ] ) @xmath236 solving for @xmath181 and inserting the result into poisson s equation ( [ poisson2 ] ) , we have @xmath237 where @xmath77 is the electron skin depth .
finally , solving for @xmath238 and inserting the result into eq .
( [ a ] ) , we obtain @xmath239 where we have used @xmath221 and , therefore , @xmath178 .
equation ( [ a2 ] ) is identical to the model of marburger and tooper @xcite for the nonlinear optical standing wave in a classical cold fluid electron plasma .
the relativistic mass increase is reflected by the ratio @xmath240 in the right - hand side of eq .
( [ a2 ] ) .
the nonlinear electron density fluctuations , which are reflected in the term proportional to @xmath241 in the right - hand side of eq .
( [ a2 ] ) can often be neglected in the weakly relativistic case @xcite .
we note that our previous model @xcite can be recovered in the weakly relativistic limit , in the following manner . assuming that , to first order , we have a balance between the ponderomotive and electrostatic pressure so that @xmath242 and that @xmath243 , and @xmath244 .
accordingly , we have @xmath245 poisson s equation @xmath246 and the cpem wave equation @xmath247 we now make a simple change of variables @xmath248 .
then , we have , by neglecting terms containing @xmath249 , the model @xcite @xmath250 @xmath251 and @xmath252 where the relativistic mass increase appears explicitly in the cpem wave equation .
, and initially a dipole field @xmath157 with @xmath178.,width=321 ] in order to study the dynamics of the nonlinear interaction between intense cpem waves and a quantum plasma , we have carried out numerical simulations of the klein - gordon - maxwell system of equations .
we have here restricted our study to one - space dimension , along the @xmath253 direction in space , and written our governing equations in the form @xmath254 @xmath255 and @xmath256 we used a periodic simulation box in space , of length @xmath257 and used of the order @xmath258 grid points to resolve the solution in space .
it is important to resolve the relatively long electron skin depth scale as well as the shorter length scale associated with accelerated electrons with the momentum @xmath259 and the associated wavelength @xmath260 . since we need at least two grid - points per wavelength to represent the solution , the required grid size is @xmath261 , which can be written @xmath262 .
for example , to represent the wave function of relativistic electrons with the momentum @xmath263 , we need a spatial grid with @xmath264 to represent the wave function for @xmath71 .
the solution was advanced in time with the standard 4th - order runge - kutta scheme , using a timestep of order @xmath265 . .
top to bottom panels show a ) the electromagnetic vector potential of the laser pulse ( the arrows show the propagation directions of the pulses ) , b ) the electron charge density , c ) the electrostatic potential , and d ) the distribution of electrons in phase space in a 10-logarithmic color scale .
parameters are @xmath71 , amplitude @xmath157 and wavenumber @xmath266 .
the laser pulses excite large amplitude oscillatory potential wakes behind them , as they penetrate the plasma slab.,width=321 ] at time @xmath267 .
groups of electrons are accelerated to ultra - relativistic speeds by the large amplitude electrostatic wake field.,width=321 ] we first study the growth and nonlinear saturation of the modulational instability , which is relevant for dense matters where the plasma is overdense or close to overdense . as initial conditions
, we used a cpem pump wave of the form @xmath268 $ ] with @xmath178 and @xmath269 . a small amplitude noise ( random numbers )
was added to @xmath224 in order to seed any instability in the system . as initial conditions for the wave function
, we used @xmath270 and @xmath271 , corresponding to a pure electronic state at equilibrium .
the initially homogeneous electron density was set to @xmath272 , corresponding to @xmath71 [ cf .
( [ h ] ) ] .
in this situation , the electromagnetic wave is unstable due to the modulational instability , which has instability for small wavenumbers , as shown in fig .
3 . in the nonlinear stage ,
the em waves self - focus into localized wavepackets similar to the ones in fig .
4 . figure [ fig : modulational ] depicts the late stage of the modulational instability .
the collapse of the cpem wave packet leads to relativistically strong ponderomotive potentials that accelerate the electrons to relativistic speeds .
the relativistic electrons are associated with small - scale spatial oscillations in the wave function , where the wavelength is comparable to or even smaller than the compton wavelength .
we see in fig .
[ fig : modulational ] that the cpem wave envelope has been focused into localized wavepackets , associated with depletions in the electron density and positive localized electrostatic potentials . in order to study the distribution of electrons both in space and momentum space ,
we have performed a fourier transform of the wavefunction @xmath12 using a moving window technique ( using a hann window ) in space . the width of the window has been tuned so that it provides a good resolution both in space and in momentum space . the resulting spatial spectrogram gives a representation of the distribution of the electrons both in space and in momentum space ; see fig .
[ fig : modulational](d ) , where the color indicates the density of electrons in phase space . in fig .
6(d ) the horizontal axis shows the spatial dependence and the vertical axis shows the momentum dependence via the relation @xmath259 between the momentum @xmath273 and the wavenumber @xmath274 . in this figure , it is clear that in the collapse stage of the solitary waves , bunches of electrons are accelerated to relativistic speeds and form self - trapped , bernstein - greene - kruskal ( bgk)-like modes that propagate away from the collapsed electromagnetic wavepackets .
next , we investigate a scenario of the short em pulse propagation and the wake - field generation in a quantum plasma .
this concept is traditionally used for the electron acceleration in classical plasmas @xcite .
the numerical results are displayed in figs .
[ fig : colliding1 ] and [ fig : colliding2 ] . here , two atto - second pulses are injected from each side of a plasma slab and are allowed to collide at the center of the slab . as initial conditions , we used a cpem pump wave of the form @xmath275 $ ] with @xmath266 and envelopes of the form @xmath276 propagating into the plasma slab .
the plasma slab is initially centered between @xmath277 with equal electrons with the number densities @xmath35 , where the electron wave function was set to @xmath278 and @xmath279 . after a time @xmath280 , we see in fig . [
fig : colliding1](b ) and ( c ) that the large amplitude cpem pulses excite plasma wake oscillations associated with large - amplitude positive potentials , and with an approximate wavelength of @xmath281 , corresponding to a leading wavenumber of @xmath282 .
the positive potentials of the plasma wake oscillations are starting to capture populations of the electrons at edges of the plasma slab , at @xmath283 . a high - frequency diffraction pattern is formed in the electron density , as faster electrons overtake slower electrons .
later , at @xmath267 , the two laser pulses have collided and passed through each other .
the trapped electrons have been further accelerated to ultra - relativistic speeds , as seen in fig .
[ fig : colliding2](d ) at @xmath284 . , where the fastest electrons have reached a momentum of @xmath285
in this paper , we have developed a relativistic model for the interaction between intense electromagnetic waves and a quantum plasma .
our nonlinear model is based on the coupled klein - gordon and maxwell equations for the relativistic electron momentum and the electromagnetic fields . in our fully relativistic model ,
the electron current and charge densities are calculated self - consistently from the kge , and they enter as sources for the nonlinear em and electrostatic waves in the maxwell equation .
the kg - maxwell system of equations has been used to derive the linear dispersion relation for the electrostatic and electromagnetic waves , as well as for investigating stimulated raman scattering and modulational instabilities in the presence of relativistically intense cpem waves . in the linear regime ,
the general dispersion relation for the electrostatic waves exhibits the quantum effect associated with the overlapping wave function . at long - wave - lengths
, we have the dispersive langmuir waves with frequencies close to the electron plasma frequency , while at shorter - wavelengths , we have the oscillation frequency of free electrons . at wavelengths comparable to or larger than the compton wavelength ,
the electron motion is fully relativistic . in the nonlinear regime
, we have demonstrated the existence of fully relativistic stimulated raman scattering and modulational instabilities .
while the raman amplification is of much interest for generating a coherent electromagnetic radiation , the modulational instability gives rise to the localization and collapse of the cpem waves into localized solitary em wavepackets .
indeed , numerical simulations of the coupled kg - maxwell equations reveal the collapse and acceleration of the electrons in the nonlinear stage of the modulational instability , as well as the possibility of wake - field acceleration of the electrons to relativistic speeds by short laser pulses at nanometer scales . in conclusion
, we stress that the present investigation of nonlinear effects dealing with intense em wave interactions with quantum plasmas is relevant for the compression of x - ray free - electron laser pulses to attosecond duration @xcite , as well as to the understanding of localized intense x - ray and @xmath1-ray bursts that emanate from compact astrophysical objects @xcite
. 99 e. hand , nature ( london ) * 461 * , 708 ( 2009 ) .
s. h. glenzer _
et al . _ ,
lett . * 98 * , 065002 ( 2007 ) ; p. neumayer _
_ , _ ibid .
_ * 105 * , 075003 ( 2010 ) ; s. h. glenzer and r. redmer , rev .
phys . * 81 * , 1625 ( 2009 ) .
a. v. andreev , jetp lett .
* 72 * , 238 ( 2000 ) .
g. mourou _
et al . _ , rev .
phys . * 78 * , 309 ( 2006 ) .
m. marklund and p. k. shukla , rev .
phys . * 78 * , 591 ( 2006 ) .
j. f. drake , p. k. kaw , y. c. lee , g. schmidt , c. s. liu , and m. n. rosenbluth , phys .
fluids * 17 * , 778 ( 1974 ) .
r. p. sharma and p. k. shukla , phys .
fluids * 26 * , 87 ( 1983 ) .
g. murtaza and p. k. shukla , j. plasma phys , * 31 * , 423 ( 1984 ) .
et al . _ , phys .
138 * , 1 ( 1986 ) . c. j. mckinstrie and r. bingham , phys
. fluids b * 1 * , 230 ( 1989 ) .
l. n. tsintsadze , sov . j. plasma phys .
* 17 * , 872 ( 1991 ) . c. j. mckinstrie and r. bingham , phys
. fluids b * 4 * , 2626 ( 1992 ) . a. s. sakharov and v. i. kirsanov , phys .
e * 49 * , 3274 ( 1994 ) .
s. gurin , g. laval , p. mora , j. c. adam , a. hron , and a. bendib , phys .
plasmas * 2 * , 2807 ( 1995 ) .
j. c. adam , a. hron , g. laval , and p. mora , phys .
lett . * 84 * , 3598 ( 2000 ) .
b. quesnel , p. mora , j. c. adam , a. hron , and g. laval , phys .
plasmas * 4 * , 3358 ( 1997 ) .
l. stenflo , phys .
* 14 * , 320 ( 1976 ) . l. stenflo , phys .
* 21 * , 831 ( 1980 ) .
l. stenflo and h. wilhelmsson , phys .
rev . a * 24 * , 1115 ( 1981 )
. v. m. malkin , n. j. fisch , and j. s. wurtele , phys .
e * 75 * , 026404 ( 2007 ) .
a. serbeto , j. t. mendona , k. h. tui _
et al . _ , phys .
plasmas * 15 * , 013110 ( 2008 ) .
a. serbeto , l. f. monteiro , k. h. tsui , and j. t. mendona , plasma phys .
fusion * 51 * 124024 ( 2009 ) .
n. piovella , m. m. cola , l. volpe , a. schiavi , r. bonifacio , phys .
lett . * 100 * , 044801 ( 2008 ) .
g. chabrier _
_ , j. phys .
: condens .
matter * 14 * , 9133 ( 2002 ) ; j. phys .
a : math . gen . * 39 * , 4411 ( 2006 ) .
m. j. coe _
_ , nature ( london ) * 272 * , 37 ( 1978 ) ; d. k. galloway and j. l. sokoloski , astrophys . j. * 613 * , l61 ( 2004 ) . k. hurley _
et al . _ , nature ( london ) * 434 * , 1098 ( 2005 ) ; a. k. harding and d. lai , rep .
. phys . * 69 * 2631 ( 2006 ) .
g. manfredi and f. haas , phys .
b * 64 * , 075316 ( 2001 ) ; p. k. shukla and b. eliasson , phys . rev
. lett . * 96 * , 245001 ( 2006 ) ; phys .
usp . * 53 * , 51 ( 2010 ) .
m. g. fuda and e. furlani , am .
* 50 * , 545 ( 1982 ) .
r. gerritsma _ et al .
_ , nature ( london ) * 463 * , 68 ( 2010 ) .
a. i. akhiezer and r. v. polovin , sov .
jetp * 3 * , 696 ( 1956 )
[ zh . eksp .
. fiz . * 30 * , 915 ( 1956 ) ] ; p. kaw and j. dawson , phys .
fluids * 13 * , 472 ( 1970 ) ; c. max and f. perkins , phys .
* 27 * , 1342 ( 1971 ) .
j. h. marburger and r. f. tooper , phys .
* 35 * , 1001 ( 1975 ) .
j. i. gersten and n. tzoar , phys .
* 35 * , 934 ( 1975 ) .
p. k. shukla and b. eliasson , phys .
lett . * 99 * , 096401 ( 2007 ) .
p. k. kaw , a. sen , and t. katsouleas , phys .
lett . * 68 * , 3172 ( 1992 ) . v. saxena , a. das , a. sen , and p. kaw , phys .
plasmas * 13 * , 032309 ( 2006 ) .
k. shukla , n. n. rao , m. y. yu , and n. l. tsintsadze , phys . rep .
* 138 * , 1 ( 1986 ) .
t. tajima and j. m. dawson , phys .
* 43 * , 267 ( 1979 ) .
r. bingham _
et al . _ , plasma phys .
fusion * 46 * , r1 ( 2004 ) .
_ , science * 292 * , 1689 ( 2001 ) .
r. m. g. m. trines , f. fiuza , r. bingham _ et al .
_ , nature phys .
* 6 * , 1793 doi:10.1038/nphys1793 ( 2010 ) . | we investigate the nonlinear interaction between a relativistically strong laser beam and a plasma in the quantum regime .
the collective behavior of the electrons is modeled by a klein - gordon equation , which is nonlinearly coupled with the electromagnetic wave through the maxwell and poisson equations .
this allows us to study the nonlinear interaction between arbitrarily large amplitude electromagnetic waves and a quantum plasma .
we have used our system of nonlinear equations to study theoretically the parametric instabilities involving stimulated raman scattering and modulational instabilities .
a model for quasi - steady state propagating electromagnetic wavepackets is also derived , and which shows the possibility of localized solitary structures in the quantum plasma .
numerical simulations demonstrate the collapse and acceleration of the electrons in the nonlinear stage of the modulational instability , as well as the possibility of wake - field acceleration of the electrons to relativistic speeds by short laser pulses at nanometer length scales .
the study has importance for the nonlinear interaction between a super - intense x - ray laser light and a solid - density plasma , where quantum effects are important . |
consider a process which is repeated @xmath1 times and each repetition has @xmath0 possible outcomes . for concreteness we may think of assigning @xmath1 balls to @xmath0 labelled boxes , where each box can hold any number of balls .
the first ball can go into any box , the second ball can go into any box , ... , and the @xmath1th ball can go into any box .
each assignment or allocation is thus a sequence of @xmath1 box labels and results in some number @xmath3 of balls in box 1 , @xmath4 in box 2 , etc . , where the @xmath5 are @xmath6 and sum to @xmath1
. there are @xmath7 possible assignments in all , and many of them can lead to the same count vector @xmath8 .
we refer to these assignments as the _ realizations _ of the count vector .
the arrangement of @xmath1 balls into @xmath0 boxes can represent the construction of any object consisting of @xmath0 distinguishable parts out of @xmath1 identical units .
so if the balls represent pixels of an image , the attributes of color and ( suitably discretized ) intensity are ascribed to the boxes to which the pixels are assigned
. then the count vector is thought of as a 2-dimensional matrix with rows labelled by intensity and columns by color .
other examples are people categorized by age , height , and weight , vehicles classified by weight , size , and fuel economy , packets in a communications network with attributes of origin , destination , size , and timestamp , etc .
the object can even be a ( discrete ) probability distribution .
when the process simply represents the classification of @xmath1 units by @xmath0 discrete or discretized attibutes , it is known as a ( multi - dimensional ) contingency table . now consider imposing constraints @xmath9 on the allowable assignments , expressed as a set of _ linear _ relations on the elements of the _ frequency _ vector @xmath10 corresponding to the counts @xmath11
e.g. @xmath12 , @xmath13 , etc .
as @xmath1 grows , the frequency vectors of more and more of the assignments that satisfy the constrtaints will have _ entropy _ closer and closer to that of a particular @xmath0-vector @xmath14 , the vector of _ maximum entropy _
@xmath15 subject to the constraints @xmath9 .
( we denote this vector by @xmath14 , as opposed to @xmath16 , to emphasize that its entries are , in general , not rational . )
this result is known , more or less , in many forms : the original is e.t .
jaynes s `` entropy concentration theorem '' @xcite , @xcite , in the information theory literature it is the `` conditional limit theorem '' @xcite , and in computer science there is `` strong entropy concentration '' @xcite , @xcite .
all these results involve limits or asymptotics in one way or another , i.e. in the statement [ cols= " < , < " , ] to interpret the first line of table [ tab : q ] , suppose we choose @xmath17 .
we then find @xmath18by corollary [ cor : pd ] , this rational approximation to the maxent p.d .
( [ eq : phisq ] ) has at least @xmath19 times more realizations than the entire set of p.d.s which satisfy the constraints to accuracy @xmath20 but differ from the p.d .
( @xmath21 ) by more than 0.01 in @xmath22 norm .
the phenomenon of entropy concentration appears when a large number of units is allocated to containers subject to constraints that are linear functions of the numbers of units in each container : most allocations will result in frequency ( normalized count ) vectors with entropy close to that of the vector of maximum entropy that satisfies the constraints .
asymptotic proofs of this phenomenon are known , beginning with the work of e. t. jaynes , but here we presented a formulation entirely devoid of probabilities and provided explicit bounds on how large the number of units must be for concentration to any desired degree to occur .
our formulation also deals with the fact that constraints can not be satisfied exactly by rational frequencies , but only to some prescribed tolerances , and also eliminates ( as opposed to `` resolves '' ) the well - known issue of expectations vs. measurements in constraints .
in addition , we established a perhaps more useful version of the concentration result , in terms of deviation from the maximum entropy vector , instead of the usual maximum entropy value , as well as results that pertain to the maximum entropy vector itself and not to a whole set of vectors around it . because of its conceptual simplicity and minimality of assumptions ,
entropy concentration is a powerful justification of the widely - used discrete maxent method ( the other being axiomatic formulations ) , and we believe that the explicit , non - asymptotic bounds strengthen it considerably .
all of our results were illustrated with detailed numerical examples .
thanks to david applegate for discussing with me what can be done , howard karloff for telling me what ca nt be done , steve korotky for our many discussions on entropy and networks , and neil sloane for many informative discussions , answering many questions , in particular about latttice points , and for his careful reading of the paper .
rounding ensures that @xmath23 . from the explanation after definition [ def : fstar ] ,
the adjustment of @xmath24to @xmath25 ensures @xmath26 , which establishes the @xmath27 claim . in more detail ,
the 0 elements of @xmath25 coincide with those of @xmath14 , and this adjustment causes at most @xmath28 of the non - zero elements of @xmath25 to differ from the corresponding elements of @xmath29 by @xmath30 , so @xmath31 .
hence @xmath32 , which establishes the claim for the @xmath22 norm .
beginning with the equality constraints ( [ eq : constr2e ] ) , note that @xmath33 .
set @xmath34 .
then we have @xmath35 , with @xmath36 .
now @xmath37 , where @xmath38 denotes the matrix infinity norm ( also known as the `` maximum row sum '' norm ) . the inequality holds because the vector norm @xmath39 is _ compatible _ with the ( rectangular ) matrix norm @xmath40 ( see @xcite , 5.7 ) .
so if we make @xmath41 , we will have @xmath42 , as required by ( [ eq : constr2e ] ) . the inequality constraints ( [ eq : constr2i ] ) are handled in exacty the same way .
theorem 16.3.2 of @xcite is a similar result , but in terms of the @xmath22 norm .
the function @xmath45 , @xmath46 $ ] , is concave and has a maximum at @xmath47 .
let @xmath48 be @xmath49 , and consider the difference of the values of @xmath50 at two points that are @xmath51 apart : @xmath52 . since @xmath53 always , the maximum of @xmath54 occurs at @xmath55 and equals @xmath56 .
so if @xmath57 , @xmath58 ( this is tighter than what we would get by simply applying the defining inequality of concavity to @xmath59 . ) we have now shown that if @xmath60 , then @xmath61 .
the result of the proposition follows . using proposition [ prop : norment0 ] , we want to find a @xmath62 s.t .
@xmath63 for all @xmath64 . setting @xmath65 and @xmath66 ,
we want to find a @xmath67 s.t . for all @xmath68 , @xmath69 .
we claim that this inequality , where @xmath70 and @xmath71 is expected to be @xmath72 , is satisfied by @xmath73 , for any @xmath74 .
indeed , @xmath75 which is possible for any @xmath74 if @xmath76 is large enough . with @xmath77 ,
this condition is @xmath78 . but
this holds for @xmath79 , a very mild requirement .
finally , the function @xmath80 is increasing for @xmath81 , so the l.h.s . of ( [ eq : haty ] ) will hold for all @xmath68 as desired .
we have now shown that @xmath82 will hold if @xmath59 is s.t . @xmath83 where the `` @xmath84 '' can be tightened to @xmath85 .
begin with @xmath86 and use the fact that @xmath87 which is defined for all @xmath88 by @xmath89 .
then we find @xmath90 finally , for the upper bound in prop .
[ prop : s ] , the sum of the last two terms is maximized when @xmath91 . for the lower bound
, it is minimized when @xmath92 .
we begin by observing that the sum over @xmath93 is bounded above by the sum over all of @xmath94 and then use the bound of proposition [ prop : s ] on @xmath95 to find @xmath96 to evaluate the last sum , let @xmath97 be the subset of @xmath94 consisting of vectors with @xmath98 non - zero elements . since the @xmath97 form a partition of @xmath94 , @xmath99 where the @xmath100 comes from the fact that as pointed out in proposition [ prop : s ] , @xmath95 depends only on the non - zero elements and not on their positions .
thus @xmath101 we now need an auxiliary result on the inner sum in ( [ eq : ss ] ) : proposition [ prop : sumint ] is proved separately later . using this result in ( [ eq : ss ] ) , @xmath104 for the first inequality we used @xmath105 and for the second we assumed that @xmath106 and @xmath107 . combining the above with ( [ eq : nrb1 ] ) , and again assuming @xmath106 we obtain the result of the lemma . ignoring the rational requirement for the moment and denoting @xmath59 by @xmath108 ,
only @xmath109 are independent , so our set is the subset of @xmath110 belonging to @xmath111 we will construct inside this set an @xmath112-dimensional rectangular parallelepiped @xmath113 whose intersection with @xmath94 is easy to count . to construct @xmath113
we will determine its two extreme points , the one with the largest coordinates , @xmath114 , and the one with the smallest , @xmath115 , where @xmath116 .
if @xmath114 satisfies ( [ eq : krs ] ) , then @xmath117 the 4th inequality is true , and the 2nd implies the first .
the 2nd and 3d inequalities are satisfied if @xmath118 , and @xmath119 will be maximized if @xmath120 since @xmath14 has @xmath121 non - zero elements , we can assume w.l.o.g .
that @xmath122 . similarly , for the other extreme point @xmath115 we must have @xmath123 and @xmath124 . if some @xmath125 are 0 the corresponding @xmath126 are 0 , and w.l.o.g .
we can take the non - zero @xmath126 to be @xmath127 .
then the @xmath126 that satisfy the inequalities and maximize the product @xmath128 are @xmath129 but this needs @xmath130 for all the non - zero @xmath125 , which we have assumed .
thus from ( [ eq : py ] ) and ( [ eq : pz ] ) @xmath131 sides of @xmath113 have length @xmath132 , and the other @xmath133 sides have length @xmath134 . again
. we can take @xmath135 to be @xmath136 , the largest element of @xmath14 .
so if we assume that @xmath137 , @xmath113 has @xmath138 sides of length @xmath139 and @xmath133 sides of length @xmath140 .
now a @xmath141-dimensional parallelepiped with sides of lengths @xmath142 , irrespective of its location in @xmath143 , contains at least @xmath144 lattice points , i.e. points in @xmath145 .
( this can be established by induction on @xmath141 . for @xmath146
it says that a segment of length @xmath147 on the real axis contains at least @xmath148 integers . )
applying this to @xmath113 with all its @xmath149 dimensions scaled up by @xmath1 , the scaled @xmath113 must contain at least @xmath150 points whose coordinates are rational numbers with denominator @xmath1 , i.e. vectors in @xmath94 .
the first factor and its attendant condition @xmath151 are absent if @xmath152 .
let @xmath153 be some constant whose purpose is expained later , in the proof of theorem [ th:1 ] .
we begin by deriving a lower bound on the size of @xmath154 , a subset of @xmath155 .
consider the set @xmath156 . since @xmath157 , proposition [ prop : theta_inf ] implies that any@xmath59 in this set also belongs to @xmath158 .
further , by the middle expression in the definition of @xmath159 , proposition [ prop : norment ] implies that any such @xmath59 also has entropy at least @xmath160 .
thus all @xmath59 in the set @xmath161 belong to @xmath162 .
finally @xmath159 satisfies the conditions of proposition [ prop : anlb ] , hence the size of @xmath161 is bounded from below by @xmath163 .
but @xmath164 , so we established the first claim of the lemma .
now suppose that all @xmath59 in @xmath154 have at least @xmath165 non - zero elements ; for the purposes of this proof we may take these to be the first @xmath98 elements .
then by proposition [ prop : s ] , if @xmath166 is an arbitrary element of @xmath154 , @xmath167 let @xmath168 ; this vector has integral entries , all positive , and summing to @xmath1 .
the maximum of @xmath169 equals @xmath170 , occurring when @xmath171 and @xmath172 .
thus the exponential in ( [ eq : a ] ) is at least @xmath173 .
further , the maximum of @xmath174 subject to @xmath175 occurs at @xmath176 , so the last factor in ( [ eq : a ] ) is at least @xmath177 .
finally @xmath178 , and so ( [ eq : a ] ) implies the second result of the lemma , but with the number @xmath98 still undetermined . by requiring @xmath179 to be less than the smallest non - zero element of @xmath14 , we can ensure that there is no element of @xmath59 which is 0 while the corresponding element of @xmath14 is positive ; this is accomplished by the last term on the r.h.s . of ( [ eq : theta0 ] ) .
thus we can take @xmath98 equal to @xmath180 , the number of non - zero elements of @xmath14 .
the upper bound on @xmath181 and the lower bound on @xmath182 are given by lemmas [ le : nrbn ] and [ le : nran ] .
both these bounds _ increase _ when the entropy tolerance @xmath183 decreases towards 0 , as makes sense . to simplify the proof
we assume that @xmath184 . then combining the two bounds and unifying
some numerical constants @xmath185 when everything else is fixed , this lower bound on @xmath186 ( eventually ) increases as @xmath187 , as we want it to .
in general , this behavior would have been impossible if @xmath188 were 1 .
this is why we introduced @xmath188 and required it to be @xmath189 : it serves to strictly separate our bounds on @xmath182 and @xmath181 .
there is freedom in choosing the value of @xmath188 , which we exploit below . to establish ( [ eq : ratio ] ) we need the l.h.s . of ( [ eq : nranrb ] ) to be @xmath190 .
this reduces to requiring @xmath191 where the constants @xmath192 are @xmath193 we will now show that ( [ eq : n1 ] ) is satisfied by @xmath194 first , assume @xmath195 . setting @xmath196 in ( [ eq : n1 ] ) with @xmath146
we reduce to @xmath197 the r.h.s . of this condition
is a convex combination of @xmath198 and 1 , so the first condition will hold if @xmath199 , which is true when @xmath200 .
now let @xmath201 .
putting @xmath202 in ( [ eq : n1 ] ) , we reduce to establishing @xmath203 , which will hold if @xmath204 , always true . the r.h.s . of ( [ eq : n11 ] )
depends on @xmath205 , which has up to this point been left unspecified .
we finally need @xmath206 and @xmath207 , and we observe that as @xmath208 , the first of these bounds increases while the second decreases .
further , the first bound is finite at 0 and infinite at 1 , whereas the second is infinite at 0 and finite at 1 .
thus there is an optimal @xmath188 which makes the two bounds equal , the @xmath209 which solves @xmath210 .
consider the simplest case @xmath212 first .
we can bound the sum as follows : @xmath213 to see this , note that @xmath214 because the sum is a lower riemann sum for the integral . since the summand is symmetric about @xmath215
, doubling this produces the desired result .
now consider the case of even @xmath98 , i.e. @xmath216 . divide the @xmath5 into @xmath217 pairs , each of which sums to some number @xmath218 and these numbers in turn sum to @xmath1 : @xmath219 here the inequality follows by applying ( [ eq : mu=2 ] ) , which does not depend on @xmath1 , to each of the inner sums .
further , @xmath220 where in the first equality we assume w.l.o.g . that @xmath221 , and the 2nd equality follows from the fact that the number of compositions of @xmath222 into @xmath223 parts ( i.e. the solutions of @xmath224 , @xmath225 ) , is @xmath226
. finally we bound the binomial coefficient by @xmath227 , to arrive at @xmath228 now we turn to the case of odd @xmath98 , i.e. @xmath229 . similarly to what we did above , @xmath230 by ( [ eq : mueven ] ) , the r.h.s
. does not exceed @xmath231 and this last sum can be bounded by the integral @xmath232 we have thus shown that for @xmath229 , @xmath233 eqs .
( [ eq : mueven ] ) and ( [ eq : muodd ] ) establish the proposition for all @xmath102 .
first we show that if @xmath234 then @xmath235 . by proposition [ prop : n1 ] ,
@xmath236 , so by proposition [ prop : theta_inf ] @xmath237 .
further , @xmath238 means that @xmath239 by proposition [ prop : norment ] .
therefore @xmath240 implies that @xmath16 belongs to the set @xmath241 as claimed .
next we put a lower bound on @xmath242 . applying proposition [ prop : s ]
we see that @xmath95 is @xmath243 the r.h.s .
of ( [ eq : a ] ) in the proof of lemma [ le : nran ] with @xmath244 , so @xmath95 is @xmath243 the bound of lemma [ le : nran ] on @xmath245 with @xmath246 .
then from the proof of theorem [ th:1 ] , we see that @xmath247 is @xmath243 the r.h.s . of ( [ eq : nranrb ] ) , but with the condition on @xmath1 being @xmath248 .
the rest of the proof of theorem [ th:1 ] then applies , to the point where @xmath1 has to satisfy @xmath206 and @xmath249 .
@xmath209 equalizes these bounds , and the completion of the proof of theorem [ th:1 ] then establishes that if @xmath250 , @xmath251 .
the function @xmath252 , @xmath107 , is increasing for @xmath253 $ ] .
the first implication in the proposition then follows immediately from theorem 16.3.2 of @xcite , the @xmath22 norm bound on entropy , which states that if two @xmath0-vectors @xmath254 are s.t .
@xmath255 , then @xmath256 .
to prove the second implication we use pinsker s inequality and the `` triangle inequality '' for cross- or relative entropy , or divergence @xmath257 .
applied to @xmath59 and @xmath14 , pinsker s inequality states that @xmath258 ( see @xcite , lemma 12.6.1 ) .
then the triangle inequality , using the uniform distribution as the prior or reference distribution ( @xcite , theorem 12.6.1 ) , implies that @xmath259 . what we want to prove follows from the above two inequalities .
pinsker s inequality can be tightened in two ways : @xcite show that the @xmath260 can be replaced by a factor @xmath261 , and @xcite give right - hand sides that are polynomials involving powers of the norm beyond the square . entirely analogous to that of lemma [ le : nrbn ] , except that the set @xmath262 is defined by ( [ eq : newanbn ] ) instead of ( [ eq : bn ] ) , and the factor @xmath263 in ( [ eq : nrb1 ] ) , coming from the upper bound on @xmath95 of proposition [ prop : s ] , is replaced by the factor @xmath264 of proposition [ prop : norment2 ] .
the proof follows that of lemma [ le : nran ] : first we lower - bound the size of @xmath265 and then the entropy of the @xmath59 in it .
the basic difference is that here we have @xmath22 norms .
if @xmath266 , so is @xmath267 , and then proposition [ prop : anlb ] says that the size of @xmath268 is at least @xmath269 .
second , concerning the entropy of @xmath270 , by proposition [ prop : norment2 ] @xmath271 implies that @xmath272 is at least @xmath273 .
the proof then follows that of lemma [ le : nran ] , except that the term @xmath274 in ( [ eq : a ] ) is replaced by @xmath275 .
the proof uses lemmas [ le : nrbn1 ] and [ le : nran1 ] and is completely analogous to that of theorem [ th:1 ] .
the main feature is that @xmath15 falls out of the new ( [ eq : nranrb ] ) , the exponential is @xmath281 with @xmath282 , and the condition on @xmath1 is now @xmath283 .
@xmath192 are the same as in theorem [ th:1 ] , except for the denominators . finally , @xmath284 is finite at @xmath285 and increases to @xmath286 at @xmath287 , whereas @xmath288 is infinite at @xmath285 and decreases to a finite value at @xmath289 .
thus the equation @xmath290 has a root @xmath209 between 0 and @xmath291 , which equates the two sides and is therefore the optimal @xmath188 .
the proof is analogous to that of lemma [ cor:1 ] .
first , by proposition [ prop : n1 ] , @xmath292 implies that @xmath293 .
second , if @xmath294 then @xmath295 by the 2nd claim of proposition [ prop : n1 ] .
hence if @xmath296 , @xmath16 belongs to the set @xmath265 .
next , by the argument in the proof of lemma [ cor:1 ] , @xmath242 can be lower - bounded by the bound of lemma [ le : nran1 ] with @xmath246 .
so @xmath297 is lower - bounded by the new ( [ eq : nranrb ] ) as in the proof of theorem [ th:2 ] but the condition on @xmath1 is now @xmath298 .
the rest follows as in the proof of theorem [ th:2 ] . in this case
there are no constraints , so by proposition [ prop : theta_inf ] @xmath299 . also , @xmath300 and @xmath301
further , if @xmath302 , then @xmath303 , so the condition of corollary [ cor : pd ] on @xmath1 is @xmath304 .
the conditions @xmath302 and @xmath278 are satisfied if @xmath305 . finally , @xmath306 . m. benoit and s. verdoolage .
polynomial approximations in the polytope model : bringing the power of quasi - polynomials to the masses . in _ proceedings of 6th workshop on optimizations for dsp and embedded systems ( odes-6 )
_ , pages 4554 .
boston , massachusetts , april 2008 .
s. verdoolaege , k. woods , m. bruynooghe , and r. cools .
computation and manipulation of enumerators of integer projections of parametric polytopes . technical report , katholieke universiteit leuven
, department of computer science , celestijnenlaan 200a , b-3001 heverlee ( belgium ) , march 2005 . | consider the construction of an object composed of @xmath0 parts by distributing @xmath1 units to those parts .
for example , say we are assigning @xmath1 balls to @xmath0 boxes .
each assignment results in a certain count vector , specifying the number of balls allocated to each box .
if only assignments satisfying a set of constraints that are linear in these counts are allowable , and @xmath0 is fixed while @xmath1 increases , most assignments that satisfy the constraints result in frequency vectors ( normalized counts ) whose entropy approaches that of the maximum entropy vector satisfying the constraints . this phenomenon of `` entropy concentration '' is known in various forms , and is one of the justifications of the maximum entropy method , one of the most powerful tools for solving problems with incomplete information .
the appeal of entropy concentration comes from the simplicity of the argument : it is based purely on counting and does not need probabilities .
existing proofs of the concentration phenomenon are based on limits or asymptotics . here
we present non - asymptotic , explicit lower bounds on @xmath1 for a number of variants of the concentration result to hold to any prescribed accuracies , taking into account the fact that allocations of discrete units can satisfy constraints only approximately .
the results are illustrated with examples on die tossing , vehicle or network traffic , and the probability distribution of the length of a @xmath2 queue . |
colour - magnitude diagrams ( cmds ) are one of the most important tools available to astronomers for studying stellar evolution , stellar populations and star clusters . and yet , there remain several features found in cmds whose origins are still a mystery .
examples include horizontal branch ( hb ) morphology , the presence of extended horizontal branch ( ehb ) stars , and blue stragglers ( bss ) ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
previous studies have shown that the observed differences in the hbs of milky way globular clusters ( gcs ) are related to metallicity @xcite , however at least one additional parameter is required to explain the spread in their colours .
many cluster properties have been suggested as possible second and third parameters , including age , central density and cluster luminosity , although no definitive candidates have been identified ( e.g. * ? ? ?
* ; * ? ? ?
an explanation to account for the existence of bss has proved equally elusive .
many bs formation mechanisms have been proposed , including stellar collisions ( e.g. * ? ? ?
* ; * ? ? ?
* ) and binary mass - transfer @xcite
. however , no clear evidence has yet emerged in favour of a dominant formation mechanism . in short ,
we still do not understand how many of the physical processes operating within star clusters should affect the appearance of cmds ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
in general , the importance of these processes can be constrained by looking for correlations between particular features in cmds and cluster properties that serve as proxies for different effects .
for example , the central density can be used as a rough proxy for the frequency with which close dynamical encounters occur .
similarly , the cluster mass can be used as a proxy for the rate of two - body relaxation .
once the relevant effects are accounted for , cmds can continue to provide an ideal tool to further our understanding of stellar evolution , stellar populations and star clusters .
it is now clear that an important interplay occurs in clusters between stellar dynamics and stellar evolution .
for example , dynamical models have shown that star clusters expand in response to mass - loss driven by stellar evolution , particularly during their early evolutionary phases when massive stars are still present ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ?
mass - loss resulting from stellar evolution has also been proposed to cause horizontal branch stars to exhibit more extended radial distributions relative to red giant branch and main - sequence turn - off ( msto ) stars in globular clusters having short central relaxation times relative to the average hb lifetime ( e.g. * ? ? ?
* ; * ? ? ?
this can be understood as follows .
red giant branch ( rgb ) stars should be more mass segregated than other stellar populations since they are among the most massive stars in gcs .
hb stars , on the other hand , are much less massive since rgb stars experience significant mass loss upon evolving into hb stars .
consequently , two - body relaxation should act to re - distribute hb stars to wider orbits within the cluster potential relative to rgb and msto stars @xcite , provided the average hb lifetime is shorter than the central relaxation time .
studies have shown that the radial distributions of the hb populations in some gcs could differ from those of other stellar populations .
for instance , @xcite presented evidence that blue hb stars could be more centrally concentrated than red hb and sub - giant branch stars in the gc ngc 1851 .
conversely , @xcite showed that blue hb stars could be centrally depleted relative to other stellar types in the gc ngc 6205 . to date , no clear evidence has been found linking the spatial distributions of hb stars to any global cluster properties .
peculiar trends have also been reported for the radial distributions of rgb stars .
for example , a deficiency of bright red giants has been observed in the gc ngc 1851 ( e.g. * ? ? ? * ) .
@xcite discussed the possibility that this deficiency could be the result of strong mass loss on the rgb .
alternatively , some authors have suggested that dynamical effects could deplete red giants .
for instance , giants could experience collisions more frequently than other stellar populations due to their larger cross - sections for collision @xcite
. one important example of the interplay that occurs in clusters between stellar evolution and stellar dynamics can be found in the study of blue stragglers . found commonly in both open and globular clusters , bss
are thought to be produced by the addition of hydrogen to the cores of low - mass main - sequence ( ms ) stars , and therefore appear as an extension of the msto in cluster cmds @xcite .
this can occur via multiple channels , most of which involve the mergers of low - mass ms stars since a significant amount of mass is typically required to reproduce the observed locations of bss in cmds ( e.g. * ? ? ?
stars in close binaries can merge if enough orbital angular momentum is lost , which can be mediated by dynamical interactions with other stars , magnetized stellar winds , tidal dissipation or even an outer triple companion ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
alternatively , ms stars can collide directly , although this is also thought to usually be mediated by multiple star systems ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
finally , bss have also been hypothesized to form by mass - transfer from an evolving primary onto a normal ms companion via roche lobe overflow @xcite .
whatever the dominant bs formation mechanism(s ) operating in dense star clusters , it is now thought to somehow involve multiple star systems .
this was shown to be the case in even the dense cores of gcs @xcite where collisions between single stars are thought to occur frequently @xcite . in @xcite
, we showed that the numbers of bss in the cores of a large sample of gcs correlate with the core masses .
we argued that our results are consistent with what is expected if bss are descended from binary stars .
@xcite also showed that at least @xmath0 of the bss in the old open cluster ngc 188 have binary companions .
although the nature of these companions remains unknown , it is clear that binaries played a role in the formation of these bss .
dynamical interactions occur frequently enough in dense clusters that they are also expected to be at least partly responsible for the observed properties of bss .
it follows that the current properties of bs populations should reflect the dynamical histories of their host clusters . as a result
, bss could provide an indirect means of probing the physical processes that drive star cluster evolution . in this paper , we present a homogeneous catalogue for red giant branch , main - sequence turn - off , horizontal branch and blue straggler stars in a sample of 35 milky way ( mw ) gcs taken from the advanced camera for surveys ( acs ) survey for globular clusters @xcite . with this catalogue ,
we investigate two important issues related to stellar populations in gcs .
first , we test the observational correlation found for bss presented in @xcite by re - doing the study with newer and more accurate photometry . the larger spatial coverage considered in our new sample offers an important additional constraint for the origin of this correlation .
second , we perform the same statistical comparison for rgb , hb and msto stars in order to study their radial distributions .
this will allow us to test some of the results and hypotheses introduced in @xcite , where we first presented this technique for studying stellar populations . in particular , we found evidence for a surplus of rgb stars in low - mass gc cores relative to msto stars .
however , we concluded that the study needed to be re - done with better photometry .
the acs data are of sufficiently high quality to address this issue . in section [ method ] , we present our selection criteria to determine the numbers of bs , rgb , hb and msto stars located in the central cluster regions .
the spatial coverage of the photometry extends out to several core radii from the cluster centre for most of the clusters in our sample . for these clusters ,
we have obtained number counts within several circles centred on the cluster centres provided in @xcite for various multiples of the core radius .
this catalogue is presented in section [ results ] . in this section ,
we also present a comparison between the sizes of the different stellar populations and the total stellar masses contained within each circle and annulus .
finally , we discuss our results for both bss and the other stellar populations in section [ discussion ] .
in this section , we present our sample of cmds and define our selection criteria for each of the different stellar populations .
we also discuss the spatial coverage offered by the acs sample , and describe how we obtain estimates for several different fractions of the total cluster mass from king models .
the data used in this study consists of a sample of 35 mw gcs taken from the acs survey for globular clusters @xcite .
the acs survey provides unpecedented deep photometry in the f606w ( @xmath1 v ) and f814w ( @xmath1 i ) filters that is nearly complete down to @xmath2 m@xmath3 . in other words ,
the cmds extend reliably from the hb all the way down to about 7 magnitudes below the msto .
we have confirmed that the photometry is nearly complete above at least 0.5 magnitudes below the msto for every cluster in our sample .
this was done using the results of artificial star tests taken from @xcite , and confirms that the photometric quality of the stellar population catalogue presented in this paper is very high .
we have also considered foreground contamination by field stars , and it is negligible .
each cluster was centred in the acs field , which extends out to several core radii from the cluster centre in most clusters and , in a few cases , beyond even 15 core radii .
coordinates for the cluster centres were taken from @xcite .
these authors found their centres by fitting a series of ellipses to the density distributions within the inner 2 of the cluster centre , and computing an average value .
the core radii were taken from @xcite . in order to select the number of stars belonging to each stellar population
, we define a series of lines in the ( f606w - f814w)-f814w plane that act as boundaries enclosing each of the different stellar populations .
to do this , we fit theoretical isochrones taken from @xcite to the cmds of every cluster in our sample .
each isochrone was generated using the metallicity and age of the cluster , and fit to its cmd using the corresponding distance modulus and extinction provided in @xcite .
the msto was then defined using our isochrone fits by selecting the bluest point along the ms .
this acts as our primary point of reference for defining the boundaries in the cmd for the different stellar populations .
consequently , the selection criteria provided in this paper are a significant improvement upon the criteria presented in @xcite , and our new catalogue for the different stellar populations is highly homogeneous
. two additional points of reference must also be defined in order for our selection criteria to be applied consistently from cluster - to - cluster .
first , the selection criteria for the hb are determined by fitting a line through the approximate mid - point of the points that populate it in the cmd .
this line is then used to define upper and lower boundaries for the hb .
theoretical isochrones become highly uncertain at the hb , so it is necessary to specify this additional criterion by eye .
second , the lower boundary of the rgb is defined for each cluster as the point along its isochrone corresponding to a helium core mass of 0.08 m@xmath3 .
we do not include rgb stars brighter than the hb since the tilt of the upper rgb varies significantly from cluster - to - cluster , presenting a considerable challenge for the consistency of our selection criteria .
moreover , the distinction between rgb and asymptotic giant branch stars in the cmd is often ambiguous .
example selection criteria for each of the different stellar populations are shown in figure [ fig : fig1 ] .
formal definitions for the boundaries in the cluster cmd that define the bs , rgb , hb and msto populations are provided in appendix [ appendix ] .
we note that the sizes of our selection boxes have been chosen to accomodate the photometric errors , which contribute to broadening the various evolutionary sequences in the cmd . with superior photometry
, the sizes of our selection boxes could therefore be reduced .
this would further decrease contamination from field stars in our samples . ]
the acs field of view extends out to several core radii from the cluster centre for nearly every cluster in our sample .
consequently , we have obtained estimates for the number of stars contained within four different circles centred on the central cluster coordinates provided in @xcite .
we list these numbers only for clusters for which the indicated circle is completely sampled by the field of view .
the radii of the circles were taken to be integer multiples of the core radius , and we focus our attention on the inner four core radii since the field of view extends beyond this for only a handful of the clusters in our sample .
an example of this is shown in figure [ fig : fig2 ] .
the numbers we list are cumulative , so that entries for each circle include stars contained within all smaller circles . ] in order to obtain accurate estimates for the total stellar mass contained within each circle , we generated single - mass king models calculated using the method of @xcite to obtain luminosity density profiles for the majority of the clusters in our sample .
the profiles were obtained using the concentration parameters of @xcite and the central luminosity densities of @xcite for each cluster in @xcite that overlaps with our sample .
we then integrated the derived luminosity density profiles numerically in order to estimate the total stellar light contained within each circle .
after removing clusters with high concentration parameters @xcite for which king models are known to provide a poor fit , we multiplied the total stellar light by a mass - to - light ratio of 2 in order to obtain estimates for the total stellar mass contained within each circle . calculating the total stellar mass contained within each circle from king models requires a number of assumptions that we will discuss fully in section [ discussion ] .
in this section , we present our catalogue along with the results of our comparisons between the sizes of the different stellar populations and the total stellar mass contained within each circle and annulus . the numbers of bs , rgb , hb and msto stars found within several different circles are shown for all clusters in table [ table : catalogue ] , along with the total number of stars with magnitudes brighter than 0.5 mag below the msto .
number counts are only shown whenever the spatial coverage is complete within the indicated circle .
how can we use our catalogue to learn which , if any , cluster properties affect the appearance of cmds ?
one way to accomplish this is by plotting the size of a given stellar population in a particular circle versus the total stellar mass contained within it . from this
, lines of best fit can be found that provide equations relating the size of each stellar population to the total stellar mass contained within each circle .
as described below , this is ideal for probing the effects of the cluster dynamics on the appearance of cmds .
the rate of two - body relaxation for a cluster can be approximated using the half - mass relaxation time @xcite : @xmath4^{3/2}n^{1/2}[m / m_{\odot}]^{-1/2 } years,\ ] ] where @xmath5 is the half - mass radius , @xmath6 is the total number of stars within @xmath5 and @xmath7 is the average stellar mass .
the half - mass radii of mw gcs are remarkably similar independent of mass , and simulations have shown that @xmath5 changes by a factor of at most a few over the course of a cluster s lifetime @xcite .
the gcs that comprise our sample show a range of masses spanning roughly 3 orders of magnitude , and have comparably old ages @xcite .
therefore , equation [ eqn : t - rh ] suggests that the degree of dynamical evolution ( due to two - body relaxation ) experienced by a cluster is primarily determined by the total cluster mass for the gcs in our sample . in particular , more massive clusters
are less dynamically evolved , and vice versa .
consequently , if the size of a given stellar population is affected by two - body relaxation , the effects should be the most pronounced in the least massive clusters in our sample .
additionally , the rate of ( direct ) stellar collisions increases with increasing cluster mass ( e.g. * ? ? ?
this suggests that , if a given stellar population is affected by collisions , the effects should be the most pronounced in the most massive clusters in our sample .
therefore , by comparing the size of each stellar population to the total stellar mass contained within a given circle , the effects of the cluster dynamics can be quantified .
this technique also ensures a normalized and consistent comparison since it accounts for cluster - to - cluster differences in the fractional area sampled by the acs field of view .
that is , we are consistently comparing the same structural area for each cluster .
the validity and implications of all of these assumptions will be discussed further in section [ discussion ] .
plots showing the number of stars belonging to each stellar population as a function of the total stellar mass contained within each circle are shown in figure [ fig : mshell_vs_npop_2x2_cum ] .
uncertainties for the number of stars belonging to each stellar population were calculated using poisson statistics .
we also plot in figure [ fig : mshell_vs_npop_2x2_noncum ] the number of stars belonging to each stellar population as a function of the total stellar mass contained in each annulus outside the core .
that is , we considered the populations for each annulus individually , as opposed to considering every star with a distance from the cluster centre smaller than the radius of the outer - most circles . recall that we have neglected clusters for which our theoretical king models provide a poor description of the true density distributions .
this was the case for clusters in our sample having a high concentration parameter , most of which are labelled as post - core collapse in @xcite . , @xmath8 , @xmath9 and @xmath10 .
blue corresponds to blue stragglers , red to red giant branch stars , green to horizontal branch stars and black to main - sequence turn - off stars .
estimates for the total stellar mass contained within each circle were found using single - mass king models , as described in section [ king ] .
[ fig : mshell_vs_npop_2x2_cum ] ] .
[ fig : mshell_vs_npop_2x2_noncum ] ] we performed a weighted least - squares fit for every relation in figure [ fig : mshell_vs_npop_2x2_cum ] and figure [ fig : mshell_vs_npop_2x2_noncum ] . slopes and y - intercepts for these lines are shown in table [ table : mshell_vs_npop_cum ] and table [ table : mshell_vs_npop_noncum ] , respectively .
uncertainties for the slopes and y - intercepts were found using a bootstrap methodology in which we generated 1,000 fake data sets by randomly sampling ( with replacement ) number counts from the observations .
we obtained lines of best fit for each fake data set , fit a gaussian to the subsequent distribution and extracted its standard deviation .
as shown in table [ table : mshell_vs_npop_cum ] , the power - law index is sub - linear for bss within the core at much better than the @xmath11 confidence level , and it is consistent with the slope obtained in our earlier analysis presented in @xcite .
the slopes for the bss are also sub - linear at better than the @xmath11 confidence level for all circles outside the core .
this is also the case for all annuli outside the core , as shown in table [ table : mshell_vs_npop_noncum ] .
note , however , that the uncertainties for the bs slopes are very large for all annuli outside the core , whereas this is not always the case for corresponding circles outside the core .
this is the result of the fact that the number of bss drops off rapidly outside the core in several clusters so that the corresponding poisson uncertainties , which are given by the square - root of the number of bss , are significant .
the rapid decline of bs numbers with increasing distance from the cluster centre in these clusters has also contributed to an increased degree of scatter in the relations for annuli outside the core relative to the corresponding relations for circles outside the core .
the slopes are consistent with being linear for all other stellar populations in the core within their respective @xmath11 confidence intervals .
this agrees with the results of our earlier analysis presented in @xcite when we performed the comparison using the total core masses .
the slopes are also consistent with being linear for all circles outside the core for both hb and msto stars .
the power - law indices are sub - linear at the @xmath11 confidence level only for rgb stars , and this is only the case for circles outside the core .
the power - law index is nearly unity for the core rgb population , yet the associated uncertainty is very large . upon closer inspection ,
the distribution of power - law indices obtained from our bootstrap analysis for rgb stars in the core is strongly bi - modal , with comparably - sized peaks centred at @xmath12 and @xmath13 .
this bi - modality is most likely an artifact of our bootstrap analysis caused by a chance alignment of data points in the log m@xmath14-log n@xmath15 plane . upon performing the comparison for only those stars found within particular annuli
, our results suggest that the slopes are consistent with being sub - linear for all stellar populations at the @xmath11 confidence level in only the annulus immediately outside the core ( i.e. @xmath16 ) .
the slopes are consistent with being linear for rgb , hb and msto stars in all other annuli .
we also tried performing the same comparisons using the total number of stars in each circle and annulus as a proxy for the total stellar mass . in this case , the slopes are sub - linear for bss within all circles and annuli at the @xmath11 confidence level .
once again , the uncertainties are very large for all annuli outside the core , whereas this is not the case for corresponding circles outside the core . the slopes are consistent with being linear within the @xmath17 confidence interval for all other stellar populations in all circles and annuli .
our results are therefore inconsistent with those presented in @xcite for the core rgb populations , in which we found that rgb numbers scale sub - linearly with the number of stars in the core at the @xmath11 confidence level .
we will discuss the implications of these new results in section [ discussion ] . as a check of our previous results reported in @xcite
, we also looked for a correlation between the observed number of bss in the cluster core and the number predicted from single - single ( 1 + 1 ) collisions .
the results of this comparison are shown in figure [ fig : ncoll_vs_nbs ] .
we define the predicted number of bss formed from 1 + 1 collisions as @xmath18 , where @xmath19 is the average bs lifetime and @xmath20 is the average time between 1 + 1 collisions in the cluster core .
we adopt the same definition for @xmath20 as used in @xcite , and assume @xmath21 gyrs as well as an average stellar mass and radius of @xmath22 m@xmath3 and @xmath22 r@xmath3 , respectively .
we also adopt a constant mass - to - light ratio of @xmath23 for all clusters .
central luminosity densities and velocity dispersions were taken from @xcite and @xcite , respectively . upon performing a weighted line of best fit for every cluster in our sample that overlaps with the catalogue of @xcite ,
we find a power - law index of @xmath24 ( the uncertainty was found using the bootstrap methodology described in section [ statistics ] ) . for the subset of dense clusters having a central luminosity density satisfying log @xmath25
, we find a power - law index of @xmath26 .
as before , we find no significant correlation with collision rate , even for the subset of dense clusters .
although we do find a weak dependence of bs numbers on collision rate for the entire sample , this is not unexpected since the collision rate and the core mass are themselves correlated , and our results suggest that there exists a strong correlation between bs numbers and the core masses . versus the logarithm of the observed number of bss in the core n@xmath27 .
filled circles correspond to clusters having central luminosity densities satisfying log @xmath28 , whereas open triangles correspond to dense clusters for which log @xmath25 .
the adopted definition for n@xmath29 has been provided in the text . [
fig : ncoll_vs_nbs ] ]
in this paper , we have presented a catalogue for bs , rgb , hb and msto stars in a sample of 35 gcs .
our catalogue provides number counts for each stellar population within several different circles centred on the cluster centre .
the radii of the circles were taken to be integer multiples of the core radius , and we have focussed on the inner four core radii since the field of view extends beyond this for only a handful of the clusters in our sample .
particular consideration was given to our selection criteria for the different stellar populations in order to ensure that they were applied consistently from cluster - to - cluster .
in particular , we have improved upon our previous selection criteria @xcite by fitting theoretical isochrones to the cluster cmds .
this provides an unambiguous definition for the location of the msto , which acts as the primary point of reference for the application of our selection criteria . as a result ,
our new catalogue is highly homogeneous .
we have used our catalogue to quantify the dependence of the size of each stellar population on the total stellar mass enclosed within the same radius . as described in section [ statistics ] , this provides a means of quantifying the effects , if any , had by the cluster dynamics in shaping the appearance of cmds above the msto .
below , we summarize the implications of our results for each of the different stellar populations . we have confirmed our previous result that the numbers of bss in the cores of gcs scale sub - linearly with the core masses @xcite .
that is , we find proportionately larger bs populations in low - mass gcs .
there exist several possibilities that could explain the origin of this sub - linear dependence .
first , we previously suggested that this could be an artifact of an anticorrelation between the binary fraction and the cluster ( or core ) mass @xcite .
this assertion stems from the fact that , if bss have a binary origin , we expect their numbers to scale with the core mass as @xmath30 , where @xmath31 is the binary fraction in the core .
as before , we find that @xmath32 .
our result could therefore be explained if @xmath33 .
second , we suggested in @xcite that the fact that the least massive gcs in our sample should be more dynamically evolved than their more massive counterparts could be contributing to the observed sub - linearity for bss . in particular , the low - mass clusters in our sample should have experienced a significant depletion of their very low - mass stars as a result of stellar evaporation induced by two - body relaxation ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . in turn
, this could contribute to a higher fraction of merger products having masses that exceed that of the msto in low - mass gcs . as a result
, more merger products would appear brighter and bluer than the msto in these clusters cmds , leading to more merger products being identified as bss .
finally , mass segregation could also be contributing to the observed sub - linear dependence for bss .
again , this is the result of the fact that the rate of two - body relaxation , and therefore dynamical friction , is in general the fastest in low - mass clusters .
bss are among the most massive stars in gcs , so they should rapidly migrate into the core via dynamical friction in clusters for which the half - mass relaxation time is shorter than the average bs lifetime .
it follows that proportionately more bss could have drifted into the core via dynamical friction in low - mass gcs .
this would also contribute to the observed sub - linear dependence of bs numbers on the core masses .
this last hypothesis can be tested by comparing our scaling relations for progressively larger circles outside the core .
if mass segregation is indeed the cause of the observed sub - linear dependence of bs numbers on the core masses , then we might expect the power - law index to systematically increase as we consider progressively larger circles .
that is , we could be including more bss that have not yet migrated into the core via dynamical friction , particularly in the most massive clusters in our sample . however , our results suggest that the power - law index remains roughly constant for all circles .
this is the case for both the comparison to the total stellar masses as well as to the total number of stars contained in each circle .
the fact that these relations are comparably sub - linear within all circles can be interpreted as evidence that mass segregation is not the dominant effect contributing to the observed sub - linear dependence of bs numbers on the total stellar masses ( or the total number of stars ) .
we note that for many of the clusters in our sample , the spatial coverage is comparable to or exceeds the half - mass radius .
this is a sufficiently large fraction of the total cluster area for our comparison to be sensitive to the effects of mass segregation . on the other hand ,
several gcs are known to show evidence for a bi - modal bs radial distribution ( e.g. * ? ? ?
* ; * ? ? ?
that is , in these clusters the number of bss is highest in the central cluster regions and decreases with increasing distance from the cluster centre until a second rise in bs numbers occurs in the cluster outskirts .
this seconary outer peak has been shown to occur at a distance from the cluster centre that exceeds 20 core radii in several cases .
consequently , the spatial coverage provided by the acs data likely does not extend sufficiently far in most clusters to detect any bi - modality in the bs radial distribution . nonetheless , if applied to a statistically - significant sample for which the spatial coverage is complete out to the tidal radius , the technique we have presented in this paper could provide a powerful constraint for the origin of the bi - modal bs radial distribution observed in several mw gcs by addressing the role played by mass segregation .
we also tried to correlate the number of bss observed in the cluster core with the number predicted from single - single collisions .
as in @xcite , we find that bs numbers depend strongly on core mass , but not on collision rate .
this also proved to be the case for the subset of dense clusters satisfying log @xmath25 .
our previous interpretation that our results provide strong evidence for a binary , as opposed to a collisional , origin for bss in gcs therefore remains the same .
the technique used in this paper to compare the sizes of the different stellar populations was first presented in @xcite . in that study
, we introduced our method and applied it to a sample of 56 gcs taken from @xcite in order to study their rgb populations .
our results were consistent with a sub - linear dependence of rgb numbers on the core masses . in particular , we found evidence for a surplus of rgb stars relative to msto stars in the cores of low - mass gcs .
we considered several possible causes for this result , but concluded that our analysis should ideally be repeated with superior photometry in order to properly assess effects such as completeness .
given the high - quality of the acs data , we are now in a position to reassess our previous result for rgb stars . upon applying our technique to the acs sample , we find that the numbers of rgb stars scale linearly with the core masses to within one standard deviation .
this is also the case for our comparison to the total number of stars in the core .
this suggests that we should reject our previous conclusions for this stellar population reported in @xcite .
specifically , if we take a strict @xmath11 limit as our criterion for whether or not the slopes are sub - linear at a statistically significant level , then the core rgb slope reported in this paper is consistent with being linear whereas this was not the case for the core rgb slope reported in @xcite . however , if we take a more stringent criterion for statistical significance , then there is no inconsistency between our new and old results for rgb stars , and both slopes are consistent with being linear .
we also tried comparing the rgb catalogue presented in this paper with the one presented in @xcite .
this showed that the old rgb numbers are slightly deficient relative to the new numbers at the high - mass end . although this difference is not sufficiently large to completely account for the difference in slopes ( for the comparison with the core mass ) found between our new and old rgb catalogues , it works in the right direction and is likely a contributing factor . if our uncertainties are also factored in , then our new and old slopes agree to within one standard deviation ( due mainly to the large uncertainty for the new slope ) .
the source of the disagreement between our old and new rgb catalogues is unclear , and we can not say whether or not incompleteness ( in the old data set ) is the culprit .
the results of our artificial star tests have at least confirmed that incompleteness is not an issue for our new catalogue , however it could certainly have contributed to the lower rgb numbers reported in @xcite .
indeed , the central cluster density tends to be higher in more massive clusters , which should negatively affect completeness .
it is not clear , however , why this would have affected rgb stars more than msto stars in the old data set . given all of these considerations , we feel that our new results show that this issue needs to be looked at in more detail before any firm conclusions can be drawn .
the evidence in favour of rgb numbers scaling linearly with the core masses is interesting .
for one thing , it suggests that two - body relaxation does not significantly affect rgb population size relative to other stellar populations of comparable mass in even the dense central regions of gcs .
this is not surprising , since two - body relaxation is a long - range effect for which the stellar radius plays a negligible role .
second , it suggests that collisions do not significantly deplete rgb stars relative to other stellar populations despite their much larger radii .
this is because the collision rate increases with increasing cluster mass , so we would expect rgb stars to appear preferentially depleted in massive clusters if they are significantly affected by collisions ( e.g. * ? ? ?
* ; * ? ? ?
third , it suggests that the sub - linear relation found for bss does not contribute to a sub - linear relation for rgb stars despite the fact that bss should eventually evolve to occupy our rgb selection box , as discussed in @xcite .
this is likely the result of the relatively small sizes of the bs populations in our sample when compared to the numbers of rgb stars since the rate at which evolved bss ascend the rgb is thought to be comparable to the rgb lifetimes of regular msto stars @xcite .
alternatively , this result could , at least in part , be explained if a smaller fraction of bss end up sufficiently bright and blue to be identified as bss in the cmds of massive gcs . in other words , it could be that a larger fraction of bss are hidden along the ms in massive clusters , as discussed in section [ bss ] . in this case
, the contributions to rgb populations from evolved bss could be comparable in all clusters , in which case a linear relationship between rgb numbers and the core masses would be expected .
finally , evolved bss would be expected to have a negligible impact on rgb population size if the average bs lifetime is considerably longer than the lifetimes of rgb stars .
this effect is difficult to quantify , however , given that bs lifetimes are poorly constrained in the literature ( e.g. * ? ? ?
* ; * ? ? ?
our results suggest that hb numbers scale linearly with the core masses .
this can be interpreted as evidence that two - body relaxation does not significantly affect the radial distributions of hb stars in gcs relative to the other stellar populations above the msto .
one reason to perhaps expect that two - body relaxation should affect the spatial distributions of hb stars stems from the fact that rgb stars are among the most massive stars in clusters , and they undergo significant mass loss upon evolving into hb stars .
consequently , the progenitors of hb stars should be heavily mass segregated .
hb stars themselves , however , have relatively low - masses so that two - body relaxation and strong dynamical encounters should act to re - distribute them to wider orbits within the cluster potential .
the hb lifetime is roughly constant at @xmath34 years @xcite and it is comparable to or exceeds the core relaxation times for most of the low - mass clusters in our sample @xcite .
therefore , we might expect the hb populations in these clusters to exhibit more extended radial profiles relative to more massive clusters .
this would contribute to a sub - linear relationship between the numbers of hb stars and the core masses .
our uncertainties are sufficiently large that this possibility can not be entirely ruled out , however our results are consistent with a linear relationship between hb numbers and the core masses .
recent observations have revealed the presence of multiple stellar populations in a number of mw gcs ( e.g. * ? ? ?
the majority of these cases have been reported in very massive clusters .
moreover , their existence is thought to be related to the chemical properties of gcs , in particular an observed anticorrelation between their sodium and oxygen abundances . in turn , these chemical signatures have been argued to be linked to the cluster metallicity , mass and age @xcite .
we identified clusters in our sample currently known to host multiple stellar populations , but none of these were clear outliers in our plots .
consequently , the effects had on our results by multiple stellar populations remains unclear .
it is certainly possible that multiple stellar populations have contributed to the uncertainties for the weighted lines of best fit performed for the relations in figure [ fig : mshell_vs_npop_2x2_cum ] and figure [ fig : mshell_vs_npop_2x2_noncum ] .
it is difficult to quantify the possible severity of this effect , however , given the limited evidence linking multiple stellar populations to cluster properties .
although the uncertainties are sufficiently large that the slopes are consistent with being linear at the @xmath11 confidence level for all stellar populations when performing the comparisons with the total stellar mass , the reported slopes are typically less than unity within the @xmath17 , and often even the @xmath35 , confidence interval . this does not appear to result from the fact that we have obtained our estimates for the total stellar masses by numerically integrating 3-dimensional density distributions and are comparing to number counts , which are projected quantities . to address this , we also tried obtaining the total stellar masses by numerically integrating 2-dimensional surface brightness profiles so that we are consistently comparing only projected quantities . despite this , our results remain unchanged and the new slopes agree with the old ones to within one standard deviation for all stellar populations .
another possibility to account for this trend that is perhaps worth considering is a systematic dependence of the mass - to - light ratios of clusters on their total mass .
there are two ways this could have affected our analysis .
first , stellar remnants have been shown to affect the dynamical evolution of clusters , and therefore the sizes of their cores ( e.g. * ? ? ?
* ; * ? ? ?
it follows that , if the number of stellar remnants depends on the cluster mass , then this could contribute to an additional underlying dependence of the core radius on the cluster mass .
this could perhaps arise as a result of the fact that the ratio of the rate of stellar evolution to the rate of dynamical evolution is larger in more massive clusters , since in general the rate of two - body relaxation decreases with increasing cluster mass whereas the rate of stellar evolution is independent of the cluster mass .
coupled with their deeper gravitational potential wells , this could contribute to more massive clusters retaining more stellar remnants .
second , variations in the mass - to - light ratios of clusters can also occur as a result of changes in the average stellar mass ( not including stellar remnants ) @xcite .
that is , we can approximate the total stellar mass contained in the core as : @xmath36 where @xmath37 is the mass - to - light ratio , @xmath38 is the central luminosity density , @xmath7 is the average stellar mass and @xmath39 is the total number of stars in the core . based on our results , @xmath40 , where we have used the total number of stars in the core with magnitudes brighter than 0.5 mag below the msto as a proxy for @xmath39 .
this could suggest that @xmath41 .
in other words , the average stellar mass in the core decreases weakly with increasing core mass .
this could in part be due to the fact that more massive clusters should be less dynamically evolved than their less massive counterparts , and should therefore be less depleted of their low - mass stars due to stellar evaporation induced by two - body relaxation ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
similarly , mass segregation should also tend to operate more rapidly in low - mass clusters , which acts to migrate preferentially massive stars into the core @xcite .
alternatively , differences in the stellar mass function in the core could result from variations in the degree of primordial mass segregation , or even variations in the initial stellar mass function .
we have assumed throughout our analysis that the core mass is a suitable proxy for the total cluster mass .
we have checked that these two quantities are indeed correlated , however this does not tell the whole story since we are also using the total cluster mass as a proxy for the degree of dynamical evolution .
the central concentration parameter , defined as the logarithm of the ratio of the tidal to core radii , describes the degree to which a cluster is centrally concentrated .
previous studies have shown that there exists a weak correlation between the concentration parameter and the total cluster mass ( e.g. * ? ? ?
* ; * ? ? ?
in order to better use our technique to reliably probe the effects of the cluster dynamics on the sizes and radial distributions of the different stellar populations , the concentration parameter should ideally be accounted for when applying our normalization technique in future studies .
it is not yet clear how the concentration parameter can be properly absorbed into the normalization , however its effect on our analysis should be small given the weak dependence on cluster mass .
the assumption that the degree of dynamical evolution experienced by a given cluster depends only on its mass is also incorrect .
two - body relaxation has been shown to dominate cluster evolution for a significant fraction of the lives of old mw gcs ( e.g. * ? ? ?
* ) , however other effects can also play a significant role .
for example , stellar evolution is known to affect the dynamical evolution of star clusters , although its primary role is played during their early evolutionary phases ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
tidal effects from the galaxy have also been shown to play an important role in deciding the dynamical fates of clusters by increasing the rate of mass loss across the tidal boundary ( e.g. * ? ? ?
consequently , clusters with small perigalacticon distances should appear more dynamically evolved than their total mass alone would suggest .
this effect can be significant , and has likely contributed to increasing the uncertainties found for the comparisons to the total stellar mass .
therefore , tidal effects from the galaxy should also ideally be absorbed into our normalization technique in future studies .
this can be done by using the perigalacticon distances of clusters as a rough proxy for the degree to which tides from the galaxy should have affected their internal dynamical evolution @xcite .
interestingly , tides could also help to explain why the uncertainties for the comparisons to the total number of stars in each circle are considerably smaller than for the comparisons to the total stellar mass .
we have used the number of stars with magnitudes brighter than 0.5 mag below the msto as a proxy for the total number of stars .
consequently , we are comparing stars within a very narrow mass range , so that all populations of interest should have been comparably affected by two - body relaxation ( except , perhaps , for hb stars ) independent of tidal effects from the galaxy .
in other words , tides should affect all stars above the msto more or less equally , and this is consistent with our results .
it is also worth mentioning here that our king models consider only a single stellar mass .
this assumption is not strictly true and could also be contributing to increasing the uncertainties found for the comparisons to the total stellar masses .
an additional concern is that we do not know if the clusters in our sample are currently in a phase of core contraction or expansion .
this has a direct bearing on the recent history of the stellar density in the core , and therefore the degree to which stars in the core should have been affected by close dynamical interactions .
these effects are independent of two - body relaxation and occur on a time - scale that is typically much shorter than the half - mass relaxation time @xcite
. the effects could be significant in clusters that were recently in a phase of core - collapse but have since rebounded back out of this highly concentrated state .
this could occur , for example , as a result of binary formation induced by 3-body interactions combined with their subsequent hardening via additional encounters @xcite .
in general , binaries play an important role in the dynamical evolution of clusters , and could have affected our results in a number of ways .
this is a difficult issue to address even qualitatively given how little is currently known about the binary populations in globular clusters .
theoretical models suggest , however , that the time - scale for core contraction is often longer than a hubble time , and that this evolutionary phase will only come to an end once the central density becomes sufficiently high for hardening encounters involving binaries to halt the process @xcite .
it follows that the cores of most mw gcs are expected to currently be in a phase of core contraction .
this process is ultimately driven by two - body relaxation , so that our assumption that the total cluster mass provides a suitable proxy for the degree of dynamical evolution is still valid . in summary , our results suggest that effects related to the cluster dynamics do not significantly affect the relative sizes of the different stellar populations above the msto .
this is the case for at least rgb , hb and msto stars .
bss , on the other hand , show evidence for a sub - linear dependence of population size on the total stellar mass contained within the same radius .
whether or not the cluster dynamics is responsible for this sub - linearity is still not clear .
notwithstanding , our results have provided evidence that mass segregation is not the dominant cause for this result , although it will be necessary to redo the comparison performed in this study with a larger spatial coverage in order to fully address this question .
further insight into the origin of the sub - linearity found for bss will be provided by reliable binary fractions for the clusters in our sample , which are forthcoming ( sarajedini 2010 , private communication ) .
.stellar population catalogue [ cols="<,<,<,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ]
in this section , we present our selection criteria for bs , rgb , hb and msto stars .
our method is similar to that described in @xcite , and we have used this as a basis for the selection criteria presented in this paper .
first , we define a location for the msto in the ( f606w - f814w)-f814w plane using our isochrone fits .
the msto is chosen to be the bluest point along the ms of each isochrone , which we denote by ( ( v - i)@xmath42,i@xmath42 ) . in order to distinguish bss from msto stars , we impose the conditions : @xmath43 where the slope of this line is @xmath44 and its y - intercept is given by : @xmath45 similarly , we distinguish bss from hb stars by defining the following additional boundaries : @xmath46 where @xmath47 , @xmath48 and @xmath49 .
we also define : @xmath50 and @xmath51 , where @xmath52 roughly corresponds to the mid - point of points that populate the hb and is chosen by eye for each cluster so that our selection criteria best fits the hb in all of the cmds in our sample .
we apply a similar set of conditions to the rgb in order to select stars belonging to this stellar population .
these boundary conditions are : @xmath53 where @xmath54 , @xmath55 is defined as the f814w magnitude corresponding to a core helium mass of 0.08 m@xmath3 and : @xmath56 core helium - burning stars , which we refer to as hb stars , are selected if they satisfy one of the following sets of criteria : @xmath57 @xmath58 or @xmath59 we define @xmath60 on a cluster - by - cluster basis in order to ensure that we do not over- or under - count the number of hb stars .
this is because the precise value of ( f606w - f814w ) at which the hb becomes the rgb varies from cluster - to - cluster .
in addition , the precise location of the transition in the cluster cmd between hb and ehb stars remains poorly understood . to avoid this ambiguity ,
we consider hb and ehb stars together throughout our analysis , and collectively refer to all core helium - burning stars as hb stars throughout this paper .
finally , msto stars are selected according to the following criteria : @xmath61
we would like to thank ata sarajedini , aaron dotter and roger cohen for providing the data on which this study was based and for their extensive support in its analysis .
we would also like to thank evert glebbeek for useful discussions .
this research has been supported by nserc and ogs .
ambartsumian v. a. 1938 , ann .
leningrad state univ . , 22 , 19 ; english translation : ambartsumian v.a . , 1985 , in dynamics of star clusters , iau symp .
113 , eds .
goodman j. & hut p. ( dordrecht : reidel ) , 521 anderson j. , sarajedini a. , bedin l. r. , king i. r. , piotto g. , reid i. n. , siegel m. , majewski s. r. , paust n. e. q. , aparicio a. , milone a. p. , chaboyer b. , rosenberg a. 2008 , aj , 135 , 2055 applegate j. h. 1986 , apj , 301 , 132 beccari g. , ferraro f. r. , possenti a. , valenti e. , origlia l. , rood r. t. 2006 , aj , 131 , 2551 beers m. e. , davies m. b. 2004 , mnras , 348 , 679 buonanno r. , corsi c. , bellazzini m. , ferraro f. r. , fusi pecci f. 1997 , aj , 113 , 706 carretta e. , bragaglia a. , gratton r. g. , recio - blanco a. , lucatello s. , dorazi v. , cassisi s. 2010 , a&a , 516 , 55 chernoff d. f. , weinberg m. d. 1990 , apj , 351 , 121 cohen r. l. , guhathakurta p. , yanny b. , schneider d. p. , bahcall j. n. 1997 , aj , 113 , 669 davies m. b. , piotto g. , de angeli f. 2004 , mnras , 348 , 129 de angeli f. , piotto g. , cassisi s. , busso g. , recio - blanco a. , salaris m. , aparicio a. , rosenberg a. 2005 , aj , 130 , 116 de marchi g. , paresce f. , portegies zwart s. 2010 , apj , 718 , 105 dervisoglu a. , tout c. a. , ibanoglu c. 2010 , mnras , 406 , 1071 djorgovski s. , meylan g. 1994 , aj , 108 , 1292 dotter a. , chaboyer b. , jevremovic d. , baron e. , ferguson j. w. , sarajedini a. , anderson j. 2007 , aj , 134 , 376 dotter , a. , sarajedini a. , anderson j. , aparicio a. , bedin l. r. , chaboyer b. , majewski s. , marin - franch a. , milone a. , paust n. , piotto g. , reid n. , rosenberg a. , siegel m. 2010 , apj , 708 , 698 farouki r. t. , salpeter e. e. 1982 , apj , 253 , 512 ferraro f. r. , messineo m. , fusi pecci f. , de palo m. a. , straniero o. , chieffi a. , limongi m. 1999 , aj , 118 , 1738 fregeau j. m. , cheung p. , portegies zwart s. f. , rasio f. a. 2004 , mnras , 352 , 1 fregeau j. m. , ivanova n. , rasio f. a. 2009 , apj , 707 , 1533 fukushige t. , heggie d. c. 1995 , mnras , 276 , 206 fusi pecci f. , ferraro f. r. , corsi c. e. , cacciari c. , buonanno r. 1992 , aj , 104 , 1831 fusi pecci f. , ferraro f. r. , bellazzini m. , et al .
1993 , aj , 105 , 1145 gieles m. , baumgardt h. , heggie d. , lamers h. 2010 , mnras , 408 , l16 gieles m. , heggie d. , zhao h. 2011 , mnras , accepted goldsbury r. , richer h. b. , anderson j. , dotter a. , sarajedini a. , woodley k. 2010 , aj , 140 , 1830 harris , w. e. 1996 , aj , 112 , 1487 ( 2010 update ) heggie d. c. , hut p. 2003
, the gravitational million - body problem : a multidisciplinary approach to sar cluster dynamics ( cambridge : cambridge university press ) henon m. 1960 , annales dastrophysique , 23 , 668 henon m. 1973 , dynamical structure and evolution of dense stellar systems , ed . l. martinet & m. mayor ( geneva obs . ) hut p. , bahcall j. n. 1983 , apj , 268 , 319 iannicola g. , monelli m. , bono g. , stetson p. b. , buonanno r. , calamida a. , zoccali m. , caputo f. , castellani m. , corsi c. e. , dallora m. , cecco a. di , deglinnocenti s. , ferraro i. , nonino m. , pietrinferni a. , pulone l. , moroni p. g. prada , romaniello m. , sanna n. , walker a. r. 2009 , apj , 696 , l120 iben i. jr .
1991 , apjs , 76 , 55 king i. r. , sosin c. , cool a. m. 1995 , apj , 452 , l33 knigge c. , leigh n. w. , sills a. 2009 , nature , 457 , 288 kruijssen j. m. d. , mieske s. 2009 , a&a , 500 , 785 lanzoni b. , dalessandro e. , perina s. , ferraro f. r. , rood r. t. , sollima a. 2007 , apj , 670 , 1065 lee h. m. , fahlman g. g. , richer h. b. 1991 , apj , 366 , 455 leigh n. w. , sills a. , knigge c. 2007 , apj , 661 , 210 leigh n. w. , sills a. , knigge c. 2008 , apj , 678 , 564 leigh n. w. , sills a. , knigge c. 2009 , mnras , 399 , l179 leigh n. w. , sills a. 2010 , mnras , 410 , 2370 leonard p.
j. t. 1989 , aj , 98 , 217 leonard p. j. t. , linnell a. p. 1992
, aj , 103 , 1928 li l. , zhang f. 2006 , mnras , 369 , 2001 mapelli m. , sigurdsson s. , ferraro f. r. , colpi m. , possenti a. , lanzoni b. 2006 , mnras , 373 , 361 mathieu r. d. , geller a. r. 2009 , nature , 462 , 1032 mccrea w. h. 1964 , mnras , 128 , 147 mclaughlin d. e. 2000 , apj , 539 , 618 mclaughlin d. e. , van der marel r. p. 2005 ,
apjs , 161 , 304 meylan g. , heggie d. c. 1997 , a&ar , 8 , 1 murray n. 2009 , apj , 691 , 946 pancino e. , seleznev a. , ferraro f. r. , bellazzini m. , piotto g. 2003 , mnras , 345 , 683 perets h. b. , fabrycky d. c. 2009 , apj , 697 , 1048 peterson r. c. , carney b. w. , dorman b. , green e. m. , landsman w. , liebert j. , oconnell r. w. , rood r. t. 2003 , apj , 588 , 299 piotto g. , king i. r. , djorgovski s. g. , sosin c. , zoccali m. , saviane i. , de angeli f. , riello m. , recio - blanco a. , rich r. m. , meylan g. , renzini a. 2002 , a&a , 391 , 945 portegies zwart s. f. , hut p. , makino j. , mcmillan s. l. w. 1998 , a&a , 337 , 363 rood r. t. 1973 , apj , 184 , 815 sandage a. r. 1953 , aj , 58 , 61 sandage a. , wallerstein g. 1960 , apj , 131 , 598 sandquist e. l. , bolte m. , hernquist l. 1997 , apj , 477 , 335 sandquist e. l. , martel a. r. 2007 , apjl , 654 , l65 sarajedini a. , bedin l. r. , chaboyer b. , dotter a. , siegel m. , anderson j. , aparicio a. , king i. , majewski s. , marin - franch a. , piotto g. , reid i. n. , rosenberg a. , steven m. 2007 , aj , 133 , 1658 saviane i. , piotto g. , fagotto f. , zaggia s. , capaccioli m. , aparicio a. 1998 , a&a , 333 , 479 shara m. m. , drissen l. , bergeron l. e. , paresce f. 1995 , apj , 441 , 617 sigurdsson s. , phinney e. s. 1995 , apjs , 99 , 609 sills a. , bailyn c. d. 1999 , apj , 513 , 428 sills a. r. , faber j. a. , lombardi j. c. , rasio f. a. , waren a. r. 2001 , apj , 548 , 323 sills a. , karakas a. , lattanzio j. 2009 , apj , 692 , 1411 spitzer l. jr .
, harm r. 1958 , apj , 127 , 544 spitzer l. jr .
1969 , apj , 168 , l139 spitzer l. jr .
, hart m. h. 1971 , apj , 166 , 483 spitzer l. jr .
, shull j. m. 1975 , apj , 200 , 339 spitzer l. 1987 , dynamical evolution of globular clusters ( princeton : princeton university press ) trenti m. , vesperini e. , pasquato m. 2010 , apj , 708 , 1598 webbink r. f. structure parameters of galactic globular clusters . in goodman , j. & hut , p. ( eds . )
dynamics of star clusters , vol .
113 of iau symposium , 541 - 577 ( 1985 ) .
zinn r. , barnes s. 1996 , aj , 112 , 1054 | we present a homogeneous catalogue for blue straggler , red giant branch , horizontal branch and main - sequence turn - off stars in a sample of 35 clusters taken from the acs survey for globular clusters . as a result of the superior photometry and relatively large field of view offered by the acs data ,
this new catalogue is a significant improvement upon the one presented in @xcite . using our catalogue ,
we study and compare the radial distributions of the different stellar populations .
we have confirmed our previous result @xcite that there is a clear , but sub - linear , correlation between the number of blue stragglers found in the cluster core and the total stellar mass contained within it . by considering a larger spatial extent than just the core
, our results suggest that mass segregation is not the dominant effect contributing to the observed sub - linearity .
we also investigate the radial distributions of the different stellar populations in our sample of clusters .
our results are consistent with a linear relationship between the number of stars in these populations and the total mass enclosed within the same radius .
therefore , we conclude that the cluster dynamics does not significantly affect the relative distributions of these populations in our sample .
[ firstpage ] stars : blue stragglers globular clusters : general stellar dynamics stars : statistics catalogues . |
the first year wmap data has confirmed the `` concordance '' values of the cosmological parameters with unprecedented accuracy and given important information on the primordial spectrum of density perturbations @xcite .
their results favor gaussian , purely adiabatic fluctuations and a spectral index that runs from @xmath4 on large scales to @xmath5 on small scales .
moreover , wmap has confirmed earlier cobe dmr observations that there is a lower amount of power on the largest scales when compared to that predicted by the standard @xmath6cdm models . although these results are not yet firmly established ( for an analysis of wmap results which finds no evidence of running see ref .
@xcite ) , it seems worthwhile to reexamine inflationary models in light of wmap results , as they may give us further insight into the very early universe .
supergravity inflationary models are particularly important as supersymmetry ( or its local version , supergravity ) is the only known way to avoid the hierachy problem , i.e. the fact that the high energy scale of inflation communicates to other sectors of the theory driving the electroweak scale much above its observed value via radiative corrections .
however , supergravity inflationary models also suffer from a kind of hierarchy problem as supersymmetry is broken by the large cosmological constant during inflation giving all scalars , including the inflaton , a soft mass of the order of the hubble parameter @xcite . as a result ,
the curvature of the inflaton potential , as measured by the @xmath7 slow - roll parameter , becomes too large to allow for a sufficiently long period of inflation to take place - the so - called @xmath7 problem .
recently , it has been shown that this problem can be avoided within the randall - sundrum type ii braneworld scenario @xcite , at least for a class of supergravity models in which the evolution of the inflaton dynamics is controlled by a single power at the point where the observed density fluctuations are produced and the inflationary potential can , therefore , be approximately given by v ^4 , [ eq : gpot ] where @xmath8 is the reduced planck mass .
in the braneworld context , the friedmann equation acquires an additional term quadratic in the energy density @xcite h^2 = 8 3 m_p^2 , [ eq : h22 ] where @xmath9 is the brane tension , which relates the four and five - dimensional planck scales through m_p = .
[ eq : mp ] it is precisely the new parameter @xmath10 that plays a crucial role in the resolution of the @xmath7-problem in supergravity inflation . as shown in ref .
@xcite , for the case where the first term in eq .
( [ eq : gpot ] ) is dominant and @xmath11 or @xmath12 , this problem can be avoided provided the five - dimensional planck mass satisfies , respectively , the condition @xmath13 and @xmath14 .
the case where the second term is dominant and @xmath11 , corresponding to chaotic inflation , has been studied in refs .
@xcite , where it is shown that it is possible to achieve successful inflation with sub - planckian field values , thereby avoiding well known difficulties with higher order non - renormalizable terms . in this paper , we reexamine this class of supergravity models for the quadratic and cubic cases in light of wmap results .
the case of chaotic inflation has already been analysed in ref .
@xcite , with the conclusion that the quadratic potential is allowed at two - sigma for any value of the brane tension and the quartic potential is very constrained , particularly in the case where the inflationary energy scale is close to the brane tension . here
we concentrate on the case where the first term in eq .
( [ eq : gpot ] ) is dominant .
we find that although the running of the scalar spectral index is within the bounds determined by wmap for both the quadratic and cubic models and for a wide range of potential parameters , a spectral index running from @xmath4 on large scales to @xmath5 on small scales is not possible in either case as precisely the opposite trend is found .
we first consider the case where the potential is quadratic in the inflaton field , @xmath15 , and we rewrite it as v= v_0 + 12 m^2 ^2 , [ eq : v2 ] and assume that the first term is dominant . in supergravity ,
effective mass squared contributions of fields are given by m^2= 8 3 h^2 , [ scond ] since the horizon of the inflationary de sitter phase has a hawking temperature given by @xmath16 @xcite .
contributions like the ones of eq .
( [ scond ] ) lead to @xmath17 ; however , the onset of inflation requires @xmath18 . within the braneworld scenario , however , @xmath7 and the remaining `` slow - roll '' parameters @xmath19 and @xmath20 ,
are modified , at high energies , by a factor proportional to @xmath21 @xcite @xmath22 in the high energy approximation , @xmath23 , we obtain , for this model @xmath24 where we have also used the approximation @xmath25 during inflation and the definition @xmath26 . as shown in ref .
@xcite , if @xmath27 is sufficiently large , the @xmath7-problem is automatically solved by the brane correction .
the number of @xmath28-folds during inflation , @xmath29 , in the braneworld scenario , is given by @xcite [ eq : n ] n - 8m_p^2__i^ _ fvv d , in the slow - roll approximation .
we see that , as a result of the modification in the friedmann equation , the expansion rate is increased , at high energies , by a factor @xmath30 . for this model , we get , in the high - energy approximation , n = & & 4 ( _ i_f ) + 2 m_p^2(_i^2-_f^2 ) + & & + 16 ^2 m_p^4(_i^4-_f^4 ) .
[ eq : n2 ] notice that , for sub - planckian field values , the second and third terms are negligible
. the value of @xmath15 at the end of inflation can be obtained from the condition \{(_f),|(_f)|}= 1 .
[ eq : phif ] however , we shall consider @xmath31 as a free parameter since quadratic potentials of the type we are studying arise typically in the context of hybrid inflation , once some other field is held at the origin by its interaction with @xmath15 . in these scenarios
, inflation may end due to instabilities triggered by the dynamics of the other field and , therefore , the amount of inflation strongly depends on the value of the inflaton field at the end of inflation , @xmath32 , at the time the instabilities arise .
actually , these instabilities are necessary in order to end inflation as @xmath33 for @xmath34 and sub - planckian field values .
0.1 cm the amplitude of scalar perturbations is given by @xcite [ eq : as ] a_s^2 .
( 51275 m_p^6 ) v^3
v^2 ^ 3 |_k = ah , where the right - hand side should be evaluated as the comoving wavenumber equals the hubble radius during inflation , @xmath35 .
thus the amplitude of scalar perturbations is increased relative to the standard result at a fixed value of @xmath15 for a given potential . using the high energy approximation and @xmath36 in eq .
( [ eq : as ] ) , we obtain a_s^2 (-8 n_k ) , [ eq : as2 ] where @xmath37 is the number of e - folds between the time the scales of interest leave the horizon and the end of inflation .
the scale - dependence of the perturbations is described by the spectral tilt @xcite n_s - 1 -6 + 2 , [ eq : ns ] which , for this model , gives @xmath38 the `` running '' of the scalar spectral index is given by _
s = 16 - 24 ^ 2 - 2 , [ eq : alfas ] and we get @xmath39~.\ ] ] the amplitude of tensor perturbations is given by @xcite a_t^2=.64150m_p^4 v ( 1+v2 ) f^2 |_k = ah , where f^2=^-1 , and s^1/2 .
in the low energy limit ( @xmath40 ) , @xmath41 , whereas @xmath42 in the high energy limit . defining ( we choose the normalization of ref .
@xcite ) r_s16 a_t^2a_s^2 , we obtain r_s0.06 m_p^4 m^4 ^ 2v_0 ^ 3 ( 8 n_k ) .
= 7.5 cm 0.1 cm wmap bounds on the above inflationary observables are , for this class of models ( case @xmath43 , class d in ref .
@xcite ) & & 0.99 < n_s < 1.28 , -0.09_s0.03 , + & & r_s 1.10 .
[ eq : observations2 ] these bounds refer to the scale best probed by the cmb observations i.e. @xmath44 mpc@xmath45 ; accordingly , we set @xmath46 . on the other hand , bounds on @xmath47 from other experiments
are less blue , e.g. the combined data sets from boomerang , cbi , dasi , dmr , maxima , toco and vsa give @xcite : 0.955 < n_s < 1.05 .
[ eq : comb ] in figure [ fig : figure1 ] , we show contours of the observational bounds on the inflationary observables @xmath47 , @xmath48 and @xmath49 in the ( @xmath27 , @xmath50 ) parameter space , for different values of @xmath51 ; in these plots , we have taken the bounds of eq .
( [ eq : observations2 ] ) except for the upper bound on @xmath47 , for which we took the bound of eq .
( [ eq : comb ] ) instead since such a blue spectrum is not to be expected .
we also plotted the @xmath52 contour , which shows that @xmath49 is required to be positive for this model .
the shaded area corresponds to the allowed region in parameter space .
we would also like to mention that we have checked that it is possible to obtain sufficient inflation with sub - planckian field values e.g @xmath53 for @xmath54 and @xmath55 within the range specified in figure 1 .
0.1 cm notice that the contour corresponding to the upper bound on @xmath49 is too small to be visible in figure [ fig : figure1 ] , hence we show it in figure [ fig : figure2 ] , for @xmath56 ( similar behavior is obtained for other values of @xmath51 ) , which makes it clear that this bound plays an important role in constraining the parameter space .
we obtain lower bounds on @xmath27 and @xmath50 , namely @xmath57 , pratically independent of @xmath51 , whereas the lower bound on @xmath50 ranges from @xmath58 to @xmath59 as @xmath51 ranges from 50 to 75 . in figures [ fig : figure2 ] and [ fig : figure3 ] , we have superposed contours of the scale @xmath10 , as derived from eqs .
( [ eq : as2 ] ) and ( [ eq : mp ] ) , where we have used the cobe normalization i.e. @xmath60 for @xmath61 . as the allowed region is quite narrow for low values of @xmath27 , it allows us to find an upper bound on @xmath10 , @xmath62 m , which we have checked is almost independent of @xmath51 . combining the above results
, we find a lower bound on the scale @xmath63 , namely @xmath64 m. notice that we have chosen to vary @xmath51 since , although a wide variety of assumptions about @xmath51 can be found in the literature , the determination of this quantity requires a model of the entire history of the universe .
however , while from nucleosynthesis onwards this is now well established , at earlier epochs there are considerable uncertanties such as the mechanism ending inflation and details of the reheating process .
this issue was recently reviewed in ref .
@xcite ( see also ref .
@xcite for similar results ) , where a model - independent upper bound was derived , namely @xmath65 ; in fact , @xmath56 is found to be a reasonable fiducial value with an uncertainty of around 5 around that value ; however , the authors stress that there are several ways in which @xmath51 could lie outside that range , in either direction .
moreover , in the braneworld context , one expects @xmath51 to depend on the brane tension .
actually , one expects to obtain larger values of @xmath51 because , in the high - energy regime , the expansion laws corresponding to matter and radiation domination are slower than in the standard cosmology , which implies a greater change in @xmath66 relative to the change in @xmath67 , therefore requiring a larger value of @xmath51 .
this is confirmed by the results of ref .
@xcite , where the bound @xmath68 is found for brane inspired cosmology .
we have studied the dependence of @xmath47 on @xmath69 to see whether @xmath47 can vary from @xmath4 to @xmath5 from large to small scales . from the condition @xmath70 , assuming that h is approximatly constant during inflation , we obtain the relation ( n_k)=k_fk , [ eq : nkvk ] where @xmath71 is the value of @xmath69 at the end of inflation . inserting this relation is eq .
( [ eq : ns2 ] ) , we obtain n_s1 + 8 - 192 ^2 ^ 2(k_f k)^8/ , [ eq : nsk ] from which we conclude that , for @xmath55 fixed , @xmath47 increases with @xmath69 .
hence , it is not possible to obtain the desired behaviour , i.e. @xmath47 decreasing from @xmath4 to @xmath5 as @xmath69 increases .
we shall now consider the case where , due to some cancellation mechanism @xcite , the quadratic term is absent and the potential is cubic in @xmath15 : v=^4 .
[ eq : v3 ] as mentioned before , we shall assume that the first term is dominant .
the parameter @xmath72 is expected to be of order unity and negative @xcite ; the model of ref .
@xcite corresponds to precisely this case , with @xmath73 .
we start by computing the slow - roll parameters : & & 18 ^2 ( m)^4 , + & & 12 ( m ) , + & & 72 ^2 ^2 , [ eq : slowr3 ] where @xmath74 . the value of @xmath15 at the end of inflation can be obtained from eq .
( [ eq : phif ] ) ; we get , from @xmath75 _ f ( 18 ^2)^1/4 m , [ eq : phif3 ] while , from @xmath76 , we obtain _ f ( 12 || ) m .
[ eq : phif4 ] hence , the prescription to be used depends on the value of @xmath27 . for @xmath77 , we see that the two prescriptions coincide for @xmath78 .
the number of @xmath28-foldings , @xmath29 , is given by : n = m 6 || .
[ eq : n3 ] therefore , sufficient inflation to solve the cosmological horizon / flatness problems , that is @xmath79 , is achieved , for instance for @xmath73 , if @xmath80 . for @xmath81
, we obtain , in the high energy regime , a_s^2 , [ eq : as3 ] where @xmath82 is the value of @xmath15 at horizon - crossing . the scalar spectral index and its running can be readily computed from the slow - roll parameters , eq .
( [ eq : slowr3 ] ) , via eqs.([eq : ns ] ) and ( [ eq : alfas ] ) .
notice that the inflationary observables can , of course , be written as a function of @xmath51 , as for the quadratic model , using eq .
( [ eq : n3 ] ) with @xmath83 , but one has to bear in mind that the prescription to use for @xmath32 depends on @xmath27 .
wmap bounds on the inflationary observables are , for this class of models ( case @xmath84 , class a in ref .
@xcite ) & & 0.94 < n_s < 1.00 , -0.02_s0.02 , + & & r_s 0.14 , [ eq : observations ] again for the scale @xmath44 mpc@xmath45 . in figure
[ fig : figure4 ] , we show contours of the inflationary observable @xmath47 , @xmath49 in the ( @xmath85 ) plane .
we have checked that neither @xmath49 nor @xmath48 give further constraints on the parameter space .
we also show contours corresponding to different values of @xmath10 , as given by eq .
( [ eq : as3 ] ) , again cobe normalized .
the upper left panel corresponds to the results for the low energy regime , @xmath86 , where the brane corrections are negligible , and the remaining three panels correspond to the high energy regime , @xmath23 , for different values of @xmath87 .
we see that it is the lower bound on @xmath47 that most constrains the model and , clearly , @xmath88 can not be obtained .
it is also clear that , for the model to work , @xmath89 is required if brane corrections are not included and @xmath90 if those corrections are included ; in the latter case , however , this bound increases outside the range @xmath91 , for @xmath73 ( this range is slightly @xmath87-dependent , see figure [ fig : figure4 ] ) .
moreover , the running parameter is always negative although it can be quite small .
finally , @xmath92 , for @xmath73 ( however , these bounds do not change significantly with @xmath87 , see figure 4 .
clearly , the spectral index can not run from @xmath4 on large scales to @xmath5 on small scales , since @xmath93 for this model . in ref .
@xcite , a very strict bound on @xmath27 was derived for this model from the requirement that the reheating temperature is small enough to avoid the gravitino problem .
we should like to point out that there was a numerical error in that computation and , in fact , the bound is much weaker and pratically meaningless .
we have analysed the implications of wmap results , in particular the bounds on the inflationary observables , for a class of supergravity inflationary models , eq .
( [ eq : gpot ] ) with @xmath94 .
we find that , for the quadratic potential , the main constraints come from the wmap s bounds on @xmath47 and upper bound on @xmath49 .
we have obtained lower bounds on parameters @xmath27 and @xmath95 , namely @xmath96 ( pratically independent of @xmath51 ) and the lower bound on @xmath50 ranges from @xmath97 to @xmath59 as @xmath51 varies between 50 and 75 .
we have also found an upper bound on @xmath10 , @xmath62 m , pratically independent of @xmath51 .
moreover , we conclude that @xmath98 is required for this model . for the cubic potential , in the low energy regime i.e. without the brane correction , a relatively high value of @xmath51 , @xmath99 , is required so as to meet wmap s lower bound on @xmath47 . in the high energy regime , when brane corrections are significant , the allowed region in the ( @xmath85 ) parameter space changes with @xmath87 and the main constraints come from wmap s lower bound on @xmath47 .
moreover , we find that @xmath100 for this model m.c.b .
acknowledges the partial support of fundao para a cincia e a tecnologia ( portugal ) under the grant pocti/1999/fis/36285 . the work of a.a .
sen is fully financed by the same grant .
santos is supported by fct grant sfrh / bd/4797/2001 . | we study a class of @xmath0 supergravity inflationary models in which the evolution of the inflaton dynamics is controlled by a single power in the inflaton field at the point where the observed density fluctuations are produced , in the context of the braneworld scenario , in light of wmap results .
in particular , we find that the bounds on the spectral index and its running constrain the parameter space both for models where the inflationary potential is dominated by a quadratic term and by a cubic term in the inflaton field .
we also find that @xmath1 is required for the quadratic model whereas @xmath2 for the cubic model .
moreover , we have determined an upper bound on the five - dimensional planck scale , @xmath3 m , for the quadratic model . on the other hand , a running spectral index with @xmath4 on large scales and @xmath5 on small scales
is not possible in either case
. 0.5 cm 0.5 cm |
a family of geometrically integral algebraic varieties defined over @xmath0 is said to satisfy the hasse principle if any variety in the family has a point in @xmath0 as soon as it has points in every completion of @xmath0 .
quadrics are among the first examples of families satisfying this property . in dimension
@xmath1 , chtelet surfaces constitute a family of varieties for which the hasse principle is known to fail in general .
the chtelet surfaces that we consider are proper smooth models of affine surfaces @xmath2 where @xmath3 $ ] is a separable polynomial of degree @xmath4 or @xmath5 . for these surfaces it follows from work of colliot - thlne , sansuc and swinnerton - dyer @xcite that all failures of the hasse principle are accounted for by the brauer manin obstruction , a cohomological obstruction based on the brauer group of the surface .
in particular it is known that the hasse principle holds whenever @xmath6 is irreducible over @xmath0 , which is the generic situation , or when @xmath6 has a linear factor over @xmath0 . in the remaining case ,
when @xmath6 is a product of two irreducible quadratic polynomials over @xmath0 , counter - examples to the hasse principle can arise .
a great deal of recent work has been directed at the quantitative arithmetic of rational varieties , the aim being to count @xmath0-rational points of bounded height on the variety , assuming that it contains a zariski dense set of @xmath0-rational points . in this paper
we seek instead to vary the varieties in a family and measure how often counter - examples to the hasse principle emerge . in the setting of chtelet surfaces
we have already noted that a random surface of the form will satisfy the hasse principle . with this in mind , we place ourselves in the family where @xmath6 factorises as a product of two diagonal quadratic polynomials over @xmath0 , which will generically be irreducible over @xmath0 . for any @xmath7 ,
let @xmath8 denote the chtelet surface defined by the equation @xmath9 with @xmath10 the fact that the brauer manin obstruction is the only obstruction to the hasse principle for this family goes back to earlier work of colliot - thlne , coray and sansuc @xcite , wherein it is shown that @xmath11 .
furthermore , in ( * ? ? ?
c ) , a @xmath12-parameter family of counter - examples to the hasse principle is exhibited .
this family , which we will meet again in
[ s : ct ] , is given by @xmath13 for any positive integer @xmath14 .
it generalises the particular case @xmath15 first discovered by iskovskikh @xcite .
our chief object is to give a finer quantitative treatment of this circle of ideas .
as we vary over rational coefficients , producing surfaces @xmath8 that are defined over @xmath0 , we will investigate the proportion of surfaces that * have points everywhere locally , * have @xmath0-rational points , * fail the hasse principle .
one of the issues that we shall need to address is how best to parameterise the chtelet surfaces that are of interest to us .
poonen and voloch @xcite have discussed similar questions in the setting of projective hypersurfaces @xmath16 of degree @xmath17 , assuming that @xmath18 and @xmath19 .
let @xmath20 denote the set of valuations of @xmath0 and let @xmath21 be the completion of @xmath0 at @xmath22 , following the convention that @xmath23 . in ( * ?
3.6 ) it is shown that the proportion of @xmath24 that are everywhere locally soluble converges to an euler product @xmath25 , where each @xmath26 is the proportion of @xmath24 for which @xmath27 .
this is the exact analogue of our theorem [ c : cor1 ] for hypersurfaces .
they conjecture , furthermore , that when @xmath28 the proportion of @xmath24 that have @xmath0-rational points should converge to this constant @xmath29 .
this is shown ( see ( * ? ? ?
3.4 ) ) to follow from the conjecture of colliot - thlne that the brauer manin obstruction to the hasse principle is the only one for smooth , proper , geometrically integral varieties over @xmath0 which are geometrically rationally connected .
in fact , for non - singular @xmath30 , with @xmath31 , the natural map @xmath32 is an isomorphism .
thus the brauer manin obstruction to the hasse principle is generically empty when @xmath31 .
bhargava @xcite has undertaken an extensive investigation of the case @xmath33 , where it is well - known that the hasse principle can fail . in this setting one
gets different behaviour to that predicted for fano hypersurfaces .
the main result of @xcite shows that that the hasse principle fails for a positive proportion of plane cubics defined over @xmath0 .
a hypersurface of degree @xmath17 in @xmath34 has @xmath35 possible coefficients .
poonen and voloch use affine space @xmath36 to parameterise these hypersurfaces , associating to each vector in @xmath37 a hypersurface @xmath24 defined over @xmath0 .
one disadvantage of this approach is that two different vectors may produce the same @xmath24 .
in our work we will handle this issue by identifying a parameter space for the chtelet surfaces @xmath8 in .
consider the open set of @xmath7 for which holds .
it will be convenient to work with the affine equation @xmath38 rather than .
we will consider two equations as leading to the same chtelet surface if there is @xmath39-transformation in the variables @xmath40 which takes one into the other . on replacing @xmath41 by @xmath42 , for example
, we see that @xmath43 , for any @xmath44 .
thus we are led to consider instead the open set @xmath45\in { \mathbb{p}}^3 : \mbox{\eqref{eq : stipulate } holds}\}.\ ] ] in a similar fashion , it is not hard to see that @xmath46 , for any @xmath47 .
furthermore , for @xmath48 , we have @xmath49 , where @xmath50 is the norm of any @xmath51 .
the underlying transformation in the latter case has matrix @xmath52 , with @xmath53 in conclusion , we have produced an action of @xmath54 on @xmath55 , where @xmath56 denotes the linear algebraic group @xmath57 $ ] associated to a field @xmath58 . building on this , we note that @xmath59 , where @xmath60 this leads to a further action on @xmath55 by the @xmath61-dimensional algebraic group @xmath62 .
we therefore obtain a group action @xmath63 , where @xmath64 is the algebraic @xmath0-group @xmath65 given @xmath66\in u$ ] , the orbit of @xmath67 under @xmath68 produces the same surface @xmath8 .
in our analysis we will restrict attention to elements of @xmath69
. we will be interested in three basic subsets , given by @xmath70\in { \mathcal{m } } : x_{a , b , c , d}({\mathbb{q}}_v ) \neq \emptyset ~\forall v\in \omega \right\},\\ { \mathcal{m}}_{\mathrm{glob } } & = \left\ { [ a , b , c , d]\in { \mathcal{m } } : x_{a , b , c , d}({\mathbb{q } } ) \neq \emptyset \right\},\\ { \mathcal{m}}_{\mathrm{br } } & = { \mathcal{m}}_{\mathrm{loc}}\setminus { \mathcal{m}}_{\mathrm{glob}}.\end{aligned}\ ] ] the elements of @xmath71 are precisely the surfaces for which there is a non - empty brauer manin obstruction to the hasse principle . using the naive exponential height function @xmath72 on @xmath73 , with norm @xmath74 on @xmath75
, we wish to study the cardinalities @xmath76 as @xmath77 .
our first result is the following .
[ t : m - local ] we have @xmath78 for any @xmath79 , where if @xmath80 is the density of points in @xmath81 for which the associated surface has @xmath21-points then @xmath82 here , as throughout our work , the implied constant is allowed to depend at most on the choice of parameter @xmath79 .
theorem [ t : m - local ] will be established in
[ s : asymptotic - nloc ] , where an explicit description of the factors @xmath80 will appear . once combined with an estimation of the total number @xmath83 of elements in @xmath84 with height at most @xmath85
, we will also establish the following result .
[ c : cor1 ] we have @xmath86 where @xmath87 in particular 83.3% of the elements of @xmath84 are everywhere locally soluble .
in [ s : s - global ] we will characterise when a surface @xmath88 in @xmath89 actually belongs to @xmath90 .
the brauer group @xmath91 will play a fundamental rle in this analysis , for which purpose we recall some basic facts here .
for any field @xmath92 , each element @xmath93 gives rise to an evaluation map @xmath94 .
class field theory gives the exact sequence @xmath95 for any @xmath93 we then have the commutative diagram @xmath96 \ar[d]^{\mathrm{ev}_{\mathcal{a } } } & x({\mathbb{a}}_{\mathbb{q } } ) \ar[d]^{\mathrm{ev}_{\mathcal{a } } } \\
\operatorname{br}({\mathbb{q } } ) \ar[r ] & \bigoplus_v \operatorname{br}({\mathbb{q}}_v ) \ar[r ] & { \mathbb{q}}/{\mathbb{z}}}\ ] ] where @xmath97 denotes the adles and @xmath98 .
let @xmath99 denote the composed map .
then it follows that @xmath100 for all @xmath93 .
we write @xmath101 in our setting , @xmath102 has order @xmath1 and so @xmath103 takes the values @xmath61 or @xmath104 in @xmath105 an obstruction to the hasse principle arises if and only if @xmath106 , that is to say , if and only if for any @xmath107 there exists @xmath93 such that @xmath108 moreover , a generator for @xmath109 is given by the quaternion algebra @xmath110 on .
fundamental to our work is the observation that is impossible if there exists a valuation @xmath22 for which @xmath103 takes both values @xmath61 and @xmath104 , as a function of @xmath111 .
thus , in order to obtain counter - examples to the hasse principle , it will be necessary for @xmath112 to be locally constant .
in [ s : lower - upper ] we shall build on [ s : s - global ] to establish the following asymptotic formula for the density of surfaces which provide counter - examples to the hasse principle .
[ thm2 ] let @xmath79 .
there exists @xmath113 such that @xmath114 it follows from theorems [ t : m - local ] and [ thm2 ] that @xmath115 satisfies the asymptotic behaviour predicted by poonen and voloch @xcite for fano hypersurfaces .
moreover , once coupled with theorem [ c : cor1 ] , theorem [ thm2 ] shows that 83.3% of the elements of @xmath84 are soluble over @xmath0 .
it should be stressed , moreover , that our argument provides a completely explicit algorithm for determining whether or not a given chtelet surface @xmath8 of the form gives a counter - example to the hasse principle , without explicitly needing to work with elements of @xmath116 .
the authors are grateful to professors colliot - thlne and fouvry for suggesting this problem to us , to pierre le boudec for comments on an earlier draft and to tim dokchitser for help with verifying numerically the proof of lemma [ lem : tau-2 ] . while working on this paper the first author
was supported by an _
iuf junior _ and _ anr project ( pepr ) _ , while the second author was supported by _
erc grant _ 306457 .
we will need to choose representative coordinates for the parameter space @xmath84 that we met in the introduction .
two sets that will feature heavily in our work are @xmath117 where @xmath118 denotes the mbius function .
let @xmath119 denote the set of relatively prime @xmath5-tuples of integers .
we define @xmath120 it is clear , in view of the invariance under the action of the maps , that we have a @xmath121 bijection between @xmath84 and our set of representatives @xmath122 .
when it comes to determining whether or not elements of @xmath122 produce surfaces @xmath8 with @xmath1-adic points , a tedious number of subcases arise , depending on the residue classes modulo @xmath5 of the coefficients . in order to reduce the number of cases that need to be considered we note that for any @xmath123 , we have @xmath124 . but
@xmath125 is left invariant under the action of the map @xmath126 defined in .
hence we may write @xmath127 the elements of @xmath122 have non - zero coordinates .
since we identify @xmath128 with @xmath129 in @xmath122 , it will suffice to take representative coordinates in which @xmath130 .
the sets in which we are interested therefore take the shape @xmath131 for @xmath132 . among the elements of these sets
we will be interested primarily in those coefficients which give rise to chtelet surfaces @xmath8 which have points everywhere locally , or points globally .
let us therefore define the sets @xmath133 for @xmath132 .
having defined suitable representatives for the spaces parameterising our chtelet surfaces , we proceed to introduce some further notation . given @xmath134
, we define the binary quadratic forms @xmath135 let @xmath136 and @xmath137 .
then @xmath138 , with @xmath139 .
we henceforth write @xmath140 where @xmath141 and @xmath142 in particular we have @xmath143 and @xmath144 where @xmath145 . returning to the family of surfaces @xmath8 arising as proper smooth models of , in gauging solubility over the completed field @xmath21 , for any @xmath146
, it will suffice to work with a non - empty zariski open subset @xmath147 of @xmath8 .
indeed , according to ( * ? ? ?
* lemme 3.1.2 ) , the set @xmath148 has dense image in @xmath149 for the topology defined by the natural topology of @xmath21 . in this way
we see that it suffices to examine the local solubility of the affine equation @xmath150 where @xmath151 are as in .
our first order of business is a precise characterisation of when an element can locally be written as the sum of two squares .
let @xmath152 and let @xmath153 denote the non - zero elements in @xmath21 which can be written as a sum of two squares in @xmath21 .
firstly we note that @xmath154 .
when @xmath22 is a prime @xmath155 then @xmath156 .
next when @xmath22 is a prime @xmath157 then @xmath158 is the set of @xmath159 for which the @xmath160-adic valuation @xmath161 of @xmath162 is even .
finally when @xmath163 we have @xmath164 , where @xmath165 frequent use will be made of this characterisation in
[ s : s - local ] and [ s : s - global ] .
we will reserve the letter @xmath160 for denoting a prime number .
our work will involve various standard arithmetic functions , including @xmath166 , the generalised divisor function @xmath167 and the euler totient function @xmath168 .
in addition to this we will also meet the function @xmath169 for any @xmath170 , so that @xmath171 .
note that @xmath172 . finally , let @xmath173 we will typically abbreviate @xmath174 by @xmath175 and @xmath176 by @xmath177 .
we will also require an estimate for the average order of the divisor function , as it ranges over the values of a binary linear form .
a crucial aspect of the following result is the uniformity in the coefficients of the linear form that it enjoys .
[ lem : divisor ] let @xmath178 with @xmath179 , let @xmath180 and let @xmath79 .
we have @xmath181 where the implied constant depends at most on @xmath182 .
suppose without loss of generality that @xmath183 .
the bound follows from previous work of the authors ( * ? ? ?
1 ) when @xmath184 .
alternatively it follows from the trivial bound @xmath185 .
we close this section with a modified result of le boudec ( * ? ? ?
* lemma 2 ) , concerning pairs of integers constrained to satisfy congruences modulo @xmath186 and modulo @xmath187 , with @xmath188 .
we are interested in the cardinality @xmath189 for suitable @xmath190 .
the following result allows us to approximate @xmath191 by the expected main term .
[ lem : lb ] let @xmath178 with @xmath179 and @xmath192 , let @xmath193 and let @xmath194 .
we have @xmath195 where @xmath196 as we have mentioned , this result is based on work of le boudec ( * ? ? ?
* lemma 2 ) , which corresponds to @xmath197 and general closed sets @xmath198 in place of @xmath199 $ ] and @xmath200.$ ] we have decided to include a sketch of the proof for completeness , the key idea being to break the @xmath201 into congruence classes modulo @xmath202 and then use the orthogonality of additive characters @xmath203 to detect these congruences .
since @xmath186 and @xmath187 are coprime , there exist @xmath204 such that @xmath205 .
we therefore have @xmath206 where @xmath207 here , recalling that @xmath208 , we have @xmath209 where @xmath210 is the ramanujan sum , satisfying @xmath211 . in particular @xmath212 and @xmath213 are seen to be independent of @xmath214 .
we denote by @xmath215 the contribution from @xmath216 or @xmath217 . then bounding the sums over @xmath218 using the standard bound for linear exponential sums
, we see that @xmath219 in the notation of the lemma . summing over @xmath220 we deduce that @xmath221 moreover , it is clear that @xmath222 note that @xmath223 . combining this with we swiftly arrive at the statement of the lemma .
as we have seen , in order to determine whether or not @xmath8 is soluble over @xmath21 , for any @xmath146 , it will suffice to work with the equation .
furthermore , when considering solubility over @xmath224 , it suffices to establish the existence of solutions @xmath225 for which @xmath226 are coprime . in order to summarise the local solubility criterion for @xmath8 it will be convenient to introduce a new symbol . for any odd prime @xmath160 and any @xmath227 with @xmath228 , we set @xmath229}= \begin{cases } \left(\frac{n / p^{\nu}}{p}\right ) , & \mbox{if $ 2\mid \nu$},\\ 1 , & \mbox{if $ 2\nmid \nu$ } , \end{cases}\ ] ] where @xmath230 is the usual legendre symbol .
this symbol takes values in @xmath231 .
finally , let @xmath232 denote the sign of any real number @xmath162 , which we extend to vectors in the obvious way .
we may now record the following result , in which @xmath136 and @xmath137 .
[ lem : x - local ] let @xmath134 . then the following hold : * @xmath233 if and only if @xmath234 ; * @xmath235 if @xmath236 ; * if @xmath237 for @xmath157 then @xmath235 if and only if * * @xmath238 or @xmath239 only when these @xmath160-adic orders are zero , * * either @xmath240}=1 $ ] or @xmath241}=1 $ ] ; * @xmath242 if and only if @xmath243 coprime @xmath244 such that @xmath245
. one recalls here that @xmath138 , with @xmath139 .
thus @xmath246 for any prime @xmath160 and @xmath235 whenever @xmath247 . that the three conditions are necessary is self - evident .
turning to sufficiency , we begin by establishing ( i ) , noting that @xmath248 for any @xmath134 . the case in which @xmath249 or @xmath250 is positive
is trivial , so we study only the case @xmath251 and @xmath252 , with ( i ) being satisfied . in this case @xmath253 and @xmath254 , whence @xmath255 condition ( iv ) is a direct consequence of the criterion for solubility of @xmath8 over @xmath256 .
solubility over @xmath224 is automatic for @xmath155 .
it therefore remains to consider solubility over @xmath224 when @xmath257 .
for ease of notation we will henceforth write @xmath258 for @xmath259 .
we claim that there exist coprime @xmath226 such that @xmath260 .
this will clearly suffice to show that @xmath261 is non - empty when @xmath236 , as required for condition ( ii ) . recall that @xmath262 and @xmath263 . if @xmath264 then the claim is satisfied by taking @xmath265 .
similarly , if @xmath266 then we may take @xmath267 .
the remaining cases to consider are when @xmath268 with @xmath269 , and lastly when @xmath270 with @xmath271 . for both of these
the choice @xmath272 is satisfactory , which thereby concludes the proof of the claim . for the remainder of the proof
we suppose that @xmath273 , with @xmath257 .
we wish to show that under the hypotheses of the lemma we can always find coprime @xmath226 such that precisely one of @xmath274 or @xmath275 is odd .
this will then imply that @xmath276 is even as required for condition ( iii ) .
suppose first that @xmath277 . then by hypothesis one of either @xmath278 or @xmath279 is a non - zero square modulo @xmath160 .
there are two cases to consider .
suppose first that @xmath280 but @xmath281 .
we can find coprime @xmath226 such that @xmath282 . on the other hand @xmath275
is even since @xmath279 is not a square modulo @xmath160 .
next suppose that @xmath283 , with @xmath284 .
we may choose coprime @xmath226 so that @xmath285 . then if @xmath286 one sees that @xmath287 is a divisor of @xmath288 by .
it therefore follows from that @xmath289 , as required .
we now suppose that @xmath290 but @xmath291 . if @xmath292 then it is clear that we can find @xmath226 , with @xmath293 , such that @xmath275 is odd .
moreover , we will have @xmath294 since @xmath160 divides precisely one of @xmath295 or @xmath296 .
if @xmath281 then @xmath297 for any coprime @xmath298 suppose @xmath299 , with @xmath300 and @xmath301 .
if @xmath302 is odd then we can ensure that @xmath274 odd by taking @xmath265 .
alternatively , if @xmath302 is even then @xmath303 by hypothesis .
hence we can force @xmath274 to be odd by considering solutions of the form @xmath304 .
the case in which @xmath305 but @xmath306 follows by symmetry . it remains to consider the possibility @xmath290 and @xmath306 .
we will consider the cases @xmath307 or @xmath308 , the remaining cases following by symmetry .
suppose first that @xmath300 and @xmath309 with @xmath310 . if @xmath302 ( resp .
@xmath311 ) is odd then we are done on taking @xmath265 ( resp .
@xmath267 ) .
suppose that @xmath302 and @xmath311 are both even .
then our hypothesis ensures that one of the legendre symbols @xmath312 or @xmath313 is equal to @xmath12 . supposing without loss of generality that it is the former , we easily make @xmath274 odd with @xmath297 .
turning to the case @xmath308 , we write @xmath300 and @xmath314 with @xmath315 and @xmath316 not both even .
if @xmath317 are of opposite parities then the situation is easy and we can proceed by taking @xmath265 . finally , if @xmath316 are both odd then by hypothesis they must be unequal .
suppose without loss of generality that @xmath318 .
then in fact @xmath319 .
taking @xmath320 we easily deduce that @xmath321 and @xmath322 , as required .
this concludes the proof of the lemma .
lemma [ lem : x - local ] gives us a means of characterising the elements of @xmath323 , in the notation of . before proceeding to the proof of theorem [ t : m - local ]
, we will need a finer description of case ( iv ) in lemma [ lem : x - local ] .
this will depend intimately on the possible residue classes of @xmath324 modulo @xmath5 , with @xmath325 odd .
the various constraints are obtained by taking @xmath326 in the second part of lemmas [ lem:1 - 1][lem:3 - 2 ] in [ s:2adic ] , where the description of @xmath327 exactly provides necessary and sufficient conditions for the non - nullity of @xmath328 .
throughout this section we will assume familiarity with the notation introduced in
[ s : prelim ] . among the set of @xmath329 we require a means of sifting for those @xmath128 such that @xmath8 actually has points in @xmath0 .
our principal tool comes from the theory of descent , as formulated by colliot - thlne , coray and sansuc ( * ? ? ?
this ensures that given any @xmath329 , we have @xmath330 if and only if there exists @xmath331 , with @xmath332 , such that the smooth variety @xmath333 has points in @xmath21 for every @xmath152 .
this in turn is equivalent to the existence of @xmath334 , with @xmath335 , such that the corresponding pair of equations with @xmath336 replaced by @xmath337 is everywhere locally soluble .
note that , by ( * ? ? ?
* lemme 3.1.2 ) , we see that in checking local solubility for the pair of equations it suffices to replace @xmath338 and @xmath339 by @xmath340 and @xmath341 , respectively , as given by .
we proceed to simplify the set of allowable @xmath342 somewhat . on carrying out a suitable change of variables it suffices , without loss of generality , to consider the existence of @xmath334 for which @xmath343 for @xmath344 .
recall that @xmath138 , with @xmath139 , in the notation of .
suppose that @xmath335 for @xmath334 satisfying the above @xmath160-adic constraints .
if @xmath345 then we must have @xmath346 .
alternatively , if @xmath347 , then @xmath348 or @xmath349 .
given such @xmath342 let us put @xmath350 to be the product of primes @xmath347 for which @xmath351 .
then we may assume that @xmath352 and @xmath353 , where @xmath354 are square - free integers comprised of primes @xmath345 .
on multiplying the first equation by @xmath355 and the second by @xmath356 , and making a further change of variables , we see that it suffices to check local solubility with @xmath357 replaced by @xmath358 , where @xmath359 .
in particular , we have the relation @xmath360 .
we summarise our remarks in the following result .
[ lem:1 ] let @xmath329 .
then @xmath361 if and only if there exists @xmath362 and @xmath363 , with @xmath364 and @xmath365 such that the smooth variety @xmath366 has solutions in @xmath21 for every @xmath146 .
the thrust of lemma [ lem:1 ] is that our task of determining elements of @xmath367 in has become a purely local problem .
in fact , the varieties @xmath368 are known to satisfy the hasse principle and any @xmath0-point on @xmath368 gives rise to a @xmath0-point on @xmath8 .
we shall henceforth refer to @xmath368 as torsor equations .
in reality , as explained in @xcite , any universal torsor over @xmath8 is birational to @xmath369 , for a smooth conic @xmath370 defined over @xmath0 . we proceed to study the sets @xmath371 for each @xmath152 . when determining solubility over @xmath224 it will suffice to consider the existence of solutions @xmath372 with @xmath226 coprime .
recall from
[ s : intro ] our discussion of the brauer group @xmath116 , which is generated by a single element , given on by the quaternion algebra @xmath373 .
given any local point @xmath374 in @xmath375 , we have @xmath376 according to our discussion of , in order for counter - examples to the hasse principle to arise , we will need @xmath103 to be locally constant for each @xmath377 , in which case we will show that there is a unique torsor @xmath378 which has @xmath21-points .
it will be convenient to retain our shorthand notation @xmath258 for @xmath259 in what follows .
the case in which @xmath160 is an odd prime not appearing in the factorisation of @xmath379 is straightforward , as the following result shows .
[ lem : p1 ] let @xmath362 and let @xmath380 , with @xmath381 . then @xmath382 .
if @xmath383 then the conclusion is trivial . suppose now that @xmath257 .
the second paragraph in the proof of lemma [ lem : x - local ] shows that there exists @xmath226 such that @xmath260 .
but then @xmath384 for @xmath344 , which concludes the proof .
since @xmath385 are square - free and entirely composed of primes congruent to @xmath4 modulo @xmath5 , lemma [ lem : p1 ] completely handles the solubility question over @xmath224 for @xmath386 . turning to @xmath387 , it remains to consider the possibility that @xmath388 , together with the possibility that @xmath347 and @xmath389 .
we may assume that @xmath390 with @xmath391 and @xmath392 .
then the condition demands that @xmath393 in particular we have @xmath394 when @xmath236 and @xmath395 when @xmath237 .
recall that @xmath145 .
we proceed to establish the following result .
[ lem : p2 ] let @xmath329 and let @xmath257 with @xmath396 .
let @xmath397 with @xmath392 .
then one of the following must hold : * if @xmath398+[\frac{-c'd'}{p}]{\leqslant}0 $ ] then there exists a unique @xmath399 with @xmath400 such that @xmath401 ; * if @xmath398=[\frac{-c'd'}{p}]=1 $ ] and @xmath402 , then @xmath403 ; furthermore , @xmath404 if and only if @xmath405 ; * if @xmath398=[\frac{-c'd'}{p}]=1 $ ] and @xmath406 , then @xmath407 unless @xmath408 is even and @xmath409 , or @xmath410 is even and @xmath411 ; likewise , @xmath412 unless @xmath413 is even and @xmath414 , or @xmath415 is even and @xmath416 .
in particular , the nullity of @xmath417 depends only on @xmath418 and @xmath419 .
[ rem:1 ] let @xmath160 be an odd prime .
lemmas [ lem : p1 ] and [ lem : p2 ] together show that whenever @xmath235 , there exists @xmath420 , with @xmath421 , such that @xmath422 .
our arguments will have a similar flavour to the proof of lemma [ lem : x - local ] .
beginning with case ( i ) we suppose without loss of generality that @xmath423=-1 $ ] .
now any solution @xmath218 must have @xmath424 , whence the choice @xmath425 ( resp .
@xmath426 ) is not admissible when @xmath402 ( resp .
@xmath406 ) .
finally we note that the case @xmath427 is admissible when @xmath402 by lemma [ lem : p1 ] , and the case @xmath428 is admissible when @xmath406 since lemma [ lem : x - local ] implies that @xmath398=1 $ ] . let us consider case ( ii ) , wherein we have @xmath429 and @xmath398=[\frac{-c'd'}{p}]=1 $ ] .
let us set @xmath430 with @xmath431 and @xmath432 .
it is clear that @xmath427 leads to a torsor with @xmath160-adic points , by lemma [ lem : p1 ] .
the choice @xmath433 is permissible if and only if there exist coprime @xmath218 such that @xmath434 for @xmath344 .
if @xmath435 then implies that there are no such solutions . alternatively
, suppose that @xmath436 .
suppose first that @xmath437 .
in particular we have @xmath280 .
if @xmath438 then we may choose @xmath218 such that @xmath282 , in which case implies that @xmath439 . if @xmath440 then we choose @xmath218 so that @xmath441 and it will follow that @xmath439 . when @xmath442 we argue according to parity . without loss of generality
we may suppose that @xmath443 and @xmath444 .
if @xmath445 then it suffices to take @xmath265 . if @xmath316 are of opposite parity , with say @xmath446 , we consider @xmath447 chosen so that @xmath448
. then it follows that @xmath449 since @xmath450 . finally , the case @xmath451 is impossible . in case ( iii )
we must have @xmath428 or @xmath452 . by symmetry
it will suffice to consider the case @xmath428 .
we wish to determine when there exist coprime @xmath218 such that @xmath274 is odd and @xmath275 is even .
suppose first that @xmath305 , so that @xmath280 .
let @xmath453 .
if @xmath454 is even we may choose @xmath218 such that @xmath285 , in which case it follows from that @xmath455 is even .
likewise , if @xmath454 is odd then @xmath291 and so we also have @xmath292 .
thus we may choose @xmath218 such that @xmath456 is even , in which case @xmath457 is odd .
we suppose now that @xmath458 , say , with @xmath300 and @xmath301 .
if @xmath459 , write @xmath309 with @xmath460 .
if @xmath302 is odd then we may take @xmath461 if @xmath462 is even then we choose @xmath463 such that @xmath464 , which is satisfactory . if , on the other hand , we write @xmath314 for @xmath465 and @xmath466 , then the situation is more complicated .
suppose that @xmath302 is odd and @xmath467 is even .
then it suffices to take @xmath265 .
if @xmath302 and @xmath467 are both odd then @xmath468 by lemma [ lem : x - local ] . supposing that @xmath318 then the argument used in the last part of the proof of lemma [ lem : x - local ] shows that @xmath320 suffices .
if @xmath302 and @xmath467 are both even then @xmath469 and @xmath470 , resulting in a case that is easy to handle .
finally if @xmath302 is even and @xmath467 is odd , a case that requires @xmath471 and @xmath472 , then @xmath473 and it is impossible to find suitable @xmath218 .
the case in which @xmath474 is symmetric .
this completes the proof of the lemma .
we now turn to the set @xmath476 , given @xmath329 .
recall the notation for the sign function @xmath232 .
we may assume that @xmath477 and ( i ) holds in lemma [ lem : x - local ] .
on recalling that @xmath478 and @xmath479 are positive integers , we see that the constraint demands that @xmath480 and @xmath481 share the same sign .
let @xmath482 .
we wish to determine when there exist @xmath483 such that @xmath484 for @xmath344 , which will then ensure that @xmath476 is non - empty .
we write @xmath485 for @xmath476 with @xmath486 .
suppose first that @xmath487 .
taking @xmath488 and @xmath489 shows that @xmath490 .
if , furthermore , @xmath491 or @xmath492 then @xmath493 . finally , if @xmath494 then @xmath495 , as can be seen by taking @xmath496 and @xmath497 .
next we consider the case @xmath498 . here , for @xmath499 , we seek the existence of @xmath218 with @xmath497 such that @xmath500 this occurs if and only if @xmath492 and @xmath501 . for @xmath502
we require instead the existence of @xmath218 with @xmath497 such that @xmath503 this occurs if and only if @xmath504 and @xmath505 .
finally we note that the case in which @xmath506 and @xmath507 does not enter into consideration .
we summarise the situation in table [ table1 ] , in which the final column lists the possible signs of @xmath508 which give rise to a non - empty set @xmath509 .@xmath476 for @xmath510 [ cols="^,^,^,^,^,^",options="header " , ] once combined with our formula for @xmath511 , this therefore concludes the proof of the lemma .
the numerical value of @xmath511 is @xmath512 .
let @xmath513 denote the cardinality on the right hand side of when @xmath514 .
tim dokchitser has kindly implemented a computer algorithm for calculating the ratios @xmath515 for small values of @xmath58 .
taking @xmath516 leads to the numerical value @xmath517 , which agrees quite closely with lemma [ lem : tau-2 ] .
the conclusion of theorem [ c : cor1 ] is now available . combining theorem [ t : m - local ] and lemma [ lem : m - total ]
, we conclude that @xmath518 with @xmath519 as in .
it follows from and that @xmath520 applying and lemma [ lem : tau-2 ] , we deduce that @xmath521 with @xmath522 given by .
the latter euler product converges very rapidly and has numerical value @xmath523 whence @xmath524 , up to 7 decimal places .
this completes the proof of theorem [ c : cor1 ] .
the goal of this section is to establish theorem [ thm2 ]
. the argument will begin along similar lines to the treatments of @xmath83 and @xmath526 , but the analysis is ultimately more involved . whereas our earlier work relied upon the basic estimate in lemma [ lem : basic ] for the number of integers in an interval which are coprime to a given integer , the treatment of @xmath527 will require more sophisticated tools from analytic number theory .
let @xmath528 .
we define an arithmetic function @xmath529 multiplicatively on prime powers via @xmath530 let @xmath531 where @xmath532 is given by .
one finds that @xmath533 we begin with an estimate for the average order of @xmath534 .
applying the burgess bound for character sums , it follows that @xmath538 for any integer @xmath539 . if @xmath540 then implies that @xmath541 hence @xmath542 the remaining sum over @xmath543 is easily seen to be @xmath544 , by an application of the selberg
delange method ( see @xcite , for example ) .
this concludes the proof . at a certain point in our argument
it will be useful to approximate the function @xmath545 in by the simpler arithmetic function @xmath546 we will need an asymptotic formula for the average order of @xmath547 , where @xmath548 is an arbitrary multiplicative arithmetic function satisfying @xmath549 for primes @xmath160 .
this is achieved in the following result .
[ lem : z - sum ] let @xmath79 and let @xmath548 be a multiplicative arithmetic function satisfying .
for @xmath550 and @xmath551 , we have @xmath552 where @xmath553 and @xmath554 is the convergent euler product @xmath555 we consider the associated dirichlet series @xmath556 say , for @xmath557 . since @xmath558
, one sees that @xmath559 where @xmath560 is the dirichlet @xmath561-function associated to the real character @xmath562 modulo @xmath5 and @xmath64 may be continued as a holomorphic function to the half - plane @xmath563 and is bounded absolutely in this region .
hence it follows that @xmath564 , where @xmath565 in particular @xmath566 in the notation of the lemma .
our starting point in the proof of theorem [ thm2 ] is , followed by the changes of variables , and . in the present situation
a further change of variables will be expedient .
define @xmath569 we now write @xmath570 for @xmath571 satisfying @xmath572 .
the union of this constraint with and is equivalent to @xmath573 and @xmath574 with @xmath575 constrained to satisfy @xmath576 for @xmath577 .
note that since @xmath578 it automatically follows that @xmath579 is coprime to @xmath580 . in
what follows it will be convenient to redefine under our various transformations we may now write @xmath582 with @xmath583 in this new notation we must proceed to consider the constraints recorded in
[ s : s - global ] which are both necessary and sufficient to have @xmath584 for all @xmath152 , but @xmath585 .
our key tool is lemma [ lem:1 ] and the associated calculations . for the infinite valuation we must have @xmath586 by part ( i ) of lemma [ lem : x - local ] and lemma [ lem : easy ] . for the primes
@xmath247 there are no additional constraints arising .
the situation is more complicated for the primes @xmath587 . according to lemma [ lem : easy - ish ] we seek constraints under which there exists a unique choice of @xmath399 with @xmath421 and @xmath422 , where @xmath378 is given in lemma [ lem:1 ] .
suppose that @xmath345 , so that @xmath237
. then lemma [ lem : x - local](iii ) implies that does nt hold , that @xmath588 and that @xmath589 if @xmath590 and @xmath591 , for @xmath592 . the constraint in lemma [ lem : p2](iii ) gives either @xmath593 and @xmath594 for @xmath595 , or @xmath596 and @xmath597 for @xmath595 .
the constraint in lemma [ lem : p2](i ) translates as @xmath598}+ { \left[}\frac{-{\varepsilon}_3{\varepsilon}_4 \ell_{3}\ell_{4}c''d''}{p}{\right]}{\leqslant}0.\ ] ] it follows from that @xmath599 and @xmath600 can not both be odd . on recalling from that @xmath601
are coprime , we see that @xmath602 if and only if @xmath603 .
similarly , @xmath604 if and only if @xmath605 , since @xmath606 are coprime . combining all these conditions therefore leads to the description that @xmath588 , with one of the following * @xmath593 and @xmath594 for @xmath595 ; * @xmath596 and @xmath597 for @xmath595 ; * @xmath603 and @xmath605 , with @xmath607 * @xmath603 and @xmath608 , with @xmath609 * @xmath610 and @xmath605 , with @xmath611 for given parameters @xmath612 , the constraints arising from primes @xmath587 for which @xmath345 force the vector @xmath613 to lie in one of finitely many congruence classes modulo @xmath160 .
let us denote the set of possible classes by @xmath614 here @xmath615 and @xmath616 for each prime @xmath160 .
although @xmath617 also depends on @xmath618 and @xmath619 , its cardinality does not , as the following result shows .
next we suppose that @xmath625 , still with @xmath626 recall that @xmath627 , with @xmath628 given by . then according to lemma [ lem :
p2](ii ) we must avoid the the case @xmath629 if @xmath405 and @xmath630 note that @xmath405 and @xmath277 if and only if @xmath631 and @xmath632 .
but for such primes @xmath633 and @xmath634 are equal .
hence , since @xmath587 , we see that the constraint is equivalent to @xmath635 if @xmath631 .
suppose now that @xmath347 and @xmath636 .
in particular @xmath405 and lemma [ lem : p2 ] implies that we must have @xmath637+[\frac{-{\varepsilon}_3{\varepsilon}_4 c'd'}{p}]{\leqslant}0 $ ] .
we can not have both @xmath638 and @xmath639 being odd .
moreover , we must have @xmath640 .
we deduce that precisely one of @xmath641 or @xmath642 is odd .
hence we have the pair of conditions @xmath643=-1,\ ] ] or @xmath644=-1.\ ] ] in accordance with this we are now led to introduce the set @xmath645 for given parameters @xmath612 , with @xmath646 , the constraints arising from primes @xmath587 for which @xmath636 force the vector @xmath613 to lie in one of finitely many congruence classes modulo @xmath160 .
let us denote the set of possible classes by @xmath647 the analogue of lemma [ lem : card - t ] is the following easy result .
for any @xmath128 constrained as above , our work so far has shown that there exists a unique @xmath362 satisfying for some @xmath363 with @xmath364 , such that @xmath649 for every valuation @xmath377 .
suppose that @xmath650 for @xmath651 .
for the prime @xmath514 , it remains to distinguish precisely when @xmath328 is non - empty but @xmath652 is empty .
this is equivalent to demanding that @xmath653 in the notation of [ s:2adic ] .
an inspection of lemmas [ lem:1 - 1][lem:3 - 2 ] shows that for a given choice of @xmath654 , the @xmath1-adic constraints placed on @xmath655 take the shape of a union of particular congruence classes modulo @xmath656 .
let us denote by @xmath657 the set of all possible classes modulo @xmath656 that can arise .
then the pair of constraints @xmath242 and @xmath658 are equivalent to demanding that @xmath659 for given @xmath654 . in particular
this constraint implies that @xmath660 is odd .
although @xmath657 depends on numerous parameters , including the residue class of @xmath342 modulo @xmath5 , its cardinality is independent of all of these .
we will set@xmath661 this constant is equal to the density of points @xmath128 for which there exist coprime integers @xmath244 for which @xmath662 , yet for every choice of coprime @xmath244 one never has both @xmath663 and @xmath664 . using our work in
[ s:2adic ] , as an analogue of lemma [ lem : tau-2 ] , it is possible to calculate a numerical value for @xmath665 .
we have chosen not to pursue this here , however , it being sufficient to note that @xmath666 since @xmath667 .
we are now ready to return to the expression for @xmath527
. it will be convenient to put @xmath668 let us define the function @xmath669 where @xmath670 are given by .
one notes that @xmath671 and @xmath672 in the notation of , with @xmath673 and @xmath674 .
in particular @xmath545 is only supported on positive integers coprime to @xmath675 , which are built from primes congruent to @xmath4 modulo @xmath5 .
summarising our investigation so far , our analogue of is @xmath676 where @xmath677 ( resp .
@xmath678 , @xmath679 ) is given by ( resp . , ) and @xmath680 the conditions of summation here are restricted to @xmath681 such that , and hold , with @xmath682 and @xmath683 the definitions of @xmath617 and @xmath684 ensure that the product @xmath685 is coprime to @xmath686 .
likewise implies that @xmath685 is odd .
hence is redundant . in our analysis of it
will frequently be useful to reduce the allowable ranges for the various parameters appearing in the outer summations .
two quantities that will feature heavily in this process are @xmath687 recall the definitions of @xmath688 .
we proceed to establish the following result .
let us denote the expression that is to be estimate by @xmath694 suppose first that @xmath695 .
then there are @xmath696 choices for @xmath697 .
recall the definition of @xmath698 , for any @xmath170 .
rankin s trick and allow us to deduce that @xmath699 as required .
let us next consider the case @xmath700 . then
our argument in
[ s : asymptotic - nloc ] , which was used to restrict the size of the parameters appearing in , easily gives @xmath701 , again using rankin s trick and to handle the sum over @xmath702 and @xmath619 .
finally we must consider the case @xmath703 in this case , since @xmath704 and the sum over @xmath705 is absolutely convergent , we obtain @xmath706 using rankin s trick again , we see that @xmath707 since @xmath708 for @xmath709 .
hence @xmath710 as required .
a trivial upper bound is given by @xmath711 applying lemma [ lem : r1r2 ] with @xmath700 and @xmath712 , we may henceforth restrict attention to parameters in for which @xmath713 with satisfactory overall error @xmath714 .
we wish to break the sum into congruence classes modulo @xmath715 .
we shall introduce the set
@xmath716 , say , of @xmath717 for which @xmath718 these two conditions imply that @xmath719 we may therefore write @xmath720 note from that @xmath721 . invoking we obtain a sum over @xmath543 which a priori runs over all integers up to order @xmath722 . define the @xmath1-dimensional lattice @xmath723 we then have @xmath724 where @xmath725 it will be crucial to show that there is a negligible contribution from large values of @xmath543 in this sum .
this will be achieved using lemma [ lem : g ] . according to heath - brown (
* lemma 2 ) we have @xmath726 recall the definition of @xmath727 .
the overall contribution to @xmath728 from @xmath729 is therefore seen to be @xmath730 note here that @xmath731 .
we break the summation over @xmath543 into @xmath732 dyadic intervals @xmath733 .
lemma [ lem : g ] therefore yields the contribution @xmath734 it remains to sum this over the remaining parameters subject to . taking @xmath735 and @xmath736 in lemma [ lem : r1r2 ] ,
one arrives at the overall contribution @xmath737 , which is satisfactory for theorem [ thm2 ] .
we may now focus our attention on the contribution from @xmath738 in , which we denote by @xmath739 .
we will need to take care of the coprimality conditions @xmath740 in @xmath741 , retaining the condition @xmath742 .
recall that modulo @xmath715 the first condition is already implied by the definition of @xmath716 , as is the fact that @xmath743 . in this way we deduce that @xmath744 where a change of variables yields @xmath745x , [ k_1,k_3]y)\in \mathsf{\lambda}_{z}\\ x { \leqslant}p/([k_1,k_2]c),~ y { \leqslant}p/([k_1,k_3]d)\\ ( x , y ) { \equiv}(x_0,y_0){\,(\operatorname{mod}{m})}\\ \gcd(xy , z)=1 \end{array } \right\}.\ ] ] here ,
if @xmath746}\in { \mathbb{z}}$ ] denotes the multiplicative inverse of @xmath747 $ ] modulo @xmath715 for @xmath748 , then @xmath749}c_0 $ ] and @xmath750}d_0 $ ] .
finally , we note that @xmath751 is automatically coprime to @xmath752 , since @xmath753 satisfies this property .
we will also need to reduce the ranges of summation for @xmath754 in this expression .
recall from that @xmath755 for any @xmath543 . to achieve our goal
we invert the summation over @xmath543 and @xmath756 , finding that the overall contribution to @xmath739 from @xmath757 is @xmath758 where @xmath759y - { \varepsilon}_2{\varepsilon}_3\ell_2\ell_2''\ell_3\ell_3''2^{\beta+\gamma } \tilde b [ k_1,k_2]x.\ ] ] recall . summing over @xmath760 , we may bound this contribution using lemma [ lem : divisor ] by @xmath761[k_1,k_3]cd } + \frac{p^{1+{\varepsilon}}}{k_1}\right)\\ & \ll \frac{t_1^{\varepsilon}p^4(\log p)^3}{abcd \sqrt{t_2 } } + p^{3+{\varepsilon}}.\end{aligned}\ ] ] applying lemma [ lem : r1r2 ] with @xmath735 and @xmath700 , this therefore shows that @xmath757 contribute @xmath762 to @xmath527 .
this is satisfactory for theorem [ thm2 ] .
likewise , the same argument shows that parameters with @xmath763 make a satisfactory overall contribution to @xmath527 . our work
so far has therefore shown that we can approximate @xmath739 by @xmath764 with acceptable error . to handle @xmath765
we call upon lemma [ lem : lb ] , noting that @xmath766 in , with @xmath767 and @xmath768,\\ a_2 & = { \varepsilon}_4\ell_1\ell_1''\ell_4\ell_4''2^\delta \tilde a [ k_1,k_3],\end{aligned}\ ] ] and furthermore , @xmath769c } , \quad v=\frac{p}{[k_1,k_3]d}.\ ] ] in particular it is clear that @xmath770 and so all the conditions are met for an application of lemma [ lem : lb ] .
this gives @xmath771[k_1,k_3]cd } \right|\ll \frac{\tau_3(z)p}{k_1z}+ \tau_3(z)(\log zm)^3 + e,\ ] ] where @xmath772 we need to sum these three error terms over the remaining parameters to check that they ultimately make a negligible contribution .
the first term on the right of is easy to deal with .
lemma [ lem : r1r2 ] with @xmath695 and @xmath736 easily shows that it contributes @xmath773 overall .
recall that @xmath755 and @xmath774 has average order @xmath775 .
the contribution to @xmath776 from the second term on the right of is therefore seen to be @xmath777 we note here that @xmath778 .
hence once summed over the remaining parameters using lemma [ lem : r1r2 ] with @xmath695 and @xmath736 , we obtain the satisfactory overall contribution @xmath779 .
it remains to handle the contribution to @xmath776 from @xmath780 .
let us write @xmath781 and @xmath782 .
then since @xmath783 it follows that @xmath784 .
taking @xmath755 and carrying out the sums over @xmath754 and @xmath785 , this contribution is seen to be @xmath786 we will need to sort the sum according to the greatest common divisor @xmath787 , writing @xmath788 . the final condition on @xmath789 forces the vectors @xmath790 in which we are interested to lie on a rank @xmath1 integer sublattice of determinant @xmath791 .
heath - brown ( * ? ? ?
* lemma 2 ) therefore shows that the number of @xmath790 is @xmath792 substituting this into the above the second term here is seen to contribute @xmath793 to @xmath776 , which once combined with lemma [ lem : r1r2 ] therefore leads to a satisfactory overall contribution .
finally , the first term contributes @xmath794 to @xmath776 , since @xmath778 .
this too is found to be satisfactory once summed over all the remaining parameters using lemma [ lem : r1r2 ] .
having handled the contribution from the error terms in , we are now free to approximate @xmath765 by the expected main term .
having done so , furthermore , it is convenient to extend the summations over @xmath754 to infinity , which we may do with acceptable error using lemma [ lem : r1r2 ] .
it henceforth suffices to consider the quantity @xmath795[k_1,k_3]},\end{aligned}\ ] ] in place of @xmath776 .
we may carry out the summations over @xmath754 , finding that @xmath796[k_1,k_3]}=\frac{6}{\pi^2 { \varphi}_2^*(mz)\psi(\tilde a \tilde b)},\ ] ] in the notation of and .
hence @xmath797 since @xmath798 . rather than carrying out the sum over @xmath543 directly , which intimately depends on @xmath799
, we shall first show that @xmath545 can be approximated by the function @xmath800 defined in . to this end
we exame the sum @xmath801 for given parameters @xmath802 and @xmath803 .
our aim is to show that @xmath804 which leads to a satisfactory overall contribution once summed over the remaining parameters , via lemma [ lem : r1r2 ] with @xmath695 and @xmath805 .
to establish the desired bound for @xmath806 we deduce from and that @xmath807 with @xmath808 .
this allows us to write @xmath809 for @xmath810 we let @xmath811 ( resp .
@xmath812 ) denote the contribution to @xmath806 from @xmath813 ( resp .
@xmath814 ) .
we begin by estimating @xmath811 . for this it will be convenient to write @xmath815 , where @xmath6 is given multiplicatively at prime powers by @xmath816 for given @xmath817 with @xmath539 , this allows us to deduce that @xmath818 by the burgess bound for character sums .
equipped with this it is straightforward to conclude that @xmath819 turning to the contribution from @xmath814 and writing @xmath820 , we have @xmath821 note that @xmath822 is necessarily odd in this summation . rearranging terms we are led to an inner sum of the form @xmath823 for suitable real numbers @xmath824 with modulus at most @xmath12 . breaking the @xmath822 sum into dyadic intervals and applying heath - brown s large sieve for real characters ( * ? ? ?
* cor . 4 ) ,
we obtain @xmath825 taking @xmath826 and combining this with our estimate for @xmath811 , we therefore arrive at the desired bound for @xmath806 in .
having shown that we may safely approximate @xmath545 by @xmath827 in @xmath828 , we proceed to swap the sums over @xmath829 and @xmath543 . in this way
we are led to consider the sum @xmath830 the asymptotic evaluation of @xmath831 is the object of the following result .
[ lem : calculate - s ] let @xmath832 where @xmath833 then we have latexmath:[\[\sum_{m , n}\sum_{{\ensuremath{\boldsymbol\ell } } } \sum_{m'',n '' } \sum_{\beta,\gamma,\delta}\sum_{{\ensuremath{\boldsymbol\ell } } '' } \frac{p^2}{cdm^2 } \sum_{t_0}\sum_{z } \frac{1}{z\psi(z ) } since @xmath835 and @xmath836 are coprime , we have @xmath837 .
writing @xmath815 , where @xmath6 is given by , we obtain @xmath838 we recall that @xmath839 . making the obvious change of variables , and and using the mbius function to detect coprimality conditions ,
the inner sum becomes @xmath840 where @xmath841 denotes the number of positive integers @xmath842 and @xmath843 such that @xmath844\mid x , \quad [ k , k_1]\mid y,\ ] ] and @xmath845 .
note here that @xmath846 and @xmath847 are automatically both coprime to @xmath752 since @xmath480 and @xmath481 are . making a further change of variables , @xmath841 is equal to the number of positive integers @xmath848e_1a)$ ] and @xmath849e_2b)$ ] such that @xmath850 and @xmath851 , for integers @xmath852e_1}a_0 $ ] and @xmath853e_2}b_0 $ ] .
taking a trivial upper bound for @xmath841 , we find that @xmath854[k , k_2]e_1e_2ab}.\end{aligned}\ ] ] note that @xmath855 for any @xmath856 . in the usual way
, combining these estimates with lemma [ lem : r1r2 ] shows that the overall contribution from @xmath857 is @xmath858 , for any @xmath79 .
hence we may assume that @xmath859 in .
we estimate @xmath841 using lemma [ lem : basic ] twice , with @xmath860 and @xmath861 .
this is valid provided that @xmath862e_1a)\gg t_1mz^{\varepsilon}$ ] and @xmath863e_2b)\gg t_1mz^{\varepsilon}$ ] .
recall from and that @xmath864 and @xmath865 .
it therefore suffices to have @xmath866 , which obviously holds .
hence we may conclude that @xmath867[k , k_2]e_1e_2abm^2 } \left(1+o \left(\frac{1}{t_1^{1-{\varepsilon}}}\right)\right).\ ] ] the overall contribution from the error term here , once substituted into and summed over the remaining parameters , is easily found to be @xmath868 .
similarly , the shape of the main term allows us to extend the summations over @xmath58 and @xmath385 to infinity with acceptable error .
we are therefore led to approximate @xmath831 by @xmath869[k , k_2]}.\end{aligned}\ ] ] the inner sum over @xmath870 here has already been calculated in .
finally , by calculating euler factors , one sees that the sums over @xmath871 evaluate to @xmath872 it now follows that @xmath873 agrees with expression in the statement of the lemma .
armed with lemma [ lem : calculate - s ] , we now return to our expression for @xmath828 .
let @xmath874 furthermore , we observe that @xmath875 say . then , in place of @xmath828 , our work so far has shown that we may work with @xmath876 where @xmath877 is given in the statement of lemma [ lem : calculate - s ] .
we have now arrived at the final stages of the argument , where it is necessary to evaluate the sum over @xmath543 asymptotically .
this is achieved by taking @xmath878 of order @xmath879 in lemma [ lem : z - sum ] . in particular
it is clear that @xmath880 .
let us write @xmath881 and @xmath882 .
since clearly holds , we deduce that @xmath883 one notes that @xmath884 and @xmath885 .
hence the overall contribution from the error term here is satisfactory for theorem [ thm2 ] , once summed over the remaining parameters satisfying using lemma [ lem : r1r2 ] . finally ,
using this result a final time we can eliminate and so extend the summation over the outer parameters to infinity , with acceptable error . returning to , we have therefore shown that the statement of theorem [ thm2 ] holds , with @xmath886 we proceed to study this constant , with the goal of showing that it is positive .
the set @xmath716 was defined to be the @xmath888 for which holds .
since @xmath656 , @xmath889 and @xmath890 are pairwise coprime , we clearly have @xmath891 furthermore , lemma [ lem : card - t ] implies that @xmath892 whereas @xmath893 is given by lemma [ lem : card - t ] .
recall the expression for the non - zero constant @xmath665 . returning to our formula for the constant , in view of the definition of @xmath688
, we sum over the six possible choices for @xmath894 to get @xmath895 next we undertake the summation over @xmath619 , where @xmath679 is given by .
a simple calculation reveals that @xmath896 likewise , with reference to the definition of @xmath677 and lemma [ lem : card - t ] , it is straightforward to check that the sum over @xmath702 evaluates to @xmath897 where @xmath898 is given by and @xmath899 noting that @xmath900 and @xmath901 , we may therefore conclude that @xmath902 these sums over @xmath903 are absolutely convergent and positive , as one checks by considering the associated euler products .
this therefore shows that @xmath904 , as required to complete the proof of theorem [ thm2 ] . | chtelet surfaces provide a rich source of geometrically rational surfaces which do not always satisfy the hasse principle .
we investigate the frequency that such counter - examples arise over the rationals . |
the temperature induced phase transition in the @xmath3 scalar field model with a spontaneous symmetry breaking ( ssb ) has a long history of investigations .
it was studied either by analytic methods of the quantum field theory or in lattice simulations ( see refs.@xcite and references therein ) .
it was recently observed by analytic calculations within the perturbation theory ( pt ) in the daisy , super daisy and some type beyond resummations @xcite that a phase transition of the first order could occur in the @xmath4-model .
however , the lack of the expansion parameter happens near the phase transition temperature @xmath5 for various kind resummations .
so , it is impossible to draw a reliable conclusion about the transition type even for small values of the coupling constant @xmath2 . in ref.@xcite
some extended kind of resummations were used for the @xmath3-model , and a phase transition of the second order was determined independently of the coupling value .
analogous results have been obtained in monte carlo ( mc ) simulations on a lattice . as a result , nowadays the general believe is that the phase transition is of the second order and the pt fails in this problem .
however , in the @xmath3-models , the results of pt calculations coincide with the lattice mc ones in the limit of @xmath6 , only @xcite .
recently , a new powerful computational platform general purpose computing on graphics processing units ( gpgpu ) technology has been put in force @xcite that gives a possibility to generate extremely large amount of mc data .
therefore the accuracy of calculations can be essentially increased and it becomes possible to shed light upon hidden peculiarities and details of different processes of interest studied by mc simulations .
one of such unsolved problems is the kind of the temperature phase transition in the @xmath4-model for small values of @xmath7 and the reliability of the pt results .
this is because there are no estimates what coupling values should be considered as small ones . as a rule
, the small values are chosen to be of the order @xmath8 .
however , to make the correct choice some physical motivation is needed .
one of possible reasons is the so - called linde - weinberg bound on the scalar field mass @xcite .
many years ago these authors observed that in models with the negative mass squared , @xmath9 , the ssb did not happen for the coupling values below some scale , @xmath10 . the actual value @xmath1 depends on the mass parameter entering the lagrangian .
so , it is natural to consider the values of the coupling @xmath11 as small ones . these values appear to be much smaller than the values mentioned above . in the present paper
we investigate the temperature induced phase transition in the @xmath4-model in a wide interval of the coupling constant @xmath2 using the gpgpu technology .
we obtain the linde - weinberg bound @xmath12 .
then we compare the mc simulations with the hot and cold starts . in a narrow interval just above the linde - weinberg bound , @xmath13
, an order parameter shows a hysteresis behavior near the phase transition temperature .
this behavior means a phase transition of the first order . with the further increasing of @xmath2
the hysteresis behavior becomes less pronounced and disappears at all reflecting a phase transition of the second order .
the paper is organized as follows . in sect.2
we describe the model and its realization on a lattice . in sect.3 a necessary information on the mc simulations
is given and the obtained results are adduced .
sect.4 summarizes the results .
in order to construct a self - consistent lattice version of the @xmath0-model we start with quantum field theory in continuous space .
the thermodynamical properties of the model are described by the generating functional @xmath14},\ ] ] where @xmath15 is a real scalar field , and the action is @xmath16 the standard realization of generating functional in mc simulations on a lattice assumes a space - time discretization and the probing random values of fields in order to construct the boltzmann ensemble of field configurations .
then any macroscopic observable can be measured by averaging the corresponding microscopic quantity over this ensemble . however
, the direct lattice implementation of ( [ gen - func ] ) encounters an evident problem : the field @xmath15 is distributed uniformly in the infinite interval @xmath17 and there is no random number generator to simulate it . of course , one can try to cut the interval off , but such a procedure requires the knowledge of characteristic scales of the model in order to separate the unimportant tails from the interval of physically important values .
instead , we prefer to rewrite the initial @xmath0 model in continuum space - time in the form allowing a further self - consistent lattice realization .
first we introduce one - to - one transformation @xmath18 to a new field variable @xmath19 defined in the finite interval @xmath20 .
let @xmath21 corresponds to @xmath22 and @xmath23 .
the generating functional in terms of @xmath24 reads @xmath25}.\ ] ] the jacobian can be included in the action as @xmath26
. then the new field @xmath24 can be easily realized by a uniform random number generator .
the second step is the space - time discretization . for mc simulations we introduce a hypercubic lattice with hypertorous geometry .
we use an anisotropic lattice with the spatial and temporal spacings @xmath27 and @xmath28 , @xmath29 .
the scalar field is defined in the lattice sites . as a result , the generating functional becomes @xmath30-\sum_x\log\varphi'[u(x)]\right)\right]},\ ] ] where @xmath31 and @xmath32= \nonumber\\ \sum_x\frac{a_s^4}{\zeta } & & \left [ \left(\varphi'[u(x)]\right)^2 \frac{\partial_\mu u(x)\partial_\mu u(x)}{2 } -\frac{m^2}{2}\varphi^2[u(x)]+\frac{\lambda}{4}\varphi^4[u(x)]\right].\end{aligned}\ ] ] the lattice forward derivative is defined as usually by the finite difference operation @xmath33 where @xmath34 is the lattice spacing in the @xmath35 direction , @xmath36 is the unit vector in the direction indicated by @xmath35 . in the case of pure condensate field
the action is determined by the potential @xmath37=\left[-\log\varphi'[u]+ \frac{a_s^4}{\zeta } \left(-\frac12m^2\varphi^2[u]+\frac{\lambda}{4}\varphi^4[u]\right)\right].\end{aligned}\ ] ] this potential is topologically equivalent to the potential @xmath38 @xmath39 .
it has one local maximum at @xmath21 and two symmetric global minima at @xmath40 and @xmath41 .
the spread between the values of the potential at the local maximum and the global minima is @xmath42}{\varphi'[0.5 ] } -\frac{a_s^4}{\zeta } \left(-\frac12m^2\varphi^2[u_0]+\frac{\lambda}{4}\varphi^4[u_0]\right).\ ] ] the quantities @xmath40 and @xmath43 play a crucial role in mc simulations .
being equivalent in theory , different choices of these parameters can produce drastically different results in actual simulations .
the reason is the finite number of simulations in an actual computer experiment .
if rare but physically important events could be missed , then the mc algorithm will not converge to the boltzmann ensemble of configurations . in case of @xmath44 iterations
all the important probabilities have to be greater than @xmath45 .
considering the phase transition , one must guarantee that the mc algorithm meets the field values compatible with both the phases to choose .
if @xmath46 , then the broken phase can be missed since the corresponding field values are extremely rare events . on the other hand , in the limit @xmath47 ( or @xmath48 ) the unbroken phase is washed out .
it is also important to ensure a finite probability of the transition between those field values .
the acceptance of non - zero condensate values of the field is ruled approximately by @xmath49 at each lattice site .
if @xmath50 , then the unbroken phase never occurs in actual simulations .
if @xmath51 , then there is no broken phase . to study the phase transition in the model
, we choose the following conditions : @xmath52 thus , the half of generated field values will be between the global minima of the ` effective ' potential , and no phase will be accidentally missed .
the probability to prefer condensate or non - condensate values will be of order @xmath53 ensuring the fast convergence of mc algorithm .
of course , the choice ( [ cond - m ] ) is not optimal for temperatures far away from the critical temperature . to satisfy two conditions ( [ cond - m ] ) we use a convenient two - parameter function , @xmath54&=&m\xi\mathrm{arctanh}\left[\eta(2u-1)+(1-\eta)(2u-1)^3\right]\end{aligned}\ ] ] with @xmath55 and @xmath56 .
the values of @xmath57 and @xmath58 have to be found as the solution of equations @xmath59 and ( [ dv ] ) .
these equations can be written as @xmath60-{\cal g}[u_0]}{\left({\cal f}[u_0]-{\cal g}[u_0]\right)^2}=z = \frac{\lambda\zeta}{m^4 a_s^4},\\ & & \frac{\lambda}{m^2 z}\varphi^2[u_0 ] = { \cal f}[u_0]-{\cal g}[u_0],\label{syst-2}\end{aligned}\ ] ] where @xmath61 is a dimensionless parameter of the model , @xmath62&=&\frac{1-\eta(2u-1)-(1-\eta)(2u-1)^3}{1+\eta(2u-1)+(1-\eta)(2u-1)^3},\\ { \cal f}[u]&= & \left(\frac{{\cal k}''[u]{\cal k}[u]}{({\cal k}'[u])^2}-1\right)\log{\cal k}[u],\\ { \cal g}[u]&= & 4\left(\log\frac{-{\cal k}'[u]}{4\eta{\cal k}[u]}-\delta_v\right),\end{aligned}\ ] ] where the primes denote derivatives with respect to @xmath24 .
( [ syst-1 ] ) gives @xmath63 , then @xmath57 can be found from ( [ syst-2 ] ) .
there is no physical solution for @xmath64 .
this forbidden interval corresponds to low temperatures which can not be reached within the chosen parametrization .
finally the lattice action is @xmath65&=&\sum_x\sum\limits_{\mu } \left [ y\sqrt{\frac{z}{\zeta\lambda } } \left(\frac{{\cal k}'[u(x)]}{{\cal k}[u(x)]}\right)^2 \left(\frac{u(x+a_\mu\hat{\mu})-u(x)}{a_\mu / a_s}\right)^2\right ] \nonumber\\ & + & \sum_x\left[-\frac{1}{4}{\cal g}[u(x ) ] -y\log^2 { \cal k}[u(x ) ] + y^2 z\log^4 { \cal k}[u(x ) ] + v_0 \right],\nonumber\\ y&=&\frac{{\cal f}[u_0]-{\cal g}[u_0]}{2\log^2{\cal k}[u_0]},\\ v_0&= & -\delta_v -\log(2m\xi\eta).\end{aligned}\ ] ] the constant part of the action @xmath66 is completely unimportant for calculations and can be omitted , since mc algorithm is based on the difference between the actions of modified and initial field configurations . in the units of classical condensate @xmath67 for @xmath68 on lattice @xmath69,scaledwidth=60.0% ] by varying @xmath70 it is possible to change @xmath71 , while keeping @xmath27 fixed . consequently the temperature @xmath72 can be changed continuously at the fixed @xmath27 . at low temperatures the global discrete symmetry @xmath73
is expected to be broken due to a non - zero field condensate .
the field condensate , @xmath74 , can be measured directly as the average of @xmath75 over the lattice . as an example , in fig.[fig : lambdaz ] we plot @xmath76 in the units of the classical condensate @xmath67 for @xmath77 and @xmath69 lattice .
the lower values of @xmath70 and @xmath61 correspond to lower temperatures .
one can see the evident phase transition with the field condensate growing with the temperature decreasing .
since the clear positive or negative values of @xmath78 appear in lattice configurations in the broken phase , we conclude about the absence of domains and will use @xmath76 in plots in what follows .
the field condensate is an obvious order parameter vanishing in the high - temperature phase .
we can use it to determine the type of the phase transition . in case of the first - order transition
the overheated and supercooled states are possible .
so , the mc simulations with the hot and cold starts have to lead to different phases near the critical temperature . combining the mc simulations for the hot and cold starts we will see a hysteresis plot .
this exfoliation in the vicinity of the critical temperature has to be observed independently of the direction in the @xmath79 plane .
we consider in details two slices of the two - dimensional function in fig.[fig : lambdaz ] at fixed @xmath80 and @xmath81 .
we compute the field condensate with the hot and cold starts for different @xmath70 .
the hysteresis - type plots will mean a first order phase transition .
the exfoliation of the simulated data in the vicinity of the critical temperature is a tiny effect .
a large amount of simulation data must be prepared to observe it . in this regard , achieving the highest performance of the computational hardware is a problem of great importance . to speed up essentially the simulation process
we apply a gpu cluster of amd / ati radeon gpus : hd5870 , hd5850 , hd4870 and hd4850 .
the peak performance of the cluster is up to 8 tflops .
the low - level amd intermediate language ( amd il ) is used in order to obtain the maximal performance of the hardware . some technical details of mc simulations on the ati gpus and review of amd stream sdk are given in ref .
@xcite and references therein .
the mc simulations are realized at hypercubic lattices up to @xmath82 .
most of the obtained statistics come from the lattice @xmath69 .
we use the pseudo - random number generator in the mc kernel , but all the key results are checked with generator @xcite .
the lattice data are stored with the single precision . updating the mc configurations
are also performed with the single precision , whereas all the averaging measurements are carried out with the double precision to avoid the accumulation of errors .
it should be noted that for a rough estimate of the parameter space in the early stages of the work we used a pseudorandom number generator to speed up the calculations .
the condensate values obtained by this generator coincide with the correspondent points produced with the help of and generators .
however , the generator realized the two minima of the broken phase with an unequal frequency .
its unexpected behavior had been discovered in the first time , since good statistical properties of this generator were reported in the literature on real modeling before .
the system is thermalized by passing 5000 mc iterations for every run . for measuring we use 1024 mc configurations separated by 10 bulk updates . for the lattice @xmath69 at @xmath80 and @xmath81 for @xmath83$].,title="fig:",scaledwidth=40.0% ] for the lattice @xmath69 at @xmath80 and @xmath81 for @xmath83$].,title="fig:",scaledwidth=40.0% ] + 0.5 cm for the lattice @xmath69 at @xmath80 and @xmath81 for @xmath83$].,title="fig:",scaledwidth=40.0% ] for the lattice @xmath69 at @xmath80 and @xmath81 for @xmath83$].,title="fig:",scaledwidth=40.0% ] as it was discussed in the previous section , we collect the data for the absolute value of the averaged field @xmath76 representing the field condensate . the temperature dependence of @xmath76 for the lattice @xmath69 at @xmath80 and @xmath81 for @xmath84 $ ] is shown in fig.[fig : lambdas ] .
the whole data set for every plot is divided into 15 bins .
different initial conditions are marked with different colors : the hot start is depicted in red ( lower bins ) and the cold start is represented in blue ( upper bins ) .
the mean values and the 95% confidence intervals are shown for each bin .
every bin contains 150 simulated points .
as it is seen from fig.[fig : lambdas ] , for @xmath85 the temperature dependence of the field condensate is insensitive to the start configuration chosen .
both the cold and hot starts lead to the same behavior of the field condensate for various @xmath70 .
this means the phase transition to be of the second order , and this result is in agreement with the common opinion on the type of the phase transition stated in refs .
however , therein this value of the coupling is considered as a small one .
then , for smaller values of @xmath7 the overheated configurations occur in the broken phase for the hot start , and the supercooled states can be found for the cold start .
that is , the exfoliation of the simulated data in the vicinity of the critical temperature is observed for different start configurations .
such a hysteresis behavior corresponds to the phase transition of the first order . with further decreasing of @xmath2 to the values of order
@xmath86 the behavior of hot- and cold - started simulations becomes completely separated and independent of the temperature .
this effect is plotted in fig.[fig : lambdalw ] .
such type property means that the ssb does not happen even at zero temperature .
the corresponding value @xmath1 can be identified as the linde - weinberg low bound . at @xmath87 , @xmath80 and @xmath83$].,title="fig:",scaledwidth=40.0% ] at @xmath87 , @xmath80 and @xmath83$].,title="fig:",scaledwidth=40.0% ] comparing the plots for two different values of @xmath61 with the same set of other parameters in fig.[fig : lambdas ] , one can see that the hysteresis behavior for @xmath88 is similar to the case of @xmath89 .
this is the common feature for all the tested values of @xmath7 , but we adduce only one slide for brevity .
the independence of @xmath61 is important .
it means that the observed exfoliation of simulated points is stable within our parametrization of the action .
thus , the introduced approach is reliable at the transition temperature .
as it was discovered in mc simulations , the temperature phase transition in the @xmath4 @xmath0 model is strongly dependent on a coupling value @xmath2 .
there is the low bound @xmath90 determining the range where the ssb is not realized .
close to this value in the interval @xmath91 the phase transition is of the first order . for larger values of @xmath2
the second order phase transition happens .
these observations , in particular , may serve as a criterium for applicability of different kind resummations in perturbation theory .
in fact , we see that the daisy and super daisy resummations give qualitatively correct results for small values of @xmath2 . for larger values
they become non - adequate to the second - order nature of the phase transition . in this case
other more complicated resummation schemes should be used .
10 jean zinn - justin , _ quantum field theory and critical phenomena _ , clarendon press , 1996 .
j. berges , n. tetradis and c. witterich , phys.rept.*363 * , 223 ( 2002 ) .
p. cea , m. comsoli and l. cosmai , arxiv : hep - lat/0407024 ; arxiv : hep - ph/0311256 ; nucl .
. suppl . * 106 * , 953 ( 2002 ) [ arxiv : hep - lat/0110008 ] . m. bordag and v. skalozub , j. phys .
a * 34 * , 461 ( 2001 ) [ arxiv : hep - th/0006089 ] ; phys.rev .
d * 65 * , 085025,2002 , [ arxiv : hep - th/0107027 v2 ] .
j. baacke and s. michalski , phys .
d * 67 * , 085006 ( 2003 ) [ arxiv : hep - ph/0210060 ] . n. petropoulos arxiv : hep - ph/0402136 .
a. d. linde , jetp lett .
* 23 * , 64 ( 1976 ) [ pisma zh .
* 23 * , 73 ( 1976 ) ] .
s. weinberg , phys .
lett . * 36 * , 294 ( 1976 ) . v. demchik and a. strelchenko , arxiv:0903.3053 [ hep - lat ] .
v. demchik , comp .
* 182 * , 692 ( 2011 ) , doi : 10.1016/j.cpc.2010.12.008 ; arxiv:1003.1898 [ hep - lat ] . | the temperature induced phase transition is investigated in the one - component scalar field @xmath0 model on a lattice by using monte carlo simulations . using the gpgpu technology a huge amount of data
is collected that gives a possibility to determine the linde - weinberg low bound on the coupling constant @xmath1 and investigate the type of the phase transition for a wide interval of coupling values .
it is found that for the values of @xmath2 close to this bound a weak - first - order phase transition happens .
it converts into a second order phase transition with the increase of @xmath2 .
a comparison with analytic calculations in continuum field theory and lattice simulations obtained by other authors is given .
_ keywords : _ scalar model ; phase transitions ; gpu |
this paper is a continuation of our previous work @xcite , where we consider the one dimensional nonlinear schrdinger equation i + _
x^2 + f(||^2 ) = 0 .
this equation appears as a relevant model in condensed matter physics : bose - einstein condensation and superfluidity ( see @xcite , @xcite , @xcite , @xcite ) ; nonlinear optics ( see , for instance , the survey @xcite ) .
several nonlinearities may be encountered in physical situations : @xmath0 gives rise to the focusing / defocusing cubic ( nls ) ; @xmath1 to the so called gross - pitaevskii equation ; @xmath2 ( see @xcite for bose - einstein condensates ) ; more generally a pure power ; the `` cubic - quintic '' ( nls ) ( see @xcite ) , where @xmath3 and @xmath4 , @xmath5 and @xmath6 are positive constants such that @xmath7 has two positive roots ; and in nonlinear optics , we may take ( see @xcite ) : [ nonlin ] f ( ) = - ^- ^2 , f ( ) = - ( - ) , f ( ) = - ( 1 + tanh ( ) ) , where @xmath8 , @xmath9 , @xmath10 , @xmath11 , @xmath12 are given constants ( the second one , for instance , takes into account saturation effects ) , etc . as a consequence , as in our work @xcite , we shall consider a rather general nonlinearity @xmath13 , with @xmath13 of class @xmath14 . in the context of bose - einstein condensation or nonlinear optics ,
the natural condition at infinity appears to be @xmath15 where @xmath16 is such that @xmath17 .
+ for solutions @xmath18 of ( nls ) which do not vanish , we may use the madelung transform @xmath19 and rewrite ( nls ) as an hydrodynamical system with an additional quantum pressure [ madtwk ] \ { ll = 0 + + = 0 . \ { ll = 0 + + = 0 , . with @xmath20 .
when neglecting the quantum pressure and linearizing this euler system around the particular trivial solution @xmath21 ( or @xmath22 ) , we obtain the free wave equation @xmath23 with associated speed of sound @xmath24 provided @xmath13 satisfies the defocusing assumption @xmath25 ( that is the euler system is hyperbolic in the region @xmath26 ) , which we will assume throughout the paper . concerning the rigorous justification of the free wave regime for the gross - pitaevkii equation ( in arbitrary dimension ) , see @xcite .
the speed of sound @xmath27 enters in a crucial way in the question of existence of travelling waves for ( nls ) with modulus tending to @xmath28 at infinity ( see , _ e.g. _ , @xcite ) .
+ the nonlinear schrdinger equation formally preserves the energy @xmath29 where @xmath30 .
since @xmath31 when @xmath32 , it follows that the natural energy space turns out to be the space @xmath33 endowed with the distance @xmath34 the cauchy problem has been shown to be locally well posed in the zhidkov space @xmath35 by p. zhidkov @xcite ( see also the work by c. gallo @xcite ) . for global well - posedness results ,
see @xcite and @xcite .
more precisely , the local well - posedness we shall use is the following .
[ @xcite , @xcite ] [ cauchy ] let @xmath36 .
then , there exists @xmath37 and a unique solution @xmath38 to ( nls ) such that @xmath39 and @xmath40 .
moreover , @xmath41 does not depend on @xmath42 .
the other quantity formally conserved by the schrdinger flow , due to the invariance by translation , is the momentum .
the momentum is not easy to define in dimension one for maps that vanish somewhere ( see @xcite , @xcite ) .
however , if @xmath43 does not vanish , we may lift @xmath44 , and then the correct definition of the momentum is given by @xcite : @xmath45 where @xmath46 denotes the real scalar product in @xmath47 .
we define @xmath48 which is the open subset of @xmath49 in which we have lifting and where the hydrodynamical formulation of ( nls ) is possible through the madelung transform .
it turns out that if the initial datum belongs to @xmath50 , the solution of ( nls ) provided by theorem [ cauchy ] remains in @xmath50 for small times , and that the momentum is indeed conserved on this time interval ( see @xcite ) .
the travelling waves with speed of propagation @xmath51 are special solutions of ( nls ) of the form @xmath52 the profile @xmath53 has then to solve the ode _ x^2 u + u f(|u|^2 ) = i c _ x u together with the condition @xmath54 as @xmath55 .
these particular solutions play an important role in the long time dynamics of ( nls ) with nonzero condition at infinity . possibly conjugating ( tw@xmath56 )
, we see that we may assume that @xmath57 without loss of generality .
moreover , we shall restrict ourselves to travelling waves which belong to the energy space @xmath58 ( so that @xmath59 at @xmath60 by sobolev embedding @xmath61 ) . for travelling waves @xmath62
that do not vanish in @xmath63 , hence that may be lifted @xmath64 , the ode ( tw@xmath56 ) can be transformed ( see , _
e.g. _ , @xcite ) into the system @xmath65 and where the function @xmath66 is related to @xmath13 by the formula @xmath67 to a nontrivial travelling wave @xmath68 is associated ( see @xcite ) some @xmath69 such that @xmath70 and @xmath71 is negative between @xmath72 and @xmath73 , and @xmath74 varies between @xmath75 and @xmath76 , that is there holds @xmath77 . moreover , the only travelling wave solution ( if it exists ) that vanishes somewhere is for @xmath78 and is called the kink : it is an odd solution ( up to a space translation ) and then @xmath79 .
we have also seen in @xcite that any travelling wave in @xmath49 with speed @xmath80 is constant ; and also that any nonconstant travelling wave in @xmath49 of speed @xmath81 belongs to a unique ( up to the natural invariances : phase factor and translation ) local branch @xmath82 defined for @xmath51 close to @xmath83 . in @xcite
, we have investigated the qualitative behaviours of the travelling waves for ( nls ) with nonzero condition at infinity for a general nonlinearity @xmath13 .
a particular attention has been payed in @xcite to the transonic limit , where we have an asymptotic behaviour governed by the korteweg - de vries or the generalized korteweg - de vries equation . in order to illustrate the very different situations we may encounter when we allow a general nonlinearity @xmath13
, we give now some energy - momentum diagrams we have obtained ( one is taken from the appendix in @xcite , where we have performed numerical simulations in dimension two for the model cases we have studied in @xcite ) .
+ @xmath84 the gross - pitaevskii nonlinearity : @xmath85 ( see figure [ zgp ] ) .
diagram , title="fig:",width=226,height=264 ] diagram , title="fig:",width=226,height=226 ] @xmath84 a cubic - quintic - septic nonlinearity ( i ) : @xmath86 ( see figure [ zcqsi ] ) .
diagram , title="fig:",width=226,height=264 ] diagram , title="fig:",width=226,height=226 ] @xmath84 a cubic - quintic - septic nonlinearity ( ii ) : @xmath87 or @xmath88 .
for these two nonlinearities , the graph of @xmath89 and @xmath90 vs. speed @xmath51 is given in figure [ zcqsii ] , but the @xmath91 diagrams are respectively those in figure [ zzcqsii ] . diagrams , title="fig:",width=170,height=226 ] diagrams , title="fig:",width=170,height=226 ] @xmath84 a cubic - quintic - septic nonlinearity ( iii ) : @xmath92 ( see figure [ zcqsiii ] ) .
diagram , title="fig:",width=226,height=264 ] diagram , title="fig:",width=188,height=226 ] @xmath84 a degenerate case : @xmath93 ( see figure [ zdege ] ) .
diagram , title="fig:",width=226,height=264 ] diagram , title="fig:",width=188,height=226 ] @xmath84 a perturbation of the previous degenerate case : @xmath94 ( see figure [ zdegeper ] ) .
diagram , title="fig:",width=226,height=264 ] diagram , title="fig:",width=188,height=226 ] @xmath84 a saturated ( nls ) : @xmath95 with @xmath96 ( see figure [ zsatur ] ) .
diagram , title="fig:",width=226,height=207 ] diagram , title="fig:",width=188,height=226 ] @xmath84 another saturated ( nls ) : @xmath97 , with @xmath98 ( see figure [ zsaturtrop ] ) .
diagram , title="fig:",width=226,height=264 ] diagram , title="fig:",width=188,height=226 ] @xmath84 the cubic - quintic nonlinearity : @xmath99 ( see figure [ zcq ] ) .
diagram , title="fig:",width=226,height=264 ] diagram , title="fig:",width=188,height=226 ] through the study ( in @xcite ) of these model cases , we have shown that if the energy - momentum diagram is well - known for the gross - pitaevskii equation , the qualitative properties of the travelling waves solutions can _ not _ be easily deduced from the global shape of the nonlinearity @xmath13 . in particular , even if we restrict ourselves to smooth and decreasing nonlinearities ( as is the gross - pitaevkii one ) , we see that we may have a great variety of behaviours : multiplicity of solutions ; branches with diverging energy and momentum ; nonexistence of travelling wave for some @xmath100 ; branches of solutions that cross ; existence of sonic travelling wave ; transonic limit governed by the ( mkdv ) or more generally by the ( gkdv ) solitary wave equation instead of the usual ( kdv ) one ; existence of cusps ... + we investigate now the behaviour at infinity of the nontrivial travelling waves , which depend whether @xmath101 or not .
we consider for @xmath102 the following assumption : + @xmath103 @xmath13 is of class @xmath104 near @xmath105 .
moreover , for @xmath106 , we have @xmath107 but @xmath108 ( note that for @xmath109 , equality always holds by definition of the speed of sound @xmath110 ) .
+ [ decroiss ] let @xmath62 be a non constant travelling wave of speed @xmath111 .
+ @xmath112 if @xmath113 , then there exist @xmath114 such that @xmath115 is a real - valued function and there exist two real constants @xmath116 ( depending only on @xmath13 and @xmath117 ) and @xmath118 such that , as @xmath119 , @xmath120 @xmath121 if @xmath122 , then @xmath123 does not vanish , hence can be lifted : @xmath124 .
furthermore , there exist four real constants @xmath125 , @xmath126 ( depending only on @xmath13 , @xmath51 and @xmath76 ) , @xmath118 and @xmath127 such that , as @xmath119 , @xmath128 and @xmath129 @xmath130
if @xmath131 then @xmath132 does not vanish , hence can be lifted : @xmath133 .
we assume that there exists @xmath134 such that @xmath103 is verified and denote @xmath135 \not = 0 .\ ] ] then , we have , as @xmath119 , @xmath136 and @xmath137 and if @xmath138 , there exists @xmath139 and @xmath114 such that @xmath140 in particular , since we impose @xmath141 , we must have @xmath142 . for the gross - pitaevkii nonlinearity ( @xmath143 ) , we may compute explicitly the travelling waves for @xmath144 ( see @xcite , @xcite ) @xmath145 up to the invariances of the problem : translations and multiplications by a phase factor . on this explicit formula
, the decay of the phase and modulus can be checked . in particular , as @xmath119
, we have @xmath146 in the above statements , the constants @xmath127 and @xmath147 reflect the gauge and translation invariance . in the spirit of the model cases proposed in @xcite , for @xmath148
we obtain a smooth decreasing nonlinearity tending to @xmath149 at @xmath150 ( thus qualitatively similar to the gross - pitaevskii nonlinearity ) for which we have @xmath151 , @xmath152 , and @xmath153 . for this nonlinearity @xmath13
, there exists a nontrivial sonic travelling wave of infinite energy ( corresponding to @xmath154 ) , since @xmath155 .
the aim of this paper is to investigate the stability of the travelling waves for the one dimensional ( nls ) .
we recall the definition of orbital stability in a metric space @xmath156 for which we have a local in time existence result .
[ defstabi ] let @xmath111 and @xmath62 be a nontrivial travelling wave of speed @xmath157 .
we say that @xmath123 is orbitally stable in @xmath156 , where @xmath158 , if for any @xmath159 , there exists @xmath160 such that for any initial datum @xmath161 such that @xmath162 , any solution @xmath38 to ( nls ) with initial datum @xmath163 is global in @xmath164 and @xmath165 in the sequel , @xmath68 will always stands for a nontrivial travelling wave , and we freeze the translation invariance by imposing that @xmath166 is even .
moreover , the solutions of ( nls ) we consider will always be those given by theorem [ cauchy ] .
the first stability result for the travelling waves for ( nls ) with nonzero condition at infinity is due to z. lin @xcite .
the analysis relies on the hydrodynamical form of ( nls ) , which is valid for solutions that never vanish .
the advantage is to work with a fixed functional space since @xmath168 , whereas the travelling waves have a limit @xmath169 ( up to a phase factor ) at @xmath170 depending on the speed @xmath51 .
the result of z. lin @xcite establishes rigorously the stability criterion found in @xcite , @xcite .
[ @xcite ] [ stab ] assume that @xmath171 is such that there exists a nontrivial travelling wave @xmath172
. then , there exists some small @xmath173 such that @xmath174 belongs to a locally unique continuous branch of nontrivial travelling waves @xmath68 defined for @xmath175 .
+ @xmath112 assume @xmath176 then , @xmath177 is orbitally stable in the sense that for any @xmath178 , there exists @xmath179 such that if @xmath180 verifies @xmath181 then the solution @xmath18 to ( nls ) such that @xmath182 never vanishes , can be lifted @xmath183 , and we have @xmath184 @xmath185 assume @xmath186 then , @xmath177 is orbitally unstable in the sense that there exists @xmath159 such that , for any @xmath179 , there exists @xmath180 verifying @xmath181 but such that if @xmath18 denotes the solution to ( nls ) with @xmath39 , then there exists @xmath187 such that @xmath18 does not vanish on the time interval @xmath188 $ ] but @xmath189 by one dimensional sobolev embedding @xmath190 , it is clear that since @xmath172 does not vanish in @xmath191 , by imposing @xmath192 small , @xmath163 does not vanish in @xmath193 and thus can be lifted .
we point out that c. gallo in @xcite fills two gaps in the proof of @xcite : the first one concerns the local in time existence for the hydrodynamical system ( see in section [ sexremarkpf ] ) and the second one is about the conservation of the energy and the momentum .
furthermore , we make two additional remarks on the proof of @xcite in section [ sexremarkpf ] .
theorem [ stab ] is stability or instability in the open set @xmath194 for the hydrodynamical distance @xmath195 which is not the energy distance . here ,
@xmath196 $ ] is the principal argument . for the stability , it suffices to consider the phase @xmath197 such that @xmath198 is zero at @xmath199 , where @xmath200 is the translation parameter . for the instability ,
no matter what is the phase @xmath197 .
the result of @xcite is based on the application of the grillakis - shatah - strauss theory @xcite ( see also @xcite , @xcite ) to the hydrodynamical formulation of ( nls ) ( see section [ sexremarkpf ] ) .
one difficulty is to overcome the fact that the hamiltonian operator @xmath201 is not onto . on the energy - momentum diagrams ,
the stability can be checked either on the graphs of @xmath89 and @xmath90 with respect to @xmath51 , either on the concavity of the curve @xmath202 .
indeed , we have seen in @xcite that the so called hamilton group relation @xmath203 holds , where the derivative is computed on the local branch .
therefore , @xmath204 this means that we have stability when @xmath205 is concave , that is @xmath206 , and instability if @xmath205 is convex ,
_ i.e. _ @xmath207 . actually , the proof of @xcite , @xcite provides an explicit control , as shown in the following lemma . [ souskontrol ] under the assumptions of theorem [ stab ] and in the case @xmath112 of stability , we have , provided @xmath208 is small enough , @xmath209 as well as the control [ kontrlin ] _ t0 _ d_hy ( ( t ) , ^i u_c _ * ( - y ) ) k d_hy ( ^ , u_c _ * ) .
the second estimate is not a simple consequence of the control , but relies on a comparison to @xmath123 for some @xmath51 close to @xmath210 instead on a comparison to @xmath172 ( this idea has also been used in @xcite ) .
it follows that , in the definition of stability for @xmath211 , one can take @xmath212 .
let us lay the emphasis on the fact that theorem [ stab ] of z. lin is given in the hydrodynamical distance @xmath213 , which is not the energy distance @xmath214 . as a matter of fact ,
the madelung transform @xmath215 where @xmath216 , @xmath217 and @xmath218 is not so well behaved .
[ cestfaux ] @xmath112 the mapping @xmath219 is an homeomorphism .
+ @xmath185 there exists @xmath220 such that @xmath221 and a sequence @xmath222 of functions in @xmath223 such that , when @xmath224 , @xmath225 therefore , @xmath226 is not locally lipschitz continuous in general .
however , for the stability issues , we compare the @xmath227 and the @xmath213 distances to some fixed travelling wave @xmath228 , which enjoys some nice decay properties at infinity .
let us now stress the link between the two distances @xmath229 and @xmath227 in this case .
[ equivdistance ] let @xmath230 and assume that @xmath231 is a non constant travelling wave with speed @xmath210 that does not vanish .
if @xmath232 , we further assume that assumption @xmath233 is verified
. then , there exists some constants @xmath234 and @xmath235 , depending only on @xmath236 , such that for any @xmath237 verifying @xmath238 , we have @xmath239 an immediate corollary of lemma [ equivdistance ] is that theorem [ stab ] is also a stability / instability result in the energy distance .
if one wishes only a stability / instability result , it is sufficient to invoke the fact that the mapping @xmath240 is an homeomorphism . however , the use of lemma [ equivdistance ] provides a stronger explicit control similar to the one obtained in remark [ souskontrol ] ( see ) . in particular , in the definition of stability for @xmath211 in @xmath241
, one can take @xmath212 .
[ stabenergy ] assume that @xmath171 is such that there exists a nontrivial travelling wave @xmath172 .
then , there exists some small @xmath173 such that @xmath174 belongs to a locally unique continuous branch of nontrivial travelling waves @xmath68 defined for @xmath175 .
+ @xmath112 if @xmath242 , then @xmath177 is orbitally stable in @xmath243 . furthermore ,
if @xmath244 is the ( global ) solution to ( nls ) with initial datum @xmath163 , then we have , for some constant @xmath234 depending only on @xmath245 and provided @xmath246 is sufficiently small , @xmath247 as well as the control @xmath248 @xmath185 if @xmath249 , then @xmath177 is orbitally unstable in @xmath243 . for the gross - pitaevskii nonlinearity ( @xmath250 ) ,
the stability ( for the energy distance @xmath251 ) of the travelling waves with speed @xmath122 was proved by f. bthuel , p. gravejat and j .- c .
saut in @xcite through the variational characterization that these solutions are minimizers of the energy under the constraint of fixed momentum .
however , in view of the energy momentum diagrams in section [ sextravelo ] , this constraint minimization approach can not be used in the general setting we consider here . indeed
, this method provides only stability , but there may exist unstable travelling waves .
moreover , it follows from the proof of theorem [ stab ] that stable waves are local minimizers of the energy at fixed momentum but not necessarily global minimizers .
finally , we emphasize that the spectral methods allow to derive an explicit ( lipschitz ) control in case of stability . another way to prove the orbital stability is to find a liapounov functional .
by liapounov functional , we mean a functional which is conserved by the ( nls ) flow and for which the travelling wave @xmath123 is a local minimum ( for instance , a critical point with second derivative @xmath252 i d for some @xmath160 ) .
such a liapounov functional always exists in the grillakis - shatah - strauss theory when @xmath242 , as shown by theorem a in appendix a. its direct application to our problem leads us to define the functional in @xmath50 @xmath253 where @xmath254 is some positive parameter .
it turns out that @xmath255 is such a liapounov functional when @xmath254 is sufficiently large .
since the proof relies on the grillakis - shatah - strauss framework , we have to work in the hydrodynamical variables .
by lemma [ equivdistance ] , we recover the case of the energy distance .
[ liapoupou ] assume that for some @xmath256 and @xmath257 small , @xmath258 \ni c \mapsto
u_c \in \mathcal{z } $ ] is a continuous branch of nontrivial travelling waves with @xmath242 . if @xmath259 there exists @xmath260 and @xmath234 , depending only on @xmath172 , such that for any @xmath261 with @xmath262 , we have @xmath263 and analogously with @xmath213 replaced by @xmath214 .
consequently , @xmath264 is orbitally stable in @xmath265 and in @xmath243 .
furthermore , if @xmath244 is the ( global ) solution to ( nls ) with initial datum @xmath163 , then we have @xmath266 provided @xmath267 is sufficiently small , and analogously with @xmath213 replaced by @xmath214 . for the travelling waves for ( nls ) in dimension one
, this type of liapounov functional appears for the first time in the paper @xcite by i. barashenkov .
however , in @xcite , the problem is treated directly on the wave function @xmath38 , whereas the correct proof holds on the hydrodynamical variables , in particular because of the gauge invariance @xmath268 .
for instance , the work @xcite suggests that we have stability for @xmath269 perturbations , whereas it holds only for perturbations in the energy space . finally , we fill some gaps in the proof of @xcite . in the grillakis - shatah - strauss theory @xcite ,
the instability is not shown by proving the existence of a unstable eigenvalue for the linearized ( nls ) and then a nonlinear instability result ( see however @xcite when the hamiltonian skew - adjoint operator is onto ) .
there exists , however , some general results that prove the existence of unstable eigenvalues . for the instability of bound states for ( nls ) ( and also for nonlinear klein - gordon equation ) , that is solutions of the form @xmath270 , the work @xcite by m. grillakis shows that the condition @xmath271 is sufficient for the existence of such an unstable eigenvalue . however , the proof relies on the fact that the bound states are real valued functions ( up to a phase factor ) and it is not clear whether it extends to the case of travelling waves we are studying . indeed , since we have to work in hydrodynamical variables in order to have a fixed functional space , the linearized operator does not have ( for @xmath272 ) the structure required for the application of @xcite .
another general result is due to o. lopes @xcite but it assumes that the linearized equation can be solved using a semigroup .
this is not the case for our problem once it is written in hydrodynamical variables ( see below ) .
finally , the paper @xcite by z. lin proposes an alternative approach for the existence of unstable eigenvalues .
the method has the advantage of allowing pseudo - differential equations ( like the benjamin - ono equation ) .
however , the results are given for three model equations involving a scalar unknown , and it is not clear whether the proof can be extended to the case of systems .
the linearization of ( nls ) near the travelling wave @xmath172 in the frame moving with speed @xmath83 is [ nlslin ] i - i c _ * _ x + ^2_x + f(|u_c_*|^2 ) + 2 growing modes @xmath273 leads to the eigenvalue problem [ pbvp ] i w - i c _ * _ x w + ^2_x w + w f(|u_c_*|^2 ) + 2 w | u_c _ * f(|u_c_*|^2 ) u_c _ * = 0 , with @xmath274 and @xmath275 .
for one dimensional problems , the linear instability is commonly shown through the use of evans functions ( see the classical paper @xcite and also the review article @xcite ) . for our problem , we look for an unstable eigenvalue for the equation written in hydrodynamical variables , namely we look for exponentially growing solutions @xmath276 of the linear problem ( written in the moving frame ) [ eklin ] \ { ll = 0 + + = 0 , . where @xmath277 is the reference solution .
the advantage is here again to work with a fixed functional space in variables @xmath278 . due to the term
@xmath279 , this equation can not be solved using a semigroup , except in the trivial case where @xmath280 is constant , hence the result of @xcite does not apply .
however , system is a particular case of the euler - korteweg system for capillary fluids ( see @xcite for a survey on this model ) . we may then use a linear instability result already shown for the euler - korteweg system with the evans function method , as in @xcite by k. zumbrun for a simplified system , and more recently in @xcite by s. benzoni - gavage for the complete euler - korteweg system .
[ vpinstable ] assume that for some @xmath256 and @xmath257 small , @xmath258 \ni c \mapsto u_c \in \mathcal{z } $ ] is a continuous branch of nontrivial travelling waves with @xmath186 then , there exists exactly one unstable eigenvalue @xmath281 for and @xmath282 , that is ( nls ) is ( spectrally ) linearly unstable .
once we have shown the existence of an unstable eigenvalue for the linearized ( nls ) equation , we can prove a nonlinear instability result as in @xcite , @xcite .
note that here , we no longer work in the hydrodynamical variables , where the high order derivatives involve nonlinear terms , but on the semilinear ( nls ) equation .
[ instabh1 ] under the assumptions of theorem [ vpinstable ] , @xmath172 is unstable in @xmath283 ( endowed with the natural @xmath269 distance ) : there exists @xmath284 such that for any @xmath160 , there exists @xmath285 such that @xmath286 but if @xmath287 denotes the maximal solution of ( nls ) , then there exists @xmath288 such that @xmath289 . since the proof is very similar to the one in @xcite , @xcite ,
we omit it .
we may actually prove a stronger instability result , since the above one is not proved by tracking the exponentially growing mode . in @xcite ,
a spectral mapping theorem is shown and used to show the nonlinear instability by tracking this exponentially growing mode , which is a natural mechanism of instability . in appendix b , we show that this spectral mapping theorem holds for a wide class of hamiltonian equation .
the direct application of corollary b.2 in appendix b gives the following nonlinear instability result .
[ instabamelioree ] we make the assumptions of theorem [ vpinstable ] , so that there exists an unstable eigenmode @xmath290 , @xmath291 .
there exists @xmath292 such that for any solution @xmath293 of the linearized equation , we have the growth estimate of the semigroup @xmath294 moreover , @xmath172 has also the following instability property : there exists @xmath295 , @xmath296 and @xmath297 , such that for any @xmath298 , the solution @xmath299 to ( nls ) with initial datum @xmath300 exists at least on @xmath301 $ ] and verifies @xmath302 in particular , for @xmath303 and @xmath304 , we have @xmath305 which implies @xmath306 as well as @xmath307 with the above result , we then show the nonlinear instability also in the energy space , and thus recover the instability result of z. lin but this time by tracking the unstable growing mode . in this section ,
we investigate the question of stability in the degenerate case @xmath308 . in @xcite ( see also @xcite ) , a stability result for the wave of speed @xmath210 is shown when the action @xmath309 ( on the local branch ) is such that , for instance , @xmath310 is positive for @xmath311 but vanishes for @xmath312 . in the energy - momentum diagrams of section [ sextravelo ] ,
the situation is different since @xmath313 changes sign _ at _ the cusps , or , equivalently , the action @xmath309 ( on the local branch ) changes its concavity _ at _ the cusp .
the work @xcite by a. comech and d. pelinovsky shows that for the nonlinear schrdinger equation , a bound state associated with a cusp in the energy - charge diagram is unstable .
the proof relies on a careful analysis of the linearized equation , which is spectrally stable , but linearly unstable ( with polynomial growth for the linear problem ) .
a similar technique was used by a. comech , s. cuccagna and d. pelinovsky in @xcite for the generalized korteweg - de vries equation .
then , m. ohta in @xcite also proves the nonlinear instability of these `` bound states '' using a liapounov functional as in @xcite .
however , in @xcite , it is assumed that @xmath314 and is onto , which are both not true here ( and there are further restrictions due to the introduction of an intermediate hilbert space ) .
recently , m. maeda has extended in @xcite the above instability result , removing some assumptions in @xcite .
we show the instability of travelling waves associated with a cusp in the energy - momentum diagram in the generic case where @xmath315 .
our approach follows the lines of @xcite , but with some modifications since our problem does not fit exactly the general framework of this paper . in particular , we can not find naturally a space `` @xmath316 '' , and some functions appearing in the proof do not lie in the range of the skew - adjoint operator @xmath317 involved in the hamiltonian formalism .
we overcome this difficulty using an approximation argument ( similar to the one used in @xcite ) .
[ stabcusp ] assume that for some @xmath256 and @xmath257 small , @xmath258 \ni c \mapsto u_c
\in \mathcal{z } $ ] is a continuous branch of nontrivial travelling waves with @xmath318 and assume in addition that @xmath13 is of class @xmath319 .
then , @xmath172 is orbitally unstable in @xmath320 . when @xmath322 , we have two types of stationary waves : the bubbles , when @xmath323 , are even functions ( up to a translation ) that do not vanish , and the kinks , when @xmath324 , are odd functions ( up to a translation )
. the instability of stationary bubbles has been shown by a. de bouard @xcite ( and is true even in higher dimension ) .
the proof in @xcite relies on the proof of the existence of an unstable eigenvalue for the linearized ( nls ) , and then the proof of a nonlinear instability result .
an alternative proof of the linear instability of the bubbles is given in @xcite ( theorem 3.11 @xmath185 ) .
[ @xcite ] [ unstabdebouard ] assume that there exists a bubble , that is a nontrivial stationary ( @xmath78 ) wave @xmath325 which does not vanish .
then , @xmath325 is ( linearly and nonlinearly ) unstable in @xmath326 ( endowed with the natural @xmath269 metric ) , that is there exists @xmath284 such that for any @xmath160 , there exists @xmath327 such that @xmath328 but if @xmath329 denotes the maximal solution of ( nls ) , then there exists @xmath288 such that @xmath330 .
actually , in the same way that corollary [ instabamelioree ] is a better instability result than corollary [ instabh1 ] , we have the following stronger instability result , which is a direct consequence of corollary b.2 in appendix b. [ unstabbubble ] assume that there exists a bubble , that is a nontrivial stationary ( @xmath78 ) wave @xmath325 which does not vanish .
then , @xmath325 is ( nonlinearly ) unstable in @xmath326 , @xmath331 and @xmath265 in the same sense as in corollary [ instabamelioree ] .
finally , we would like to emphasize that we may recover the instability result for bubbles from the proof of theorem [ stab ] , relying on the hydrodynamical form of ( nls ) , which holds true here since bubbles do not vanish .
our result holds in the energy space and for the hydrodynamical distance .
[ unstab ] assume that there exists a bubble , that is a nontrivial stationary ( @xmath78 ) wave @xmath325 which does not vanish
. then , there exists some small @xmath173 such that @xmath332 belongs to a locally unique continuous branch of nontrivial travelling waves @xmath68 defined for @xmath333 .
then , @xmath334 has a derivative at @xmath335 , @xmath336 and @xmath337 is orbitally unstable for the distances @xmath338 and @xmath213 .
_ we give a proof based on the argument of z. lin @xcite , which is possible since @xmath325 is a bubble hence does not vanish and the spectral decomposition used in @xcite still holds when @xmath78 .
moreover , it is clear that the mapping @xmath339 is smooth up to @xmath78 , using the uniform exponential decay at infinity near @xmath78 and arguing as in @xcite .
therefore , it suffices to show that @xmath340 . from the expression of the momentum given in subsection 1.2 in @xcite
, we have , for @xmath341 , @xmath342 since @xmath323 .
indeed , we are allowed to pass to the limit in the integral once it is written with the change of variables @xmath343 : @xmath344 since @xmath323 .
therefore , @xmath345 since @xmath346 ( @xmath325 is not trivial ) .
the conclusion follows then from the proof of theorem [ stab ] .
+ when we know that @xmath347 , we may also use the evans function as in theorem [ vpinstable ] to show the existence of an unstable eigenmode .
however , due to the fact that the kink @xmath348 is real - valued , we can use the arguments in @xcite , @xcite .
we now turn to the case of the kinks ( @xmath324 and @xmath348 is odd up to a translation ) .
since @xmath325 vanishes at the origin , the hydrodynamical form of ( nls ) can not be used . the stability of the kink as attracted several recent works .
l. di menza and c. gallo @xcite have investigated the linear stability through the vakhitov - kolokolov function vk , defined by @xmath349^{-1 } ( \p_x u_0 ) \big ) ( \p_x u_0 ) \ dx , \ ] ] where @xmath325 is the kink , for @xmath350 for some @xmath351 .
they show that the vakhitov - kolokolov function vk has a limit vk@xmath352 when @xmath353 . if vk@xmath354 , then the linearization of ( nls ) around the kink has an unstable real positive eigenvalue .
when vk@xmath355 , the linearization of ( nls ) around the kink has a spectrum included in @xmath356 ( spectral stability ) .
note that the approach of @xcite ( extending @xcite ) does not give directly unstable eigenvalues in the case @xmath357 .
recently , the link between the quantity @xmath313 and the sign of vk@xmath352 has been given by d. pelinovsky and p. kevrekidis @xcite ( proof of lemma 3.10 there , the factor @xmath358 coming from the coefficients of the ( nls ) equation in @xcite ) : [ jolie ] 2 _ 0 = _ c 0 ; and they also prove , in a different way from @xcite that we have spectral stability when @xmath359 and existence of an unstable eigenvalue ( in @xmath360 ) if @xmath361 .
it is shown in @xcite that the limit @xmath362 does exist .
actually , they prove that the function @xmath363 is of class @xmath364 and that the derivative at @xmath78 is also given by ( see ) @xmath365^{-1 } ( \p_x u_0 ) \big ) ( \p_x u_0 ) \ dx \nonumber \\ = & \ , 2 \sqrt{2 } \int_\r { \rm i m } \big ( \frac{\p u_c}{\p c}_{|c=0 } \big ) \p_x u_0 \ dx .\end{aligned}\ ] ] our next lemma gives an explicit formula of the expression , involving only the nonlinearity @xmath13 .
[ prolongement ] assume that @xmath348 is a kink
. then , there exists @xmath366 such that @xmath348 belongs to the ( locally ) unique branch @xmath367 .
moreover , @xmath368 as @xmath369 and the continuous extension @xmath363 has a derivative at @xmath370 given by @xmath371 the advantage of the formula given in lemma [ prolongement ] compared to is that it allows a direct computation of @xmath372 when @xmath13 is known , which does not require to compute numerically @xmath325 and @xmath373 .
for instance , it is quite well adapted to the stability analysis as in @xcite .
let us observe that it may happen that a kink is unstable ( see @xcite , @xcite ) .
+ in the case of linear instability , @xcite shows that then , nonlinear instability holds .
actually , c. gallo and l. di menza prove in @xcite a stronger result , where they show that the @xmath374 norm ( and not only the @xmath375 norm ) does not remain small .
[ @xcite ] [ unstabdmg ] assume that there exists a kink , that is a nontrivial stationary ( @xmath78 ) wave @xmath325 vanishing somewhere , and satisfying @xmath376 .
then , @xmath325 is ( linearly and nonlinearly ) unstable in the sense that there exists @xmath284 such that for any @xmath160 , there exists @xmath327 such that @xmath328 but if @xmath329 denotes the maximal solution of ( nls ) , then there exists @xmath288 such that @xmath377 .
the proof in @xcite relies on the tracking of the exponentially growing eigenmode .
one may actually improve slightly the result as this was done in corollary [ instabamelioree ] . as a matter of fact ,
this was the result in theorem [ unstabdmg ] that has motivated us for corollary [ instabamelioree ] .
we focus now on the nonlinear stability issue when there is linear ( spectral ) stability , that is when @xmath378 . concerning the gross - pitaevskii nonlinearity ( @xmath379 ) , for which we have @xmath378 , we quote two papers on this question .
the first one is the work of p. grard and z. zhang @xcite where the stability is shown by inverse scattering , hence in a space of functions sufficiently decaying at infinity .
the analysis then relies on the integrability of the one - dimensional ( gp ) equation .
the other work is by f. bthuel , p. gravejat , j .- c .
saut and d. smets @xcite .
they prove the orbital stability of the kink of the gross - pitaevskii equation by showing that the kink is a global minimizer of the energy under the constraint that a variant of the momentum is fixed ( recall that the definition of the momentum has to be clarified for an arbitrary function in the energy space ) , and that the corresponding minimizing sequences are compact ( up to space translations and phase factors ) . in this approach
, it is crucial ( see @xcite , @xcite ) that @xmath380 in order to prevent the dichotomy case for the minimizing sequences .
however , since the energy of the kink is equal to @xmath381 whereas its momentum is always equal to @xmath382 , it is clear that the condition @xmath383 does not hold in general , as shown in the following example . +
* example . * for @xmath384 ,
consider @xmath385 which is smooth and decrease to @xmath386 as the gross - pitaevskii nonlinearity .
we have @xmath387 , @xmath388 , @xmath389 and @xmath390 for @xmath391 large ( the left - hand side tends to @xmath392 ) , and numerical computations show that it is the case for @xmath393 . furthermore , lemma [ prolongement ] gives [ cestpaslemoment ] _ |c=0 = - + 12 _ 0 ^ 1 ( - 1 ) d . since @xmath394
, it can be easily checked that the right - hand side of is a decreasing function of @xmath395 tending to @xmath396 when @xmath397 ( by monotone convergence ) . in particular , for any @xmath384 , we have @xmath398 , that is the kink is always ( linearly ) stable .
the energy - momentum diagram for this type of nonlinearity with @xmath399 large is as in on the right of figure [ zzcqsii ] ( the left part correspond to @xmath391 smaller ) .
+ in comparison with the constraint minimization approach as in @xcite , which allows to establish a global minimization result , the spectral methods as in @xcite allow to put forward locally minimizing properties , which turn out to be useful for the stability analysis in dimension 1 . in the stability analysis of the kink , one issue is the definition of the momentum @xmath90 , which was up to now given only for maps in @xmath400 , that is for maps that never vanish , but the kink vanishes at the origin . in @xcite , the notion of momentum was extended to the whole energy space @xmath49 , hence including maps vanishing somewhere , as a quantity defined mod @xmath401 , and was called `` untwisted momentum '' .
this notion will be useful for our stability result .
[ @xcite ] if @xmath402 , the limit @xmath403\ ] ] exists in @xmath404 .
the mapping @xmath405 is continuous and if @xmath402 verifies @xmath406 ( i.e. @xmath407 ) , then @xmath408 . finally , if @xmath409 is a solution to ( nls ) , then @xmath410 does not depend on @xmath42 . _
_ for sake of completeness , we recall the proof of @xcite .
let @xmath411 and let us verify the cauchy criterion . since @xmath412 at @xmath413 , we may lift @xmath414 in @xmath415 and in @xmath416 for some @xmath417 sufficiently large . for @xmath418
, we thus have in @xmath404 @xmath419 \\ & \quad \quad - \big [ \int_{-r}^{+r } \langle i \psi | \p_x \psi \rangle \ dx - r_0 ^ 2 \big ( \arg(\psi(+r ) ) - \arg(\psi(-r ) ) \big ) \big ] \\ & = \int_{r}^{r ' } \langle i \psi | \p_x \psi \rangle \ dx + \int_{- r'}^{- r } \langle i \psi | \p_x \psi \rangle \ dx \\ & \quad \quad \quad - r_0 ^ 2 \big ( \arg(\psi(r ' ) ) - \arg(\psi(r ) ) \big ) + r_0 ^ 2 \big ( \arg(\psi ( - r ' ) ) - \arg(\psi(-r ) ) \big ) \\ & = \int_{r}^{r ' } a_+^2 \p_x \phi_+ \ dx + \int_{- r'}^{- r } a_-^2 \p_x \phi_- \
& \quad \quad \quad - r_0 ^ 2 \big ( \phi_+(r ' ) - \phi_+(r ) \big ) + r_0 ^ 2 \big ( \phi_-(- r ' ) - \phi_-(-r ) \big ) \\ & = \int_{r}^{r ' } ( a_+^2 - r_0 ^ 2 ) \p_x \phi_+ \
dx + \int_{- r'}^{- r } ( a_-^2 - r_0 ^ 2 ) \p_x \phi_- \ dx .\end{aligned}\ ] ] the absolute value of each term is @xmath420 thus tends to zero if @xmath421 .
thus , @xmath422 is well - defined .
the proof of the continuity follows the same lines , and allows to show that @xmath423 is actually locally lipschitz continuous .
let @xmath411 . if @xmath424 and @xmath425 is small enough , we have @xmath426 as small as we want . in particular , if @xmath427 is large enough so that @xmath428 for @xmath429 , we have @xmath430 for @xmath431 . as a consequence ,
writing @xmath414 and @xmath432 in @xmath433 and in @xmath434 , we have , in @xmath435 and for @xmath436 , @xmath437 \\ & \quad \quad - \big [ \int_{-r}^{+r } \langle i \tilde{\psi } | \p_x \tilde{\psi } \rangle \ dx - r_0 ^ 2 \big ( \arg(\tilde{\psi}(+r ) ) - \arg(\tilde{\psi}(-r ) ) \big ) \big ] \\ & = \int_{-r_0}^{+r_0 } \langle i ( \psi - \tilde{\psi } ) | \p_x \psi \rangle + \langle i \tilde{\psi } | \p_x ( \psi - \tilde{\psi } ) \rangle \ dx \\ & \quad \quad \quad + \int _
{ r_0 } ^r a_+^2 \p_x \phi_+ - \tilde{a}_+^2 \p_x \tilde{\phi}_+ \ dx - r_0 ^ 2 \big ( \phi_+(+r ) - \tilde{\phi}_+(+r ) \big ) \\ & \quad \quad \quad + \int_{-r}^ { - r_0 } a_-^2 \p_x \phi_- - \tilde{a}_-^2 \p_x \tilde{\phi}_- \ dx + r_0 ^ 2 \big ( \phi_-(- r ) - \tilde{\phi}_-(-r ) \big ) \\ & = \int_{-r_0}^{+r_0 } \langle i ( \psi - \tilde{\psi } ) | \p_x \psi \rangle + \langle
i \tilde{\psi } | \p_x ( \psi - \tilde{\psi } ) \rangle \ dx \\ & \quad \quad \quad + r_0 ^ 2 \big ( \phi_+(+r_0 ) - \tilde{\phi}_+(+r_0 ) \big ) + r_0 ^ 2 \big ( \phi_-(-r_0 ) - \tilde{\phi}_-(-r_0 ) \big ) \\ & \quad \quad \quad + \int _ { r_0 } ^r ( a_+^2 -r_0 ^ 2 ) \p_x \phi_+ - ( \tilde{a}_+^2 -r_0 ^ 2 ) \p_x \tilde{\phi}_+ \ dx + \int _ { - r}^{- r_0 } ( a_-^2 -r_0 ^ 2 ) \p_x \phi_- - ( \tilde{a}_-^2 -r_0 ^ 2 ) \p_x \tilde{\phi}_- \ dx .\end{aligned}\ ] ] we now estimate all the terms . for the last line , we use cauchy - schwarz to get @xmath438 , and similarly for the other terms . moreover , using that @xmath439 , we get by cauchy - schwarz @xmath440 ) } \leq | ( \psi - \tilde{\psi})(0 ) | + \sqrt { r_0 } |\!| \p_x \psi - \p_x \tilde{\psi } |\!|_{l^2(\r ) } \leq k(r_0 ) d_\mathcal{z } ( \psi , \tilde{\psi } ) $ ] .
thus , the terms of the second line can be estimated by @xmath441 , and for those of the first line , they can also be bounded by @xmath442 . passing to the limit
as @xmath443 then gives @xmath444 this completes the proof for the definition of @xmath423 . to show that @xmath423 is constant under the ( nls ) flow
, we use that @xmath445 and the approximation by smoother solutions ( see proposition 1 in @xcite ) . + for the stability of the kink
, we can no longer use the grillakis - shatah - strauss theory applied to the hydrodynamical formulation of ( nls ) , since the kink vanishes at the origin .
therefore , it is natural to consider the liapounov functional @xmath255 introduced in section [ sex0csliap ] , which becomes in the stationary case @xmath113 : @xmath446 since the momentum @xmath90 is not well - defined in @xmath58 , we have to replace it by the untwisted momentum @xmath423 , which is defined modulo @xmath447 .
consequently , it is natural to define the functional in @xmath49 @xmath448 which is well - defined and continuous in @xmath49 since @xmath449 is @xmath450-periodic .
in addition , @xmath451 is conserved by the ( nls ) flow as @xmath89 and @xmath423 .
[ statiomini ] assume that there exists a kink , that is a nontrivial stationary ( @xmath78 ) wave @xmath325 which is odd .
assume also that @xmath452 then , there exists some small @xmath453 such that @xmath348 is a local minimizer of @xmath451 .
more precisely , denoting @xmath454 we have , for any @xmath455 , @xmath456 the crucial point in this result is to prove that the functional @xmath457 controls the infimum @xmath458 . from this locally minimizing property of the kink when @xmath459 , we infer its orbital stability , provided we can prove some compactness on the minimizing sequences .
our method allows to infer a control on the distance of the solution to ( nls ) to the orbit of the kink , but it is much weaker than those obtained by spectral methods in lemma [ souskontrol ] or corollary [ stabenergy ] for instance .
[ stabmini ] assume that there exists a kink , that is an odd nontrivial stationary ( @xmath78 ) wave @xmath325 and that @xmath452 then , @xmath325 is orbitally stable in @xmath320 .
moreover , if @xmath299 is the ( global ) solution to ( nls ) with initial datum @xmath163 , we have the control @xmath460{\mathscr{k}(\psi^\ini ) - e(u_0 ) } \leq k \sqrt[4]{d_{\mathcal{z } } ( \psi^\ini , u_0 ) } \ ] ] provided that the right - hand side is sufficiently small .
this result settles the nonlinear stability under the condition @xmath461 for a general nonlinearity @xmath13 .
in particular , it may be applied to the nonlinearity @xmath13 given in the example above .
it shows that the stability of the kink holds with @xmath462 .
we do not claim that the exponent @xmath463 is optimal . for a complete study of the stability of the travelling waves
, it would remain to investigate the case of the sonic ( @xmath131 ) travelling waves ( when they exist ) .
the methods we have developed do not apply directly , and we give in section [ sexsonic ] some of the difficulties associated with this critical situation .
for simplicity , we shall denote @xmath464 we freeze the invariance by translation by imposing @xmath465 ( hence also @xmath466 ) even , so that we can use the formulas in @xcite .
in particular , it suffices to show the asymptotics for @xmath467 : the case @xmath468 follows by symmetry .
we start with the proof of case @xmath469 . under assumption @xmath103 and since @xmath470
, we infer the taylor expansion @xmath471 \xi^{m+3 } + \bo ( \xi^{m+4 } ) = \lambda_m \xi^{m+3 } + \bo ( \xi^{m+4 } ) \end{aligned}\ ] ] since when @xmath103 holds , all the terms @xmath472 cancel out .
the coefficient @xmath473 is not zero by assumption .
note that the existence of a nontrivial sonic wave , which depends on the global behaviour of @xmath474 , imposes that @xmath475 when @xmath476 is small and has the sign of @xmath477 .
therefore , from the formula ( following from the hamiltonian equation @xmath478 , see @xcite for example ) @xmath479 and since there holds , as @xmath480 ( with the sign of @xmath481 ) , @xmath482 } \
d\xi \\ = & \ , - \frac{2 \ { \rm sgn}(\xi_{\cs } ) } { m+1 } \big ( \frac{1 } { \sqrt{- \lambda_m \eta^{m+1 } } } - \frac{1 } { \sqrt{- \lambda_m \xi_{\cs}^{m+1 } } } \big ) + \left\{\begin{array}{ll } \ds { \bo(1 ) } & \quad { \rm if } \ m = 0 \\ \ \\ \ds { \bo ( | \ln | \eta | | ) } & \quad { \rm if } \ m = 1 \\ \ \\
\ds { \bo ( \eta^{- \frac{m-1}{2 } } ) } & \quad { \rm if } \ m \geq 2 \end{array}\right .
\end{aligned}\ ] ] ( here , we use that the last integrand is @xmath483 as @xmath484 ) , it follows that , as @xmath467 , @xmath485 this shows the asymptotics for the modulus , or @xmath486 . the asymptotic expansion for @xmath487 is easily deduced from the equation on the phase @xmath488 , and the phase @xmath489 is then computed by integration , which completes the proof of case @xmath469 .
the proof of @xmath185 is easier . indeed , in this case , the function @xmath474 has the expansion @xmath490 hence @xmath491 as a consequence , the result follows from the expansion , for @xmath492 @xmath493 } \
d\xi \\ = & \ , { \rm sgn ( \xi_c ) } \frac { \ln ( \eta / \xi_c ) } { \sqrt { \cs^2 - c^2 } } + \int_{\xi_c}^{0 } \frac{\bv(\xi ) } { \sqrt{-\bv_c(\xi ) } \sqrt { ( \cs^2 - c^2 ) \xi^2 } \big[\sqrt{-\bv_c(\xi ) } + \sqrt { ( \cs^2 - c^2 ) \xi^2 } \big ] } \
d\xi + \bo ( \eta ) \end{aligned}\ ] ] since the integrand for the last integral is continuous at @xmath494 . this yields the desired expansion for the modulus : @xmath495 } \ d\xi \big ) + \bo \big [ \exp \big ( - 2 x \sqrt { \cs^2 - c^2 } \big ) \big ] \\ = & \ , m_c \exp \big ( - x \sqrt { \cs^2 - c^2 } \big ) + \bo \big [ \exp \big ( - 2 x \sqrt { \cs^2 - c^2 } \big ) \big ] , \end{aligned}\ ] ] with @xmath496 } \ d\xi \big ) \not = 0,\ ] ] and hence for the phase by similar computations to those above .
the proof of case @xmath112 is similar , separating the case @xmath497 of the kink ( even solution ) from the case @xmath498 of the bubble ( odd solution ) and is omitted .
* ( i ) the mapping @xmath499 is an homeomorphism . *
let @xmath500 , @xmath501 such that @xmath502 for @xmath213 . then , @xmath503 in @xmath269 , @xmath504 in @xmath505 and we may assume ( possibly adding some multiple of @xmath401 to @xmath506 , that @xmath507 .
we write , using the embedding @xmath508 for the before last line , @xmath509 from the convergences we have .
now observe that @xmath510 pointwise , hence , by the dominated convergence theorem , @xmath511 and similarly for the other term .
therefore , @xmath512 as wished .
let now @xmath500 , @xmath501 such that @xmath502 for @xmath214 . then , @xmath513 in @xmath505 and @xmath514 .
since @xmath515 is @xmath516-lipschitz continuous , we infer for the modulus @xmath517 moreover , @xmath518 and this implies @xmath519 .
therefore , it suffices to show that @xmath504 in @xmath520 .
we use the formula @xmath521 , which yields @xmath522 hence [ lipp ] _ x _
l^2 a_n - a _ l^2 + _ x - _ x _ n _ l^2 + | _ n - | _ x _ l^2 .
the first two terms tend to zero as @xmath224 . for the last term
, we use here again the dominated convergence theorem since @xmath523 pointwise .
this concludes for @xmath112 .
+ * proof of ( ii ) .
* let us define @xmath524 by @xmath525 . then , straightforward computations give @xmath526 and , for @xmath527 , by monotonicity of @xmath528 , [ lo ] _
x^2x ( _ x _ * ) ^2 d x x .
we now consider @xmath529 defined by @xmath530 if @xmath531 or @xmath532 , @xmath533 if @xmath534 , @xmath535 if @xmath536 and @xmath537 if @xmath538 .
then , we easily obtain @xmath539 moreover , @xmath540 and , by our choice of @xmath541 and using , @xmath542 since @xmath543 , it follows that , as wished @xmath544 we do not know whether the mapping @xmath240 is locally lipschitz , but is probably not .
note first that since @xmath236 does not vanish , if @xmath545 is sufficiently small and @xmath238 , then @xmath546 , hence @xmath547 in @xmath63 , thus @xmath548 does not vanish , may be lifted @xmath549 , and we may further assume @xmath550 $ ] . in , we can easily check that the terms leading to the `` @xmath551 '' are indeed controlled by @xmath552 .
in other words , we have @xmath553 provided @xmath554 is small enough . in order to bound the two remaining terms
, we write , for @xmath555 , @xmath556 which implies , using that @xmath557 is @xmath516-lipschitz and the cauchy - schwarz inequality , [ integ ] | 1 - ^ i ( _ * ( x ) - ( x ) ) | | ( 0 ) - _ * ( 0 ) | + u - u _ * _ l^2 .
consequently , @xmath558 and @xmath559 both terms are @xmath560 .
indeed , @xmath561 is a travelling wave , hence @xmath562 , @xmath563 , @xmath528 are bounded functions which decay at infinity exponentially if @xmath564 ( _ cf . _ proposition [ decroiss ] @xmath112 or @xmath185 ) . if @xmath131 , since assumption @xmath233 is satisfied , we invoke proposition [ decroiss ] @xmath469 , which ensures that @xmath528 and @xmath563 decay at the rate @xmath565 ( @xmath563 decays faster actually ) .
therefore , @xmath566 .
gathering these estimates provides @xmath567 on the other hand , from and the estimate @xmath568 ( see the proof of @xmath112 ) , we infer @xmath569 using here again the estimate @xmath570 , we deduce @xmath571 the proof is complete .
we would like to point out two minor points concerning the proof of theorem [ stab ] by z. lin .
we recall that the proof of z. lin @xcite relies on the grillakis - shatah - strauss theory @xcite once we have written ( nls ) under the hydrodynamical form , denoting @xmath572 , @xmath573 : @xmath574 or more precisely , with @xmath575 and denoting @xmath576 , @xmath577 the variational derivative , [ hydrohamilto ] ( c + u ) = j ( c + ) , j ( cc 0 & _ x + _ x & 0 ) .
the first remark is that the scalar product in the hilbert space @xmath578 can not be @xmath579 as used in @xcite , but the natural one is @xmath580 .
this requires to make some minor changes in the proof , especially not to identify @xmath581 with @xmath269 .
for instance , a linear mapping @xmath582 is associated with the momentum through the formula @xmath583 for the ( non hilbertian ) scalar product @xmath584 .
the correct definition is actually @xmath585 where @xmath586 is the canonical injection . as already mentioned in [ sex0cs ] , the two points in the proof of @xcite that have been completed by c. gallo in @xcite are that : @xcite uses a local in time existence for the hydrodynamical system in @xmath587 , and not only in @xmath588 ; and that the energy and the momentum are indeed conserved for the local solution if the initial datum does not vanish .
+ the second point is that in the proof of stability ( theorem 3.5 in @xcite ) , it is made use of the fact that if @xmath589 and @xmath590 is a sequence such that @xmath591 and @xmath592 , then there exists a sequence @xmath593 such that @xmath594 in @xmath595 , @xmath596 and @xmath597 . in the context of bound states ,
the existence of such a sequence @xmath593 follows by simple scaling in space , since then the momentum or charge is simply @xmath598 . however , for the one dimensional travelling waves for ( nls ) , the momentum @xmath90 is scaling invariant .
we do not know if the existence of such a sequence holds in a general framework , but for the problem we are studying , we can rely on the following lemma , which is an adaptation of lemma 6 in @xcite ( see also lemma in @xcite ) .
[ raccord ] there exists @xmath599 and @xmath600 , depending only on @xmath13 such that for any @xmath601 and @xmath602 with @xmath603 , there exists @xmath604 verifying @xmath605 and @xmath606 _ proof . _ if @xmath607 , we simply take @xmath608 .
we then assume @xmath609 , since the case @xmath610 will follow by complex conjugation .
we then define , for some small @xmath611 to be determined later , @xmath612 = a \ex^{i \vp } .\ ] ] it is clear that @xmath613 , \c ) $ ] and that @xmath614 , thus @xmath615 provided we choose @xmath616 . moreover , using that the phase @xmath617 has compact support @xmath618 $ ] , @xmath619 for the last integral ,
the first factor is equal to @xmath75 if @xmath620 and the second factor is equal to @xmath621 when @xmath622 .
hence , direct computation gives @xmath623 for the energy part , notice first that @xmath624 is as small as we want if @xmath625 and @xmath626 are chosen sufficiently small .
therefore , @xmath627 by simple computations , we have @xmath628 since @xmath629 , which concludes the proof .
+ we then consider a sequence @xmath630 and show the existence of the desired sequence @xmath631 .
we recall that @xmath632 ( resp .
@xmath633 ) is associated with a mapping @xmath634 ( resp .
@xmath236 ) that does not vanish .
we have @xmath635 , thus for @xmath636 large enough , @xmath637 . for @xmath636 fixed , we now pick @xmath638 large enough so that @xmath639 ^ 2 .\ ] ] in particular , by sobolev embedding , @xmath640 we are now in position to apply ( for @xmath641 large ) lemma [ raccord ] with @xmath642 .
this provides the mapping @xmath643 . since @xmath644 , for @xmath641 large enough
, there exists @xmath645 such that @xmath646 .
we then consider the mapping @xmath647 defined by @xmath648 from the construction of @xmath649 and the phase factor @xmath650 , @xmath651 is well - defined and continuous .
it is clear that @xmath652 for every ( large ) @xmath641 , and that @xmath653 converges to @xmath654 as @xmath224 . denoting @xmath655 the hydrodynamical expression of @xmath656 , it remains to show that @xmath594 in @xmath657 .
we thus compute , with the definition of @xmath656 , @xmath658 ^ 2 + k | p ( \psi_n ) - p(u _ * ) | \to 0 .\end{aligned}\ ] ] for the before last inequality , we have used that for @xmath659 , @xmath660 has modulus uniformly close to @xmath661 , hence @xmath662 . note that the construction still holds for the energy distance , the computations being similar .
* proof of estimate @xmath663 .
* instead of concluding the stability proof as in @xcite , we can notice that we have actually the bound [ minicontr ] e_hy ( ) - e_hy ( _ c _ * ) _ y || - _ c _ * ( - y ) @xmath664 and @xmath665 for some small @xmath666 .
if @xmath163 does not have momentum equal to @xmath667 , we use lemma [ raccord ] to infer that there exists @xmath668 , with momentum equal to @xmath669 , and such that @xmath670 and @xmath671 . therefore , for @xmath672 , denoting @xmath673 and @xmath674 the hydrodynamical variables for @xmath38 and @xmath675 , @xmath676 \\ \leq & \ , \sqrt{k } \sqrt{e ( \tilde{\psi}(t ) ) - e ( u_{c _ * } ) } + \bo ( \sqrt{| p ( \psi(t ) ) - p(u_{c _ * } ) | } ) \\ \leq & \ , k \big [ \sqrt { | e ( \psi(t ) ) - e ( u_{c _ * } ) |
+ | p ( \psi^\ini ) - p(u_{c _ * } ) | } + \sqrt{| p ( \psi^\ini ) - p(u_{c _ * } ) | } \big ] , \end{aligned}\ ] ] which yields .
+ the above estimate is optimal when @xmath677 since @xmath172 is a critical point of the action @xmath678 .
this bound shows that , in the definition of stability , one has to take @xmath679 in general .
the estimate @xmath663 shows that one can actually take @xmath212 .
+ * proof of estimate @xmath680 . *
the point is to compare @xmath681 to @xmath682 with @xmath683 such that @xmath684 instead of comparing to @xmath172 . in other words , we replace @xmath685 by @xmath682 .
note first that since @xmath686 , there exists , by the implicit function theorem , such a @xmath683 .
we then proceed as follows .
let @xmath687 be close to @xmath172 .
then , there exists @xmath688 such that @xmath689 .
moreover , since @xmath690 , it follows [ crapule ] || _ c - _ c _ * || k | c - c _ * | k | p ( u_c ) - p(u_c _ * ) | = k |
p(^ ) - p(u_c _ * ) | k || _ hy^- _ c _ * || k d_hy ( ^ , u_c _ * ) . from , it comes @xmath691 as soon as @xmath692 .
the fact that the constant @xmath234 can be taken uniform with respect to @xmath51 for @xmath51 close to @xmath210 comes directly from the proof in @xcite .
therefore , for @xmath672 , @xmath693 \\ \leq & \ , \sqrt{k } \sqrt{e ( \psi ( t ) ) - e ( u_{c } ) } + \bo ( | p ( \psi^\ini ) - p(u_{c _ * } ) | ) .\end{aligned}\ ] ] using that @xmath694 and that @xmath695 is a critical point of the action @xmath696 , we infer @xmath697 ( \psi_{\rm hy}^\ini ) - [ e_{\rm hy } - c p_{\rm hy } ] ( \bu_{c } ) = \bo ( |\!| \psi_{\rm hy}^\ini - \bu_{c } |\!|^2 ) $ ]
. consequently , @xmath698 by .
this gives .
in this section , we set @xmath699 and we assume @xmath700 the approach is reminiscent to the proof of m. maeda @xcite .
several modifications are necessary since for the skew - adjoint operator @xmath701 , we can not find the required hilbert space @xmath316 .
more degenerate cases can probably be considered as in @xcite .
we shall denote @xmath702 and @xmath703 the riesz isomorphisms and @xmath704 and @xmath705 , endowed with its canonical scalar product .
they are the corresponding hilbert spaces needed in @xcite .
we consider the symmetric matrix @xmath706 which is such that @xmath707 and @xmath708 .
our assumption @xmath709 will simplify a little the computations in @xcite .
the functions @xmath710 and @xmath711 used in @xcite become now @xmath712 and @xmath713 in order to clarify the dualities used in @xcite , we provide some elements of the proof adapted to our context .
[ maeda1 ] there exists @xmath714 small and @xmath715 with @xmath716 and such that , for any @xmath717 , @xmath718 _ proof .
_ we have @xmath719 since @xmath720 , the conclusion follows from an easy implicit function argument near @xmath721 . in @xcite ,
the linear mapping @xmath582 is seen from @xmath595 to @xmath722 , but here , there is no confusion to define @xmath723 .
+ we define , for @xmath717 , @xmath724 which then verifies by construction @xmath725 .
[ maeda2 ] as @xmath726 , there holds @xmath727 .
_ using that @xmath728 , @xmath729 and @xmath730 , we have by taylor expansion @xmath731 as wished .
+ we recall that we have defined the tubular neighbourhood @xmath732 .
[ maeda3 ] for @xmath733 small enough , there exist four @xmath734 mappings @xmath735 , @xmath736 , @xmath737 and @xmath738 , satisfying , for @xmath739 , @xmath740 and the orthogonality relations @xmath741_{|c = c _ * + \bar{\gamma } ( \bu ) } ) _ h = ( \vartheta ( \bu ) , \mathbb{b } \bu_{c _ * + \bar{\gamma } ( \bu ) } ) _ h = 0 .\ ] ] finally , @xmath742 and @xmath743 .
we consider the mapping @xmath744 defined by @xmath745_{|c = c _ * + \gamma } ) _ h \\
( \bu ( \cdot -y ) - \bw(\gamma ) - \alpha \mathbb{b } \bu_{c _ * + \gamma } , \mathbb{b } \bu_{c _ * + \gamma } ) _ h \end{array}\right ) .\ ] ] then @xmath746 since @xmath747 . in order to show that @xmath748 is of class @xmath364
, we have to pay attention to the translation term @xmath749 , since differentiation in @xmath200 requires @xmath750 whereas we only have @xmath751 .
it thus suffices to write @xmath752_{|c = c _ * + \gamma } ( \cdot + y ) ) _ h - ( \bw(\gamma ) + \alpha \mathbb{b } \bu_{c _ * + \gamma } , [ \p_c \bu_c]_{|c = c _ * + \gamma } ) _ h \\ ( \bu , \mathbb{b } \bu_{c _ * + \gamma } ( \cdot + y ) ) _ h - ( \bw(\gamma ) + \alpha \mathbb{b } \bu_{c _ * + \gamma } , \mathbb{b } \bu_{c _ * + \gamma } ) _ h \end{array}\right)\ ] ] to see that @xmath748 is indeed of class @xmath734 on @xmath753 since @xmath754 is smooth . moreover , using that @xmath755_{|c = c _ * } $ ] , we infer @xmath756_{|c = c _ * } , \p_x \bu _ * ) _ h & - ( \mathbb{b } \bu _ * , \p_x \bu _ * ) _ h \\ - ( \bu _ * , \p_x [ \p_c \bu_c]_{|c = c _ * } ) _ h & - |\!| [ \p_c \bu_c]_{|c = c _ * } |\!|^2_h & - ( \mathbb{b } \bu _ * , [ \p_c \bu_c]_{|c = c _ * } ) _ h \\ - ( \bu _ * , \mathbb{b } \p_x \bu _ * ) _ h & - ( [ \p_c \bu_c]_{|c = c _ * } , \mathbb{b } \bu _ * ) _ h & - |\!| \mathbb{b } \bu _ * |\!|_h^2 \end{array}\right ) .\ ] ] here , @xcite uses assumption 2 @xmath469 , that reads for us @xmath757_{|c = c _ * } , \p_x \bu _ * ) _ h = - ( \p_x [ \p_c \bu_c]_{|c = c _ * } , \bu _ * ) _ h = 0 $ ] .
it is indeed the case since @xmath758 is chosen even for any @xmath51 ( close to @xmath210 ) .
furthermore , @xmath759 by integration by parts , @xmath760 since @xmath761 is skew - adjoint , and @xmath762_{|c = c _ * } ) _ h = \p_c [ p_{\rm hy } ( \bu_c ) ] _ { |c = c _ * } = 0 $ ] by hypothesis .
therefore , @xmath763_{|c = c _ * } |\!|^2_h & 0 \\ 0 & 0 & - |\!| \mathbb{b } \bu _ * |\!|_h^2 \end{array}\right)\ ] ] is invertible , thus the implicit function theorem provides three real - valued functions @xmath764 , @xmath765 and @xmath766 , defined near @xmath767 ( in @xmath595 ) and with @xmath768 , such that @xmath769 .
these functions are extended to @xmath770 ( for @xmath771 small enough ) by the formulas @xmath772 , @xmath773 and @xmath774 for any @xmath775 such that @xmath776 lies in the neighbourhood of @xmath767 where @xmath764 , @xmath765 and @xmath766 are defined .
consequently , the mapping @xmath777 is orthogonal in @xmath778 to @xmath779 , @xmath780_{|c = c _ * + \bar{\gamma } ( \bu ) } $ ] and @xmath781 , as desired . since @xmath782 is assumed of class @xmath319 , we have @xmath783 and the regularities @xmath742 and @xmath743 follow easily .
+ we would like to point out that in @xcite ( lemma 3 ) , it is claimed that `` @xmath784 '' is orthogonal to `` @xmath785 '' ( we refer to the notations in @xcite ) .
however , since `` @xmath786 '' is already orthogonal to `` @xmath785 '' by construction , this is equivalent to `` @xmath787 '' , or `` @xmath788 = 0 $ ] '' at `` @xmath789 '' .
we have not understood why this should happen since in general , for the function @xmath790 , the point @xmath791 is the only local critical point .
for this reason , we have added a component to the original mapping @xmath748 in @xcite .
let us observe that then , lemma 3 in @xcite uses the assumption
@xmath792 . on the other hand ,
the derivative of @xmath748 in @xcite assumes `` @xmath793 '' , for otherwise the expression `` @xmath794 '' , for instance , is meaningless .
we have therefore given some details showing clearly the smoothness of @xmath748 .
we now prove a lemma which shows that the quadratic functional associated with @xmath795 gives a good control on @xmath796 thanks to the orthogonality conditions on this function .
this result is in the spirit of lemma 7 in @xcite .
[ maeda4 ] there exist @xmath797 and @xmath798 such that if @xmath799 and if @xmath800 verifies @xmath801_{|c = c _ * + \gamma } ) _ h = ( \vartheta , \mathbb{b } \bu_{c _ * + \gamma } ) _ h = 0 , \ ] ] then @xmath802 . _
_ as a first step , we prove that if @xmath800 verifies @xmath803 , @xmath804_{|c = c _ * } ) _ h = ( \vartheta , \mathbb{b } \bu _ { * } ) _ h = 0 , \ ] ] then @xmath805 .
indeed , assume that @xmath806 .
let @xmath807 be a negative eigenvector of @xmath808 .
we claim that can not have @xmath809_{|c = c _ * } , \chi ) _ h ) = ( 0,0 ) $ ] . for
otherwise , @xmath810 implies that @xmath811 is @xmath505-orthogonal to @xmath812 , which is the eigenvector associated with the only negative eigenvalue @xmath813 of @xmath795 seen as an unbounded operator on @xmath505 , thus @xmath814 , and since we assume equality , this means that @xmath811 belongs to the kernel of @xmath815 , which is spanned by @xmath816 , but the condition @xmath817 then implies @xmath818 , a contradiction .
therefore , there exists @xmath819 such that @xmath820 and @xmath821_{|c = c _ * } + b \vartheta , \chi ) _ h = 0 $ ] .
the nonzero vector @xmath822_{|c = c _ * } + b \vartheta $ ] then verifies @xmath823 and @xmath824_{|c = c _ * } , j \bu _ * ) _ h + b ( \vartheta , j \bu _ * ) _ h = 0 $ ] , so that @xmath825 . here
, we have used once again that @xmath757_{|c = c _ * } , j \bu _ * ) _ h = 0 $ ] since the left vector is an even function and the right vector an odd function . however , in view of the equality @xmath826_{|c = c _ * } , \phi \rangle_{x^*,x } = ( \mathbb{b } \bu _ * , \phi ) _ h $ ] , valid for any @xmath827 ( which follows from differentiation of @xmath828 at @xmath829 ) ) , we have @xmath830_{|c = c _ * } , \vartheta \rangle_{x^*,x } = ( \mathbb{b } [ \p_c \bu_c]_{|c = c _ * } , \vartheta ) _ h = 0 .\ ] ] as a consequence , @xmath831_{|c = c _ * } , [ \p_c \bu_c]_{|c = c _ * } \rangle_{x^*,x } + b^2 \langle \bf _ * '' ( \bu _ * ) \vartheta , \vartheta \rangle_{x^*,x } \\
= & \ , a^2 ( \mathbb{b } \bu _ * , [ \p_c \bu_c]_{|c = c _ * } ) _ h + b^2 \langle \bf _ * '' ( \bu _ * ) \vartheta , \vartheta \rangle_{x^*,x } = b^2 \langle \bf _ * '' ( \bu _ * ) \vartheta , \vartheta \rangle_{x^*,x } , \end{aligned}\ ] ] since @xmath762_{|c = c _ * } ) _ h = \p_c [ p_{\rm hy } ( \bu_c ) ] _ { |c = c _ * } = 0 $ ] in our situation .
we reach a contradiction since the right - hand side is supposed @xmath832 .
we now prove the lemma by contradiction , and then assume that there exists sequences @xmath833 and @xmath834 such that @xmath835 , @xmath836 and [ orthogo ] ( _ n , _ x _ c _ * + _ n ) _ h = ( _ n , [ _ c _ c]_|c = c _ * + _ n ) _ h = ( _ n ,
_ c _ * + _ n ) _ h = 0 , but @xmath837 .
possibly passing to a subsequence , we may assume the existence of some @xmath838 such that @xmath839 in @xmath840 .
we then show the lower semicontinuity of @xmath841 .
this is roughly a verification of part of assumption ( a3 ) in @xcite , used in lemma 7 there . by compact sobolev embedding
, we may assume @xmath842 in @xmath843 .
a straightforward computation gives @xmath844 since @xmath845 remains bounded away from zero uniformly and @xmath846 in @xmath847 as @xmath224 , the weak convergence @xmath848 in @xmath375 implies @xmath849 for the remaining terms , we write , for some @xmath850 to be determined later , @xmath851 ^ 2 \ dx \\ & \quad + \int_{|x| \leq r } + \int_{|x| \geq
r } \frac12 \big ( - \frac { ( 2 u_{c _ * + \gamma_n } - ( c _ * + \gamma_n ) ) ^2 } { r_0 ^ 2 + \eta_{c _ * + \gamma_n } } - 2 f ' ( r_0 ^ 2 + \eta_{c _ * + \gamma_n } ) \big ) \zeta_n ^2 \ dx .\end{aligned}\ ] ] for the first integral , we may use that @xmath852 in @xmath853 and the fact that @xmath854 converges to @xmath855 uniformly to deduce [ cachalot ] ( r_0 ^ 2 + _ c _ * + _ n ) ^1/2 _ n + ( r_0 ^ 2 + _ * ) ^1/2 + l^2 , hence , @xmath856 ^ 2 \ dx \nonumber \\ & \leq \varliminf_{n \to + \ii } \int_\r 2 \big [ ( r_0 ^ 2 + \eta_{c _ * + \gamma_n } ) ^{1/2 } \upsilon_n + \frac { ( 2 u_{c _ * + \gamma_n}-(c _ * + \gamma_n ) ) \zeta_n } { 2 ( r_0 ^ 2 + \eta_{c _ * + \gamma_n } ) ^{1/2 } } \big]^2 \ dx .\end{aligned}\ ] ] since @xmath842 in @xmath857 ) $ ] and @xmath858 uniformly , it follows that @xmath859 for the last integral , we have to use the decay at infinity of @xmath860 and @xmath861 uniformly for @xmath862 small .
this gives @xmath863 as @xmath864 , uniformly in @xmath641 .
since @xmath171 , there exist some small @xmath235 and some @xmath865 large such that , for any @xmath641 and any @xmath866 with @xmath867 , @xmath868 in particular , since @xmath848 in @xmath520 , @xmath869 in @xmath520 , thus @xmath870 combining these three @xmath871 inequalities , we deduce [ semi ] _ * ( _ * ) , _
n + _ c _ * + _ n ( _ c _ * + _ n ) _ n , _
n _ x^*,x = 0 . turning back to our sequence @xmath872
, we may pass to the limit in : @xmath873_{|c = c _ * } ) _ h = ( \vartheta , \mathbb{b } \bu _ * ) _ h = 0 .\ ] ] comparing with , we deduce from our first claim that @xmath818 .
this means that we must have equality in all the above @xmath874 inequalities .
in particular , the weak convergence is actually strong , thus @xmath875 in @xmath876 ( the strong convergence in @xmath877 being already known since @xmath842 in @xmath878 ) .
going back to the equality in thus provides @xmath879 in @xmath880 , since @xmath881 remains uniformly bounded away from zero and by weak convergence , @xmath882 .
finally , the equality in means that is actually a strong convergence , that is @xmath883 in @xmath520 since @xmath842 in @xmath520 .
the contradiction then follows : @xmath884 .
this result is also lemma 7 in @xcite , and is said to be lemma 7 in @xcite .
however , the hypothesis of lemma 7 in @xcite are not verified , and in particular assumption ( b3 ) there .
it is natural to believe that this assumption is verified is most physical situations , but it is not clear whether it always holds true in the general framework of @xcite without further hypothesis .
the next lemma provides a control for @xmath885 .
[ maeda5 ] assume @xmath733 small enough .
then , there exists @xmath600 such that for any @xmath739 satisfying @xmath886 , there holds @xmath887 _ proof .
_ it is the same as in @xcite , lemma 8 , but give it for completeness .
we expand and use that @xmath888 and the definition @xmath889 for the second line : @xmath890 since @xmath891 , we infer @xmath892 = \sigma(\bar{\gamma}(\bu ) ) ( \bu_{c_*+\bar{\gamma } ( \bu ) } , \vartheta ( \bu ) ) _ h + p_{\rm hy } ( \vartheta ( \bu ) ) \ ] ] and the conclusion follows since @xmath893 by lemma [ maeda1 ] . + now , we give a lemma useful to estimate @xmath894 .
[ maeda6 ] assume @xmath733 small enough .
then , there exists @xmath600 such that for any @xmath739 satisfying @xmath886 and @xmath895 , there holds @xmath896 in particular , @xmath897 . _ proof .
_ it is the same as in @xcite , lemma 9 .
note first that the last assertion is a direct consequence of the first one and lemma [ maeda5 ] .
next , we argue by contradiction and assume that there exists a sequence @xmath898 in @xmath595 such that @xmath899 and @xmath900 . for simplicity ,
we denote @xmath901 , @xmath902 , @xmath903 .
then , by lemma [ maeda5 ] , we have @xmath904 .
therefore , by taylor expansion and lemma [ maeda3 ] , it holds @xmath905 however , by lemma [ maeda2 ] , @xmath906 , and since @xmath907 , there holds @xmath908 . furthermore , using @xmath909 , the third orthogonality condition in lemma [ maeda3 ] and that @xmath910 , we deduce @xmath911 for the last line , we have used another taylor expansion with @xmath912
finally , lemma [ maeda4 ] yields @xmath913 .
reporting these expansions in yields @xmath914 for @xmath641 sufficiently large , which contradicts our assumption .
+ we now need to find an extension of the functionals `` @xmath915 '' and `` @xmath90 '' used in @xcite ( and also in @xcite ) . in these works
, these functionals are built on what should be here `` @xmath916 '' , but unfortunately , @xmath917 does not have vanishing integral over @xmath63 ( for instance , @xmath918 has constant sign ) .
we rely instead on a construction of a suitable approximation of `` @xmath919 '' .
a similar construction is used in @xcite .
[ maeda7 ] for any @xmath920 , there exists a @xmath14 mapping @xmath921 such that , for any @xmath922 , @xmath923 is an odd function verifying @xmath924_{|c = c_*+ \gamma } \n_x \leq \kappa .\ ] ] _ proof .
_ we fix an even function @xmath925 such that @xmath926 . for @xmath927 to be fixed later , but independent of @xmath928 and @xmath391 , we set @xmath929 and @xmath930_{|c = c_*+ \gamma } ( y ) - \frac{1}{t_\kappa } \theta_0 \big ( \frac{y}{t_\kappa } \big ) \int_\r [ \p_c \bu_c]_{|c = c_*+ \gamma } ( z ) \
dz \big ] \ dy .\ ] ] it is clear that @xmath931 and that , since @xmath932 and @xmath933 , @xmath934_{|c = c_*+ \gamma } = \frac{1}{t_\kappa } \theta_0 \big ( \frac{\cdot}{t_\kappa } \big ) \int_\r [ \p_c \bu_c]_{|c = c_*+ \gamma } ( z ) \
dz\ ] ] in particular , @xmath935_{|c = c_*+ \gamma } \n_x^2 = \big [ \frac{1}{t_\kappa } \n \theta_0 \n_{l^2}^2 + \frac{1}{t_\kappa^3 } \n \p_x \theta_0 \n_{l^2}^2 \big ] \big ( \int_\r [ \p_c \bu_c]_{|c = c_*+ \gamma } ( z ) \ dz \big)^2 \leq \kappa^2\ ] ] if we choose @xmath936 sufficiently large and @xmath937 smaller if necessary .
moreover , @xmath938 is odd since @xmath758 and @xmath939 are even .
in addition , the even function @xmath940_{|c = c_*+ \gamma } ( y ) - \frac{1}{t_\kappa } \theta_0 \big ( \frac{y}{t_\kappa } \big ) \int_\r [ \p_c \bu_c]_{|c = c_*+ \gamma } ( z ) \ dz $ ] decays exponentially at infinity ( since @xmath939 has compact support and @xmath941 decays exponentially ) , and has zero integral ( since @xmath939 has integral equal to one ) , hence @xmath942_{|c
= c_*+ \gamma } ( y ) - \frac{1}{t_\kappa } \theta_0 \big ( \frac{y}{t_\kappa } \big ) \int_\r [ \p_c \bu_c]_{|c = c_*+ \gamma } ( z ) \ dz \big ] \ dy\ ] ] and decays exponentially at infinity .
it follows easily from these two equalities that @xmath943 is well defined and continuous , hence also @xmath944 . by the same type of arguments , @xmath945_{|c = c_*+ \gamma } ( y ) - \frac{1}{t_\kappa } \theta_0 \big ( \frac{y}{t_\kappa } \big ) \int_\r [ \p^2_c \bu_c]_{|c = c_*+ \gamma } ( z ) \
dz \big ] \ dy\ ] ] is well - defined and is a continuous function of @xmath928 with values into @xmath946 , and similarly for the second derivative .
+ we now define , in the tubular neighbourhood @xmath770 of @xmath947 , the functional ( corresponding to `` @xmath915 '' in @xcite ) @xmath948 depending on @xmath949 , which will be determined later . the first properties of @xmath950 are given below .
[ maeda8 ] for any @xmath920 , @xmath951 is of class @xmath364 . in addition
, there exists some bounded mapping @xmath952 such that if @xmath953 is a solution to that remains in @xmath770 , then @xmath954 where @xmath955 is defined by @xmath956 _ proof .
_ the fact that @xmath950 is of class @xmath364 follows directly from the second expression and the fact that @xmath957 and @xmath735 are @xmath364 ( in @xcite formula ( 3.11 ) , the same remark as for the smoothness of @xmath748 after lemma [ maeda3 ] holds , since it requires `` @xmath793 '' ) .
if @xmath958 is a solution to that remains in @xmath770 , we therefore have , denoting @xmath959 and @xmath960 , @xmath961 we now observe that the invariance of @xmath950 by translation provides by differentiation the equality , for @xmath739 , @xmath962 in particular , the second term in vanishes .
in addition , since @xmath963 a solution to that remains in @xmath770 , there holds , denoting by @xmath964 the variational derivative , @xmath965 by .
in addition , since @xmath966 and @xmath967 , @xmath968 the second term vanishes since @xmath735 is invariant by translation ( by definition , see the proof of lemma [ maeda3 ] ) . as a consequence , @xmath969 the first term is simply @xmath970 .
we then define @xmath952 by @xmath971 ( see the regularity shown for @xmath972 in lemma [ maeda3 ] ) , so that integration by parts yields @xmath973 inserting these relations into then gives @xmath974 which is the desired equality .
if @xmath975 is just a continuous in time solution to that remains in @xmath770 , then the integrated relation @xmath976 holds , as can be seen by using the continuity of the flow and the approximation of such a solution by smoother ones ( see @xcite ) .
+ we now compute the asymptotics of @xmath977 for @xmath726 and small @xmath395 . [ maeda9 ] we have @xmath978 _ proof . _
the proof follows the one of lemma 5 in @xcite . as a first step ,
notice that @xmath979 , @xmath980 , as it can be seen from the equality @xmath981 and the local uniqueness of the solution to @xmath982 .
therefore , since @xmath983 and @xmath984 , @xmath985 + o _ { \gamma \to 0 } ( \gamma^2 ) \\ = & \ , - \frac{\gamma^2 \ddot{p } _ * } { 2 |\!| \bu _ * |\!|^2_h } \bf '' _ * ( \bu _ * ) [ \mathbb{b } \bu _ * ] + o _ { \gamma \to 0 } ( \gamma^2 ) .\end{aligned}\ ] ] in addition , since @xmath986 is even and @xmath987 is odd , we deduce @xmath988 consequently , @xmath989 , j \upsilon_\kappa ( \gamma ) \rangle_{x^*,x } + o _ { \gamma \to 0 } ( \gamma^2 ) , \ ] ] where `` @xmath990 '' does not depend on @xmath391 .
moreover , lemma [ maeda7 ] provides @xmath991_{|c = c_*+ \gamma } |\!|_x \leq \kappa $ ] independently of @xmath799 , hence @xmath992 , [ \p_c \bu_c]_{|c = c_*+ \gamma } \rangle_{x^*,x } + o _ { ( \gamma , \kappa ) \to ( 0,0 ) } ( \gamma^2 ) \\
= & \ , \frac{\gamma^2 \ddot{p } _ * } { 2 |\!| \bu _ * |\!|^2_h } \langle \bf '' _ * ( \bu _ * ) [ \mathbb{b } \bu _ * ] , [ \p_c \bu_c]_{|c = c _ * } \rangle_{x^*,x } + o _ { ( \gamma , \kappa ) \to ( 0,0 ) } ( \gamma^2 ) .\end{aligned}\ ] ] finally , using once again the equality ( for @xmath827 ) @xmath993_{|c = c _ * } , \phi \rangle_{x^*,x } = ( \mathbb{b } \bu _ * , \phi ) _ h $ ] and that @xmath795 is self - adjoint , we infer @xmath994 , [ \p_c \bu_c]_{|c = c _ * } \rangle_{x^*,x } = \langle \bf '' _ * ( \bu _ * ) ( [ \p_c \bu_c]_{|c = c _ * } ) , \mathbb{b } \bu _ * \rangle_{x^*,x } = \n \mathbb{b } \bu _ * \n^2_h = \n \bu _ * \n^2_h , \ ] ] and reporting this into the previous expression gives the result .
+ we now compute the asymptotics of @xmath995 for more general functions . [ maeda10 ] let @xmath996 be small enough .
if @xmath739 satisfies @xmath886 and @xmath997 , then , we have @xmath998 uniformly for @xmath999 . _ proof . _
first , we may apply lemma [ maeda6 ] and infer that @xmath1000 .
then , we write @xmath1001 and , recalling the expression @xmath1002 we wish to make a taylor expansion . first , note that @xmath1003 + \bo ( | \bar{\gamma}(\bu ) |^3 ) \\
= & \ , \bf_{c_*+ \bar{\gamma}(\bu ) } ' ( \bw ( \bar{\gamma } ( \bu ) ) ) + \bf_{c_*+ \bar{\gamma}(\bu ) } '' ( \bu_{c_*+ \bar{\gamma } ( \bu ) } ) [ \vartheta ( \bu ) ] + \bo ( | \bar{\gamma}(\bu ) |^3 ) \end{aligned}\ ] ] hence , since @xmath1004 and ( lemma [ maeda1 ] ) @xmath1005 , we have @xmath1006 , thus @xmath1007 , \big\ { j \upsilon_\kappa ( \bar{\gamma}(\bu ) ) ( \cdot + \bar{y}(\bu ) ) + ( \bu , \p_\gamma \upsilon_\kappa ( \bar{\gamma}(\bu ) ) ( \cdot + \bar{y}(\bu ) ) ) _ h \bn_{\bar{\gamma } } ( \bu ) \big\ } \big\rangle_{x^*,x } .\end{aligned}\ ] ] now , in the bracket term , we may replace @xmath1008 by @xmath1009 ( since @xmath1010 ) . by the computations of lemma [ maeda9 ] and the equalities @xmath979 , @xmath980 , this gives @xmath1011 , j \upsilon_\kappa ( \bar{\gamma}(\bu ) ) \big\rangle_{x^*,x } \\ = & \ , \bo ( | \bar{\gamma}(\bu ) |^3 ) - \big\langle \bf_{c_*+ \bar{\gamma}(\bu ) } '' ( \bu_{c_*+ \bar{\gamma}(\bu ) } ) [ \vartheta ( \bu ) ] , \ { \p_c \bu_c\}_{| c= c_*+ \bar{\gamma}(\bu ) } \big\rangle_{x^*,x } + \bo(\kappa ) \\
= & \ , \bo ( | \bar{\gamma}(\bu ) |^3 ) - \big\langle \bf_{c_*+ \bar{\gamma}(\bu ) } '' ( \bu_{c_*+ \bar{\gamma}(\bu ) } ) [ \ { \p_c \bu_c\}_{| c= c_*+ \bar{\gamma}(\bu ) } ] , \vartheta ( \bu ) \big\rangle_{x^*,x } + \bo(\kappa ) \end{aligned}\ ] ] using lemma [ maeda7 ] and the self - adjointness of @xmath1012 . choosing @xmath1013 and from the equality ( for @xmath827 ) @xmath1014 , \phi \rangle_{x^*,x } = ( \mathbb{b } \bu_c , \phi ) _ h $ ] , we infer @xmath1015 by the orthogonality condition in lemma [ maeda4 ] . inserting the expansion of @xmath1016 given in lemma [ maeda9 ]
yields the conclusion . + * proof of theorem [ stabcusp ] .
* we have to show that there exists @xmath733 such that , for any @xmath160 , we can choose an initial datum at distance @xmath1017 from @xmath947 but that escape from @xmath770 . since @xmath1018 in @xmath1019
, we shall take the initial datum to be @xmath1020 for some small @xmath10 , and denote @xmath1021 the corresponding solution . in view of lemma [ maeda2 ] , we have @xmath1022 , hence we can choose @xmath1023 with the sign of @xmath1024 so that @xmath1025 we now assume that @xmath1021 is globally defined and remains in @xmath770 , where @xmath666 is as in lemma [ maeda6 ] . by conservation of energy and momentum and the construction of @xmath1026 , we deduce @xmath1027 , and @xmath1028 .
the first step is to have a control on @xmath1029 .
we denote @xmath1030 , @xmath1031 and @xmath1032 . applying lemma [ maeda6 ]
, we obtain @xmath1033 .
in addition , lemma [ maeda2 ] and taylor expansion gives @xmath1034 where we have used that @xmath1035 ( for the terms involving @xmath1036 ) and lemma [ maeda2 ] .
furthermore , by the orthogonality relations in lemma [ maeda3 ] and using that @xmath1037 and @xmath1038 , it holds @xmath1039 in addition , by lemma [ maeda4 ] , the before last term is @xmath1040 . as a consequence , by conservation of @xmath1041 , we infer , for small @xmath1023 , @xmath1042 in particular , this forces @xmath1043 to always be of the sign of @xmath1044 and to satisfy @xmath1045 ( provided @xmath666 and @xmath1023 are small enough ) .
since now , we have a good upper bound for @xmath1046 , we can choose @xmath1047 , which is such that , for any @xmath672 , @xmath1048
. in particular , we can apply lemma [ maeda10 ] and get @xmath1049 with this choice @xmath1050 , we deduce from lemma [ maeda8 ] that @xmath1051 since @xmath1045 , it follows that , when @xmath1052 ( the case @xmath1053 is analogous ) , @xmath1054 hence @xmath1055 is unbounded as @xmath1056 goes to @xmath1057
. however , by definition of @xmath1058 , we have by cauchy - schwarz @xmath1059 for @xmath739 .
we have reached a contradiction .
the proof of theorem [ stabcusp ] is complete .
* existence of at least one unstable eigenvalue . *
the proof of the existence of at least one unstable eigenvalue relies on evans function technique , as in @xcite , @xcite .
we shall actually use theorem 1 in @xcite when observing ( see _ e.g. _ @xcite ) that the euler - korteweg system \ { ll = 0 + + = 0 , .
where @xmath1061 is the ( smooth enough ) capillarity , reduces to ( where , we recall , @xmath1062 , @xmath1063 and @xmath1064 ) , namely @xmath1065 for the capillarity @xmath1066 , as can be shown by straightforward computations .
the associated eigenvalue problem in the moving frame is [ rouge ] \ { ll = 0 + + = 0 . .
the link with the original eigenvalue problem is done through the formula [ saumon ] w = u _ * ( + i _
-^x ) , since this corresponds to @xmath1067 .
notice indeed that the second equation in gives @xmath1068 .
it then follows from theorem 1 in @xcite that under the assumption @xmath1069 , there exists at least one unstable eigenvalue @xmath1070 .
* existence of at most one unstable eigenvalue .
* the fact that there exists at most one unstable eigenvalue follows from arguments as in @xcite ( appendix b ) and is a direct consequence of theorem 3.1 in @xcite , that we recall now .
[ @xcite ] [ atmost ] let @xmath1071 and @xmath1072 be two two operators on a real hilbert space @xmath595 , with @xmath1072 self - adjoint and @xmath1071 skew - symmetric .
then , the number of eigenvalues , counting algebraic multiplicities , of @xmath1073_\c $ ] in the right - half plane @xmath1074 is less than or equal to the number of negative eigenvalues of @xmath1075 , counting multiplicities . in order to apply this result to our problem ,
let us write the eigenvalue problem under the form @xmath1076 where @xmath1077 is the self - adjoint sturm - liouville operator @xmath1078 ( which is bounded from below ) on @xmath1079 and with @xmath1080 we are in the setting of theorem [ atmost ] with @xmath1081 skew - symmetric and @xmath1082 self - adjoint .
we thus show that @xmath1082 has at most one negative eigenvalue .
since @xmath1083 remains bounded away from zero , it is clear that , for @xmath1084 and @xmath1085 given , @xmath1086 if and only if [ dupropre ] ^- = , ^- , since we may express @xmath1087 in terms of @xmath1088 with the second equation .
we observe that the translation invariance shows that @xmath1089 belongs to the kernel of @xmath1075 , that is , using once again the relation @xmath1090 , @xmath1091 .
furthermore , @xmath1092 has the same continuous spectrum as its constant coefficient limit as @xmath1093 , namely @xmath1094 that is @xmath1095 , since @xmath1096 .
since @xmath1097 has exactly one zero ( at @xmath1098 ) , it follows from standard sturm - liouville theory that @xmath1092 has precisely one negative eigenvalue @xmath1099 and that the second eigenvalue is @xmath75 . taking the scalar product with yields @xmath1100 } \
dx = \s |\!| \zeta |\!|_{l^2}^2 .\ ] ] now , for @xmath1101 , we consider the self - adjoint operator @xmath1102 clearly , @xmath1103 , @xmath1104 , and @xmath1105 is decreasing .
let us assume now that the self - adjoint operator @xmath1082 has at least two negative eigenvalues .
then , we denote @xmath1106 the two smallest eigenvalues of @xmath1082 ( necessarily simple ) , and @xmath1107 , @xmath1108 two associated eigenvectors . since @xmath1109 is self - adjoint , @xmath1110 . furthermore , @xmath1111 , hence , by monotonicity , @xmath1112 for any @xmath1113 .
therefore , @xmath1114 has at least one negative eigenvalue for @xmath1113 .
we denote @xmath1115 the smallest eigenvalue of @xmath1116 .
then , @xmath1117 and @xmath1118 decreases in @xmath1119 $ ] .
moreover , we may choose a positive eigenvector @xmath1120 for the eigenvalue @xmath1121 , with @xmath1122 .
since @xmath1123 , it follows from the monotonicity that for any @xmath1124 , we have @xmath1125 . when @xmath1126 , we then have @xmath1127 , and thus @xmath1128 , which implies that we may choose @xmath1129 without loss of generality .
similarly , if @xmath1130 , we see that we may choose @xmath1129 .
we obtain a contradiction since then @xmath1131 and thus @xmath1107 and @xmath1108 can not be orthogonal in @xmath505 .
we have thus shown that @xmath1082 has at most one negative eigenvalue , and then theorem [ atmost ] shows that @xmath1132 has at most one eigenvalue in @xmath1074 , as wished . in this section , we drop the `` @xmath1133 '' for the travelling wave we are considering .
when linearizing the ( nls ) equation in the moving frame with speed @xmath51 , we obtain [ eqlin ] i - i c _ x + ^2_x + f(|u|^2 ) + 2 , u f(|u|^2 ) u = 0 , or @xmath1134 we wish to show that this linear equation can be solved using a continuous semigroup . in order to handle later the nonlinear terms , we work in @xmath1135 instead of @xmath1136 . therefore , we consider the unbounded operator @xmath1137 on @xmath1138 defined by @xmath1139 it follows easily that for @xmath1140 , @xmath1141 \psi_1 , \psi_2 \rangle_{h^1(\r , \c ) } - \langle [ f(|u|^2 ) + 2 f'(|u|^2 ) u_{1}^2 ] \psi_2 , \psi_1 \rangle_{h^1(\r , \c ) } \big ) \\ \leq & \ , k |\!|
\psi |\!|^2_{h^1(\r , \c^2 ) } .\end{aligned}\ ] ] moreover , the spectrum of @xmath1142 is included in the half - space @xmath1143 , hence @xmath1142 generates a continuous semigroup @xmath1144 on @xmath1145 . in order to estimate the growth of the semigroup @xmath1144 on @xmath1145 ,
we could try to use the same approach as @xcite , which relies on the proof of the spectral mapping theorem in @xcite .
however , our situation is slightly different since in these studies , the reference solutions is real - valued ( it is a bound state in @xcite and the kink in @xcite ) . therefore , @xmath1146 and @xmath1142 has no diagonal term , and the system is much more decoupled than in our situation . as a matter of fact , it is not very clear whether the arguments of @xcite carry over to our problem .
we thus have chosen to use the general approach given in appendix b. we thus verify the assumptions of theorem b.4 ( see also corollary b.2 ) there , which are easy : @xmath1147 generates a semigroup in @xmath1135 and the spectrum of @xmath1147 is of the form @xmath1148 , where @xmath1149 is the essential spectrum and @xmath1150 two simple eigenvalues . moreover ,
the eigenvector associated with @xmath1151 belongs to @xmath1152 .
therefore , theorem b.4 in appendix b applies and the growth estimate for the linearized problem follows . for the nonlinear instability result , we argue as for corollary b.2 in appendix b , since the manifold @xmath1153 is transverse to the curve @xmath1154 in @xmath1155 .
indeed , it follows from that @xmath1156 .
assume that @xmath1157 , with @xmath1158 .
then , integration of the first equation of provides @xmath1159 hence , using that @xmath1160 and the equality @xmath1161 , we infer @xmath1162 since @xmath1163 and @xmath1164 has constant sign in @xmath191 , integrating over @xmath191 then implies @xmath1165 , which in turn yields @xmath1166 and @xmath1167 , a contradiction .
consequently , @xmath1168 and the manifold @xmath1153 is indeed transverse to the curve @xmath1154 in @xmath1155 .
let us recall that the momentum @xmath1170 , for @xmath1171 , has the expression @xmath1172 since @xmath1173 .
therefore , we decompose @xmath1174 with two integrals : [ split ] p(u_c ) = c _ _
c^0 + c _ _ c^0 ( - ) d . using the change of variables
@xmath343 , the second integral in is equal to @xmath1175 the passage to the limit @xmath369 being justified by the dominated convergence theorem since the absolute value of the integrand is @xmath1176 for @xmath1177 for small @xmath51 and for @xmath1178 , since @xmath1179 , @xmath1180 and hence is equal to @xmath1181 } \big| \\ & \leq k \frac { ( 1-t)^2 } { ( 1-t ) \sqrt{1-t } \sqrt{1-t } \sqrt{1-t } } = \frac{k}{\sqrt{1-t } } \in l^1 ( ( 1/2,1 ) ) .\end{aligned}\ ] ] furthermore , letting @xmath1182 , @xmath1183 , the first integral in is equal to @xmath1184 - 2 r_0 ^ 2 ( r_0 ^ 2 + \xi_c ) \sqrt { \frac{- \xi_c } { r_0 ^ 2 + \xi_c } } \\ & \quad + ( r_0 ^ 2 + \xi_c ) ^2 \big [ \sqrt { \frac{- \xi_c } { r_0 ^ 2 + \xi_c } } + \frac13 \sqrt { \frac{- \xi_c } { r_0 ^ 2 + \xi_c } } ^3 \big ] \big\ } .\end{aligned}\ ] ] by direct computation .
since @xmath1185 is a simple zero of @xmath1186 , we have @xmath1187 thus @xmath1188 and @xmath1189 as a consequence , the first integral in is equal to @xmath1190 gathering these two relations , we obtain @xmath1191 as wished . since we have a kink solution @xmath325 for @xmath78 , this implies that @xmath1192 is negative in @xmath1193 and that @xmath73 is a simple zero of @xmath1194 , that is @xmath1195 .
then , @xmath1196 in @xmath1197 and @xmath1198 for @xmath1199 , it follows that there exists @xmath798 such that @xmath1200 we consider for @xmath1201 the quantity @xmath1202 the study of @xmath1203 is easy . [ minokink ] there holds @xmath1204 more precisely , for any @xmath1205 , @xmath1206 finally , if @xmath1205 , @xmath1207 and @xmath1208 , then there exists @xmath775 and @xmath197 such that @xmath1209 .
_ taking @xmath1210 as a comparison map , we see that @xmath1211 .
moreover , if @xmath1205 and @xmath1212 , we may assume , up to a translation , that @xmath1213 . then
, denoting @xmath1214 we have the inequalities @xmath1215 | \ dx \geq \big| \int_0^{+\ii } \p_x [ g(|u| ) ] \ dx \big| = \big| g(|u|(+\ii ) ) - g(|u|(0 ) ) \big| \\ = & \ , \big| g(r_0 ) - g(\mu )
\big| = 2 \int_\mu^{r_0 } \sqrt { f(s^2 ) } \ ds .\end{aligned}\ ] ] arguing similarly in @xmath1216 , we get @xmath1217 for the kink @xmath325 , which is real - valued , we have the first integral @xmath1218 , hence , using the change of variables @xmath1219 , @xmath1220 if @xmath1221 , we have then @xmath1222 , hence @xmath1223 as wished .
assume finally that @xmath1205 verifies @xmath1207 and @xmath1224
. then @xmath1225 and all the above inequalities are equalities .
in particular , we must have @xmath1226 and equality in @xmath1227 , which means that @xmath1228 . combining this ode with the condition @xmath1229
, we see that @xmath1230 , since @xmath1231 solves @xmath1232 .
finally , the fact that @xmath1233 implies that the phase is constant in @xmath1234 and in @xmath1235 : there exist two constants @xmath1236 satisfying @xmath1237 for @xmath1238 . therefore , @xmath1239 mod @xmath447 , and then @xmath1240 implies @xmath1241 mod @xmath401 , that is @xmath1242 in @xmath63 , which is the desired result .
+ we recall the expansion @xmath1243 as @xmath1244 , where @xmath1245 . from the hamilton group relation @xmath1246
, we also infer by integration @xmath1247 . as a first step ,
we define the small parameter @xmath453 .
the key point is to prove the following result .
[ supercool ] there exist some constant @xmath1248 and a small @xmath453 such that , for any @xmath1249 , @xmath1250 _ proof .
_ notice first that for @xmath1171 small , there exists @xmath123 travelling wave of speed @xmath51 and that @xmath1251 with @xmath72 a smooth function in @xmath51 such that @xmath1252 , hence @xmath1253 and is smooth .
therefore , there exists , for @xmath1254 small , a unique @xmath1255 , with @xmath1256 , such that @xmath1257 .
in particular , taking @xmath1258 as a comparison map in @xmath1259 , we have @xmath1260 in particular , it follows that , for some positive constant @xmath234 and for @xmath1261 small enough , [ paraudessus ] _ min ( ) e(u_0 ) + k ^2 e(u_0 ) .
consider now @xmath51 small , a bounded open interval @xmath1262 and @xmath1263 a solution to the newton equation @xmath1264 in @xmath1262 , with @xmath1265 , @xmath1266 and @xmath1267 .
as @xmath1268 , @xmath71 converges to @xmath1194 in @xmath1269 ) $ ] .
moreover , @xmath1270 is negative in @xmath1271 and has a simple zero at @xmath1272 .
therefore , if @xmath51 and @xmath1273 are sufficiently small , we must have @xmath1274 .
consequently , if @xmath1275 solves ( tw@xmath56 ) on a bounded interval @xmath1262 , satisfies @xmath1276 ( @xmath1277 ) and if @xmath1278 is @xmath1279 at @xmath1280 and at @xmath1281 , with @xmath1282 , then [ chero ] _
x_-^x_+ |_x v|^2 + f(|v|^2 ) dx 12 e(u_0 ) .
here , we use that the newton equation on the modulus @xmath1283 actually holds true in @xmath1262 .
since @xmath1196 in @xmath1197 and @xmath1284 when @xmath1285 , there exists @xmath600 and @xmath1286 such that @xmath1287 for @xmath1288 .
hence , if @xmath1289 , then [ garcia ] | p(v ) | e ( v ) .
moreover , arguing as in the proof of proposition [ minokink ] , we show that there exists @xmath1290 such that if @xmath1205 and @xmath1291 takes values @xmath1279 and @xmath1292 , then @xmath1293 in particular , since @xmath1294 , we may choose @xmath1261 sufficiently small so that if @xmath1205 and @xmath1295 , then @xmath1296 .
this means that for the mappings we are considering , @xmath1287 . +
* step 1 : construction of a suitable minimizing sequence .
* there exists a sequence @xmath1297 in @xmath49 such that @xmath1298 , @xmath1299 , @xmath1300 $ ] , @xmath1301 and @xmath1302 since @xmath1303 , the maps @xmath1304 we consider may be lifted @xmath1305 .
therefore ( with @xmath1306 ) , @xmath1307 the infimum in @xmath1308 may be written @xmath1309 for each @xmath1310 , we minimize in @xmath1308 a quadratic functional on an affine hyperplane , with minimizer given by @xmath1311 as a consequence , the infimum in @xmath1308 in is @xmath1312 = \inf _ { p \in \r } \big [ p^2 \big ( \int_\r \frac{(a^2-r_0 ^ 2)^2}{a^2 } \ dx \big)^{-1 } + 2 m r_0 ^ 4 \sin^2 \big ( \frac{p - \pi r_0 ^ 2 } { 2 r_0 ^ 2 } \big ) \big ] .\ ] ] it is clear that this last infimum is achieved only for @xmath1313 inside @xmath1314 $ ] .
indeed , the second term is @xmath1315-periodic and if @xmath1316 , then @xmath1317 is a better competitor .
moreover , the function @xmath1318 is continuous and even , hence we may consider some @xmath1319 $ ] ( depending on @xmath915 ) , which is a minimizer for this last infimum .
the corresponding @xmath1320 is then a minimizer for the infimum in @xmath1308 in .
writing that @xmath1321 = 2 p \big ( \int_\r \frac{(a^2-r_0 ^ 2)^2}{a^2 } \ dx \big)^{-1 } + 2 m r_0 ^ 2 \sin \big ( \frac { p - \pi r_0 ^ 2}{2r_0 ^ 2 } \big ) = 0 , \ ] ] we deduce the relations @xmath1322 we conclude by considering a minimizing sequence @xmath1323 in , and translating in space so that @xmath1324 .
+ since @xmath1325 in @xmath1326 , we have @xmath1327 for @xmath641 large . therefore , by compact sobolev embedding @xmath1328 ) \hookrightarrow l^\ii([-r,+r])$ ] , we may assume , up to a possible subsequence , that there exists @xmath1329 such that for any @xmath1330 , @xmath1331 in @xmath1328 ) $ ] and @xmath1332 uniformly on @xmath1333 $ ] . moreover , by lower semicontinuity and fatou s lemma , @xmath1334 . since @xmath1335 in @xmath63 , @xmath1336 in @xmath191 and we may lift @xmath1337 . furthermore ,
@xmath1338 , hence @xmath1339 .
we also know that @xmath1340 $ ] for all @xmath641 , hence we may assume , up to another subsequence , that @xmath1341 converges to some @xmath1342 $ ] .
we also set @xmath1343 in view of step 1 , and the convergence @xmath1344 uniformly on any compact interval @xmath1333 $ ] , it follows that [ zorro ] 2 a^2 _ x = c ( a^2 - r_0 ^ 2 ) _ x _ n _ x l^_loc ( ) .
note that @xmath1345 hence @xmath1346 at @xmath60 .
in particular , there exist @xmath1347 such that @xmath1348 in @xmath1349 and in @xmath1350 and @xmath1351 . + * step 2 .
* there exist @xmath1352 such that @xmath1353 we work for @xmath1354 , the other case being similar .
we consider @xmath1355 , @xmath1356 small such that @xmath1357 verifies @xmath1358 in @xmath1350 .
this is possible since @xmath1359 . then , @xmath1360 in @xmath191 and @xmath1361 ,
hence @xmath1362 is then a comparison map for @xmath1259 , and in view of the equality @xmath1363 , it follows that @xmath1364 letting @xmath1365 and using the weak and strong convergences for @xmath1366 , we infer @xmath1367 dividing by @xmath1368 and letting @xmath1369 and then @xmath1370 , we deduce that @xmath1371 solves ( tw@xmath56 ) in @xmath1350 and @xmath1304 has finite energy .
moreover , @xmath1372 is small , thus @xmath1373 in @xmath1350 for some constants @xmath1374 and @xmath1375 , and the speed @xmath51 is such that @xmath1376 , hence @xmath1377 . since @xmath1283 has finite energy in @xmath63 and solves ( tw@xmath56 ) in @xmath1350 , @xmath1304 is @xmath734 in @xmath1378 .
moreover , @xmath1283 reaches a minimum at @xmath1379 , thus we must have @xmath1380 , which imposes @xmath1381 .
note that @xmath1382 being even , it is possible to translate @xmath1304 so that @xmath1383 and @xmath1384 .
observe that @xmath1385 , hence @xmath1386 .
this yields [ cacoute ] _
|x| r |_x v|^2 + f(|v|^2 ) dx = _ |x| z |_x u_c|^2 + f(|u_c|^2 ) dx e(u_0 ) - k .
in particular , we deduce from @xmath1387 hence @xmath1388 for @xmath1389 small ( @xmath1390 ) . * step 3 .
* we prove that @xmath1391 in @xmath1392 .
+ indeed , if it is not the case , there exists a bounded interval @xmath1393 such that @xmath1394 in @xmath1395 and @xmath1396 , with @xmath1397 .
therefore , we can make perturbations of the amplitude @xmath1398 localized in @xmath1399 .
hence , arguing as in step 2 , we see that then , @xmath1304 solves ( tw@xmath56 ) in @xmath1395 , with @xmath1400 and @xmath1396 , @xmath1397 .
we then are in position to apply , yielding @xmath1401 but the combination with provides @xmath1402 which is not possible if @xmath1261 is sufficiently small .
+ * step 4 .
* we have @xmath1403 or ( @xmath1404 and @xmath1405 ) .
+ indeed , assume @xmath1406 , and consider @xmath1407 , @xmath1408 , @xmath672 and @xmath1409 , so that @xmath1410 in @xmath191 .
since @xmath1406 , we actually have @xmath1411 and @xmath1362 is a comparison map for @xmath1259 . arguing as before , we thus have @xmath1412 by
, we may pass to the limit as @xmath1365 in all the terms and deduce @xmath1413 at this stage , we see the relevance of taking a minimizing sequence as chosen in step 1 , since it allows to pass to the limit in the nonlinear terms involving @xmath1414 . as a consequence , using , @xmath1415 in the distributional sense in @xmath1416 .
the term @xmath1417 is continuous in @xmath191 .
however , since @xmath1418 for @xmath1419 and @xmath1420 for @xmath1421 , we infer @xmath1422 plus a piecewise continuous function in the distributional sense in @xmath1416 .
since @xmath1423 ( recall that @xmath1424 ) , this forces to have @xmath1425 , that is @xmath1426 .
consequently , @xmath1427 and then @xmath1405 . + in the next step , we take into account the loss in the weak convergence @xmath1331 . +
* there exists @xmath600 such that @xmath1428 let @xmath159 be fixed but small , and pick some @xmath1429 large so that @xmath1430 we claim that there exists some small @xmath1431 , independent of @xmath284 such that @xmath1432 for @xmath1433 and @xmath641 large . indeed ,
otherwise , we may argue as in step 3 and show , as in the beginning of the proof there , that @xmath1434 .
this is not possible since @xmath1435 and @xmath1436 is close to @xmath1437 as @xmath1438 .
therefore , as for , @xmath1439 consequently , @xmath1440 passing to the liminf and using the weak convergence in @xmath1441 $ ] , we infer @xmath1442 however , implies @xmath1443 so that @xmath1444 letting @xmath1445 , the conclusion follows .
+ * step 6 .
* there exists @xmath295 such that , if @xmath1446 , then @xmath1447 we recall the expansion @xmath1243 as @xmath1244 , where @xmath1245 . from the hamilton group relation @xmath1246
, we also infer by integration @xmath1448 . on the other hand , by definition of @xmath1449 , @xmath1450 \\
= & \ , m r_0 ^ 4 \big [ 1 - \sqrt{1 - \sin^2 \big ( \frac{p(v_n ) - r_0 ^ 2 \pi}{r_0 ^ 2 } \big ) } \big ] = m r_0 ^ 4 \big [ 1 - \sqrt{1 - \frac{c_n^2 } { m^2 } } \big ] \end{aligned}\ ] ] for @xmath641 large . here , we have used that @xmath1451 $ ] ( _ cf . _ step 2 ) , thus @xmath1452 , for otherwise , we would have , by proposition [ minokink ] @xmath1453 \\
\geq & \ , e(u_0 ) - k \mu _ * + 2 m r_0 ^ 4 + \bo(\mu_*^2 ) , \end{aligned}\ ] ] but this contradicts if @xmath1261 is sufficiently small .
we assume @xmath1406 , so that , by step 4 , @xmath1454 and @xmath1455 .
we recall the expansion @xmath1456 . by definition of @xmath1457
, one has @xmath1458
\leq \varliminf_{n \to + \ii } e(v_n ) + \lim_{n \to + \ii } 2 m r_0 ^ 4 \sin^2 \big ( \frac{p(v_n ) - r_0 ^ 2 \pi}{2 r_0 ^ 2 } \big ) = \varliminf_{n \to + \ii } \mathscr{k } ( v_n ) = \mathscr{k}_{\rm min}(\mu)\ ] ] since @xmath1459 is minimizing for @xmath1259 .
moreover , from the expression of @xmath1304 , it holds ( for @xmath1446 ) @xmath1460 \quad \quad \quad { \rm and } \quad \quad \quad p(v ) = p ( u_{\s_\mu } ) + r \s_\mu \frac{(r_0 ^ 2- \mu^2)^2 } { \mu^2 } .\ ] ] furthermore , @xmath1461 and @xmath1462 with @xmath1463 $ ] and @xmath1464 , thus @xmath1465 combining this with the expansion of @xmath1466 and @xmath1467 gives @xmath1468 + 2 r \big [ \frac{\s_\mu^2(r_0 ^ 2- \mu^2)^2}{4\mu^2 } + f(\mu^2 ) \big ] \\
\geq & \ ,
e(u_0 ) + \frac{| p_\sharp | } { k } + \frac{\s^2_\mu}{2 } \big [ \dot{p}_0 + \frac{1}{m } \big ] + o(\mu^2 ) + 4 r f(0 ) \\
\geq & \ , e(u_0 ) + \frac{1 } { k } \big| r_0 ^ 2 \arcsin ( \s_\mu / ( m r_0 ^ 2 ) ) + \s_\mu \dot{p}_0 + r \s \frac{(r_0 ^ 2- \mu^2)^2 } { \mu^2 } + o(\s_\mu ) \big| + \frac{\s^2_\mu}{2 } \big [ \dot{p}_0 + \frac{1}{m } \big ] + o(\mu^2 ) + 4 r f(0 ) \\
\geq & \ , e(u_0 ) + \frac{1 } { k } \big| \frac { \s_\mu } { m } + \s_\mu \dot{p}_0 + r \s_\mu \frac{(r_0 ^ 2- \mu^2)^2 } { \mu^2 } + o(\s_\mu ) \big| + \frac{\s^2_\mu}{2 } \big [ \dot{p}_0 + \frac{1}{m } \big ] + o(\mu^2 ) + 4 r f(0 ) \\
\geq & \ , e(u_0 ) + \frac { \s_\mu}{k } \big| \dot{p}_0 + \frac{1 } { m } + r \frac{(r_0 ^ 2- \mu^2)^2 } { \mu^2 } \big| + \frac{\s^2_\mu}{2 } \big [ \dot{p}_0 + \frac{1}{m } \big ] + 4 r f(0 ) + o(\mu^2 ) .\end{aligned}\ ] ] the right - hand side is a continuous piecewise affine function of @xmath1469 ( the `` @xmath1470 '' does not depend on @xmath1469 ) . since @xmath1471 and @xmath1472 ( since @xmath1473 by hypothesis )
, it follows that the right - hand side is a function of @xmath1469 which is decreasing in @xmath1474 $ ] and increasing in @xmath1475 , with @xmath1476 therefore , using once again that @xmath1477 , @xmath1478 + 4 r_0(\mu ) f(0 ) + o(\mu^2 ) \\
= & \ , e(u_0 ) + \big [ \dot{p}_0 + \frac{1}{m } \big ] \frac{2 \mu^2 f(0)}{r_0 ^ 4 } - \big [ \dot{p}_0 + \frac{1}{m } \big ] \frac{4
\mu^2 f(0)}{r_0 ^ 4 } + o(\mu^2 ) \\ = & \ , e(u_0 ) - \mu^2 \frac{2 f(0)}{r_0 ^ 4 } \big [ \dot{p}_0 + \frac{1}{m } \big ] + o(\mu^2 ) .\end{aligned}\ ] ] in view of our hypothesis @xmath1479 , we infer that @xmath1480 for @xmath1481 sufficiently small and some positive constant @xmath234 , as wished . if the assumption @xmath1479 is not verified , but if @xmath1482 for instance , then the function of @xmath1469 above is increasing in @xmath1483 , with minimum value achieved at @xmath1484 and equal to @xmath1485 + o(\mu^2 ) = e(u_0 ) + \frac{2 \mu^2 f(0)}{r_0 ^ 4 } \big [ \dot{p}_0 + \frac{1}{m } \big ] + o(\mu^2 ) \geq e(u_0 ) + \frac{\mu}{k } .\ ] ] we then would have concluded a stronger estimate , which is actually in contradiction with , hence we are necessarily in the case @xmath1406 .
the assumption @xmath1479 is however crucial for the last step . + * step 7 .
* we assume @xmath1479 .
then , for @xmath1261 sufficiently small , the case @xmath1403 does not occur .
+ we argue in a similar way , but since @xmath1403 , the expressions for @xmath1486 and @xmath1487 are given by @xmath1488 here , we have used that @xmath1489 since @xmath68 solves ( tw@xmath56 ) .
combining this here again with the expansion of @xmath1490 and @xmath1491 gives , using that @xmath1492 , @xmath1493 - 4 \int_0^z f(|u_c|^2 ) \ d x \\
\geq & \ , e(u_0 ) + \frac{| p_\sharp | } { k } + \frac{c^2}{2 } \big [ \dot{p}_0 + \frac{1}{m } \big ] - 4 z f(0 ) + o(\mu^2 ) \\
\geq & \ , e(u_0 ) + \frac{1 } { k } \big| \arcsin ( c / m ) + c \dot{p}_0 - c \int_0^z \frac{(r_0 ^ 2- a_c^2)^2 } { a_c^2 } \ d x + o(c ) \big| + \frac{c^2}{2 } \big [ \dot{p}_0 + \frac{1}{m } \big ] + o(\mu^2 ) - 4 z f(0 ) .\end{aligned}\ ] ] following the lines of the proof of lemma [ prolongement ] , we have [ jdevlop ] c _ 0^z d x = 2 ( ) + ( ^2 ) . indeed , noticing that @xmath1494 in
@xmath1495 $ ] with @xmath1386 , we write , expanding the square , @xmath1496 then , using the change of variable @xmath1497 , @xmath1498 by computations similar to those for the proof of lemma [ prolongement ] .
this proves .
therefore , [ pasmal ] _ min ( ) e(u_0 ) + | c - 2 ( ) + o(c ) | + + o(^2 ) - 4 z f(0 ) . by , the left - hand side is @xmath1499 .
since @xmath1479 , @xmath1500 , @xmath1501 and @xmath1390 , this implies @xmath1502 - 2 \arctan \big ( \sqrt{\frac{\mu^2}{r_0 ^ 2 + \xi_c } } \big ) + o(c ) \big| \leq k \mu , \ ] ] thus @xmath1503 and finally , for @xmath1261 small enough , @xmath1504 combining this with the equality @xmath1505 seen during the proof of lemma [ prolongement ] , we infer @xmath1506 in particular , going back to and since for @xmath1507 , @xmath1508 this implies @xmath1509 which provides ( since @xmath1510 ) @xmath1511 inserting this into and keeping in mind that the left - hand side is @xmath1499 , we deduce @xmath1512 - 2 \arctan \big ( \sqrt{\frac{\mu^2}{r_0 ^ 2 + \xi_c } - 1 } \big ) = o ( \mu ) .\ ] ] however , since @xmath1513 , this gives @xmath1514 \sim \frac{2 \mu \sqrt{f(0)}}{r_0 ^ 2 } \big [ \dot{p}_0 + \frac{1 } { m } \big ] , \ ] ] yielding a contradiction for small @xmath1515 since we have @xmath1516 by assumption
. therefore , the case @xmath1403 does not occur for sufficiently small @xmath1261 .
if we had @xmath1482 , we would not have been able to show that @xmath1259 gives a control on @xmath1517 .
+ the proof of proposition [ supercool ] is complete .
+ * proof of theorem [ statiomini ] .
* let @xmath1518 .
if @xmath1519 , then proposition [ supercool ] gives @xmath1520 .
if @xmath1207 , we deduce from proposition [ minokink ] that @xmath1521 .
hence @xmath1522 except if @xmath1523 . from the study of the equality case in proposition [ minokink ]
, it follows that @xmath1524 , as claimed . as a first step
, we shall need a quantified version of proposition [ minokink ] .
[ minokinkameliore ] there exists @xmath1525 and @xmath600 , depending only on @xmath13 such that , for any @xmath1205 verifying @xmath1526 there holds @xmath1527 _ proof . _
first , we translate the problem in space so that @xmath1528 and shall choose the phase factor later .
we follow the lines of the proof of proposition [ minokink ] and actually get ( writing @xmath216 locally in @xmath1529 ) @xmath1530 ^ 2 \ dx \\ & \ , + 2 \int_0^{+\ii } | \sqrt { f(|u|^2 ) } \p_x |u| | \
dx \\ \geq & \ , \int_0^{+\ii } { \bf 1}_{|u| > 0 } a^2 ( \p_x \phi ) ^2 \ dx + \int_0^{+\ii } \big [ \sqrt { f(|u|^2 ) } - | \p_x |u| \ , | \big]^2 \ dx + 2 \int_\mu^{r_0 } \sqrt { f(s^2 ) } \ ds .\end{aligned}\ ] ] arguing similarly in @xmath1216 , we get [ merou ] e(u ) e(u_0 ) + _ _
|u| > 0 a^2 ( _ x ) ^2 dx + _ ^2 dx - 4 _
the gradient of the phase is controlled using .
we shall now estimate the modulus part .
let us denote @xmath1531 and @xmath1532 for which we have , by , [ bornh ] h _
l^2()^2 e(u ) - e(u_0 ) + 4 _ 0^
ds e(u ) - e(u_0 ) + k . recall that @xmath325 verifies @xmath1533 in @xmath63 , hence @xmath1232 in @xmath1534 . setting @xmath1535 ,
we infer @xmath1536 we set , for @xmath1537 , @xmath1538 since @xmath325 verifies @xmath1539 and @xmath1540 in @xmath1326 , it follows that @xmath1541 moreover , by taylor expansion , we infer the existence of @xmath1542 and @xmath1543 such that , for @xmath1544 , @xmath1545 , @xmath1546 the estimate is clearly uniform in view of the exponential decay of @xmath1547 at infinity . therefore , [ jolieedo ] _ x = + g ( x , ) + h(x ) .
we view this ode as a linear ode with source term @xmath1548 .
since @xmath1547 solves the homogeneous equation , we infer from duhamel s formula and the fact that @xmath1549 that for @xmath1550 , [ duhamel ] ( x ) = + _ x u_0 ( x ) _ 0^x dz .
we shall prove that this equation implies that if @xmath1389 and @xmath1551 are sufficiently small , then [ normelinf ] _
l^2(_+ ) k ( h _ l^2(_+ ) + ) .
we assume @xmath1552 .
note that since @xmath325 is a kink , we have the decays given in proposition [ decroiss ] .
hence , there exists two positive constants @xmath1553 and @xmath1554 such that @xmath1555 in particular , if @xmath1556 in the interval @xmath1557 $ ] , then implies , for @xmath1558 $ ] , @xmath1559 ) } |\theta ( z)| + |h| ( z ) \big ] \ dz \\ \leq & \ , \mu + \frac{k k_1 k_2}{\cs } \n \theta \n_{l^\ii([0,r])}^2 + \frac{k_1 k_2 } { \sqrt { 2\cs } } \n h \n_{l^2(\r_+)}\end{aligned}\ ] ] by cauchy - schwarz .
we thus choose @xmath1560 sufficiently small so that @xmath1561 then , we consider the set @xmath1562 of all @xmath1406 such that @xmath1563 in the interval @xmath1557 $ ] . since @xmath1564 is continuous by sobolev embedding and @xmath1565 , @xmath1566 and
is closed in @xmath1567 .
moreover , the above estimate shows that for @xmath1568 , @xmath1569 ) } \leq \mu + \frac{k k_1 k_2}{\cs } \n \theta \n_{l^\ii([0,r])}^2 + \frac{k_1 k_2 } { \sqrt { 2\cs } } \n h \n_{l^2(\r_+ ) } , \ ] ] which gives @xmath1569 ) } \big ( 1 - \frac{k k_1 k_2}{\cs } \n \theta \n_{l^\ii([0,r ] ) } \big ) \leq \mu + \frac{k_1 k_2 } { \sqrt { 2\cs } } \n h \n_{l^2(\r_+ ) } , \ ] ] and then [ borninfi ] _
l^([0,r ] ) 2 < _ 0 . consequently , @xmath1562 is open in @xmath1567 . by connexity , @xmath1570 , proving . in
what follows , we assume @xmath1571 sufficiently small so that @xmath1572 , thus @xmath1573 . in particular , @xmath1574 ) } |\theta ( z)| + |h| ( z ) \big ] \ dz .\ ] ] for @xmath1406 to be determined later , we then deduce from classical convolution estimates that @xmath1575 ) } \leq \mu \sqrt{r } + k_3 \n \theta \n_{l^\ii(\r_+ ) } \n \theta \n_{l^2([0,r ] ) } + k_3 \n h \n_{l^2(\r_+ ) } .\ ] ] imposing @xmath1571 smaller if necessary , we may assume that @xmath1576 so that we get @xmath1575 ) } \leq k_4 \big ( \mu \sqrt{r } + \n h \n_{l^2(\r_+ ) } \big ) .\ ] ] reporting this into provides @xmath1577 ) } \leq k_5 \big ( \mu^2 r + \n h \n_{l^2(\r_+)}^2 \big ) .\ ] ] arguing similarly in @xmath1578 $ ] and using , we obtain an @xmath269 estimate for @xmath1579 in @xmath1580 $ ] : [ near ] _
h^1([-r , r])^2 k_6 ( e(u ) - e(u_0 ) + ^2 r + ) .
we now turn to the estimate in @xmath1581 . for that purpose , we write @xmath1582 since @xmath325 decays exponentially ( see proposition [ decroiss ] ) , it follows that @xmath1583 furthermore , by integration by parts , @xmath1584 ^ 2 ) - f ( u_0 ^ 2 ) -2u_0 f'(u_0 ^ 2 ) \theta \ dx \\
\leq & \ , \int_{|x| \leq r } 2 \theta [ \p^2_x u_0 + u_0 f ( u_0 ^ 2 ) ] \ dx - 2 \theta ( + r ) \p_x u_0 ( + r ) + 2 \theta ( -r ) \p_x u_0 ( -r ) + k \n \theta \n_{h^1([-r,+r])}^2 \\
\leq & \ , k \ex^ { - \cs r } + k \big ( e(u ) - e(u_0 ) + \mu^2 r + \mu \big ) .\end{aligned}\ ] ] for the before last line , we have used that @xmath1585 ^ 2 ) - f ( u_0 ^ 2 ) - 2 u_0 f ' ( u_0 ^ 2 ) \theta $ ] is @xmath1586 as @xmath1587 and for the last line , that @xmath325 solves @xmath1588 , the exponential decay of @xmath1589 and the uniform bound on @xmath1590 .
reporting these estimates into provides @xmath1591 .\end{aligned}\ ] ] combining this with , we deduce that for any @xmath1446 , there holds @xmath1592 .\ ] ] we then choose @xmath1593 if @xmath1303 or @xmath1594
if @xmath1221 , and get @xmath1595 notice that if @xmath1596 everywhere , then we may give a quick proof of the above estimate , since , using here again integration by parts and that @xmath1597 , we may deduce that @xmath1598 and since @xmath1596 , @xmath1599 by taylor expansion , providing the desired @xmath1600 bound on @xmath1590 .
observe now that @xmath1601 ^ 2 \ dx , \ ] ] hence @xmath1602^{1/2 } .\end{aligned}\ ] ] we distinguish now the cases @xmath1603 and @xmath1303 , and begin with the assumption @xmath1303 .
then , we have a global lifting @xmath1604 and @xmath1605^{1/2 } .\end{aligned}\ ] ] now , we notice that @xmath1606 and that [ chapon ] ( u ) - e(u_0 ) 2 m r_0 ^ 4 ^2 ( ) ( p(u ) - r_0 ^ 2 mod 2 r_0 ^ 2 ) ^2 .
we define @xmath1607 . by cauchy - schwarz
, we have @xmath1608 inserting this into gives @xmath1609 \\
\frac{k}{\inf_{|x| \geq \d } a } \big ( \mathscr{k}(u ) - e(u_0 ) + \mu \big)^{1/2 } .\end{aligned}\ ] ] in addition , by cauchy - schwarz , @xmath1610 consequently , @xmath1611 \big ( \mathscr{k}(u ) - e(u_0 ) + \mu \big)^{1/2 } .\end{aligned}\ ] ] from our choice @xmath1612 and since @xmath1613 , we infer @xmath1614 . similarly , @xmath1615 . reporting this into yields
@xmath1616 we now freeze the gauge invariance by imposing @xmath1617 .
note that then @xmath1618 .
furthermore , since @xmath1617 , @xmath1619 which implies , for @xmath1620 , @xmath1621 and similarly , since @xmath1622 , for @xmath1623 , @xmath1624 we turn back to and infer @xmath1625 inserting these estimates in , it follows that @xmath1626 we now turn to the case @xmath1603 . without loss of generality , we may assume that @xmath1627 in @xmath1628 ( since @xmath1629 at @xmath413 ) , and let @xmath1630 be such that @xmath1631 and @xmath1627 in @xmath1632 .
we first estimate @xmath1633 by writing that @xmath1634 thus @xmath1635 moreover , we have two local liftings @xmath1636 in @xmath1637 and @xmath1638 in @xmath1639 .
going back to , we then deduce @xmath1640^{1/2 } .\end{aligned}\ ] ] arguing as for the case @xmath1641 , we obtain @xmath1642 in @xmath1643 and in @xmath1644 . by definition of @xmath423
, we have @xmath1645 in @xmath404 , hence the same arguments as in the case @xmath1646 provide @xmath1647 since the integral @xmath1648 is bounded by @xmath1649 by cauchy - schwarz . imposing @xmath1650 for the gauge invariance , we infer @xmath1651 and @xmath1652 , hence @xmath1653 .
therefore , we conclude as before that @xmath1654 which finishes the proof of the proposition . + in order to prove theorem [ stabmini ] , we use proposition [ supercool ] , which provides @xmath1655 thus @xmath1656 inserting this bound in proposition [ minokinkameliore ] then gives @xmath1657^{1/4 } \leq k \sqrt[8]{\mathscr{k}(u ) - e(u_0 ) } , \ ] ] and the proof is complete .
we have left aside in our study the case of the sonic waves ( @xmath1659 ) , but would like to say a few words on the difficulties associated with this critical case .
we note that if there exists a sonic nontrivial travelling wave , it does not vanish , hence we may use the hydrodynamical formulation of ( nls ) as in @xcite .
the point is that the sturm - liouville operator ( see section 4 in @xcite ) @xmath1660 with @xmath1661 has , by weyl s theorem , essential spectrum @xmath1662 when @xmath131 .
indeed , we know from proposition [ decroiss ] that @xmath486 and its derivatives tend to zero at infinity , hence , as @xmath119 , @xmath1663 since @xmath1664
. therefore , there does not exist @xmath160 such that @xmath1665 for any @xmath1313 orthogonal to the subspace spanned by the negative and the zero eigenvalue , and thus the grillakis - shatah - strauss theory does not apply . in the case
@xmath1666 , where it is natural to expect stability , a natural thing would be to try to work with the functional @xmath1667 and to follow the lines of the proof of theorem [ statiomini ] .
indeed , the spectral analysis shall not give positive definiteness of the hessian due to presence of essential spectrum down to @xmath75 .
therefore , we may study @xmath255 at fixed @xmath1668 close to @xmath1669 .
when @xmath171 and @xmath1670 , the infimum of @xmath166 contains a neighborhood of @xmath1671 for @xmath51 close to @xmath210 . for @xmath232 ,
this is no longer the case : we have only a one sided neighborhood of @xmath1669 .
it is plausible that the study for @xmath1515 in this one sided neighborhood of @xmath1669 can be done as in the proof of theorem [ statiomini ] , but for the remaining values of @xmath1515 , we have to find a sharp ansatz , which is not very easy to find .
furthermore , for the linear instability which is expected if @xmath1672 , let us mention the following point . for the eigenvalue problem studied in @xcite , the characteristic equation for the constant coefficient limit at infinity , namely @xmath1673 becomes , when @xmath232 , [ karakcs ] r^4 - 2
r + ^2 = 0 .
the behaviour of the roots for small @xmath1674 is then different from the case @xmath171 .
indeed , there exists a root @xmath1675 for @xmath1676 , and for the three other roots , we use the variable @xmath1677{\lambda } z $ ] , which transforms @xmath1678 into @xmath1679 .
this last equation has , for @xmath1676 , three roots @xmath1680{2\cs } $ ] , where @xmath1681 and @xmath1682 , @xmath516 , @xmath1683 .
in particular , has three roots @xmath1680{2\cs \lambda } $ ] , @xmath1682 , @xmath516 , @xmath1683 .
the value @xmath1684 is then a branching point , and we shall have a smooth problem not in @xmath1685 but in @xmath1686 { \lambda } $ ] . since analyticity is not necessary for our purpose , we may define an evans function @xmath1687 in @xmath1688 , smooth , and such that , for @xmath1689 , @xmath1690 { \lambda } ) = 0 $ ] if and only if @xmath1685 is an unstable eigenvalue for .
another difficulty comes from the fact that it will be difficult to find an analytic extension of the evans function @xmath1687 near @xmath75 since , by proposition [ decroiss ] , for @xmath232 , @xmath1691 and @xmath1692 decay only at an algebraic rate and not an exponential rate .
consequently , we can not use the gap lemma of @xcite and @xcite .
finally , as a straightforward computation shows , the stable and unstable subspaces for the eigenvalue problem are transverse for @xmath1693 but their continuous extensions at @xmath1684 have a nontrivial intersection .
therefore , both stability and instability requires some further analysis , and the situation is then much more delicate than the one studied in subsection [ sexins ] .
we work with the notations of @xcite , and recall them briefly .
we consider a hamiltonian equation in a real hilbert space @xmath1694 , with scalar product @xmath1695 , under the form = j e(u ) , where @xmath1696 is a closed linear operator with dense domain and skew - symmetric .
assume that @xmath1697 is a @xmath1698-group of unitary operators in @xmath1694 generated by @xmath1699 , which is skew - adjoint and with dense domain , and that @xmath89 is invariant by @xmath1697 , that is @xmath1700 for any @xmath1701 , @xmath1702 .
assume moreover that @xmath1703 for any @xmath1701 and that there exists @xmath1704 linear and bounded such that @xmath1705 and @xmath1706 is an extension of @xmath1699 .
we then set @xmath1707 the basic assumptions of @xcite are the following ones .
+ _ assumption 1 ( existence of solutions ) : _ for any @xmath1708 there exists @xmath1709 , depending only on @xmath1710 , such that for any @xmath1711 , there exists a @xmath1712 with @xmath1713 solution of ( @xmath1714 ) in the sense that for any @xmath1715 , @xmath1716 and verifying @xmath1717 and @xmath1718 for @xmath1719 .
+ _ assumption 2 ( existence of `` bound states '' ) : _ there exists an interval @xmath1720 , not reduced to a singleton , and a mapping @xmath1721 of class @xmath364 such that , for any @xmath1722 , @xmath1723 _ assumption 3 ( spectral decomposition ) : _ for each @xmath1722 , the operator @xmath1724 has its kernel spanned by @xmath1725 , has one negative simple eigenvalue and the rest of its spectrum is positive and bounded away from zero .
+ under assumption 2 , we consider some @xmath1726 and the associated bound state @xmath1727 , and then define , for @xmath1728 , the functional @xmath1729 it is clear that @xmath1727 is a critical point of @xmath1730 : @xmath1731 .
we denote by @xmath1732 its second derivative , which is a self - adjoint operator .
the main result of this appendix is the following . + * theorem a * _ we make assumptions 2 and 3 and suppose moreover that the operator @xmath1733 is a compact perturbation of @xmath1734 .
if @xmath1735 and @xmath1736 there exists @xmath160 such that @xmath1737 in particular , for any @xmath1738 with @xmath1739 , we have @xmath1740 therefore , when assumption 1 is moreover verified , the ( global ) solution @xmath1741 to ( @xmath1714 ) with initial datum @xmath1742 verifies @xmath1743 provided the right - hand side is sufficiently small .
_ we point out that the condition that the operator @xmath1733 is a compact perturbation of @xmath1734 is not very restrictive , since in many cases coming from pde s , it involves less derivatives than @xmath1744 and @xmath1745 tends to zero at spatial infinity .
+ this type of liapounov functional has been used by i. barashenkov in @xcite to prove that the travelling waves of ( nls ) in dimension one are stable when @xmath1746 .
the proof follows basically the one in @xcite , but some points have to be clarified .
the interest of this type of liapounov functional is that the saddle point @xmath1727 is now a non degenerate local minimum for @xmath1730 .
this is a great advantage for numerical simulation of the `` bound states '' , since a gradient flow method on @xmath1730 can be used .
this approach has been used , with a very similar functional , in @xcite by n. papanicolaou and p. spathis for the numerical simulation of the travelling waves for a planar ferromagnets model . in the same spirit , in @xcite
, we also use a gradient flow method on this type of functional for the numerical simulation of the travelling waves for ( nls ) in two dimensions .
+ _ proof of theorem a. _ recall that the spectrum of @xmath1734 is , by assumption 3 , such that @xmath1747 , @xmath1748 and @xmath1749 for some @xmath1750 . since we assume that @xmath1751 is a compact perturbation of @xmath1734 , the essential spectrum of @xmath1752 is the same as the one of @xmath1734 , hence is included in @xmath1753 .
furthermore , @xmath1754 and @xmath1755 by assumption 3 .
since @xmath1756 and @xmath1757 is an extension of @xmath1758 , we have @xmath1759 , hence @xmath1760 . noticing that @xmath1733 is a nonnegative operator
, we infer that @xmath1761 is one - dimensional .
therefore , it suffices to show that @xmath1752 has no eigenvalues in @xmath1628 .
as we have seen that @xmath1762 is a nonnegative operator , we deduce that @xmath1763 .
let us first show that @xmath1764 by contradiction . if @xmath1765 is an eigenvalue of @xmath1752 , then there exists @xmath1766 , @xmath1767 , such that @xmath1768 . taking the duality product with @xmath1769 yields @xmath1770
since the two terms in the sum are nonnegative , this implies @xmath1771 and @xmath1772 , which in turn implies @xmath1773 ( here , @xmath812 is a negative eigenvector of @xmath1774 for the eigenvalue @xmath1775 . as a consequence
, we must have @xmath1776 . on the other hand , differentiating the equality @xmath1777 at @xmath1778 yields @xmath1779 , where @xmath1780 .
thus we must have @xmath1781 .
therefore , @xmath1782 is orthogonal to @xmath812 and this gives @xmath1783 .
however , this is not possible if @xmath1784 , since we have @xmath1785 . as a consequence ,
if @xmath1685 is a negative element of the spectrum of @xmath1752 , then @xmath1786 and @xmath1787 is an eigenvalue : there exists @xmath1788 such that @xmath1767 and @xmath1789 since @xmath1786 , we then infer v = - m q ( _ _ * ) , v _ ^ * , ( h _ _ * - ) ^-1 q ( _ _ * ) . since @xmath1790 , we can not have @xmath1791 .
then , taking the scalar product of ( a.1 ) with @xmath1792 ( here , @xmath1793 is the usual riesz isomorphism ) gives @xmath1794 it is clear that @xmath1795 is smooth in @xmath1796 and that @xmath1797 we now study the limit of @xmath1795 at @xmath1798 .
let us recall that @xmath1799 and that we have already seen that @xmath1800 , _ i.e. _ @xmath1792 is orthogonal to @xmath1801 .
therefore , as @xmath1802 , @xmath1803 and it thus comes @xmath1804 since @xmath1805 by hypothesis , it follows that if @xmath1806 , the function @xmath1795 increases in @xmath1807 and tends to some negative limit at @xmath1798 .
in particular , @xmath1795 is negative , hence we can not have @xmath1808 with @xmath1809 .
we have therefore shown that the spectrum of @xmath1752 consists in a simple eigenvalue @xmath75 with eigenspace spanned by @xmath1810 and the rest of the spectrum is positive and bounded away from @xmath75 .
this concludes the proof .
+ we would like to point out that in the proof of @xcite , the fact that @xmath1764 was not shown , the kernel of @xmath1752 was not studied and the essential spectrum was not considered
. moreover , the functional spaces are not given , hence we do not know for which perturbations stability holds .
we still consider in this appendix an abstract hamiltonian equation in the framework of @xcite = j e(u ) on the real hilbert space @xmath1694 , with scalar product @xmath1695 . here
@xmath1811 is of class @xmath14 and @xmath1696 is a closed linear operator with dense domain and skew - symmetric in the sense that @xmath1812 for @xmath1702 , @xmath1813 .
we assume that there exists a @xmath1698-group @xmath1697 of unitary operators in @xmath1694 generated by @xmath1699 , which is skew - adjoint and with dense domain , and that @xmath89 is invariant by @xmath1697 , that is @xmath1814 for any @xmath1815 , @xmath1702 .
assume moreover that @xmath1816 for any @xmath1815 and that there exists @xmath1704 linear and bounded such that @xmath1705 and @xmath1706 is an extension of @xmath1699 .
we then set @xmath1817 which is invariant by the flow ( @xmath1714 ) ( see @xcite ) . by `` bound state '' , we mean a particular solution @xmath1818 of ( @xmath1714 ) of the form @xmath1819 for some @xmath1815 and where @xmath1820 , @xmath1821 . in other words , @xmath1822 .
there exists an open interval @xmath1720 , not reduced to a singleton , and a mapping @xmath1823 of class @xmath364 such that , for any @xmath1722 , @xmath1723 the solution @xmath1819 is said to be stable in @xmath1694 if for any @xmath733 , there exists @xmath160 such that any solution to ( @xmath1714 ) with initial datum @xmath1824 is global in time and remains in @xmath1825 for @xmath672 .
otherwise , it is said unstable .
this supposes some knowledge on the cauchy problem for ( @xmath1714 ) ( at least existence of solutions ) .
if we are given some banach space @xmath1826 with continuous imbedding @xmath1827 , we may also say that the solution @xmath1819 is said to be stable from @xmath1694 to @xmath1828 if for any @xmath733 , there exists @xmath160 such that any solution to ( @xmath1714 ) with initial datum @xmath1824 remains in @xmath1829 for @xmath672 . clearly , a solution stable in @xmath1694 is precisely a solution able from @xmath1694 to @xmath1694 , and is also stable from @xmath1694 to @xmath1828 , hence instability from @xmath1694 to @xmath1828 is a stronger statement that instability in @xmath1694 . in our framework ,
the notion of orbital stability is more relevant
. let us consider @xmath1830 a group and @xmath1831 a unitary representation of @xmath1832 on @xmath164 , extending @xmath1833 and leaving @xmath89 and @xmath1834 invariant .
then , @xmath1819 is said to be orbitally stable in @xmath1694 ( for the group @xmath1830 ) if for any @xmath733 , there exists @xmath160 such that any solution to ( @xmath1714 ) with initial datum @xmath1835 is global in time and remains in @xmath1836 for @xmath672 .
we may also define orbital instability from @xmath1694 to @xmath1826 in a natural way . in @xcite , a general framework for the stability analysis for the `` bound state '' has been given .
in particular , the nonlinear orbital instability in proved in @xcite through the construction of a liapounov type functional . however , this method does not give a clear understanding neither of how we get farther from the `` bound state '' , nor on which timescale it occurs . the need for allowing an additional group of invariances @xmath1830
can be seen in the case of bound state solutions , that is @xmath1837 to the nonlinear schrdinger equation i _
t + + f(||^2 ) = 0 , or the nonlinear klein - gordon equation in @xmath1838 ^2_t = + f(||^2 ) , since then , the invariance by translation in space must be taken into account in the definition of orbital stability , and we are in a case where @xmath1839 acts naturally by translation .
the translations are taken into account in @xcite . in @xcite and @xcite ,
the notion of orbital stability is for @xmath1830 trivial .
it is clear from the definition that orbital stability for @xmath1840 implies orbital stability for arbitrary @xmath1830 . for the instability in the nonlinear schrdinger equation or the nonlinear klein - gordon equation , @xcite and @xcite work with radial @xmath269 functions .
the fact that this also implies the orbital instability with the action of @xmath1839 by translations follows immediately from the fact that for any @xmath1841 $ ] the manifold @xmath1842 is orthogonal to @xmath1843 . for the stability analysis of a `` bound state '' @xmath1844 ,
it is natural to consider the linearization of ( @xmath1714 ) near @xmath1845 .
more precisely , we linearize according to the ansatz @xmath1846 , so that the `` bound state '' becomes stationary .
the linearized problem then becomes = j ( e ( ) - q ( ) ) v = v , where , @xmath1847 denoting the riesz isomorphism , @xmath1848 is skew - adjoint .
the purpose of this appendix is to give a general result , for hamiltonian equations , showing that linear instability implies nonlinear ( orbital ) instability . by linear instability , we mean that the complexification of [ @xmath1849@xmath1850 has at least one eigenvalue in the right - half space @xmath1074 .
the argument follows ideas from the works of f. rousset and n. tzvetkov @xcite . showing the existence of an unstable eigenvalue
can be done through various techniques : @xcite ( in the framework of @xcite when @xmath1851 is onto ) , @xcite ( assuming a special structure of the hamiltonian equation ) ; for uses of the vakhitov - kolokolov function , see @xcite , @xcite or @xcite .
when @xmath1851 is not onto , we quote @xcite .
for one dimensional partial differential equations , one may also use evans function ( see the survey @xcite ) as in @xcite , @xcite , @xcite , @xcite .
the paper @xcite proposes another approach which allows to treat pseudo - differential equations , such as the bbm equation , the benjamin - ono equation , regularized boussinesq equations , the intermediate long wave equation ... in order to pass from linear to nonlinear instability , the following result is standard .
we refer to the paper @xcite by d. henry , j. perez and w. wreszinski .
it can also be found in @xcite and @xcite . + * theorem b.1 * ( @xcite , @xcite , @xcite ) _ we assume that @xmath1142 generates a continuous semigroup on @xmath595 and that @xmath1852 meets the right - half space @xmath1074 .
we assume moreover that @xmath1853 is locally lipschitz continuous and verifies , for some @xmath1854 , @xmath1855 as @xmath1856 .
then , the solution @xmath1857 is unstable for the equation @xmath1858 . _
+ in @xcite , it is claimed that an orbital instability result can also be established .
theorem b.1 shows nonlinear instability without assuming that the equation is hamiltonian .
however , if ( @xmath1859 ) can be solved using a semigroup , it does not give the growth of its norm .
moreover , it does not say that if the initial datum is in a most unstable direction , that is an eigendirection of @xmath1142 corresponding to an eigenvalue of maximal positive real part ( plus the complex conjugate if necessary ) , then you can track the exponential growth of the solution .
in particular , it does not explain the mechanism of instability and does not give any information on the timescale on which you see the instability .
for instance , some strong instability results are shown by proving blow - up in finite time ( see @xcite ) , but the instability due to an exponentially growing mode holds on a much smaller timescale .
we wish to provide here some results clarifying the instability mechanism by tracking the exponentially growing mode .
when we want to prove a nonlinear instability result from a linear instability one , one needs some information on the growth of the semigroup @xmath1860 , when such a semigroup @xmath1861 exists , which we shall assume in this appendix .
the growth estimate on @xmath1861 relies classically on the following spectral mapping result due to j. prss @xcite , which generalizes the work of l. gearhart @xcite . +
* theorem b.2 * ( @xcite ) _ let @xmath595 be a complex hilbert space and @xmath1142 an unbounded operator on @xmath595 which generates a continuous semigroup @xmath1862 on @xmath595 . for @xmath1863 , we have _
@xmath1864 the following result is an immediate corollary . + * corollary b.1 * _ let @xmath595 be a complex hilbert space and @xmath1142 an unbounded operator on @xmath595 which generates a continuous semigroup @xmath1862 on @xmath595 .
assume that for any @xmath1865 , we have @xmath1866 and that there exists @xmath1867 such that @xmath1868 .
then , for any @xmath1863 , the spectral mapping holds : @xmath1869 .
_ + _ proof . _ since @xmath1868 , we have @xmath1870 .
if @xmath1871 does not have modulus one , then note that when @xmath1872 , the supremum for @xmath1873 in theorem b.2 can be @xmath1874 only when @xmath1875 , and we conclude with our hypothesis . + the fact that we exclude @xmath75 in the spectral mapping theorem just comes from the fact that we consider a semigroup and not a group .
however , in most hamiltonian pde s , we have time reversibility and we have actually a continuous group and not only a semigroup . in most cases , we work with @xmath1876 where @xmath316 is a real hilbert space , thus for applying theorem b.2 or corollary b.1 we have to consider , as usual , the complexified operator @xmath1877 defined by @xmath1878 .
it seems that the first time theorem b.2 is used to prove a growth estimate on a semigroup was by t. kapitula and b. sandstede in @xcite .
later , the work @xcite by f. gesztesy _
_ also uses this result for bound states for ( nls ) .
the bounds on the resolvent in @xcite were proved using the particular structure of the linearized operator . in @xcite ,
the computations are more involved and rely on suitable kernel estimates of some hilbert - schmidt operators .
the same type of estimates have also been used in @xcite .
the main objective of this appendix is to provide a generalization of these results to a wide class of hamiltonian equations .
indeed , the approaches in @xcite and @xcite seem specific to the problem .
in addition , it is not clear whether the computations in @xcite and @xcite can be extended to other types of equations . in particular , in @xcite and in the present paper , we have a situation similar to the one studied in @xcite , namely travelling wave solutions to a nonlinear schrdinger equation with nonzero condition at infinity , but for nonzero propagation speeds , the travelling wave is not real - valued ( as it is the case in @xcite for stationary waves or for bound state solutions ) , and the bloc diagonal structure of the linearized hamiltonian disappears .
an additional difficulty is that in @xcite and the present work , the limits of the travelling waves at @xmath1057 and @xmath149 differ .
the proof we give is based on ideas from @xcite and make very few spectral assumptions on @xmath1082 .
+ _ assumption _ ( a ) : the spectrum of @xmath1082 consists in a finite number ( possibly zero ) of nonpositive eigenvalues @xmath1879 , ... ,
@xmath1880 in @xmath1881 $ ] , each one with finite multiplicity , and the rest of the spectrum is positive and bounded away from @xmath75 . furthermore that for any @xmath1882 , we have @xmath1883 and @xmath1884 \subset d(\bl ) $ ]
. finally , there exists @xmath1867 such that @xmath1885 .
+ the first hypothesis on the location of the spectrum of @xmath1082 is quite weak , since it is verified when @xmath1082 is bounded from below and has essential spectrum positive and bounded away from zero . indeed , if @xmath160 is such that @xmath1886 , then the eigenvalues of @xmath1082 in @xmath1887 $ ] are isolated , of finite multiplicity , and are bounded from below by assumption .
the second hypothesis @xmath1888 is a regularity assumption on the eigenvectors .
let us recall that theorem [ atmost ] ensures that the number of eigenvalues ( with algebraic multiplicities ) of @xmath1889 in the right - half space @xmath1074 is less than or equal to the number of negative eigenvalues of @xmath1109 , hence is finite under assumption ( a ) .
let us now state our main result , the proof of which is given in section b.3.1 .
+ * theorem b.3 * _ we make assumption ( a ) and suppose that @xmath1860 generates a continuous semigroup .
then , for any @xmath1863 , the spectral mapping holds : @xmath1890_\c } ) \setminus \ { 0 \ } = \ex^ { t \s ( [ \bj \bl]_\c ) } $ ] .
furthermore , denoting @xmath1891_\c ) \cap \ { { \rm re } \geq 0 \ } \big\ } \in [ 0 , + \ii ) , \ ] ] for any @xmath1892 , there exists @xmath1893 such that , for any @xmath672 , we have @xmath1894 assume in addition @xmath1895 and denote @xmath1896_\c ) \ \text{s.t . } \ { \rm re } \ ,
\lambda = \gamma_0 \big\ } \in \n^ * .\ ] ] then , there exists @xmath1897 such that , for any @xmath672 , we have _ @xmath1898 in particular , theorem b.3 provides a very simple proof of the spectral mapping theorem used in @xcite and @xcite .
indeed , the self - adjoint operator @xmath1082 involved in these papers is block diagonal : @xmath1899 and both @xmath1900 , @xmath1901 have at most two nonnegative eigenvalues . more generally , if @xmath1900 and @xmath1901 are closed self - adjoint operators on @xmath1019 verifying assumption ( a ) and if @xmath1902 is a linear bounded operator which is compact with respect to @xmath1903 and @xmath1904 , then the self - adjoint operator @xmath1905 also verifies assumption ( a ) .
indeed , @xmath1082 is bounded from below ( since @xmath1906 is bounded ) and its essential spectrum is @xmath1907 for some positive @xmath611 , since @xmath1908 is compact with respect to @xmath1903 and @xmath1904 ) .
in @xcite ( section 7.1 there ) and @xcite ( see proposition 10 there ) , a spectral mapping theorem is used for such an operator . in @xcite ,
the specific algebra of the problem was used , and for @xcite , the proof relies on the arguments in @xcite , but here again , in both cases , we may use theorem b.3 to show the same result .
we start with a classical result for `` semilinear '' equations , proved in section b.3.3 . + * theorem b.4 * _ let @xmath595 be a real hilbert space , and consider an evolution equation of the form @xmath1909 where @xmath1910 is a locally lipschitz mapping satisfying @xmath1911 as @xmath1856 and @xmath1142 is a linear operator which generates a semigroup .
we assume that @xmath1912 has an unstable eigenvalue in the right half plane @xmath1074 and a finite number of eigenvalues in @xmath1074 .
we denote @xmath1913_\c ) \cap \ { { \rm re } > 0 \ } \big\ } \in ( 0 , + \ii)\ ] ] and fix @xmath1914 with @xmath1915 and an associated eigenvector @xmath1916 such that @xmath1917 .
assume furthermore that there exists @xmath1918 and @xmath1897 such that @xmath1919 then , @xmath75 is an unstable solution .
more precisely , there exist @xmath600 , @xmath1920 and @xmath1921 such that , for any @xmath298 , the solution @xmath1769 with initial datum @xmath1922 exists at least on @xmath1923 $ ] and verifies , for @xmath1924 , @xmath1925 in particular , for @xmath1926 , we see the instability for @xmath1927 .
if @xmath316 is a banach space containing @xmath1019 and with continuous imbedding @xmath1928 , the trivial solution @xmath1929 is also unstable from @xmath1019 to @xmath1930 . _
+ let us observe that it is always possible to choose the ( complex ) eigenvector @xmath1931 so that @xmath1932 since for any @xmath1933 , @xmath1934 is also an eigenvector .
the following corollary deals with the orbital instability . we recall
that under assumption ( a ) , @xmath1935_\c $ ] has a finite number of eigenvalues in @xmath1074 . +
* corollary b.2 * _ we make assumption ( a ) and suppose that @xmath1860 generates a continuous semigroup .
let @xmath1936 be a banach space containing @xmath164 and with continuous imbedding @xmath1827 .
assume moreover that @xmath1935_\c $ ] has at least one eigenvalue in @xmath1074 and choose @xmath1871 with @xmath1937_\c ) \cap \ { { \rm re } > 0 \ } \big\ } \in ( 0 , + \ii)\ ] ] and @xmath1916 an associated eigenvector such that @xmath1938 .
we assume moreover that @xmath1939 is a @xmath364 submanifold of @xmath164 .
we finally suppose that the equation ( @xmath1714 ) is semilinear in the sense that there exists @xmath1940 locally lipschitz continuous such that @xmath1941 as @xmath1856 and @xmath1942 + \phi ( v ) .\ ] ] then , there exists @xmath1248 , @xmath1920 and @xmath1921 , depending only on @xmath1943 and @xmath1944 , with the following properties .
for any @xmath298 , the solution @xmath1308 to ( @xmath1714 ) with initial datum @xmath1945 exists at least on @xmath1923 $ ] and verifies , for @xmath1924 , @xmath1946 in particular , the `` bound state '' solution @xmath1947 is nonlinearly orbitally unstable from @xmath1694 to @xmath1936 and , for @xmath1948 , we see the instability for @xmath1949 . _
+ in @xcite , a similar assertion is made for the orbital instability in the remark after theorem 2 there , but with @xmath1950 . for applications to pde s , the
space @xmath1694 may be a sobolev space @xmath1951 , and @xmath1952 a space like @xmath505 or @xmath374 for instance .
the framework of @xcite is the single energy space ( for instance @xmath375 ) , but an instability result established by tracking exponentially growing modes allows to prove instability from the regular space @xmath1694 ( @xmath375 ) to the nonregular space @xmath1936 ( @xmath505 or @xmath374 ) . here
, we may obtain instability in @xmath520 . + * remark b.1 * in the framework of @xcite , where a lyapounov type functional is used , it follows that the instability is seen for a time at most equal to @xmath1953 , where @xmath234 is some positive constant .
this timescale is much larger than the natural one @xmath1954 .
+ we may apply our result to the nonlinear schrdinger equation i _
t + + f(||^2 ) = 0 , or the nonlinear klein - gordon equation ^2_t = + f(||^2 ) in @xmath1838 .
we shall consider a nonlinearity @xmath13 at least @xmath734 , so that we are in the framework of @xcite .
+ @xmath84 a bound state solutions for these two equations is a particular solution of the form @xmath1837 .
the instability is in general linked to the fact that _ ^d |_|^2 dx < 0 ( _ ^d |_|^2 dx ) < 0 the existence of at least one unstable eigenvalue has been shown under assumption ( b.1 ) by @xcite for radial bound states with an arbitrary number of nodes and in @xcite for radial ground states .
corollary b.2 may be applied with @xmath1955 , where @xmath1956 , @xmath1957 and assuming that the nonlinearity verifies @xmath1958 , and @xmath1959 or @xmath1960 .
the result in @xcite shows the instability of linearly unstable bound states for ( nls ) ( in dimension @xmath1961 ) with @xmath1962 by showing the exponential growth of an unstable eigenmode .
our result gives a simple proof of this result , but restricted to the sufficiently smooth cases , namely @xmath1313 an odd integer or @xmath1963 .
for non smooth nonlinearities , the situation is more delicate ( see @xcite ) .
an alternative approach is to combine strichartz estimates with the growth estimate on the semigroup @xmath1964 given in theorem b.3 , as in @xcite .
+ @xmath84 corollary b.2 also applies to the discrete nonlinear schrdinger equation
n , i _ t _
n + ( _ n+1 - 2 _ n + _
n-1 ) + _ n f(|_n|^2 ) = 0 , as studied in @xcite with the saturated nonlinearity @xmath1965 , @xmath1966 ( existence of travelling wave solution ) and in @xcite ( defocusing cubic ( dnls ) , _ i.e. _ @xmath1967 for some @xmath1966 ) .
the numerical analysis in the paper @xcite shows the existence of linearly unstable bound state solutions .
the travelling wave solutions numerically obtained in @xcite are linearly stable , but it may happen that for other nonlinearities @xmath13 , some are linearly unstable . for quasilinear problems , we shall not make restrictions on the smoothness of the nonlinearity .
the result relies on the strategy of e. grenier @xcite and the works @xcite .
we consider the evolution equation = j ( l_0 u + f ( u ) ) for @xmath1968 , where @xmath1969 , with the following hypothesis .
the operator @xmath1851 is a fourier multiplier , skew - symmetric on @xmath505 , into and with domain containing @xmath269 .
there exists @xmath1970 such that the operator @xmath1971 is a fourier multiplier with domain containing @xmath1972 , symmetric and having a self - adjoint realization on @xmath1973 .
moreover , for some @xmath1974 , @xmath1975 verifies : @xmath1976 the framework proposed in @xcite was for @xmath1971 coercive in @xmath375 , that is @xmath1977 .
for the examples below , we shall have @xmath1978 or @xmath1979 , which requires very few modifications to the proof of @xcite .
we still assume that for some group @xmath1830 , there exists a unitary representation of @xmath1830 on @xmath164 , @xmath1980 , leaving the equation ( e ) invariant .
we consider a stationary solution of the evolution equation ( e ) , that is some @xmath1981 such that @xmath1982 .
we are interested in the stability of this solution .
we assume that the commutator @xmath1983 $ ] is bounded in @xmath520 , which is the case when @xmath1984 is bounded in @xmath520 or when @xmath1985 and @xmath701 .
we suppose that for the problem @xmath1986 where @xmath1987 is smooth , bounded as well as its derivatives and @xmath1988 for every @xmath1989 , we have local well - posedness for @xmath1990 large enough : there exists a time @xmath927 and a unique solution in @xmath1991 , h^s ) $ ] .
we moreover assume that for some continuous non - decreasing function @xmath1992 with @xmath1993 , the tame estimate @xmath1994 \ } , \p_x^\alpha v ) _
{ l^2 } | \leq \kappa ( \n w \n_{w^{s+1,\ii } } + \n v \n _ { h^s } )
\n v \n_{h^s}^2 , \ ] ] with @xmath1995 holds true . in order to control high order derivatives
, we finally require that for @xmath1990 large enough , there exists a self - adjoint operator @xmath1996 and @xmath1997 such that @xmath1998 and @xmath1999 ( for a criterion which ensures the existence of such a multiplier , see lemma 5.1 in @xcite ) .
+ adapting the strategy of @xcite , we may deduce the following result . since the proof is very similar ,
we omit it . + * theorem b.5 * _ we make the above assumptions and moreover that @xmath2000 verifies hypothesis ( a ) in @xmath520 .
we assume furthermore that @xmath2001_\c $ ] has an unstable eigenvalue in the right half - plane @xmath1074 , denote @xmath2002_\c ) \cap \ { { \rm re } > 0 \ } \big\ } \in ( 0 , + \ii)\ ] ] and fix @xmath2003_\c ) $ ] with @xmath1915 and an associated eigenvector @xmath2004_\c ) $ ] such that @xmath2005 . there exists @xmath2006 such that , if @xmath2007 , @xmath1834 is nonlinearly unstable from @xmath2008 to @xmath520 and to @xmath374 : there exists @xmath600 , @xmath1920 and @xmath1921 such that , for any @xmath298 , the @xmath1951 solution @xmath1308 to ( e ) with initial datum @xmath2009 exists at least on @xmath1923 $ ] and verifies , for @xmath1924 , @xmath2010 hence @xmath2011 if , in addition , @xmath2012 is a @xmath364 submanifold of @xmath2008 , then we also have @xmath2013 in particular , for @xmath1948 , we see the nonlinear orbital instability for @xmath2014 .
_ some model quasilinear equations are given by wave equations ( in one space dimension ) such as the generalized korteweg - de vries equation _ t u + _
x(f(u ) ) + _ x^3 u = 0 , the generalized regularized korteweg - de vries equation , also called benjamin - bona - mahony equation or peregrine equation when @xmath2015 , _ t
u + _ x u + _
x(f(u ) ) - _ t _
x^2 u = 0 , the generalized regularized boussinesq equation _ t^2 u - ^2_x u - ^2_x(f(u ) ) - _ t^2 _ x^2 u = 0 . each of these equations admit nontrivial solitary wave solutions @xmath2016 for @xmath51 in @xmath2017 , @xmath2018 and @xmath2019 respectively .
for these solitary wave solutions , the momentum is , respectively , @xmath2020 the existence of exactly one unstable eigenvalue has been shown with the use of an evans function by r. pego and m. weinstein in @xcite for these three equations under the condition @xmath2021 .
the paper @xcite by o. lopes also gives a linear instability result .
equations ( gbbm ) and ( grbsq ) turn out to be semilinear due to the regularization effect . indeed
, they may be written @xmath2022 therefore , corollary b.2 applies to these two models and this shows the nonlinear instability when linear instability holds . in @xcite , some generalizations of the equations ( gkdv ) , ( gbbm ) and ( gbbm ) have been proposed that take into account pseudodifferential operators .
these are respectively _ t u + _
x(f(u ) ) - _ x u = 0 , _
t u + _ x u + _
x(f(u ) ) + _ t u = 0 and _ t^2 u - ^2_x u - ^2_x(f(u ) ) + _ t^2 u = 0 . here
, @xmath2023 is a fourier multiplier of symbol @xmath2024 : @xmath2025 ( here , @xmath2026 denotes fourier transform ) .
we assume @xmath2027 ( otherwise , see @xcite ) . when @xmath2028 , these equations reduce to ( gkdv ) , ( gbbm ) and ( gbbm ) respectively .
the benjamin - ono equation ( @xmath2029 ) , the smith equation ( @xmath2030 ) and the intermediate long wave ( or whitham ) equation ( @xmath2031 , for some constant @xmath2032 ) are common models of dispersive wave equations that are of type ( i ) .
we refer to @xcite for references on these models and the existence of solitary waves .
the associated momentum is @xmath2033 for these models , evans function type arguments do not work since we no longer have a differential equation ( it is nonlocal ) . the paper @xcite by z. lin proposes another approach than the evans function technique for establishing the existence of unstable eigenvalues .
however , it is not completely clear whether this method extends easily to the case of systems such as the euler - korteweg system ( ek ) ( given at the beginning of section [ sexins ] ) .
+ * theorem b.6 ( @xcite ) * _ we consider one of the equations ( i ) , ( ii ) or ( iii ) with @xmath7 of class @xmath734 satisfying @xmath2034 and @xmath2035 for @xmath2036 .
we assume moreover that @xmath2024 is even , nonnegative , and verifies , for some @xmath2037 , @xmath2038 .
assume that @xmath2039 is a @xmath364 branch of travelling wave solution to ( i ) , ( ii ) or ( iii ) with @xmath2040 defined near @xmath83 and suppose that the linearized operator @xmath1082 has exactly one negative eigenvalue , that @xmath2041 is spanned by @xmath2042 and that @xmath2043 .
then , @xmath172 is linearly unstable . _ + it is not easy to determine whether the hypothesis of theorem b.6 hold true when @xmath2023 is not a ( differential ) sturm - liouville operator .
see however @xcite on this question .
it is clear that if the assumptions of theorem b.6 are verified , then assumption ( a ) is also satisfied . as for the ( gbbm ) and the ( grbsq ) equations , the equations ( ii ) and ( ii ) turn out to be semilinear , thus we may prove nonlinear orbital instability by applying corollary b.2 . the kawahara equation ( or fifth order ( kdv ) equation ) _ t u + _
x(f(u ) ) + _ x^3 u + _
x^5 u = 0 , with @xmath736 , @xmath2044 two real constants , is another relevant dispersive model . for this equation
, it may happen that the linearized equation around the solitary wave has more than one negative eigenvalues , in which case the works @xcite do not give a clear necessary and sufficient condition for stability .
the paper @xcite by t. bridges and g. derks gives a sufficient condition for linear instability for solitary wave solutions , but also for other types of travelling solutions .
this condition is probably not necessary since it may happen that there exist at least two unstable eigenvalues , or two complex conjugate eigenvalues . instead of stating a general result for nonlinear orbital instability
, we shall consider several model cases on which we will verify the hypothesis of theorem b.5 , in particular the question of the existence of the multiplier @xmath1996 .
+ * proposition b.1 * _ we consider the equation ( i ) , namely @xmath2045 with @xmath7 of class @xmath734 satisfying @xmath2034 and @xmath2046 for @xmath2036 .
we assume that @xmath2024 is one of the following functions : @xmath2047 @xmath2048 there exists @xmath2049 such that , if there exists @xmath2050 such that ( i ) has a nontrivial solitary wave @xmath2051 which is linearly unstable , then for any @xmath2007 , it is also nonlinearly unstable from @xmath2008 to @xmath2008 , to @xmath520 and to @xmath2052 . _ + by application of theorem b.5 , we are thus able to show the nonlinear instability from @xmath1951 to @xmath505 or @xmath374 by tracking the exponentially growing mode ( this question was left open in @xcite and also in @xcite ) .
in particular , we obtain the @xmath505 nonlinear instability of the linearly unstable solitary waves for these models .
+ _ proof .
_ all the assumptions for theorem b.5 for these types of models are verified in section 8.1 in @xcite , except the existence of the multiplier @xmath1996 . for ( kdv ) ,
where @xmath1977 , we shall take ( for @xmath2053 integer ) @xmath2054 as the computations from @xcite , section 8.1 show .
for the kawahara equation , with @xmath1979 , we take ( for @xmath2055 integer ) @xmath2056 and since the computations are very similar , we omit them . for the benjamin - ono equation , we have @xmath2057 and @xmath1978 , and we will then have to deal with pseudo - differential operator which are fourier multipliers with homogeneous symbol . for this type of operator , we shall need some commutator estimates .
we denote by @xmath2058 or @xmath2059 the fourier transform of @xmath1931 , and @xmath2060 the fourier multiplier with symbol @xmath2061 ( this is the hilbert transform ) . + * lemma b.1 * _ ( i ) let @xmath2062 with @xmath2063 ( for instance , @xmath2064 for some @xmath2065 )
. then , there exists @xmath1974 such that , for any @xmath2066 , @xmath2067 ( ii ) let @xmath2068 with @xmath2069 ( for instance , @xmath2070 for some @xmath2071 ) .
then , there exists @xmath1974 such that , for any @xmath2072 , @xmath2073 \bm^{\frac12 } \mathscr{h } v \no_{l^2(\r ) } \leq & \ , c \n v \n_{l^{2}(\r ) } .\end{aligned}\ ] ] ( iii ) let @xmath2068 with @xmath2074 ( for instance , @xmath2070 for some @xmath2071 ) .
then , there exists @xmath1974 such that , for any @xmath2075 , @xmath2076 \bm^{\frac12 } v \no_{l^2(\r ) } \leq & \ , c \n v \n_{l^{2}(\r ) } .\end{aligned}\ ] ] _ + _ proof .
_ we have @xmath2077 using the inequality @xmath2078 , we thus obtain @xmath2079 .
this argument does not provide the sharpest bound in @xmath2080 , since it involves @xmath2081 , whereas the use of paradifferential calculus will use only @xmath2082 .
however , we shall to use this refinement here .
the starting point for the second inequality is @xmath2083 to prove this , note that by homogeneity @xmath2084 , @xmath197 , it suffices to prove @xmath2085 which is easy .
therefore , @xmath2086 \bm^{\frac12 } \mathscr{h } v \big ) ( \xi ) \big| \\ & = \big| \int_\r | \xi |^{\frac32 } \hat{h } ( \xi - \zeta ) \hat{v}(\zeta ) \ d \zeta - \int_\r | \zeta |^{\frac32 } \hat{h } ( \xi - \zeta ) \hat{v}(\zeta ) \
d \zeta - \int_\r \frac32 | \zeta |^{\frac12 } { \rm sgn}(\zeta ) ( \xi - \zeta ) \hat{h } ( \xi - \zeta ) \hat{v}(\zeta ) \ d \zeta \big| \\ & \leq c \int_\r | \xi - \zeta |^{\frac32 } | \hat{h } ( \xi - \zeta ) | \cdot | \hat{v}(\zeta ) | \ d \zeta \\ & = c | \mathscr{f } ( \bm^{\frac32 } h ) | * | \hat{v } | , \end{aligned}\ ] ] and we conclude as before . for the third inequality , we argue in a similar way with the estimate @xmath2087 the proof is complete .
+ for the benjamin - ono equation , @xmath2088 , @xmath2089 and the index @xmath1990 will be half an integer : @xmath2090 .
therefore , we set @xmath2091 + \ { s \ } $ ] , with @xmath2092 $ ] integer and @xmath2093 .
let us define , for @xmath2090 , @xmath2094 , @xmath2095 } \p_x^{2[s ] } \bm + \gamma_s \p_x^{[s ] } \ { f'(q ) \p_x^{[s ] } \cdot \ } } & \quad { \rm if \ } \ { s \ } = \frac12 , \end{array}\right.\ ] ] for some real constant @xmath2096 to be determined later .
it is clear that @xmath1996 is self - adjoint on @xmath505 and that there exists @xmath2097 such that @xmath2098 in order to verify the assumptions for the multiplier @xmath2099 , it remains to study @xmath2100 .
when @xmath2101 , _ i.e. _
@xmath2102 , this quantity is @xmath2103 , ( -1)^s \p_x^{2s } u ) _ { l^2 } + \gamma_s { \rm re } ( \p_x [ f'(q ) u ] , \bm^{\frac12 } \p_x^{s-1 } \ { f'(q ) \p_x^{s-1 } \bm^{\frac12 } u \ } ) _ { l^2 } \nonumber \\ \tag{b.1 } & \quad + c { \rm re } ( \p_x u , \mathbb{m}_s u ) _ { l^2 } .\end{aligned}\ ] ] by skew - adjointness , the first and last scalar products are zero . by integration by parts and leibniz formula ,
we deduce , since @xmath2104 , @xmath2105 , ( -1)^s \p_x^{2s } u ) _ { l^2 }
= & \ , { \rm re } ( \p_x^{s+1 } [ f'(q ) u ] , \p_x^{s } u ) _ { l^2 } \\
\leq & \ , { \rm re } ( f'(q ) \p_x^{s+1 } u , \p_x^{s } u ) _ { l^2 } + ( s + 1 ) { \rm re } ( \p_x [ f'(q ) ] \p_x^{s } u , \p_x^{s } u ) _ { l^2 } + c_s \n u \n_{h^s } \n u \n_{h^{s-1 } } \\
\leq & \ , \big ( s + \frac12 \big ) { \rm re } ( \p_x [ f'(q ) ] \p_x^{s } u , \p_x^{s } u ) _ { l^2 } + c_s \n u \n_{h^s } \n u \n_{h^{s-1 } } .\end{aligned}\ ] ] similarly , using the easy estimates @xmath2106 and @xmath2107 for @xmath2108 with @xmath2109 ( this is an immediate consequence of lemma b.1 ) @xmath2110 , & \ , \bm^{\frac12 } \p_x^{s-1 } \ { f'(q ) \p_x^{s-1 } \bm^{\frac12 } u \ } ) _ { l^2 } = \gamma_s ( -1)^{s-1 } { \rm re } ( \bm^{\frac12 } \p_x^{s } [ f'(q ) u ] , f'(q ) \p_x^{s-1 } \bm^{\frac12 } u ) _ { l^2 } \\ \leq & \ , \gamma_s ( -1)^{s-1 } { \rm re } ( \bm^{\frac12 } [ f'(q ) \p_x^s u ] , f'(q ) \p_x^{s-1 } \bm^{\frac12 } u ) _ { l^2 } +
c \n u \n_{h^{s-\frac12 } } ^2 .\end{aligned}\ ] ] using lemma b.1 , we deduce @xmath2111 - f'(q ) \bm^{\frac12 } \p_x^s u \n_{l^2 } \leq c ( q ) \n u \n _ { h^s } $ ] , thus @xmath2110 , & \ , \bm^{\frac12 } \p_x^{s-1 } \ { f'(q ) \p_x^{s-1 } \bm^{\frac12 } u \ } ) _ { l^2 } \\ \leq & \ , \gamma_s ( -1)^{s-1 } { \rm re } ( f'(q ) \p_x^s \bm^{\frac12 } u , f'(q ) \p_x^{s-1 } \bm^{\frac12 } u ) _ { l^2 }
+ c \n u \n_{h^s } \n u \n_{h^{s-\frac12 } } \\ = & \ , \frac { \gamma_s}{2 } ( -1)^{s } { \rm re } ( \p_x [ f'(q ) ] \p_x^{s-1 } \bm^{\frac12 } u , f'(q ) \p_x^{s-1 } \bm^{\frac12 } u ) _ { l^2 } + c \n u \n_{h^s } \n u \n_{h^{s-\frac12 } } \\ \leq & \ , c \n u \n_{h^{s-\frac12 } } ^2 + c \n u \n_{h^s } \n u \n_{h^{s-\frac12 } } \leq c \n
u \n_{h^s } \n u \n_{h^{s-\frac12 } } .\end{aligned}\ ] ] we now turn to the term @xmath2112 using lemma b.1 , we write @xmath2113 \bm^{\frac12 } \mathscr{h } \ { \p_x^{s-1 } \bm^{\frac12 } u \ } \no_{l^2 } \leq & \ , c ( q ) \n \p_x^{s-1 } \bm^{\frac12 } u \n_{l^{2 } } \\
\leq & \ , c ( q ) \n u \n_{h^{s-\frac12 } } , \end{aligned}\ ] ] which implies @xmath2114 \bm^{\frac12 } \mathscr{h } \ { \p_x^{s-1 } \bm^{\frac12 } u \ } ) _ { l^2 } \\ & \quad \quad + c \n u \n_{h^s } \n u \n_{h^{s-\frac12 } } . \end{aligned}\ ] ] noticing that @xmath2115 and @xmath2116 ( since @xmath2117 has symbol equal to @xmath2118 ) , we infer @xmath2119 \p_x^{s } u ) _ { l^2 } % \\ & \quad \quad + c \n u \n_{h^s } \n u \n_{h^{s-\frac12 } } \\ & = \gamma_s ( -1)^{s } { \rm re } ( \p_x^s u , \p_x [ f'(q ) ] \p_x^{s } u ) _ { l^2 } + c \n u \n_{h^s } \n u \n_{h^{s-\frac12 } } \end{aligned}\ ] ] by integration by parts . reporting these estimates into ( b.1 ) , we infer @xmath2120 \p_x^{s } u , \p_x^{s } u ) _ { l^2 } + \gamma_s ( -1)^{s } { \rm re } ( \p_x^s u , \p_x [ f'(q ) ] \p_x^{s } u ) _
{ l^2 } \\ & + c \n u \n_{h^s } \n u \n_{h^{s-\frac12 } } .\end{aligned}\ ] ] therefore , the choice @xmath2121 provides the desired control @xmath2122 when @xmath2123 , the computations are similar : ( b.1 ) becomes now @xmath2124 } \p_x^{2[s ] } \bm u ) _ { l^2 } + \gamma_s { \rm re } ( \p_x \bm u , \p_x^{[s ] } \ { f'(q ) \p_x^ { [ s ] } u \ } ) _ { l^2 } \nonumber \\
& \quad + { \rm re } ( \p_x [ f'(q ) u ] , ( -1)^{[s ] } \p_x^{2[s ] } \bm u ) _ { l^2 } + \gamma_s { \rm re } ( \p_x [ f'(q ) u ] , \p_x^{[s ] } \ { f'(q ) \p_x^ { [ s ] } u \ } ) _ { l^2 } \nonumber \\
\tag{b.2 } & \quad + c { \rm re } ( \p_x u , \mathbb{m}_s u ) _ { l^2 } , \end{aligned}\ ] ] and the first and last scalar product still vanish .
moreover , by integration by parts and leibniz formula , we deduce , since @xmath2104 , @xmath2110 , & \ , \p_x^{[s ] } \ { f'(q ) \p_x^{[s]}u \ } ) _ { l^2 } = \gamma_s ( -1)^{[s ] } { \rm re } ( \p_x^ { [ s ] + 1 } [ f'(q ) u ] , f'(q ) \p_x^{[s ] } u ) _ { l^2 } \\ \leq & \ , \gamma_s ( -1)^ { [ s ] } { \rm re } ( f'(q ) \p_x^{[s ] + 1 } u , f'(q ) \p_x^{[s]}u ) _ { l^2 } + c \n u \n_{h^ { [ s ] } } ^2 \\ \leq & \ , \gamma_s ( -1)^ { [ s ] -1 } { \rm re } ( \p_x [ f'(q ) ] \p_x^{[s ] } u , f'(q ) \p_x^{[s ] } u ) _ { l^2 } + c \n u \n_{h^ { [ s ] } } ^2 \\ \leq & \ , c \n u \n_{h^ { [ s ] } } ^2 = c \n u \n_{h^ { s - \frac12 } } ^2 .\end{aligned}\ ] ] furthermore , @xmath2105 , ( -1)^{[s ] } \p_x^{2[s ] } \bm u ) _ { l^2 }
= & \ , { \rm re } ( \bm^{\frac12 } \p_x^{[s]+1 } [ f'(q ) u ] , \p_x^{[s ] } \bm^{\frac12 } u ) _ { l^2 } \\
\leq & \ , { \rm re } ( \p_x \bm^{\frac12 } \ { f'(q ) \p_x^{[s ] } u \ } , \p_x^{[s ] } \bm^{\frac12 } u ) _ { l^2 } \\ & \ , + [ s ] { \rm re } ( \p_x \bm^{\frac12 } \ { \p_x [ f'(q ) ] \p_x^{[s]-1 } u \ } , \p_x^{[s ] } \bm^{\frac12 } u ) _ { l^2 } + c \n u \n_{h^ { [ s ] - \frac12 } } \n u \n_{h^ { [ s ] + \frac12 } } . \end{aligned}\ ] ] for the second scalar product , we write , by lemma b.1 , @xmath2125 \p_x^{[s]-1 } u \ } , & \ , \p_x^{[s ] } \bm^{\frac12 } u ) _ { l^2 } \\
= & \ , { \rm re } ( \bm^{\frac12 } \ { \p_x^2 [ f'(q ) ] \p_x^{[s]-1 } u \ } , \p_x^{[s ] } \bm^{\frac12 } u ) _ { l^2 } + { \rm re } ( \bm^{\frac12 } \ { \p_x [ f'(q ) ] \p_x^{[s ] } u \ } , \p_x^{[s ] } \bm^{\frac12 } u ) _ { l^2 } \\ \leq & \ , c \n u \n_{h^ { [ s ] - \frac12 } } \n u \n_{h^ { [ s ] + \frac12 } } + { \rm re } ( \p_x [ f'(q ) ] \p_x^{[s ] } \bm^{\frac12 } u , \p_x^{[s ] } \bm^{\frac12 } u ) _ { l^2 } + c \n u \n_{h^ { [ s ] } } \n u \n_{h^ { [ s ] + \frac12 } } \\ \leq & \ , { \rm re } ( \p_x [ f'(q ) ] \p_x^{[s ] } \bm^{\frac12 } u , \p_x^{[s ] } \bm^{\frac12 } u ) _ { l^2 } + c \n u \n_{h^ { s- \frac12 } } \n u \n_{h^ { s } } .\end{aligned}\ ] ] for the first scalar product , we use lemma b.1 once again : @xmath2126 } u \ } , \p_x^{[s ] } \bm^{\frac12 } u ) _ { l^2 } \\ \leq & \ , { \rm re } ( f'(q ) \p_x \bm^{\frac12 } \p_x^{[s ] } u , \p_x^{[s ] } \bm^{\frac12 } u ) _ { l^2 } + \frac32 { \rm re } ( \p_x [ f'(q ) ] \p_x^{[s ] } \bm^{\frac12 } u , \p_x^{[s ] } \bm^{\frac12 } u ) _ { l^2 } + c \n u \n_{h^ { s- \frac12 } } \n u \n_{h^ { s } } \\ \leq & \ , { \rm re } ( \p_x [ f'(q ) ] \p_x^{[s ] } \bm^{\frac12 } u , \p_x^{[s ] } \bm^{\frac12 } u )
_ { l^2 } + c \n u \n_{h^ { s- \frac12 } } \n u \n_{h^ { s } } .\end{aligned}\ ] ] as a consequence , since @xmath2092 = s - \frac12 $ ] , @xmath2127 , ( -1)^{[s ] } \p_x^{2[s ] } \bm u ) _ { l^2 } \leq \big ( s + \frac12 \big ) { \rm re } ( \p_x [ f'(q ) ] \p_x^{[s ] } \bm^{\frac12 } u , \p_x^{[s ] } \bm^{\frac12 } u ) _ { l^2 } + c \n u \n_{h^ { s- \frac12 } } \n u \n_{h^ { s } } .\ ] ] we turn finally to the term @xmath2128 } \ { f'(q ) \p_x^ { [ s ] } u \ } ) _ { l^2 } = \gamma_s ( -1)^{[s ] } { \rm re } ( \p_x^{[s ] } \bm^{\frac12 } u , \p_x \bm^{\frac12 } \ { f'(q ) \p_x^ { [ s ] } u \ } ) _ { l^2 } , \ ] ] and infer , by lemma b.1 , @xmath2129 } \ { f'(q ) \p_x^ { [ s ] } u \ } ) _ { l^2 } \leq & \ , \gamma_s ( -1)^{[s ] } { \rm re } ( \p_x^{[s ] } \bm^{\frac12 } u , f'(q ) \p_x \bm^{\frac12 } \p_x^ { [ s ] } u ) _ { l^2 } \\ & \quad + \frac32 \gamma_s ( -1)^{[s ] } { \rm re } ( \p_x^{[s ] } \bm^{\frac12 } u , \p_x [ f'(q ) ] \bm^{\frac12 } \p_x^ { [ s ] } u ) _ { l^2 } + c \n u \n_{h^ { s- \frac12 } } \n u \n_{h^ { s } } \\ = & \ , \gamma_s ( -1)^{[s ] } { \rm re } ( \p_x^{[s ] } \bm^{\frac12 } u , \p_x [ f'(q ) ] \bm^{\frac12 } \p_x^ { [ s ] } u ) _ { l^2 } + c \n u \n_{h^ { s- \frac12 } } \n u \n_{h^ { s } } .\end{aligned}\ ] ] therefore , @xmath2130 } \big ) { \rm re } ( \p_x^{[s ] } \bm^{\frac12 } u , \p_x [ f'(q ) ] \bm^{\frac12 } \p_x^ { [ s ] } u ) _ { l^2 } +
u \n_{h^ { s- \frac12 } } \n u \n_{h^ { s } } , \ ] ] hence choosing @xmath2131- 1 } ( s + \frac12 ) $ ] gives the result .
it remains to study the cases of the intermediate long wave equation and the smith equation , for which @xmath2132 is , respectively , @xmath2133 we denote @xmath2134 the operator with symbol @xmath2135 ( the one of the benjamin - ono equation ) , and define @xmath1996 as for the benjamin - ono case ( hence with `` @xmath2023''@xmath2136 ) .
we observe that in both cases , @xmath2137 is bounded on @xmath520 . indeed ,
its symbol is continuous in @xmath191 and , for @xmath2138 , @xmath2139 and @xmath2140 respectively . in the quantity @xmath2141
, we then have to bound from above the extra term @xmath2142 , that is ( using the skew - adjointness for the higher order derivatives in @xmath1996 ) , @xmath2143 @xmath2144 } \ { f'(q ) \p_x^{[s ] } u \ } ) _ { l^2 } = \gamma_s ( -1)^{[s ] } { \rm re } ( \p_x^{[s]+1 } ( \tilde{\bm } u ) , f'(q ) \p_x^{[s ] } u ) _ { l^2 } \quad \quad { \rm if \ } \ { s \ } = \frac12 .\ ] ] we then note that in both cases , one may actually split @xmath2145 , where @xmath2146 is the multiplication by @xmath2147 ( resp .
@xmath2148 ) and @xmath2149 has a symbol which is continuous in @xmath191 and @xmath2150 at infinity , so that @xmath2149 is bounded from @xmath2151 to @xmath2152 if @xmath2153 . therefore , when @xmath2101 , we easily get @xmath2154 \p_x^{s-1 } \bm_0^{\frac12 } u ) _ { l^2 } + c \n u \n_{h^{s - \frac12 } } ^2 \leq
c \n u \n_{h^{s - \frac12 } } ^2 \end{aligned}\ ] ] and similarly when @xmath2155 . therefore , the estimate @xmath2156 remains true for the intermediate long wave equation and the smith equation .
the proof of proposition b.1 is thus complete by applying theorem b.6 .
we shall prove the resolvent estimate required in corollary b.1 .
let us consider @xmath2157 with @xmath2158 and the resolvent equation @xmath2159 , or ( + i ) v = ( v ) - . by hypothesis ,
the essential spectrum of @xmath1860 is of the form @xmath2160 $ ] .
moreover , we have seen that @xmath1860 has a finite number of eigenvalues in the half - space @xmath1074 , hence , for @xmath2161 sufficiently large , we know that there exists a unique solution @xmath1769 to ( b.3 ) . by taking the scalar product with @xmath2162 , we deduce the conservation law ( v , ( v ) ) _ = - re ( , ( v ) ) _ . by our assumption , there exists a finite ( possibly empty ) number of eigenvalues in @xmath2163 $ ] , @xmath2164 , each one of finite multiplicity . for any @xmath1882 , we fix an orthonormal basis @xmath2165 of the eigenspace @xmath2166 . by assumption ( a ) , any eigenvector @xmath2167 is smooth in the sense that @xmath2168 and @xmath2169 .
we then make a spectral orthogonal decomposition @xmath2170 where @xmath2171 and @xmath2172 for some positive @xmath611 . in the double sum
, we have a finite number ( independent of @xmath1769 ) of terms .
inserting this into ( b.4 ) yields @xmath2173 \leq k |\!| \sigma |\!|_\bx |\!| v |\!|_\bx + k \sum_{k,\ell } |\alpha_{k,\ell } |^2\ ] ] using the inequality @xmath2174 with @xmath2175 , @xmath2176 and @xmath2177 , the equality @xmath2178 and incorporating the term @xmath2179 in the left - hand side , we infer || v_+ ||_^2 k _ k , |_k , |^2 + k
|| ||_^2 . on the other hand , since @xmath2168 and @xmath2169 by assumption ( a ) , taking the scalar product of ( b.1 ) with @xmath2167 provides @xmath2180 consequently , @xmath2181 thus @xmath2182 which implies , if @xmath2183 , @xmath2184 reporting this into ( b.5 ) gives @xmath2185 if @xmath2186 , we deduce @xmath2187 and it follows that @xmath2188 where @xmath1554 does not depend on @xmath2189 ( large enough ) as wished .
* proof of the semigroup estimate when @xmath2190 . * here , we assume @xmath714 . as a consequence ,
the spectrum of @xmath2191_\c $ ] is of the form @xmath2192 , where @xmath2193_\c ) \subset \s_{\rm s } \subset \ { { \rm re } \leq 0 \ } $ ] and @xmath2194 consists in a finite number of eigenvalues of finite algebraic multiplicities .
therefore , we may define ( see , _ e.g. _
@xcite , @xcite ) the spectral riesz projection @xmath2195_\c - z ) ^{-1 } \ dz , \ ] ] where @xmath2196 is any simple ( positively oriented ) closed curve enclosing @xmath2197 . as a consequence ,
@xmath2198 is bounded , commutes with @xmath2199_\c $ ] on @xmath2200_\c ) $ ] and verifies @xmath2201_\c \mathbb{p } ) = \s_{\rm u } $ ] , @xmath2202_\c ( { \rm i d } - \mathbb{p } ) ) = \s_{\rm s } $ ] .
moreover , @xmath2191_\c \mathbb{p } $ ] is bounded , hence generates a continuous semigroup , @xmath2203_\c \mathbb{p } } $ ] , given by the exponential series @xmath2204_\c \mathbb{p } } = \sum_{n=0}^{+\ii } \frac{t^n ( [ \bj \bl]_\c \mathbb{p})^n}{n ! } .\ ] ] in addition , @xmath2191_\c ( { \rm i d } - \mathbb{p } ) = [ \bj \bl]_\c - [ \bj \bl]_\c \mathbb{p } $ ] also generates a continuous semigroup and that we have @xmath2205_\c } = \ex^{t \bj \bl \mathbb{p } } \ex^{t [ \bj \bl]_\c ( { \rm i d } - \mathbb{p } ) } $ ] . the semigroup generated by the bounded operator @xmath2191_\c \mathbb{p } $ ] is easily analyzed .
we shall now apply the spectral mapping theorem of j. prss ( theorem b.2 ) to @xmath2191_\c ( { \rm i d } - \mathbb{p } ) $ ] in order to control the growth of its norm . by corollary b.1
, it suffices to estimate its resolvent @xmath2206_\c ( { \rm i d } - \mathbb{p } ) - ( \gamma + i \tau ) ] ^{-1 } $ ] for large @xmath2207 ( note that @xmath2201_\c ( { \rm i d } - \mathbb{p } ) ) = \s_{\rm s } \subset \ { { \rm re } \leq 0 \ } $ ] .
if @xmath2208 and @xmath2209 is large , it is clear that the solution @xmath2210 to @xmath2206_\c ( { \rm i d } - \mathbb{p } ) - ( \gamma + i \tau ) ] u = \sigma $ ] is given by @xmath2211_\c - ( \gamma + i \tau ) ] ^{-1 } ( { \rm i d } - \mathbb{p } ) \sigma - \frac{1 } { \gamma + i \tau } \mathbb{p } \sigma , \ ] ] thus , for @xmath2209 large , @xmath2212_\c ( { \rm i d } - \mathbb{p } ) -(\gamma + i \tau ) ] ^{-1 } \no_{\mathscr{l}_c(\bx_\c ) } \leq \no [ [ \bj \bl]_\c - ( \gamma + i \tau ) ] ^{-1 } \no_{\mathscr{l}_c(\bx_\c ) } \n { \rm i d } - \mathbb{p } \n_{\mathscr{l}_c(\bx ) } + \frac{1 } { |\gamma + i \tau | } \n \mathbb{p } \n_{\mathscr{l}_c(\bx_\c)}\ ] ] is bounded .
consequently , by theorem b.2 and since @xmath2213 , @xmath2214_\c ( { \rm i d } - \mathbb{p } ) } ) = \ex^{t \s([\bj \bl]_\c ( { \rm i d } - \mathbb{p } ) ) } = \ex^{t \s_{\rm s } } \subset \bar{d}(0,1 ) $ ] .
it follows that for any @xmath159 , there exists @xmath2215 such that @xmath2216_\c ( { \rm i d } - \mathbb{p } ) } \n_{\mathscr{l}_c(\bx_\c ) } \leq k_\epsilon \ex^ { \epsilon t } .\ ] ] since @xmath2203_\c \mathbb{p } } $ ] is given by the exponential series , we also have the optimal estimate @xmath2216_\c \mathbb{p } } \n_{\mathscr{l}_c(\bx_\c ) } \leq k_0 ( 1 + t)^{m-1 } \ex^ { \gamma_0 t } \ ] ] by definition of @xmath2217 .
we conclude by taking @xmath2218 for instance .
since @xmath1147 generates a continuous semigroup , @xmath1769 is a solution to @xmath2219 if and only if it is a mild solution : @xmath2220 there exists @xmath16 such that @xmath2221 if @xmath2222 .
we choose @xmath2223 , where @xmath2224 and @xmath2225 is an eigenvector for the eigenvalue @xmath1685 , and write the solution under the form @xmath2226 .
if @xmath2227 , we can choose @xmath2228 .
then , @xmath2229 let us denote @xmath2230 and let @xmath2231 be the maximal time such that @xmath2232 and @xmath2233 in @xmath2234 , where @xmath2235 will be determined later .
we shall work for @xmath2236 , so that @xmath2237 .
then , @xmath2238 since @xmath2239 .
applying now the gronwall inequality to @xmath2240 then gives , since @xmath2241 , @xmath2242 \d^2 \ex^ { 2 t \gamma_0 } = k \d^2 \ex^ { 2 t \gamma_0 } .\ ] ] we now choose @xmath2243 , so that the right - hand side is @xmath2244 , and this implies that @xmath1308 exists at least on @xmath2245 $ ] .
in addition , for @xmath2246 , @xmath2247 as desired .
we conclude choosing @xmath297 so small that @xmath2248 .
we pick some @xmath2249 ( for instance @xmath2250 ) in order to have the semigroup estimate required in theorem b.4 .
the solution @xmath2251 verifies , for @xmath2252 , @xmath2253 hence , @xmath2254 remains at distance @xmath2255 from @xmath2256 and therefore @xmath2257 assume @xmath2227 .
then , we observe that the straight line @xmath2258 is transverse to the tangent space t@xmath2259 of the manifold @xmath2260 , since @xmath2225 is an eigenvector of @xmath1860 for @xmath2261 , hence does not belong to the kernel of @xmath1082 .
therefore , @xmath2262 for small @xmath2263 .
thus , @xmath2264 if @xmath2265 , the equation @xmath2191_\c ( w ) = \lambda w $ ] splits as @xmath2266 and @xmath2267 .
therefore , @xmath2268 and @xmath2269 do not belong to @xmath2270 .
consequently , the surface @xmath2271 is transverse to the tangent space t@xmath2259 of the manifold @xmath2260 , and we conclude as before that @xmath2272 structure of korteweg models and stability of diffuse interfaces .
_ interfaces free bound . *
7 * , no . 4 ( 2005 ) , 371 - 414 .
_ instabilit des tats
stationnaires dans les quations de schrdinger et de klein - gordon non linaires . _ c. r. acad .
paris sr .
* 293 * , no . 9
( 1981 ) , 489 - 492 . _ existence and properties of travelling waves for the gross - pitaevskii equation . _
stationary and time dependent gross - pitaevskii equations , 55 - 103 , contemp .
, * 473 * , amer . math .
soc . , providence , ri , ( 2008 ) .
_ nonlinear schrdinger equation as a model of superfluid helium .
_ in `` quantized vortex dynamics and superfluid turbulence '' edited by c.f .
barenghi , r.j .
donnelly and w.f .
vinen , lecture notes in physics , volume 571 , springer - verlag , 2001 . | we study the stability / instability of the subsonic travelling waves of the nonlinear schrdinger equation in dimension one .
our aim is to propose several methods for showing instability ( use of the grillakis - shatah - strauss theory , proof of existence of an unstable eigenvalue via an evans function ) or stability .
for the later , we show how to construct in a systematic way a liapounov functional for which the travelling wave is a local minimizer .
these approaches allow to give a complete stability / instability analysis in the energy space including the critical case of the kink solution .
we also treat the case of a cusp in the energy - momentum diagram .
+ * key - words : * travelling wave , nonlinear schrdinger equation , gross - pitaevskii equation , stability , evans function , liapounov functional . + *
msc ( 2010 ) : * 35b35 , 35c07 , 35j20 , 35q40 , 35q55 . |
witten s formulation @xcite of an intrinsic definition of the jones polynomial @xcite based on physical models lead to the more rigorous mathematical definitions via representations of quantum groups that were given by reshetikhin , turaev , and viro @xcite , @xcite .
these quantum invariants are speculated to generalize to higher dimensions .
such putative invariants have their origins in a theory of quantum gravity @xcite and higher categories @xcite . in relation to the current work
, the following progress has been made .
quantum spin networks were generalized by crane - yetter @xcite to give @xmath0-manifold invariants that were based on ooguri s proposal @xcite .
the invariants can be used to compute the signature as shown in @xcite .
birmingham - rakowski @xcite generalized the dijkgraaf - witten @xcite invariant of 3-manifolds , defined by group 3-cocycles , to triangulated 4-manifolds using pairs of cocycles .
crane and frenkel @xcite constructed hopf categories to define 4-manifold invariants , and they gave examples using canonical bases of quantum groups . in @xcite crane and yetter used cocycles to construct hopf categories . in this paper
we provide direct relations between the cocycle conditions of @xcite and pachner moves of @xmath0-manifolds , thus constructing a generalization of the dijkgraaf - witten invariants to dimension 4 .
the relations are established diagrammatically , providing connections between hopf categorical structures and triangulations via dual graphs and their movies .
the current paper is self - contained , but the reader might enjoy our introduction to the subject given in @xcite , where many of our ideas and motivations are introduced in a more leisurely fashion .
for the diagrammatic foundation of the invariants in dimension 3 see , @xcite , and . for the algebraic approach
see @xcite .
finally , there is a relation to higher dimensional knot theory as found in @xcite .
let us continue our motivational remarks . in @xmath1 dimensions
, planar diagrams played a key role in the definitions of both knot invariants and manifold invariants .
such diagrams are convenient since they help one grasp the categorical and algebraic structures needed for defining invariants .
one of the difficulties in generalizing to dimension @xmath0 or higher is the lack of such visualizations and diagrammatic machinery .
the purpose of this paper is to provide basic diagrammatic tools to study @xmath0-manifold triangulations , and to use such formulations to define invariants .
in particular , we formulate the crane - frenkel approach in terms of cocycles as initial data and prove the invariance under pachner moves in a diagrammatic way .
we introduce spin networks for the study of such invariants .
we hope that the present work serves as a basic tool in exploring the possibilities in higher dimensions .
there is a close relationship bewteen certain physical models in statistical mechanics and quantum field theory and the formulation of `` quantum '' invariants of knots , links and three - manifolds .
we hope that this relationship continues into dimension four .
in particular , one can hope that a four dimensional toplogical field theory ( such as we study here ) would be related to the calculation of amplitudes in quantum gravity .
the naive reason for this hope is quite simple : an amplitude for quantum gravity must sum over the possible metrics on the four - dimensional spacetime . averaging over metrics
should in priniciple produce basic numbers that are metric - independent . in other words
a valid process of averaging over metrics should produce topological invariants of the underlying four - space .
of course on the physical side it will be neccessary to extricate the topological part from the complexities of the model .
on the pure mathematical side it will be neccessary to see the relevance of the mathematics .
nevertheless this hope for an application to quantum gravity is one of the forces that drives our project .
the organization of the paper is as follows . in section
[ 23 ] , we review state sum invariants for triangulated manifolds in dimensions @xmath2 and @xmath1 .
we emphasize diagrammatic relations between triangulations and algebraic structures . at the end of the section
, we summarize the idea of categorification in relation to the construction of higher dimensional invariants . in section [ pac4sec ]
we present diagrams of pachner moves in dimension @xmath0 .
we also introduce singular moves in dimension @xmath0 , and prove that singular moves together with @xmath1-dimensional pachner moves imply @xmath0-dimensional pachner moves . these lemmas will be used to prove the well - definedness of our invariants . in section [ t&dsec ]
we give generalization of spin networks to dimension @xmath0 .
triangulations are represented by movies of graphs , and these graph movies will be used to give a direct relation between hopf category structures and triangulations .
cocycle conditions defined by @xcite will be reviewed in section [ data ] .
symmetry of cocycles are defined . in section [ bzman ]
the state sum invariants will be defined , and will be proved to be well - defined in section [ invsec ] .
our proofs are diagrammatic .
they provide the basic machinery necessary to define other invariants defined via hopf categories .
the axioms of hopf categories are related to moves on triangulations of 4-manifolds in a manner similar to the relationship between hopf algebras and moves on 3-manifolds .
section [ hopfsec ] reviews the definition of hopf category and make the connection with the rest of the paper .
in this section , we review topological lattice field theories in dimension @xmath1 and explain how they are generalized from those in dimension @xmath2 .
first we review dimension @xmath2 following @xcite where semi - simple algebras are used .
an alternative approach is given by lawrence in @xcite in which the algebra is assumed to be frobinius .
next the turaev - viro @xcite theory is reviewed following @xcite and @xcite .
invariants of 3-manifolds derived from hopf algebras are presented following @xcite .
alternative approaches are found in kuperburg @xcite and kauffman - radford @xcite .
some of the summary appeared in @xcite .
we summarize wakui s defintion @xcite of the dijkgraff - witten invariants @xcite , but here we show invariance using the pachner theorem .
this section closes with a conceptual scheme for generalizing to dimension 4 . * topological lattice field theories in dimension @xmath2 . * * pachner moves in dimension @xmath1 . *
[ pac23 ] * turaev - viro invariants . * * invariants defined from hopf algebras . * in this section we review invariants defined by chung - fukuma - shapere @xcite and kuperberg @xcite ( we follow the description in @xcite ) .
we note that the invariants obtained in this section are also very closely related to the invariants defined and studied by hennings , kauffman , radford and otsuki ( see @xcite for example ) . for background material on hopf algebras see @xcite or @xcite , for example . * definition ( bialgebras ) . *
a _ bialgebra _ over a field @xmath3 is a quintuple @xmath4 such that 1 .
@xmath5 is an algebra where @xmath6 is the multiplication and @xmath7 is the unit .
( _ i.e. _ , these are @xmath3-linear maps such that @xmath8 , @xmath9 ) .
@xmath10 is an algebra homomorphism ( called the _ comultiplication _ ) satisfying @xmath11 , 3 .
@xmath12 is an algebra homomorphism called the _
counit _ , satisfying @xmath13 .
* definition ( hopf algebras ) . * an _ antipode _ is a map @xmath14 such that @xmath15 .
a _ hopf algebra _ is a bialgebra with an antipode . the image of the comultiplication is often written as @xmath16 for @xmath17 .
the image in fact is a linear combination of such tensors but the coefficients and the summation are abbreviated ; this is the so - called sweedler notation @xcite .
the most important property from the present point of view is the compatibility condition between the multiplication and the comultiplication ( _ i.e. _ , the condition that the comultiplication is an algebra homomorphism ) , and we include the commuting diagram for this relation in figure [ coh ] .
the condition is written more specifically @xmath18 where @xmath19 denotes the permutation of the second and the third factor : @xmath20 . in the sweedler notation , it is also written as @xmath21 . the definition of invariants defined in @xcite is similar to the @xmath2-dimensional case . given a triangulation @xmath22 of a @xmath1-manifold @xmath23 , give spins to edges with respect to faces ( triangles ) .
the weights then are assigned to edges and to faces .
the structure constants @xmath24 ( resp .
@xmath25 ) of multiplication ( resp .
comultiplication ) are assigned as weights to faces ( resp .
if an edge is shared by more than three faces , then a composition of comultiplications are used .
for example for four faces sharing an edge , use the structure constant for @xmath26 .
the coassociativity ensures that the other choice @xmath27 gives the same constant @xmath28 .
thus the partition function takes the form @xmath29 .
this formula exhibits the form of the partition function for this model , but is not technically complete .
the full formula uses the antipode in the hopf algebra to take care of relative orientations in the labellings of the simplicial complex .
see @xcite for the details . in @xcite
the well definedness was proved by using singular triangulations these generalize triangulations by allowing certain cells as building blocks . in this case
the move called the _ cone move _ for a singular triangulation plays an essential role .
this move is depicted in figure [ 3dcone ] with a dual graph to illustrate the relationship to the compatibility condition .
let us now explain the relation of this move to the compatibility condition verbally . in the left hand side of fig .
[ 3dcone ] there are distinct and parallel triangular faces sharing the edge @xmath30 and @xmath31 ; these triangles have different edges connecting the vertices @xmath2 and @xmath1 .
one of these is shared by the face @xmath32 while the other is shared by the face @xmath33 .
the parallel faces ( 123 ) and ( 123) are collapsed to a single face to obtain the right hand side of fig . [ 3dcone ] .
now there is a single face with edges @xmath30 , @xmath34 , and @xmath35 , and the edge @xmath34 is shared by three faces @xmath36 , @xmath32 , and @xmath33 .
the thick segments indicate part of the dual graph .
each segment is labeled by hopf algebra elements .
reading from bottom to top , one sees that the graphs represent maps from @xmath37 to itself . the left - hand - side of the figure represents @xmath38 @xmath39 @xmath40 while the right - hand - side represents @xmath41 and these are equal by the consistency condition between multiplication and comultiplication .
this shows that the hopf algebra structure gives solutions to the equation corresponding to the cone move . that the partition function in this case does not depend on the choice of triangulation
is proved by showing that the pachner moves follow from the cone move and other singular moves .
figure [ chfs ] explains why the @xmath42-move follows from singular moves ( this figure is basically the same as a figure in @xcite ) .
let us explain the figure .
the first polyhedron is the right - hand - side of the @xmath42-move .
there are three internal faces and three tetrahedra .
perform the cone move along edge @xmath43 thereby duplicating face @xmath44 .
internally , we have face @xmath44 glued to face @xmath33 along edge @xmath43 and face @xmath45 glued to face @xmath46 along edge @xmath47 .
these faces are depicted in the second polyhedron . by associativity
these faces can be replaced by four faces parallel to four faces on the boundary ; @xmath36 , @xmath48 , @xmath49 , @xmath50 .
this is the configuration in the third polyhedron .
then there are two @xmath1-cells bounded by these parallel faces .
collapse these cells and push the internal faces onto the boundary ( this is done by singular moves ) .
the result is the fourth polytope which now is a single polytope without any internal faces .
this is the middle stage in the sense that we have proved that the right - hand - side of the @xmath42-move is in fact equivalent to this polytope .
now introduce a pair of internal faces parallel to the faces @xmath48 and @xmath50 to get the fifth polytope ( the left bottom one ) .
perform associativity again to change it to a pair of faces @xmath51 and @xmath52 to get the sixth polytope .
perform a cone move along the pair of faces with vertices @xmath53 these faces share edges @xmath54 and @xmath55 edge @xmath56 is duplicated . ) the last picture which is the left - hand - side of the @xmath42-move . in summary , we perform cone moves and collapse some @xmath1-cells to the boundary and prove that both sides of the pachner move is in fact equivalent to the polyhedral @xmath1-cell without internal faces .
we give a generalization of this theorem to dimension @xmath0 in lemmas [ pac33 ] , [ pac24 ] , and [ pac15 ] .
* dijkgraaf - witten invariants . * * summary : going up dimensions . *
in section [ pac23 ] we reviewed the pachner moves for triangulations in dimensions @xmath2 and @xmath1 and their relations to associativity of algebras . in this section , we describe pachner moves in dimension @xmath0 .
relations of these moves to the stasheff polytope was discussed in @xcite . in general , an @xmath57-dimensional pachner move of type @xmath58 , where @xmath59 , is obtained by decomposing the ( spherical ) boundary of an @xmath60-simplex into the union of two @xmath57-balls such that one of the balls is the union of @xmath61 @xmath57-simplices , the other ball is the union of @xmath62 @xmath57-simplices , and the intersection of these balls is an @xmath63-sphere . by labeling the vertices of the @xmath60-simplex
these moves are easily expressed .
for example , the table below indicates the lower dimensional pachner moves : [ cols="<,^,^",options="header " , ] the cone move to @xmath64 and @xmath65 ( which is illustrated schematically in fig .
[ tacolem5 ] ) followed by the pillow move to @xmath66 and @xmath67 collapses @xmath68 to @xmath69 as claimed .
thus we get two polytopes @xmath70 and @xmath71 .
next perform a pachner move to @xmath72 which shares @xmath73 . as a result
we get three new tetrahedra @xmath74 .
the @xmath75-move is illustrated in fig .
[ tacolem6 ] ; the labels on the polytopes indicate their antecendents .
thus we obtain @xmath76 bounded by @xmath77 , @xmath78 , @xmath79 , @xmath66 , @xmath64 , @xmath80 , @xmath81 , @xmath82 , and @xmath83 , and @xmath84 bounded by @xmath85 , @xmath86 , @xmath87 , and @xmath74 .
hence we now can collapse @xmath84 to the tetrahedra @xmath85 , @xmath86 , and @xmath87 in the same manner as we did to @xmath68 .
the collapsing is indicated in fig .
[ tacolem7 ] .
the result is a single polytope @xmath88 resulted from @xmath76 which has the same boundary tetrahedra as those of the left hand side of the 4-dimensional pachner move .
figure [ tacolem7 ] indicates the resulting polytope at the bottom of the figure . in fig .
[ tacolem8 ] the 3-dimensional boundary is illustrated .
notice the following : ( 1 ) triangle ( ace ) is no longer present ; ( 2 ) among the nine tetrahedra illustrated , neither triangle @xmath89 nor triangle @xmath90 appears ; ( 3 ) these are all of the tetrahedral faces of the 5 simplex that contain neither @xmath89 nor @xmath90 .
thus we can apply the same method starting with @xmath90 to get to this polytope .
this proves that @xmath91-move is described as a sequence of singular moves ( cone , taco , and pillow moves ) and pachner moves .
@xmath92 * lemma . *
[ pac24 ] the @xmath93-move is described as a sequence of cone moves , pillow moves , taco moves , and @xmath1-dimensional pachner moves .
_ we use the following labeling for the @xmath93-move in this proof : @xmath94 perform a @xmath91-move ( which was proved to be a sequence of the singular moves in the preceeding lemma ) to @xmath95 to get @xmath96 .
then the polytope now consists of @xmath97 , @xmath98 , @xmath99 , and @xmath100 .
perform a pachner move to the tetrahedra @xmath69 , that are shared by @xmath99 and @xmath100 , to get @xmath101 .
this changes @xmath102 to a @xmath0-cell bounded by @xmath103 , @xmath104 , @xmath105 , and @xmath106 .
the cone move followed by the pillow move collapses this polytope yielding @xmath107 , the left - hand - side of the @xmath93-move . @xmath92
* lemma . * [ pac15 ] the @xmath108-move is described as a sequence of cone moves , pillow moves , taco moves , and @xmath1-dimensional pachner moves .
_ we use the following labelings : @xmath109 perform the @xmath91-move to @xmath110 to get @xmath111 .
the @xmath0-simplices @xmath100 and @xmath99 share all their tetrahedral faces except @xmath64 ( and @xmath65 ) . perform a @xmath112-move to each of these shared tetrahedra to get @xmath0-cells bounded by copies of @xmath64 sharing all the @xmath2-faces .
thus the pillow moves will collapse @xmath100 and @xmath99 .
the same argument collapses @xmath113 to get the left - hand - side of the @xmath108-move . @xmath92
* remark . * in @xcite crane and frenkel proposed constructions of @xmath0-manifold quantum invariants using hopf categories .
hopf categories generalize the definition of hopf algebra to a categorical setting in the same way that modular categories generalize modules .
one of the conditions in their definition is called the coherence cube which generalizes the compatibility condition of hopf algebras between mutiplication and comutiplication ( see section [ hopfsec ] ) .
they showed that this condition corresponds to the cone move .
thus lemmas in this section can be used to prove the well - definedness of invariants they proposed by showing that their definition is invariant under pachner moves .
in dimension 3 , quantum spin networks are used on the one hand to provide calculations of identities among representations of quantum groups @xcite . on the other hand they are cross sections of the dual complex of a triangulated 3-manifold ( see section [ pac23 ] ) .
in this section , we use similar graphs to relate them to the dual complex of triangulated 4-manifold .
we begin the discussion on the local nature of triangulated 4-manifolds near 2-dimensional faces . * graphs , 2-complexes , and triangulations . * [ graphdef ] * definition . *
the dual complex @xmath114 of @xmath115 is defined as follows .
pick a vertex @xmath116 of @xmath114 in the interior of each @xmath0-simplex of @xmath115 .
connect two vertices @xmath117 and @xmath118 of @xmath114 if and only if the corresponding @xmath0-simplices of @xmath115 share a @xmath1-face .
thus each edge of @xmath114 is dual to a tetrahedron of @xmath115 .
edges @xmath119 of @xmath114 bound a face @xmath120 if and only if the corresponding tetrahedra share a @xmath2-face of @xmath115 . a set of @xmath2-faces @xmath121 of @xmath114 bounds a @xmath1-face ( a polyhedron ) if and only if the corresponding faces of @xmath122 share an edge of @xmath115 .
finally a set of @xmath1-faces of @xmath123 bounds a @xmath0-face if and only if the corresponding edges of @xmath122 share a vertex .
thus @xmath123 gives a cw - complex structure to the @xmath0-manifold .
* definition .
* let @xmath115 be a triangulation of a @xmath0-manifold @xmath23 , and let @xmath123 be the dual complex .
each @xmath1-face of @xmath123 is a polytope .
choose a triangulation of each @xmath1-face into tetrahedra so that it defines a triangulation of the @xmath1-skeleton of @xmath123 .
we require that such a triangulation does not have interior vertices in the @xmath2-faces of @xmath123 .
thus the restriction on each @xmath57-agonal @xmath2-face consists of @xmath124 triangles .
such a choice of triangulation is called a _ @xmath1-face triangulation _ ( a _ triangulation _ for short ) of @xmath114 .
a @xmath1-face triangulation is denoted by @xmath125 .
* definition ( carrier surface ) .
* in each tetrahedron of the triangulation @xmath115 , we embed the dual spine to the tetrahedron .
the intersection of the dual spine with a triangular face is a graph consisting of a 3-valent vertex with edges intersecting the edges of the tetrahedron .
there is a vertex in the center of the 2-complex at which four edges ( corresponding to the faces of the tetrahedron ) and six faces ( corresponding to the edges ) intersect .
the union ( taken over all tetrahedra in the triangulated 4-manifold ) of these 2-complexes form a 2-complex , @xmath126 , that we call the
_ carrier surface_. let us examine the incidence relations of the carrier surface along faces and edges of the triangulation .
consider a 2-face , @xmath120 , of the triangulation @xmath115 .
suppose that @xmath57 tetrahedra are incident along this triangle @xmath120 .
then the dual face @xmath127 is an @xmath57-gon .
the 4-manifold in a neighborhood of the face @xmath120 looks like the cartesian product @xmath128 .
the carrier surface in this neighborhood then appears as @xmath129 where @xmath130 is the 1-complex that consists of the cone on @xmath57-vertices ( _ i.e. _ , the @xmath57-valent vertex ) , and @xmath131 is the graph that underlies that alphabet character ( a neighborhood of a trivalent vertex ) .
for example @xmath132 is an interval , @xmath133 , @xmath134 , _ etc .
_ we can think of @xmath130 being embedded in @xmath127 with the edges of @xmath130 intersecting the centers of the edges of @xmath127 and the vertex of @xmath130 lying at the `` center '' of @xmath127 ( _ i.e. _ we may assume that @xmath127 is a regular polygon ) . consider an edge , @xmath135 , of @xmath115 , and the 3-cell , @xmath136 , that is dual to @xmath135 .
the faces of @xmath136 are @xmath57-gons , @xmath127 , that are dual to the triangular faces , @xmath120 , which are incident to @xmath135 .
the carrier surface intersects a face @xmath127 in the graph @xmath130 .
the carrier surface intersects @xmath136 in a 2-complex that is the cone on the union of the @xmath130 where the union is taken over all the faces of @xmath136 .
the situation is depicted in fig .
[ face ] in which three tetrahedra intersect along a triangular face . on the right hand side of the figure
, we illustrate a _ graph movie_. the two graphs that are drawn there represent the intersection of the carrier complex with the boundary of @xmath128 . in a neighborhood of this face
the carrier complex looks like @xmath137 . in this and subsequent figures ,
the vertices that are labeled with open circles correspond to the dual faces @xmath127 . in this figure ,
three such circled vertices appear since the dual face appears on each of the duals to the three edges . *
faces and diagrams . * [ f&d ] * defintion . * we perturb the carrier surface to construct a 2-dimensional complex that has the following properties : 1 .
the vertices of the complex all have valence 4 or valence 6 . 2 .
exactly three sheets meet along an edge ; 3 .
the set of edges can be partitioned into two subsets ; we color the edges accordingly .
4 . a valence 4 vertex has 4 edges of the same color incident to it ; 5 . a valence 6 vertex has 3 edges of each color incident to it . 6 .
thus , the 2-complex has a tripartite graph as its 1-complex and a bipartition on the set of edges .
such a 2-complex will be called a _
perturbed carrier_. * lemma . * [ corresplemma ] a perturbed carrier can be constructed from the carrier surface by means of a 3-face triangulation .
_ consider a 3-face triangulation ; recall this is a triangulation of the dual 3-cells of the triangulation @xmath115 , and a 3-face is the dual to an edge @xmath136 . a @xmath57-agonal face of @xmath136 is divided into @xmath124 triangles .
the graph @xmath130 in the @xmath57-agonal face is replaced with the dual to the triangulation . in @xmath136 , the cone on the union of the @xmath138s is replaced by the union of the duals to the tetrahedra in the triangulation .
these are the surfaces with 6 faces , 1 vertex , and 4 edges ; they glue together in @xmath136 to form the subcomplex in which all of the vertices have one color .
an example is illustrated in fig .
[ pertcube ] in which the dual of an edge is a cube .
the vertices that have two different colored edges incident to them are found on the triangular faces of the 3-face triangulation .
three of the edges are coming from the dual face , the other three edges are coming from the dual complex of the original tetrahedra .
the local structure at the 6-valent vertices was explained in detail above .
this completes the proof . @xmath92
* definition . *
a _ graph movie _ is a sequence of graphs that appear as cross sections of a portion of the perturbed carrier when a height function is chosen , such that the stills of movies are graphs having trivalent ( circled and uncircled ) vertices and between two stills , the movie changes in one of the following ways : 1 .
the change of the movie at a face ( a @xmath139-valent vertex of a carrier surface ) is as defined above ( the change of graphs shown in fig .
[ face ] ) .
2 . the change of the movie at a @xmath0-valent vertex is as depicted in fig .
[ alpha ] bottom .
the changes of the movie at critical points of edges and faces of the carrier surface is generic .
they are depicted in fig .
[ othermovies ] . in the graphs we use circled vertices and uncircled vertices .
these are cross section of two types of edges . in the figures of carrier surfaces ( fig .
[ surfcubelhs ] , [ surfcuberhs ] , and [ pertcube ] ) , the edges corresponding to circled vertices are depicted by thin tubes .
the graph movie defined here includes definitions given above ( which are clearly equivalent ) .
the graph movie allows us to view the perturbed carrier via a sequence of 2-dimensional cross - sections whereas the carrier surface itself does not embed in 3-dimensional space . * taco moves and graph movies . *
in this section , we list cocycles and their equalities that will be used in the following sections . these cocycles are given in @xcite in relation to hopf categories ( see section [ hopfsec ] ) .
some non - trivial examples are given therein .
first , we mention that two of the cocycle conditions are depicted in figs .
[ cocymovie1 ] and [ cocymovie2 ] as relations to graph movies where the edges of the graph have been colored with pairs of group elements and dual group elements .
these graph movies correspond to the dual graphs that correspond to the taco move ( fig .
[ tacomovielhs ] and [ tacomovierhs ] ) .
the coloring will be explained in the subsequent section .
let @xmath140 be a finite group and @xmath141 be the multiplicative group of a field @xmath142 .
let @xmath143 denote the abelian group of all functions from @xmath144 to @xmath141 where @xmath145 we need the following functions ( called cocycles if they satisfy the conditions given in the next section ) .
* @xmath146 , * @xmath147 , * @xmath148 . *
cocycle conditions .
* * @xmath149 + @xmath150 * @xmath151 + @xmath152 , * @xmath153 + @xmath154 , * @xmath155 + @xmath156 . * cocycle symmetries . * * definition . * [ cocysym ] the following are called the _
cocycle symmetries_. * @xmath157 + @xmath158 + where @xmath159 .
* @xmath160 + @xmath161 * @xmath162 + @xmath163
* labeling . * * definition . * [ deflabel ] a _ labeling _ ( or _ color _ ) of @xmath115 with oriented edges with respect to a finite group @xmath140 is a function @xmath164 where @xmath165 and @xmath166 here @xmath167 denotes the set of oriented edges , and @xmath168 is the set of tetrahedra . we require the following compatibility condition . if @xmath169 forms an oriented boundary of a face of a tetrahedron @xmath170 , @xmath171 , and @xmath172 , then @xmath173 , where @xmath174 , @xmath175 , and @xmath176 .
we call this rule _ the local rule of colors at a triangle _ ( or simply a _ local rule _ ) .
the situation is depicted on the left of fig .
[ rule ] .
when an order of vertices is given , the edges are oriented by ascending order of vertices ( if the vertices @xmath116 and @xmath177 of an edge have the order @xmath178 , then the edge is oriented from @xmath116 to @xmath177 ) . however in this definition the order on vertices is not required , although orientations on edges are required . for an oriented edge @xmath135 , the same edge with the opposite orientation
is denoted by @xmath179 .
consider the edge @xmath180 on the left of fig .
[ rule ] , and reverse the orientation of @xmath180 to get @xmath181 .
then the color of @xmath181 is required to be @xmath182 where @xmath183 as depicted in the figure .
in other words , the color for an edge with reversed orientation is defined to satisfy the local requirement of the left of fig .
[ rule ] .
we often use sets of non - negative integers to represent simplices of @xmath115 .
for example , fix a @xmath1-face ( or tetrahedron ) @xmath22 of @xmath115 .
let @xmath184 , @xmath185 , @xmath2 , and @xmath1 denote the vertices of @xmath22 .
for a pair of an oriented edge @xmath186 and a tetrahedron @xmath187 a labeling assigns a pair @xmath188 which we sometimes denote by @xmath189 .
when a total order is fixed , the integers are assumed to have the compatible order ( @xmath190 ) . we will show ( lemma [ 6in1 ] ) that there is a coloring of each tetrahedron satisfying the local rule .
furthermore , we will show that changing the orientations of edges of a colored tetrahedron results in a unique coloring
. * definition . *
a _ labeling _ ( or _ color _ ) of @xmath125 with oriented dual edges is a function @xmath191 where @xmath192 here @xmath193 ( resp .
@xmath194 ) denotes the set of oriented edges ( resp .
3-polytopes ) of @xmath125 ( resp .
@xmath123 ) . the following compatibility conditions are required .
if @xmath169 form an oriented boundary of a face of a tetrahedron @xmath170 of @xmath125 , then the first factors of colors ( group elements ) coincide , and if they are @xmath195 , and @xmath196 , then @xmath197 , where it is required that @xmath198 .
when an order of vertices is given , the edges are oriented by ascending order of vertices as before .
consider the edge @xmath180 in the fig .
[ rule ] right , and reverse the orientation of @xmath180 to get @xmath181 .
then the color of @xmath181 is required to be @xmath199 where @xmath200 as depicted in the figure .
in other words , the color for an edge with reversed orientation is defined to satisfy the local requirement of the right side of fig .
[ rule ] . in the figure ,
dual graphs in triangles are also depicted .
we put a small circle around a trivalent vertex for the dual faces .
as in the case for tetrahedra , dual tetrahedra can be colored , and changing the orientation for colored dual tetrahedra gives a unique new coloring ( lemma [ 6in1 ] ) .
note that there is a pair @xmath201 which is dual to a pair @xmath202 , in the sense that @xmath203 is dual to the tetrahedron @xmath170 and @xmath204 is dual to the edge @xmath135 .
however there are pairs in @xmath205 that are not to dual to pairs in @xmath206 .
* definition . * a _ labeling _ ( or _ color _ ) of @xmath207 is a function @xmath208 such that @xmath209 if @xmath201 is dual to @xmath202 .
this function is also called a @xmath210 . for a particular pair @xmath211 ( resp .
@xmath212 ) , the image @xmath213 ( resp .
@xmath214 ) is also called a _
spin_. this is sometimes denoted by @xmath215 . * definition .
* we say that two simplices are _ adjacent _ if they intersect .
we say a simplex @xmath216 and a dual simplex @xmath217 are _ adjacent _ if @xmath216 intersects the polyhedron of the dual complex in which @xmath217 is included .
* lemma . * [ 6in1 ] ( 1 ) for a tetrahedron or dual tetrahedron , there are colors satisfying the local rule at every face or dual face .
\(2 ) there are colors on the edges and dual edges adjacent to a given face satisfying the local rules .
\(3 ) let @xmath126 be a color assigned to the oriented edges ( or dual edges ) of a tetrahedron ( or dual tetrahedron ) .
let @xmath218 be a color assigned to the same tetrahedron ( or dual tetrahedron ) with orientations reversed on some of the edges
. then @xmath218 is uniquely determined .
if the color @xmath126 is assigned to oriented edges and dual edges that are adjacent to a face , then the color @xmath218 is uniquely determined when some of the edges or dual edges have their orientations reversed .
_ we prove ( 1 ) and ( 2 ) in the case of tetrahedra .
the proof for dual tetrahedra is similar and follows from @xcite . for tetrahedra
, the situation is depicted in fig .
[ alpha ] . in the top of the figure ,
a tetrahedron with colors on three edges is depicted .
other edges receive compatible colors that are determined by these three . in the middle ,
pairs of front faces and back faces are depicted in the left and right respectively , together with dual graphs . in the bottom of the figure ,
only the dual graphs are depicted , together with colors on all the edges .
we check that the first factors of colors match in multiplication convention in fig .
the second factors are also checked as follows : in the bottom left figure , @xmath219 and @xmath220 . in the bottom right figure , @xmath221 ( the same relation as above ) and @xmath222 which reduces to @xmath223 , the same relation as above .
thus the requirements on faces match at a tetrahedron .
to prove ( 3 ) for tetrahedra , first consider the case where the orientation of the edge @xmath224 is reversed .
then the face @xmath36 forces the change @xmath225 .
the other face @xmath226 forces the change @xmath227 .
these are equal since @xmath159 .
the situation is depicted in the top right of fig .
[ alphasymmetry ] where the reversed orientations are depicted by a small circle on the edge .
the top left figure indicates the original colors . in the figure ,
the `` hats '' on the dual group elements are abbreviated for simplicity .
the other cases when the orientation of a single edge is reversed , are also depicted for the cases @xmath228 , @xmath229 .
the general case follows because all cases are obtained by compositions of these changes .
next consider statement ( 2 ) . in fig .
[ phi ] the colors are depicted using dual graphs ( identify this graph with the graph in fig .
[ face ] ) .
first we check the orientation conventions in the figure .
identify the circled vertex of the right - hand - side graph of bottom of fig .
[ face ] with the right - hand - side of fig .
then the orientation conventions of dual edges coincide where @xmath230 of fig .
[ rule ] corresponds to the edge @xmath231 , @xmath180 to @xmath232 , @xmath233 to @xmath234 .
the tetrahedron @xmath235 is shared by @xmath236 and @xmath237 , that are ordered as @xmath238 among three @xmath0-simplices .
thus this correspondence to @xmath230 matches with the definition of the orientation of dual edges .
now we check the constraints . in the left bottom of fig .
[ face ] the following relations must hold : @xmath239 ( from the top left vertex ) , @xmath240 ( from the top right vertex ) , @xmath241 ( from the bottom right vertex ) , and @xmath242 ( from the bottom left vertex ) .
the last relation is reduced by substitution to @xmath243 both sides , so that the weight is compatible . in the right of the figure
, we get the relation @xmath241 from the top vertex , which is the same as above , and the condition for the bottom vertex is already incorporated ( by using @xmath244 in bottom left ) .
thus the colors around a face are compatible .
now let us check that the orientation conventions are compatible in fig .
the orientations on the edges are the orientations from the vertex ordering as seen in the figure .
the orientations on dual edges are checked as follows . in the figure the face @xmath245
is shared by three @xmath0-simplices , @xmath236 , @xmath246 , and @xmath237 .
the dual edge labeled by @xmath247 is dual to the tetrahedron @xmath235 and oriented from @xmath236 to @xmath237 , corresponding to the edge @xmath230 on the right of fig .
[ rule ] . respectively , the one labeled by @xmath248 goes from @xmath237 to @xmath246 corresponding to @xmath180 , the one labeled @xmath249 goes from @xmath236 to @xmath246 corresponding to @xmath233 .
thus the orientations defined from the order on vertices match the convention in fig .
[ rule ] .
to prove part ( 3 ) , we check the cases of interchanging the orientation of some of the edges
. the general case will follow from the cases depicted in fig .
[ phisymmetry1 ] and [ phisymmetry2 ] since the orientation changes depicted therein generate all the orientation changes .
first consider the case where the edge @xmath30 has reversed orientation .
this corresponds to changing the vertex order from @xmath250 in the figure to @xmath251 .
( the orientations on dual edges do not change . )
then three colors change : @xmath252 to @xmath253 , @xmath254 to @xmath255 , and @xmath256 to @xmath257 .
these changes are forced by the rules at faces ( uncircled trivalent vertices of the graphs ) .
hence we check the rules at circled trivalent vertices .
the only relevant vertex is the one on the left bottom in the figure .
it must hold that @xmath258 .
this indeed follows from @xmath259 .
the other cases are similar .
@xmath92 * lemma . * [ 2compcolor ] the colors define a function @xmath260 where @xmath126 is a perturbed carrier of @xmath115 and @xmath261 is the set of @xmath2-faces of @xmath262 .
conversely , a function @xmath260 defines a color @xmath262 defined for the triangulation @xmath115 and @xmath263 .
_ this follows from lemma [ corresplemma ] and the definition of @xmath262 .
@xmath92 * weighting .
* * definition ( weights for tetrahedra)*. let @xmath264 be a tetrahedron with vertices @xmath184 , @xmath185 , @xmath2 , and @xmath1 .
suppose @xmath265 , @xmath266 , and @xmath267 .
the weight of @xmath22 with respect to the given labelings of edges is a number ( an element of the ground field ) defined by @xmath268 here @xmath269 is @xmath270 and is defined as follows .
let @xmath271 be the tetrahedron in consideration where @xmath272 .
then @xmath22 is shared by two @xmath0-simplices , say , @xmath273 and @xmath274 . here
we ignore the given labels of @xmath116 and @xmath177 , and consider the orders written above ( @xmath116 and @xmath177 coming last ) .
then exactly one of these two @xmath0-simplices , say , @xmath262 , with this order @xmath275 , matches the orientation of the @xmath0-manifold , and the other has the opposite orientation .
consider the label on @xmath116 induced by the ordering on the vertices . if the integer index of @xmath116 is such that the oriented simplex @xmath276 is obtained from the order induced by labeling by an even permutation , then @xmath277 .
otherwise , @xmath278 .
( sometimes we represent the order of vertices by labeling the vertices by integers . ) * definition ( weights for faces ) .
* suppose that a face @xmath279 is shared by three tetrahedra @xmath235 , @xmath280 , and @xmath281 .
suppose @xmath282 , @xmath283 , @xmath284 , @xmath285 , and @xmath286 .
the situation is depicted in fig .
[ phi ] . then the weight for the face @xmath287 is defined as the number @xmath288 here the sign @xmath289 is defined as follows . if the local orientation defined by @xmath245 in this order together with the orientation of the link @xmath290 of @xmath291 in this order gives the same orientation as that of the @xmath0-manifold , then @xmath292 , otherwise @xmath293 .
suppose the face @xmath245 is shared by @xmath57 ( more than three ) tetrahedra @xmath294 , @xmath295 .
note that the vertices of these tetrahedra other than @xmath296 form a link of the face @xmath245 .
assume that these vertices are cyclically ordered by @xmath297 , and that the link of @xmath245 and @xmath298 matches the orientation of the @xmath0-manifold .
suppose @xmath299 , @xmath300 , @xmath301 , @xmath302 , @xmath303 , ( @xmath304 ) , @xmath305 .
then @xmath306 @xmath307 if the cyclic order of vertices is not as above , then it can be obtained from the above by transpositions .
when a transposition between @xmath3-th and @xmath308-st vertex occurs , change the argument of @xmath3-th weight @xmath309 to @xmath310 .
notice that by the conditions in the definition of a triangulation each face must be shared by at least three tetrahedra .
however in the course of computation we may have to deal with _
singular triangulations_. in this case it can happen that only two tetrahedra share a face .
let @xmath311 and @xmath312 be such two terahedra sharing the face @xmath313 . then the weight assigned to the face @xmath313 in this case is the product of kronecker s deltas : @xmath314 there are other cases of singular triangulations that appear in our proofs of well - definedness , and their weights are defined as follows .
suppose that two tetrahedra share vertices @xmath315 and share all of their faces except @xmath316 .
meanwhile , suppose that the face @xmath316 on each of the respective tetrahedra is shared with tetrahedra @xmath317 and @xmath318 .
the other faces are shared by the tetrahedra @xmath319 , @xmath320 , and @xmath321 .
the situation is depicted in fig .
[ conemove ] top . in the figure ,
colors and weights are also depicted .
the signs for each weight @xmath322 are also depicted in the figure in this situation , by indicating the power @xmath323 on one of the @xmath324s .
suppose in another situation that two tetrahedra share all vertices and all @xmath2-faces ( triangles ) as depicted in fig .
[ pillow ] .
the signs for this situation are also depicted in the figure . in general
if the order of vertices are different , then they are obtained from the above specific situations by compositions of permutations .
then the weights and signs are defined by applying lemma [ 6in1 ] .
* definition ( weights for edges ) .
* consider the triangulation of the 3-sphere @xmath325 consisting of 5 tetrahedra @xmath326 , @xmath327 , @xmath328 , @xmath329 , and @xmath330 , where integers represent the vertices .
this triangulation is depicted in figure [ edgecone ] by solid lines . here , the subdivided tetrahedron with vertices @xmath331 and @xmath332 with the interior vertex @xmath185 is a triangulation of a 3-ball , and together with the `` outside '' tetrahedron @xmath330 they form a triangulation of @xmath325 .
now take a cone of this triangulation with respect to the vertex @xmath184 to obtain a triangulation of a @xmath0-ball consisiting of 5 @xmath0-simplices @xmath236 , @xmath237 , @xmath246 , @xmath333 , and @xmath334 .
this is depicted in figure [ edgecone ] also , where edges having @xmath184 as end point are depicted by dotted lines .
( regard dotted lines as lying in the interior of the @xmath0-ball . )
suppose an edge @xmath335 has this particular triangulation as the neighborhood .
suppose @xmath336 , @xmath337 , and @xmath338 .
then the weight for the edge @xmath186 is defined by @xmath339 the sign @xmath340 is defined in the same manner as @xmath341 simply taking the dual orientations . if the neighborhood of an edge has a different triangulation , then the weight is defined as follows .
let @xmath342 be the set of tetrahedra of the polytope @xmath343 , and let @xmath344 be the set of edges of @xmath345 , @xmath346 . then the weight is defined by @xmath347 where each @xmath348 is assigned to a tetrahedron of the above triangulation following the order convention of vertices .
the product of the above expression is taken over all the shared edges , and the sum is taken over all the possible states on shared edges .
the exponent , @xmath349 , on the normalization factor , @xmath350 , is the number of verticies in the interior of the polyhedron dual to the given edge . * partition function .
* * definition .
* the partition function @xmath351 for a triangulation @xmath352 with a total order on vertices is defined by @xmath353 where the product ranges over tetrahedra , faces and edges of the triangulation @xmath115 , the summation ranges over all the possible states , and the exponent @xmath349 on the normalization factor is the number of vertices in the triangulation .
* main theorem . * [ ftn ] the partition function @xmath354 defined above for triangulations @xmath355 of a 4-manifold @xmath23 is independent of the choice of the triangulation @xmath115 and @xmath125 and independent of choice of order on vertices .
therefore the partition function @xmath356 defines an invariant of a 4-manifold @xmath23 .
section [ invsec ] is devoted to giving the proof of this theorem . * diagrams , cocycles , and triangulations .
* [ dc&t ] here we explain relations among diagrams , cocycles , and triangulations . figure [ rule ] illustrates the coloring rules at triangles and dual triangles . in these triangles and dual triangles graphs
are embedded ; the verticies of the graphs in the dual triangles are labeled by small circles .
the cocycles @xmath357 are assigned to tetrahedra ; the cocycles @xmath348 are assigned to dual tetrahedra , and the cocycles @xmath322 are assigned to triangular faces .
each such figure also corresponds to a graph movie which depicts a part of the perturbed carrier surface .
we can think of the cocycles as being assigned to the vertices of the perturbed carrier surface which has a tripartition on its vertex set . indeed
the many scenes that constitute the graph movie are found on the boundary of a regular neighborhood of the vertices of the carrier surface . in this way we can directly visualize the construction of the invariant as a colored surface with weighted vertices or as a colored graph movie with weights associated to the scenes .
simliarly , the cocycle conditions can be described as relations on movies of tree diagrams .
figures [ cocymovie1 ] and [ cocymovie2 ] depict these relations .
each change of a tree diagram ( scene in the movie ) corresponds to a cocycle as indicated .
when we multiply the left - hand - side and the right - hand - side of cocycles in the movies , we obtain cocycle conditions among @xmath357 , @xmath322 , and @xmath348 .
the cocycle conditions can also be understood in terms of certain singular surfaces that are embedded in the 4-manifold .
these surfaces are depicted in figs .
[ surfcubelhs ] and [ surfcuberhs ] . in these figures
the cocycles @xmath322 corresponds to the surface @xmath358 and the cocycle @xmath357 corresponds to the surface that is dual to a tetrahedron .
the assignment of @xmath322 to @xmath137 is indicated in the weights on fig .
the reasons for these assignments is that the cocycle @xmath322 is found when three tetrahedra share a triangular face , and the cocycle @xmath357 is assigned to a tetrahedra . in figs .
[ surfcubelhs ] and [ surfcuberhs ] some edges are denoted as tubes .
a tube of the form @xmath359 corresponds to a triangle that is shared by three tetrahdra as in fig .
recall the notation in section [ bzman ] : @xmath115 denotes a triangulation of a 4-manifold @xmath23 , @xmath123 its dual complex , @xmath125 a 3-face triangulation of @xmath123 . in section [ ordersec ]
we show that the partition function defined is independent of the order on vertices . in section [ invp ] , we show that the partition function is independent of the triangulation . in section [ nodual ]
we show that the partition function is independent of the choice of dual triangulation .
* independence on order of vertices . * [ ordersec ] in this section we prove * lemma . * [ invlemma1 ] the cocycle symmetries imply the independence of the partition function on the order on vertices of the triangulation . _
_ for tetrahedra and dual tetrahedra , the weights are the cocycles @xmath357 and @xmath348 respectively . as in @xcite
, it is sufficient to check how weights change when the order of vertices are changed from @xmath235 to @xmath360 , @xmath228 , and @xmath229 .
such changes are illustated in fig .
[ alphasymmetry ] for @xmath357 , and the corresponding conditions are listed for @xmath357 , @xmath348 and @xmath322 in section [ cocysym ] . for a face
, we check as follows . in fig .
[ phi ] an order of vertices are given , where the face is given by @xmath245 , and the other vertices are given labels @xmath1 ,
@xmath0 , and @xmath332 .
the changes of orders of vertices are generated by the changes from @xmath250 to @xmath251 , @xmath361 , ( for the face ) @xmath362 , @xmath363 , ( for the dual face ) since only the relative orders among the vertices of the face and those of the dual faces are in consideration .
for these changes , the colors are listed in fig .
[ phisymmetry1 ] and fig .
[ phisymmetry2 ] .
they are depicted in terms of dual graphs , and on the right hand side , the orientations of edges of faces / dual faces are shown .
the small circles indicate reversed orientations .
the corresponding conditions are listed in section [ cocysym ] .
if more than three tetrahedra share a face , then a change in order of the vertices can be achieved by such pairwise switches .
futhermore , in order to affect such changes , we may have to group the vertices in sets of 3 .
this grouping is achieved by a 3-face triangulation .
so the proof will follow once we have shown invariance under the 3-face triangulation . @xmath92 * independence under pachner moves . *
[ invp ] in this section , we explicitly relate the cone move , taco move and pillow move to the cocycle conditions . since these moves and lower dimensional moves generate the pachner moves , we will use the cocycle conditions to show that the partition function is invariant under the pachner moves .
* lemma . * [ coneinv ] the partition function is invariant under the cone move for a local triangulation with a specific choice of order depicted in fig .
[ conemove ] . _ proof .
_ let @xmath364 and @xmath365 be tetrahedra sharing the same faces @xmath245 , @xmath316 and @xmath226 , but having different faces @xmath366 and @xmath367 , such that ( 1 ) the union of the triangles @xmath366 @xmath368 @xmath367 bounds a 3-ball @xmath369 in the 4-manifold , ( 2 ) the union of @xmath369 , @xmath364 and @xmath365 is diffeomorphic to the @xmath1-sphere bounding a @xmath0-ball @xmath370 in the 4-manifold .
( see figs .
[ 4dcone ] , [ conemove ] . ) in these figures , the movies of dual graphs are depicted where each of the faces @xmath316 , @xmath226 , @xmath36 is shared by another tetrahedron ( @xmath280 , @xmath281 , @xmath319 , respectively ) .
we prove the invariance in this case .
the general case follows from such computations together with the pentagon identity of @xmath348 .
[ conemove ] shows the colors and cocycles assigned to this local triangulation ( again note the direct relation between this assignment and those for the top graph in fig .
[ cocymovie1 ] ) .
the left hand side of the cone move ( top of fig .
[ conemove ] ) has the local contribution @xmath371 ( note that the orientation of the face @xmath372 is opposite ) , and the right hand side of the cone move ( bottom of the figure ) has the local contribution @xmath373 thus the partition function is invariant under the cone move because the cocycle condition is satisfied .
@xmath92 * lemma .
* the partition function is independent under the pillow move for a specific local triangulation with the order depicted in fig .
[ pillow ] .
_ in fig . [ pillow ]
the assignments of colors and cocycles are shown .
the left hand side of the pillow move ( top of fig . [ pillow ] ) has the local contribution @xmath374 and the right hand side of the pillow move ( bottom of the figure ) has the local contribution @xmath375 this follows from the cocycle condition used in the above lemma .
@xmath92 * lemma . *
the partition function is independent under the taco move for a specific local triangulation with the order depicted in fig .
[ taco ] . _ proof . _ in fig .
[ taco ] the assignments of colors and cocycles are shown .
the left hand side of the taco move ( top of fig .
[ taco ] ) has the local contribution @xmath376 and the right hand side of the taco move ( bottom of the figure ) has the local contribution @xmath377 this is exactly one of the cocycle conditions .
@xmath92 observe that the diagrammatics of the graph movie move that results from the taco move match exactly the graph movie move that represents the cocycle condition fig .
[ cocymovie1 ] .
similar graph movies can be drawn for the cone and pillow moves and the correspondence with the move and the coycle conditions can be worked out via the graph movies .
making such correspondence shows explicitly the method of constructing invariants via hopf categories where , instead of cocycle conditions , coherence relations are used .
the coherence relations can be expressed by such graph movie moves ( see section [ hopfsec ] ) .
since the partition function is invariant under the cone , taco , and pillow moves , and since @xmath357 satisfies a pentagon relation , we have the partition function is invariant under the pachner moves .
* independence on triangulations of the dual complexes . * [ nodual ] in this section , we complete the proof that the partition function is well - defined by showing that the partition function does not depend on the 3-face triangulation , @xmath125 .
* lemma . * [ relpaclemma ] if @xmath378 and @xmath379 are triangulations of a @xmath1-dimensional polytope which is diffeomorphic to a @xmath1-ball such that @xmath378 and @xmath379 restrict to the same triangulation on the boundary , then they are related by a finite sequence of pachner moves . _ proof .
_ we first prove the corresponding statement in dimension 2 , then use the result in dimension 2 to achieve the result in dimension 3 . in the proof we use the notation @xmath380-move to indicate the move in which @xmath381 simplices are replaced by @xmath382 simplices .
so the @xmath383-move is the inverse move , and the order of @xmath381 and @xmath382 matter . in dimension 2
, we have two triangulations of the disk that agree on the boundary , and we are to show that they they can be arranged by pachner moves fixing the boundary to agree on the interior .
we prove the result by induction on the number of vertices on the boundary .
recall @xcite @xcite , that the _ star of a @xmath3-simplex _ ( in a simplicial complex ) is the union of all the simplices that contain the @xmath3-simplex .
the _ link of a @xmath3-simplex _ is the union of all the simplices in the star that do not contain the @xmath3-simplex .
we will examine the stars and links of vertices on the boundary of a disk ( and later on the bounday of a 3-ball ) .
so denote the star of @xmath116 with respect to the boundary by : @xmath384 .
similarly , the link of @xmath116 with respect to the boundary is @xmath385 while these sets with respect to the interior are @xmath386 and @xmath387 , respectively . in dimension 2
, @xmath384 is a pair of edges that share the vertex @xmath116 .
meanwhile , @xmath388 is a polygonal path properly embedded in the disk that is the most proximate to @xmath116 among all paths in the interior that join the points of @xmath385 .
we fix our consideration on one of the triangulations , say @xmath378 of @xmath389 .
we want to alter this triangulation so that @xmath388 is an edge ( so it has no interior vertices ) .
if we can achieve this alteration , then we can perform similar moves to @xmath379 .
the vertex @xmath116 on either triangulation then will become the vertex of a triangle that is attached to the disk along a single edge .
we can remove such a triangle ( or alternatively , work in the interior ) and apply induction on the number of vertices on the boundary .
consider an interior vertex , @xmath390 in @xmath369 .
if the star of @xmath390 in @xmath369 is the union of three triangles at @xmath390 , then we can remove this vertex from @xmath391 by means of a @xmath392-move .
perform such moves until there are no interior vertices of valence 3 . in this way
we may assume that a vertex , @xmath393 has valence larger than 3 .
if the valence of @xmath394 is greater than @xmath1 , then there are a pair of triangles in @xmath386 sharing edge @xmath395 upon which a pachner move of type @xmath396 can be performed .
such a move removes @xmath394 from the link of @xmath116 .
after such a move , check for interior vertices of valence 3 and remove them by type @xmath392-moves . in this way we can continue until the link of @xmath116 is an edge .
if @xmath391 is a triangle , then the process will reduce the triangulation until there are no interior vertices .
now we mimic the proof given in dimension 2 , to dimension 3 .
first , assume that an interior vertex @xmath390 in @xmath397 has as its star the union of 4 tetrahedra .
then we may eliminate such an interior vertex by means of a type @xmath398 pachner move .
consider the link , @xmath388 of a vertex , @xmath116 , on the boundary .
if this link is a triangle , then we may eliminate the vertex from the boundary , as in the 2-dimensional case .
for the star of @xmath116 is a single tetrahedron that is glued to the ball along a single face .
more generally , @xmath384 is a union of triangles forming a polygon , so @xmath384 is the cone on the polygon @xmath385 where @xmath116 is the cone point . consider the disk properly embedded in @xmath369 that is the link of @xmath116 .
this link , @xmath388 , is a triangulated disk .
there is a sequence of 2-dimensional pachner moves that change @xmath388 to a triangulation of an @xmath57-gon , with no interior vertices .
we use these 2-dimensional moves to determine @xmath1-dimensional moves performed in a neighborhood of @xmath386 as follows .
suppose that a @xmath392-move is used to simplify the disk that is the link of @xmath116 .
then consider the vertex @xmath390 at which such a move is performed . by our first step ,
its star is not the union of 4 tetrahedra .
three tetrahedra intersect along the edge , @xmath399 , and a @xmath400-move can be performed in the star of @xmath116 to remove the vertex @xmath390 from the link .
after such a move , then check for vertices in the interior whose valence is @xmath0 .
remove these by @xmath398-moves , until no such vertices remain .
potentially , some vertices from the link of @xmath116 are removed , and the effect of such a removal on the link is to perform a @xmath392-move .
in general a @xmath392-move to the link corresponds to a @xmath400-move to the star , or a @xmath398-move to a part of the star and a tetrahedron on the other side of the link .
if a @xmath396-move is used to simplify @xmath388 , then there is either a @xmath400-move or a @xmath75-move to that ball which induces it .
specifcally , if an edge in @xmath388 has as its star the union of 3 tetrahedra , then two of these are found in @xmath386 and the other one is on the other side of the link of @xmath116 . in this case
perform a @xmath400-move to @xmath369 .
the link of @xmath116 changes by a @xmath396-move .
if the link of the edge to be changed is more than 3 tetrahedra , then perform a @xmath75-move to @xmath386 . in this way
, a triangulation always results from these moves . after each such move , one must go and check for vertices of valence @xmath0 and remove them by @xmath398-moves .
eventually , we can remove all interior vertices from @xmath388 and we can further make sure that the link of @xmath116 is in some standard position .
we can remove @xmath116 from the boundary of @xmath369 by removing @xmath116 and the @xmath124-tetrahedra in its star where @xmath57 is the valence of @xmath116 with repsect to the boundary .
the result follows by induction .
@xmath92 * lemma . * [ inv3d ] the partition function @xmath356 does not depend on the choice of @xmath1-face triangulations of @xmath123 .
_ first let us analyze the case when a face @xmath245 is shared by four tetrahedra . then we will discuss the general case .
figures [ fourface ] and [ 4faces ] depict the case where a face @xmath245 is shared by four tetrahedra @xmath235 , @xmath280 , @xmath281 and @xmath319 .
then the dual complex has a rectangular @xmath2-face @xmath401 which is dual to the face @xmath245 .
there are two triangulations of a rectangle , say @xmath402 and @xmath403 , for @xmath401 .
( these are the triangulations that have no interior vertices ) .
the @xmath1-polytopes in @xmath123 that share @xmath401 are duals @xmath404 , @xmath405 and @xmath406 .
let @xmath378 and @xmath379 be @xmath1-face triangulations of @xmath123 that restrict to @xmath402 and @xmath403 respectively and restrict to the same triangulation on all the other @xmath2-faces of @xmath123 .
we show that the partition functions defined from @xmath378 and @xmath379 give the same value .
recall ( fig .
[ 3ass ] ) that two pairs of faces of a tetrahedron give two triangulation of a rectangle .
we attach a tetrahedron in between @xmath401 and @xmath404 and change the triangulations on the face .
more specifically , attach a pair of adjacent faces of a tetrahedron onto the face @xmath401 of @xmath404 along the triangulation @xmath378 restricted to @xmath402 .
perform the same attachment for @xmath405 and @xmath406 .
then we get a new triangulation @xmath407 of @xmath123 which restricts to @xmath408 on @xmath401 .
thus @xmath409 and @xmath379 have the same triangulation on the @xmath2-skeleton of @xmath123 by the assumption that all the other faces have the same triangulation .
thus @xmath409 and @xmath379 are related by a finite sequence of pachner moves fixing the boundary triagulation by lemma [ relpaclemma ] which does not change the partition function by the pentagon identity and the orthogonality of the cocycle @xmath348 .
hence it remains to prove that @xmath378 and @xmath409 give the same value of the partition function .
if the dual face @xmath401 is in general a polygon of more than four faces ( say @xmath57-gon ) , then triangulations consist of @xmath124 triangles ( by the condition of the definition of @xmath1-face triangulation ) .
such trianglations are related by only @xmath410-moves , which are realized by attaching a pair of faces of tetrahedra once at a time as above .
thus the above argument is applied to general cases by repeating the argument .
now we prove that @xmath378 and @xmath409 in the case @xmath401 is a rectangle give the same value of the partition function .
figure [ 4faces1 ] depicts the graphs for the triangulation . in fig .
[ 4faces1 ] the perturbations of these graphs to trivalent graphs are also depicted .
these perturbations correspond to @xmath411 ( triangulations of a rectangle @xmath401 in this case ) as depicted in the bottom of fig .
[ 4faces1 ] .
thus the colors assigned near the face @xmath245 with triangulations @xmath378 and @xmath409 are also assigned to edges of the perturbed graphs ( the right pictures of arrows in the figure , marked ( a ) , ( a ) , ( b ) , and ( b ) ) .
these graphs are identified by the following graphs in fig .
[ cocymovie2 ] : ( a ) corresponds to top graph , ( a ) to top left , ( b ) to bottom right , ( b ) to bottom .
the weights @xmath322 assigned to each triangulation are thus @xmath412 for @xmath378 and @xmath413 for @xmath409 ( or vice versa ) , if the group elements assigned are as indicated in fig .
[ cocymovie2 ] . now since @xmath409 is obtained from @xmath378 by attaching three tetrahedra , and they receive the weights @xmath414 , @xmath415 , @xmath416 , therefore the cocycle condition depicted in fig .
[ cocymovie2 ] shows that they are equal .
* overview of hopf categories .
* * review of 2-categories . *
a ( small ) strict @xmath2-category consists of the following data : \(1 ) a set of _ objects _ * obj * , \(2 ) a set of
_ @xmath185-morphisms _ * @xmath185-mor * , whose elements have _ source _ and _ target _ objects , \(3 ) a set of
_ @xmath2-morphisms _ * @xmath2-mor * , whose elements have _ source _ and _ target _ @xmath185-morphisms .
explicitly , given any two objects @xmath417 , there is a set of @xmath185-morphisms * @xmath185-mor*@xmath418 between them ; the object @xmath419 is called the source object , @xmath369 is the target .
such a @xmath185-morphism @xmath120 is also denoted by @xmath420 .
given any two 1-morphisms @xmath120 , @xmath421 * @xmath185-mor*@xmath418 there is a set of 2-morphisms between them ; @xmath120 is the _ source arrow ( @xmath185-morphism ) _ , and @xmath422 is _ the target arrow ( @xmath185-morphism ) . _
the object @xmath419 is the _ source object _ and the object @xmath369 is the _ target object _ ( of the @xmath2-morphism ) .
therefore both the sources and targets of @xmath120 and @xmath422 are required to coincide .
such a @xmath2-morphism is also denoted by @xmath423 .
( the _ smallness _ is the condition that these are sets . )
the following compositions of morphisms are defined and related as indicated in items ( 1 ) through ( 5 ) .
\(1 ) for any @xmath185-morphisms @xmath424 and @xmath425 , a @xmath185-morphism @xmath426 is the composite .
\(2 ) for any @xmath185-morphism @xmath427 and a @xmath2-morphism @xmath428 between @xmath185-morphisms @xmath429 , there is a composition ( a @xmath2-morphism ) @xmath430 . similarly , there is a 2-morphism @xmath431 when @xmath432 .
these compositions are depicted on the left and middle of fig .
[ compos ] .
\(3 ) for any @xmath2-morphisms @xmath423 and @xmath433 where @xmath434 , there is a composition @xmath435 .
this is depicted in fig .
[ compos ] right .
the composition @xmath436 is called the _ vertical composition _ of 2-morphisms .
\(4 ) the composition depicted in fig .
[ twosq ] is unambiguous in the sense that @xmath437 the result @xmath438 is called the _ horizontal composition _ of the 2-morphisms @xmath357 and @xmath348 .
\(5 ) the composition depicted in fig .
[ 4sq4real ] is also unambiguous in the sense that @xmath439 as 2-morphisms from @xmath440 to @xmath441 .
we assume further that @xmath442 and @xmath443 where @xmath444 are 1-morphisms , @xmath357 is a 2-morphism and these composites are defined . for any object @xmath419 ( resp .
@xmath185-morphism @xmath120 ) , the identity @xmath185-morphism @xmath445 ( resp .
@xmath2-morphism @xmath446 ) is defined .
the identity @xmath2-morphism satisfies @xmath447 , @xmath448 for any @xmath2-morphism @xmath357 . for any @xmath2-morphism @xmath449 where @xmath450 , @xmath451 and @xmath452 .
the following are the conditions for the _
strictness_. \(1 ) the compositions of @xmath185- and @xmath2-morphisms are associative ( @xmath453 , @xmath454 ) .
\(2 ) for any @xmath185-morphism @xmath420 , @xmath455 .
this concludes the definition of a small strict @xmath2-category .
note that by composing morphisms we can represent @xmath2-morphisms by planar polygons .
more on 2-categories can be found also in @xcite .
we follow the definitions in @xcite for the rest of this section .
a _ monoidal @xmath2-category _ consists of a @xmath2-category @xmath456 together with the data : ( 1 ) an object @xmath457 , ( 2 ) for any object @xmath419 two @xmath2-functors @xmath458 and @xmath459 such that @xmath460 for any objects @xmath419 , @xmath369 , and ( 3 ) for any two @xmath185-morphisms @xmath461 and @xmath462 a @xmath2-morphism @xmath463 these data satisfy 8 conditions that we omit here . a _ @xmath2-vector space _ is a @xmath3-linear additive category @xmath464 that admits a subset @xmath465 such that ( 1 ) any object @xmath466 is simple , ( 2 ) for any @xmath466 , @xmath467 , and ( 3 ) for any object @xmath468 there is a unique finite subset @xmath469 such that @xmath470 . the set @xmath369 is called a basis and a @xmath471vector space is finite if @xmath472 is finite .
the @xmath2-category of finite dimensional @xmath2-vector spaces is denoted by @xmath473 .
the set of @xmath185-morphisms are @xmath3-linear functors , and @xmath2-morphisms are natural transformations .
this is a subcategory of the @xmath2-category @xmath474 of small @xmath3-linear additive categories with the @xmath3-linear functors as @xmath185-morphisms .
theorem 2.7 of @xcite says that @xmath473 admits the structure of a strongly involutory moniodal 2-category in such a way that for @xmath475vect , obj@xmath476 are pairs of objects of @xmath477 and @xmath478 , mor@xmath479 . a _ comoidal category _
is a category @xmath477 in @xmath474 with the following data .
two @xmath3-linear functors @xmath480 and @xmath481 where vec = the category of vector spaces , and some natural isomorphisms , including @xmath482 , satisfying certain relations , including the pentagon relation for @xmath357 .
a _ @xmath2-bialgebra _ or a _ bimonoidal category _ is a category @xmath483 which has both monoidal and comoidal category structures and with additional data including a natural isomorphism @xmath484 satisfying certain conditions , where @xmath168 denotes permutations , including figures depicted in [ sjcube ] , [ coh1 ] , and [ coh2 ] .
the faces in the figures are @xmath2-morphisms , and these cubes are relations among @xmath2-morphisms in the hopf category . * cocycle conditions and hopf categories . * * diagrams for morphisms
. * * coherence cubes and diagrams . *
in this paper we established diagrammatic machinery for the study of @xmath0-manifold invariants using triangulations and graphs .
in particular , invariance under pachner moves of crane - frenkel invariants for cocycles constructed by crane - yetter is proved by using graphs .
this strongly suggests generalizations of the dikgraaf - witten invariant to @xmath0-manifolds using cocycles defined in @xcite .
we have shown direct relations among algebraic structures ( hopf categories ) , triangulations , and ( graph ) diagrams in dimension @xmath0 , generalizing spin network theory in @xmath1-dimensions .
further study on higher dimensional tqfts and higher algebraic structures are anticipated .
we expect that our diagrammatic machinery established in this paper serve as tools for further developments in the area .
open questions remain . which finite groups contain cocycles that satisfy the symmetry conditions ?
can other examples of hopf categories be constructed .
what do the invariants from this construction measure ? can these invariants be related to invariants that arise from donaldson theory ?
we are grateful to j barrett , l. crane , and d. yetter for valuable conversations .
the first named author was supported by the nsa while some of the research for this paper was being conducted .
the second named author is partially supported by nsf dms-2528707 .
the third named author is partially supported by the university of south florida research and creative scholarship grant program under grant number 1249932r0 .
barrett , j. and westbury , b. _ invariants of piecewise - linear 3-manifolds , _ 24 june 1994 , hep - th/9311155v3 8aug 1995 . and _ invariants of piecewise - linear @xmath1-manifolds _ , trans .
( 1996 ) , no .
10 , 39974022 .
carter , scott , kauffman , louis h. , and saito , masahico , _ diagrammatics , singularities , and their algebraic interpretations , _ to appear in proceedings of x encontro brasileiro de topologia , brasilian mathematical society . | crane and frenkel proposed a state sum invariant for triangulated 4-manifolds .
they defined and used new algebraic structures called hopf categories for their construction .
crane and yetter studied hopf categories and gave some examples using group cocycles that are associated to the drinfeld double of a finite group . in this paper
we define a state sum invariant of triangulated @xmath0-manifolds using crane - yetter cocycles as boltzmann weights .
our invariant generalizes the 3-dimensional invariants defined by dijkgraaf and witten and the invariants that are defined via hopf algebras .
we present diagrammatic methods for the study of such invariants that illustrate connections between hopf categories and moves to triangulations .
[ section ] [ sect ] |
besides the many existing radioactive beam facilities and their upgrades , many more are being constructed or under planning , including the cooling storage ring ( csr ) facility at hirfl in china @xcite , the radioactive ion beam ( rib ) factory at riken in japan @xcite , the fair / gsi in germany @xcite , spiral2/ganil in france @xcite , and the facility for rare isotope beams ( frib ) in the usa @xcite .
these new facilities offer the possibility to study the properties of nuclear matter or nuclei under the extreme condition of large isospin asymmetry . as a result ,
the study of the isospin degree of freedom in nuclear physics has recently attracted much attention .
the ultimate goal of such study is to extract information on the isospin dependence of in - medium nuclear effective interactions as well as the equation of state ( eos ) of isospin asymmetric nuclear matter , particularly its isospin - dependent term or the density dependence of the nuclear symmetry energy .
this knowledge , especially the latter , is important for understanding not only the structure of radioactive nuclei , the reaction dynamics induced by rare isotopes , and the liquid - gas phase transition in asymmetric nuclear matter , but also many critical issues in astrophysics @xcite .
unfortunately , the density dependence of the nuclear symmetry energy , especially its behavior at high densities , is largely unknown and is regarded as the most uncertain among all the properties of isospin asymmetric nuclear matter . although the nuclear symmetry energy at normal nuclear matter density @xmath1 @xmath2 is known to be around @xmath3 mev from the empirical liquid - drop mass formula @xcite , its values at other densities are poorly known @xcite .
various microscopic and phenomenological models , such as the relativistic dirac - brueckner - hartree - fock ( dbhf ) @xcite and the non - relativistic brueckner - hartree - fock ( bhf ) @xcite approach , the relativistic mean - field ( rmf ) model based on nucleon - meson interactions @xcite , and the non - relativistic mean - field model based on skyrme - like interactions @xcite , have been used to study the isospin - dependent properties of asymmetric nuclear matter , such as the nuclear symmetry energy , the nuclear symmetry potential , the isospin - splitting of nucleon effective mass , etc . , but the predicted results vary widely .
in fact , even the sign of the symmetry energy above @xmath4 is uncertain @xcite .
the theoretical uncertainties are mainly due to the lack of knowledge about the isospin dependence of in - medium nuclear effective interactions and the limitations in the techniques for solving the nuclear many - body problem . as to the incompressibility of asymmetric nuclear matter , it is essentially undetermined @xcite , even after about 30 years of studies . for comparison , the incompressibility of symmetric nuclear matter at its saturation density
@xmath5 has been determined to be @xmath6 mev from the nuclear gmr @xcite and the eos at densities of @xmath7 has also been constrained by measurements of collective flows in nucleus - nucleus collisions @xcite . as a phenomenological approach ,
the rmf model has achieved great success during the last decade in describing many nuclear phenomena @xcite .
for example , it provides a novel saturation mechanism for the nuclear matter , an explanation of the strong spin - orbit interaction in finite nuclei , a natural energy dependence of the nucleon optical potential , etc .
the rmf approach is generally based on effective interaction lagrangians that involve nucleon and meson fields . in this approach
, a number of parameters are adjusted to fit the properties of many nuclei . as such
, these models usually give excellent descriptions of nuclear properties around or below the saturation density .
since the original lagrangian proposed by walecka more than @xmath3 years ago @xcite , there have been a lot of different treatments , extensions , and applications of the rmf model .
the three main versions are the nonlinear models @xcite , models with density - dependent meson - nucleon couplings @xcite , and point - coupling models without mesons @xcite . for each version of the rmf model ,
there are also many different parameter sets with their values fitted to the binding energies and charge radii of a large number of nuclei in the periodic table .
including isovector mesons in the effective interaction lagrangians further allows the rmf model to describe successfully the properties of nuclei far from the @xmath8-stability line . with recent developments in constraining the isospin - dependent properties of asymmetric nuclear matter , especially the density dependence of the nuclear symmetry energy ,
it is of great interest to see to what extend the results from different versions of the rmf model are consistent with these constrains . in the present work , based on
commonly used @xmath0 different parameter sets in three different versions of the rmf model , we carry out a systematic study of the isospin - dependent bulk and single - particle properties of asymmetric nuclear matter . in particular , we study the density dependence of the nuclear symmetry energy from these rmf models and compare the results with the constraints recently extracted from analyses of the isospin diffusion data from heavy - ion collisions based on the isospin and momentum - dependent ibuu04 transport model with in - medium nucleon - nucleon ( nn ) cross sections @xcite , the isoscaling analyses of isotope ratios in intermediate energy heavy ion collisions @xcite , and measured isotopic dependence of the giant monopole resonances ( gmr ) in even - a sn isotopes @xcite . among these @xmath0
commonly used interactions in nuclear structure studies , only a few are found to give symmetry energies that are consistent with the empirically extracted one .
furthermore , we study the nuclear symmetry potential and the isospin - splitting of the nucleon effective mass in isospin asymmetric nuclear matter .
our results indicate that the nuclear symmetry potential at fixed baryon density may increase or decrease with increasing nucleon momentum depending on the definition of the nucleon optical potential and the interactions used .
this dependence is also seen in the isospin - splitting of the nucleon effective mass in neutron - rich nuclear matter .
in addition , the isospin - splitting of the nucleon scalar density in neutron - rich nuclear matter is also studied .
the paper is organized as follows . in section [ isospin ]
, we discuss some isospin - dependent bulk and single - particle properties of asymmetric nuclear matter , such as the nuclear symmetry energy , the nuclear symmetry potential , and the isospin - splitting of nucleon effective mass as well as current experimental and/or empirical constraints on these quantities .
the theoretical frameworks for the different versions of rmf models , i.e. , the nonlinear rmf models , the models with density - dependent nucleon - meson coupling , and the nonlinear and density - dependent point - coupling models , are briefly reviewed in section [ rmf ] .
results on the isospin - dependent properties of asymmetric nuclear matter , i.e. , the nuclear symmetry energy , the nuclear symmetry potential , and the isospin - splitting of nucleon effective mass and the nucleon scalar densities in neutron - rich nuclear matter , from different versions of rmf models are presented and discussed in section [ results ] . a summary
is then given in section [ summary ] .
for completeness , the isospin- and momentum - dependent mdi interaction , which will be used as a reference in some cases for comparison , is briefly described in appendix [ mdi ] .
the eos of isospin asymmetric nuclear matter , given by its binding energy per nucleon , can be generally written as @xmath9where @xmath10 is the baryon density with @xmath11 and @xmath12 denoting the neutron and proton densities , respectively ; @xmath13 is the isospin asymmetry ; @xmath14 is the binding energy per nucleon in symmetric nuclear matter , and @xmath15 is the nuclear symmetry energy .
the absence of odd - order terms in @xmath16 in eq .
( [ esympara ] ) is due to the exchange symmetry between protons and neutrons in nuclear matter when one neglects the coulomb interaction and assumes the charge symmetry of nuclear forces .
the higher - order terms in @xmath17 are negligible , e.g. , the magnitude of the @xmath18 term at @xmath19 is estimated to be less than @xmath20 mev @xcite . neglecting the contribution from higher - order terms in eq .
( [ esympara ] ) leads to the well - known empirical parabolic law for the eos of asymmetric nuclear matter , which has been verified by all many - body theories to date , at least for densities up to moderate values . as a good approximation
, the density - dependent symmetry energy @xmath21 can be extracted from @xmath22 , i.e. , the energy change per nucleon when all protons in the symmetric nuclear matter are converted to neutrons while keeping the total nuclear density fixed . in this sense
, the nuclear symmetry energy gives an estimation of the binding energy difference between the pure neutron matter without protons and the symmetric nuclear matter with equal numbers of protons and neutrons .
it should be mentioned that the possible presence of the higher - order terms in @xmath23 at supra - normal densities can significantly modify the proton fraction in @xmath24-equilibrium neutron - star matter and the critical density for the direct urca process which can lead to faster cooling of neutron stars @xcite . around the nuclear matter saturation density @xmath19 , the nuclear symmetry energy @xmath21 can be expanded to second - order in density as @xmath25 where @xmath26 and @xmath27 are the slope and curvature parameters of the nuclear symmetry energy at @xmath19 , i.e. , @xmath28 the @xmath26 and @xmath27 characterize the density dependence of the nuclear symmetry energy around normal nuclear matter density , and thus carry important information on the properties of nuclear symmetry energy at both high and low densities .
in particular , the slope parameter @xmath26 has been found to correlate linearly with the neutron - skin thickness of heavy nuclei and thus can in principle be determined from measured thickness of the neutron skin of such nuclei @xcite .
unfortunately , because of the large uncertainties in the experimental measurements , this has not yet been possible so far . at the nuclear matter saturation density and around @xmath29 , the isobaric incompressibility of asymmetric nuclear matter can also be expressed to second - order in @xmath16 as @xcite @xmath30 where @xmath31 is the incompressibility of symmetric nuclear matter at the nuclear matter saturation density and the isospin - dependent part @xcite @xmath32characterizes the density dependence of the nuclear symmetry energy .
information on @xmath33 can in principle be extracted experimentally by measuring the gmr in neutron - rich nuclei .
earlier attempts based on this method have given , however , widely different values .
for example , a value of @xmath34 mev with a large uncertainty was obtained in ref .
@xcite from a systematic study of the gmr in the isotopic chains of sn and sm . in this analysis
, the value of @xmath31 was found to be @xmath35 mev , which is somewhat larger than the commonly accepted value of @xmath36 mev . in a later study ,
an even less stringent constraint of @xmath37 mev was extracted from the gmr of finite nuclei , depending on the mass region of nuclei and the number of parameters used in parameterizing the incompressibility of finite nuclei @xcite .
most recently , a much stringent constraint of @xmath38 mev has been obtained in ref .
@xcite from measurements of the isotopic dependence of the gmr in even - a sn isotopes . besides studies of nuclear structure , heavy - ion reactions , especially those induced by radioactive beams , also provide a useful means to investigate in terrestrial laboratories the isospin - dependent properties of asymmetric nuclear matter , particularly the density dependence of the nuclear symmetry energy .
indeed , significant progress has recently been made both experimentally and theoretically in extracting the information on the behaviors of nuclear symmetry energy at sub - saturation density from the isospin diffusion data in heavy - ion collisions from the nscl / msu @xcite . using the isospin and momentum - dependent ibuu04 transport model with in - medium nn cross sections ,
the isospin diffusion data were found to be consistent with a density - dependent symmetry energy of @xmath39 with @xmath40 at subnormal density @xcite , which has led to the extraction of a value of @xmath41 mev for the slope parameter of the nuclear symmetry energy at saturation density and a value of @xmath42 mev for the isospin - dependent part of the isobaric incompressibility of isospin asymmetric nuclear matter @xcite .
this has further imposed stringent constraints on both the parameters in the isospin - dependent nuclear effective interactions and the neutron skin thickness of heavy nuclei . among the @xmath43 sets of skyrme interactions commonly used in nuclear structure studies ,
only the @xmath44 sets siv , sv , g@xmath45 , and r@xmath45 have been found to give symmetry energies that are consistent with above extracted one .
using these skyrme interactions , the neutron - skin thickness of heavy nuclei calculated within the hartree - fock approach is consistent with available experimental data @xcite and also that from a relativistic mean - field model based on an accurately calibrated parameter set that reproduces the gmr in @xmath46zr and @xmath47pb as well as the isovector giant dipole resonance of @xmath47pb @xcite .
the extracted symmetry energy further agrees with the symmetry energy @xmath48 recently obtained from the isoscaling analyses of isotope ratios in intermediate energy heavy ion collisions @xcite , which gives @xmath49 mev and @xmath50 mev .
the extracted value of @xmath42 mev from the isospin diffusion data is also consistent with the value @xmath38 mev obtained from recently measured isotopic dependence of the gmr in even - a sn isotopes @xcite .
we note that the gmr only allows us to extract the value of @xmath33 but not that of @xmath26 .
these empirically extracted values for @xmath26 and @xmath51 represent the best and most stringent phenomenological constraints available so far on the nuclear symmetry energy at sub - normal densities .
although the behavior of the symmetry energy at high densities is presently largely undetermined , much of this information is expected to be obtained from future high energy radioactive beam facilities .
the nuclear symmetry potential refers to the isovector part of the nucleon mean - field potential in isospin asymmetric nuclear matter . besides the nuclear density
, the symmetry potential of a nucleon in nuclear matter also depends on the momentum or energy of the nucleon .
the nuclear symmetry potential is different from the nuclear symmetry energy as the latter involves the integration of the isospin - dependent mean - field potential of a nucleon over its momentum .
the nuclear symmetry potential is thus a dynamical quantity while the nuclear symmetry energy is a thermodynamic quantity , and both are important for understanding many physics questions in nuclear physics and astrophysics .
various microscopic and phenomenological models have been used to study the symmetry potential @xcite , and the predicted results vary widely as in the case of the nuclear symmetry energy . in particular , whereas most models predict a decreasing symmetry potential with increasing nucleon momentum albeit at different rates , a few nuclear effective interactions used in some models give an opposite behavior .
there are different methods to derive the real part of the non - relativistic optical potential based on the dirac equation with lorentz covariant nucleon vector and scalar self - energies .
the most popular one is the so - called schrdinger - equivalent potential " ( sep ) . from the nucleon scalar self - energy @xmath57 and the time - like component of the vector self - energy @xmath58
, the schrdinger - equivalent potential is given by @xcite : @xmath59+\frac{\sigma _ { \tau } ^{0 } } { m_{\tau } } e_{\tau } \notag \\ & = & \sigma _ { \tau } ^{s}+\sigma _ { \tau } ^{0}+\frac{1}{2m_{\tau } } [ ( \sigma _ { \tau } ^{s})^{2}-(\sigma _ { \tau } ^{0})^{2 } ] + \frac{\sigma _ { \tau } ^{0}}{m_{\tau } } e_{\mathrm{kin}},\notag\\ \label{usep}\end{aligned}\]]where @xmath60 is the kinetic energy of a nucleon , i.e. , @xmath61 with @xmath62 being its total energy .
eq.([usep ] ) shows that @xmath63 increases linearly with the nucleon energy @xmath62 or kinetic energy @xmath60 if the nucleon self - energies are independent of energy .
we note that by construction solving the schrdinger equation with above sep gives same bound - state energy eigenvalues and elastic phase shifts as the solution of the upper component of the dirac spinor in the dirac equation with same nucleon scalar self - energy and time - like component of the vector self - energy @xcite .
the above sep thus best represents the real part of the nucleon optical potential in non - relativistic models @xcite .
the corresponding nuclear symmetry potential is given by @xmath64 with @xmath16 being the isospin asymmetry .
another popular alternative for deriving the non - relativistic nucleon optical potential in relativistic models is to take it as the difference between the total energy @xmath62 of a nucleon with momentum @xmath65 in the nuclear medium and its energy at the same momentum in free space @xcite , i.e. , @xmath66 in obtaining the last step in above equation , the dispersion relation @xmath67has been used .
this definition for the nucleon optical potential has also been extensively used in microscopic dbhf calculations @xcite and transport models for heavy - ion collisions @xcite . for energy - independent nucleon self - energies ,
@xmath68 approaches a constant value of @xmath69 when @xmath70 , unlike the linear increase of @xmath63 with the nucleon energy . for @xmath71
, we have @xmath72 while @xmath73 .
therefore , @xmath68 displays a more reasonable high energy behavior than @xmath63 .
we note that unlike @xmath63 , @xmath68 does not give the same bound - state energy eigenvalues and elastic phase shifts as the solution of the upper component of the dirac equation . as in the case of @xmath63 , the symmetry potential in this approach
is defined by @xmath74 in ref .
@xcite , another optical potential was introduced based on the second - order dirac ( sod ) equation , and it corresponds to multiplying eq.([usep ] ) by the factor @xmath75 , i.e. , @xmath76+\frac{\sigma _
{ \tau } ^{0 } } { m_{\tau } } e_{\tau } ] \frac{m_{\tau } } { e_{\tau } } \notag \\ & = & \sigma _ { \tau } ^{0}+\frac{m_{\tau } } { e_{\tau } } \sigma _ { \tau } ^{s}+\frac{1}{2e_{\tau } } [ ( \sigma _ { \tau } ^{s})^{2}-(\sigma _ { \tau } ^{0})^{2 } ] .
\label{usd}\end{aligned}\ ] ] for energy - independent nucleon self - energies , @xmath77 has the same asymptotical value of @xmath58 as @xmath68 when @xmath78 . for @xmath71
, we have @xmath79 $ ] . the symmetry potential based on the optical potential of eq .
( [ usd ] ) is given by @xmath80 above discussions thus show that the optical potentials defined in eqs .
( [ uopt ] ) and ( [ usd ] ) have similar high energy behaviors , but they may be very different from that defined in eq .
( [ usep ] ) .
if we assume that @xmath81 and @xmath82 , which has been shown to be generally valid in the rmf model even at higher baryon densities , we have , however , @xmath83 at low momenta ( @xmath84 ) , indicating that above three definitions for the optical potential in the rmf model behave similarly at low energies . however , it should be stressed that , among the three optical potentials defined above , only @xmath63 is obtained from a well - defined theoretical procedure and is schrdinger - equivalent while @xmath68 and @xmath85 are used here for heuristic reasons as they are of practical interest in microscopic dbhf calculations , transport models for heavy - ion collisions , and dirac phenomenology .
empirically , a systematic analysis of a large number of nucleon - nucleus scattering experiments and ( p , n ) charge - exchange reactions at beam energies up to about @xmath86 mev has shown that the data can be very well described by the parametrization @xmath87 with @xmath88 mev and @xmath89 @xcite .
although the uncertainties in both parameters @xmath90 and @xmath91 are large , the nuclear symmetry potential at nuclear matter saturation density , i.e. , the lane potential @xmath92 @xcite , clearly decreases approximately linearly with increasing beam energy @xmath60 .
this provides a stringent constraint on the low energy behavior of the nuclear symmetry potential at saturation density .
as we will see in the following , although the predicted energy dependence of nuclear symmetry potential at low energy from the rmf models does not agree with the empirical lane potential , it is consistent with results from microscopic dbhf @xcite , the extended bhf with 3-body forces @xcite , and chiral perturbation theory calculations @xcite , which give a lane potential that also stays as a constant or increases slightly with momentum for nucleons with momenta less than about @xmath93 mev / c or with kinetic energies @xmath94 but decreases with momentum when the momentum is larger than about @xmath93 mev / c .
recently , the high energy behavior of the nuclear symmetry potential has been studied in the relativistic impulse ( @xmath95-@xmath96 ) approximation based on the empirical nn scattering amplitude @xcite .
the results indicate that the nuclear symmetry potential derived from the schrdinger - equivalent potential at a fixed density becomes almost constant when the nucleon kinetic energy is greater than about @xmath97 mev , independent of the parameters used in the analysis .
it is further shown that for such high energy nucleons the nuclear symmetry potential is slightly negative at baryon densities below about @xmath98 @xmath2 and then increases almost linearly to positive values at high densities .
these results provide important constraints on the high energy behavior of the nuclear symmetry potential in asymmetric nuclear matter .
furthermore , with the love - franey nn scattering amplitude developed by murdock and horowitz @xcite , the intermediate - energy ( @xmath99 @xmath100 mev ) behavior of the nuclear symmetry potential constructed from the schrdinger - equivalent potential in isospin asymmetric nuclear matter has also been investigated recently @xcite .
it shows that the nuclear symmetry potential at fixed baryon density decreases with increasing nucleon energy .
in particular , the nuclear symmetry potential at saturation density changes from positive to negative values at nucleon kinetic energy of about @xmath101 mev .
such an energy and density dependence of the nuclear symmetry potential is consistent with those from the isospin- and momentum - dependent mdi interaction with @xmath102 ( see appendix [ mdi ] for details on this interaction ) .
these results thus provide an important consistency check for the energy / momentum dependence of the nuclear symmetry potential in asymmetric nuclear matter , particularly the mdi interaction with @xmath102 . on the other other ,
the low energy behavior of the nuclear symmetry potential at densities away from normal nuclear density is presently not known empirically .
experimental determination of both the density and momentum dependence of the nuclear symmetry potential is thus of great interest , and heavy - ion reactions with radioactive beams provides a unique tool to extract this information in terrestrial laboratories .
many different definitions for the nucleon effective mass can be found in the literature @xcite . in the present work
, we mainly focus on the following three effective masses : the dirac mass @xmath103 ( also denoted as @xmath104 in the present work ) , the landau mass @xmath105 , and the lorentz mass @xmath106 .
the dirac mass @xmath103 is defined through the nucleon scalar self - energy in the dirac equation , i.e. , @xmath107 it is directly related to the spin - orbit potential in finite nuclei and is thus a genuine relativistic quantity without non - relativistic correspondence .
we note that the difference between the nucleon vector and scalar self - energies determines the spin - orbit potential , whereas their sum defines the effective single - nucleon potential and is constrained by the nuclear matter binding energy at saturation density . from the energy spacings between spin - orbit partner states in finite nuclei ,
the constraint @xmath108 @xmath109 has been obtained on the value of the dirac mass @xcite .
the landau mass @xmath105 is defined as @xmath110 in terms of the single - particle density of state @xmath111 at energy @xmath62 and thus characterizes the momentum dependence of the single - particle potential . in the relativistic model ,
it is given by @xcite @xmath112 since @xmath113 is in principle measurable , the landau mass from the relativistic model should have a comparable value as that in the non - relativistic model .
empirically , based on non - relativistic effective interactions such as the skyrme - type interactions , calculations of the ground - state properties and the excitation energies of quadrupole giant resonances have shown that a realistic choice for the nucleon landau mass is @xmath114 = @xmath115 @xcite .
the smaller landau mass than that of nucleon free mass would lead to a smaller level density at the fermi energy and much spreaded single - particle levels in finite nuclei @xcite .
the lorentz mass @xmath116 characterizes the energy dependence of the schrdinger - equivalent potential @xmath63 in the relativistic model and is defined as @xcite @xmath117 it has been argued in ref . @xcite
that it is the lorentz mass @xmath116 that should be compared with the usual non - relativistic nucleon effective mass extracted from analyses carried out in the framework of non - relativistic optical and shell models .
it can be easily seen that in the non - relativistic approximation ( @xmath118 ) , the lorentz mass @xmath116 reduces to the landau mass @xmath119 . in relativistic models , the nucleon effective mass has sometimes also been introduced via the energy dependence of the optical potential in eq .
( [ uopt ] ) and the second - order dirac optical potential in eq .
( [ usd ] ) , i.e. , @xmath120 and @xmath121 , \label{msod}\end{aligned}\ ] ] respectively .
the isospin - splitting of nucleon effective mass in asymmetric nuclear matter , i.e. , the difference between the neutron and proton effective masses is currently not known empirically @xcite .
previous theoretical investigations have indicated that most rmf calculations with the isovector @xmath122 meson predict @xmath123 while in the microscopic dbhf approach , @xmath124 can be larger or smaller than @xmath125 depending on the approximation schemes and methods used for determining the lorentz and isovector structure of the nucleon self - energy @xcite . for the nucleon lorentz mass ,
the microscopic dbhf or bhf approach and most non - relativistic skyrme - hartree - fock calculations predict @xmath126 , while most rmf and a few skyrme - hartree - fock calculations give opposite predictions .
for completeness , we briefly introduce in the following the main ingredients in the nonlinear rmf model , the density - dependent rmf model , the nonlinear point - coupling rmf model , and the density - dependent point - coupling rmf model .
we neglect the electromagnetic field in the following since in the present work we are interested in the properties of the infinite nuclear matter . furthermore , besides the mean - field approximation in which operators of meson fields are replaced by their expectation values ( the fields are thus treated as classical c - numbers ) , we also use the non - sea approximation which neglects the effect due to negative energy states in the dirac sea .
the lagrangian density in the nonlinear rmf model generally includes the nucleon field @xmath127 , the isoscalar - scalar meson field @xmath128 , the isoscalar - vector meson field @xmath129 , the isovector - vector meson field @xmath130 , and the isovector - scalar meson field @xmath122 , i.e. , @xmath131 \psi \notag \\ & & + \frac{1}{2}(\partial _ { \mu } \sigma \partial ^{\mu } \sigma -m_{\sigma } ^{2}\sigma ^{2})-\frac{1}{4}\omega _ { \mu \nu } \omega ^{\mu \nu } + \frac{1}{2 } m_{\omega } ^{2}\omega _ { \mu } \omega ^{\mu } \notag \\ & & -\frac{1}{3}b_{\sigma } m{(g_{\sigma } \sigma ) } ^{3}-\frac{1}{4}c_{\sigma } { \ ( g_{\sigma } \sigma ) } ^{4}+\frac{1}{4}c_{\omega } { ( g_{\omega } ^{2}\omega _ { \mu } \omega ^{\mu } ) } ^{2 } \notag \\ & & + \frac{1}{2}(\partial _ { \mu } \vec{\delta}\cdot \partial ^{\mu } \vec{\delta } -m_{\delta } ^{2}\vec{\delta}^{2})+\frac{1}{2}m_{\rho }
^{2}\vec{\rho}_{\mu } \cdot \vec{\rho}^{\mu } -\frac{1}{4}\vec{\rho}_{\mu \nu } \cdot \vec{\rho } ^{\mu \nu } \notag \\ & & + \frac{1}{2}(g_{\rho } ^{2}\vec{\rho}_{\mu } \cdot \vec{\rho}^{\mu } ) ( \lambda _ { s}g_{\sigma } ^{2}\sigma ^{2}+\lambda _ { v}g_{\omega } ^{2}{\omega _ { \mu } \omega ^{\mu } } ) \notag \\ & & -g_{\rho } \vec{\rho}_{\mu } \cdot \bar{\psi}\gamma ^{\mu } \vec{\tau}\psi + g_{\delta } \vec{\delta}\cdot \bar{\psi}\vec{\tau}\psi \ ; , \label{lagnl}\end{aligned}\ ] ] where the antisymmetric field tensors @xmath132 and @xmath133 are given by @xmath134 and @xmath135 , respectively , and other symbols have their usual meanings .
also , vectors in isospin space are denoted by arrows .
this model also contains cross interactions between the isovector meson @xmath136 and isoscalar @xmath137 and @xmath129 mesons through the cross - coupling constants @xmath138 and @xmath139 @xcite .
in addition , we include the isovector - scalar channel ( @xmath122 meson ) which is important for the saturation of asymmetric nuclear matter and has also been shown to be an important degree of freedom in describing the properties of asymmetric nuclear matter @xcite .
the above lagrangian density is quite general and allows us to use most of presently popular parameter sets in the nonlinear rmf model . from the standard euler - lagrange formalism
, we can deduce from the lagrangian density equations of motion for the nucleon and meson fields .
the resulting dirac equation for the nucleon field is @xmath140 \psi = 0\;,\ ] ] with the following nucleon scalar and vector self - energies : @xmath141 for the isoscalar meson fields @xmath137 and @xmath129 , they are described by the klein - gordon and proca equations , respectively , i.e. , @xmath142\ ; , \\
\partial _ { \mu } \omega ^{\mu \nu } + m_{\omega } ^{2}\omega ^{\nu } & = & g_{\omega } [ \bar{\psi}\gamma ^{\nu } \psi -c_{\omega } g_{\omega } ^{3}(\omega _ { \mu } \omega ^{\mu } \omega ^{\nu } ) \notag \\ & & -\lambda _ { v}g_{\rho }
^{2}\vec{\rho}_{\mu } \cdot \vec{\rho}^{\mu } g_{\omega } \omega ^{\nu } ] \;.\end{aligned}\ ] ] analogous equations for the isovector @xmath122 and @xmath136 meson fields are @xmath143.\end{aligned}\ ] ] for a static , homogenous infinite nuclear matter , all derivative terms drop out and the expectation values of space - like components of vector fields vanish ( only zero components @xmath144 and @xmath145 survive ) due to translational invariance and rotational symmetry of the nuclear matter .
in addition , only the third component of isovector fields ( @xmath146 and @xmath147 ) needs to be taken into consideration due to the rotational invariance around the third axis in the isospin space . in the mean - field approximation ,
meson fields are replaced by their expectation values , i.e. , @xmath148 , @xmath149 , @xmath150 , and @xmath151 , and the meson field equations are reduced to @xmath152 , \\
m_{\omega } ^{2}\bar{\omega}_{0 } & = & g_{\omega } [ \rho _ { b}-c_{\omega } { ( g_{\omega } \bar{\omega}_{0})}^{3 } \notag \\ & & -\lambda { ( g_{\omega } \bar{\omega}_{0})(}g_{\rho } \bar{\rho } _ { 0}^{(3)})^{2 } ] , \label{omgnl } \\
m_{\delta}^{2}{\bar{\delta}}^{(3 ) } & = & g_{\delta } ( \rho _ { s , p}-\rho _ { s , n } )
. \label{delnl } \\
m_{\rho } ^{2}\bar{\rho}_{0}^{(3 ) } & = & g_{\rho } [ \rho _ { b , p}-\rho _ { b , n}-\lambda _ { s}{(}g_{\rho } \bar{\rho}_{0}^{(3)}){(g_{\sigma } \sigma ) } ^{2 } \notag \\ & & -\lambda _
{ v}{(}g_{\rho } \bar{\rho}_{0}^{(3)}){(g_{\omega } \bar{\omega}_{0})}^{{2 } } ] .
\label{rhonl}\end{aligned}\ ] ] in the above , the nucleon scalar density @xmath153 is defined as @xmath154 with the proton ( @xmath155 ) and neutron ( @xmath156 ) scalar densities given by @xmath157 ,
, n \notag\\ \label{rhosnp}\end{aligned}\ ] ] where @xmath158 with @xmath159 and @xmath160 denoting the proton and neutron dirac masses , respectively , i.e. , @xmath161 the nucleon scalar and vector self - energies are then given by @xmath162 with @xmath163 and @xmath164 for protons and neutrons , respectively .
the set of coupled equations for the nucleon and meson fields can be solved self - consistently using the iteration method , and the properties of the nuclear matter can then be obtained from these fields . from the resulting energy - momentum tensor , we can calculate the energy density @xmath165 and pressure @xmath166 of asymmetric nuclear matter , and the results are given by @xmath167 \notag \\ & & + \frac{1}{3}b_{\sigma } m{(g_{\sigma } \bar{\sigma})}^{3}+\frac{1}{4}c_{\sigma } { ( g_{\sigma } \bar{\sigma})}^{4}+\frac{3}{4}c_{\omega } { ( g_{\omega } \bar{\omega}_{0})}^{4 } \notag \\ & & + \frac{1}{2}(g_{\rho } \bar{\rho}_{0}^{(3)})^{2}[\lambda _ { s}{(g_{\sigma } \bar{\sigma})}^{2}+3\lambda _ { v}{(g_{\omega } \bar{\omega}_{0})}^{2}]\end{aligned}\ ] ] and @xmath168 \notag \\ & & -\frac{1}{3}b_{\sigma } m{(g_{\sigma } \bar{\sigma})}^{3}-\frac{1}{4 } c_{\sigma } { ( g_{\sigma } \bar{\sigma})}^{4}+\frac{1}{4}c_{\omega } { ( g_{\omega } \bar{\omega}_{0})}^{4 } \notag \\ & & + \frac{1}{2}(g_{\rho } \bar{\rho}_{0}^{(3)})^{2}[\lambda _ { s}{(g_{\sigma } \bar{\sigma})}^{2}+\lambda _ { v}{(g_{\omega } \bar{\omega}_{0})}^{2}].\end{aligned}\ ] ] in the above , @xmath169 and @xmath170 are , respectively , the kinetic contributions to the energy densities and pressure of protons and neutrons in nuclear matter , and they are given by @xmath171,\quad i = p , n,\end{aligned}\ ] ] and @xmath172,\quad i = p , n.\end{aligned}\ ] ] the binding energy per nucleon can be obtained from the energy density via @xmath173 while the symmetry energy is given by @xmath174 } , \label{esymnl}\end{aligned}\ ] ] with the effective @xmath136-meson mass given by @xcite @xmath175\ ] ] and @xmath176 where @xmath177 and @xmath178 is the nucleon dirac mass in symmetric nuclear matter . in the density - dependent rmf model , instead of introducing terms involving self - interactions of the scalar meson field and cross - interactions of meson fields as in the nonlinear rmf model ,
the coupling constants are density dependent . the lagrangian density in this model is generally written as @xmath179\psi \notag \\ & & + \frac{1}{2}(\partial _
{ \mu } \sigma \partial ^{\mu } \sigma -m_{s}^{2}\sigma ^{2})+\frac{1}{2}(\partial _
{ \mu } \vec{\delta}\cdot \partial ^{\mu } \vec{\delta}-m_{\delta } ^{2}\vec{\delta}^{2 } ) \notag \\ & & -\frac{1}{4}\omega _ { \mu \nu } \omega ^{\mu \nu } + \frac{1}{2}m_{\omega } ^{2}\omega _ { \mu } \omega ^{\mu } \notag \\ & & -\frac{1}{4}\vec{\rho}_{\mu \nu } \cdot \vec{\rho}^{\mu \nu } + \frac{1}{2 } m_{\rho } ^{2}\vec{\rho}_{\mu } \cdot \vec{\rho}^{\mu } \label{lagdd}\end{aligned}\ ] ] the symbols used in above equation have their usual meanings as in eq.([lagnl ] ) but the coupling constants @xmath180 , @xmath181 , @xmath182 and @xmath183 now depend on the ( baryon ) density , which are usually parametrized as @xmath184 with @xmath185 and @xmath186 being the saturation density of symmetric nuclear matter . in some parameter sets , @xmath187\ ] ]
is used for the @xmath136 meson . since the coupling constants in the density - dependent rmf model depend on the baryon fields @xmath188 and @xmath127 through the density , additional terms besides the usual ones in the nonlinear rmf model appear in the field equations of motion when the partial derivatives of @xmath189 are carried out with respect to the fields @xmath188 and @xmath127 in the euler - lagrange equations .
the resulting dirac equation for the nucleon field now reads : @xmath140 \psi = 0,\ ] ] with the following nucleon scalar and vector self - energies : @xmath190 the new term @xmath191 in the vector self - energy , which is called the _ rearrangement _ self - energy @xcite , is given by @xmath192 with @xmath193 .
the rearrangement self - energy plays an essential role in the applications of the theory since it guarantees both the thermodynamical consistency and the energy - momentum conservation @xcite . for the meson fields , the equations of motion are @xmath194 in the static case for an infinite nuclear matter , the meson equations of motion become @xmath195 so the nucleon scalar and vector self - energies are @xmath196 with @xmath197 from the energy - momentum tensor
, the energy density and pressure of nuclear matter can be derived , and they are given by @xmath198\end{aligned}\ ] ] and @xmath199 .\end{aligned}\ ] ] it is seen that the rearrangement self - energy does not affect the energy density but contributes explicitly to the pressure .
furthermore , the symmetry energy can be written as @xmath200 } , \label{esymdd}\end{aligned}\ ] ] with notations similarly defined as in the nonlinear rmf model .
the point - coupling model is defined by a lagrangian density that consists of only nucleon fields . in the present study ,
we use the lagrangian density of the nonlinear point - coupling model of refs.@xcite , i.e. , @xmath201 with @xmath202^{2 } \notag \\ & & -{{\frac{1}{4}}}\,\gamma _ { \mathrm{tv}}[(\bar{\psi}\vec{\tau}\gamma _ { \mu } \psi ) \cdot ( \bar{\psi}\vec{\tau}\gamma ^{\mu } \psi ) ] ^{2 } , \\
\mathcal{l}^{\mathrm{der } } & = & -{{\frac{1}{2}}}\,\delta _ { \mathrm{s } } ( \partial _ { \nu } \bar{\psi}\psi ) ( \partial ^{\nu } \bar{\psi}\psi ) \notag \\ & & -{{\frac{1}{2}}}\,\delta _ { \mathrm{v}}(\partial _ { \nu } \bar{\psi}\gamma _ { \mu } \psi ) ( \partial ^{\nu } \bar{\psi}\gamma ^{\mu } \psi ) \notag \\ & & -{{\frac{1}{2}}}\,\delta _ { \mathrm{ts}}(\partial _ { \nu } \bar{\psi}\vec{\tau}\psi ) \cdot ( \partial ^{\nu } \bar{\psi}\vec{\tau}\psi ) \notag \\ & & -{{\frac{1}{2}}}\,\delta _ { \mathrm{tv}}(\partial _ { \nu } \bar{\psi}\vec{\tau } \gamma _ { \mu } \psi ) \cdot ( \partial ^{\nu } \bar{\psi}\vec{\tau}\gamma ^{\mu } \psi ) .\end{aligned}\ ] ] in the above , @xmath203 is the kinetic term of nucleons and @xmath204 describes the four - fermion interactions while @xmath205 and @xmath206 contain , respectively , higher - order terms involving more than four fermions and derivatives in the nucleon field . for the twelve coupling constants in the lagrangian density , @xmath207 , @xmath208 , @xmath209 , @xmath210 , @xmath211 , @xmath212 , @xmath213 , @xmath214 , @xmath215 , @xmath216 , @xmath217 , and @xmath218 ,
the subscripts denote the tensor structure of a coupling with s v and tstand for scalar , vector , and isovector , respectively .
the symbols @xmath219 , @xmath220 , @xmath221 , and @xmath222 refer , respectively , to four - fermion or second - order terms , derivative couplings , third- and fourth order terms @xcite . from the variation of the lagrangian density
( [ lagnlpc ] ) with respect to @xmath188 , we obtain the following dirac equation for the nucleon field : @xmath223\psi = 0,\ ] ] where the nucleon scalar ( @xmath224 ) and vector ( @xmath225 ) self - energies are @xmath226 respectively , with @xmath227 in the above , @xmath228 denotes the four - dimensional dalembertian . in the translationally invariant infinite nuclear matter , all terms involving derivative couplings drop out and the spatial components of the four - currents also vanish . in terms of the baryon density @xmath229 and scalar density @xmath230 as well as the isospin baryon density @xmath231 and the isospin scalar density @xmath232 ,
the nucleon scalar and vector self - energies in asymmetric nuclear matter can be rewritten as @xmath233 the energy density @xmath165 and the pressure @xmath166 derived from the energy - momentum tensor in the nonlinear point - coupling rmf model are given by @xmath234 @xmath235 where @xmath236 and @xmath237 are defined as in eq .
( ef ) with the nucleon dirac masses @xmath238 furthermore , the symmetry energy in this model can be expressed as @xmath239 } , \label{esymnlpc}\end{aligned}\ ] ] with notations again similarly defined as in the nonlinear rmf model .
for the density - dependent point - coupling rmf model , we use the lagrangian density of refs.@xcite , i.e. , @xmath240 with @xmath241 in the above , @xmath203 is the kinetic term of the nucleons and @xmath242 is a four - fermion interaction while @xmath206 represents derivatives in the nucleon scalar densities . unlike in the nonlinear point - coupling rmf model ,
the density - dependent point - coupling rmf model used here includes only second - order interaction terms with density - dependent couplings @xmath243 and @xmath244 that are determined from finite - density qcd sum rules and in - medium chiral perturbation theory @xcite .
variation of the lagrangian eq.([lagddpc ] ) with respect to @xmath188 leads to the single - nucleon dirac equation @xmath223\psi = 0,\ ] ] with the nucleon scalar and vector self - energies given , respectively , by @xmath245 where @xmath246 and @xmath247 in the above , we have @xmath248 , where the four - velocity @xmath249 is defined as @xmath250 with @xmath251 being the three - velocity vector , and @xmath252 and @xmath253 represent the rearrangement contributions resulting from the variation of the vertex functionals with respect to the nucleon fields in the density operator @xmath254 . the latter coincides with the baryon density in the nuclear matter rest frame . in the translationally invariant infinite asymmetric nuclear matter ,
the nucleon scalar and vector self - energies become @xmath255 with the rearrangement contribution to the self - energy @xmath256.\ ] ] for asymmetric nuclear matter , the energy density @xmath165 and the pressure @xmath166 derived from the energy - momentum tensor in the density - dependent point - coupling rmf model are @xmath257 and @xmath258 where @xmath236 and @xmath237 are defined as in eq .
( [ ef ] ) with the effective nucleon masses @xmath259 as in the density - dependent rmf model , _ rearrangement _ contributions appear explicitly in the expression for the pressure .
finally , the symmetry energy can be written as @xmath260 } , \label{esymddpc}\end{aligned}\ ] ] with similar notations as in the nonlinear rmf model .
using above models , we have studied the isospin - dependent properties of asymmetric nuclear matter . in the following ,
we focus on results regarding the nuclear symmetry energy , the nuclear symmetry potential , the isospin - splitting of nucleon effective mass , and the isospin - dependent nucleon scalar density in asymmetric nuclear matter . for different versions of the rmf model considered in the present work
, we mainly consider parameter sets commonly and successfully used in nuclear structure studies .
in particular , we select the parameter sets nl1 @xcite , nl2 @xcite , nl3 @xcite , nl - sh @xcite , tm1 @xcite , pk1 @xcite , fsu - gold @xcite , ha @xcite , nl@xmath136 @xcite , nl@xmath261 @xcite for the nonlinear rmf model ; tw99 @xcite , dd - me1 @xcite , dd - me2 @xcite , pkdd @xcite , dd @xcite , dd - f @xcite , and ddrh - corr @xcite for the density - dependent rmf model ; and pc - f1 @xcite , pc - f2 @xcite , pc - f3 @xcite , pc - f4 @xcite , pc - la @xcite , and fkvw @xcite for the point - coupling rmf model . there are totally @xmath0 parameter sets , and most of them can describe reasonably well the binding energies and charge radii of a large number of nuclei in the periodic table except the parameter set ha , for which to our knowledge there are no calculations for finite nuclei .
we note that all selected parameter sets include the isovector - vector channel involving either the isovector - vector @xmath136 meson or the isovector - vector interaction vertices in the lagrangian .
the ha parameter set further includes the isovector - scalar meson field @xmath262 and fits successfully some results obtained from the more microscopic dbhf approach @xcite . the parameter sets nl@xmath261 and ddrh - corr also include the isovector - scalar meson field @xmath262 , while pc - f2 , pc - f4 , pc - la , and fkvw include the isovector - scalar interaction vertices . the parameter sets nl@xmath261 as well as nl@xmath136 are obtained from fitting the empirical properties of asymmetric nuclear matter @xcite and describe reasonably well the binding energies and charge radii of a large number of nuclei @xcite . for the ddrh - corr ,
its parameters are determined from the density - dependent meson - nucleon vertices extracted from the self - energies of asymmetric nuclear matter calculated in the microscopic dbhf approach with momentum corrections , and it reproduces satisfactorily the properties of finite nuclei and the eos from the dbhf approach @xcite . in
the parameter sets pc - f1 , pc - f2 , pc - f3 , pc - f4 and pc - la for the nonlinear point - coupling model , their coupling constants are determined in a self - consistent procedure that solves the model equations for representative nuclei simultaneously in a generalized nonlinear least - squares adjustment algorithm @xcite .
the parameter set fkvw for the density - dependent point - coupling model are determined by the constraints derived from the finite - density qcd sum rules , in - medium chiral perturbation theory , and experimental data of a number of finite nuclei @xcite .
[ esymden ] displays the density dependence of the nuclear symmetry energy @xmath21 for the @xmath0 parameter sets in the nonlinear , density - dependent , and point - coupling rmf models . for comparison
, we also show in fig .
[ esymden ] results from the phenomenological parametrization of the momentum - dependent nuclear mean - field potential based on the gogny effective interaction @xcite , i.e. , the mdi interactions with @xmath263 ( open squares ) and @xmath264 ( solid squares ) , where different @xmath265 values correspond to different density dependence of the nuclear symmetry energy but keep other properties of the nuclear eos the same @xcite ( see appendix [ mdi ] for details ) . from analyzing the isospin diffusion data from nscl / msu using the ibuu04 transport model with in - medium nn cross sections , it has been found that the mdi interactions with @xmath263 and @xmath264 give , respectively , the upper and lower bounds for the stiffness of the nuclear symmetry energy at densities up to about @xmath266 @xcite . for the parameter sets nl1 , nl2 , nl3 , nl - sh , tm1 , pk1 , fsu - gold , ha , nl@xmath267 , and nl@xmath268 in the nonlinear rmf model ( a ) ; tw99 , dd - me1 , dd - me2 , pkdd , dd , dd - f , and ddrh - corr in the density - dependent rmf model ( b ) ; and pc - f1 , pc - f2 , pc - f3 , pc - f4 , pc - la , and fkvw in the point - coupling rmf model ( c ) . for comparison , results from the mdi interaction with @xmath263 ( open squares ) and @xmath264(solid squares )
are also shown . ]
it is seen from fig .
[ esymden ] that the density dependence of symmetry energy varies drastically among different interactions . in the nonlinear rmf model , while the dependence on density is almost linear for most parameter sets , it is much softer for the parameter sets fsu - gold and ha .
the softening of the symmetry energy from the latter two parameter sets is due to the mixed isoscalar - isovector couplings @xmath138 and @xmath139 @xcite which modifies the density dependence of symmetry energy as seen in eq .
( [ esymnl ] ) . for the parameter set nl@xmath261
, it gives a symmetry energy that depends linearly on density at low densities but becomes stiffer at high densities due to inclusion of the isovector - scalar @xmath122 meson .
the approximate linear density - dependent behavior of the symmetry energy for other parameter sets in the nonlinear rmf model can also be understood from eq .
( [ esymnl ] ) , which shows that the symmetry energy at high densities is dominated by the potential energy that is proportional to the baryon density if the mixed isoscalar - isovector coupling and the isovector - scalar @xmath122 meson are not included in the model .
the density dependence of the symmetry energy in the density - dependent rmf model is essentially determined by the density dependence of the coupling constants @xmath183 and @xmath182 of isovector mesons .
most parameter sets in this case give similar symmetry energies except the parameter sets pkdd and ddrh - corr . compared with other parameter sets in the density - dependent rmf model
, pkdd gives a very large while ddrh - corr gives a very small value for the symmetry energy at saturation density .
for point - coupling models , all parameter sets ( pc - f1 , pc - f2 , pc - f3 , pc - f4 and pc - la ) in the nonlinear point - coupling rmf model predict almost linearly density - dependent symmetry energies while the parameter set fkvw in the density - dependent point - coupling rmf model gives a somewhat softer symmetry energy .
.bulk properties of nuclear matter at the saturation point : @xmath269 ( mev ) , @xmath270 ( @xmath2 ) , @xmath31 ( mev ) , @xmath271 ( mev ) , @xmath272 ( mev ) , @xmath26 ( mev ) , and @xmath273 ( mev ) using the @xmath0 parameter sets in the nonlinear , density - dependent , and point - coupling rmf models .
the last column gives the references for corresponding parameter sets .
[ cols="^,^,^,^,^,^,^,^,^",options="header " , ] for the different nucleon effective masses in symmetric nuclear matter at saturation density , we show in table [ effmass ] the results from the @xmath0 parameter sets in the nonlinear , density - dependent , and point - coupling rmf models . it is seen that the values of @xmath274 , @xmath114 , @xmath275 , @xmath276 , and @xmath277 are in the range of @xmath278 , @xmath279 , @xmath280 , @xmath281 , and @xmath282 , respectively .
the parameter sets nl2 , ha , nl@xmath136 and nl@xmath261 seem to give too large values , i.e. , @xmath283 , @xmath284 , @xmath285 , and @xmath285 , respectively , for the @xmath286 as values in the range of @xmath287 are needed to describe reasonably the spin - orbit splitting in finite nuclei using the rmf models . on the other hand ,
the larger dirac masses leads to larger landau masses @xmath114 of @xmath288 , @xmath289 , @xmath290 , and @xmath291 , respectively , for the parameter sets nl2 , ha , nl@xmath136 and nl@xmath261 , which are consistent with the empirical constraint of @xmath292 = @xmath115 @xcite . , @xmath293 , @xmath294 , @xmath276 , and @xmath277 in symmetric nuclear matter as well as their corresponding isospin splittings in neutron - rich nuclear matter with isospin asymmetry @xmath295 for the parameter sets nl1 , nl2 , nl3 , nl - sh , tm1 , pk1 , fsu - gold , ha , nl@xmath296 , and nl@xmath268 in the nonlinear rmf model . ]
but for tw99 , dd - me1 , dd - me2 , pkdd , dd , dd - f , and ddrh - corr in the density - dependent rmf model . ]
but for pc - f1 , pc - f2 , pc - f3 , pc - f4 , pc - la , and fkvw in the point - coupling rmf model .
] the density dependence of the different nucleon effective masses in symmetric nuclear matter and corresponding isospin splitting @xmath297 in asymmetric nuclear matter with isospin asymmetry @xmath298 are shown in fig .
[ mstardennl ] for the parameter sets nl1 , nl2 , nl3 , nl - sh , tm1 , pk1 , fsu - gold , ha , nl@xmath136 , and nl@xmath261 in the nonlinear rmf model . figs .
[ mstardendd ] and [ mstardenpc ] display the same results as in fig .
[ mstardennl ] but for the parameter sets tw99 , dd - me1 , dd - me2 , pkdd , dd , dd - f , and ddrh - corr in the density - dependent rmf models and for pc - f1 , pc - f2 , pc - f3 , pc - f4 , pc - la , and fkvw in the point - coupling rmf model , respectively .
it is seen that different parameter sets in the nonlinear rmf model give significantly different density dependence for the nucleon effective masses while the different parameter sets in the density - dependent and point - coupling rmf models predict roughly the same density dependence for the nucleon effective masses except that the parameter set pc - la gives very large values for the nucleon effective masses at high densities .
this unusual behavior for pc - la was also observed in ref .
@xcite , and it is due to the fact that the coupling constant @xmath299 for the higher - order interaction term in pc - la is positive @xcite and dominates at high density , leading thus to the very large nucleon effective mass . for the landau mass at a fixed baryon density ,
its value @xmath114 is generally larger than @xmath286 .
this can be seen from eq .
( [ mlandau ] ) if it is rewritten as @xmath300 which shows that @xmath301 if nucleon self - energies are independent of momentum / energy . for the lorentz mass @xmath116 , @xmath302 depends almost linearly on density and thus has a stronger density dependence than the dirac and landau masses .
we note from eqs .
( [ dispersion ] ) and ( [ mlandir ] ) that eq .
( [ mlorentz ] ) can be reduced to @xmath303 if nucleon self - energies are independent of momentum / energy .
therefore , the density dependence of @xmath304 is determined uniquely by the density dependence of nucleon vector self - energy . in the nonlinear rmf model ,
most of the parameter sets , except for tm1 , pk1 and fsu - gold which include the self - coupling of the @xmath129 meson field , give a linear density dependence for @xmath58 , leading thus to the observed linear density dependence of @xmath116 .
as to the nonlinear density dependence of @xmath116 in the density - dependent rmf model and point - coupling models , it is due to the nonlinear density dependence of the coupling constant or the inclusion of higher - order couplings . for @xmath276 and @xmath305 , they are seen to have roughly same magnitude and also same density dependence as @xmath114 .
this feature can be understood from the fact that with the dispersion relation of eq .
( [ dispersion ] ) , eq .
( [ mopt ] ) and eq .
( [ msod ] ) can be re - expressed as @xmath306 and @xmath307 , \label{msod2}\ ] ] respectively .
since @xmath308 ( for example , @xmath309 mev / c at @xmath310 @xmath2 ) , we have @xmath311 ( with an error of a few percent ) and thus @xmath312 .
furthermore , the second term in eq .
( [ msod2 ] ) can be neglected compared with the first term as @xmath313 ( it is a good approximation at low densities and with an error of about @xmath314 at high densities , e.g. , @xmath310 @xmath2 ) . as a result , we have @xmath315 . from the dirac equation , one sees that condensed scalar fields ( scalar self - energies ) lead to a shift of nucleon mass such that the nuclear matter is described as a system of pseudo - nucleons with masses @xmath104 ( dirac mass ) moving in classical vector fields with @xmath122 meson field or isovector - scalar potential further generating the splitting of the proton and neutron dirac masses in asymmetric nuclear matter . for the isospin
splitting of @xmath103 in neutron - rich nuclear matter , it is interesting to see that the parameter sets ha , nl@xmath261 , ddrh - corr , and pc - f4 give @xmath316 while pc - f2 , pc - la , and fkvw exhibit the opposite behavior of @xmath317 .
this feature implies that the isospin - dependent scalar potential can be negative or positive depending on the parameter sets used . in the nonlinear rmf model ,
we obtain from eqs .
( [ delnl ] ) and ( [ mdiracnl ] ) @xmath318 which indicates that we always have @xmath319 in the neutron - rich nuclear matter where @xmath320 .
this argument is also applicable to the density - dependent rmf model by replacing @xmath321 with the density dependent @xmath182 .
for the nonlinear point - coupling models , we have , on the other hand , @xmath322 a similar equation can be obtained for the density - dependent point - coupling models with the replacement of @xmath323 by the density dependent @xmath324 .
therefore , the isospin splitting of @xmath103 in neutron - rich nuclear matter depends on the sign of the isovector - scalar coupling constant @xmath325 and @xmath324 in the point - coupling models .
since the value of @xmath326 in pc - f2 and pc - la as well as the value of @xmath324 in fkvw are positive , these parameter sets lead to the isospin - splitting @xmath327 in neutron - rich nuclear matter , which is opposite to that in other parameter sets considered here .
the isospin splitting of @xmath103 is directly related to the isovector spin - orbit potential that determines the isospin - dependent spin - orbit splitting in finite nuclei .
unfortunately , there are no clear experimental indication about the isospin dependence of the spin - orbit splitting in finite nuclei @xcite , so detailed experimental data on the single - particle energy levels in exotic nuclei are needed to pin down the isospin splitting of @xmath103 in asymmetric nuclear matter . for the isospin splitting of @xmath105 in neutron - rich nuclear matter , most parameter sets give @xmath328 ,
which is consistent with the usual results in non - relativistic models . the parameter sets nl@xmath261 and ddrh - corr give , however , the opposite result due to the strong isospin - splitting of @xmath103 with @xmath329 for nl@xmath261 and ddrh - corr and the fact that @xmath105 is related to the fermi momentum and @xmath103 according to eq .
( [ mlandir ] ) .
the isospin - splitting @xmath330 implies that neutrons have a larger level density at the fermi energy and thus more compressed single - particle levels in finite nuclei than protons . for the isospin splitting of @xmath116 in neutron - rich nuclear matter ,
all parameter sets give @xmath331 except that the pc - l3 gives @xmath332 at high densities . from eq .
( [ mlorentz2 ] ) , we have @xmath333 which leads to the observed isospin - splitting @xmath334 as we generally have @xmath335 as discussed above . for the parameter set pc - l3
, it includes a higher - order isovector - vector term through the parameter @xmath336 .
since the latter has a negative value and dominates at high densities according to eq .
( [ sig0nlpc ] ) , it leads to @xmath337 and thus @xmath338 at high densities .
the isospin splitting of @xmath276 and @xmath277 in neutron - rich nuclear matter show a similar behavior as @xmath105 as expected from the discussions below eqs .
( [ mopt2 ] ) and ( [ msod2 ] ) .
the nucleon scalar density as defined in eq . ( [ rhos ] ) is the source for the nucleon scalar self - energy ( scalar potential ) . in the rmf model , the isospin - dependent nucleon scalar density is uniquely related to the nucleon dirac mass as shown in eq .
( [ rhosnp ] ) .
the latter equation also shows that the scalar density is less than the baryon density due to the factor @xmath339 which causes a reduction of the contribution of rapidly moving nucleons to the scalar source term .
this mechanism is responsible for nuclear matter saturation in the mean - field theory and essentially distinguishes relativistic models from non - relativistic ones . in practice ,
the isospin - dependent nucleon scalar density is also an essential ingredient for evaluating the relativistic optical potential for neutrons and protons in the relativistic impulse approximation ( see , e.g. , refs .
@xcite and references therein ) . and @xmath340 for the parameter sets nl1 , nl2 , nl3 , nl - sh , tm1 , pk1 , fsu - gold , ha , nl@xmath296 , and nl@xmath268 of the nonlinear rmf model ( a ) ; tw99 , dd - me1 , dd - me2 , pkdd , dd , dd - f , and ddrh - corr of the density - dependent rmf model ( b ) ; pc - f1 , pc - f2 , pc - f3 , pc - f4 , pc - la , and fkvw of the point - coupling rmf model ( c ) . ] in fig
. [ rhosrhob ] , we show the neutron and proton scalar densities as functions of the baryon density @xmath341 in nuclear matter with isospin asymmetry @xmath29 and @xmath340 for the @xmath0 parameter sets from the nonlinear , density - dependent , and point - coupling rmf models . it is seen that the neutron scalar density is larger than that of protons in neutron - rich nuclear matter at a fixed baryon density .
although results for different parameter sets are almost the same at lower baryon densities , they become different when @xmath342 @xmath2 , and this is consistent with the conclusions of refs .
in particular , different parameter sets in the nonlinear rmf model predict a larger uncertainty for the value of the nucleon scalar density at high baryon density while all the parameter sets ( except pc - la ) in the density - dependent rmf model and point - coupling models give roughly same results for the nucleon scalar density .
these features are consistent with the results for the density dependence of nucleon dirac mass shown in figs .
[ mstardennl ] , [ mstardendd ] , and [ mstardenpc ] . at low baryon densities ,
neutron and proton scalar densities are seen to increase roughly linearly with baryon density , and this can be easily understood from eq .
( [ rhosnp ] ) , which is reduced to the following expression at low densities ( @xmath343 due to @xmath344 ) : @xmath345 therefore , neutron and proton scalar densities generally approach their respective baryon densities in asymmetric nuclear matter at low baryon densities .
using different versions of relativistic mean - field models that are commonly used in current nuclear structure studies , i.e. , the nonlinear model , the model with density - dependent nucleon - meson coupling , and the point - coupling model , we have investigated systematically the isospin - dependent bulk and single - particle properties of isospin - asymmetric nuclear matter . in particular , we considered @xmath0 parameter sets commonly and successfully used in nuclear structure studies , i.e. , nl1 , nl2 , nl3 , nl - sh , tm1 , pk1 , fsu - gold , ha , nl@xmath136 , nl@xmath261 for the nonlinear rmf model ; tw99 , dd - me1 , dd - me2 , pkdd , dd , dd - f , and ddrh - corr for the density - dependent rmf model ; and pc - f1 , pc - f2 , pc - f3 , pc - f4 , pc - la , and fkvw for the point - coupling rmf model .
most of the parameter sets are obtained from fitting the binding energies and charge radii of a large number of nuclei in the periodic table or the results from the microscopic dbhf approach , which have been shown to describe successfully a number of the properties of finite nuclei . using these models ,
we have studied the density dependence of nuclear symmetry energy and compared the results with the symmetry energy recently extracted from the analyses of the isospin diffusion data from heavy - ion collisions based on an isospin- and momentum - dependent transport model with in - medium nn cross sections , the isoscaling analyses of isotope ratios in intermediate energy heavy ion collisions , and measured isotopic dependence of the giant monopole resonances in even - a sn isotopes .
these analyses have led to the extraction of @xmath41 mev for the slope parameter of the nuclear symmetry energy at saturation density and @xmath346 mev or @xmath347 mev for the isospin - dependent part of the isobaric incompressibility of isospin asymmetric nuclear matter , which may represent the most stringent phenomenological constraints available so far on the nuclear symmetry energy at sub - saturation densities . using these constraints ,
we have found that , among the @xmath0 parameter sets considered in the present work , only six sets , i.e. , tm1 , nl@xmath136 , nl@xmath261 , pkdd , pc - la , and fkvw , have nuclear symmetry energies that are consistent with the extracted @xmath26 value of @xmath348 mev while fifteen sets , i.e. , nl3 , nl - sh , tm1 , pk1 , ha , nl@xmath136 , nl@xmath261 , tw99 , pkdd , dd - f , pc - f1 , pc - f2 , pc - f3 , pc - f4 , and fkvw , have nuclear symmetry energies that are consistent with the extracted @xmath33 value of @xmath349 mev or @xmath347 mev .
furthermore , we have found surprisingly that only five parameter sets , i.e. , tm1 , nl@xmath136 , nl@xmath261 , pkdd , and fkvw , in the @xmath0 parameter sets have nuclear symmetry energies that are consistent with the extracted values for both @xmath26 and @xmath33 .
we have noted that most parameter sets in the nonlinear and point - coupling rmf models predict stiffer symmetry energies while those in the density - dependent rmf model give softer symmetry energies .
these features are probably related to the rather limited flexibility in the parametrization of the isovector channel in all rmf models and also the fact that most of the parameter sets are obtained from fitting properties of finite nuclei which are mostly near the @xmath8-stability line and thus have less constraint on the isospin - dependent properties of asymmetric nuclear matter .
moreover , we have focused here on the behavior of the symmetry energy around saturation density while the parameter sets in rmf models are fitted to the properties of finite nuclei that are more sensitive to the properties of the nuclear symmetry energy at sub - saturation densities .
we have also investigated the energy dependence of three different nucleon optical potentials , i.e. , the schr dinger - equivalent potential@xmath350 ( eq . ( [ usep ] ) ) , the optical potential from the difference between the total energy of a nucleon in nuclear medium and its energy at the same momentum in free space @xmath351 ( eq . ( [ uopt ] ) ) , and the optical potential based on the second - order dirac equation @xmath352 ( eq . ( [ usd ] ) ) , as well as their corresponding symmetry potentials @xmath353 , @xmath354 , and @xmath355 as functions of momentum .
the results indicate that different optical potentials in symmetric nuclear matter exhibit similar energy dependence at low energies but have different high energy behaviors . in particular , at high energies , @xmath350 continues to increase linearly with momentum while @xmath351 and @xmath352 seem to saturate and thus display a more satisfactory high - energy limit compared to the optical potentials extracted from proton - nucleus scatterings using the dirac phenomenology . on the other hand , the nuclear symmetry potential at a fixed baryon density can increase or decrease with increasing nucleon momentum depending on the definition for the nucleon optical potential and the interactions used . for @xmath356 at @xmath357 @xmath2 ,
results from all @xmath0 parameter sets show that it increases with momentum , which is consistent with the predictions of microscopic dbhf and chiral perturbation calculations at low momenta ( less than about 300 mev / c ) but is inconsistent with the experimental result that the nuclear symmetry potential at saturation density ( the lane potential @xmath92 ) decreases at low energies ( beam energy @xmath60 above @xmath264 mev and less than about @xmath86 mev and corresponding momentum values are from about @xmath358 mev / c to @xmath359 mev / c ) and ria predictions at higher energies . for @xmath354 and @xmath360 , they can , however , decrease with momentum for some parameter sets , which is qualitatively consistent with the experimental constraint .
again , we emphasize that , for the three definitions of the optical potential and thus their corresponding nuclear symmetry potentials , only @xmath63 is well - defined theoretically and is schrdinger - equivalent while @xmath68 and @xmath85 are used here for references as @xmath68 has been extensively used in microscopic dbhf calculations @xcite and transport models for heavy - ion collisions @xcite and @xmath85 has been used in analyses of the relativistic optical potential based on the dirac phenomenology @xcite .
we have further explored different nucleon effective masses , i.e. , @xmath103 , @xmath105 , @xmath361 , @xmath362 , and @xmath363 in symmetric nuclear matter as well as their isospin - splittings in neutron - rich nuclear matter .
most of the parameter sets are found to give reasonable values for @xmath103 as required by the spin - orbit splitting data in finite nuclei but too small values for @xmath364 , implying that they would give too small a level density at the fermi energy and too large a spread of the single - particle levels in finite nuclei . for @xmath116 , it is found to display the strongest ( almost linear ) density dependence even at high densities .
interestingly , including the isovector - scalar channel leads to the isospin - splitting of @xmath103 , and @xmath365 is always obtained in neutron - rich nuclear matter for the nonlinear and density - dependent rmf models but an opposite result can be observed in the point - coupling model . for @xmath105 ,
most parameter sets give the isospin splitting @xmath366 in neutron - rich nuclear matter , which is consistent with usual results in non - relativistic models , while an opposite isospin - splitting is observed for @xmath116 .
in addition , @xmath362 , and @xmath367 are found to display similar behaviors as @xmath368 .
finally , we have studied the baryon density dependence of the nucleon scalar density and its isospin - splitting in neutron - rich nuclear matter .
the results indicate that the neutron scalar density is larger than that of proton in neutron - rich nuclear matter at a fixed baryon density . at low baryon densities ,
the neutron and proton scalar densities generally approach their respective baryon densities in asymmetric nuclear matter . in the present work ,
we have focused on three versions of standard rmf models , i.e. , the nonlinear , density - dependent , and point - coupling rmf models .
we note that there are some recent works @xcite in which the standard rmf models are extended to include density - dependent hadron masses and meson coupling constants via the brown - rho ( br ) scaling @xcite .
in particular , the parameter sets slc and slcd constructed in ref .
@xcite are not only consistent with current experimental results for symmetric matter at normal and supra - normal densities and the symmetry energy constrained by the isospin diffusion data at sub - saturation densities , but also give a fairly satisfactory description of the ground state properties of finite nuclei , including binding energies , charge radii , and neutron skin thickness . in all standard rmf models ,
the nucleon self - energies are independent of momentum / energy . as a result
, the dirac mass and the landau mass obtained from these models can not be simultaneously consistent with experimental data ( see , e.g. , eq .
( [ mlandir ] ) ) . also , the schrdinger - equivalent potential@xmath350 ( eq . ( [ usep ] ) ) in these models increases linearly with nucleon energy even at high energies .
recently , momentum - dependent nucleon self - energies have been introduced in the rmf model by including in the lagrangian density the couplings of meson fields to the derivatives of nucleon densities @xcite , and the results indicate that a reasonable energy dependence of the schrdinger - equivalent potential in symmetric nuclear matter at saturation density can be obtained , and the landau mass can also be increased to a more reasonable value while keeping the dirac mass unchanged , which further leads to an improved description of @xmath8-decay half - lives of neutron - rich nuclei in the @xmath369 and @xmath370 regions @xcite . in the framework of density - functional theory , including the couplings of meson fields to the derivatives of nucleon densities in the lagrangian density provides an effective way to take into account higher - order effects .
another way to introduce the momentum - dependence in nucleon self - energies is to include the fock exchange terms by means of the relativistic hartree - fock ( rhf ) approximation , even though in practice the inclusion of the fock terms would increase significantly the numerical complexity such that it is very difficult to find appropriate effective lagrangians for the rhf model to give satisfactory quantitative description of the nuclear structure properties compared with standard rmf models @xcite .
recently , there have been some developments in the density - dependent rhf approach @xcite .
it is shown that the density - dependent rhf model can describe the properties of both finite nuclei and nuclear matter with results comparable to those from standard rmf models . a more phenomenological way to improve the results of rmf models
is to introduce momentum- as well as isospin - dependent form factors in the meson - nucleon coupling constants .
it has been shown in refs .
@xcite that the empirically observed energy dependence of the nuclear optical potential in symmetric nuclear matter at saturation density can be reproduced by relativistic mean - field models with momentum - dependent form factors . finally , to better understand the isospin - dependent properties of asymmetric nuclear matter it is crucial to investigate the density and momentum dependence of underlying isovector nuclear effective interaction . to reach this ultimate goal , we need not only more advanced theoretical approaches but also more experimental data both on finite nuclei , especially those far from @xmath371-stability line , and from heavy - ion reactions induced by high energy neutron - rich nuclei .
we would like to thank paolo finelli , wei - zhou jiang , plamen g. krastev , and gao - chan yong for helpful discussions or communications .
l.w.c . also thanks the hospitality of texas a&m university at commerce and college station where part of the work was done .
this work was supported in part by the national natural science foundation of china under grant nos .
10575071 and 10675082 , moe of china under project ncet-05 - 0392 , shanghai rising - star program under grant no .
06qa14024 , the srf for rocs , sem of china , the china major state basic research development program under contract no .
2007cb815004 ( l.w.c . ) , the us national science foundation under grant no .
phy-0457265 , the welch foundation under grant no .
a-1358 ( c.m.k . ) , the us national science foundation under grant no .
phy-0652548 , and the research corporation under award no .
7123 ( b.a.l . ) .
the isospin- and momentum - dependent mdi interaction is based on a modified finite - range gogny effective interaction @xcite . in the mdi interaction ,
the potential energy density @xmath372 of an asymmetric nuclear matter at total density @xmath136 and isospin asymmetry @xmath16 is expressed as follows @xcite , @xmath373 in the mean - field approximation , eq .
( [ mdiv ] ) leads to the following single - particle potential for a nucleon with momentum @xmath65 and isospin @xmath374 in asymmetric nuclear matter @xcite : @xmath375 in the above @xmath376 ( @xmath377 ) for neutrons ( protons ) ; @xmath378 ; @xmath379 is the phase - space distribution function at coordinate @xmath380 and momentum @xmath65 .
the parameters @xmath381 and @xmath382 are obtained by fitting the momentum - dependence of @xmath383 to that predicted by the gogny hartree - fock and/or the brueckner - hartree - fock calculations , the saturation properties of symmetric nuclear matter and the symmetry energy of @xmath384 mev at normal nuclear matter density @xmath385 @xmath2 @xcite .
the incompressibility @xmath31 of cold symmetric nuclear matter at saturation density @xmath19 is set to be @xmath386 mev .
the parameters @xmath387 and @xmath388 depend on the @xmath265 parameter according to @xmath389the different @xmath265 values in the mdi interaction are introduced to vary the density dependence of the nuclear symmetry energy while keeping other properties of the nuclear equation of state fixed @xcite , and they can be adjusted to mimic the predictions of microscopic and/or phenomenological many - body theories on the density dependence of nuclear matter symmetry energy .
the last two terms in eq .
( [ mdiu ] ) contain the momentum - dependence of the single - particle potential .
the momentum dependence of the symmetry potential stems from the different interaction strength parameters @xmath390 and @xmath391 for a nucleon of isospin @xmath374 interacting , respectively , with unlike and like nucleons in the background fields .
more specifically , we use @xmath392 mev and @xmath393 mev . @xmath396 \right .
\nonumber \\ & & + 4\lambda \left [ ( p_{f}^{3}(\tau ) -p_{f}^{3}(\tau ^{\prime } ) ) \tan ^{-1}\frac{p_{f}(\tau ) -p_{f}(\tau ^{\prime } ) } { \lambda } -(p_{f}^{3}(\tau ) + p_{f}^{3}(\tau ^{\prime } ) ) \tan ^{-1}\frac{p_{f}(\tau ) + p_{f}(\tau ^{\prime } ) } { \lambda } \right ] \nonumber \\ & & \left .
+ \frac{1}{4}\left [ \lambda ^{4}+6\lambda ^{2}(p_{f}^{2}(\tau ) + p_{f}^{2}(\tau ^{\prime } ) ) -3(p_{f}^{2}(\tau ) -p_{f}^{2}(\tau ^{\prime } ) ) ^{2}\right ] \ln \frac{(p_{f}(\tau ) + p_{f}(\tau ^{\prime } ) ) ^{2}+\lambda ^{2}}{(p_{f}(\tau ) -p_{f}(\tau ^{\prime } ) ) ^{2}+\lambda ^{2}}\right\}\end{aligned}\ ] ] we note that the mdi interaction has been extensively used in the transport model for studying isospin effects in intermediate energy heavy - ion collisions induced by neutron - rich nuclei liba04a , che04,che05a , liba05a , liba05b , liba06b , yon06a , yon06b , yon07 and the study of the thermal properties of asymmetric nuclear matter @xcite .
in particular , the isospin diffusion data from nscl / msu have constrained the value of @xmath265 to between @xmath264 and @xmath164 for nuclear matter densities less than about @xmath398 @xcite . | using various relativistic mean - field models , including the nonlinear ones with meson field self - interactions , those with density - dependent meson - nucleon couplings , and the point - coupling models without meson fields , we have studied the isospin - dependent bulk and single - particle properties of asymmetric nuclear matter . in particular , we have determined the density dependence of nuclear symmetry energy from these different relativistic mean - field models and compare the results with the constraints recently extracted from analyses of experimental data on isospin diffusion and isotopic scaling in intermediate - energy heavy ion collisions as well as from measured isotopic dependence of the giant monopole resonances in even - a sn isotopes . among the @xmath0 parameter sets in the relativistic mean - filed model that are commonly used for nuclear structure studies , only a few
are found to give symmetry energies that are consistent with the empirical constraints .
we have also studied the nuclear symmetry potential and the isospin - splitting of the nucleon effective mass in isospin asymmetric nuclear matter .
we find that both the momentum dependence of the nuclear symmetry potential at fixed baryon density and the isospin - splitting of the nucleon effective mass in neutron - rich nuclear matter depend not only on the nuclear interactions but also on the definition of the nucleon optical potential . |
in studying the motion of molecular motors we introduced in @xcite a solvable model : we think of the molecular motors as diffusion particles traveling in a narrow random channel . based on the model suggested in @xcite , we consider in this paper wave front propagation for a reaction - diffusion in narrow random channels .
problems of this type naturally appear in the theory of nerve impulse propagation and in combustion theory .
our analysis relies on techniques in large deviations similar to that of ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* chapter 7 ) and @xcite , @xcite , @xcite , @xcite , ( * ? ? ?
* ; * ? ? ?
* chapter 5 ) .
we shall note that problems of this type are mentioned in ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* chapter 7 ) , @xcite , @xcite .
it is also interesting to note that similar problems are considered in @xcite , @xcite , @xcite but from different points of view .
let us first briefly recall the model introduced in @xcite . let @xmath0 be a pair of piecewise smooth functions with @xmath1 .
let @xmath2 be a tubular @xmath3-d domain of infinite length , i.e. it goes along the whole @xmath4-axis . at the discontinuities of @xmath0 , we connect the pieces of the boundary via straight vertical lines .
the domain @xmath5 models the `` main '' channel in which the motor is traveling .
let a sequence of `` wings '' @xmath6 ( @xmath7 ) be attached to @xmath5 .
these wings are attached to @xmath5 at the discontinuities of the functions @xmath0 .
consider the union @xmath8 .
an example of such a domain @xmath9 is shown in fig.1 , in which one can see four `` wings '' @xmath10 .
we assume that , after adding the `` wings '' , for the domain @xmath9 , the boundary @xmath11 has two smooth pieces : the upper boundary and the lower boundary .
let @xmath12 be the inward unit normal vector to @xmath11 .
we make same assumptions as in @xcite .
* assumption 1 . *
the set of points @xmath13 for which there are points @xmath14 at which the unit normal vector @xmath15 is parallel to the @xmath4-axis : @xmath16 has no limit points in @xmath17 .
each such point @xmath4 corresponds to only one point @xmath14 for which @xmath16 .
* assumption 2 .
* for every @xmath4 the cross - section of the region @xmath9 at level @xmath4 , i.e. , the set of all points belonging to @xmath9 with the first coordinate equal to @xmath4 , consists of either one or two intervals that are its connected components .
that is to say , in the case of one interval this interval corresponds to the `` main channel '' @xmath5 ; and in the case of two intervals one of them corresponds to the `` main channel '' @xmath5 and the other one corresponds to the wing .
the wing will not have additional branching structure .
also , for some @xmath18 we have @xmath19 .
let us take into account randomness of the domain @xmath9 . keeping the above assumptions in mind
, we can assume that the functions @xmath0 and the shape of the wings @xmath20 ( @xmath21 ) are all random .
thus we can view the shape of @xmath9 as random .
we introduce a filtration @xmath22 , @xmath23 as the smallest @xmath24-algebra corresponding to the shape of @xmath25\}$ ] .
we introduce stationarity and mixing assumptions .
let us consider some @xmath26 , @xmath27 .
the set @xmath28 consists of some shapes of the domain @xmath29\}$ ] .
let @xmath30 ( @xmath31 ) be the operator corresponding to the shift along @xmath4-direction : @xmath32 consists of the same shapes as those in @xmath28 but correspond to the domain @xmath33\}$ ] .
* assumption 3 . * ( stationarity ) we have @xmath34 .
* assumption 4 . *
( mixing ) for any @xmath26 and any @xmath35 we have @xmath36 exponentially fast .
for instance , we can assume that there exists some @xmath37 such that @xmath38 for @xmath39 .
in particular , the mixing assumption implies that the transformation @xmath30 is ergodic . here and below the symbols @xmath40 and @xmath41 etc .
refer to probabilities and expectations etc .
with respect to the filtration @xmath42 .
let @xmath43 .
the parameter @xmath44 is small .
the domain @xmath45 models the narrow random channel .
let us consider the following reaction - diffusion equation in the domain @xmath45 : @xmath46 here @xmath47 is the inward unit normal vector field on @xmath48 ; @xmath49 is the velocity field ; the function @xmath50 is smooth and is of kpp type : @xmath51)$ ] , @xmath52 , @xmath53 for all @xmath54 , e.g. @xmath55 .
the initial function @xmath56 ( not identically equal to @xmath57 ) is smooth and compactly supported in @xmath58 : @xmath59 .
we notice that the initial function @xmath60 depends only on the variable @xmath13 and is independent of @xmath61 .
alternatively , problem ( 1.1 ) can be considered on the domain @xmath9 with a change of variable @xmath62 .
the equivalent problem takes the form @xmath63 here @xmath64 is the inward unit co - normal vector field on @xmath11 corresponding to the operator @xmath65 .
the diffusion process @xmath66 corresponding to problem ( 1.2 ) takes the form of equation ( 2 ) in @xcite .
we have @xmath67 let @xmath68 , @xmath69 denote probabilities and expectations with respect to the filtration generated by @xmath70 ( and henceforth @xmath71 ) .
we note that as in @xcite the motion of @xmath70 is independent of the random shape of @xmath9 ( and henceforth @xmath45 ) .
we have , in the same way as @xcite , the following .
* assumption 5 . *
the process @xmath71 is independent of the filtration @xmath72 corresponding to the shape of @xmath9 .
we shall also make some assumptions parallel to assumptions 7 and 8 in @xcite . to this end
we let @xmath73 be the random variable distributed the same as the distance along @xmath4-axis between two wings : @xmath73 is the distance along @xmath4-axis between two cross - sections of @xmath9 where there is a branching .
let @xmath74 be the cross - section width of the wing .
let @xmath75 be the projection length of a wing onto @xmath4-axis ( @xmath75 can be positive or negative ; compare with ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* assumptions 7 and 8) ) .
we assume the following .
* assumption 6 .
* with @xmath40 probability @xmath76 we have ( 1 ) @xmath77 for some constants @xmath78 ; ( 2 ) @xmath79 and @xmath80 for some constant @xmath81 .
our goal in this paper is to study the asymptotic wave front propagation properties for the generalized solution of ( 1.2 ) . to be precise , by a
_ generalized solution _ of ( 1.2 ) we mean the one defined via the path integral representation ( feynmann - kac ) formula : @xmath82 \ . \eqno(1.3)\ ] ] here @xmath83 for @xmath84 and @xmath85 .
the latter equality is due to the kpp nonlinearity assumption .
we shall also suppose that @xmath86 , @xmath87 $ ] .
the proof of existence , uniqueness and regularity of the generalized solution to the integral equation ( 1.3 ) is close to ( * ? ? ?
* ; * ? ? ? * chapter 5 , section 3 ) . for the reader
s convenience we will prove it in section 3 of this paper .
we introduced in @xcite the metric graph @xmath88 corresponding to the domain @xmath9 ( see fig.1 ) .
let @xmath89 be defined on @xmath88 as in ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* section 2 ) .
the construction of the graph @xmath88 and the process @xmath90 as well as some basic convergence results in @xcite will be recalled in section 2 .
the process @xmath90 is a diffusion process on @xmath88 with a generator @xmath28 and the domain of definition @xmath91 .
we consider the reaction - diffusion equation associated with the markov process @xmath90 .
this equation takes the form @xmath92 the initial function @xmath93 is the same as in ( 1.2 ) .
we require that for each fixed @xmath94 , @xmath95 .
this requirement ia a kind of boundary condition .
for details we refer to @xcite and ( * ? ? ?
* ; * ? ? ?
* chapter 8) and the references therein . the generalized solution to ( 1.4 )
is defined as the solution to the integral equation @xmath96 \ . \eqno(1.5)\ ] ] the existence and regularity of the solution can be proved in a same way as those for ( 1.3 ) .
we will briefly mention this in section 3 .
we will show , in section 3 of this paper , that as @xmath97 , the solution @xmath98 of ( 1.3 ) will converge to @xmath99 of ( 1.5 ) in the strong sense . here
@xmath100 and the mapping @xmath101 is an identification map that will be recalled in section 2 .
after we get convergence results we will focus on the study of the solution @xmath99 of ( 1.5 ) .
we will show that , as @xmath102 , the solution @xmath99 behaves asymptotically as a traveling wave .
this wave is traveling in both positive and negative directions along @xmath4-axis .
the paper is organized as follows : in section 2 we recall some necessary basic set up of @xcite , such as the construction of the graph @xmath88 , the process @xmath90 , etc . ; in section 3 we show the convergence as @xmath97 of the solution @xmath98 to @xmath99 ; in section 4 we obtain some auxiliary results that will be used in later sections ; in section 5 we derive the large deviation principle ; in section 6 we study the wave front propagation properties of the solution @xmath99 .
we shall first recall some basic facts in @xcite ( also see @xcite ) . let us work with a fixed shape of @xmath9 .
first of all we need to construct a graph @xmath88 related to the domain @xmath9 ( see fig.1 ) . for @xmath103 ,
let @xmath104 be the intersection of the domain @xmath9 with the line @xmath105 .
the set @xmath106 may have several connected components .
we identify all points in each connected component and the set thus obtained , equipped with the natural topology , is homeomorphic to a graph @xmath88 .
we label the edges of this graph @xmath88 by @xmath107 ( there might be infinitely many such edges ) .
we see that the structure of the graph @xmath88 consists of many edges ( such as @xmath108 , ... in fig.1 ) that form a long line corresponding to the domain @xmath5 and many other short edges ( such as @xmath109 , ... in fig.1 ) attached to the long line in a random way .
we will henceforth denote by @xmath110 the long line corresponding to the domain @xmath5 .
a point @xmath111 can be characterized by two coordinates : the horizontal coordinate @xmath4 , and the discrete coordinate @xmath112 being the number of the edge @xmath113 in the graph @xmath88 to which the point @xmath114 belongs .
let the identification mapping be @xmath115 .
we note that the second coordinate is not chosen in a unique way : for @xmath114 being an interior vertex @xmath116 of the graph @xmath88 we can take @xmath112 to be the number of any of the several edges meeting at the vertex @xmath116 .
the distance @xmath117 between two points @xmath118 and @xmath119 belonging to the same edge of the graph @xmath88 is defined as @xmath120 ; for @xmath121 belonging to different edges of the graph it is defined as the geodesic distance @xmath122 , where the minimum is taken over all chains @xmath123 of vertices @xmath124 connecting the points @xmath125 and @xmath126 . for an edge @xmath127
we consider the `` tube '' @xmath128 in @xmath9 .
the `` tube '' @xmath129 can be characterized by the interval @xmath130 $ ] and the `` height functions '' @xmath131 : @xmath132 . for @xmath133 $ ] ,
we denote the set @xmath134 to be the connected component of @xmath135 that corresponds to the `` tube '' @xmath129 : @xmath136 $ ] .
let @xmath137 for all @xmath13 .
we notice that each @xmath138 , etc .
is smooth .
the vertices @xmath139 correspond to the connected components containing points @xmath14 with @xmath16 .
there are two types of vertices : the interior vertices ( in fig.1 they are @xmath140 ) are the intersection of three edges ; the exterior vertices ( in fig.1 they are @xmath141 ) are the endpoints of only one edge . using the ideas in @xcite with a little modification we can establish the weak convergence of the process @xmath142 ( which is not markov in general ) as @xmath97 in the space @xmath143}({\gamma})$ ] to a certain markov process @xmath90 on @xmath88 . a sketch of the proof of this fact is in ( * ?
* ; * ? ? ?
* ; * ? ? ?
* section 2 ) .
the process @xmath90 is a diffusion process on @xmath88 with a generator @xmath28 and the domain of definition @xmath91 .
we are going now to define the operator @xmath28 and its domain of definition @xmath91 . for each edge
@xmath113 we define an operator @xmath144 : @xmath145 here @xmath146 is the average of the velocity field @xmath49 on the connected component @xmath134 , with respect to lebesgue measure in @xmath61-direction . at places where @xmath147 , the above expression for @xmath148 is understood as a limit as @xmath149 : @xmath150 we will assume throughout this paper the following .
* assumption 7 . * the function @xmath151 .
we notice that this is a bit different from the corresponding one in ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* assumption 6 ) .
we point out that the vanishing mean drift assumption is crucial for the method of our analysis to work .
thus under our assumption 7 we have @xmath152 the operator @xmath144 can be represented as a generalized second order differential operator ( see @xcite ) @xmath153 where , for an increasing function @xmath154 , the derivative @xmath155 is defined by @xmath156 , and @xmath157 is the scale function , @xmath158 is the speed measure .
the operator @xmath28 is acting on functions @xmath159 on the graph @xmath88 : for @xmath160 being an interior point of the edge @xmath113 we take @xmath161 .
the domain of definition @xmath91 of the operator @xmath28 consists of such functions @xmath159 satisfying the following properties . @xmath162 the function @xmath159 is a continuous function that is twice continuously differentiable in @xmath4 in the interior part of every edge @xmath113 ; @xmath162 there exist finite limits @xmath163 ( which are taken as the value of the function @xmath164 at the point @xmath116 ) ; @xmath162 there exist finite one - sided limits @xmath165 along every edge ending at @xmath166 and they satisfy the gluing conditions @xmath167 where the sign `` @xmath168 '' is taken if the values of @xmath4 for points @xmath169 are @xmath170 and `` @xmath171 '' otherwise . here
@xmath172 ( when @xmath116 is an exterior vertex ) or @xmath173 ( when @xmath116 is an interior vertex ) .
for an exterior vertex @xmath166 with only one edge @xmath113 attached to it the condition ( 2.1 ) is just @xmath174 .
such a boundary condition can also be expressed in terms of the usual derivatives @xmath175 instead of @xmath176 .
it is @xmath177 .
we remark that we are in dimension 2 so that these exterior vertices are accessible , and the boundary condition can be understood as a kind of ( not very standard ) instantaneous reflection . in dimension 3 or higher
these endpoints do not need a boundary condition , they are just inaccessible . for an interior vertex the gluing condition ( 2.1 ) can be written with the derivatives @xmath175 instead of @xmath176 . for @xmath112 being one of the @xmath178 we define @xmath179 ( for each edge @xmath113 the limit is a one - sided one ) . then the condition ( 2.1 )
can be written as @xmath180 it can be shown as in ( * ? ? ?
* ; * ? ? ?
* section 2 ) that the process @xmath90 exists as a continuous strong markov process on @xmath88 .
we fix the shape of @xmath9 . for every @xmath44 , every @xmath181 and every @xmath182
let us consider the distribution @xmath183 of the trajectory @xmath142 starting from a point @xmath184 in the space @xmath143}({\gamma})$ ] of continuous functions on the interval @xmath185 $ ] with values in @xmath88 : the probability measure defined for every borel subset @xmath186}({\gamma})$ ] as @xmath187 .
similarly , for every @xmath111 and @xmath188 let @xmath189 be the distribution of the process @xmath90 in the same space : @xmath190 .
the following theorem is basic for our analysis .
* theorem 2.1 . * _ for every @xmath191 and every @xmath188 the distribution @xmath183 converges weakly to @xmath192 as @xmath97 . _ in other words we have @xmath193 for every bounded continuous functional @xmath194 on the space @xmath143}({\gamma})$ ] .
the proof of this theorem follows from @xcite and there is a sketch in ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* section 3 ) .
we omit duplicating the details here .
we recall that our definition of the generalized solutions to ( 1.2 ) and ( 1.4 ) are the solutions of the integral equations ( 1.3 ) and ( 1.5 ) , respectively .
that is to say , we have @xmath82\ ] ] and @xmath197 \ .\ ] ] * theorem 3.1 .
* _ there exist unique bounded measurable generalized solutions @xmath98 , @xmath198 and @xmath99 , @xmath199 for ( 1.2 ) and ( 1.4 ) , respectively .
these solutions are continuous for all @xmath94 .
_ * proof .
* we take ( 1.2 ) as an example .
the proof for ( 1.4 ) is exactly the same .
we shall prove the existence and regularity by using a contraction mapping principle ( compare with @xcite ) . to this end
we consider the banach space @xmath200 of bounded measurable functions on @xmath185\times [ d]$ ] with norm @xmath201\times [ d]}|v(t , x)|$ ] .
consider in @xmath200 the following operator @xmath202=f[v](t,(x , z))={\mathbf e}_{(x , z)}^w\left[\exp\left(\int_0^t c(v(t - s , { \boldsymbol{x}}_s^{\varepsilon}))ds\right)g(x_t^{\varepsilon})\right ] \ , \
v\in b_t \ .\ ] ] it is then checked that we have , for @xmath203 , that @xmath204-{\mathbf e}_{(x , z)}^w\left[\exp\left(\int_0^t c(v(t - s,{\boldsymbol{x}}_s^{\varepsilon})ds\right)g(x_t^{\varepsilon})\right]\right| } \\ {
\displaystyle}{\leq \|g\|\left|{\mathbf e}_{(x , z)}^w\left[\exp\left(\int_0^t c(u(t - s,{\boldsymbol{x}}_s^{\varepsilon}))ds\right)-\exp\left(\int_0^t c(v(t - s,{\boldsymbol{x}}_s^{\varepsilon}))ds\right)\right]\right| } \\ { \displaystyle}{\leq \|g\|\exp(c(0)t)\cdot\text{lip}(c)\cdot t \cdot \|u - v\| } \\=c(t_0)\|u - v\| \end{array}\ ] ] for @xmath205 and @xmath206 .
this guarantees that @xmath194 is a contraction provided that @xmath207 . by contraction mapping theorem
we have existence and uniqueness of generalized solution in the space of bounded measurable functions to the problem ( 1.2 ) on the interval @xmath208 $ ] . since the solution @xmath209 we can use the same @xmath210 and work with intervals @xmath211 $ ] ,
... , up to @xmath212 $ ] ( @xmath213 ) provided that @xmath214 .
this gives existence `` in the large '' for a unique generalized solution @xmath215 for ( 1.2 ) in the space of bounded measurable functions .
the continuity of the solution @xmath215 in the variables @xmath216 and @xmath217 is provided by ( 1 ) the continuity of @xmath218 ; ( 2 ) the lipschitz continuity of @xmath219 ; ( 3 ) the continuity and continuous dependence of the process @xmath70 on @xmath220 .
@xmath221 we shall then show the approximation of the generalized solution @xmath98 as @xmath222 is small by the generalized solution @xmath223 .
we prove this via a sequence of auxiliary results .
* lemma 3.1 . * _ we have _ @xmath224 * proof . *
we consider the stopping time @xmath225 . by strong markov property of the process
@xmath70 we have @xmath226 } \\ = { \displaystyle}{{\mathbf e}_{(x , z_2)}^w\left[\exp\left(\int_0^{\tau^{\varepsilon } } c(u^{\varepsilon}(t - s , x_s^{\varepsilon } , z_s^{\varepsilon}))ds\right)u^{\varepsilon}(t-\tau^{\varepsilon } , ( x_{\tau^{\varepsilon}}^{\varepsilon } , z_1))\right ] } \ . \end{array}\ ] ] since the motion @xmath227 is moving very fast as @xmath97 we see that @xmath228 almost surely as @xmath97 .
this immediately implies the convergence .
@xmath221 let @xmath229 .
we introduce a new function @xmath230 and we see from lemma 3.1 that we have the following .
* corollary 3.1 . * _ we have _
@xmath231 we are going now to prove that the function @xmath232 has a uniform in @xmath222 bounded first derivative in the variable @xmath4 .
* lemma 3.2 . *
_ we have an a - priori estimate _ @xmath233 _ where _ @xmath234 _ is independent of _ @xmath222 .
* by ( 1.3 ) we have @xmath235 \ .\ ] ]
differentiating with respect to @xmath4 we have @xmath236 } \ . \end{array}\ ] ] note that @xmath237 therefore if we let @xmath238 , we get @xmath239 where @xmath240 are bounded with their bound depending on the regularity of @xmath60 , @xmath219 and the shape parameter @xmath241 , yet independent of @xmath222 .
we then apply a gronwall inequality to conclude that we have an a - priori estimate @xmath242 where @xmath234 is independent of @xmath222 . from the above estimate and taking into account the smoothness of the shape parameter @xmath241
, we see that the a - priori estimate in the statement of the lemma holds .
in fact , by the definition of @xmath243 , we have @xmath244 thus @xmath245 we notice that @xmath246 here @xmath247 .
this implies ( 3.3 ) . @xmath221 making use of theorem 2.1 , corollary 3.1 and lemma 3.2 we can prove the following .
* theorem 3.2 . * _ we have _ @xmath248 * proof . * an outline of this proof is mentioned at the end of ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* chapter 7 ) .
we fulfill the details here .
let @xmath100 . from ( 1.3 ) and ( 1.5 ) we see that @xmath249-{\mathbf e}_{(x , k)}^w\left[\exp\left(\int_0^t c(u(t - s , y_s))ds\right)g(x_t)\right]\right| } \\ \leq \|g\|\cdot[(i)+(ii)+(iii ) ] \ . \end{array}\ ] ] here @xmath250
@xmath251 @xmath252 thus we see that @xmath253 as @xmath97 due to ( 3.2 ) ; @xmath254 as @xmath97 due to the weak convergence of the processes @xmath142 to @xmath90 on @xmath88 as @xmath97 in @xmath143}({\gamma})$ ] ( theorem 2.1 ) and ( 3.3 ) ; @xmath255 can be bounded by a constant multiple of @xmath256 plus a term going to @xmath57 as @xmath97 ( due to ( 3.2 ) ) .
we then apply a standard technique via gronwall s inequality and we can conclude .
this section will be devoted to obtaining some auxiliary results which will be used in sections 56 .
let the random variable @xmath257 for @xmath258 . intuitively , @xmath259 is the first time that the process @xmath90 , starting from @xmath260 , comes back to @xmath110 with the value of its @xmath4-component @xmath261 .
we recall that @xmath110 is the long line in @xmath88 corresponding to the domain @xmath5 . in the same way we define @xmath262 for @xmath263 .
let @xmath264 .
let the function @xmath265 \eqno(4.2)\ ] ] for @xmath13 .
we remind the reader of a small notational convention here . in this section for convenience of notation
we have a minus sign in front of the stopping time @xmath266 in ( 4.2 ) . in the sections
56 we will drop this minus sign and instead we will be mainly working with @xmath267 .
let us first consider the case when @xmath268 .
the function @xmath269 is the solution of the following sturm - liouville problem @xmath270 to solve the above problem we shall first recall the basic theory of feller ( @xcite ) .
we follow here @xcite and we also refer the reader to ( * ? ? ?
* ; * ? ? ?
* lemma 2.10 ) . without loss of generality
let us first work with some interval @xmath271 $ ] for some @xmath272 .
we consider the eigenvalue problem associated with the generalized second - order differential operator @xmath273 : @xmath274 on an interval @xmath275 $ ] . here
@xmath276 is the speed measure and @xmath277 is the scale function .
the function @xmath278 is a strictly increasing continuous function on @xmath279 and the function @xmath280 is a strictly increasing function on @xmath279 continuous to the right .
the generalized derivatives are defined as @xmath281 where @xmath282 and @xmath159 is a real function defined in a neighborhood of @xmath4 .
there are two basic solutions @xmath283 , @xmath284 of the equation ( 4.4 ) with @xmath285 and @xmath286 ; the function @xmath283 is increasing in @xmath4 and @xmath284 is decreasing in @xmath4 ; the derivatives @xmath287 , @xmath288 are increasing functions .
moreover , an explicit representation of the functions @xmath289 is available ( @xcite ) .
we set , for @xmath290 , @xmath291 $ ] , @xmath292 let @xmath293 it could be justified that the above series converges .
we can easily check that @xmath294 let @xmath295 @xmath296 moreover , we can calculate the derivative @xmath297 @xmath298 then we have @xmath299 and @xmath300 in terms of generalized derivatives we see that the above is equivalent to @xmath301 a general solution of ( 4.4 ) can be represented as a linear combination @xmath302 the constants @xmath303 and @xmath304 are determined by boundary conditions to be specified . in the case when @xmath305 situation is similar and we have to make small changes accordingly . to be more precise
, we can treat the point @xmath75 as the point @xmath57 and the point @xmath57 as the point @xmath75 . the formulas ( 4.5.0 )
( ( 4.5.0@xmath306 ) ) , ( 4.5.1 ) and ( 4.5.2 ) ( ( 4.5.2@xmath306 ) ) have to be changed accordingly . in the rest of this section we will be mainly performing detailed steps in the calculations assuming @xmath272 and we will present corresponding results when @xmath305 without a detailed calculation .
let us come back to our problem ( 4.3 ) .
first of all we note that the structure of the graph @xmath88 consists of two types of edges : the first type of edges are lined up together forming the edge @xmath110 and we label them as @xmath307 , @xmath308 ; the second type of edges correspond to the wings and we label them as @xmath309 , @xmath308 .
these edges are labeled in a consecutive way ( see fig.1 ) .
let the projection of the second type of edges @xmath309 onto the @xmath4-direction be isomorphic to @xmath310 $ ] for @xmath311 and @xmath312 $ ] for @xmath313 .
let the interval @xmath307 be isomorphic to @xmath314 $ ] .
we solve the problem @xmath315 on each edge @xmath316 $ ] ( the first type ) , @xmath317 $ ] for @xmath311 and @xmath318 $ ] for @xmath313 ( the second type ) .
we notice that in this case when we represent the operator @xmath144 as a generalized second order derivative operator @xmath319 we will have @xmath320 and @xmath321 the general solution is represented as @xmath322 . here
@xmath323 and @xmath324 are the two basic solutions corresponding to the interval @xmath113 and we identify @xmath4 with some @xmath325 ( or its projection onto the @xmath4-axis , anyway ) .
we note that they are random solutions .
the constants @xmath326 and @xmath327 are to be determined .
we shall seek for a solution @xmath328 whenever @xmath329 .
thus we have @xmath330 in the above @xmath331 are the corresponding cross - section width of the channel at the junctions .
we have @xmath332 .
* lemma 4.1 . *
_ we have _
@xmath333 _ here _
@xmath334 _ with _ @xmath335 @xmath336 _ if @xmath311 ; and _ @xmath337 _ if @xmath313 .
_ * proof . * the first three equalities in ( 4.6 ) will give us @xmath338 where we can calculate the random matrix @xmath339 .
let us first consider the case when @xmath311 . from the first equation of ( 4.6 )
we see that we have @xmath340 we have @xmath341 , @xmath342 , @xmath343 .
so from the second equality of ( 4.6 ) we get @xmath344 the third equality in ( 4.6 ) gives us @xmath345 from ( 4.10 ) , ( 4.11 ) and ( 4.12 ) we can conclude that @xmath346 here the random variables @xmath347 and @xmath348 are defined by ( 4.8 ) and ( 4.9.1 ) . in the case when @xmath313 we just have to replace in ( 4.11 ) the coefficient @xmath349 by @xmath350 and we have to change the sign in front of @xmath351 in ( 4.12 ) to minus .
we thus get ( 4.9.2 ) .
@xmath221 let @xmath352 .
due to shift invariance ( stationarity ) we see that without loss of generality we can assume that @xmath353 and @xmath73 have the same distribution .
* lemma 4.2 . * _ the random variables _ @xmath354 , @xmath308 _ are identically distributed .
_ * proof . *
we have , by strong markov property of the process @xmath90 on @xmath88 , that for @xmath264 , @xmath355={\mathbf e}^w[e^{-{\lambda}t^{l_{k+1}}_{l_{k}}}]{\mathbf e}^w[e^{-{\lambda}t^{l_k}_0 } ] \ .\ ] ] by stationarity we see that @xmath356 $ ] has the same distribution as @xmath357 $ ] .
since @xmath355=u(l_{k+1})=c_{2k-1}^+\ ] ] and @xmath358=u(l_k)=c_{2k-1}^- \ , \ ] ] we see that the distribution of @xmath354 is independent of @xmath308 .
@xmath221 * lemma 4.3 . * _ for any @xmath308 we have @xmath359 almost surely with respect to @xmath40 . _ * proof . * this is because we have @xmath360\in [ 0,1]$ ] and lemma 4.2 .
@xmath221 it is convenient to introduce the notation @xmath361 here @xmath362 is given by the formula ( 4.5.0 ) specified in the interval @xmath113 .
* lemma 4.4 . * _ for any @xmath308 we have @xmath363 @xmath364 almost surely with respect to @xmath40 .
_ * proof .
* by lemma 4.1 we have @xmath365 making use of ( 4.5.2 ) and ( 4.5.3 ) we see that @xmath366 . thus @xmath367 and since we have lemma 4.2 and lemma 4.3 we see that @xmath364 , almost surely with respect to @xmath40 . @xmath221 * lemma 4.5 . * _ we have @xmath368 and thus @xmath369 almost surely with respect to @xmath40 . _ * proof .
* this is a simple consequence of lemma 4.1 , lemma 4.3 and lemma 4.4 .
@xmath221 * lemma 4.6 . * _ we have _
when @xmath311 ; and _ @xmath371 _ when @xmath313 .
_ * proof .
* let us first consider the case when @xmath311 . by ( 4.9.1 ) we can write @xmath372 here @xmath373 @xmath374 by making use of ( 4.5.2 ) and @xmath375 as well as @xmath376 it is straightforward to calculate that @xmath377 @xmath378 @xmath379 @xmath380 @xmath381 so we get @xmath382 thus we get ( 4.13.1 ) . the equality ( 4.13.2 )
is obtained in a similar way .
@xmath221 making use of lemma 4.6 and basic calculations ( 4.5.0)(4.5.4 ) , as well as our assumptions 2 and 6 , we see that we have the following .
* corollary 4.1 . * @xmath383 combining lemmas 4.5 and 4.6 we colculde that @xmath384 .
thus we have the following .
* theorem 4.1 .
* @xmath385|]<\infty \ ; \ { \mathbf e}[|\ln { \mathbf e}^w[e^{-{\lambda}t_0^{-l}}]|]<\infty \ .\ ] ] in the above theorem the second inequality is estimated in a similar fashion as the first one .
* lemma 4.7 . *
_ we have @xmath386 and @xmath387 almost surely with respect to @xmath40 .
_ * proof . *
we take @xmath388 as an example .
the case for @xmath389 is similar .
let @xmath390}^{(x , k)}<\infty)$ ] where @xmath229 .
here @xmath391}^{(x , k)}$ ] is the first time the process @xmath90 , starting from @xmath392 , hits the point @xmath393 or @xmath394 .
we set @xmath395 where we identify @xmath229 with some @xmath325 ( or projection of @xmath113 onto the @xmath4-axis , anyway ) .
then we have @xmath396 this gives @xmath397 @xmath398 @xmath399 thus @xmath400 we see that @xmath401 .
and we have recursively that @xmath402 thus we see that @xmath403 thus @xmath404 by assumption 1 ( @xmath405 ) .
in particular , @xmath386 almost surely with respect to @xmath40 .
@xmath221 * lemma 4.8 . *
_ we have @xmath406=+\infty$ ] and @xmath407=+\infty$ ] almost surely with respect to @xmath40 .
_ * proof .
* we take @xmath388 as an example .
the proof for @xmath408 is similar .
we show by comparison . to this end
we construct the _ part _ of the process @xmath409 within the domain @xmath5 ( compare with ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* section 4.2 ) ) .
let @xmath410 be an additive functional , which is called _ the proper time _ of the domain @xmath5 .
we introduce the time @xmath411 inverse to @xmath412 and continuous on the right .
let @xmath413 .
one can show that as @xmath97 the weak convergence of @xmath414 to @xmath415 .
the process @xmath415 is described as a one - dimensional diffusion process on @xmath17 with gluing conditions ( see ( * ? ? ?
* ; * ? ? ?
* section 4.2 ) ) .
we have @xmath416 where @xmath417 is the proportion of time of process @xmath90 spent inside @xmath110 .
we see that @xmath418 where @xmath419 .
it is not hard to prove , via an approximation similar as in ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* section 4.2 ) , that @xmath420=+\infty$ ] .
more precisely , let @xmath421 then we have @xmath422=\lim{\limits}_{a{\rightarrow}\infty}v(l ; a)$ ] .
since we can estimate @xmath423 and we have our assumption 1 , we see that @xmath406=+\infty$ ] .
we are interested in describing the wave front propagation corresponding to the solution @xmath99 of ( 1.5 ) . to this end
we study the quenched large deviation principle for the random variable @xmath424 . here
@xmath425 , @xmath426 and @xmath427 is the first component of the process @xmath89 on @xmath88 starting from a point @xmath428 . here
@xmath112 may be @xmath57 or some other integer @xmath429 depending on the structure of @xmath88 .
this is in essence an adaptation of the arguments of @xcite and @xcite . here and below , for notational convenience we will use the symbol @xmath430 to denote the process @xmath431 ( which is the first component of the process @xmath432 ) starting from a point @xmath433 on @xmath88 with an arbitrary choice of @xmath112 .
the fact that the large deviation results for the random variable @xmath424 are independent of the choice of @xmath112 will be revealed in the proof of theorem 5.2 .
let @xmath434 and we introduce @xmath435 \ .
\eqno(5.1)\ ] ] recall that @xmath73 is the distance between two consecutive vertices @xmath116 and @xmath139 at which there is an edge corresponding to a wing .
we see that @xmath73 is a random variable measurable with respect to the filtration @xmath436 generated by the shape of @xmath9 . for each fixed shape of @xmath9
the random variables @xmath437 and @xmath389 are well defined and they are measurable with respect to the filtration generated by the wiener process @xmath71 . notice that by our assumption 6 we have @xmath438 where @xmath439 are constants .
* lemma 5.1 . * _ suppose that @xmath434 is such that @xmath440|\right)<\infty \ ; \ { \mathbf e}\left(|\ln { \mathbf e}^w[e^{{\lambda}t_0^{-l}}{\mathbf 1}_{t_0^{-l}<\infty}]|\right)<\infty \ .\ ] ] _ _ let @xmath434 and @xmath441 . then almost surely with respect to @xmath40 the limits _
e}l}{\mathbf e}\left(\ln { \mathbf e}^w[e^{{\lambda}t_0^l}{\mathbf 1}_{t_0^l<\infty}]\right)\ ] ] @xmath443=\dfrac{1}{{\mathbf e}l}{\mathbf e}\left(\ln { \mathbf e}^w[e^{{\lambda}t_0^{-l}}{\mathbf 1}_{t_0^{-l}<\infty}]\right)\ ] ] _ hold .
the convergence is uniform with respect to @xmath444 and @xmath445 as @xmath444 and @xmath445 vary in a set that is bounded and @xmath446 is bounded away from zero .
moreover , @xmath447 is independent of @xmath444 and @xmath445 . _
* let us just work with @xmath448 .
the proof of this fact is essentially the same as that of ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* section 2 , proposition 1 ) provided we make small modifications . in fact , by the strong markov property of the process @xmath90 on @xmath88 it is easy to deduce that for @xmath449 we have @xmath450 let there be located @xmath451 edges that correspond to the `` wings '' in the interval @xmath452 $ ] .
we see that @xmath453 holds @xmath40a.s .. on the other hand , we have , by the ergodic theorem , that @xmath454\right)\ ] ] holds @xmath40a.s .. thus we see that @xmath455\right)\ ] ] provided that @xmath440|\right)<\infty \ .\ ] ] the rest of the argument is the same as in ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* section 2 , proposition 1 ) .
@xmath221 we note that by theorem 4.1 the requirements of lemma 5.1 always hold for @xmath456 .
let @xmath457 our theorem 4.1 implies that @xmath458 .
* lemma 5.2 . *
( properties of the function @xmath459 ) \(1 ) _ @xmath460 ; _
\(2 ) _ @xmath461 for @xmath267 ; _
\(3 ) _ @xmath462 as @xmath463 ; _ \(4 ) _ @xmath464 as @xmath465 ; _
\(5 ) _ @xmath459 is convex for @xmath466 ; _
\(6 ) _ for @xmath467 , @xmath459 is differentiable and @xmath468 } { { \mathbf e}^w[e^{{\lambda}t_0^{\pm l}}{\mathbf 1}_{t_0^{\pm l}<\infty}]}\right ] \ ; \ ] ] in particular , @xmath469\in ( 0,\infty ] \ ; \ ] ] _
\(7 ) _ @xmath470 is monotonically strictly increasing for @xmath471 ; _
\(8 ) _ @xmath472 and therefore @xmath473 . _
* proof . *
the proof of this lemma is the same as in ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* lemma 2.2 , proposition 4.2 ) .
the last statement ( 8) follows from our lemmas 4.7 and 4.8 . @xmath221 we define @xmath474 * lemma 5.3 . * ( properties of the function @xmath475 ) \(1 ) _ @xmath476 for @xmath477 ; _
\(2 ) _ @xmath475 is convex and decreasing in @xmath478 for @xmath479 ; _
\(3 ) _ @xmath480 and @xmath481 . _
* proof . *
the proof of this lemma is the same as in @xcite .
@xmath221 * theorem 5.1 . * ( large deviation principle for hitting time ) _ almost surely with respect to @xmath40 the following estimates hold .
let @xmath482 and @xmath441 .
for any closed set @xmath483 we have @xmath484@xmath485 and for any open set @xmath486 we have @xmath487 @xmath488 _ * proof . *
the proof of this theorem is the same as the proof of ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* theorem 2.3 ) . for the sake of completeness
we shall briefly repeat it here .
we prove the first and third bounds for example .
the second and fourth estimates are the same .
let @xmath489 .
let us consider the upper bound first .
we have , by chebyshev s inequality , that @xmath490 thus we see that @xmath491 since @xmath492 .
we now derive the lower bound .
let @xmath493 and @xmath494 .
let @xmath495 be the @xmath496-ball centered at @xmath196 .
let @xmath497 be such that @xmath498 now we make use of a cramr s change of measure .
let @xmath499 @xmath500 \ .\ ] ] then we get @xmath501 } \ . \end{array}\ ] ] one can show in the same way as in @xcite and @xcite , that @xmath502 suppose we already have ( 5.2 ) , then we can conclude that we have @xmath503-{\delta}|{\lambda}_u| } \\ { \displaystyle}{=(v - c)i^+\left(\dfrac{u}{v - c}\right)-{\delta}|{\lambda}_u| } \end{array}\ ] ] which implies the lower bound .
* theorem 5.2 . * ( large deviation principle ) _ almost surely with respect to @xmath40 the following estimates hold .
let @xmath504 and @xmath505 $ ] .
for any closed set @xmath506 we have _
@xmath507 _ and for any open set @xmath508 we have _
@xmath509 _ for any closed set @xmath510 $ ] we have _ @xmath511 _ and for any open set @xmath512 $ ] we have _ @xmath513 * proof . *
we show the first two estimates as an example .
the last two estimates are the same .
we shall make use of the duality @xmath514 here @xmath515 .
we have @xmath516 here @xmath24 is the first time that the process @xmath90 , starting from @xmath517 , arrives at @xmath110 ; @xmath518 is the first time that the process @xmath90 arrives at the first branching point @xmath519 on @xmath88 with @xmath4-coordinate @xmath520 ; @xmath521 is the first time that the process @xmath90 , starting from @xmath522 , arrives at @xmath523 .
we note that by our assumption 6 in probability @xmath76 the distances @xmath524 , @xmath525 are finite . on the other hand , as @xmath102 , by law of large numbers for stationary sequences we have @xmath526 and @xmath527 almost surely .
thus for fixed @xmath528 we have @xmath529 from here we have @xmath530 which proves the upper bound .
we now derive the lower bound .
we have , for @xmath531 , @xmath532 the second term in the above formula can be estimated by using space reversal invariance and the corresponding large deviation principle , in the same way as ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* proof of theorem 2.4 ) and ( * ? ? ?
* section 5 ) , provided that we have ( 5.3 ) .
it turns out that @xmath533 so then we have , by ( 5.3 ) again and theorem 5.1 , @xmath534 this proves the upper bound . @xmath221
after we get the quenched large deviation principle we study the wave front propagation of the solution @xmath99 of ( 1.5 ) making use of the arguments of @xcite , @xcite and ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* chapter 7 ) .
we apply theorem 5.2 with @xmath548 and @xmath549 .
as @xmath102 we see that , for @xmath550 such that @xmath551 and @xmath44 ; or for @xmath552 such that @xmath553 and @xmath44 we have @xmath554 .
this proves the lemma . @xmath221
* this lemma is proved in the same way as ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* lemma 4.1 ) , ( * ? ? ?
* ; * ? ? ?
* corollary 1 ) , provided that we have the estimates ( 6.3 ) and ( 6.4 ) in the following lemma .
@xmath221 * lemma 6.3 . * _ for any @xmath504 and @xmath559 we have _
@xmath560}|x^x(s)-x|\geq \eta t\right ) = 0 \ . \eqno(6.3)\ ] ] _ also , for a given @xmath37 there exists @xmath561 sufficiently small so that _ @xmath562 .
_ comets , f. , gantert , n. , zeitouni , o. , quenched , annealed and functional large deviations for one - dimensional random walk in random environment , _ probability theory and related fields _ , * 118 * ( 2000 ) , pp . | we consider a reaction - diffusion equation in narrow random channels .
we approximate the generalized solution to this equation by the corresponding one on a random graph .
by making use of large deviation analysis we study the asymptotic wave front propagation .
_ keywords _ : reaction - diffusion equation , wave front propagation , diffusion processes on graphs , random environment .
_ 2010 mathematics subject classification numbers _ : 35k57 , 35a18 , 60j60 , 60k37 . |
the trembling motion ( zitterbewegung , zb ) , first devised by schroedinger for free relativistic electrons in a vacuum @xcite , has become in the last two years subject of great theoretical interest as it has turned out that this phenomenon should occur in many situations in semiconductors @xcite .
whenever one deals with two or more energy branches , an interference of the corresponding upper and lower energy states results in the trembling motion even in the absence of external fields . due to a formal similarity between two interacting bands in a solid and the dirac equation for relativistic electron in a vacuum one can use methods developed in the relativistic quantum mechanics for non - relativistic electrons in solids @xcite .
most of the zb studies for semiconductors took as a starting point plane electron waves ( see , however , refs .
@xcite ) . on the other hand
, lock @xcite in his important paper observed : such a wave is not localized and it seems to be of a limited practicality to speak of rapid fluctuations in the average position of a wave of infinite extent. using the dirac equation lock showed that , when an electron is represented by a wave packet , the zb oscillations do not remain undamped but become transient .
in particular , the disappearance of oscillations at sufficiently large times is guaranteed by the riemann - lebesgue theorem as long as the wave packet is a smoothly varying function .
since the zb is by its nature not a stationary state but a dynamical phenomenon , it is natural to study it with the use of wave packets .
these have become a practical instrument when femtosecond pulse technology emerged ( see ref .
@xcite ) . in the following we study theoretically the zitterbewegung of mobile charge carriers in three modern materials : bilayer graphene , monolayer graphene and carbon nanotubes .
we have three objectives in mind .
first , we obtain for the first time analytical results for the zb of gaussian wave packets which allows us to study not only the trembling motion itself but also effects that accompany this phenomenon .
second , we describe for the first time the transient character of zb in solids , testing on specific examples the general predictions of ref .
third , we look for observable phenomena and select both systems and parameters which appear most promising for experiments .
we first present our analytical results for bilayer graphene and then quote some predictions for observable quantities in monolayer graphene and carbon nanotubes .
two - dimensional hamiltonian for bilayer graphene is well approximated by @xcite @xmath0 where @xmath1 .
the form ( [ bg_h ] ) is valid for energies @xmath2mev in the conduction band .
the energy spectrum is @xmath3 , where @xmath4 , i.e. there is no energy gap between the conduction and valence bands .
the position operator in the heisenberg picture is a @xmath5 matrix @xmath6 we calculate @xmath7,\ ] ] where @xmath8 .
the third term represents the zitterbewegung with the frequency @xmath9 , corresponding to the energy difference between the upper and lower energy branches for a given value of @xmath10 .
we want to calculate the zb of a charge carrier represented by a two - dimensional wave packet @xmath11 the packet is centered at @xmath12 and is characterized by a width @xmath13 .
the unit vector @xmath14 is a convenient choice , selecting the [ 11 ] component of @xmath15 , see eq .
( [ bg_x_11 ] ) .
an average of @xmath16 is a two - dimensional integral which we calculate analytically @xmath17 where @xmath18 $ ] , and @xmath19 , \ \ \ \ \\end{aligned}\ ] ] in which @xmath20 contains the time dependence .
we enumerate the main features of zb following from eqs .
( [ bg_x(t ) ] ) and ( [ bg_xz(t ) ] ) .
first , in order to have the zb in the direction @xmath21 one needs an initial transverse momentum @xmath22 .
second , the zb frequency depends only weakly on the packet width : @xmath23 , while its amplitude is strongly dependent on the width @xmath13 .
third , the zb has a transient character , as it is attenuated by the exponential term . for small @xmath24
the amplitude of @xmath25 diminishes as @xmath26 with @xmath27 fourth , as @xmath24 ( or @xmath28 ) increases the cosine term tends to unity and the first term in eq .
( [ bg_xz(t ) ] ) cancels out with the second term , which illustrates the riemann - lebesgue theorem ( see ref .
after the oscillations disappear , the charge carrier is displaced by the amount @xmath29 , which is a remnant of zb .
fifth , for very narrow packets ( @xmath30 ) the exponent in eq .
( [ bg_xz(t ) ] ) tends to unity , the oscillatory term is @xmath31 and the last term vanishes . in this limit
we recover undamped zb oscillations .
zitterbewegung of a charge carrier in bilayer graphene _ versus _ time , calculated for a gaussian wave packet width @xmath32 and @xmath33m@xmath34 : a ) displacement , b ) electric current , c ) probability densities for upper and lower components of the wave function , d ) dispersion @xmath35 .
the decay time is @xmath36 fs , see eq .
( [ bg_gammaz]).,width=321,height=428 ] next , we consider observable quantities related to the zb , beginning by the current .
the latter is given by the velocity multiplied by charge .
the velocity is simply @xmath37 , where @xmath38 is given by eq .
( [ bg_xz(t ) ] ) .
the calculated current is plotted in fig .
[ fig1]b , its oscillations are a direct manifestation of zb .
the zitterbewegung is also accompanied by a time dependence of upper and lower components of the wave function . to characterize this evolution
we define probability densities for the upper and lower components @xmath39 where @xmath40 , and the time - dependent wave function is @xmath41 .
we have @xmath42 where @xmath43 . for @xmath44
( [ bg_psit ] ) reduces to eq .
( [ bg_psi ] ) .
the calculated probability densities are @xmath45,\end{aligned}\ ] ] where @xmath46 . the time dependence of @xmath47 is illustrated in fig .
clearly , there must be @xmath48 at any time , but it is seen that the probability density flows back and forth between the two components .
it is clear that the oscillating probability is directly related to zb . since in bilayer graphene
the upper and lower components are associated with the first or second layer , respectively @xcite , it follows that , at least for this system , the trembling motion represents oscillations of a charge carrier between the two graphene layers . for sufficiently long times
there is @xmath49 , so that the final probability is equally distributed between the two layers . for a very narrow packet ( @xmath50 )
we have @xmath51 $ ] , which indicates that the probability oscillates without attenuation .
for @xmath52 there is @xmath53 $ ] , i.e. there are no oscillations and the initial probability @xmath14 simply decays into @xmath54 . the above phenomenon can be considered from the point of view of the entropy : @xmath55@xmath56 . at the beginning
( the carrier is in one layer ) the entropy is zero , at the end ( when the probability is equally distributed in two layers ) the entropy is @xmath57 .
however , the entropy increases in the oscillatory fashion ( see ref .
@xcite ) .
the transient character of zb is accompanied by a temporal spreading of the wave packet .
in fact , the question arises whether the damping of zb is not simply _ caused _ by the spreading of the packet . to study this question we calculate an explicit form of the wave function given by integrals ( [ bg_psit ] ) .
the result is @xmath58 , \ \
\\end{aligned}\ ] ] @xmath59 @xmath60 , \ \\end{aligned}\ ] ] where @xmath61 and @xmath62 .
it is seen that the packet , which was gaussian at @xmath44 ( see eq .
( [ bg_psi ] ) ) , is _ not _ gaussian at later times ( see discussion ) .
the upper and lower components have the same decay time , oscillation period , etc . in order to characterize the spreading ( or dispersion ) of the packet we calculate its width @xmath35 as a function of time
@xmath63 ^ 2 = \langle \psi(\bm r , t)|\hat{\bm r}^2 - \langle \hat{\bm r } \rangle^2 | \psi(\bm r , t)\rangle,\ ] ] where @xmath64 is the above two - component wave function and @xmath65 .
the calculated width @xmath66 is plotted versus time in fig .
it is seen that during the initial 80 femtoseconds the packet s width increases only twice compared to its initial value , while the zb and the accompanying effects disappear almost completely during this time .
we conclude that the spreading of the packet is _ not _ the main cause of the transient character of the zb . in fact , also the spreading oscillates a little , but this effect is too small to be seen in fig . [ fig1]d .
it is well known that the phenomenon of zb is due to an interference of wave functions corresponding to positive and negative eigen - energies of the initial hamiltonian .
looking for physical reasons behind the transient character of zb described above , we decompose the total wave function @xmath64 into the positive ( @xmath67 ) and negative ( @xmath68 ) components @xmath69 and @xmath70 .
we have @xmath71 where @xmath72 and @xmath73 are the eigen - functions of the hamiltonian ( [ bg_h ] ) in @xmath74 space corresponding to positive and negative energies , respectively .
further @xmath75 after some manipulations we finally obtain @xmath76 the function @xmath70 is given by the identical expression with the changed signs in front of @xmath77 and @xmath78 terms .
there is @xmath79 and @xmath80 .
now we calculate the average values of @xmath81 and @xmath82 using the positive and negative components in the above sense .
we have @xmath83 so that we deal with four integrals .
a direct calculation gives @xmath84 @xmath85 where @xmath29 and @xmath25 have been defined in eq .
( [ bg_x(t ) ] ) .
thus the integrals involving only the positive and only the negative components give the constant shift due to zb , while the mixed terms lead to the zb oscillations .
all terms together reconstruct the result ( [ bg_x(t ) ] ) .
next we calculate the average value @xmath82 .
there is no symmetry between @xmath81 and @xmath82 because the wave packet is centered around @xmath86 and @xmath87 .
the average value @xmath82 is again given by four integrals .
however , now the mixed terms vanish since they contain odd integrands of @xmath88 , while the integrals involving the positive and negative components alone give @xmath89 this means that the positive and negative subpackets move in the opposite directions with the same velocity @xmath90 .
the relative velocity is @xmath91 .
each of these packets has the initial width @xmath13 and it ( slowly ) spreads in time . after the time @xmath92 the distance between the two packets equals @xmath13 , so the integrals ( [ bg_x_mix ] ) are small , resulting in the diminishing zitterbewegung amplitude .
this reasoning gives the decay constant @xmath93 , which is exactly what we determined above from the analytical results ( see eq . ( [ bg_gammaz ] ) ) .
thus , _ the transient character of the zb oscillations is due to the increasing spacial separation of the subpackets _ corresponding to the positive and negative energy states .
this confirms our previous conclusion that it is not the packet s slow spreading that is responsible for the attenuation ( see discussion ) .
however , as we show below , also spreading may possibly play this role in some cases . to conclude our analytical discussion of the zb in bilayer graphene we consider an interesting property of the velocity squared .
if @xmath94 and @xmath95 are calculated directly from the hamiltonian ( [ bg_h ] ) , then it is easy to show that @xmath96 and @xmath97 , so that @xmath98 does not depend on time . in the heisenberg picture we split the velocity components into classical and zb parts @xmath99 and similarly for @xmath100 . noting that @xmath101 we have @xmath102 .
( [ bg_vx(t ) ] ) we show that each of these terms is time independent : @xmath103 and @xmath104 , and similarly for @xmath105 and @xmath106 .
thus , the velocity squared of the zb component @xmath107=@xmath108=@xmath109 is equal to that of the classical component @xmath110=@xmath111=@xmath109 .
now we turn to monolayer graphene . the two - dimensional band hamiltonian describing
its band structure is @xcite @xmath112 where @xmath113cm / s .
the resulting energy dispersion is linear in momentum : @xmath114 , where @xmath115 . the quantum velocity in the schroedinger picture is @xmath116 , it does not commute with the hamiltonian ( [ mg_h ] ) . in the heisenberg picture we have @xmath117 . using eq .
( [ mg_h ] ) we calculate @xmath118 the above equation describes the trembling motion with the frequency @xmath119 , determined by the energy difference between the upper and lower energy branches for a given value of @xmath10 . as before , the zb in the direction @xmath21 occurs only if there is a non - vanishing momentum @xmath120 .
we calculate an average velocity ( or current ) taken over the wave packet given by eq .
( [ bg_psi ] ) . the averaging procedure amounts to a double integral .
the latter is not analytical and we compute it numerically .
the results for the current @xmath121 are plotted in fig .
[ fig2 ] for @xmath122m@xmath123 and different realistic packet widths @xmath13 ( see ref .
it is seen that the zb frequency does not depend on @xmath13 and is nearly equal to @xmath124 given above for the plane wave . on the other hand ,
the amplitude of zb does depend on @xmath13 and we deal with decay times of the order of femtoseconds . for small @xmath13
there are almost no oscillations , for very large @xmath13 the zb oscillations are undamped .
these conclusions agree with our analytical results for bilayer graphene .
the behavior of zb depends quite critically on the values of @xmath125 and @xmath13 , which is reminiscent of the damped harmonic oscillator .
oscillatory electric current caused by the zb in monolayer graphene _ versus _ time , calculated for a gaussian wave packet with @xmath122m@xmath123 and various packet widths @xmath13.,width=321,height=321 ]
finally , we consider monolayer graphene sheets rolled into single semiconducting carbon nanotubes ( cnt ) .
the band hamiltonian in this case is similar to eq .
( [ mg_h ] ) except that , because of the periodic boundary conditions , the momentum @xmath126 is quantized and takes discrete values @xmath127 , where @xmath128 , @xmath129 , @xmath130 , and @xmath131 is the length of circumference of cnt @xcite . as a result ,
the free electron motion can occur only in the direction @xmath132 , parallel to the tube axis .
the geometry of cnt has two important consequences .
first , for @xmath130 there _ always _ exists a non - vanishing value of the quantized momentum @xmath133 .
second , for each value of @xmath134 there exists @xmath135 resulting in the same subband energy @xmath3 , where @xmath136 the time dependent velocity @xmath100 and the displacement @xmath137 can be calculated for the plane electron wave in the usual way and they exhibit the zb oscillations ( see ref .
@xcite ) . for small momenta
@xmath138 the zb frequency is @xmath139 , where @xmath140 .
the zb amplitude is @xmath141 .
however , we are again interested in the displacement @xmath142 of a charge carrier represented by a one - dimensional wave packet analogous to that described in eq .
( [ bg_psi ] ) @xmath143 the average displacement is @xmath144 where @xmath145 and @xmath146\exp(b^2)$ ] , where @xmath147 and @xmath148 is the error function . the zb oscillations of @xmath142 are plotted in fig .
[ fig3 ] for @xmath149 , @xmath130 and @xmath150 .
it can be seen that , after the transient zb oscillations disappear , there remains a shift @xmath151 .
thus the zb separates spatially the charge carriers that are degenerate in energy but characterized by @xmath152 and @xmath153 quantum numbers .
the current is proportional to @xmath154 , so that the currents related to @xmath155 and @xmath156 cancel each other . to have a non - vanishing current one needs to break the above symmetry , which can be achieved by applying an external magnetic field parallel to the tube axis @xcite .
zitterbewegung of two charge carriers in the ground subband of a single carbon nanotube of @xmath150 _ versus _ time ( logarithmic scale ) , calculated for gaussian wave packets of two different widths @xmath13 .
after the zb disappears a constant shift remains .
the two carriers are described by different quantum numbers @xmath157 . at higher times
the amplitude of zb oscillations decays as @xmath158 ( see text).,width=321,height=321 ] above we considered the situation in which a non - vanishing value of transverse momentum @xmath159 is built in by the tube topology .
however , it is also possible to prepare a wave packet with an initial non - vanishing momentum @xmath125 . using the method presented above for bilayer graphene ( see eq .
( [ bg_psit_s ] ) ) we can decompose the total wave packet ( [ cnt_psi ] ) into the positive and negative subpackets with the result @xmath160 the function @xmath161 is given by a similar expression with the changed signs in front of @xmath77 and @xmath162 terms .
here we use the notation @xmath163 and @xmath164 . now
the oscillating part of @xmath82 is , as before @xmath165 for @xmath52 the above @xmath166 reduces to @xmath167 given by eq .
( [ cnt_y_osc ] ) .
the average contributions of positive ( or negative ) terms alone are @xmath168 where @xmath169 is the packet function .
the sum of the first terms for @xmath170 and @xmath161 in eq .
( [ cnt_psip2y ] ) gives @xmath171 , as before .
for @xmath52 the second term vanishes which physically means that the relative velocity of the two subpackets is zero , so that they stay together in time .
it is for this reason that the decay of zb is slow ( see fig .
[ fig3 ] ) .
if @xmath172 , the second term in eq . ( [ cnt_psip2y ] ) does not vanish , the two subpackets run away from each other , their overlap diminishes and the zb disappears much more quickly .
the question remains : what is the physical reason for the ( slow ) damping of the zb electron shown in fig .
[ fig3 ] , if the subpackets stay together ?
( as we mentioned in the introduction , the mathematical expression for the damping phenomenon is the riemann - lebesgue theorem . ) trying to answer this question we calculated the spreading of the wave subpackets ( [ cnt_psit_pos ] ) in time . for the initial width @xmath173
the subpackets reach the width @xmath174 after the time of @xmath175 fs ( see fig . [ fig3 ] ) .
thus , we would be tempted to say that it is the spreading of the packets that is responsible for the attenuation of zb .
however , it should be noted that , while at higher times the packet s dispersion is linear in time ( see ref . @xcite and fig .
[ fig3]d ) , the amplitude of zb oscillations decays as @xmath158 .
a similar slow damping of zb occurs for one - dimensional relativistic electrons in a vacuum if the average momentum of the subpackets is zero ( see discussion ) .
it is of interest that the zb phenomena similar to those described above occur also for wave packets representing free relativistic electrons in a vacuum governed by the dirac equation .
this confirms again the strong similarity of the two - band models for non - relativistic electrons in solids to the description of free relativistic electrons in a vacuum , see refs .
in contrast to bilayer graphene , the kinematics of the one - dimensional relativistic wave packets may not be described analytically , so the solutions were computed numerically and visualized graphically by thaller @xcite .
it was shown that : 1 ) an initial relativistic gaussian wave packet after spreading is not gaussian any more .
this is analogous to our eqs .
( [ bg_psi_up ] ) and ( [ bg_psi_low ] ) .
2 ) if an average momentum of the initial positive and negative subpackets is zero , the overlap of the two subpackets remains almost constant in time and the resulting zb decays quite slowly .
this corresponds to our considerations of cnt with @xmath52 , see fig .
( it is to be reminded that the two overlapping subpackets are orthogonal to each other . )
3 ) if the initial average momentum of both subpackets is nonzero , the two subpackets quickly run away from each other and the zb falls quickly since it is sustained only when the subpackets have some overlap in the position space .
this corresponds to our considerations of bilayer graphene , see fig .
[ fig1 ] .
the transient zb of free relativistic wave packets in a vacuum was also studied numerically by braun _
it was shown that , for example , the decay times of a typical wave packet having the width @xmath176 and the initial wave vector @xmath177 a.u .
is @xmath178 fs .
this should be compared with our predicted decay times of @xmath36 fs for bilayer graphene .
it turns out once again that solids are much more promising media for an observation of zitterbewegung than a vacuum .
the zitterbewegung phenomenon described above should not be confused with the bloch oscillations of charge carriers in superlattices , although the latter occur at picosecond frequencies and have comparable picoseconds decay times ( see e.g. refs .
however , the bloch oscillations are basically a one - band phenomenon , they have been realized in superlattices ( although this is in principle not the condition _ sine qua non _ ) and , most importantly , they require an external electric field driving electrons all the way to the brillouin zone boundary . on the other hand , the zb needs at least two bands and it is a no - field phenomenon . on the other hand , narrow - gap superlattices could provide a suitable medium of its observation . in view of our results
it is clear that , in order to observe the transient zitterbewegung , it is necessary to prepare simultaneously a sufficient number of charge carriers in the form of wave packets .
if one wants to detect the current , the trembling motion of all carriers must have the same phase . on the other hand ,
if one wants to see only the remnant displacement , the phase coherence is not necessary . as we said above , the zb frequency is to a good accuracy given by the corresponding energy difference between the upper and lower energy branches while the amplitude depends strongly on packet s width .
for the two graphene materials considered above one needs an initial momentum in one direction to have the zb along the transverse direction ( see also ref .
@xcite ) . for nanotubes
the initial momentum is automatically there due to the circular boundary conditions .
the oscillatory motion between two graphene layers , as illustrated in fig .
[ fig1]c , appears to be a promising phenomenon for observation .
as far as the detection is concerned , one needs sensitive current meters or scanning probe microscopy , both working at infrared frequencies and femtosecond to picosecond decay times ( see refs .
in summary , using the two - band structure of bilayer graphene , monolayer graphene and carbon nanotubes we show that charge carriers in these materials , localized in the form of gaussian wave packets , exhibit the transient zitterbewegung with the decay times of femtoseconds in graphene and picoseconds in nanotubes .
observable dynamical zb effects , most notably the electric current , are described .
it is demonstrated that , after the trembling motion disappears , there remains its trace in the form of a persistent charge displacement .
it is emphasized that the described zb in solids is in close analogy to that of the relativistic electron in a vacuum .
99 e. schroedinger , sitzungsber .
* 24 * , 418 ( 1930 ) .
schroedinger s derivation is reproduced in a.o .
barut and a.j .
bracken , phys .
d * 23 * , 2454 ( 1981 ) .
f. cannata , l. ferrari and g. russo , sol .
. comun . * 74 * , 309 ( 1990 ) ; l. ferrari and g. russo , phys . rev .
b * 42 * , 7454 ( 1990 ) .
w. zawadzki , phys .
b * 72 * , 085217 ( 2005 ) .
w. zawadzki , phys .
b * 74 * , 205439 ( 2006 ) . j. schliemann , d. loss and r.m .
westervelt , phys .
* 94 * , 206801 ( 2005 ) ; phys .
b * 73 * , 085323 ( 2006 ) .
katsnelson , europ . phys
. j. b * 51 * , 157 ( 2006 ) .
rusin and w. zawadzki , j. phys .
matter * 19 * , 136219 ( 2007 ) .
j. cserti and g. david , phys .
b * 74 * , 172305 ( 2006 ) .
r. winkler , u. zuelicke and j. bolte , cond - mat/0609005 ( 2006 ) .
b. trauzettel , y.m .
blanter and a.f .
morpurgo , phys .
b * 75 * , 035305 ( 2007 ) .
w. zawadzki , in _ optical properties of solids _ , edited by e.d .
haidemenakis ( gordon and breach , new york , 1970 ) , p. 179 .
w. zawadzki , in _ high magnetic fields in the physics of semiconductors ii _ , edited by g. landwehr and w. ossau ( world scientific , singapore , 1997 ) , p. 755 .
k. huang , am . j. phys . *
20 * , 479 ( 1952 ) .
lock , am .
* 47 * , 797 ( 1979 ) .
b.m . garraway and k.a .
suominen , rep .
58 * , 365 ( 1995 ) .
e. mccann and v.i .
falko , phys . rev .
* 96 * , 086805 ( 2006 ) .
novoselov , e. mccann , s.v .
morozov , v.i .
falko , m.i .
katsnelson , u. zeitler , d. jiang , f. schedin and a.k .
geim , nature physics * 2 * , 177 ( 2006 ) . m. novaes and m.a.m .
de aguiar , phys .
e * 70 * , 045201(r ) ( 2004 ) .
wallace , phys . rev . * 71 * , 622 ( 1947 ) . j.c .
slonczewski and p.r .
weiss , phys . rev . * 109 * , 272 ( 1958 ) . j.w .
mcclure , phys .
rev . * 104 * , 666 ( 1956 ) .
novoselov , a.k .
geim , s.v .
morozov , d. jiang , m.i .
katsnelson , i.v .
grigorieva , s.v .
dubonos and a.a .
firsov , nature * 438 * , 197 ( 2005 ) .
y. zhang , y.w .
tan , h.l .
stormer and p. kim , nature * 438 * , 201 ( 2005 ) .
sadowski , g. martinez , m. potemski , c. berger and w.a .
de heer , phys .
* 97 * , 266405 ( 2006 ) .
j. schliemann , phys .
b * 75 * , 045304 ( 2007 ) .
r. saito , g. dresselhaus and m.s .
dresselhaus , _ physical properties of carbon nanotubes _ ( imperial college press , london , 1999 ) .
h. ajiki and t. ando , j. phys .
jpn * 62 * , 2470 ( 1993 ) .
b. thaller , arxiv : quant - ph/0409079 ( 2004 ) .
braun , q. su and r. grobe , phys .
a * 59 * , 604 ( 1999 ) .
r. martini , g. klose , h.g .
roskos , h. kurz , h.t .
grahn and r. hey , phys .
b * 54 * , r14325 ( 1996 ) .
lyssenko , g. valusis , f. loser , t. hasche , k. leo , m.m .
dignam and k. kohler , phys .
lett . * 79 * , 301 ( 1997 ) .
kosevich , a.b .
hummel , h.g .
roskos and k. kohler , phys .
lett . * 96 * , 137403 ( 2006 ) .
topinka , b.j .
leroy , s.e.j .
shaw , e.j .
heller , r.m .
westervelt , k.d .
maranowski and a.c .
gossard , science * 289 * , 2323 ( 2000 ) .
leroy , j. ph .
matter * 15 * , r1835 ( 2003 ) . | observable effects due to trembling motion ( zitterbewegung , zb ) of charge carriers in bilayer graphene , monolayer graphene and carbon nanotubes are calculated .
it is shown that , when the charge carriers are prepared in the form of gaussian wave packets , the zb has a transient character with the decay time of femtoseconds in graphene and picoseconds in nanotubes .
analytical results for bilayer graphene allow us to investigate phenomena which accompany the trembling motion .
in particular , it is shown that the transient character of zb in graphene is due to the fact that wave subpackets related to positive and negative electron energies move in opposite directions , so their overlap diminishes with time .
this behavior is analogous to that of the wave packets representing relativistic electrons in a vacuum . |
while magnetic t - tauri type models for low - mass star formation have been reasonably well accepted , there is currently no such picture for the early evolutionary phases of higher mass stars .
many questions , such as whether higher mass stars form via disc accretion , and issues relating to the importance of magnetic fields during high - mass star formation , remain open .
herbig ae / be stars are of considerable interest in attempts to resolve these questions , as they are young stars of intermediate mass ( 2 - 15m@xmath3 ) at the interface between low and high - mass star formation .
in addition , herbig ae / be stars are the most massive stars with a visible pre - main sequence phase .
t tauri stars have relatively strong magnetic fields @xcite and these fields have important physical effects on the accretion flow .
it is generally believed that magnetic fields play a lesser role for stars more massive than the sun , as stars at spectral types a and earlier lack convective outer mantles .
the picture of the role of magnetic accretion as a function of spectral type changed when @xcite found that a large fraction ( 9/11 ) of herbig ae stars show changes in the linear polarisation intrinsic to the h@xmath0 line ( extended to 10/13 with data from * ? ? ?
* ) similar to the fraction ( 9/10 ) of such intrinsic line effects in t tauri stars @xcite .
the origin for these h@xmath0 line effects in both t tauri and herbig ae stars is believed to be the result of the fact that _ emission line _ photons from the central object are scattered off a rotating accretion disc with an inner hole @xcite . from here
the material is probably funnelled along magnetic field lines @xcite , onto the stellar surface .
@cccccccc@ object name & & sp . type & & + & @xmath4 & @xmath5 & & @xmath4 & @xmath5 & @xmath4 & @xmath5 + mwc 166 & 6.90&7.35 & b0 iv & 30/12/96 + 01/01/97@xmath6 & 29/09/04 & @xmath7 & 480 + mwc 1080 & 11.02&13.28 & b0 & @xmath7 & 28/09/04 & @xmath7 & 4320 + & & & & 29/09/04 & 29/09/04 & 2880 & 3600 + gu cma & 6.64&6.67 & b2 v & 11/07/95@xmath6 & 28/09/04 & @xmath7 & 720 + mwc 361 & 7.27&7.69 & b2v & 18/12/99@xmath8 & 28/09/04 & @xmath7 & 1600 + il cep & 9.03&9.80 & b2 iv@xmath7v & 19/12/99@xmath8 & 29/09/04 & @xmath7 & 1440 + bd + 404124&10.53&11.37 & b2 & 19/12/99@xmath8 & 28/09/04 & @xmath7 & 2160 + @xmath9 ori & 4.56&4.38 & b3 iii@xmath7iv & 11/01/95@xmath6 & 28/09/04 & @xmath7 & 144 + mwc 147 & 8.73&9.06 & b6 v & 18/12/99@xmath8 & 28/09/04&@xmath7 & 1920 + ab aur & 7.03&7.15 & a0 v & 18/12/99@xmath8 & 30/09/04 & @xmath7 & 1920 + mwc 120 & 7.92&7.95 & a0 & 29/09/04 & 29/09/04 & 1080 & 2640 + mwc 480 & 7.63&7.91 & a2 & 30/09/04 & 30/09/04 & 1800 & 1320 + xy per & 9.41&10.04 & a2 ii & 20/12/99@xmath8 & 29/09/04 & @xmath7 & 1200 + + [ tab : observations ] early herbig be stars ( spectral types b0-b5 , see * ? ? ?
* ) show very different behaviour in their h@xmath0 polarimetry .
although the frequency of observed effects is high ( 7/12 ; * ? ? ?
* ) , and wholly consistent with all herbig be stars being embedded in flattened circumstellar media ( c.f .
classical be stars ) , their h@xmath0 polarisation behaviour is very different from that of herbig ae / t tauri stars , as it is not the line but the continuum that is found to be polarised , with the h@xmath0 line `` depolarised '' with respect to the continuum @xcite .
so , in summary , the emission line photons are intrinsically polarised compared to the continuum for herbig ae and t tauri stars , but are depolarised relative to the continuum for herbig be stars .
the difference in the behaviour of the linear polarisation across h@xmath0 between the herbig be and ae stars may be an indication that there is a transition in the hertzsprung - russell diagram from magnetic accretion at spectral type a to disc accretion at spectral type b. however , alternatively , @xcite considered the option that the compact polarised h@xmath0 emission occurring in the herbig ae and t tauri stars may be masked in the herbig be stars due to their much higher levels of h@xmath0 emission .
we therefore designed an observational test to distinguish between these two options .
the test involves linear spectropolarimetry across less opaque , higher excitation emission lines that can be expected to arise in the same location as the compact h@xmath0 component that causes the line polarisation in the herbig ae stars .
if the extension and optical depth of h@xmath0 emission masks the evidence for a compact source in herbig be stars , then higher balmer ( h@xmath1 and h@xmath2 ) observations should reveal position angle rotations ( @xmath10 loops ) in the herbig be group .
however , if these angle rotations are not observed at these lines in the herbig be population , while they do appear at h@xmath0 among the herbig ae stars , this would suggest that the circumstellar environments of herbig be stars are physically different from those of herbig ae stars .
the goal of this paper is thus to perform spectropolarimetry of h@xmath1 and h@xmath2 observations in early herbig be stars .
the paper proceeds as follows : we begin by discussing the observations and data reduction in @xmath112 , then we discuss the results in @xmath113 before providing discussion of these results and concluding remarks in @xmath114 .
detailed comments of the results for each object in our sample are presented in appendix a , divided between herbig be stars ( @xmath11 [ s_a_hbe ] ) and herbig ae stars ( @xmath11 [ s_a_hae ] ) in our sample .
the linear spectropolarimetric data were taken using the 4.2 m william herschel telescope ( wht ) , la palma over 3 nights in late september 2004 .
the isis spectrograph , a calcite block , and a rotating half - wave plate were used for the observations , as well as a dekker and slit .
the dekker had three 5 holes , separated by 18 , while a 1 slit width was used for all object observations .
the calcite block acted to separate the incoming light into two perpendicular rays , the o ( ordinary ) and e ( extraordinary ) rays , so one image contains a set of o and e rays for the image and two sets for the sky .
rotation of the half - wave plate then allowed measurement of the polarisation at the angles 0@xmath12 , 45@xmath12 , 22.5@xmath12 and 67.5@xmath12 .
one complete data set is therefore formed by combining four images , one at each of the four angles mentioned above .
all bright early herbig be stars ( from the catalogue by * ? ? ?
* ) visible from la palma in september 2004 were observed .
we note that we strictly stick to their criteria , even though the evolutionary status of objects like @xmath9 ori is debated .
these data were supplemented by four herbig ae stars with similar criteria , as a basis for comparison .
@cccccc@ object name & sp . type & & + & & @xmath4 & @xmath5 & @xmath4 & @xmath5 + mwc 166 & b0 iv & 0.49 @xmath13 0.01@xmath6 & 0.37 @xmath13 0.01 & 44.9 @xmath13 0.1@xmath6 & 37.2 @xmath13 0.5 + mwc 1080 & b0 & @xmath7 & 1.71 @xmath13 0.02 & @xmath7 & 72.2 @xmath13 0.3 + & & 1.51 @xmath13 0.01 & 1.99 @xmath13 0.02 & 77.2 @xmath13 0.1 & 72.1 @xmath13 0.3
+ gu cma & b2 v & 1.15 @xmath13 0.01 @xmath6 & 1.25 @xmath13 0.01 & 18.9 @xmath13 0.1@xmath6 & 17.2 @xmath13 0.1 + mwc 361 & b2v & 0.82 @xmath13 0.01@xmath8 & 0.88 @xmath13 0.01 & 95.6 @xmath13 0.1@xmath8 & 96.4 @xmath13 0.1 + il cep & b2 iv@xmath7v & 4.24 @xmath13 0.01@xmath8 & 4.18 @xmath13 0.02 & 101.6 @xmath13 0.1@xmath8 & 100.7 @xmath13 0.1 + bd + 404124 & b2 & 1.22 @xmath13 0.01@xmath8 & 1.37 @xmath13 0.01 & 8.7 @xmath13 0.2@xmath8 & 17.4 @xmath13 0.2 + @xmath9 ori & b3 iii@xmath7iv & 0.27 @xmath13 0.01@xmath6 & 0.77 @xmath13 0.01 & 55.7 @xmath13 0.6@xmath6 & 43.9 @xmath13 0.1 + mwc 147 & b6 v & 1.05 @xmath13 0.01@xmath8 & 0.85 @xmath13 0.01 & 100.0 @xmath13 0.1@xmath8 & 102.4 @xmath13 0.1 + ab aur & a0 v & 0.11 @xmath13 0.01@xmath8 & 0.24 @xmath13 0.01 & 53.3 @xmath13 0.9@xmath8 & 54.5 @xmath13 0.6 + mwc 120 & a0 & 0.41 @xmath13 0.01 & 0.40 @xmath13 0.01 & 115.4 @xmath13 0.2 & 117.0 @xmath13 0.1 + mwc 480 1 & a2 & 0.20 @xmath13 0.01 & 0.33 @xmath13 0.01 & 64.9 @xmath13 1.3 & 64.3 @xmath13 0.6 + xy per & a2 ii & 1.60 @xmath13 0.01@xmath8 & 1.60 @xmath13 0.01 & 131.6 @xmath13 0.1@xmath8 & 125.5 @xmath13 0.1 + + [ tab : continuum ] the data includes images in both the @xmath5 ( 4295 @xmath74955 ) and @xmath4 ( 6150 @xmath76815 ) bands , with mean wavelength dispersions of 0.22 / pixel and 0.45 / pixel respectively .
for the @xmath5 band data , the r1200b grating and the 4096x2048 pixel eev12 ccd detector on the isis blue arm were used , resulting in a spectral resolution ( measured from arc lines ) of 51 km / s around h@xmath1 . in the @xmath4 band
, observations were taken using the r1200r grating and the marconi2 ccd on the isis red arm , resulting in a spectral resolution of 34 km / s around h@xmath0 . in order to eliminate instrumental asymmetry ,
observations for each object were taken with the object in the a ( left most ) and b ( central ) dekker holes .
the dekker hole c was not used for object observations , but the sky was always measured at either a or b , whichever was not being used for the target .
care was taken not to saturate the ccd for objects with particularly strong emission lines by taking short exposures .
the seeing on the first two nights was fair ( @xmath14 ) but became poor ( @xmath15 ) on the third night due to clouds .
data reduction was carried out on each image frame using the figaro software maintained by starlink , and consisted of bias subtraction , cosmic ray removal , bad pixel correction , spectrum straightening and flat fielding .
wavelength calibration was performed using several calibration spectra ( obtained by observing a copper - argon - neon lamp ) taken throughout the observing runs .
the images were then imported into the starlink program ccd2pol , which is part of the time - series / polarimetry ( tsp ) package , in order to obtain the stokes parameters for each polarisation set .
the resultant data was then fed into the polmap package also maintained by starlink , in order to measure the polarisation and polarisation angle , and to produce plots of the data .
the objects observed included several polarisation and zero - polarisation standards .
these were observed so that the instrumental polarisation and instrumental polarisation angle could be measured by comparing the initial results for these objects with the literature . from these data
, we found that the polarisation p has a systematic error of @xmath16 0.1@xmath17 and the polarisation angle @xmath18 has a systematic error of @xmath19 1@xmath12 . while in principle
the achieved accuracy is governed by photon statistics alone as @xmath20 ( typically 0.01@xmath17 ) , it can be seen that these calibration and telescope - induced errors are the main limiting factor .
we do not correct for instrumental or interstellar polarisation because these only add a wavelength - independent constant to the stokes @xmath21 and @xmath22 parameters . due to this wavelength independence
, we concentrate on the @xmath23 representation of the hydrogen emission line polarisation signatures when classifying the observed behaviour .
furthermore , it is often not possible to distinguish the effects of the interstellar medium from that of the material immediately surrounding the target object such as dust .
( 1,0)475 [ f : mwc361 ]
we have listed the observational properties for both the @xmath4 and @xmath5 band observations for all herbig ae and be stars in table @xmath24 . for 9 of the 12 objects , additional @xmath4 band data from @xcite and @xcite have been listed . for 3 objects ,
h@xmath0 data were taken in 2004 .
the continuum polarisation was measured for all targets in both @xmath5 and @xmath4 , and is listed in table @xmath25 .
this acts as a reasonable indicator of variability , although some intrinsic difference is expected between the two bands . in some cases ,
the data from different epochs show little variation and a direct comparison is possible . for those cases
in which time - variability might be an issue , we will discuss the possible implications .
line characteristics , polarisation effect properties and the measured intrinsic polarisation , where possible , are summarised for all objects in table @xmath26 . in the following section , we first provide a brief discussion regarding the definition of observed line effects ( @xmath11 [ s_charac ] ) , before we present the polarisation data of our herbig ae / be objects , highlighting representative data for some individual objects in @xmath11 [ s_hbe ] and @xmath11 [ s_hae ] , while we present the remaining data with accompanying notes in the appendix .
we finish the section with a global comparison between herbig ae and herbig be stars ( @xmath11 [ s_overview ] ) . in order to characterise the observed line effects in terms of `` depolarisation '' or `` @xmath10 loop '' ,
we briefly discuss the definitions , following the method first used in @xcite . for depolarisation , the line effect needs to be broad compared to the emission line .
if a pa ` flip ' occurs , i.e. a sharp change in the profile from above to below continuum levels , or vice versa , then this can not be caused by depolarisation , but the line itself must be intrinsically polarised . in @xmath10 space , the following classification can be followed : where no change in polarisation or pa is observed , the @xmath10 loci appear as a ` spot ' . in the case of a depolarisation , a linear excursion from , and returning to ,
the central spot is observed ( e.g. see the top three panels in fig .
@xmath27 ) , whilst in the case of a pa ` flip ' , a loop is observed in @xmath10 space ( e.g. see the bottom three panels of fig .
@xmath27 ) .
figure @xmath28 shows first the polarisation data for mwc 361 in the top panels .
this object has the clearest depolarisation line effects in the sample and is therefore an excellent example to start the description .
the top half of the figure shows , from left to right , the data on h@xmath2 , h@xmath1 and h@xmath0 respectively .
the triplots show from bottom to top the intensity spectrum , the polarisation percentage and polarisation angle rebinned to a constant accuracy per pixel , which is shown to the top right of each plot .
below the triplots , the _ qu _ graphs are shown . as expected , the line emission is stronger when going to lower order recombination lines .
indeed , h@xmath2 emission is obviously present , but barely exceeds the continuum , as it fills in the underlying photospheric absorption line .
nevertheless , even the h@xmath2 line shows a depolarisation similar to h@xmath0 , despite its much smaller transition probability .
the intrinsic polarisation measured from the line excursions is @xmath29 and 3@xmath30 ( with errors of order 3@xmath30 ) respectively for the @xmath5 and @xmath4 band data . in all bands ,
the line - effect is best described as depolarisation , although there is a somewhat more complex structure in h@xmath0 ( see * ? ? ?
the continuum polarisation and pa across the @xmath4 and @xmath31 bands are quite similar , with the difference being mainly due to the slight difference in intrinsic pa between the epochs of observation .
the behaviour in _ qu _ space is also very similar across all three lines , although the blue - to - red line ratio has changed between the two observations .
based on the similar depolarisations in the polarisation spectra as well as the similar excursion in _ qu _ space , we conclude that the ( de- ) polarisation mechanism responsible for the appearance at h@xmath0 is also acting at h@xmath1 and h@xmath2 .
therefore , a small - scale electron scattering disc , seen close to edge - on , is the most plausible explanation for the phenomena observed in mwc 361 .
let us now move to the data for @xmath9 ori , which are plotted in the lower panels of fig .
this object shows depolarisation in our data at both h@xmath1 and h@xmath2 , but it does not show any appreciable effect in the @xmath4 band data of @xcite taken nine years earlier .
however , @xcite do observe a polarisation change at h@xmath0 and closer examination of the _ qu _ plots shows that the end of the excursions at both h@xmath1 and h@xmath2 lie near the continuum cluster for h@xmath0 of @xcite .
this corroborates their suggestion that the decreased emission level at h@xmath0 at the time of observation resulted in a marked drop in the percentage of linearly polarised scattered starlight from the star . also noteworthy
is the fact that we see a line effect at h@xmath2 in the first place , although the line is in absorption . for an ordinary main - sequence star of the same spectral type as @xmath9 ori , h@xmath1 and h@xmath2 are in absorption .
the line effect across h@xmath1 appears to coincide only with the emission peaks , and not with the absorption at line centre .
there is some weak emission at h@xmath2 as well - there is just not enough to fill in the absorption . never before has such a weak emission line
been observed to have a polarisation effect .
this indicates that , as long as the data is of a high enough quality , only moderate amounts of emission are needed to cause a detectable line effect .
we obtained data in both the _ b _ and _ r _ band during the same run of the herbig ae star mwc 480 , and these are shown in fig .
mwc 480 shows no line effects in the @xmath5 band , but a strong line effect at h@xmath0 is present .
contrary to the objects discussed above , the appearance in _ qu _ space is that of a loop .
there is evidence of emission at both h@xmath1 and h@xmath2 , and as we showed in the previous subsection , if there would have been a line effect , our fairly high signal - to - noise data might have revealed it . in the case of the herbig
be stars , the polarisation is due to electron scattering in a disc , which does not result in a loop in the _ qu _ diagram .
the non - detection in these lines , while we do find an intrinsic line effect at h@xmath0 data taken nearly simultaneously is a further indication that the polarisation mechanism in herbig ae stars may be different from that in the earlier objects . for five out of the eight herbig be stars that have been observed at h@xmath1 , we observe depolarisations .
a comparison with existing h@xmath0 data shows that the intrinsic polarisation angle is similar at both h@xmath0 and h@xmath1 .
in one case , no line - effect was detected previously at h@xmath0 , but there does seem to be one at shorter wavelengths now .
this is most likely due to the intrinsic variability of the hydrogen recombination emission . the detection statistic , that more than half , but not all , of the objects are found to show a line effect is consistent with previous studies such as @xcite and @xcite .
only one of the herbig ae type stars displays a line effect at h@xmath1 ( mwc 120 , with a spectral type of a0 , the earliest herbig ae star ) .
all four objects show polarisation effects at h@xmath0 in the form of a _ qu _ loop .
as has been shown by @xcite , and additional data of @xcite , the vast majority of herbig ae stars ( 10 out of 13 ) that display a line - effect at h@xmath0 show a loop in the _ qu _ diagram .
our new data on herbig ae stars indicate that the h@xmath0 signatures are again of the loop variety , but we note that in the present data the line - effect for mwc 120 is not as clear as when it showed a loop in the older data of @xcite . on the other hand , the overwhelming majority of herbig be objects ( five out of seven ) show depolarisations at h@xmath0 .
the only herbig be star that shows a loop ( hd 58647 in * ? ? ?
* ) has a late - b spectral type ( b9 ) .
the h@xmath0 line of herbig be stars is very strong , with the line - to - continuum contrasts of these objects generally exceeding 10 ( see * ? ? ?
* ) , whereas the h@xmath0 line in the herbig ae stars is not so strong , with typical line - to - continuum contrasts in the range 3 - 4 .
@cccccccccccc@ object name & & & & + & & & & & + & h@xmath0&h@xmath1&h@xmath2&h@xmath0&h@xmath1&h@xmath2&h@xmath0&h@xmath1&h@xmath2 & @xmath4 & @xmath5 + mwc 166&@xmath714&2.9&3.5&y&n&n&exc&@xmath7&@xmath7 & 136 @xmath13 4 & @xmath7 + mwc 1080@xmath33&@xmath7&@xmath714&@xmath71.1&@xmath7&y&y&@xmath7&smear&smear&@xmath7 & 158 @xmath13 14@xmath34 + mwc 1080@xmath35&@xmath7100&@xmath715&@xmath71.4&y&y&y&smear&smear&smear & 165 @xmath13 4 & 164 @xmath13 12@xmath34 + gu cma&@xmath714&3.8&4.8&n&n&n&@xmath7&@xmath7&@xmath7&@xmath7&@xmath7 + mwc 361&@xmath765&@xmath72.5&3.2&y&y&y&exc&exc&exc & 3 @xmath13 2 & 177 @xmath13 4@xmath34 + il cep&@xmath721&3.4&4.5&n&n&n&@xmath7&@xmath7&@xmath7&@xmath7&@xmath7 + bd + 404124&@xmath7120&@xmath710&@xmath70.60&y&y&n&exc&exc&@xmath7 & 36 @xmath13 3 & 37 @xmath13 6 + @xmath9 ori&@xmath72.9&1.1&2.6&n&y&y&@xmath7&exc&exc&@xmath7 & 36 @xmath13 2@xmath34 + mwc 147&@xmath765&@xmath72.3&2.4&y&y&n&exc / loop?&exc&@xmath7 & 168 @xmath13 4 & 21 @xmath13 13 + ab aur&@xmath732&15&16&y&n&n&loop&@xmath7&@xmath7&@xmath7&@xmath7 + mwc 120&@xmath720&7.6&9.5&y&y&y&loop&loop&loop&@xmath7&@xmath7 + mwc 480&@xmath711&16&16&y&n&n&loop&@xmath7&@xmath7&@xmath7&@xmath7 + xy per&1.9&16&13&y&n&n&loop&@xmath7&@xmath7&@xmath7&@xmath7 + + [ tab : measurements ] for
the herbig be stars , our main finding is that for all cases where a line effect was found , the herbig be stars show the _ same _ type of line effect at h@xmath1 and h@xmath2 as was found earlier at h@xmath0 .
mwc 147 with a spectral type of b6 may be an exception ; the results are inconclusive .
an illustration of the differences between herbig be and ae stars is provided in fig .
[ f : convince1 ] .
here we compare the h@xmath1 line in herbig be stars , and the h@xmath0 line in herbig ae stars , as these lines are rather comparable in strength , yet we find a marked difference in the character of their line effects .
this suggests that transition probability and optical depth do not appear to affect the type of line effect observed , and the h@xmath0 feature alone should be sufficient to decide between intrinsic line polarisation ( present in the herbig ae / t tauri stars ) and depolarisation ( seen in the herbig be stars ) .
these line polarisation effects may then be used as a tool to determine whether the underlying accretion process is likely to be magnetospheric , or more characteristic of disc accretion , as motivated in the following .
we attribute the @xmath10 loops of the herbig ae / t tauri stars to magnetospheric accretion for a number of reasons .
first of all , the line is only then expected to be intrinsically polarised if the photons originate _ interior _ to the scatterers .
i.e. a compact source of line photons should be available .
secondly , the @xmath10 loops indicate that the scatterers are situated in a rotating geometry , most likely the accretion disc surrounding the central star , but this is not all : @xcite recently showed that a disc that reaches the star at its equator would show _ double _ pa flips rather than _ single _ ones .
this indicates that the line polarimetry behaviour observed in the herbig ae / t tauri stars is only expected when the accretion disc has an inner hole .
even more so , @xcite constrained the sizes of these inner holes and found these to be compatible with magnetospheric radii .
the depolarisation data seen in the herbig be stars hint at an altogether different circumstellar geometry .
first of all , the compact source of line photons is missing , whilst secondly , the continuum polarisation suggests the occurrence of electron scattering within a few stellar radii .
all in all , the depolarisation data seen in the early herbig be stars are consistent with the presence of a disc , and so suggest formation by disc accretion .
observations in the spectral range of b4-b9 could help us understand what could cause this possible shift from magnetospheric accretion in herbig ae stars to disc accretion in early herbig be stars .
we presented spectropolarimetric data on eight herbig be stars at h@xmath0 , h@xmath1 and h@xmath2 , and we found five cases where the behaviour could be described as `` depolarisation '' .
two herbig be stars show no polarisation line effects , and only one transitional object ( at spectral type b6 ) may show hints of a @xmath23 loop .
furthermore , the observed line effects are consistent across all three observed hydrogen lines , indicating that optical depth is not a major factor in the observed linear polarimetry across emission lines in herbig stars .
this leads us to conclude that it is most likely that a different accretion mechanism is operating in herbig be stars than in herbig ae stars .
the fact that the role of magnetic fields may be diminishing when going to earlier spectral types could , at first sight , be considered at odds with the fact that up to 30 - 50% of herbig stars were found to be significant emitters of x - rays @xcite , some of which have very early b - types . however , the higher spatial resolution of current x - ray satellites , in particular chandra , can be utilised to investigate whether the x - ray photons are intrinsic to the herbig star , or whether they are originating from a low - mass companion ( see e.g. * ? ? ?
in a recent infrared ( ir ) coronographic imaging and chandra study of the extremely early herbig be star mwc 297 , @xcite and @xcite have argued that the x - rays likely originate from a low - mass companion .
interestingly , an increase in the number of low - mass companions ( `` clustering '' ) with stellar mass and spectral type amongst herbig stars was found by @xcite , which @xcite considered as yielding a higher chance for misidentification of x - rays towards earlier type herbig stars .
therefore , the current status on the origin of x - rays of herbig ae / be stars suggests that t tauri and some herbig ae stars may be intrinsic x - ray emitters ( e.g. * ? ? ? * ) , while early herbig be stars are likely not .
this picture on the x - rays from herbig stars as a function of spectral type can be reconciled with the linear spectropolarimetry results presented in this paper if there is a `` fading '' of the importance of magnetic fields at spectral types earlier than b4-b9 .
it is interesting , however , that @xcite recently reported the detection of a significant magnetic field in the herbig be star mwc 361 , as this is also the object for which @xcite found a narrow pa rotation in the red wing of h@xmath0 , on top of an otherwise well - defined depolarisation .
we finally mention that infrared interferometry observations by @xcite suggest that the early herbig be stars ( b0-b3 ) are discrepant in terms of the `` disc size - source luminosity relationship '' that works so well for herbig ae and late herbig be systems .
this discrepancy could possibly be caused by the fact that different accretion scenarios are operating in the early herbig be stars and the herbig ae stars .
this appears to be in line with the line polarimetry data of herbig ae / be stars currently available : the herbig ae stars show intrinsic line polarisation , similar to t tauri stars , and consistent with magnetospheric accretion , while the majority of herbig be stars show depolarisation , consistent with disc accretion .
we thank the referee for comments that improved the clarity and contents of this paper and prof .
janet drew for many fruitful discussions .
jcm would like to thank ben davies for help with polmap .
the william herschel telescope is operated on the island of la palma by the isaac newton group in the spanish observatorio del roque de los muchachos of the instituto de astrofisica de canarias .
the allocation of time on the wht was awarded by patt , the united kingdom allocation panel .
data analysis facilities were provided by the starlink project , which is run by cclrc on behalf of pparc .
jcm is funded by the particle physics and astronomy research council of the united kingdom .
jsv acknowledges rcuk for an academic fellowship .
99 baines d. , oudmaijer r.d .
, porter j.m . , pozzo m. , 2006 , mnras , 367 , 737 damiani f. , micela g. , sciortino s. , harnden f.r . , 1994 , apj , 436 , 807 damiani f. , micela g. , sciortino s. , 2006 , a&a , 447 , 1041 hubrig s. , yudin r.v .
, schller m. , pogodin m.a . , 2006 , a&a , 446 , 1089 johns - krull c.m .
, valenti j.a . , hatzes a.p . , kanaan a. , 1999 , apj , 510 , l41 monnier j.d . , millan - gabet r. , billmeier r. et al . , 2005 ,
apj , 624 , 832 monet d.g . ,
levine s.e . , canzian b. et al . , 2003 , aj , 125 , 984 natta a. , prusti t. , neri r. , wooden d. , grinin v.p . ,
mannings v. , 2001 , a&a , 371 , 186 oudmaijer r.d . , drew j.e .
, 1999 , mnras , 305 , 166 poeckert r. , marlborough j.m . , 1976 , apj , 206 , 182 stelzer b. , micela g. , hamaguchi k. , schmitt j.h.m.m . , 2006 , a&a , 457 , 223 swartz d.a .
, drake j.j . , elsner r.f .
et al . , 2005 ,
apj , 628 , 811 symington n.h . , harries t.j . , kurosawa , r. , naylor t. , 2005 , mnras , 358 , 977 testi l. , palla f. , natta a. , 1998 , a&as , 133 , 81 th p.s .
, de winter d. , perez m.r . , 1994 , a&as , 104 , 315 vink j.s . , drew j.e . , harries t.j .
, oudmaijer r.d . , 2002 , mnras , 337 , 356 vink j.s . , drew j.e . , harries t.j .
, oudmaijer r.d . , unruh y. , 2005a , mnras , 359 , 1049 vink j.s . , harries t.j .
, drew j.e . , 2005b , a&a , 430 , 213 vink , j.s .
, oneill p.m. , els s.g . , drew j.e .
, 2005c , a&a , 438 , l21 wade g.a . , drouin d. , bagnulo s. et al . , 2005 ,
a&a , 442l , 31 wade g.a . ,
drouin d. , bagnulo s. , et al .
, 2006 , astro - ph/0601624 zinnecker h. , preibisch th . , 1994 ,
a&a , 292 , 152
these objects show line effects at all observed lines . for mwc 1080 ( fig .
@xmath37 ) though these are shallow and quite broad , except in the case of h@xmath2 on 29/04/2004 .
there is an increase in @xmath5 band continuum polarisation between the two nights of observation , but despite the large difference in continuum polarisation between @xmath5 and @xmath4 , the intrinsic pa agrees with the data taken on the 29th .
the shape of the depolarisation effect and ( q , u ) smear are similar to those observed by @xcite , although the @xmath4 band continuum polarisation they record is noticeably higher .
they attribute the smear in the ( q , u ) diagram to changing continuum polarisation , which is notable between @xmath4 and @xmath5 in our observations and , to a lesser extent , across the bands themselves .
mwc 1080 would seem to be quite variable on short timescales , and is a known binary system ( * ? ? ? * and references therein ) so will probably exhibit long term variations as well .
these objects show line effects at only some of the observed lines . bd + 404124 shows ( fig .
@xmath39 ) depolarisation at h@xmath0 and h@xmath1 , with corresponding excursions in the ( q , u ) plots , but there are no more than vague suggestions of a line effect at h@xmath2 .
both the continuum polarisation and pa increase slightly in the @xmath5 band from the @xmath4 band observations , possibly causing the slight change in shape of the line effect between h@xmath0 and h@xmath1 .
there is , however , clear and quite strong emission at all three lines , so the lack of a clear effect at h@xmath2 is most likely due to the faintness of bd + 404124 and may become apparent with longer exposure observations .
the first point of interest for mwc 147 is the fact that the emission lines have different shapes in the @xmath4 and @xmath5 bands ( fig .
@xmath40 ) . h@xmath0 ( observed 18/12/1999 ) is double peaked with a v / r peak ratio @xmath41 1 , i.e. v @xmath41 r , while h@xmath1 ( observed 28/09/2004 ) is double peaked with v @xmath42 r. h@xmath2 appears single peaked , with the emission slightly displaced to the r side of the absorption feature .
this difference between h@xmath0 and h@xmath1 is probably due to time variability .
there are definitely line effects at h@xmath0 and h@xmath1 , with little sign of effects at h@xmath2 , but it is unclear if the effects are depolarisation effects , or qu loops .
this makes it difficult to say if mwc 147 is more likely accreting via magnetospheric accretion or via disc accretion . for mwc 166 ,
a depolarisation is observed at h@xmath0 while no line effects are observed in the @xmath5 band ( fig .
@xmath43 ) .
the continuum polarisation in the @xmath5 band is lower than at @xmath4 , to the point where the continuum polarisation in @xmath5 is near the bottom of the depolarisation at h@xmath0 .
in addition , there are only slight indications of emission at h@xmath1 or h@xmath2 .
overall , it is to be expected that we do not see line effects for mwc 166 in @xmath5 , with the differences possibly caused by variations in the object between the two observations .
neither gu cma ( fig .
@xmath44 ) , nor il cep ( fig .
@xmath45 ) show any sign of line effects in either the @xmath5 or @xmath4 band .
although the @xmath4 band data was taken at a different epoch to the @xmath5 band data , the continuum polarisation percentage and pa agree , which suggests little change in the circumstellar medium around the object between these observations .
the high level of continuum polarisation of the objects , @xmath461.2@xmath17 for gu cma and @xmath464.2@xmath17 for il cep , and the high quality of the data means that it is unlikely that these objects exhibit undetected line effects , and their electron scattering discs are expected to be relatively pole - on .
these objects have line effects at only some of the observed lines .
ab aur shows clear line effects in the h@xmath0 data of @xcite but we observe no line effects in our @xmath5 band data ( fig .
@xmath47 ) .
the continuum polarisation has increased considerably , although @xcite see line effects at h@xmath0 in dec 2003 with a level of continuum polarisation in @xmath4 similar to ours .
therefore , while the polarisation level around ab aur has increased , it should not have prevented detection of the line effect .
there are signs of emission in the h@xmath1 and h@xmath2 profiles , but it is possible that it is simply not high enough for an a type star , which tend to have broader and deeper absorption lines and lower continuum emission at these wavelengths than b type stars .
mwc 120 shows the suggestion of a line effect at h@xmath2 and clear line effects at the other two observed lines ( fig .
@xmath48 ) .
the v / r ratio appears to be similar in both h@xmath1 and h@xmath0 , and there are signs of some emission at h@xmath2 .
the data is of quite a high signal - to - noise , so longer exposure may not improve the definition of the possible effect at h@xmath2 . xy per shows no sign of line effects in @xmath5 but does show line effects at h@xmath0 ( fig .
@xmath49 ) .
the continuum polarisation is consistent between the two observed bands , and there are few signs of emission in the @xmath5 band line profiles , suggesting that there are too few photons for a signature to be observed . | we present linear spectropolarimetric data for eight herbig be and four herbig ae stars at h@xmath0 , h@xmath1 and h@xmath2 .
changes in the linear polarisation are detected across all balmer lines for a large fraction of the observed objects , confirming that the small - scale regions surrounding these objects are flattened ( i.e. disk - like ) .
furthermore , all objects with detections show similar characteristics at the three spectral lines , despite differences in transition probability and optical depth going from h@xmath0 to h@xmath2 . a large fraction of early herbig
be stars ( b0-b3 ) observed show line depolarisation effects .
however the early herbig ae stars ( a0-a2 ) , observed for comparison , show intrinsic line polarisation signatures .
our data suggest that the popular magnetic accretion scenario for t tauri objects may be extended to herbig ae stars , but that it may not be extended to early herbig be stars , for which the available data are consistent with disc accretion .
[ firstpage ] techniques : spectropolarimetry - circumstellar matter - stars : emission - line , herbig , be - stars : formation - stars : pre - main - sequence |
in this article , all graphs are finite and simple .
we denote by @xmath5 the _ chromatic number _ of a graph @xmath1 , that is the minimum number of colors needed to give a color to each vertex in such a way that any two adjacent vertices receive different colors .
we denote by @xmath6 the maximum size of a set of pairwise adjacent vertices ( that we call a _ clique _ ) .
it is clear that for every graph , @xmath7 , while the converse inequality is false in general ( the smallest example is the chordless cycle on five vertices ) .
let @xmath8 be any function from @xmath9 to @xmath9 .
a graph @xmath1 is _
@xmath4-bounded by @xmath8 _ if every induced subgraph @xmath10 of @xmath1 satisfies @xmath11 .
this notion first appeared in an article of gyrfs @xcite .
a class of graphs is
_ @xmath4-bounded _
if for some function @xmath8 , every graph of the class is @xmath4-bounded by @xmath8 .
it is well known that the class of all graphs is not @xmath4-bounded , this follows from the existence of graphs with @xmath12 and arbitralily large chromatic number , see for instance @xcite .
graphs that are @xmath4-bounded by the indentity function are known as _
perfect graphs_. they were the object of much research ( see @xcite for a survey ) , and the notion of @xmath4-boundedness was invented to try to have some insight on them . in his seminal paper ,
gyrfs @xcite made many conjectures , and some of them were claiming that excluding chordless cycles with various constraints on their length should lead to @xmath4-bounded classes .
recently , much progress has been reported toward these conjectures , see for instance @xcite . in this paper , we focus on excluding cycles with contraints on their chords .
a _ unichord _ in a graph is an edge that is the unique chord of some cycle ( note that the cycle has length at least 4 because of the chord ) .
a graph is _ unichord - free _ if it does not contain any unichord .
a _ long unichord _ in a graph is an edge that is the unique chord of some cycle of length at least 5 .
the _ house _ is the graph on five vertices @xmath13 with the following edges : @xmath14 ( so the house is the smallest graph that contains a long unichord ) .
house * _ is any graph obtained from the house by repeatedly subdividing edges .
a graph is _
house*-free _ if it does not contain any house * as an induced subgraph .
it is straightforward to check that long - unichord - free graphs form a generalisation of unichord - free graphs and of house*-free graphs .
they also form a generalisation of the classical class of chordal graphs ( a graph is _ chordal _ if it contains no chordless cycle of length at least 4 ) . as classical result states that chordal
graphs are perfect ( equivalently , they are @xmath4-bounded by the indentity function ) . in @xcite
, it is proved that unichord - free graphs are @xmath4-bounded by the function @xmath15 , and in @xcite , it is proved that house*-free graphs are @xmath4-bounded by some exponential function .
we generalise these theorems and we provide a better bound for the last one by showing that long - unichord - free graphs are @xmath4-bounded by a polynomial function of degree 3 ( see theorem [ the : cluc ] ) .
our proof relies on a decomposition theorem ( see theorem [ th : maindec ] ) that is easily obtained from @xcite and @xcite .
it should be pointed out that as observed by esperet ( personal communication ) , no counter - example to the following statement is known : every @xmath4-bounded class is @xmath4-bounded by some polynomial .
it must be stressed out that applying decomposition theorems to prove @xmath4-boundedness is not at all straightforward .
there are several papers dealing with the following question : when a prescribed operation is applied repeatedly to some graphs from a class @xmath4-bounded by @xmath8 , is the larger class that is obtained @xmath4-bounded by a possibly different function @xmath16 ?
this has been answered positively for several operations , most notably for one that we use in our decomposition theorem , the so called amalgam operation ( to be defined in the next section ) , see @xcite .
but the theorem from @xcite is not enough for our purpose because it leads to an exponential @xmath4-bounding function . here , to obtain @xmath4-boundedness we prove a stronger property for the sake of induction , roughly we find graphs with a special structure that intersect all inclusion - wise maximal cliques of the graph to be colored , and we apply this procedure inductively on what remains ( where the maximum clique is smaller ) .
our decomposition theorem turns out to be a complete structural description of long - unichord - free graphs : it tells how all long - unichord - free graphs can be constructed from simple pieces . as a byproduct of this description
we obtain a polynomial time recognition algorithm for long - unichord - free graphs .
this was an open question mentioned in @xcite ( where a similar algorithm is given for unichord - free graphs ) .
it should be pointed out that in @xcite , a problem of the very same flavour is proved to be np - complete : deciding whether a graph contains a cycle @xmath17 with a unique chord @xmath18 such that @xmath19 and @xmath20 are at distance at least 4 along the cycle .
this shows that naive attempts to obtain our recognition algorithm are likely to fail . in section [ sec :
opkn ] , we define all the decompositions and operations on graphs that we need , and we survey several results about how they preserve perfection and @xmath4-boundedness . in section [ sec : subst ]
, we study a technic to prove that an operation on graphs ( namely , the substitution operation , everything is defined in the next section ) preserves @xmath4-boundedness .
this technic is analogous to the one used in the proof of the replication lemma of lovsz ( see @xcite ) .
it consists in identifying the existence of a particular subgraph that intersects all maximal cliques of a graph , and in showing that this existence is preserved by the operation .
our technic yields short proofs of known results and may provide good bounds in some situations . for intance
we prove that the closure of 3-colourable graphs under substitution is @xmath4-bounded by a quadratic function , a seemingly new result .
note that the results from section [ sec : subst ] are not used in the rest of the article .
they illustrate our method and are of independent interest . in section [ sec : amalgam ]
, we apply a similar technic to a larger set of operations ( namely , we consider 1-joins , amalgams and proper 2-cutsets ) .
the price to pay for that is that the classes of graphs where the technic can be applied are even more restricted .
but fortunately , it does not vanish to nothing as shown afterward . in section [ sec : struc ] ,
we prove the structure theorem for long - unichord - free graphs . in section [ sec : cluc ]
, we apply the results of the previous sections to prove that long - unichord - free graphs are @xmath4-bounded .
finally , in section [ sec : reco ] we provide a polynomial time algorithm to recognize long - unichord - free graphs , based on the decomposition theorem .
we now define several classical decompositions for graphs , that are all partition of the vertex - set with some structural constraints . for each of them ,
we explain how it allows to obtain smaller graphs called _ blocks of decomposition _ , and how the decomposition can be reversed into an operation that allows building a graph from smaller pieces .
a vertex @xmath21 in a graph @xmath1 is _ complete _ to @xmath22 if for all @xmath23 , @xmath24 .
we also say that @xmath21 is _ @xmath25-complete_.
a set @xmath26 is _ complete _ to a set @xmath27 disjoint from @xmath25 if every vertex of @xmath25 is @xmath28-complete .
a vertex @xmath21 in a graph @xmath1 is _ anticomplete _ to @xmath22 if for all @xmath23 , @xmath29 .
we also say that @xmath21 is _ @xmath25-anticomplete_. a set @xmath26 is _ anticomplete _ to a set @xmath27 disjoint from @xmath25 if every vertex of @xmath25 is @xmath28-anticomplete .
a ( possibly empty ) clique @xmath30 of a graph @xmath1 is a clique cutset of @xmath1 if there exists a partition @xmath31 of @xmath32 such that @xmath33 and there are no edges of @xmath1 between @xmath34 and @xmath35 .
we then say that @xmath31 is a _
split _ for this clique cutset , and that @xmath36 $ ] and @xmath37 $ ] are the _ blocks of decomposition _ of @xmath1 with respect to this split .
note that @xmath38 , and we say that @xmath1 is obtained from @xmath39 and @xmath40 by _ gluing along a clique_. this operation can be performed for any pair of graphs @xmath41 such that @xmath42 is a clique . if @xmath43 , this operation can also be refered to as _
disjoint union_. when @xmath44 , the operation can be refered to as _ gluing along a vertex _ , and the unique vertex in @xmath30 is called a _
cutvertex_. a set @xmath45 of vertices of a graph @xmath1 is a _
homogeneous set _ if @xmath46 , @xmath47 , and every vertex of @xmath48 is either complete or anticomplete to @xmath45 .
we then denote by @xmath49 the graph obtained from @xmath1 by deleting @xmath45 and adding a vertex @xmath20 adjacent to all @xmath45-complete vertices of @xmath1 . the graphs @xmath50 $ ] and @xmath49 are the _ blocks of decomposition of @xmath1 _ with respect to the homogeneous set @xmath45 . when @xmath1 is graph on at least two vertices , @xmath20 a vertex of @xmath1 and @xmath10 a graph on at least two vertices vertex - disjoint from @xmath1 , then the graph @xmath51 obtained from @xmath1 by deleting @xmath20 , adding @xmath10 and all possible edges between vertices of @xmath10 and the neighbors of @xmath20 in @xmath1 is called the graph obtained from @xmath1 by _ substituting _
@xmath10 for @xmath20 .
we also say that @xmath51 is obtained from @xmath1 and @xmath10 by a _
substitution_. observe that @xmath52 is a homogeneous set of @xmath51 .
a _ 1-join _ of a graph @xmath1 is a partition of @xmath32 into sets @xmath34 and @xmath35 such that there exist sets @xmath53 satisfying : * @xmath54 , @xmath55 ; * @xmath56 and @xmath57 ; * there are all possible edges between @xmath58 and @xmath59 ; * there are no other edges between @xmath34 and @xmath35 .
we say that @xmath60 is a _ split _ of this 1-join . for @xmath61 , the _ block of decomposition _
@xmath62 with respect to this split is the graph obtained from @xmath63 $ ] by adding a vertex @xmath64 complete to @xmath65 .
the operation that is the reverse of the 1-join decomposition is defined as follows .
start with two vertex - disjoint graphs @xmath39 and @xmath40 on at least 3 vertices .
for some vertex @xmath66 ( resp .
@xmath67 ) of @xmath39 ( resp .
@xmath40 ) such that @xmath68 ( resp .
@xmath69 ) is non - empty , @xmath1 is obtained from the disjoint union of @xmath70 and @xmath71 by adding all possible edges between @xmath68 and @xmath69 .
we say that @xmath1 is obtained from @xmath39 and @xmath40 by a _
1-join composition_. an _ amalgam _ of a graph @xmath1 is a partition @xmath72 of @xmath32 such that @xmath30 is a ( possibly empty ) clique , @xmath73 is a 1-join of @xmath1 with a split @xmath60 and @xmath30 is complete to @xmath74 ( possibly , vertices of @xmath30 have neighbors in @xmath75 ) .
we say that @xmath76 is a _
split _ of the amalgam defined above . for @xmath61 , the _ block of decomposition _
@xmath62 with respect to this split is the graph obtained from @xmath77 $ ] by adding a vertex @xmath64 complete to @xmath78 .
the operation that is the reverse of the amalgam decomposition is defined as follows .
start with two graphs @xmath39 and @xmath40 whose intersection forms a clique @xmath30 with @xmath79 and such that for @xmath61 there is a vertex @xmath64 in @xmath80 whose neighborhood is @xmath81 where @xmath65 is non - empty , disjoint from @xmath30 and @xmath30-complete .
let @xmath1 be obtained from the union of @xmath70 and @xmath71 by adding all edges between @xmath58 and @xmath59 .
we say that @xmath1 is obtained from @xmath39 and @xmath40 by an _
amalgam composition_. the amalgam is obviously a generalisation of the 1-join .
if @xmath58 or @xmath59 were allowed to be empty , it could be also be seen as a generalisation of the clique cutset , but it is not ( we keep this distinction that might seem artificial , for historical reasons and compatibility of definitions with previous papers ) . if @xmath82 then @xmath34 is a homogeneous set of @xmath1 . yet ,
formally the amalgam is not a generalisation of the homogeneous set , because a homogeneous set @xmath45 such that @xmath83 ( which is allowed ) does not imply the presence of an amalgam .
setting @xmath43 , @xmath84 , and @xmath85 does not work because then @xmath86 .
however , it works for all homogeneous sets @xmath45 such that @xmath87 .
this remark leads us to consider the following trivial decomposition and lemma .
a _ universal vertex _ in a graph @xmath1 is a vertex @xmath20 complete to @xmath88 .
note that @xmath88 is then a homogeneous set of @xmath1 ( that does not yield a 1-join or an amalgam as noted above ) .
from the discussion above , the following is trivial .
[ l : easys ] if @xmath1 has a homogenous set , then either @xmath1 has a 1-join ( and therefore an amalgam ) or @xmath1 has a universal vertex .
the operation that is the reverse of `` having a universal vertex '' is simply _ adding a universal vertex _ , which means adding a vertex @xmath20 to a graph @xmath1 , and all possible edges between @xmath20 and @xmath32 .
a _ proper 2-cutset _ of a connected graph @xmath1 is a pair of non - adjacent vertices @xmath89 , such that @xmath32 can be partitioned into non - empty sets @xmath34 , @xmath35 and @xmath90 so that : @xmath91 , @xmath57 ; there are no edges between @xmath34 and @xmath35 ; and both @xmath92 $ ] and @xmath93 $ ] contain a path from @xmath94 to @xmath95 .
we say that @xmath96 is a _ split _ of this proper 2-cutset . for @xmath61 ,
the _ block of decomposition _ @xmath62 with respect to this split is the graph obtained from @xmath97 $ ] by adding a vertex @xmath98 complete to @xmath99 .
the operation that is the reverse of the proper 2-cutset is defined as follows .
start with two graphs @xmath39 and @xmath40 whose intersection is a pair of vertices @xmath89 non - adjacent in both @xmath39 and @xmath40 , and such that @xmath89 have a common neighbor @xmath100 in @xmath39 , and a common neighbor @xmath101 in @xmath40 .
suppose furthermore that for @xmath61 , there exists a path from @xmath94 to @xmath95 in @xmath102 .
let @xmath1 be the union of @xmath103 and @xmath104 .
we say that @xmath1 is obtained from @xmath39 and @xmath40 by a _
proper 2-cutset composition_. the next two lemmas are very easy to prove and we give them without proofs .
[ l : hsubst ] suppose that @xmath1 is obtained from @xmath39 by substituting @xmath40 for @xmath20 .
if @xmath51 is an induced subgraph of @xmath1 , then either @xmath51 is isomorphic to an induced subgraph of @xmath39 , or @xmath51 is an induced subgraph of @xmath40 , or @xmath51 is obtained from an induced subgraph of @xmath39 by substituting an induced subgraph of @xmath40 for @xmath20 .
[ l : ls ] suppose that @xmath1 is obtained from @xmath39 and @xmath40 by one of the operations from @xmath105gluing along a clique , substitution , 1-join composition , amalgam composition , gluing along a proper 2-cutset@xmath106 .
if @xmath51 is an induced subgraph of @xmath1 , then either @xmath51 is isomorphic to an induced subgraph of @xmath39 , or @xmath51 is isomorphic to an induced subgraph of @xmath40 , or @xmath51 is obtained from an induced subgraph of @xmath39 and an induced subgraph of @xmath40 by an operation from @xmath107 .
the following was proved by gallai @xcite ( gluing along a clique ) , lovsz @xcite ( substitutions ) , cunningham @xcite ( 1-join ) , burlet and fonlupt @xcite ( amalgams ) , cornujols and cunningham @xcite ( proper 2-cutset ) . note that in @xcite , an operation more general than gluing along a proper 2-cutset is considered ( the so - called _
2-join _ , not worth defining here ) .
note also that with our definitions , it could be that for a graph @xmath1 with a proper 2-cutset , the blocks of decompositions are not perfect .
this happens for instance with the chordless cycle @xmath108 and the proper 2-cutset @xmath109 .
the blocks of decomposition are then both isomorphic to a cycle of length 5 , a notoriously non - perfect graph . but
the converse works smoothly : if two graphs are perfect , then a perfect graph is obtained by gluing them along a proper 2-cutset .
a proof of this is implicit in @xcite .
another simple way to check this is to note that when a vertex @xmath20 has degree 2 and non - adjacent neighbors @xmath89 in a perfect graph @xmath1 , then all paths from @xmath94 to @xmath95 in @xmath1 have even length ( otherwise , @xmath1 contains an odd chordless cycle of length at least 5 ) .
such a pair @xmath89 is what is called an _ even pair _ , and it is proved in @xcite that there exists an optimal coloring of @xmath1 such that @xmath94 and @xmath95 have the same color .
the perfection of a graph obtained from two perfect graphs by gluing @xmath39 and @xmath40 along a proper 2-cutset @xmath99 is then easy to prove by a direct coloring argument : use colorings of @xmath39 and @xmath40 that both give the same color to @xmath94 and @xmath95 .
[ th : pop ] perfect graphs are closed under the following operations : gluing along a clique , substitution , 1-join composition , amalgam composition , gluing along a proper 2-cutset .
we now turn our attention to the perservation of @xmath4-boundedness under the operations , but there is an important technicality .
a class of graphs is hereditary if it is closed under taking induced subgraphs .
the _ closure _ of a class @xmath110 of graphs under a set of graph operations is the class @xmath111 obtained from the graphs of @xmath110 by perfoming the operations repeatedly and in any order .
a set of operations _
preserves @xmath4-boundedness _ if the closure of any hereditary @xmath4-bounded class under the set of operations is a @xmath4-bounded class . of course ,
the function that bounds @xmath4 needs not be the same in @xmath110 and @xmath111 , and in most cases , it is not .
this leads to a potential problem : it may happen that an operation @xmath112 preserves @xmath4-boundedness , that another operation @xmath113 also preserves @xmath4-boundedness , but that the set of operations @xmath114 does not preserves @xmath4-boundedness .
this is explained in @xcite , where an actual ( but slightly artificial ) example of this phenomenon is provided .
it is very easy to prove that gluing along a clique preserves @xmath4-boundedness . in @xcite
, it is proved that substitution preserves @xmath4-boundedness . in @xcite
, it is proved that 1-join composition preserves @xmath4-boundedness .
but it is not at all easy to prove for instance that the set of operations \{1-join composition , gluing along a clique } preserves @xmath4-boundedness or that amalgam composition preserves @xmath4-boundedness .
however , these are true statements , and corollaries of the next theorem from @xcite .
[ th : penev ] if a class of graphs is @xmath4-bounded , then its closure under the following set of operations is @xmath4-bounded : \{substitution , amalgam composition , gluing along a clique}. also gluing along a proper 2-cutset preserves @xmath4-boundedness as shown in @xcite .
[ th : cpst ] if a class of graphs is @xmath4-bounded , then its closure under the operation of gluing along a proper 2-cutset is @xmath4-bounded .
note that in theorem [ th : penev ] , if the class we start with is @xmath4-bounded by a function @xmath8 , then the closure is @xmath4-bounded by an exponential in @xmath8 ( something close to @xmath115 ) . in theorem
[ th : cpst ] , the situation is much better , and the resulting function is linear into the function @xmath8 we start with .
the function was even improved by penev , thomass and trotignon , see @xcite . note that in @xcite an operation more general than the proper 2-cutset is considered ( namely , the operation of gluing along a 2-cutset , not worth defining here ) .
say that a graph @xmath1 has property @xmath116 if it has no edges ( such a graph is called an _ independent graph _ ) . we now define inductively a property @xmath117 for all @xmath118 as follows : a graph @xmath1 has property @xmath117 if for every induced subgraph @xmath51 of @xmath1 there exists an induced subgraph @xmath10 of @xmath51 that has property @xmath119 and that intersects every maximal clique of @xmath51 . from the definition , it is clear that property @xmath117 is _ hereditary _ : if a graph has it , then so are all its induced subgraphs .
graphs with property @xmath120 could be defined as graphs @xmath1 such that for every induced subgraph @xmath10 of @xmath1 there exists a stable set of @xmath10 that intersects every maximal clique of @xmath10 ( where a _
stable set _ in a graph is a set of vertices that induces an independent graph ) .
graphs satisfying property @xmath120 are known as _ strongly perfect graphs _ , see @xcite for a survey about them .
they form a ( proper ) subclass of perfect graphs . to the best of our knowledge , for @xmath121
, graphs with property @xmath117 were not studied so far .
the following provides examples of graphs with property @xmath117 .
[ l : amkpk ] for all @xmath118 , graphs with chromatic number at most @xmath122 have property @xmath119 .
we proceed by induction on @xmath122 . for @xmath123 ,
the result is obvious .
suppose it holds for some fixed @xmath118 .
let @xmath1 be a graph with chromatic number at most @xmath124 , and @xmath51 an induced subgraph of @xmath1 .
in @xmath51 , there exists an induced subgraph @xmath10 of chromatic number at most @xmath122 that intersects all maximal cliques of @xmath51 : consider for instance the union of the first @xmath122 ( possibly empty ) colour classes in a colouring of @xmath51 with @xmath124 colours . by the induction hypothesis , @xmath10 has property @xmath119 .
this proves that @xmath1 has property @xmath125 .
the following is similar to the classical lovsz s replication lemma , stating that perfect graphs are closed under substitutions .
[ l : pkcs ] for all @xmath126 , property @xmath117 is closed under substitution .
we proceed by induction on @xmath122 .
if @xmath127 , we have to prove that substituting an independent graph for a vertex @xmath20 of an independent graph yields an independent graph , which is obvious .
so , suppose @xmath118 and suppose property @xmath119 is closed under substitution .
suppose that @xmath1 is a graph obtained from @xmath39 by substituting @xmath40 for @xmath128 and @xmath39 and @xmath40 have property @xmath125 .
we will prove that @xmath1 contains an induced subgraph with property @xmath119 that intersect all maximal cliques of @xmath1 . by lemma [ l : hsubst ]
, the same proof can be done for induced subgraphs @xmath51 of @xmath1 .
a maximal clique in @xmath1 is either a maximal clique of @xmath39 that does not contain @xmath20 ( we say that such a maximal clique has
_ type non-@xmath20 _ ) , or is equal to @xmath129 where @xmath130 is a maximal clique of @xmath131 $ ] and @xmath132 is a maximal clique of @xmath40 ( we say that such a maximal clique has _
type @xmath20 _ ) . for @xmath61
, because @xmath62 has property @xmath125 , there exists an induced subgraph @xmath133 of @xmath62 that has property @xmath119 and that intersects every maximal clique of @xmath62 .
there are now two cases .
if @xmath134 , then let @xmath10 be the graph obtained from @xmath135 by substituting @xmath136 for @xmath20 . by the induction hypothesis , @xmath10 has property @xmath119 .
let @xmath30 be a maximal clique in @xmath1 .
if @xmath30 is of type @xmath20 , then @xmath137 is a maximal clique of @xmath40 , and it is intersected by @xmath138 , so it is intersected by @xmath10 . if @xmath30 is of type non-@xmath20 , then @xmath30 is a maximum clique of @xmath39 , so it must intersect @xmath135 , and not in @xmath20 , so it intersects @xmath10 .
we proved that @xmath10 intersects all maximal cliques of @xmath1 . if @xmath139 , then we set @xmath140 .
let @xmath30 be a maximal clique in @xmath1 .
if @xmath30 is of type @xmath20 , then @xmath141 is a maximal clique of @xmath39 , and it is intersected by @xmath142 . if @xmath30 is of type non-@xmath20 , then @xmath30 is a maximum clique of @xmath39 , so it must intersect @xmath142 .
we proved again that @xmath10 intersects all maximal cliques of @xmath1 . since substitution is one of the most simple operation that preserves perfection , it is worth asking whether property @xmath117 is closed under gluing along a clique ( another simple operation that preserves perfection and @xmath4-boundedness ) .
it turns out that it is not the case for @xmath123 ( examples are provided in @xcite ) .
here we give another example showing that @xmath143 is not closed under gluing along a clique .
to check this , it is convenient to rephrase property @xmath143 : for every induced subgraph @xmath51 , there is a strongly perfect graph @xmath10 that is an induced subgraph of @xmath51 and that intersects all maximal cliques of @xmath51 .
chordal graphs are shown to be strongly perfect in @xcite .
on figure [ fig : p2 ] , three graphs are represented .
graph @xmath39 is obtained from a copy of @xmath144 and a copy of @xmath145 by adding a matching . in @xmath40 , there are five copies of @xmath145 , say @xmath135 , , @xmath146 , and there are all possible edges between @xmath133 and @xmath147 for all @xmath148 ( taken modulo 5 ) .
four of the copies have a @xmath144 matched to them .
it is easy to check that @xmath39 and @xmath40 both have property @xmath143 ( in fact , they have the stronger property that a chordal graph intersects all maximal cliques ) . for instance , in @xmath39 , by picking a vertex in the @xmath145 and by taking all its non - neighbors , we obtain a chordal graph @xmath10 that intersects all maximal cliques of @xmath39 . in @xmath40
, we can take four copies of the chordal graph used for @xmath39 that exist in the non - matched @xmath145 s .
note that in @xmath40 , no vertex of the top clique needs to be taken in chordal graph that intersects all maximal cliques .
however , @xmath149 , that is obtained by gluing @xmath39 and @xmath40 along a @xmath145 , does not have property @xmath143 . to see this , note that every matching edge is a maximal clique .
also , @xmath10 can not contain a vertex in each of the @xmath145 s ( because this would form a @xmath144 , that is not strongly perfect ) , so at least one copy of @xmath145 does not intersect @xmath10 .
the @xmath144 matched to this copy therefore has to be all in @xmath10 , a contradiction to the strong perfection of @xmath10 .
[ cols="^,^,^ " , ] we now explain how property @xmath117 is related to @xmath4-boundedness .
we define @xmath150 and @xmath151 for all integers @xmath152 . for all integers @xmath118 and @xmath153 , we set @xmath154 by an easy induction , for all integer @xmath126
, @xmath155 , @xmath156 and @xmath157 is the identity function . also , it is easy to check that @xmath158 for all integers @xmath159 is a polynomial of degree @xmath122 ( with the convention that @xmath160 ) .
[ l : pkchib ]
graphs with property @xmath117 are @xmath4-bounded by the function @xmath161 ( and therefore by a polynomial of degree @xmath122 ) . for @xmath127 ,
this is trivial .
let us prove it by induction on @xmath122 for @xmath118 .
let @xmath1 be a graph that has property @xmath117 and set @xmath162 . by property @xmath117
, @xmath1 contains an induced subgraph @xmath163 that has property @xmath119 and intersects all maximal cliques of @xmath1 , so that @xmath164 . in @xmath165 , there exists also an induced subgraph @xmath166 that has property @xmath119 and intersects all maximal cliques of @xmath1 , and continuing like that , we prove that @xmath1 can be vertex - wise partitioned into @xmath6 induced subgraphs @xmath135 , @xmath136 ,
, @xmath167 , such that for all @xmath168 , @xmath169 has property @xmath119 and @xmath170 . by the induction hypothesis ,
we have @xmath171 the same proof can be made for all induced subgraphs of @xmath1 .
[ th : closk ] the closure by substitutions of the class of @xmath122-colourable graphs is a class of graph that is @xmath4-bounded by @xmath172 ( in particular , by a polynomial of degree @xmath173 ) .
every graph in the class has property @xmath119 , either by lemma [ l : amkpk ] or by lemma [ l : pkcs ] .
so , by lemma [ l : pkchib ] , it is @xmath4-bounded by @xmath172 .
as observed by penev , for large values of @xmath122 , a stronger result was implicitly proved in @xcite .
[ th : chispoly ] if a class of graphs is @xmath4-bounded by @xmath174 , then the closure of the class under substitution is @xmath4-bounded by @xmath175 .
since @xmath122-colourable graphs are @xmath4-bounded by @xmath176 , we know by theorem [ th : chispoly ] that the closure of @xmath122-colorable graphs under substitution forms a class @xmath4-bounded by @xmath177 .
so , when @xmath122 is large , @xmath178 is smaller than @xmath179 , but for small values , theorem [ th : closk ] provides the best bound known so far . for instance , the fact that the closure of 3-colorable graphs under substitution is @xmath4-bounded by a quadratic function is seemingly a new theroem .
in the rest of the paper , we adopt the unusual convention that _ no vertex of a graph is complete to the empty set_. we call _ a constraint for a graph @xmath1 _ any pair @xmath180 such that @xmath181 and @xmath182 are disjoint sets and @xmath183 is a clique of @xmath1 .
note that @xmath181 and @xmath182 are therefore disjoint possibly empty cliques of @xmath1 .
when @xmath180 is a constraint for a graph @xmath1 , a _
splitter for @xmath184 _ is an induced subgraph @xmath10 of @xmath1 that satisfies the following .
* @xmath10 intersects all maximal cliques of @xmath1 ( except possibly @xmath182 when @xmath182 is a maximal clique of @xmath1 ) ; * @xmath10 contains all vertices of @xmath181 and @xmath10 contains no @xmath181-complete vertex ( when @xmath185 , this constraint can be forgotten ) ; * @xmath10 contains no vertex of @xmath182 .
we now define inductively a property @xmath186 for all @xmath187 .
a graph has _
property @xmath188 _ if it is perfect . for @xmath118 , a graph has _ property @xmath189 _ if for every induced subgraph @xmath51 and every constraint @xmath180 for @xmath51
, there is splitter for @xmath190 with the additionnal property that @xmath191 $ ] has property @xmath192 .
such a splitter is called a _
@xmath122-splitter_. it is obvious that if a graph @xmath1 has property @xmath186 then every induced subgraph of @xmath1 has property @xmath186 . on figure
[ f : noqk ] , we show a graph @xmath1 that does not have property @xmath186 for any @xmath122 . to see this ,
suppose for a contradiction that @xmath1 has property @xmath186 for some @xmath118 , and consider the minimum such @xmath122 .
since @xmath1 contains a @xmath144 and is therefore not perfect , we have @xmath121 .
define @xmath182 as the set of black vertices on the figure .
it is straightforward that the only splitter for @xmath193 is @xmath194 , so @xmath195 = g$ ] must have property @xmath196 , a contradiction to the minimality of @xmath122 .
the next lemma gives the taste of our main theorem on property @xmath186 ( and is a particular case of it , but we prefer proving it separately ) .
it is very easy , but it seems impossible to prove it formally without an induction .
[ l : du ] property @xmath186 is closed under disjoint union .
we prove the lemma by induction on @xmath122 .
if @xmath123 , the result follows directly from theorem [ th : pop ] ( because taking the disjoint union means gluing along an empty clique ) .
so , suppose @xmath118 and let @xmath1 be the disjoint union of two graphs @xmath39 and @xmath40 that have property @xmath189 .
let @xmath180 be a constraint for @xmath1 .
up to symmetry , we may assume that @xmath197 , and consider a @xmath122-splitter @xmath135 for @xmath198 . in @xmath40 , we consider a @xmath122-splitter for @xmath199 .
it straightforward to check that @xmath200 is a splitter for @xmath184 , and by the induction hypothesis , it is a @xmath122-splitter .
[ l : trivialsplitter ] for every graph @xmath1 and every constraint @xmath180 for @xmath1 there exist a splitter for @xmath184 .
define @xmath10 as the graph induced by all vertices of @xmath201 that are not complete to @xmath181 ( in particular , if @xmath202 then @xmath194 ) .
we claim that @xmath10 is a splitter for @xmath184 .
it intersects all maximal cliques of @xmath1 ( except possibly @xmath182 when @xmath182 is a maximal clique ) because any vertex in @xmath181 is complete to @xmath203 ( and when @xmath202 and @xmath182 is not maximal , there must be a vertex complete to @xmath182 in @xmath10 ) .
obviously , @xmath10 contains all vertices of @xmath181 , no @xmath181-complete vertex , and no vertex of @xmath182 .
[ l : monotone ] for all @xmath118 , any graph with property @xmath186 has property @xmath189 .
let @xmath180 be a constraint for a graph @xmath1 with property @xmath186 .
lemma [ l : trivialsplitter ] provides a splitter for @xmath184 .
this splitter has property @xmath192 because so does @xmath1 .
the same proof can be done for all induced subgraphs of @xmath1 , so every induced subgraph of @xmath1 has a @xmath122-splitter .
[ l : smallsplitter ] let @xmath180 be a constraint for a non - bipartite triangle - free connected graph @xmath1 .
then there exists a splitter @xmath10 for @xmath184 such that @xmath204 . since @xmath1 is triangle - free , @xmath205 . also , since @xmath1 is non - bipartite and triangle - free , @xmath206 .
we claim that @xmath1 contains a vertex @xmath20 such that : * @xmath207 ; * @xmath20 has no neighbor in @xmath182 ; * @xmath20 is not adjacent to any @xmath181-complete vertex . to prove the claim , we break into cases according to the sizes of @xmath181 and @xmath182 . if @xmath208 and @xmath209 , then any vertex @xmath20 satisfies the constraint . if @xmath208 and @xmath210 , then any vertex not in @xmath182 and non - adjacent to the unique vertex in @xmath182 satisfies the constraint ( and there is such a vertex , for otherwise @xmath1 is bipartite ) . if @xmath211 and @xmath209 , then any @xmath181-complete vertex @xmath20 satisfies the constraint ( and there is such a vertex since @xmath1 is connected ) .
if @xmath211 and @xmath210 , then any @xmath181-complete vertex not in @xmath182 satisfies the contraints . since @xmath1 is connected , we may therefore assume that @xmath181 is made of a vertex @xmath21 whose only neighbor is the vertex @xmath212 from @xmath182 .
if all vertices of @xmath201 are adjacent to @xmath212 , then since @xmath1 is triangle free , @xmath213 is a stable set , and @xmath1 is bipartite , a contradiction .
it follows that @xmath20 can be choosen among the non - neighbors of @xmath212 . if @xmath214 and @xmath209 , then no vertex in @xmath1 is @xmath181-complete since @xmath1 is triangle - free , so any vertex @xmath20 not in @xmath181 satisfies the constraint ( and there exists such a vertex since @xmath1 is not bipartite ) .
if @xmath208 and @xmath215 , then any vertex not in @xmath182 and with no neighbor in @xmath182 satisfies the constraint , so suppose that no such vertex exists .
it follows that @xmath216 where @xmath217 .
since @xmath1 is triangle - free , we see that @xmath218 and @xmath219 are stable sets , so @xmath1 is bipartite , a contradiction .
this proves the claim .
now define @xmath45 as the set of @xmath181-complete vertices . since @xmath1 is triangle - free ,
we see that @xmath220 is a stable set of @xmath1 ( except when @xmath215 ) .
it follows that @xmath221 is a splitter for @xmath184 . and because of @xmath20 , we have @xmath204 .
[ l : smallqk ] every triangle - free graph on @xmath222 vertices has property @xmath223 .
we prove the property by induction on @xmath224 .
it is well known that all graphs on at most 4 vertices are perfect ( they are all chordal except the cycle of length 4 that is bipartite ) .
so by lemma [ l : monotone ] , the property is true for @xmath225 .
suppose it holds for @xmath222 , and consider a graph @xmath1 on @xmath226 vertices and a constraint @xmath180 for @xmath1 .
by lemma [ l : monotone ] , we may assume that @xmath1 is not perfect ( and in particular not bipartite ) .
it is enough to find an @xmath227-splitter for @xmath184 , because the proper induced subgraphs of @xmath1 have property @xmath223 by the induction hypothesis .
lemma [ l : smallsplitter ] provides a splitter @xmath10 such that @xmath228 , so the induction hypothesis shows that this splitter has property @xmath223 .
[ pr : q_k ] property @xmath186 is closed under the following set of operations : \{gluing along a clique , amalgam ( and therefore 1-join composition ) , substitution and proper 2-cutset composition}. we proceed by induction on @xmath122 . for @xmath123 ,
the result follows directly from theorem [ th : pop ] .
suppose it holds for some fixed @xmath187 .
we consider a graph @xmath1 obtained from two graphs @xmath39 and @xmath40 with property @xmath189 by one of the operations , and we show that for every constraint @xmath180 , there exists a @xmath122-splitter for @xmath184 .
each time , the splitter is obtained by combining @xmath122-splitters of @xmath39 and @xmath40 with well chosen constraints ( they exist by assumption ) , and the combination has property @xmath186 by the induction hypothesis . note that by lemma [ l : ls ] , the same proof can be done for induced subgraphs of @xmath1 , so we do not need to consider induced subgraphs of @xmath1 .
let us consider the operations one by one .
we suppose that @xmath31 is a split for a clique cutset of @xmath1 , so @xmath1 is obtained from @xmath229 $ ] and @xmath230 $ ] by gluing along @xmath30 . we suppose that @xmath39 and @xmath40 have property @xmath189 . up to symmetry
, we may assume that @xmath231 .
set @xmath232 and let @xmath135 be a splitter for @xmath233 .
there are two cases .
* case 1 * : @xmath234 .
set @xmath235 and @xmath236 .
note that @xmath237 is not a maximal clique of @xmath40 since @xmath238 .
let @xmath136 be a splitter for @xmath239 .
let @xmath240 .
since @xmath136 contains no @xmath241-complete vertex , we have @xmath242 .
we now check that @xmath10 is a @xmath122-splitter for @xmath184 .
first , @xmath10 contains all vertices of @xmath181 and @xmath10 contains no vertex of @xmath182 .
vertices of @xmath135 are not complete to @xmath243 .
also a vertex @xmath244 is not complete to @xmath181 , for otherwise , @xmath245 ( because as noted already @xmath242 ) .
it follows that @xmath246 .
but then , @xmath247 , so @xmath20 is @xmath241-complete , a contradiction .
hence , @xmath10 contains no @xmath181-complete vertex . finally , @xmath10
intersects all maximal cliques of @xmath1 ( except @xmath182 when @xmath182 is a maximal clique of @xmath1 and therefore of @xmath39 ) , because all such cliques are either in @xmath39 or in @xmath40 . since @xmath248 $ ] and @xmath249 $ ] have property @xmath186 and @xmath191 $ ] is obtained from these two graphs by gluing along @xmath250 , we know by the induction hypothesis that @xmath191 $ ] has property @xmath186 .
* case 2 * : @xmath251 .
note that @xmath252 .
set @xmath253 and @xmath254 and let @xmath136 be a splitter for @xmath239 .
let @xmath240 .
so @xmath10 contains all vertices of @xmath181 , no vertex of @xmath182 and no @xmath181-complete vertex .
observe that @xmath191 $ ] is obtained from @xmath248 $ ] and @xmath255 $ ] by gluing along @xmath256 . and @xmath255 $ ] has property @xmath186 because it is an induced subgraph of @xmath249 $ ]
. so by the induction hypothesis , @xmath191 $ ] has property @xmath186 .
it remains to prove that @xmath10 intersects all maximal cliques of @xmath1 ( except @xmath182 when @xmath182 is a maximal clique of @xmath1 ) .
let @xmath257 be any maximal clique of @xmath1 . since @xmath30 is clique cutset of @xmath1
, @xmath257 is a maximal clique of @xmath39 or @xmath40 .
if @xmath258 , then @xmath257 is intersected by @xmath135 ( and therefore @xmath10 ) unless @xmath259 .
if @xmath260 , then @xmath257 is intersected by @xmath136 , unless @xmath261 .
in this last case , @xmath30 is a maximal clique of @xmath1 ( and therefore @xmath39 ) , and since it is not intersected by @xmath135 ( because @xmath262 ) , it must be that @xmath259 . in all cases , @xmath10 intersects @xmath257 except when @xmath263 .
we suppose that @xmath264 is a split for an amalgam of @xmath1 . for @xmath61 , the _ block of decomposition _ @xmath62 with respect to this split is the graph obtained from @xmath77 $ ] by adding a vertex @xmath64 complete to @xmath78 , so @xmath1 is obtained from @xmath39 and @xmath40 by an amalgam composition . we suppose that @xmath39 and @xmath40 have property @xmath189 .
let @xmath265 be a constraint for @xmath1 .
* case 1 * @xmath266 does not contain @xmath183 and @xmath267 does not contain @xmath183 . since @xmath183 is a clique , @xmath183 belongs to @xmath268 , @xmath269 , @xmath270 .
there are three subcases .
* case 1a * : @xmath271 .
set @xmath272 and @xmath273 .
note that @xmath274 is not a maximal clique of @xmath39 since @xmath271 .
let @xmath135 be a splitter for @xmath233 .
set @xmath275 and @xmath276 .
note that @xmath237 is not a maximal clique of @xmath40 since @xmath271 .
let @xmath136 be a splitter for @xmath239 .
let @xmath240 .
we now check that @xmath10 is a @xmath122-splitter for @xmath184 .
it is obvious that @xmath10 contains all vertices of @xmath181 . if a vertex @xmath277 is complete to @xmath181 , then @xmath20 is complete to @xmath278 , a contradiction .
this implies that @xmath135 contains no @xmath181-complete vertex , no vertex of @xmath182 .
similarly , @xmath136 contains no @xmath181-complete vertex , no vertex of @xmath182 .
hence , @xmath10 contains no @xmath181-complete vertex , no vertex of @xmath182 .
@xmath10 contains @xmath181 and @xmath271 so @xmath10 intersects all maximal cliques of @xmath268 .
therefore , @xmath10 intersects all maximal cliques of @xmath1 .
@xmath191 $ ] is obtained from two graphs @xmath279 $ ] and @xmath280 $ ] ( both graphs have property @xmath186 ) by an amalgam composition , so by the induction hypothesis , @xmath191 $ ] has property @xmath186 . * case 1b * : @xmath281 , @xmath282 . up to symmetry , we may assume that @xmath283 .
set @xmath284 and @xmath273 .
note that @xmath274 is not a maximal clique of @xmath39 since @xmath283 .
let @xmath135 be a splitter for @xmath233 .
set @xmath285 and @xmath286 .
note that @xmath237 is not a maximal clique of @xmath40 since @xmath287 .
let @xmath136 be a splitter for @xmath239 .
let @xmath288 $ ] .
we now check that @xmath10 is a @xmath122-splitter for @xmath184 .
similar to previous case , we have @xmath10 contains all vertices of @xmath181 , no @xmath181-complete vertex , no vertex of @xmath182 .
since @xmath135 intersects all maximal cliques of @xmath39 then @xmath289 and @xmath120 must intersect all maximal cliques of @xmath58 ( if not this clique combined with @xmath66 and @xmath30 would be a maximal clique of @xmath39 that @xmath135 does not intersect , a contradiction ) .
this implies that @xmath120 intersects all maximal cliques of @xmath268 .
hence , @xmath10 intersects all maximal cliques of @xmath1 . because @xmath270 , @xmath290
hence , @xmath191 $ ] is obtained from two graphs @xmath279 $ ] and @xmath280 $ ] ( both graphs have property @xmath186 ) by an amalgam composition , so by the induction hypothesis , @xmath191 $ ] has property @xmath186 . * case 1c * : @xmath185 . if @xmath291 , then , we choose @xmath292 , set @xmath293 and by the proof of case @xmath294 and case @xmath295 we obtain a @xmath122-splitter @xmath10 for @xmath184 .
so , suppose @xmath296 .
this means @xmath182 is a maximal clique of @xmath1 .
set @xmath297 and @xmath273 .
let @xmath135 be a splitter for @xmath233 .
set @xmath298 and @xmath276 .
let @xmath136 be a splitter for @xmath239 .
let @xmath240 .
we now check that @xmath10 is a @xmath122-splitter for @xmath184 .
it is obvious that @xmath10 contains no vertex of @xmath182 and @xmath10 intersects all maximal cliques of @xmath1 except @xmath299 .
@xmath191 $ ] is obtained from two graphs @xmath300 $ ] and @xmath280 $ ] ( both graphs have property @xmath186 ) by an amalgam composition , so by the induction hypothesis , @xmath191 $ ] has property @xmath186 .
* case 2 * we are not in case 1 , so up to symmetry , we may assume that @xmath231 .
set @xmath232 and let @xmath135 be a splitter for @xmath233 .
there are three cases . *
case 2a * : @xmath234 .
set @xmath235 and @xmath236 .
note that @xmath237 is not a maximal clique of @xmath40 since @xmath238 .
let @xmath136 be a splitter for @xmath239 .
since @xmath136 contains no @xmath241-complete vertex , we have @xmath242 , @xmath301 and @xmath302 .
let @xmath303 $ ] .
we now check that @xmath10 is a @xmath122-splitter for @xmath184 .
it is obvious that @xmath10 contains all vertices of @xmath181 and @xmath10 contains no vertex of @xmath182 .
vertices of @xmath135 are not complete to @xmath243 .
also a vertex @xmath244 is not complete to @xmath181 , for otherwise , @xmath245 ( because as noted already @xmath242 ) .
it follows that @xmath246 .
but then , @xmath304 since @xmath135 contains no @xmath278-complete vertex .
this implies @xmath305 , so @xmath20 is @xmath241-complete , a contradiction .
hence , @xmath10 contains no @xmath181-complete vertex .
@xmath10 contains @xmath234 so @xmath10 intersects all maximal cliques of @xmath268 .
hence , @xmath10 intersects all maximal cliques of @xmath1 .
observe that @xmath191 $ ] is obtained from @xmath306 $ ] and @xmath307 $ ] by gluing along @xmath308 . and @xmath306 $ ] has property @xmath186 because it is an induced subgraph of @xmath309 $ ] .
so by the induction hypothesis , @xmath191 $ ] has property @xmath186 . *
case 2b * : @xmath251 and @xmath310 .
@xmath135 contains all vertices of @xmath181 so @xmath311 . set @xmath312 and @xmath236 .
note that @xmath237 is not a maximal clique of @xmath40 since @xmath313 .
let @xmath136 be a splitter for @xmath239 . since @xmath136 contains no @xmath241-complete vertex , we have @xmath314 , @xmath301 .
let @xmath315 $ ] .
we now check that @xmath10 is a @xmath122-splitter for @xmath184 .
it is obvious that @xmath10 contains all vertices of @xmath181 and @xmath10 contains no vertex of @xmath182 .
@xmath10 contains no @xmath181-complete vertex since vertices of @xmath135 are not complete to @xmath243 and @xmath138 is anticomplete to @xmath181 ( because as noted already @xmath314 , @xmath301 and @xmath311 ) .
since @xmath135 intersects all maximal cliques of @xmath39 then @xmath289 and @xmath120 must intersect all maximal cliques of @xmath58 ( if not this clique combined with @xmath66 and @xmath30 would be a maximal clique of @xmath39 that @xmath135 does not intersect , a contradiction ) .
this implies that @xmath120 intersects all maximal cliques of @xmath268 .
hence , @xmath10 intersects all maximal cliques of @xmath1 .
observe that @xmath191 $ ] is obtained from @xmath316 $ ] and @xmath317 $ ] by gluing along @xmath318 ( possibly empty ) . and @xmath317 $ ] has property @xmath186 because it is an induced subgraph of @xmath249 $ ] .
so by the induction hypothesis , @xmath191 $ ] has property @xmath186 .
* case 2c * @xmath251 and @xmath319 .
@xmath135 contains all vertices of @xmath181 so @xmath311 .
also @xmath320 , for otherwise @xmath319 is complete to @xmath181 .
choose any vertex @xmath321 , set @xmath322 and @xmath323 .
note that @xmath237 is not a maximal clique of @xmath40 since @xmath313 .
let @xmath136 be a splitter for @xmath239 .
since @xmath136 contains no @xmath241-complete vertex , we have @xmath314 .
let @xmath324 $ ] .
we now check that @xmath10 is a @xmath122-splitter for @xmath184 .
it is obvious that @xmath10 contains all vertices of @xmath181 and @xmath10 contains no vertex of @xmath182 .
vertices of @xmath135 are not complete to @xmath243 .
also a vertex @xmath325 is not complete to @xmath181 , for otherwise , @xmath326 since @xmath311 .
but then @xmath20 is adjacent to a vertex of @xmath327 ( because as noted already @xmath320 ) , a contradiction .
hence , @xmath10 contains no @xmath181-complete vertex .
since @xmath136 intersects all maximal cliques of @xmath40 then @xmath328 and @xmath143 must intersect all maximal cliques of @xmath59 ( if not this clique combined with @xmath67 and @xmath30 would be a maximal clique of @xmath40 that @xmath136 does not intersect , a contradiction ) .
this implies that @xmath143 intersects all maximal cliques of @xmath268 .
hence , @xmath10 intersects all maximal cliques of @xmath1 .
consider @xmath329 , @xmath191 $ ] is obtained from two graphs @xmath330 $ ] and @xmath255 $ ] ( both graphs have property @xmath186 ) by gluing along a clique @xmath256 ( possibly empty ) , so by the induction hypothesis , @xmath191 $ ] has property @xmath186 . in the case
@xmath331 , @xmath191 $ ] is obtained from two graphs @xmath332 $ ] and @xmath249 $ ] ( both graphs have property @xmath186 ) by an amalgam composition , so by the induction hypothesis , @xmath191 $ ] has property @xmath186 .
we are done when @xmath231 .
hence , @xmath1 has property @xmath189 .
since the amalgam is already treated , by lemma [ l : easys ] , it is enough to prove that property @xmath186 is closed under adding a universal vertex .
let @xmath1 be a graph obtained from a graph @xmath51 by adding a universal vertex @xmath20 .
we suppose that @xmath51 has property @xmath189 .
let @xmath265 be a constraint for @xmath1 .
there are two cases .
* case 1 * : @xmath333 .
let @xmath334 $ ] .
so @xmath10 contains all vertices of @xmath181 , no vertex of @xmath182 and no @xmath181-complete vertex .
every maximal clique of @xmath1 contains @xmath20 since @xmath20 is a universal vertex , this means @xmath10 intersects all maximal cliques of @xmath1 .
@xmath191 $ ] is clique then it has property @xmath188 .
therefore , @xmath191 $ ] has property @xmath186 .
hence , @xmath10 is a @xmath122-splitter for @xmath184 .
* case 2 * : @xmath335 .
let @xmath10 be a splitter for @xmath336 .
we have @xmath337 so in @xmath1 , @xmath10 contains all vertices of @xmath181 , no vertex of @xmath182 , no @xmath181-complete vertex .
assume @xmath10 does not intersect a maximal clique @xmath338 ( @xmath339 ) of @xmath1 .
it follows that @xmath10 does not intersect a maximal clique @xmath340 of @xmath51 since @xmath20 is a universal vertex of @xmath1 , a contradiction .
hence , @xmath10 intersects all maximal cliques of @xmath1 ( except @xmath182 when @xmath182 is a maximal clique ) . if @xmath341 then @xmath342 $ ] has property @xmath186 since @xmath342= g'[v(h ) \cup k^{+}]$ ] .
if @xmath343 then @xmath342 $ ] is obtained from graph @xmath344 $ ] by adding a universal vertex @xmath20 .
so by the induction hypothesis , @xmath342 $ ] has property @xmath186 .
hence , @xmath1 has property @xmath189 .
we suppose that @xmath96 is a split for a proper 2-cutset . for @xmath61 ,
the block of decomposition @xmath62 with respect to this split is the graph obtained from @xmath97 $ ] by adding a vertex @xmath98 complete to @xmath99 , so @xmath1 is obtained from @xmath39 and @xmath40 by a proper 2-cutset composition .
we suppose that @xmath39 and @xmath40 have property @xmath189 .
let @xmath265 be a constraint for @xmath1 .
up to symmetry , we may assume that @xmath345 .
set @xmath232 and let @xmath135 be a splitter for @xmath233 .
there are two cases .
* case 1 * : @xmath346 . because @xmath182 is a clique
, we have @xmath347 . without loss of generality
, we can assume @xmath348 .
note that @xmath349 since @xmath350 is clique of @xmath1 .
it follows @xmath252 .
the set @xmath135 intersects maximal clique @xmath351 and @xmath352 since @xmath135 does not contain vertices of @xmath353 .
it follows @xmath354 . * if @xmath355 then set @xmath356 and let @xmath136 be a splitter for @xmath239 . + let @xmath357 $ ] so @xmath10 intersects all maximal cliques of @xmath1 ( except possibly @xmath182 when @xmath182 is a maximal clique ) because all such cliques are either in @xmath39 or in @xmath40 .
+ @xmath10 contains all vertices of @xmath181 and no vertex of @xmath182 .
also @xmath95 is not complete to @xmath181 since @xmath358 and @xmath359 is anticomplete to @xmath181 .
this implies that @xmath10 contains no @xmath181-complete vertex .
+ @xmath191 $ ] is obtained from two graphs @xmath360 $ ] and @xmath307 $ ] ( both graphs have property @xmath186 ) by a proper 2-cutset composition so by the induction hypothesis , @xmath191 $ ] has property @xmath186 . * if @xmath361 then set @xmath362 and let @xmath136 be a splitter for @xmath239 . + let @xmath363 $ ] so @xmath10 intersects all maximal cliques of @xmath1 ( except possibly @xmath182 when @xmath182 is a maximal clique ) because all such cliques are either in @xmath39 or in @xmath40 .
+ the graph @xmath10 contains all vertices of @xmath181 and no vertex of @xmath182 .
also @xmath136 contains no @xmath241-complete vertex so @xmath136 does not contains @xmath94 or @xmath95 .
it follows @xmath136 is anticomplete to @xmath181 .
hence , @xmath10 contains no @xmath181-complete vertex .
+ @xmath191 $ ] is obtained from two graphs @xmath364 $ ] and @xmath365 $ ] ( both graphs have property @xmath186 ) by a gluing at clique @xmath366 , so by the induction hypothesis , @xmath191 $ ] has property @xmath186 . * case 2 * : @xmath367 . * if @xmath368 then @xmath181 does not contains @xmath94 or @xmath95 . set @xmath369 and let @xmath136 be a splitter for @xmath239 . + let @xmath363 $ ] so @xmath10 intersects all maximal cliques of @xmath1 ( except possibly @xmath182 when @xmath182 is a maximal clique ) because all such cliques are either in @xmath39 or in @xmath40 . + the graph @xmath10 contains all vertices of @xmath181 and no vertex of @xmath182 .
also @xmath136 contains no @xmath241-complete vertex so @xmath136 does not contains @xmath94 or @xmath95 .
it follows @xmath136 is anticomplete to @xmath181 .
hence , @xmath10 contains no @xmath181-complete vertex .
+ the graph @xmath191 $ ] is obtained from two disjoint graphs @xmath370 $ ] and @xmath371 $ ] ( both graphs have property @xmath186 ) so by the induction hypothesis , @xmath191 $ ] has property @xmath186 . * if @xmath372 and @xmath373 then set @xmath374 .
if @xmath375 then set @xmath376 and @xmath181 is a clique so up to symmetry , we may assume that @xmath374 .
+ set @xmath377 and let @xmath136 be a splitter for @xmath239 . because @xmath136 does not contain @xmath366-complete vertices , @xmath378 , so @xmath136 contains @xmath95 since @xmath136 intersects all maximal cliques .
+ let @xmath379 $ ] so @xmath10 intersects all maximal cliques of @xmath1 ( except possibly @xmath182 when @xmath182 is a maximal clique ) because all such cliques are either in @xmath39 or in @xmath40 .
+ the graph @xmath10 contains all vertices of @xmath181 and no vertex of @xmath182 .
also @xmath136 contains no @xmath181-complete vertex since @xmath136 does not contain @xmath366-complete vertex .
hence , @xmath10 contains no @xmath181-complete vertex .
+ the graph @xmath191 $ ] is obtained from graphs @xmath380 $ ] and @xmath381 $ ] ( both graphs have property @xmath186 ) by gluing along a clique @xmath382 , so by the induction hypothesis , @xmath191 $ ] has property @xmath186 * if @xmath383 and @xmath384 then without loss of generality , we can assume that @xmath385 . because @xmath135 intersects the maximal clique @xmath386 , @xmath354 .
hence , @xmath387 ( since if @xmath388 , then @xmath389 , a contradiction ) .
+ set @xmath390 , @xmath391 and let @xmath136 be a splitter for @xmath239 . because @xmath136 does not contain @xmath392-complete vertex , @xmath136 does not contain @xmath94 or @xmath95 .
+ let @xmath393 $ ] so @xmath10 intersects all maximal cliques of @xmath1 ( except possibly @xmath182 when @xmath182 is a maximal clique ) because all such cliques are either in @xmath39 or in @xmath40 .
+ the graph @xmath10 contains all vertices of @xmath181 and no vertex of @xmath182 .
also @xmath10 contains no @xmath181-complete vertex since @xmath387 .
+ the graph @xmath191 $ ] is obtained from two graphs @xmath380 $ ] and @xmath394 $ ] ( both graphs have property @xmath186 ) by gluing at clique @xmath366 , so by the induction hypothesis , @xmath191 $ ] has property @xmath186 . recall that the function @xmath161 was defined in the previous section .
[ l : qkchib ] for all @xmath118 , graphs with property @xmath186 are @xmath4-bounded by the function @xmath161 .
the proof is similar to the proof of lemma [ l : pkchib ] . in the induction step ,
when we consider a graph with property @xmath189 , we use the constraint @xmath395 to find an induced subgraph with property @xmath192 that intersects all maximal cliques .
we can now prove the following theorem that is a seemingly new and non - trivial result .
note that @xmath4-boundedness of the class under consideration can easilly be obtained by theorem [ th : penev ] , but this approach would only provide an exponential function , while we provide a polynomial .
[ th : clossmall ] the closure by the set of operations @xmath105gluing along a clique , substitutions , 1-join composition , amalgam compositions , gluing along a proper 2-cutset@xmath106 of the class of graphs of triangle - free graphs of order at most @xmath396 is a class of graph that is @xmath4-bounded by @xmath397 ( in particular , by a polynomial of degree @xmath122 ) .
every graph in the class has property @xmath186 , either by lemma [ l : smallqk ] or by theorem [ pr : q_k ] .
so , by lemma [
l : qkchib ] , it is @xmath4-bounded by @xmath161 .
recall that a _ long - unichord _ in a graph is an edge that is the unique chord of some cycle of length at least 5 .
a graph is _ long - unichord - free _ if it does not contain any long - unichord . in this section ,
we prove a decomposition theorem for long - unichord - free graphs .
we obtain its proof somehow for free , by combining two known theorems .
the theorem below is proven in @xcite .
the original statement is slightly more precise , but this one is enough for our purpose . a _
cap _ in a graph is a cycle of length at least 5 , with a unique chord @xmath398 such that @xmath94 and @xmath95 are at distance two along the cycle .
[ th : cf ] if @xmath1 is a connected cap - free graph , then either : * @xmath1 is chordal ; * @xmath1 is triangle - free ; * @xmath1 has a universal vertex ; * @xmath1 has a cutvertex ; * @xmath1 has an amalgam .
here is useful corallary .
[ th : weakdec ] if @xmath1 is long - unichord - free , then either : * @xmath1 is chordal ; * @xmath1 is unichord - free ; * @xmath1 has a universal vertex ; * @xmath1 has a cutvertex ; * @xmath1 has an amalgam .
since a cap has a long unichord , @xmath1 is cap - free .
if @xmath1 contains a triangle , one of the outcome follows from theorem [ th : cf ] . and
if @xmath1 is triangle - free , then every unichord of @xmath1 is a long - unichord .
so @xmath1 is unichord - free .
the petersen and the heawood graphs are the graphs represented on figure [ fig : ph ] .
the following theorem is proved in @xcite .
[ th : ucf ] if @xmath1 is a connected unichord - free graph , then either : * @xmath1 is a clique ; * @xmath1 is an induced subgraph of the petersen graph ; * @xmath1 is an induced subgraph of the heawood graph ; * @xmath1 is bipartite and one side of the bipartition is made of vertices of degree at most 2 ; * @xmath1 has a cutvertex ; * @xmath1 has a 1-join ( and therefore an amalgam ) ; * @xmath1 has a proper 2-cutset .
our main decomposition theorem is the following .
[ th : maindec ] let @xmath1 be a connected long - unichord - free graph .
then either : * @xmath1 is an induced subgraph of the petersen graph ; * @xmath1 is an induced subgraph of the heawood graph ; * @xmath1 is chordal ; * @xmath1 is bipartite and one side of the bipartition is made of vertices of degree at most 2 ; * @xmath1 has a universal vertex ; * @xmath1 has a cutvertex ; * @xmath1 has an amalgam ; * @xmath1 has proper 2-cutset ; we apply theorem [ th : weakdec ] .
so either @xmath1 satisfies one of the outcomes , or @xmath1 is unichord free . in this last case
, the result follows from theorem [ th : ucf ] .
our main purpose is to prove that all long - unichord - free graphs are @xmath4-bounded .
this is a direct consequence of theorems [ th : weakdec ] , [ th : penev ] and the fact proved in @xcite that unichord - free graphs are @xmath4-bounded . but this approach would only provide an exponential bound ( because of theorem [ th : penev ] ) . here , by using property @xmath186 , we prove that the class is @xmath4-bounded by a polynomial . by theorem [ pr : q_k ] and
[ th : maindec ] , to prove that long - unichord - free graphs have property @xmath186 , it is enough to prove that the basic graphs from theorem [ th : maindec ] have property @xmath186 .
it turns out that most of these basic graphs are perfect : chordal graphs and bipartite graphs are perfect , and the heawood graph is bipartite .
so , all these graphs have property @xmath188 , and property @xmath186 for all @xmath118 by lemma [ l : monotone ] .
so , the only problem is the petersen graph , but it has property @xmath399 by lemma [ l : smallqk ] .
hence , we have a short proof that long - unichord - free graphs have all property @xmath399 .
we now prove several lemmas needed to show that in fact , long - unichord - free graphs have property @xmath400 . the only problem is to handle the petersen graph .
we rely on the labeling of the petersen graph represented on figure [ fig : p6 ] .
we first observe that @xmath400 is best possible : by setting @xmath401 and @xmath402 , it can be checked that the petersen graph does not have property @xmath403 .
indeed , the splitter @xmath10 can not contain @xmath212 and @xmath404 that are @xmath181-complete , so it would have to contain @xmath405 , @xmath406 and also @xmath407 and @xmath408 , so that @xmath409 contains a @xmath144 . ]
[ l : spq2 ] the graph induced by the petersen graph on @xmath410 has property @xmath403 all the odd cycles of the graph under consideration go through @xmath407 , @xmath21 and @xmath408 , so that whatever @xmath181 and @xmath182 , one of these three vertices is neither in @xmath182 nor in the splitter obtained by taking all other vertices .
it follows that a splitter @xmath10 such that @xmath411 induces a perfect graph exists .
[ l : pq3 ] the petersen graph has property @xmath400 .
we rely on the representation of the petersen graph @xmath1 given on figure [ fig : p6 ] .
it is well known that all pairs of adjacent vertices in the petersen graph are equivalent , that is for @xmath412 , there is an automorphism @xmath413 of @xmath1 such that @xmath414 and @xmath415 .
this property is refered to as _ the symmetry of @xmath1_. let @xmath51 be an induced subgraph of @xmath1 , and @xmath416 be a constraint for @xmath51 .
if @xmath417 ( so @xmath209 ) , then because of the symmetry of @xmath1 , we may assume that @xmath418 and @xmath419 induces a 1-splitter for @xmath190 .
if @xmath211 , then we may assume because of the symmetry of @xmath1 that @xmath420 and @xmath421 or @xmath422 . in both cases , @xmath423 induces a 2-splitter for @xmath190 .
note that @xmath424 $ ] has property @xmath403 by lemma [ l : spq2 ] .
we may therefore assume that @xmath185 .
because of the symmetry of @xmath1 , we may assume that @xmath425 .
we now observe as above that @xmath426 induces a 1-splitter for @xmath190 ( it has property @xmath403 as above ) .
[ l : clucfreeq3 ] every long - unichord - free has property @xmath400 .
as noted at the begining of the section , all basic graphs in theorem [ th : maindec ] are perfect ( and have therefore property @xmath188 , and property @xmath400 by lemma [ l : monotone ] ) .
also the petersen graph has property @xmath400 by lemma [ l : pq3 ] .
the result now follows from theorems [ th : maindec ] and [ pr : q_k ] .
[ the : cluc ] long - unichord - free graphs are @xmath4-bounded by @xmath427 ( in particular , by a polynomial of degree 3 )
. follows directly from lemmas [ l : clucfreeq3 ] and [ l : qkchib ] .
in this section , we describe a polytime algorithm that decides whether a graph contains a long - unichord .
the next lemma is straightforward to check ( while a formal proof would be very long ) , so we prefer letting it without proof . 1 . @xmath428 .
2 . @xmath429 .
3 . @xmath430 where @xmath431 and @xmath432 .
4 . @xmath433 where @xmath187 , @xmath434 , @xmath431 , @xmath435 .
@xmath436 where @xmath187 , @xmath437 , @xmath438 , @xmath439
@xmath440 where @xmath441 , @xmath442 , @xmath443
@xmath440 where @xmath441 , @xmath442 , @xmath444
@xmath445 where @xmath446 , @xmath447 , @xmath448
@xmath445 where @xmath446 , @xmath447 , @xmath449 .
if the decomposition under consideration is a cutvertex , the result is clear .
so , suppose it is an amalgam .
because @xmath39 and @xmath40 are induced subgraph of @xmath1 , if @xmath1 is long - unichord - free then both @xmath39 and @xmath40 are long - unichord - free .
conversely , suppose that @xmath39 and @xmath40 are long - unichord - free and assume @xmath17 is the cycle with unique chord with length at least @xmath451 of @xmath1 .
suppose first that the unique chord divides @xmath17 into two holes .
we say that the type of @xmath17 is @xmath452 , according to the types @xmath45 and @xmath453 of the two holes with respect to the amalgam .
based on lemma [ typeh ] , and the property that these two holes share only one edges , there are @xmath454 possible types for @xmath17 : @xmath455 , @xmath456 , @xmath457 , @xmath458 , @xmath459 , @xmath460 , @xmath461 , @xmath462 , @xmath463 , @xmath464 , @xmath465 , @xmath466 , @xmath467 .
if @xmath17 belongs to type @xmath467 then any vertex of @xmath59 ( this vertex is exist since @xmath313 ) with hole type @xmath468 induces a cycle with a long unichord in in @xmath39 , a contradiction .
if @xmath17 belongs to the remaining types , either @xmath39 or @xmath40 contains @xmath17 , contradiction .
hence , we may assume that the unique chord divides @xmath17 into a hole and a triangle ( @xmath17 is a cap ) .
this also means that @xmath17 consists of a hole @xmath10 plus a vertex @xmath21 that is adjacent to two adjacent vertices of this hole . if @xmath10 type @xmath469 , @xmath21 does not exits .
if @xmath10 is type @xmath470 , every choice of @xmath21 makes @xmath17 belong to @xmath39 and @xmath40 . if @xmath10 is type @xmath468 ( or @xmath471 ) , this hole with a vertex of @xmath59 ( or @xmath58 ) induced a cycle with a long - unichord in @xmath39 , a contradiction .
we now describe our algorithm ( similar to the algorithm to recognize cap - free graph from @xcite ) .
a graph is _ basic _ if it is chordal or unichord - free .
a _ decomposition tree _
@xmath472 of a graph @xmath1 is defined as follows : * the root of @xmath472 is @xmath1 .
* if some node @xmath10 of @xmath472 is not basic and has a universal vertex , then its unique child is @xmath473 where @xmath45 is the set of all universal vertices of @xmath10 . *
if some node of @xmath472 is not basic , has no universal vertex and has a decomposition , then its children are its blocks of decomposition . *
all nodes not handled in the previous cases are leaves of @xmath472 ( to be more specific : basic nodes and nodes that are not basic , without universal vertices and decomposition ) .
note that the definition is not fully deterministic since different decompositions can be present in a graph . in this case ,
one of the decomposition should be used ( so a graph may have different decomposition trees ) .
note that every graph has a decompostion tree , and that every decomposition tree of a graph is finite ( because the children of a given node are smaller than the node ) .
[ l : sum ] suppose that @xmath1 is a non - basic graph , with no universal vertex and with an amalgam or a cutvertex .
let @xmath39 and @xmath40 be the blocks of decompostion of @xmath1 with respect to this amalgam or cutvertex
. then @xmath475 .
if for @xmath61 we have @xmath476 , then every pair @xmath477 such that @xmath478 can be associated injectively to a similar pair in @xmath1 , so the inequality holds .
hence , we may assume that @xmath479 . if we have @xmath480 , then every pair @xmath477 such that @xmath481 can be associated injectively to a similar pair in @xmath1 .
since @xmath1 has no universal vertex , some vertex @xmath482 has a non - neighbor in @xmath1 and provides an extra non - adjacent pair in @xmath1 , so that the inequality holds .
hence , we may assume that @xmath483 .
so , we just have to check that @xmath484 .
this is the case because @xmath1 is not basic , so it is not chordal and contains a chordless cycle of length at least 4 ( that provides at least two non - adjacent pairs ) . the next lemma is implicitly proved in @xcite ( as corollary 2.16 ) , but the machinery there is much heavier and relies on many definitions , so we prefer to give our own simple proof .
note that in @xcite , it is claimed without proof that any graph with an amalgam should have an amalgam such that at least one block of decomposition has no amalgam .
such a result would imply the existence of a decomposition tree of linear size , but unfortunately , in @xcite , a counter - example to the claim is provided . if a graph @xmath10 has a universal vertex @xmath20 , then @xmath486 .
hence , by lemma [ l : sum ] the number of leaves of @xmath472 is at most @xmath487 . since removing the set of universal vertices
can be done at most once to any node of the decomposition tree , we obtain the bound @xmath485 .
the first step is to build a decomposition tree for @xmath1 .
deciding whether a graph is basic can be performed in time @xmath491 ( see @xcite for unichord - free graphs and @xcite for chordal graphs ) . finding a universal vertex ,
a cutvertex or an amalgam can be performed in time @xmath492 ( see @xcite for the amalgam and the other claims are trivial ) .
so , the tree can be constructed .
once the tree is given , the algorithm checks whether all leaves are basics , and by lemma [ l : decbas ] , this decides whether the graph is long - unichord - free .
j. fonlupt and j.p .
transformations which preserve perfectness and @xmath493-perfectness of graphs . in a.
bachem , m. grtschel , and b. korte , editors , _ bonn workshop on combinatorial optimization _ , pages 8385 .
north - holland , 1982 .
annals of discrete mathematics , 16 . | a _ long unichord _ in a graph is an edge that is the unique chord of some cycle of length at least 5 .
a graph is _ long - unichord - free _ if it does not contain any long - unichord .
we prove a structure theorem for long - unichord - free graph .
we give an @xmath0-time algorithm to recognize them .
we show that any long - unichord - free graph @xmath1 can be colored with at most @xmath2 colors , where @xmath3 is the maximum number of pairwise adjacent vertices in @xmath1 .
* key words : * amalgam , @xmath4-bounded , chords .
* ams classification : * 05c75 , 05c15 , 05c85 . |
next to photons neutrinos are the most abundant particles in the universe .
this means they have a profound impact on many different aspects of cosmology , from the question of leptogenesis in the very early universe , over big bang nucleosynthesis , to late time structure formation . in the present review
i focus mainly on late - time aspects of neutrino cosmology , and particularly on issues relevant to cosmological bounds on the neutrino mass .
the absolute value of neutrino masses are very difficult to measure experimentally . on the other hand , mass differences between neutrino mass eigenstates , @xmath0 ,
can be measured in neutrino oscillation experiments .
the combination of all currently available data suggests two important mass differences in the neutrino mass hierarchy .
the solar mass difference of @xmath1 ev@xmath2 and the atmospheric mass difference @xmath3 ev@xmath2 @xcite .
in the simplest case where neutrino masses are hierarchical these results suggest that @xmath4 , @xmath5 , and @xmath6 .
if the hierarchy is inverted @xcite one instead finds @xmath7 , @xmath8 , and @xmath9 .
however , it is also possible that neutrino masses are degenerate @xcite , @xmath10 , in which case oscillation experiments are not useful for determining the absolute mass scale .
experiments which rely on kinematical effects of the neutrino mass offer the strongest probe of this overall mass scale .
tritium decay measurements have been able to put an upper limit on the electron neutrino mass of 2.3 ev ( 95% conf . )
@xcite . however , cosmology at present yields an much stronger limit which is also based on the kinematics of neutrino mass .
very interestingly there is also a claim of direct detection of neutrinoless double beta decay in the heidelberg - moscow experiment @xcite , corresponding to an effective neutrino mass in the @xmath11 ev range .
if this result is confirmed then it shows that neutrino masses are almost degenerate and well within reach of cosmological detection in the near future .
another important question which can be answered by cosmological observations is how large the total neutrino energy density is .
apart from the standard model prediction of three light neutrinos , such energy density can be either in the form of additional , sterile neutrino degrees of freedom , or a non - zero neutrino chemical potential .
the paper is divided into sections in the following way : in section 2 i review the present cosmological data which can be used for analysis of neutrino physics . in section 3
i discuss neutrino physics around the epoch of neutrino decoupling at a temperature of roughly 1 mev , including the relation between neutrinos and big bang nucleosynthesis .
section 4 discusses neutrinos as dark matter particles , including mass constraints on light neutrinos , and sterile neutrino dark matter .
section 5 contains a relatively short review of neutrino physics in the very early universe from the perspective of leptogenesis .
finally , section 6 contains a discussion .
[ [ large - scale - structure - lss ] ] large scale structure ( lss ) + + + + + + + + + + + + + + + + + + + + + + + + + + + + + at present there are two large galaxy surveys of comparable size , the sloan digital sky survey ( sdss ) @xcite and the 2dfgrs ( 2 degree field galaxy redshift survey ) @xcite .
once the sdss is completed in 2005 it will be significantly larger and more accurate than the 2dfgrs . at present
the two surveys are , however , comparable in precision .
[ [ cosmic - microwave - background - cmb ] ] cosmic microwave background ( cmb ) + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the temperature fluctuations are conveniently described in terms of the spherical harmonics power spectrum @xmath12 .
since thomson scattering polarizes light , there are also power spectra coming from the polarization .
the polarization can be divided into a curl - free @xmath13 and a curl @xmath14 component , yielding four independent power spectra : @xmath15 , @xmath16 , @xmath17 , and the @xmath18-@xmath19 cross - correlation @xmath20 .
the wmap experiment has reported data only on @xmath15 and @xmath20 as described in refs .
foreground contamination has already been subtracted from their published data .
in addition to the wmap experiment there are a number of other current cmb experiments , both ground and balloon based .
_ @xcite have provided a compilation of various data sets , and in addition to these there is the acbar experiment @xcite which has measured the cmb at small scales .
[ [ other - data ] ] other data + +
+ + + + + + + + + + apart from cmb and lss data there are a number of other cosmological measurements of importance to neutrino cosmology .
one is the measurement of the hubble constant by the hst hubble key project , @xmath21 @xcite .
the constraint on the matter density coming from measurements of distant type ia supernovae is also important for neutrino physics .
the most recent result is from the supernova cosmology project @xcite and yields @xmath22 ( statistical ) @xmath23 ( identified systematics ) .
in the standard model neutrinos interact via weak interactions with @xmath24 and @xmath25 . in the absence of oscillations neutrino
decoupling can be followed via the boltzmann equation for the single particle distribution function @xcite @xmath26 where @xmath27 represents all elastic and inelastic interactions . in the standard model
all these interactions are @xmath28 interactions in which case the collision integral for process @xmath29 can be written @xmath30 where @xmath31 is the spin - summed and averaged matrix element including the symmetry factor @xmath32 if there are identical particles in initial or final states .
the phase - space factor is @xmath33 .
the matrix elements for all relevant processes can for instance be found in ref . @xcite . if maxwell - boltzmann statistics is used for all particles , and neutrinos are assumed to be in complete scattering equilbrium so that they can be represented by a single temperature , then the collision integral can be integrated to yield the average annihilation rate for a neutrino @xmath34 where @xmath35 this rate
can then be compared with the hubble expansion rate @xmath36 to find the decoupling temperature from the criterion @xmath37 . from this one
finds that @xmath38 mev , @xmath39 mev , when @xmath40 , as is the case in the standard model .
this means that neutrinos decouple at a temperature which is significantly higher than the electron mass .
when @xmath41 annihilation occurs around @xmath42 , the neutrino temperature is unaffected whereas the photon temperature is heated by a factor @xmath43 .
the relation @xmath44 holds to a precision of roughly one percent .
the main correction comes from a slight heating of neutrinos by @xmath41 annihilation , as well as finite temperature qed effects on the photon propagator @xcite . shortly after neutrino decoupling
the weak interactions which keep neutrons and protons in statistical equilibrium freeze out .
again the criterion @xmath45 can be applied to find that @xmath46 mev @xcite .
eventually , at a temperature of roughly 0.2 mev deuterium starts to form , and very quickly all free neutrons are processed into @xmath47he .
the final helium abundance is therefore roughly given by @xmath48 @xmath49 is determined by its value at freeze out , roughly by the condition that @xmath50 . since the freeze - out temperature is determined by @xmath51 this in turn means that @xmath51 can be inferred from a measurement of the helium abundance . however , since @xmath52 is a function of both @xmath53 and @xmath51 it is necessary to use other measurements to constrain @xmath53 in order to find a bound on @xmath51 .
one customary method for doing this has been to use measurements of primordial deuterium to infer @xmath53 and from that calculate a bound on @xmath51 .
usually such bounds are expressed in terms of the equivalent number of neutrino species , @xmath54 , instead of @xmath51 .
the exact value of the bound is quite uncertain because there are different and inconsistent measurements of the primordial helium abundance ( see for instance ref .
@xcite for a discussion of this issue ) .
the most recent analyses are @xcite where a value of @xmath55 ( 95% c.l . )
was found and @xcite which found the result @xmath56 .
the difference in these results can be attributed to different assumptions about uncertainties in the primordial helium abundance .
another interesting parameter which can be constrained by the same argument is the neutrino chemical potential , @xmath57 @xcite . at first sight
this looks like it is completely equivalent to constraining @xmath58 .
however , this is not true because a chemical potential for electron neutrinos directly influences the @xmath59 conversion rate .
therefore the bound on @xmath60 from bbn alone is relatively stringent ( @xmath61 @xcite ) compared to that for muon and tau neutrinos ( @xmath62 @xcite ) . however , as will be seen in the next section , neutrino oscillations have the effect of almost equilibrating the neutrino chemical potentials prior to bbn , completely changing this conclusion .
the bbn bound on the number of neutrino species presented in the previous section can be complemented by a similar bound from observations of the cmb and large scale structure .
the cmb depends on @xmath58 mainly because of the early integrated sachs wolfe effect which increases fluctuation power at scales slightly larger than the first acoustic peak .
the large scale structure spectrum depends on @xmath58 because the scale of matter - radiation equality is changed by varying @xmath58 .
several recent papers have analyzed wmap and 2df data for bounds on @xmath58 @xcite , and some of the bounds are listed in table [ table : nnu ] .
recent analyses combining bbn , cmb , and large scale structure data can be found in @xcite , and these results are also listed in table [ table : nnu ] .
common for all the bounds is that @xmath63 is ruled out by both bbn and cmb / lss .
this has the important consequence that the cosmological neutrino background has been positively detected , not only during the bbn epoch , but also much later , during structure formation .
.various recent limits on the effective number of neutrino species , as well as the data used . [ cols="<,<,<",options="header " , ] [ table : summary ] another cornerstone of neutrino cosmology is the measurement of the total energy density in non - electromagnetically interacting particles . for many years
big bang nucleosynthesis was the only probe of relativistic energy density , but with the advent of precision cmb and lss data it has been possible to complement the bbn measurement . at present the cosmic neutrino background is seen in both bbn , cmb and lss data at high significance . finally , cosmology can also be used to probe the possibility of neutrino warm dark matter , which could be produced by active - sterile neutrino oscillations . in the coming years the steady stream of new observational data will continue , and the cosmological bounds on neutrino will improve accordingly .
for instance , it has been estimated that with data from the upcoming planck satellite it could be possible to measure neutrino masses as low as 0.1 ev . certainly neutrino
cosmology will continue to be a prospering field of research for the foreseeable future .
i acknowledge use of the publicly available cmbfast package written by uros seljak and matthias zaldarriaga @xcite and the use of computing resources at dcsc ( danish center for scientific computing ) .
i also wish to thank georg raffelt for valuable comments on the manuscript .
99 p. aliani , v. antonelli , m. picariello and e. torrente - lujan , arxiv : hep - ph/0309156 .
p. c. de holanda and a. y. smirnov , arxiv : hep - ph/0309299 . m. maltoni , t. schwetz , m. a. tortola and j. w. f. valle , phys .
d * 68 * , 113010 ( 2003 ) [ arxiv : hep - ph/0309130 ] . v. a. kostelecky and s. samuel , phys .
b * 318 * , 127 ( 1993 ) .
g. m. fuller , j. r. primack and y. z. qian , phys .
d * 52 * , 1288 ( 1995 ) [ arxiv : astro - ph/9502081 ] .
d. o. caldwell and r. n. mohapatra , phys .
b * 354 * , 371 ( 1995 ) [ arxiv : hep - ph/9503316 ] . s. m. bilenky , c. giunti , c. w. kim and s. t. petcov , phys .
d * 54 * , 4432 ( 1996 ) [ arxiv : hep - ph/9604364 ] . s. f. king and n. n. singh , nucl .
b * 596 * , 81 ( 2001 ) [ arxiv : hep - ph/0007243 ] .
h. j. he , d. a. dicus and j. n. ng , arxiv : hep - ph/0203237 .
a. ioannisian and j. w. valle , phys .
b * 332 * , 93 ( 1994 ) [ arxiv : hep - ph/9402333 ] .
p. bamert and c. p. burgess , phys .
b * 329 * , 289 ( 1994 ) [ arxiv : hep - ph/9402229 ] .
r. n. mohapatra and s. nussinov , phys .
b * 346 * , 75 ( 1995 ) [ arxiv : hep - ph/9411274 ] .
h. minakata and o. yasuda , phys .
d * 56 * , 1692 ( 1997 ) [ arxiv : hep - ph/9609276 ] .
f. vissani , arxiv : hep - ph/9708483 .
h. minakata and o. yasuda , nucl .
b * 523 * , 597 ( 1998 ) [ arxiv : hep - ph/9712291 ] .
j. r. ellis and s. lola , phys .
b * 458 * , 310 ( 1999 ) [ arxiv : hep - ph/9904279 ] .
j. a. casas , j. r. espinosa , a. ibarra and i. navarro , nucl .
b * 556 * , 3 ( 1999 ) [ arxiv : hep - ph/9904395 ] .
j. a. casas , j. r. espinosa , a. ibarra and i. navarro , nucl .
b * 569 * , 82 ( 2000 ) [ arxiv : hep - ph/9905381 ] .
e. ma , j. phys .
g * 25 * , l97 ( 1999 ) [ arxiv : hep - ph/9907400 ] .
r. adhikari , e. ma and g. rajasekaran , phys .
b * 486 * , 134 ( 2000 ) [ arxiv : hep - ph/0004197 ] . c. kraus _ et al .
european physical journal c ( 2003 ) , proceedings of the eps 2003 - high energy physics ( hep ) conference .
d. n. spergel _ et al .
_ , astrophys .
j. suppl .
* 148 * ( 2003 ) 175 [ astro - ph/0302209 ] . c. l. bennett _ et al . _ ,
preliminary maps and basic results , astrophys .
j. suppl .
* 148 * ( 2003 ) 1 [ astro - ph/0302207 ] .
a. kogut _ et al .
_ , astrophys .
j. suppl .
* 148 * ( 2003 ) 161 [ astro - ph/0302213 ] .
g. hinshaw _ et al .
_ , astrophys .
j. suppl .
* 148 * ( 2003 ) 135 [ astro - ph/0302217 ] .
l. verde _ et al .
_ , astrophys . j. suppl .
* 148 * ( 2003 ) 195 [ astro - ph/0302218 ] .
h. v. peiris _ et al .
_ , astrophys .
j. suppl .
* 148 * ( 2003 ) 213 [ astro - ph/0302225 ] . c. l. kuo _ et al . _
[ acbar collaboration ] , astrophys .
j. * 600 * , 32 ( 2004 ) [ arxiv : astro - ph/0212289 ] .
x. wang , m. tegmark , b. jain and m. zaldarriaga , astro - ph/0212417 .
s. hannestad and j. madsen , phys .
d * 52 * , 1764 ( 1995 ) [ arxiv : astro - ph/9506015 ] .
d. a. dicus , e. w. kolb , a. m. gleeson , e. c. sudarshan , v. l. teplitz and m. s. turner , phys .
d * 26 * , 2694 ( 1982 ) .
n. c. rana and b. mitra , phys . rev .
d * 44 * , 393 ( 1991 ) .
m. a. herrera and s. hacyan , astrophys . j. * 336 * , 539 ( 1989 ) .
a. d. dolgov and m. fukugita , phys .
d * 46 * , 5378 ( 1992 ) .
s. dodelson and m. s. turner , phys .
d * 46 * , 3372 ( 1992 ) .
b. d. fields , s. dodelson and m. s. turner , phys .
d * 47 * , 4309 ( 1993 ) [ arxiv : astro - ph/9210007 ] .
a. d. dolgov , s. h. hansen and d. v. semikoz , nucl .
b * 503 * , 426 ( 1997 ) [ arxiv : hep - ph/9703315 ] .
a. d. dolgov , s. h. hansen and d. v. semikoz , nucl .
b * 543 * , 269 ( 1999 ) [ arxiv : hep - ph/9805467 ] .
n. y. gnedin and o. y. gnedin , astrophys .
j. * 509 * , 11 ( 1998 ) .
s. esposito , g. miele , s. pastor , m. peloso and o. pisanti , nucl .
b * 590 * , 539 ( 2000 ) [ arxiv : astro - ph/0005573 ] .
g. steigman , arxiv : astro - ph/0108148 . g. mangano , g. miele , s. pastor and m. peloso , arxiv : astro - ph/0111408 .
s. hannestad , physical review d * 65 * , 083006 ( 2002 ) .
v. barger , j. p. kneller , h. s. lee , d. marfatia and g. steigman , phys .
b * 566 * , 8 ( 2003 ) [ arxiv : hep - ph/0305075 ] .
a. cuoco , f. iocco , g. mangano , g. miele , o. pisanti and p. d. serpico , arxiv : astro - ph/0307213 .
k. ichikawa and m. kawasaki , phys .
d * 67 * , 063510 ( 2003 ) [ arxiv : astro - ph/0210600 ] .
s. hannestad , jcap * 0305 * ( 2003 ) 004 [ astro - ph/0303076 ] .
p. crotty , j. lesgourgues and s. pastor , phys .
d * 67 * , 123005 ( 2003 ) [ arxiv : astro - ph/0302337 ] .
e. pierpaoli , mon . not .
soc . * 342 * , l63 ( 2003 ) [ arxiv : astro - ph/0302465 ] .
a. d. dolgov , phys .
rept . * 370 * , 333 ( 2002 ) [ arxiv : hep - ph/0202122 ] . c. lunardini and a. y. smirnov , phys .
d * 64 * , 073006 ( 2001 ) [ arxiv : hep - ph/0012056 ] .
s. pastor , g. g. raffelt and d. v. semikoz , arxiv : hep - ph/0109035 .
a. d. dolgov , s. h. hansen , s. pastor , s. t. petcov , g. g. raffelt and d. v. semikoz , nucl .
b * 632 * , 363 ( 2002 ) [ arxiv : hep - ph/0201287 ] .
k. n. abazajian , j. f. beacom and n. f. bell , phys .
d * 66 * , 013008 ( 2002 ) [ arxiv : astro - ph/0203442 ] . y. y. wong , phys . rev .
d * 66 * , 025015 ( 2002 ) [ arxiv : hep - ph/0203180 ] .
m. kawasaki , k. kohri and n. sugiyama , phys .
lett . * 82 * , 4168 ( 1999 ) [ arxiv : astro - ph/9811437 ] .
d * 62 * , 023506 ( 2000 ) [ arxiv : astro - ph/0002127 ]
. g. f. giudice , e. w. kolb and a. riotto , phys .
d * 64 * , 023508 ( 2001 ) [ arxiv : hep - ph/0005123 ] .
g. f. giudice , e. w. kolb , a. riotto , d. v. semikoz and i. i. tkachev , phys .
d * 64 * , 043512 ( 2001 ) [ arxiv : hep - ph/0012317 ] .
s. hannestad , arxiv : astro - ph/0403291 . s. s. gershtein and y. b. zeldovich , jetp lett .
* 4 * , 120 ( 1966 ) [ pisma zh .
, 174 ( 1966 ) ] .
r. cowsik and j. mcclelland , phys .
* 29 * , 669 ( 1972 ) .
s. tremaine and j. e. gunn , phys .
lett . *
42 * , 407 ( 1979 ) .
j. madsen , phys .
d * 44 * , 999 ( 1991 ) .
p. salucci and a. sinibaldi , astron .
, 1 ( 1997 ) . c. p. ma and e. bertschinger , astrophys .
j. * 455 * , 7 ( 1995 ) [ arxiv : astro - ph/9506072 ] . w. hu , d. j. eisenstein and m. tegmark , phys .
* 80 * ( 1998 ) 5255 [ astro - ph/9712057 ] .
o. elgaroy and o. lahav , jcap * 0304 * ( 2003 ) 004 [ astro - ph/0303089 ] .
s. hannestad and g. raffelt , arxiv : hep - ph/0312154 .
p. crotty , j. lesgourgues and s. pastor , arxiv : hep - ph/0402049 .
k. n. abazajian and s. dodelson , phys .
* 91 * ( 2003 ) 041301 [ astro - ph/0212216 ] .
u. seljak and m. zaldarriaga , astrophys . j. * 469 * ( 1996 ) 437 [ astro - ph/9603033 ] .
see also the cmbfast website at http://cosmo.nyu.edu/matiasz/cmbfast/cmbfast.html s. perlmutter _ et al . _
[ supernova cosmology project collaboration ] , astrophys . j. * 517 * ( 1999 ) 565 [ astro - ph/9812133 ] .
a. bandyopadhyay , s. choubey , s. goswami , s. t. petcov and d. p. roy , phys .
b * 583 * , 134 ( 2004 ) [ arxiv : hep - ph/0309174 ] . m. kaplinghat , l. knox and y. s. song , phys .
lett . * 91 * , 241301 ( 2003 ) [ arxiv : astro - ph/0303344 ] .
s. kazantzidis , l. mayer , c. mastropietro , j. diemand , j. stadel and b. moore , arxiv : astro - ph/0312194 .
s. ghigna , b. moore , f. governato , g. lake , t. quinn and j. stadel , astrophys . j. * 544 * , 616 ( 2000 ) [ arxiv : astro - ph/9910166 ] .
n. yoshida , a. sokasian , l. hernquist and v. springel , astrophys . j. * 591 * , l1 ( 2003 ) [ arxiv : astro - ph/0303622 ] .
z. haiman , r. barkana and j. p. ostriker , arxiv : astro - ph/0103050 . v. avila - reese , p. colin , o. valenzuela , e. donghia and c. firmani , arxiv : astro - ph/0010525 .
p. bode , j. p. ostriker and n. turok , astrophys .
j. * 556 * , 93 ( 2001 ) [ arxiv : astro - ph/0010389 ] .
colin , v. avila - reese and o. valenzuela , astrophys .
j. * 542 * , 622 ( 2000 ) [ arxiv : astro - ph/0004115 ] .
s. hannestad and r. j. scherrer , phys .
d * 62 * , 043522 ( 2000 ) [ arxiv : astro - ph/0003046 ]
. j. sommer - larsen and a. dolgov , arxiv : astro - ph/9912166 .
s. colombi , s. dodelson and l. m. widrow , astrophys . j. * 458 * , 1 ( 1996 ) [ arxiv : astro - ph/9505029 ] . v. k. narayanan , d. n. spergel , r. dave and c. p. ma , arxiv : astro - ph/0005095 .
k. abazajian , g. m. fuller and w. h. tucker , astrophys .
j. * 562 * , 593 ( 2001 ) [ arxiv : astro - ph/0106002 ] . k. abazajian , g. m. fuller and m. patel , phys .
d * 64 * , 023501 ( 2001 ) [ arxiv : astro - ph/0101524 ] .
x. d. shi and g. m. fuller , phys .
lett . * 82 * , 2832 ( 1999 ) [ arxiv : astro - ph/9810076 ] .
s. dodelson and l. m. widrow , phys .
* 72 * , 17 ( 1994 ) [ arxiv : hep - ph/9303287 ] .
c. j. hogan and j. j. dalcanton , phys .
d * 62 * , 063511 ( 2000 ) [ arxiv : astro - ph/0002330 ] .
j. j. dalcanton and c. j. hogan , astrophys .
j. * 561 * , 35 ( 2001 ) [ arxiv : astro - ph/0004381 ] . t. j. weiler , phys .
lett . * 49 * , 234 ( 1982 ) .
t. j. weiler , astropart .
phys . *
11 * , 303 ( 1999 ) [ arxiv : hep - ph/9710431 ] .
a. ringwald , arxiv : hep - ph/0111112 .
z. fodor , s. d. katz and a. ringwald , phys .
* 88 * , 171101 ( 2002 ) [ arxiv : hep - ph/0105064 ] .
g. gelmini and g. varieschi , arxiv : hep - ph/0201273 .
g. gelmini and a. kusenko , phys .
82 * , 5202 ( 1999 ) .
o. e. kalashev , v. a. kuzmin , d. v. semikoz and g. sigl , phys .
d * 65 * , 103003 ( 2002 ) [ arxiv : hep - ph/0112351 ] .
k. greisen , phys . rev
. lett .
* 16 * , 748 ( 1966 ) ; g. t. zatsepin and v. a. kuzmin , jetp lett . * 4 * , 78 ( 1966 ) . | the current status of neutrino cosmology is reviewed , from the question of neutrino decoupling and the presence of sterile neutrinos to the effects of neutrinos on the cosmic microwave background and large scale structure .
particular emphasis is put on cosmological neutrino mass measurements . |
the purpose of this paper is to present a novel approach for efficient estimation via the monte carlo ( mc ) method .
the approach is very broadly applicable but here , to present the main ideas , we narrow the focus to ensemble monte carlo where estimation is based on stochastically independent trajectories of a system . to illustrate , we use simulation of time - dependent nonlinear processes
for which monte carlo is a particularly general and powerful numerical method compared to available alternatives .
time - dependent nonlinear processes are very general models used , among others , in statistical mechanics @xcite , data assimilation in climate , weather and ocean modeling @xcite , financial modeling @xcite , and quantitative biology @xcite . hence developing efficient mc methods
may significantly impact a wide range of applications .
a known weakness of mc is its slow rate of convergence .
assume @xmath0 is a random quantity defined on paths of a process and let @xmath1 denote its standard deviation . the convergence rate of mc for estimating the expected value of @xmath0 is @xmath2 where @xmath3 is the number of independent paths of the process .
in general the canonical @xmath4 rate of convergence can not be improved upon , hence , since the inception of the mc method , a number of variance reduction ( vr ) techniques have been devised to reduce @xmath1 ( see , @xcite for an early account and @xcite and @xcite for more recent discussions ) .
most vr techniques lead to estimators of the form @xmath5 i.e. , a weighted average of the samples .
these techniques prescribe ( i ) a recipe for selecting samples @xmath6 and ( ii ) a set of weights @xmath7 .
to arrive at these prescriptions , one must rely on the existence of specific problem features and the ability of the user of the method to discover and effectively exploit such features .
this lack of generality has significantly limited the applicability of vr techniques .
the point of departure of a new strategy , called database monte carlo ( dbmc ) , is to address this shortcoming and to devise generic vr techniques that can be generically applied @xcite .
all vr techniques bring additional information to bear on the estimation problem , however , as mentioned above , this information is problem specific and relies on exploiting special features of the problem at hand .
by contrast , as will be clarified in this paper , dbmc adds a generic computational exploration phase to the estimation problem that relies on gathering information at one ( or more ) nominal model parameter(s ) to achieve estimation efficiency at neighboring parameters .
the advantage of this approach is its generality and wide applicability : it is quite easy to implement and it can wrap existing ensemble mc codes . on the other hand , the computational exploration phase of the dbmc approach may require extensive simulations and can be computationally costly .
therefore , the initial setup cost needs justification .
the setup cost may be justified in projects that involve estimations at many model parameters and/or in projects where there is a real - time computational constraint . in the first type of project
, the setup cost may lead to efficiency gain for each subsequent estimation , and for a large enough number of subsequent estimations it can be easily justified . in projects with a real - time constraint
the setup cost is an off - line `` passive '' cost that can lead to estimates of significantly higher quality ( lower statistical error ) ; the higher quality in many such projects more than justifies the setup cost . in this paper
we limit ourselves to presenting the implementation of the vr technique of control variates ( cv ) in the dbmc setting ( see @xcite for discussion of other vr techniques ) .
the cv technique , which compared to the vr technique of importance sampling is less utilized in computational physics , requires identifying a number of random variables called _ control variates _ , say @xmath8 , that are correlated with @xmath0 and have _ known _ means .
the correlation with @xmath0 implies that @xmath9 s carry information about @xmath0 .
the cv technique is a way of utilizing the information included in the controls ( their known means ) to help with the estimation of the mean of variable @xmath0 . in the dbmc setting we assume that @xmath10 depends on a model parameter @xmath11 and use @xmath12 where @xmath13 s are in a neighborhood of @xmath11 ( @xmath14 ) . in a departure from the classical cv technique ,
we use `` high quality '' estimates of @xmath15 $ ] rather than precise values of @xmath15 $ ] to arrive at the controlled estimator of @xmath16 $ ] . as we argue in this paper ( and elsewhere @xcite ) this departure
allows for substantially broader choices of control variates and makes the cv technique significantly more flexible and effective .
the dbmc method shares a similar intent as the well - known histogram reweighing method @xcite from the markov chain monte carlo literature ( e.g. @xcite ) , but with a very different setting and implementation , and with broader applicability .
for example , it does not rely on having a boltzmann distribution or @xmath17 structure .
given its generality , it has potential applications , among others , in ensemble weather prediction , hydrological source location , climate and ocean , optimal control , and stochastic simulations of biological systems .
the remainder of the paper is organized as follows . in section [ prel ]
we discuss preliminaries , including the details of the example numerical study
the time - dependent ginzburg - landau ( tdgl ) equation as well as the method of control variates .
estimation of mean outcomes of the tdgl equation over a range of temperatures is of interest , especially considering the large difference in behavior below and above the coexistence curve . in section [ methods ]
we describe the dbmc methodology and motivation in a general context .
section [ numerical ] discusses the implementation and results of dbmc as applied to estimation of quantities generated by the tdgl equation , and the results of that numerical study .
we conclude in section [ conclusions ] .
we present aspects of our approach and numerical results in the context of the time - dependent ginzburg landau ( tdgl ) model .
it is worth noting that this model is chosen for illustrative purposes only and we do not make use of any of its specific features .
we use a canonical equation of phase - ordering kinetics @xcite the stochastic tdgl equation in two spatial dimensions .
this is written as @xmath18 where @xmath19 represents a local order parameter , e.g. a magnetization at point @xmath20 and time @xmath21 ( @xmath22 denotes transpose ) .
the noise has mean zero and covariance @xmath23 .
we choose a double - well potential @xmath24 . as in @xcite @xmath25
is a constant , and @xmath11 is a function of temperature such that a high @xmath11 corresponds to a low temperature .
we use a discrete form of ( [ eq : tdgl ] ) using a forward euler - maruyama stochastic integrator and a 5-point stencil for the laplacian ( denoted @xmath26 ) for simulation : @xmath27\\ \nonumber & + & \sqrt{2 ( \delta t/\delta x ) } n({\bf x},t)\end{aligned}\ ] ] with time step @xmath28 and lattice spacing @xmath29 , and where @xmath30 are independent and identically distributed standard normal random variables for each space - time point @xmath31 .
what follows applies to other discretization schemes as well . to cover a broad range of estimation problems , we consider the estimation of quantities related to a specific space - time point , quantities that are global ( entire lattice at a particular time ) and quantities that depend on the entire time evolution of the system . specifically , we consider the following representative quantities : 1 . _ point magnetization _ : @xmath32 , 2 .
_ total magnetization at a specific time @xmath21 _ : @xmath33 , and 3 .
_ total space - time magnetization _ : @xmath34
. the problem of estimating the expected value of any one of the above quantities can be represented by : @xmath35\ ] ] where @xmath36 is a vector of random numbers representing all the noise / uncertainty in a single complete path @xmath37 of the dynamics ; @xmath11 is the temperature related parameter ; @xmath38 is the random sample of a quantity of interest ( e.g. , the magnetization from a single sample path ) , and @xmath39 $ ] denotes expectation .
note that knowing the noise @xmath36 and parameter @xmath11 completely determines the path @xmath37 and the sample quantity of interest @xmath38 . here
we give a brief review of the classical control variate ( cv ) technique for variance reduction ( see @xcite @xcite ) .
let @xmath40 , @xmath41 $ ] .
assume @xmath8 are random variables ( called control variates ) that are correlated with @xmath0 and assume their means @xmath15 $ ] are known .
let @xmath42 , @xmath43=(e[x_1 ] , \cdots , e[x_k])^\top$ ] , and @xmath44 .
then @xmath45 , defined below , is a controlled estimator of @xmath16 $ ] @xmath46)= y+ \beta^\top({\bf x}-e[{\bf x}])\ ] ] the estimator @xmath45 uses information included in samples of the controls ( the degree of their deviation from their known means ) to `` correct / adjust '' the estimator @xmath0 and bring it closer to its unknown mean .
this is the key idea of cv .
( alternatively , @xmath45 can be viewed as the fitted value of @xmath0 when @xmath0 is linearly regressed on variables @xmath47 . in other words
, @xmath45 includes the part of the variation in @xmath0 that can not be `` explained '' by @xmath9 s . )
@xmath45 is an unbiased estimator of @xmath16 $ ] for all vectors @xmath48 .
the coefficient vector that minimizes the variance of @xmath45 is : @xmath49 where @xmath50 is the @xmath51 covariance matrix of @xmath52 and @xmath53 is the @xmath54 vector of covariances of @xmath0 and @xmath9 s .
when @xmath55 is used , the variance of @xmath45 is given by @xmath56 where @xmath57 therefore , @xmath58 and hence @xmath59 is precisely the theoretical degree of variance reduction if the controlled estimator @xmath60)$ ] is used to estimate @xmath61 as opposed to the crude mc estimator @xmath0 , and it is called the variance reduction ratio ( vrr ) statistic for control variates .
note that there is no upper limit to the degree of achievable variance reduction since @xmath62 can potentially be very close to @xmath63 when the controls are highly correlated with the estimation variable @xmath0 .
in other words , the cv technique can potentially be very effective leading to orders of magnitude of variance reduction . in practice and in general , @xmath50 and @xmath53 ( i.e. @xmath55 )
are not known exactly and need to be estimated from samples of @xmath9 s and @xmath0 .
typically , @xmath55 is estimated from the same samples used to construct the controlled estimator @xmath45 . while this practice adds some bias for small sample sizes , and
thus makes the effective decrease in estimator mean squared error not precisely equal to the variance reduction ratio @xmath59 , this bias converges to zero faster than the standard error of @xmath45 .
thus , expending computational resources into generating separate pilot samples for estimating @xmath55 is not considered to be justifiable . for an insightful and detailed discussion of the cv technique ,
see @xcite . the critical task for using the cv technique
is in finding effective controls . once the controls are selected , the rest of the procedure is fairly routine .
an effective control , say @xmath64 , needs to satisfy two requirements ( to simplify the discussion we consider a scalar control ) : 1 .
@xmath64 needs to be correlated with @xmath0 , and 2 .
@xmath65 $ ] needs to be available to the user , i.e. , known .
the main barrier to finding effective controls is the second requirement , namely the requirement of a known mean @xmath65 $ ] .
a modification of the cv technique called biased control variate ( bcv ) reduces the burden of requirement ( r2 ) by allowing for a good approximation of @xmath65 $ ] when @xmath65 $ ] can not be evaluated analytically @xcite . while bcv lowers the requirement barrier and expands the range of available choices for effective controls , it nonetheless limits its potential scope by implicitly assuming an analytic path to arriving at the approximate value . as we describe in the next section , in the dbmc approach we turn the second requirement into a computational task
; in other words , we use statistical estimation to obtain a good estimate of @xmath65 $ ] .
therefore , barrier ( r2 ) is completely removed and the range of choices of controls is dramatically expanded .
the relevant question now becomes whether the computational investment in estimating @xmath65 $ ] pays enough dividends to make the investment worthwhile .
the starting point of the dbmc approach is the observation that in many parametric estimation settings , including in the example considered in this paper , quantities @xmath66 and @xmath67 are highly correlated when the same random input @xmath36 is used to generate them and when @xmath11 and @xmath68 are close .
this suggests using control variates @xmath69 , @xmath70 , when estimating @xmath66 where @xmath13 s are `` close '' to @xmath11 .
while we have identified potentially effective controls , we do not have sufficient information about them , i.e. , @xmath71 $ ] is not known and needs to be evaluated .
this brings us to the second feature of the dbmc method that corresponds to its initial computational information gathering / setup stage .
this stage corresponds to statistical estimation of @xmath72 .
details are given below .
the dbmc approach consists of a setup stage and an estimation stage .
the dbmc setup phase involves generating a `` large '' number of input random vectors @xmath36 and obtaining `` high quality '' estimates of @xmath72 .
let @xmath73 ( @xmath74 `` large '' ) denote a large set of random inputs .
this set represents the _ database_. given the database , the averages of the controls are precisely calculated .
a schematic of this stage is given in figure [ fig : dbmc setup ] . ' '' '' 1 .
for @xmath75 1 . generate @xmath76 according to the distribution of the inputs ; 2 .
for @xmath14 1 . simulate the path @xmath77 2 .
evaluate the value of the control @xmath78 + 2 . for @xmath14 1 .
find @xmath79 , the average of the @xmath80th control on the datebase , as @xmath81 ' '' '' to estimate @xmath82 , at a @xmath11 close to @xmath13 s ( @xmath70 ) select a `` small '' sample ( say of size @xmath83 ) uniformly from the database .
for each sample @xmath76 re - simulate the equation using @xmath76 and @xmath11 to obtain @xmath84 .
for these samples the values of the controls @xmath85 are available in the database . using these
evaluate a controlled estimate of @xmath82 . a schematic version of these steps
is given in figure [ fig : dbmc estimation ] . ' '' '' 1 .
for @xmath86 1 . select @xmath76 uniformly from the database ; 2 . simulate the path @xmath87 ; 3 .
evaluate the estimation variable @xmath84 .
+ 2 . find the controlled estimator of @xmath82 : @xmath88\ ] ] ' '' '' there are two general schemes for implementation of our cv approach : ( i1 ) corresponding to what is described above , requires storing simulation inputs @xmath89 and outputs @xmath90 in a database for later resampling ; ( i2 ) does not utilize resampling , so there is no storage of data beyond recording the calculated control means .
both implementations are feasible , the first is preferable in most cases ; the second may be preferred in some cases .
we elaborate below .
implementation ( i1 ) .
* the database of random inputs , i.e. , @xmath76 s , are either directly stored or enough information about them ( e.g. input seeds of a pseudo - random number generator ) is stored to be able to regenerate @xmath76 s precisely . * the @xmath91 paths corresponding to @xmath13 , @xmath92 , i.e.
, @xmath93 are generally simulated `` in parallel '' as elements of a random vector , @xmath94 , are progressively generated .
* for each random input , say @xmath76 , the value of the controls , @xmath85 , @xmath95 , @xmath96 are stored .
implementation ( i2 ) .
* once the setup stage is completed , the only values stored are the `` high quality '' estimates of the means of @xmath72 s , i.e. , the @xmath91 values @xmath79 , @xmath14 .
* at the estimation phase , @xmath3 random input vectors @xmath76 , @xmath97 , are generated _ anew _ ; paths at @xmath11 and @xmath13 , @xmath70 , are simulated using new random inputs and for each path @xmath85 and @xmath98 are calculated ; finally , using these values , the controlled estimator is evaluated .
the promise of the approach is the following : by anchoring estimation via cv at a few high quality estimates ( at @xmath99 ) , it is possible to obtain high quality estimates at other locations in the parameter space ( at other @xmath11 ) with far fewer samples .
the actual statistical properties of the resulting estimators , and the computational efficiency of generating them , reflect choices made in implementing each given problem .
for example , how much computation should be `` invested '' in the exploration phase , and which points @xmath13 in the parameter space should be explored are two important questions that need further investigation .
such choices generally involve problem dependent tradeoffs , and we leave them to future studies . instead
, the analysis that follows is meant to provide a general and qualitative understanding of the statistical properties , computational efficiency and the tradeoffs involved .
the discussion is as general as possible , but consistent with the numerical study described in section [ numerical ] , where such implementation choices were made utilizing only a basic familiarity with the problem .
for further discussion , see @xcite .
we give the analysis for implementation ( i1 ) .
in other words , assume we are re - sampling from the database .
analysis of implementation ( i2 ) shows similar estimator statistical properties . to simplify the discussion consider a single control , say @xmath100 .
let @xmath101 $ ] , @xmath102 $ ] , @xmath103 , @xmath104 .
assume a database of input variables are generated and let @xmath105 and @xmath106 be random variables corresponding to @xmath0 and @xmath100 that are generated by re - sampling ( uniformly , with replacement ) from the database .
let @xmath107 , @xmath108 , @xmath109 , @xmath110 denote the means and variances of the re - sampled variables @xmath105 and @xmath111 .
conditioned on the database , the controlled estimator is exactly the classical cv estimator and all results from classical cv apply .
for example , for any scalar @xmath48 , @xmath112 is an unbiased estimator of @xmath107 , @xmath109 is known , and the optimal @xmath55 is what is prescribed by classical cv if we take all random variables as those defined on the database .
a measure of variance reduction due to using a controlled estimator is @xmath113 we use the controlled estimator @xmath114 as an estimator for @xmath82 .
assume optimal @xmath55 is used to define @xmath114 and assume @xmath115=j^*(\theta)$ ] .
e.g. @xcite , not to be confused with the resampling bias discussed in this section ] in general @xmath116 . therefore , @xmath114 is a biased estimator of @xmath82 where the bias is introduced by sampling from the database , i.e. , from @xmath105 , as opposed to from @xmath0 .
we have some probabilistic assessment of this bias and we can reduce it by increasing the size of the database .
specifically , for this bias we can obtain an approximate @xmath117 probability confidence interval : @xmath118 where @xmath119 is the @xmath120 quantile from the standard normal distribution .
in other words , with high probability the bias is of the order of @xmath121 .
we assume that for large @xmath74 the bias is sufficiently small to be disregarded and that we can focus on @xmath122 in ( [ eq : vrr_cv * ] ) as the key measure of computational gain in using the controlled estimator to estimate @xmath82 . generating the above large database , as we pointed out earlier ,
corresponds to an initial `` setup '' cost .
let @xmath123 be the computational cost of generating a sample of @xmath66 .
this cost involves generating an @xmath36 , simulating the path , and evaluating @xmath124 .
a reasonable assumption for many problems is that this cost is about the same for all @xmath11 and @xmath36 .
then the set - up cost of generating the database and obtaining averages of the controls is approximately @xmath125 .
let @xmath126 denote the variance reduction ratio at @xmath11 , i.e. , the ratio of the variance of an uncontrolled sample and that of a controlled sample at @xmath11 .
then , the statistical error of a controlled estimator based on @xmath3 samples is approximately the same as that of @xmath127 samples of an uncontrolled estimator .
thus , the ratios of the computational costs of the two estimators ( to arrive at the same statistical accuracy ) is @xmath128 .
therefore , @xmath129 can serve as a measure of benefit of the dbmc approach .
the setup cost of the dbmc approach can be justified in two types of applications .
the first type are those applications that require solving many instances of the estimation problem , at many @xmath11 s .
if the total number of instances is sufficiently large , and some variance reduction is achieved on the average on those instances , then the large fixed set - up cost can be dwarfed by the total computational savings from the many estimations .
the second type are real - time applications where the setup cost can be viewed as an off - line cost enabling significant efficiency gains in the critical task of real - time estimation .
typically , the `` cost '' of delay in such real - time estimation is higher and not merely computational , justifying even a much larger computational effort off - line .
the numerical results in this section are intended to give a qualitative illustration of the efficiency gains that can be achieved using the dbmc approach .
specifically , we estimate the variance reduction that can be achieved over regular ( crude ) sampling , when estimating the three quantities of interest ( a point magnetization , total magnetization at a specific time @xmath21 and the total time - space magnetization ) at a range of the parameter @xmath11 .
our choices of the size of the database , number of samples used for estimation , range of parameter values , and the controls are simply for illustration purposes .
however , we expect that the numerical results are , qualitatively , quite representative .
we simulate the tdgl dynamics on a @xmath130 lattice ( lattice spacing @xmath131 , @xmath132 ) with fixed @xmath133 . on each path , we evolve the system for a total of @xmath134 time steps ( @xmath135 ) which is sufficient for the system to exhibit behavior that is specific to its temperature region . the critical point for this system is @xmath136 @xcite , and our parameter range of interest ( @xmath137 to @xmath138 ) extends to both sides of that critical point . to build a database ,
we simulate @xmath139 paths and evaluate point magnetization , total magnetization at a specific time , and total space - time magnetization at two nominal values of @xmath11 , @xmath140 and @xmath141 . for each quantity of interest , we consider three control variate estimators .
the first two estimators , cv1.2 and cv1.35 , use single controls corresponding to @xmath142 and @xmath143 , respectively .
we chose to anchor our estimators at those two nominal values for @xmath11 because they are located on opposite sides of the phase transition line @xmath144 .
the third estimator , cv2c , uses both controls simultaneously .
we use @xmath145 samples for crude and cv estimators . to estimate the variance of these estimators , following the micro - macro simulation approach ( see , e.g. , @xcite ) , we use @xmath146 independent macro simulations consisting of @xmath147 independent micro simulations .
we obtain variance estimates from each macro simulation and average the resulting @xmath146 values to obtain an overall variance estimate .
we report the ratios of the variance estimates ( crude / controlled , as in ( [ eq : vrr_cv * ] ) ) as @xmath122 . a sampling of vrr results for the total space - time magnetization ( problem p3 ) is given in table [ tab : table1 ] and the corresponding graph is given in figure [ fig : vrr_tim_log ] .
the graph for point magnetization ( problem p1 ) are given in fig . [
fig : vrr_point_log ] , and the results for the total magnetization at a time @xmath21 ( problem p2 ) are quite similar and are excluded .
.[tab : table1]variance reduction ratios of the estimators applied to space - time integral of the magnetization , @xmath148 , at several values of @xmath11 . [ cols="<,^,^,^,^,^,^,^,^,^",options="header " , ] , over a range of values of @xmath11 ( log scale ) .
[ fig : vrr_tim_log ] ] , in log scale .
[ fig : vrr_point_log ] ] based on these results , we draw the following conclusions : * controlled estimators produce dramatic variance reduction for parameter values very close to the nominal parameters and substantial variance reduction at values moderately close to the nominal . * for all the estimation problems , adding the second control consistently improves performance , in some cases leading to substantial reduction in variance ( compared to single controls ) . of course , by incorporating information from points on both sides of the critical temperature
, cv2c is expected to give better coverage than either of the single control estimators .
however , cv2c does better than either of the single control estimators even in their own regions , which suggests that each control provides relevant information to the estimation problem in the opposite region .
* vrr values for the total space - time magnetization are somewhat larger than those for the point and total magnetization at a specific time @xmath21 we expect this to be true more generally for path integrals when compared with values at specific time instances .
in this paper we described a new strategy , database monte carlo ( dbmc ) , for improving computational efficiency of ensemble monte carlo . for a specific time - dependent nonlinear dynamics we showed that the approach can lead to significant efficiency gains for a range of estimation problems .
our selection of the controls has been ad - hoc and for illustration purposes .
further work is required to better understand the options available and the computational tradeoffs involved . to this end
, our current research is focused on ( i ) derivation of more specific guidelines for the selection of effective control variates , ( ii ) implementation of the dbmc strategy in conjunction with other variance reduction techniques , for example , stratification and importance sampling , and ( iii ) application of the method in some specific domains , for example , estimation problems in geophysical fluids and biochemical systems . | in this paper we present a new approach to control variates for improving computational efficiency of ensemble monte carlo .
we present the approach using simulation of paths of a time - dependent nonlinear stochastic equation .
the core idea is to extract information at one or more nominal model parameters and use this information to gain estimation efficiency at neighboring parameters .
this idea is the basis of a general strategy , called database monte carlo ( dbmc ) , for improving efficiency of monte carlo . in this paper
we describe how this strategy can be implemented using the variance reduction technique of control variates ( cv ) .
we show that , once an initial setup cost for extracting information is incurred , this approach can lead to significant gains in computational efficiency .
the initial setup cost is justified in projects that require a large number of estimations or in those that are to be performed under real - time constraints .
monte carlo , variance reduction , control variates s05.10.ln , 02.70.uu , 02.70.tt |
the ultraviolet ( uv ) radiation plays a crucial role in many processes of the interstellar medium ( ism ) , such as photoelectric heating , grain charging , photoionization , and photo - dissociation of molecules @xcite , in which the far - ultraviolet ( fuv : 6ev @xmath11 13.6ev ) radiation from massive stars or interstellar radiation field ( isrf ) influences the structure , chemistry , thermal balance , and evolution of neutral interstellar medium @xcite .
therefore , studying these influence helps to understand the process of star formation .
the fuv emission selectively dissociates co rare isotopologues more effectively than co owing to different levels of self - shielding effects @xcite . for the fuv emission with energy high enough to photodissociate @xmath12co , it rapidly becomes optically thick when it penetrates into molecular clouds .
in contrast , the self - shielding effect of c@xmath1o is relatively less significant because of the shift of its absorption lines and its low abundance .
therefore , the fuv emission with energy above the c@xmath1o dissociation level is expected to penetrate relatively deeper in a molecular cloud , and the abundance ratio of @xmath0co and c@xmath1o , @xmath4 , will increase when the self - shielding effect of c@xmath1o does not yet dominate . when the self - shielding effect of both @xmath0co and c@xmath1o become important , the abundance ratio should decrease toward the intrinsic abundance value which derived from abundances of the elements , @xmath12c , @xmath0c , @xmath13o , and @xmath1c . the intrinsic value may vary as the distance to the galactic center @xcite .
the @xmath4 has been observed in regions with various conditions . in the typical massive star - forming regions , where the ism is filled up with diffuse fuv flux from ob stars , the measurements of @xmath4 are usually higher than the intrinsic value .
@xcite measured @xmath4 in the orion - a giant molecular cloud ( orion - a gmc ) and found that the @xmath4 of most regions are a factor of two greater than 5.5 , the typical value in the solar system .
it is possible that besides the interstellar fuv radiation , embedded ob stars provide strong fuv radiation so that the distance of the penetration will increase .
another possibility is that the higher temperature in the massive cores will also change the fractionation of @xmath12c and @xmath0c @xcite . in the intermediate - mass star - forming regions
, @xcite showed that the intensity ratio of @xmath0co ( @xmath14=21 ) to c@xmath1o ( @xmath14=21 ) , which is equivalent to @xmath4 for optically thin case , rises to a peak up to @xmath240 at @xmath8 5mag then decreases to @xmath24.5 with increasing @xmath7 in the southeast of the california molecular cloud .
although the trend is consistent with the theoretical expectations , the peak occurs at somewhat higher extinction ; @xcite showed that the peak is at @xmath813mag for regions with various density from 10@xmath15 to 10@xmath16@xmath17 . in the low - mass star - forming regions ,
@xcite observed a part of the ic 5146 filament and found that the @xmath4 value is considerably greater than 5.5 in the outer parts ( @xmath18 10mag ) .
ldn 1551 ( hereafter l1551 ) is a relatively isolated nearby star - forming region located at a distance of 160pc @xcite at the south end of the taurus - auriga - perseus molecular cloud complex
. two small clusters of young stellar objects ( ysos ) were detected in l1551 .
one contains two embedded class i sources , l1551 irs 5 ( hereafter irs 5 ) and l1551 ne ( hereafter ne ) , and the other is a group of more evolved ysos ( hereafter hl tau group ) located at the north of irs 5 and ne .
the irs 5/ne cluster co - host a parsec - scale bipolar outflow @xcite which is likely mixed from outflows of each source @xcite .
several herbig - haro objects along the outflows have been identified with different origins @xcite .
another east - west redshifted outflow was found later @xcite , but since irs 5 and ne are two binary systems @xcite the origin of this east - west outflow is still questionable @xcite .
the hl tau group contains hl tau , xz tau , lkh@xmath19 358 , and v1213 tau . the jets and outflows of hl tau group were detected in optical observations @xcite .
in addition to the protostars , a gravitationally bound starless core , l1551 mc , is found at the north - western side of the irs 5/ne cluster which may have started gravitational collapse or could be supported by magnetic fields of @xmath2160@xmath20 g @xcite .
since l1551 dose not contain ob stars , l1551 is a suitable target for studying the fuv influence only from the interstellar radiation . in this paper
, we aim to study the variation of @xmath4 under the influence of the interstellar fuv radiation .
in contrast to previous studies , we observed @xmath12co ( @xmath14=10 ) , @xmath0co ( @xmath14=10 ) , and c@xmath1o ( @xmath14=10 ) lines using the 25-beam array receiver system ( bears ) receiver equipped on the nobeyama radio observatory ( nro ) 45 m telescope to obtain data that could resolve the sub - parsec scale with a complete coverage from the outskirts of the cloud into its dense region .
we also use @xmath6 archival data to obtain a visual extinction map up to @xmath870mag in order to examine the variation of @xmath4 with the fuv attenuation . in [ sec_obs ] , we describe our nro observations and herschel data . in [ sec_result ] , we present the @xmath12co , @xmath0co , and c@xmath1o maps of l1551 , and then derive the excitation temperature and optical depths of the @xmath0co and c@xmath1o lines and the column densities of these molecules . in [ sec_discussion ] , we discuss the dependence of the abundance ratio of @xmath0co to c@xmath1o on visual extinction , @xmath7 , the influence of fuv radiation , and the x - factor in l1551 . at the end , we summarize our results in [ sec_summary ] .
we observed the l1551 star - forming region using the 45 m telescope at nro .
three molecular lines were observed : @xmath12co ( @xmath14=10 ; 115.27120ghz ) , @xmath0co ( @xmath14=10 ; 110.20135ghz ) , and c@xmath1o ( @xmath14=10 ; 109.78218ghz ) .
the @xmath12co data has been published in @xcite .
the observations were carried out between 2007 to 2010 ( table 1 ) .
the telescope has beam sizes ( hpbw ) , @xmath21 , of 15@xmath22 at 115ghz , and of 16@xmath22 at 110ghz .
the main beam efficiencies , @xmath23 , were measured as 32@xmath24 in 20072008 season at 115ghz , 38@xmath24 in 20092010 season at 110ghz , and 43@xmath24 in 2009 season at 110ghz .
these efficiencies were measured every season by nro staff with a superconductor - insulator - superconductor ( sis ) receiver , s100 , and an acousto - optical spectrometers ( aoss ) .
the front end is bears , which has 25 beams configured in a 5-by-5 array with a beam separation of 41@xmath251 @xcite .
bears is operated under double - sideband ( dsb ) mode with each beam connected to a set of 1024 channel auto - correlators ( acs ) as the back end .
we set each set of acs to a bandwidth of 32mhz and a resolution of 37.8khz @xcite , corresponding to velocity resolutions of @xmath20.1kms@xmath26 at the rest frequencies of the three lines .
we calibrated the variations in both beam efficiency and sideband ratio of the 25 beams with the values from the nro website .
these values were measured by nro staff every season through measurements of bright sources with s100 in ssb mode . for the observations , we used the on - the - fly ( otf ) mapping technique @xcite .
the antenna is driven at a constant speed to continuously scan our observing region toward l1551 . to get rid of artificial scanning patterns in the results
, we scanned the observed regions twice , one in ra and one in dec directions , and then combined the otf maps in the two orthogonal directions by plait algorithm @xcite .
the antenna pointing was checked every @xmath26070 minutes by observing the sio maser source nml - tau , and the pointing accuracy was about 3@xmath22 during the observations .
figure [ fig_intro ] ( a ) , ( b ) , and ( c ) show the observed areas of @xmath12co , @xmath0co , and c@xmath1o , covering 44@xmath27 @xmath3 44@xmath27 , 42@xmath27 @xmath3 43@xmath27 , and 30@xmath27 @xmath3 30@xmath27 , respectively .
the data were converted in terms of the main - beam brightness temperature in units of k , @xmath28 , where @xmath29 is the antenna temperature in units of k. in order to maximize the energy concentration ratio , we applied a spheroidal function with @xmath30 and @xmath31 @xcite to convolve the otf data , and then obtained three dimensional @xmath32 final cube data . adopting nyquist sampling for the 45 m telescope , we set the spatial grid size to 7@xmath255 , and the final effective beam sizes , @xmath33 , are 21@xmath258 for @xmath12co and 22@xmath252 for @xmath0co and c@xmath1o .
the rms noise 1@xmath34 levels are 1.23k for @xmath12co , 0.94k for @xmath0co , and 0.67k for c@xmath1o in @xmath35 for a velocity resolution of 0.1kms@xmath26 . to achieve higher signal - to - noise ratios , we further convolved all maps with gaussians to make channel maps , integrated intensity maps , mean velocity maps , and velocity dispersion maps , and the effective beam size of each map is written in its figure caption .
we summarize the parameters for each line in table [ obs_parameter ] .
we used the herschel science archival spire / pacs image data ( observation i d : 1342202250/1342202251 , quality : level 2 processed ) of 160 , 250 , 350 , and 500 @xmath20 m to make dust temperature ( @xmath36 ) and column density maps in a similar way to @xcite . since the 70@xmath20 m emission seems to be not detected except towards irs 5/ne and the hl tau group , we used only 4 bands ( 500 , 350 , 250 , and 160@xmath20 m ) . at first
, we made convolutions for all @xmath6 images ( other than the 500@xmath20 m image ) to smooth their resolutions to the 500@xmath20 m resolution of 36@xmath22 by using the idl package developed by @xcite .
then , we resampled up or down all the images ( other than the 250@xmath20 m ) to the same grid size of the 250@xmath20 m image ( 6@xmath22 ) and derived a spectral energy distribution ( sed ) at each position of the 250@xmath20 m image . here , we adopted an area of @xmath37 , centered at ( ra@xmath38 , dec@xmath38 ) = ( 4@xmath3930@xmath4034@xmath419 , + 18@xmath4228@xmath2755@xmath22 ) , as an area of the zero point of the surface brightness for the l1551 cloud . assuming a single temperature of the dust emission , the gray - body sed fitting was performed with a function of @xmath43 where @xmath44 infers the frequency
, @xmath45 expresses planck s law , @xmath46 is the observed surface brightness , and @xmath47 is the dust optical depth .
@xmath47 can be expressed as @xmath48 , where @xmath49 is the dust opacity per unit mass and @xmath50 is the surface mass density .
we adopted @xmath51 where @xmath52 = 2 , following @xcite .
we carried out the sed fitting at each point where the dust emission was detected above 3 times of the rms noise level , which was measured in the area of the zero point of the surface brightness at all these 4 bands . in the case that the sed fitting failed or the surface brightness was not high enough for sed fitting at some points , we set nan there .
as the data weight of the sed fitting , we adopted @xmath53 , where @xmath54 is the square sum of the rms noise and the calibration uncertainties of surface brightness ( 15@xmath24 at 500 , 350 , and 250@xmath20 m from @xcite ;
20@xmath24 at 160@xmath20 m from @xcite ) .
the dust temperature of each pixel was derived from the above sed fitting , and the result is shown in fig.[fig_intro ] ( f ) .
we used @xmath55 statistic to determine the goodness of the sed fitting per pixel @xcite .
for the sed fitting , each @xmath55 statistic has a @xmath55 distribution with two degrees of freedom , and then the probability q gives a quantitative measure for the goodness . in most of the inner region of the l1551 cloud ,
the sed fitting seems good ( the probability q @xmath2 0.8 ) , while in the region around ysos and toward the outskirts , the fitting seems bad with small probability q. the badness around ysos may be due to multiple - temperature components presenting in the sed . to obtain a column density map with a higher resolution of 18@xmath22 ( fig.[fig_intro ] ( d ) ) , we derived the column density directly from the 250@xmath20 m surface brightness ( @xmath56 from previous fitting ) adopting the dust temperature of the sed fitting with a lower resolution of 36@xmath22 @xcite .
figure [ fig_intro ] ( a ) shows the integrated intensity map of @xmath12co ( @xmath14=10 ) integrated in the @xmath57 range from @xmath5815kms@xmath26 to 20kms@xmath26 for pixels with its signal - to - noise ratio greater than 3 . the @xmath12co emission is distributed all over the observed area .
a sharp edge at the southeast of the mapping area can be recognized .
@xcite suggested that the stellar winds from massive stars , betelgeuse and rigel in the ori - eridanus supershell , possibly compressed the edge of the l1551 molecular cloud .
two elongated structures which are extended in the southwest northeast direction and centered on the class i source irs 5 are identified as the molecular outflows ejected from irs 5 and another embedded class i source ne @xcite .
figure [ fig_intro ] ( b ) shows the integrated intensity map of @xmath0co ( @xmath14=10 ) integrated in the @xmath57 range from @xmath5815kms@xmath26 to 20kms@xmath26 for pixels with its signal - to - noise ratio greater than 3 . the overall distribution of @xmath0co is consistent with those in previous studies of @xmath0co carried out with position switching mode of the fcrao 14 m @xcite and nro 45 m @xcite telescopes .
there are a cavity structure at the northeast of ne and a u - shaped wall structure at the southwest of irs 5 .
in addition , two intensity peaks can be seen toward irs 5 and ne .
these structures are not recognized in the @xmath12co maps .
figure [ fig_intro ] ( c ) shows the integrated intensity map of c@xmath1o ( @xmath14=10 ) integrated in the @xmath57 range from @xmath5815kms@xmath26 to 20kms@xmath26 for pixels with its signal - to - noise ratio greater than 3 .
the c@xmath1o emission line is likely to trace the inner part of the regions traced by @xmath12co and @xmath0co .
the overall distribution of c@xmath1o is consistent with those in previous studies of c@xmath1o carried out with the otf mode of the kitt peak 12 m telescope @xcite and the position switching mode of the nro 45 m @xcite telescope . in order to derive the optical depths and column densities of @xmath0co and c@xmath1o
, we assume that ( 1 ) @xmath12co ( @xmath14=10 ) is optically thick , ( 2 ) @xmath12co and its rarer isotopic species trace the same component , and ( 3 ) these three lines reach local thermal equilibrium ( lte ) .
thus , the temperature of @xmath12co ( @xmath14=10 ) can be treated as the excitation temperature , @xmath59 , of @xmath0co ( @xmath14=10 ) and c@xmath1o ( @xmath14=10 ) for deriving the optical depths and column densities of @xmath0co and c@xmath1o . in order to make direct comparison between the different lines , we smooth all the data so that they have the same effective beam size of 30@xmath254 ( corresponding to 0.023pc at the distance of 160pc ) .
then , we obtain their peak intensities and full widths at half maximum ( fwhm ) by applying gaussian fitting to spectra of the @xmath12co ( @xmath14=10 ) , @xmath0co ( @xmath14=10 ) , and c@xmath1o ( @xmath14=10 ) cube data for pixels with their signal - to - noise ratio greater than 5 .
in general , the spectra of @xmath0co ( @xmath14=10 ) and c@xmath1o ( @xmath14=10 ) show single - component velocity structures , and thus we can apply single - gaussian fitting as an appropriate method .
however , since the spectra of @xmath12co ( @xmath14=10 ) often show multiple velocity components due to the prominent outflows , we adopt different fitting strategies to obtain the excitation temperature .
first , if the @xmath12co spectrum has only one velocity component , which usually happen in quiescent ambient region , the single - gaussian fitting may gives an adequate result .
if the residual of the fitting at the peak velocity is less than 3@xmath34 , we take the peak intensity to calculate the excitation temperature .
second , if the single - gaussian fitting fails , we apply double - gaussian fitting to separate the ambient and outflow components . except the regions around ne , only one redshifted or blueshifted component appears .
we identify the component of which the peak velocity is similar to the peak velocities of @xmath0co and c@xmath1o spectra as the ambient component and the other component as the outflow component , if ( 1 ) the residuals of the fitting at both the peak velocities are less than 3@xmath34 and ( 2 ) the velocity ranges of the two gaussian components at half maximum are not overlapped with each other . the later criterion is included , because the determination of the peak intensity of the ambient component could be affected by the close outflow component . in some cases , the peak @xmath12co intensity is even smaller than the peak @xmath0co intensity , which is not physical and is rejected .
third , for the remaining spectra , the velocity structure seems to contain three or more components .
since our goal is to obtain the peak intensity of the ambient gas , we only apply a single gaussian to one peak having the peak velocity closest to the ambient gas velocity determined from the @xmath0co and c@xmath1o spectra at the same position .
we fix the peak velocity of this single gaussian as the mean value of the @xmath0co and c@xmath1o peak velocities and we only fit the gaussian within a narrow velocity range bracketed by the two local minima around the peak . following the above procedure
, we obtain the estimate of the excitation temperature from the peak intensity of the fitted gaussian to the ambient @xmath12co emission .
figure [ fig_intro ] ( e ) shows the excitation temperature map deriving from the following equation with the previous assumption ( 1 ) and the beam filling factor of 1 ( e.g. , * ? ? ?
* ; * ? ? ?
* ) @xmath60^{-1}\rm\,k\\ & = 5.53 \left[\log\left(1+\frac{5.53}{t_{\rm mb,^{12}co}+0.818}\right)\right]^{-1 } \rm\,k , \end{split}\ ] ] where @xmath61 is the peak intensity of @xmath12co ( @xmath14=10 ) in units of k from the above fitting , @xmath62 is the effective radiation temperature @xcite , and @xmath63=2.7k is the temperature of cosmic microwave background radiation .
hereafter , we call this excitation temperature derived from eq.([equ_tex ] ) , @xmath64 .
we then obtain the optical depths , @xmath65 , of the @xmath0co ( @xmath14=10 ) and c@xmath1o ( @xmath14=10 ) emission and the column densities , @xmath66 , of the @xmath0co and c@xmath1o gas using the following equations ( e.g. , * ? ? ?
* ; * ? ? ?
* ) @xmath67^{-1}-0.164)}\right),\ ] ] @xmath68^{-1}-0.1666)}\right),\ ] ] @xmath69 and @xmath70 where @xmath71 and @xmath72 are the peak intensities in k , @xmath73 and @xmath74 are the fwhms in kms@xmath26 , and @xmath75 and @xmath76 are the beam filling factors .
the beam filling factors can be expressed as @xmath77 , where @xmath78 and @xmath79 are the source size and the effective beam size , respectively . in molecular clouds ,
the c@xmath1o emission is usually considered to trace the dense cores , clumps , and/or filaments . on the other hand , the @xmath0co emission traces more extended regions than c@xmath1o . in the case of the taurus molecular cloud
, @xcite and @xcite found that both the typical @xmath0co and c@xmath1o core size are @xmath20.1pc .
since the effective beam sizes we used here are 30@xmath254 ( corresponding to 0.023pc at the distance of 160pc ) which is much smaller than the typical core size , we can assume the emission of our sources fill up the beam and the beam filling factors become @xmath20.95 .
therefore , we simply adopt @xmath75 and @xmath76 as 1 .
figures [ fig_tau_n ] ( a ) and [ fig_tau_n ] ( b ) show the optical depth maps of @xmath0co ( @xmath14=10 ) and c@xmath1o ( @xmath14=10 ) , respectively . for the @xmath0co ( @xmath14=10 ) emission , its optical depth is more than 1.5 at the center of l1551 , and drops to less than 1 at the outer edge . for the c@xmath1o ( @xmath14=10 )
emission , its optical depth is less than 0.8 in the whole region , suggesting that the c@xmath1o ( @xmath14=10 ) emission is fully optically thin in l1551 .
figures [ fig_tau_n ] ( c ) and [ fig_tau_n ] ( d ) show the column density maps of @xmath0co and c@xmath1o , respectively .
the cavity structure and the u - shaped wall structure can be seen in the @xmath0co column density map .
the c@xmath1o column density map shows that c@xmath1o concentrate in the region surrounded by irs 5 , ne , and l1551 mc . moreover , the u - shaped wall structure is also seen in the c@xmath1o column density map as a low column density part .
we can derive the abundances from @xmath80 and @xmath81 with the h@xmath82 column density derived from the fir dust continuum images . because we adopt the dust temperature map with an effective beam size of 36@xmath22 ( see [ sec_obs_her ] ) , we convolve the @xmath83 and @xmath84 maps with gaussians to degrade their resolutions down to 36@xmath22 ( corresponding to 0.027pc at the distance of 160pc ) .
figures [ fig_xx ] ( a ) and [ fig_xx ] ( b ) show the abundance maps of @xmath0co and c@xmath1o , respectively .
the mean and standard deviation of @xmath85 and @xmath86 are ( [email protected])@xmath310@xmath87 and ( [email protected])@xmath310@xmath88 , respectively .
we can see @xmath89 peaks toward irs 5 , ne , and l1551 mc in fig.[fig_intro ] ( d ) and @xmath84 also concentrates in the above regions in fig.[fig_tau_n ] ( d ) but the peaks of @xmath84 are not as obvious as @xmath89 . as a result , the @xmath86 map shows low values in above three regions , which means the c@xmath1o depletion occurs at these regions .
it is reasonable for the depletion at l1551 mc because we find that l1551 mc is dense and cold in figures [ fig_intro ] ( d ) and ( f ) . for irs 5 and ne ,
the dust temperature at the envelop of protostars are higher than other regions , but the dust temperature is still less than the co evaporation temperature of 2025k and the h@xmath82 column density is very high compared with other regions ; therefore the envelop of protostars also show c@xmath1o depletion . on the other hand , the low @xmath85 value also appears at the c@xmath1o depletion region ; however , the low @xmath85 value derived in this region may not show the real situation because @xmath0co and dust emission may not trace the exactly same layer in the line of sight ( see [ temperature ] ) .
the abundance ratio of @xmath0co to c@xmath1o can be directly derived from @xmath4 = @xmath90 .
figure [ fig_xx ] ( d ) shows the spatial variation of @xmath4 .
the @xmath4 value ranges from @xmath23 to @xmath227 in the whole region , and the mean and standard deviation of the abundance ratios are [email protected] .
spatially , the @xmath4 value in the outskirts of the cloud is more than @xmath28 and drops to less than 5.5 around l1551 mc except for the hl tau group and the southwest blueshifted outflow lobe with the brightest @xmath12co integrated intensity .
to study the influence of interstellar fuv radiation on @xmath4 , we remove the outflow regions to avoid the influence from outflow activities for later discussion .
we define the outflow regions by the regions where the @xmath12co integrated intensities in either blueshifted or redshifted parts are higher than 3@xmath34 ( see fig.[fig_xx ] ( f ) ) ; here we regard @xmath5815 to 5.5kms@xmath26 and 7.9 to 20kms@xmath26 as the blueshifted and redshifted outflow velocity ranges , respectively
. moreover , to avoid the influence from underestimation of @xmath86 by c@xmath1o depletion , we also remove the region where the @xmath86 value is less than the ism standard value of 1.6@xmath310@xmath88 @xcite ( shown as the gray contour in fig .
[ fig_xx ] ( b ) ) . in the other words ,
we remove the region where c@xmath1o depletion factor = @xmath91 1 .
figure [ fig_xx ] ( e ) shows the abundance map excluding the outflow and c@xmath1o depletion regions .
we can see that the spatial variation of the @xmath4 value is outside - in deceasing and the minimum of the @xmath4 value coincides with the dense part in l1551 ( see the h@xmath82 column density map in fig.[fig_intro ] ( d ) as a comparison ) .
this spatial variation might be caused by the penetration capability of interstellar fuv radiation which is related to the visual extinction , @xmath7 . in order to investigate the dependence of the abundance ratio on the visual extinction within the molecular cloud
, we derived the @xmath7 map from the h@xmath82 column density map with a relation from @xcite , @xmath92 figure [ fig_xx ] ( c ) shows the derived @xmath7 map . since we adopt the dust temperature map with an effective beam size of 36@xmath22 ( see [ sec_obs_her ] ) , we convolve the abundance ratio maps with gaussians to degrade their resolutions down to 36@xmath22 ( corresponding to 0.027pc at the distance of 160pc ) .
figure [ fig_av_xx ] ( a ) shows the correlation between the @xmath4 value and the @xmath7 value . the @xmath4 value is higher than the typical solar system value of 5.5 in the @xmath7 range less than 10mag , and then decreases down to @xmath25.5 with increasing @xmath7 value , suggesting the dependence of the abundance ratio on @xmath7 value is significant .
although the sample of @xmath18 2mag is sparse because we derived the abundance ratio from the detections with both @xmath0co ( @xmath14=10 ) and c@xmath1o ( @xmath14=10 ) more than 5@xmath34 , we still can see that the maximum of @xmath4 value occurs between @xmath8 14mag . to figure out the averaged trend of the @xmath4 value
, we averaged @xmath4 value in 2.5mag @xmath7 bins .
the averaged @xmath4 value reaches the maximum of 10.3 which is @xmath22 times more than 5.5 in the @xmath7 range less than 2.5mag .
theoretical studies predict that in the mean isrf , the @xmath4 maximum occurs at @xmath8 13mag and reaches @xmath2510 times of the intrinsic abundance ratio , and then decreases toward the intrinsic abundance ratio when @xmath7 further increases @xcite ; that is , the selective photodissociation of c@xmath1o appears at @xmath8 13mag .
our maximum of the averaged @xmath4 value is only @xmath22 times of 5.5 , and our maximum of the non - averaged @xmath4 value just barely reaches 27.7 which is @xmath25 times of 5.5 . since the model calculations in @xcite only considered non - turbulent uniform density model , our lower maximum @xmath4 value may be caused by non - uniform density distributions .
another example is shown by @xcite who derived a similar correlation in a filament of ic 5146 ( see fig.19 in * ? ? ?
* ) , which is also a low mass star - forming region .
they found that the @xmath4 ratio at @xmath93 10mag is @xmath23 times of 5.5 and then decreases to @xmath25.5 at @xmath94 10mag .
note that the @xmath4 peak positions in the @xmath7 axes in l1551 and ic 5146 are different .
although @xcite used the near - infrared color excess ( nice ) technique to derive @xmath7 and we used the gray - body sed fitting and equ.[equ_av ] to derive @xmath7 , both the nice method and our method are based on the assumption that @xmath95 .
thus these difference may be owing to that the difference beam size between our data or there are still some different physical condition between them . as seen in fig.[fig_intro ] ( c ) , the c@xmath1o emission is rather weak or not detected at the periphery of the cloud , in which higher @xmath4 value is expected . on the other hand ,
the @xmath0co emission is strong at these regions .
since we only calculated @xmath4 value as the signal - to - noise ratio of both @xmath0co and c@xmath1o greater than 5 ( see [ result_col_abu ] ) , we may only sample the positions with high c@xmath1o abundance . in order to check this possibility
, we can estimate the upper limit of @xmath4 value by using the c@xmath1o emission with the signal - to - noise ratio of 3 and assuming fwhm@xmath96 as the value at the outskirts in fig.[fig_intro ] ( c ) ( @xmath20.18 km s@xmath26 ) at the region where the c@xmath1o emission is weak ( @xmath97 5@xmath34 level noise ) but the @xmath0co emission is still strong ( @xmath98 5@xmath34 level noise ) .
then @xmath4 value can reach @xmath245 at these regions , which is @xmath28 times of 5.5 . in order to derive the optical depths and column densities
, we assumed that the @xmath12co , @xmath0co , and c@xmath1o lines trace the same region , and derived the excitation temperature , @xmath64 , using the peak intensity of @xmath12co ( @xmath14=10 ) .
however , the three lines may trace different regions , and @xmath64 could not be used as the excitation temperatures of @xmath0co ( @xmath14=10 ) and c@xmath1o ( @xmath14=10 ) . to investigate the influence of the excitation temperature , we re - estimate the abundance ratio by using the dust temperature ( shown in fig.[fig_intro ] ( f ) ) , @xmath36 , as the excitation temperature of @xmath0co or / and c@xmath1o . since the thermal
emission from dust is usually optically thin and can traces inner regions , the @xmath0co or / and c@xmath1o may trace the same region as dust .
we found that @xmath36 is generally smaller than @xmath64 in the inner region of l1551 and @xmath36 in the area we calculated abundance ratio has a mean value of [email protected] ; thus , there is a possibility that the @xmath12co emission and the dust emission may trace different regions in the inner region of l1551 . the h@xmath82 column density traced by
the dust emission is also a very useful measure to examine whether the co lines trace the same region as the dust or not .
in fact , we can see the distributions of the @xmath89 map and the @xmath84 map are similar , compared to the @xmath83 map .
thus , we assume the excitation temperature of c@xmath1o is the dust temperature , but we try both @xmath64 and @xmath36 for @xmath0co for comparison . on the other hand , since @xcite used the two rotational transitions of @xmath12co ( @xmath14=10 and 32 ) to derive a mean excitation temperature of the main ne / irs 5 outflow to be 16.5k , which is generally higher than @xmath64 , we also use this temperature for comparison . therefore , we compare the derived abundance ratios under the following four assumptions : 1
. @xmath99 = @xmath64 and @xmath100 = @xmath64 .
( original)[tex_1 ] 2 .
@xmath99 = @xmath64 and @xmath100 = @xmath36.[tex_2 ] 3 .
@xmath99 = @xmath36 and @xmath100 = @xmath36.[tex_3 ] 4 .
@xmath101.[tex_4 ] each panel in fig.[fig_av_xx ] shows the correlation between @xmath4 and @xmath7 under one of the above four assumptions .
the mean and standard deviation of the abundance ratios are [email protected] , [email protected] , [email protected] , and [email protected] under these four assumptions respectively .
the dependence of the @xmath4 value on the @xmath7 value under the assumption ( [ tex_2 ] ) and ( [ tex_4 ] ) are consistent with that under the original assumption ( [ tex_1 ] ) .
we notice that the @xmath4 value at @xmath102 5 under the assumption ( [ tex_4 ] ) is sightly lower than that in the assumption ( [ tex_1 ] ) , which reveals that a higher excitation temperature will derive lower abundance ratio .
actually , @xmath64 used in fig.[fig_av_xx ] ( a ) has a mean value of [email protected] .
even if we use 16.5k as the excitation temperature in the assumption ( [ tex_4 ] ) , we still obtain almost the same trend in the correlation between @xmath4 and @xmath7 .
thus , this derivation of abundance ratio is not sensitive to the excitation temperature . in the case of the assumption ( [ tex_3 ] ) ,
the trend of @xmath4 is similar to those under the other assumptions in the @xmath7 range less than 10mag , but the @xmath4 value increases in the @xmath7 range more than 10mag .
in fact , the ambient region in which the @xmath7 value larger than 10mag is corresponding to l1551 mc in which the dust temperature is less than @xmath210k .
since the only difference between the assumption ( [ tex_2 ] ) and ( [ tex_3 ] ) is the excitation temperature of @xmath0co , this behavior can be interpreted as follows : the excitation temperature of @xmath0co in l1551 mc is underestimated in the case of @xmath59 = @xmath36 , and thus the column density of @xmath0co is overestimated .
this suggests that the @xmath0co line does not trace the dense region of l1551 mc traced by the dust emission : l1551 mc can not be recognized in the @xmath0co integrated intensity map of fig .
[ fig_intro ] ( b ) .
consequently , even when we take into account the uncertainties in the excitation temperature estimation , our results still lead to the same conclusion .
that is , the @xmath4 value reaches @xmath211 at the low @xmath7 value ( @xmath8 14mag ) which is indeed larger than the solar system value , and decreases toward the solar system value in the high @xmath7 value ( @xmath8 520mag ) except the case of the low ( @xmath210 k ) excitation temperature for the @xmath0co line .
orion - a gmc is located at a distance of 400pc @xcite and is under strong influence of the fuv emission from the trapezium cluster and nu ori @xcite .
several photon dissociation regions ( pdrs ) have been identified from comparison of the distributions of the 8@xmath20 m , 1.1 mm , and @xmath12co ( @xmath14=10 ) line emission @xcite .
@xcite found that the abundance ratios of @xmath0co to c@xmath1o in pdrs and non - pdrs are both higher than 5.5 , and concluded that these high abundance ratios are due to the selective fuv photodissociation of c@xmath1o . in order to investigate the influence of the different environments on the abundance ratio of @xmath0co to c@xmath1o
, we compare the abundance ratio between l1551 and orion - a .
figure [ fig_n18_xx ] shows the @xmath4 value as a function of @xmath84 for l1551 and orion - a ( both the pdr and non - pdr regions ) . the data of orion - a are from @xcite . here
, @xcite used the c@xmath1o column density to estimate the total column density . although the c@xmath1o column density and the @xmath7 value do not have a perfect linear relation because of the selective photodissociation @xcite
, we can still consider that the higher column density is generally corresponding to the higher @xmath7 value and vice versa .
our results show that the abundance ratio in l1551 is generally lower than that in orion - a .
the mean @xmath4 value in l1551 is [email protected] in contrast to [email protected] in pdrs and [email protected] in non - pdrs in orion - a @xcite . for the non - pdrs of orion - a
, the @xmath4 value has a maximum in the low @xmath84 regime ( @xmath103 @xmath104 @xmath105 ) , and then decreases to @xmath210 in the high @xmath84 regime ( see fig.7 ( d ) , ( e ) , ( f ) , and ( g ) in * ? ? ?
this trend is similar to the result of l1551 but the selective photodissociation of c@xmath1o in l1551 only occurs at the outskirts ( @xmath8 14mag , see fig.[fig_av_xx ] ( a ) ) .
however , even at the high column density , the abundance ratio is still about two times higher than 5.5 in orion - a , because of the presence of the embedded ob stars in the cloud . in the other words , the fuv strength in orion - a is higher than that in l1551 , which is the environmental difference between l1551 and orion - a . in the orion - a
, the fuv radiation from the embedded ob stars is likely to penetrate the whole cloud . for the pdrs of orion - a
, the stronger fuv radiation causes the even higher convergence value of @xmath215 .
measurements of the mass distribution in molecular clouds help to understand their physical and chemical characteristics .
however , in ism , the most abundant molecular species , h@xmath82 , is hard to be observed because h@xmath82 lacks its electric - dipole moment and its quadruple transition is hard to occur in the typical molecular cloud environment .
the secondary abundant molecular species , co , is not the case , because co is much easier to be observed and the @xmath12co ( @xmath14=10 ) emission is considered as the most available mass tracer .
thus , the co - to - h@xmath82 conversion factor , also called x - factor " , is defined as , @xmath106 where @xmath107 is the integrated intensity of the @xmath12co ( @xmath14=10 ) emission .
@xcite showed that the averaged x - factor = 2@xmath310@xmath108@xmath105k@xmath26km@xmath26s with @xmath530@xmath24 uncertainty in the milky way disk .
nevertheless , the x - factor can vary by a factor of @xmath2 100 in different regions , because @xmath12co ( @xmath14=10 ) is often optically thick @xcite .
for example , @xcite measured x - factor @xmath2 ( 1.612)@xmath310@xmath108@xmath105k@xmath26km@xmath26s in the taurus molecular cloud , @xcite measured x - factor @xmath2 3@xmath310@xmath109@xmath105k@xmath26km@xmath26s in the perseus molecular cloud , and @xcite measured x - factor @xmath2 2.53@xmath310@xmath108@xmath105k@xmath26km@xmath26s in the southeastern part of the california molecular cloud .
these discrepancies are also found in ism numerical simulations which show that x - factor is likely dependent on extinction , volume density , temperature , metallicity , turbulence , star formation feedback , and so on @xcite . even within a molecular cloud
, the x - factor can still vary with a wide range . on the other hand ,
the variation of the x - factor of the @xmath0co and c@xmath1o can be smaller ( see appendix [ appendix_x ] ) .
hereafter we calculate the x - factor for pixels in fig.[fig_intro ] ( a ) with the signal - to - noise ratio of the @xmath12co integrated intensity greater than 3 .
figure [ fig_av_w12 ] shows the correlation between the @xmath107 value and the @xmath7 value , and the points are color - coded by @xmath110 .
we can see that there are multiple trends .
the averaged x - factor of the whole region is 1.08@xmath11110@xmath108@xmath105k@xmath26km@xmath26s .
this is about a factor of two smaller than the average x - factor in milky way , but the dispersion still covers the average x - factor in milky way . in order to reveal these different trends , although we have already defined two regions , the ambient and outflow components , in the previous discussion , we divide into more regions based on the previous division here : ( a ) the diffuse component which is the ambient component excluding the l1551 mc component , ( b ) the l1551 mc component which is the green polygon in fig.[fig_xx ] ( f ) , and ( c ) the outflow component which is the same as defined in sec . 4.1 . in each panel
, we use the chi - squre fitting method with errors in both coordinates to find the best fit line .
the diffuse component ( fig.[fig_av_w12 ] ( a ) ) shows a relatively narrow distribution , compared with the other two regions , and the fitted x - factor = 1.24@xmath11210@xmath108@xmath105k@xmath26km@xmath26s is similar to the fitted x - factor of the whole region and is the same order of magnitude as the average x - factor in milky way . the l1551 mc component ( fig.[fig_av_w12 ] ( b ) ) , however , shows a more extended distribution and a larger fitted x - factor = 5.59@xmath11310@xmath108@xmath105k@xmath26km@xmath26s .
this is due to the fact that the @xmath12co emission is optically thick and thus saturated in this dense starless core .
the outflow component ( fig.[fig_av_w12 ] ( c ) ) has a fitted x - factor = 7.43@xmath11410@xmath109@xmath105k@xmath26km@xmath26s , which is smaller than that of the diffuse component , but shows three different distributions .
two of the distributions are extended and have shallow slopes in the correlation , which belong to the outflow region overlapped with irs 5 and ne shown in fig.[fig_xx ] ( f ) as two green incomplete circles and marked in the panel ( c ) of fig.[fig_av_w12 ] .
the @xmath12co emission from the irs 5 and ne region is likely saturated because the h@xmath82 column densities of irs 5 and ne are comparable to that of l1551 mc ( @xmath115@xmath105 ) .
the other narrow distribution belongs to the remaining outflows and dominates the fitting of x - factor , which may be due to the fact that the entrained energy more strongly excites the @xmath12co emission and makes its high excitation temperature .
consequently , in the low @xmath7 range ( @xmath116mag ) , the x - factor of the diffuse component is consistent with the milky way average value , and the x - factor of the outflow component is smaller .
however , in l1551 mc ( @xmath117mag ) , the x - factor becomes larger and has a large dispersion .
figure [ fig_av_x ] shows the correlation between the x - factor and the @xmath7 value , and the points are color - coded by @xmath59 .
we can see that the x - factor of the whole region is dependent on @xmath7 with a fitted power law , x - factor @xmath9 @xmath118 . especially , the l1551 mc component has a well - correlated distribution ( less dispersion of the power index ) , x - factor @xmath9 @xmath119 . a similar distribution but a shallower power law , x - factor @xmath9 @xmath120 , was also found by @xcite in the southeastern part of the california molecular cloud excluding the hot region around the massive star lkh@xmath19 101 .
however , in perseus , @xcite found two characteristic features : the x - factor steeply decreases at @xmath121mag and gradually increase at @xmath117mag .
the steeply decreasing trend is likely due to the sharp transition from cii / ci to co , but this steeply decreasing trend does not appear in our results .
the ism numerical simulations performed by @xcite suggests that the transition behavior is dependent on the volume density .
they simulated molecular clouds with uniform density distribution from 10@xmath15 to 10@xmath122@xmath17 and initial turbulence , and showed that the low density case has the decreasing trend with a wider @xmath7 range compared to the high density case .
for the highest density case in their simulation ( n @xmath2 10@xmath122@xmath17 ) , the result only shows one characteristic trend , x - factor @xmath9 @xmath89 , which means the @xmath12co emission is saturated ( optically thick ) to be constant and then the x - factor directly relates to @xmath89 , which is consistent with our results . in l1551 , we do not find the decreasing trend even at the low @xmath7 range , which indicates the fuv strength in l1551 is less than that in perseus . under the mean isrf
, @xmath12co is self - shielded at @xmath123 0.5mag @xcite , which may be the case in l1551 .
figure [ fig_tex_x ] shows the correlation between the x - factor and the @xmath59 value , and the points are color - coded by @xmath7 .
our data do not show obvious correlated distributions .
however , an inverse power law , x - factor = @xmath124 , was found in the southeastern part of california cloud @xcite , which is consistent with x - factor @xmath9 @xmath125 in simulations of @xcite .
although this relation passes through the diffuse component of l1551 , we can not confirm this relationship due to the narrow temperature range of our data .
figure [ fig_mom2_x ] shows the correlation between the x - factor and the @xmath12co velocity dispersion , @xmath126 , and the points are color - coded by @xmath59 .
the simulations predict that x - factor @xmath9 @xmath127 @xcite .
however , our data do not have a correlated distribution .
@xcite also found no correlation in the southeastern part of california cloud .
we have carried out wide - field otf observations in the @xmath12co , @xmath0co , and c@xmath1o ( @xmath14=10 ) lines toward the l1551 molecular cloud using the bears receiver of the nro 45 m telescope .
the main results are summarized as follows : 1 .
the abundance ratio of @xmath0co to c@xmath1o ( @xmath4 = @xmath90 ) is derived to be @xmath2327 . the mean value is [email protected] .
we can see a spatially decreasing trend in the @xmath4 map from the outskirts to the dense part of l1551 , the starless core l1551 mc .
we found that the @xmath4 value reaches its maximum in @xmath8 14mag , and then decreases down to about the typical solar system value of 5.5 in the high @xmath7 value , suggesting that the selective photodissociation of c@xmath1o from the interstellar fuv radiation occurs in l1551 .
pdr theoretical models with the mean isfr suggest that the selective photodissociation of c@xmath1o causes the increase of @xmath4 value up to @xmath2510 times of the intrinsic value in the low @xmath7 regions with @xmath2 13mag .
however , in l1551 , the average @xmath4 value in 2.5mag bins reaches its maximum value of 10.3 which is @xmath22 times of 5.5 in the @xmath7 range less than 2.5mag .
since the model calculations consider non - turbulent uniform cloud density , this lower maximum @xmath4 value may be caused by non - uniform cloud density .
the mean of the excitation temperatures of @xmath0co and c@xmath1o derived from @xmath12co in the non - outflow regions is [email protected] .
even if we consider the higher excitation temperature of @xmath128 16.5k or assume @xmath129 , we also find a similar trends of @xmath4 .
therefore , our results described in item 2 are not sensitive to the uncertainties in the excitation temperature determination .
4 . the trends in the @xmath4 variations in l1551 and orion - a are similar to each other . that is , both
@xmath4 values reach their maximums at low @xmath84 value , and then decrease as @xmath84 value increase ; however , the @xmath4 value in orion - a converges to a higher value than that in l1551 .
this is due to the fuv radiation from the embedded ob stars .
we calculated that the averaged x - factor = @xmath130 in l1551 is 1.08@xmath11110@xmath108@xmath105k@xmath26km@xmath26s which is somewhat smaller but consistent with the milky way average value @xmath2 2@xmath310@xmath108@xmath105k@xmath26km@xmath26s .
the x - factor of the different regions in l1551 has a large variation . for the diffuse region
, we found a x - factor = 1.24@xmath11210@xmath108@xmath105k@xmath26km@xmath26s which is similar to the milky way average value .
for the outflow region , we found a smaller x - factor = 7.43@xmath11410@xmath109@xmath105k@xmath26km@xmath26s .
the region of l1551 mc has a larger x - factor = 5.59@xmath11310@xmath108@xmath105k@xmath26km@xmath26s with a large dispersion because of the saturation of the @xmath12co emission . for the whole region
, we found a correlation between the x - factor and the @xmath7 value with x - factor @xmath9 @xmath118 at @xmath7 = @xmath20.270mag which is due to the saturation of the @xmath12co emission .
the 45 m radio telescope is operated by nobeyama radio observatory , a branch of national astronomical observatory of japan .
part of this work was achieved using the grant of naoj visiting fellow program supported by the research coordination committee , national astronomical observatory of japan ( naoj ) .
s.j.l . and s.p.l .
acknowledge support from the ministry of science and technology of taiwan with grants most 102 - 2119-m-007 - 004-my3 .
figure [ fig_ch_12co ] shows the velocity channel maps of @xmath12co ( @xmath14=10 ) . in the velocity range from @xmath588kms@xmath26 to 6kms@xmath26
, an extended elongated emission can be recognized at the southwest of irs 5 .
this component is identified as the blueshifted lobe of the outflows ejected from irs 5 and ne @xcite . in the velocity range from 4kms@xmath26 to 10kms@xmath26
, a narrower elongated structure appears at the northeast of irs 5 and through ne , which could still be a mix of the outflows from both sources . in the velocity range from 2kms@xmath26 to 6kms@xmath26 ,
a clam - shaped emission shows up around the hl tau group , which are identified as the collective outflows from the hl tau group @xcite .
this outflow structure extends to @xmath214kms@xmath26 in the redshifted lobe . in the velocity range from 10kms@xmath26 to 20kms@xmath26
, there is an emission between ne and the hl tau group .
this component is located between the redshifted outflows from irs 5 and ne and the collective outflows from the hl tau group @xcite , which could be the results of the outflow interaction . in the velocity range from 8kms@xmath26 to 16kms@xmath26 ,
another narrow elongated emission is distributed in the east west direction pointing back to irs 5 and ne , but the origin of this east west outflow is still unclear @xcite . at 10kms@xmath26 ,
a diffused emission with an intensity of @xmath224k is distributed at the southeast of the l1551 molecular cloud , and this is a part of the ambient gas in taurus @xcite . at 8kms@xmath26 , two individual small cores with an extent of @xmath235@xmath27 at the northern side probably do not associate with the l1551 molecular cloud because their velocity of @xmath28 km s@xmath26 is different from the systemic velocity of @xmath26 km s@xmath26 in l1551 ( see the 6 km s@xmath26 panel in fig.[fig_ch_12co ] )
. figures [ fig_mom ] ( a ) and ( d ) show @xmath12co mean velocity and velocity dispersion maps , respectively .
the mean velocity over whole region is [email protected]@xmath26 @xcite .
we calculate that the mean velocity dispersion toward all observed area is [email protected]@xmath26 .
note that the velocity dispersion at the area of the outflows ( @xmath20.54.0kms@xmath26 ) is relatively higher than the mean value toward the overall observed area .
figure [ fig_ch_13co ] shows the velocity channel maps of @xmath0co ( @xmath14=10 ) . in the velocity range from 4.25kms@xmath26 to 7.25kms@xmath26 , we can see the u - shaped wall clearly at the southwest of irs 5 which is distributed around the southwestern blueshifted lobe of the outflows traced in @xmath12co and extending to the outside of those outflows .
note that although the u - shaped wall seems to surround the blueshifted outflows , its velocity covers both redshifted and blueshifted ranges , which may hint that the blueshifted outflow axis is almost parallel to the plane of sky . in the velocity range from 6.25kms@xmath26 to 7.25kms@xmath26
, we can see the cavity at the northeast of ne . at the velocity of 6.25kms@xmath26
, we can see a narrow filamentary structure sticking out from irs 5 and extending toward l1551 mc and beyond . at the velocity of 6.75kms@xmath26
, the @xmath0co emission is distributed over the whole region , which is consistent to the centroid velocity of [email protected]@xmath26 measured by @xcite . at the velocity of 7.75kms@xmath26 , we can barely see an east west elongated structure corresponding to the east west outflow seen in the @xmath12co map .
figure [ fig_mom ] ( b ) shows the @xmath0co mean velocity map .
the cavity is clearly seen in the blueshifted velocities with respect to the mean velocity of 6.7 km s@xmath26 . because the cavity is on top of the redshifted outflows from irs 5/ne , we speculate that the gas on the cavity region is pushed toward us by the redshifted outflows .
there is a redshifted streamer ( @xmath131 , see the black dashed line in fig.[fig_mom ] ( b ) ) at the southwest of the narrow filamentary structure ( see the grey dashed line in fig.[fig_mom ] ( b ) ) and at the north of east west outflow .
however , the integrated intensity of this streamer is very low in fig.[fig_intro ] ( b ) , an thus this component is not discussed hereafter .
figure [ fig_mom ] ( e ) shows the @xmath0co velocity dispersion map .
the velocity dispersion of the u - shaped wall ( @xmath20.40.7kms@xmath26 ) is relatively higher than the mean value in the overall observed area ( [email protected]@xmath26 ) .
figure [ fig_ch_c18o ] shows the velocity channel maps of c@xmath1o ( @xmath14=10 ) .
these maps show a filamentary structure in the northwest southeast direction in the velocity range from 6.25kms@xmath26 to 7.25 km , and l1551 mc traced by nh@xmath132 coincides with this filament @xcite .
this filamentary structure can also be recognized in the @xmath0co channel maps ( the 6.25kms@xmath26 panel of fig.[fig_ch_13co ] ) and the h@xmath82 column density map ( fig.[fig_intro ] ( d ) ) .
a u - shaped wall structure is present in the velocity range from 6.75kms@xmath26 to 7.05 km , which coincides with the redshifted part of the u - shaped wall traced in @xmath0co .
another fainter filamentary structure can be seen in the velocity range from 7.05kms@xmath26 to 7.35kms@xmath26 , and its position is just above the upper arm of the u - shaped wall .
figure [ fig_mom ] ( c ) and ( f ) are the c@xmath1o mean velocity and the velocity dispersion maps , respectively . within the u - shaped wall ,
the integrated intensity becomes low and the gas is mostly blueshifted ( see fig.[fig_intro ] ( c ) ) .
figures [ fig_av_w13 ] and [ fig_av_w18 ] , respectively , show the correlation between the @xmath133 value , the integrated intensity of the @xmath0co ( @xmath14=10 ) emission , and the @xmath7 value , and the @xmath134 value , the integrated intensity of the c@xmath1o ( @xmath14=10 ) emission , and the @xmath7 value , in which the points are color - coded by @xmath110 . in order to make direct comparison with their column density
, we only plot the same regions where the pixels have their signal - to - noise ratio greater than 5 ( see [ result_col_abu ] ) .
the fitted x - factor of the @xmath0co and c@xmath1o emission are 7.48@xmath13510@xmath108 and 7.23@xmath13610@xmath137@xmath105k@xmath26km@xmath26s , respectively . in fig.[fig_av_w13 ] and [ fig_av_w18 ] , the distribution of points in each panel is similar to that for @xmath12co but has a narrower distribution because their optical depths are smaller than that of the @xmath12co emission . especially for the c@xmath1o emission , since @xmath138 , even the distribution of l1551 mc has a linear correlation , compared with the @xmath12co and @xmath0co .
alonso - albi , t. , fuente , a. , crimier , n. , et al .
2010 , , 518 , a52 aniano , g. , draine , b. t. , gordon , k. d. , & sandstrom , k. 2011 , , 123 , 1218 bergin , e. a. , & tafalla , m. 2007 , , 45 , 339 bertout , c. , robichon , n. , & arenou , f. 1999 , , 352 , 574 bethell , t. j. , zweibel , e. g. , & li , p. s. 2007 , , 667 , 275 bohlin , r. c. , savage , b. d. , & drake , j. f. 1978 , , 224 , 132 bolatto , a. d. , wolfire , m. , & leroy , a. k. 2013 , , 51 , 207 burrows , c. j. , stapelfeldt , k. r. , watson , a. m. , et al .
1996 , , 473 , 437 caselli , p. , walmsley , c. m. , tafalla , m. , dore , l. , & myers , p. c. 1999 , , 523 , l165 devine , d. , reipurth , b. , & bally , j. 1999 , , 118 , 972 dickman , r. l. 1978 , , 37 , 407 chou , t .- l . , takakuwa , s. , yen , h .- w . , ohashi , n. , & ho , p. t. p. 2014
, , 796 , 70 clark , p. c. , & glover , s. c. o. 2015 , , 452 , 2057 emerson , j. p. , harris , s. , jennings , r. e. , et al . 1984 , , 278 , l49 emerson , d. t. , & graeve , r. 1988 , , 190 , 353 ford , a. b. , & shirley , y. l. 2011 , , 728 , 144 frerking , m. a. , langer , w. d. , & wilson , r. w. 1982 , , 262 , 590 griffin , m. j. , abergel , a. , abreu , a. , et al .
2010 , , 518 , ll3 gutermuth , r. a. , megeath , s. t. , myers , p. c. , et al .
2009 , , 184 , 18 herbig , g. h. , andrews , s. m. , & dahm , s. e. 2004 , , 128 , 1233 hirota , t. , ando , k. , bushimata , t. , et al . 2008 , , 60 , 961 hollenbach , d. j. , & tielens , a. g. g. m. 1997 , , 35 , 179 kawamura , a. , onishi , t. , yonekura , y. , et al .
1998 , , 117 , 387 knyves , v. , andr , p. , menshchikov , a. , et al .
2010 , , 518 , ll106 kong , s. , lada , c. j. , lada , e. a. , et al .
2015 , , 805 , 58 lada , c. j. , lada , e. a. , clemens , d. p. , & bally , j. 1994 , , 429 , 694 lada , c. j. , lombardi , m. , & alves , j. f. 2009 , , 703 , 52 lee , m .- y .
, stanimirovi , s. , wolfire , m. g. , et al .
2014 , , 784 , 80 liszt , h. s. 2007 , , 476 , 291 looney , l. w. , mundy , l. g. , & welch , w. j. 1997 , , 484 , l157 menten , k. m. , reid , m. j. , forbrich , j. , & brunthaler , a. 2007 , , 474 , 515 moriarty - schieven , g. h. , & wannier , p. g. 1991 , , 373 , l23 moriarty - schieven , g. h. , johnstone , d. , bally , j. , & jenness , t. 2006 , , 645 , 357 mundt , r. , buehrke , t. , solf , j. , ray , t. p. , & raga , a. c. 1990 , , 232 , 37 pineda , j. l. , goldsmith , p. f. , chapman , n. , et al .
2010 , , 721 , 686 poglitsch , a. , waelkens , c. , geis , n. , et al .
2010 , , 518 , ll2 pound , m. w. , & bally , j. 1991 , , 383 , 705 press , w. h. , teukolsky , s. a. , vetterling , w. t. , & flannery , b. p. 2007
, numerical recipes ( cambridge : cambridge univ . press ) , 778 qian , l. , li , d. , & goldsmith , p. f. 2012 , , 760 , 147 reipurth , b. , rodrguez , l. f. , anglada , g. , & bally , j. 2002 , , 124 , 1045 rllig , m. , & ossenkopf , v. 2013 , , 550 , aa56 sandstrom , k. m. , peek , j. e. g. , bower , g. c. , bolatto , a. d. , & plambeck , r. l. 2007 , , 667 , 1161 sawada , t. , ikeda , n. , sunada , k. , et al .
2008 , , 60 , 445 schwab , f. r. 1984 , in optimal gridding of visibility data in radio interfer- ometry , indirect imaging , ed .
j. a. robert ( cambridge : cambridge univ .
press ) , 333 shetty , r. , glover , s. c. , dullemond , c. p. , & klessen , r. s. 2011 , , 412 , 1686 shetty , r. , glover , s. c. , dullemond , c. p. , et al .
2011 , , 415 , 3253 shimajiri , y. , kawabe , r. , takakuwa , s. , et al .
2011 , , 63 , 105 shimajiri , y. , sakai , t. , tsukagoshi , t. , et al .
2013 , , 774 , ll20 shimajiri , y. , kitamura , y. , saito , m. , et al .
2014 , , 564 , aa68 shimajiri , y. , kitamura , y. , nakamura , f. , et al .
2015 , , 217 , 7 snell , r. l. , loren , r. b. , & plambeck , r. l. 1980 , , 239 , l17 snell , r. l. 1981 , , 45 , 121 sorai , k. , sunada , k. , okumura , s. k. , et al .
2000 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol .
4015 , radio telescopes , ed . h. r. butcher , 86 - 95 stojimirovi , i. , narayanan , g. , snell , r. l. , & bally , j. 2006 , , 649 , 280 swift , j. j. , welch , w. j. , & di francesco , j. 2005 , , 620 , 823 swift , j. j. , welch , w. j. , di francesco , j. , & stojimirovi , i. 2006 , , 637 , 392 swift , j. j. , & welch , w. j. 2008 , , 174 , 202 sunada , k. , yamaguchi , c. , nakai , n. , et al .
2000 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 4015 , radio telescopes , ed . h. r. butcher , 237 - 246 szcs , l. , glover , s. c. o. , & klessen , r. s. 2014 , , 445 , 4055 tachihara , k. , onishi , t. , mizuno , a. , & fukui , y. 2002 , , 385 , 909 takakuwa , s. , saito , m. , saigo , k. , et al .
2014 , , 796 , 1 ulich , b. l. , & haas , r. w. 1976 , , 30 , 247 van dishoeck , e. f. , & black , j. h. 1988 , , 334 , 771 visser , r. , van dishoeck , e. f. , & black , j. h. 2009 , , 503 , 323 warin , s. , benayoun , j. j. , & viala , y. p. 1996
, , 308 , 535 wilson , t. l. 1999 , reports on progress in physics , 62 , 143 yoshida , a. , kitamura , y. , shimajiri , y. , & kawabe , r. 2010 , , 718 , 1019 co ( @xmath14=10 ) , ( b ) @xmath0co ( @xmath14=10 ) , and ( c ) c@xmath1o ( @xmath14=10 ) in @xmath57=[@xmath5815kms@xmath26 , 20kms@xmath26 ] in units of k kms@xmath26 with effective beam sizes of 24@xmath259 , 25@xmath252 , and 25@xmath252 , respectively .
the crosses denote the class i sources ne , and irs 5 , and the class i / ii source hl tau .
the circles denote xz tau , lkh@xmath19 358 and v1213 tau . the black ellipse denotes the starless core l1551 mc .
the blue and white ellipses denote a u - shaped wall structure and the magenta one a cavity structure .
the linear scale of 0.1pc is shown at the bottom - left corner of each panel .
( d ) the column density map of h@xmath82 in units of @xmath105 with an effective beam size of 18@xmath22 , derived from _
herschel _ data .
( e ) the excitation temperature map in units of k with an effective beam size of 30@xmath254 , derived from the @xmath12co ( @xmath14=10 ) peak intensities .
( f ) the dust temperature map in units of k with an effective beam size of 36@xmath22 .
each beam size is denoted as the white ellipses at the bottom - right corner of each panel . ]
co ( @xmath14=10 ) and ( b ) c@xmath1o ( @xmath14=10 ) and the column density maps of ( c ) @xmath0co and ( d ) c@xmath1o in units of @xmath105 .
the effective beam sizes are 30@xmath254.[fig_tau_n ] note that these maps are made from the data with the signal - to - noise ratio larger than 5 .
the symbols in the panels have the same meaning as in fig.1 . ] and @xmath86 , the abundance ratio , @xmath4 , and the visual extinction , @xmath7 .
( a ) the @xmath85 map and ( b ) the @xmath86 map with effective beam sizes of 36@xmath22 . the gray contour in the panel ( b ) indicates the ism standard value , @xmath86 = 1.6@xmath310@xmath88 .
( c ) the @xmath140 map in units of mag , derived from _
herschel _ data .
( d ) the @xmath4 map and ( e ) the @xmath4 map without outflow and c@xmath1o depletion regions . the effective beam sizes are 36@xmath22 .
the white , magenta , and black contours in the panels ( d ) and ( e ) indicate 5.5 , 8.25 , 11mag , respectively .
( f ) the @xmath12co ( @xmath14=10 ) integrated intensity contours with an effective beam size of 36@xmath22 superposed on the column density map of h@xmath82 with an effective beam size of 18@xmath22 . the contours indicate the blueshifted outflows ( blue , @xmath57 = [ @xmath5815kms@xmath26 , 5.5kms@xmath26 ] ) , the redshifted outflows ( red , @xmath57 = [ 7.9kms@xmath26 , 20kms@xmath26 ] ) , and the ambient component ( black , @xmath57 = [ 5.5kms@xmath26 , 7.9kms@xmath26 ] ) .
the contour levels start from 3@xmath34 with a step of 6@xmath34 where the rms noise
1@xmath34 levels are 1.44k kms@xmath26 ( red ) , 2.68k kms@xmath26 ( black ) , 1.23k kms@xmath26 ( blue ) .
the green polygon is a selected region of l1551 mc and the two incomplete green circles are selected regions of irs 5 and ne , respectively .
the symbols in the panels have the same meaning as in fig.1 . ] , and the abundance ratio , @xmath4 .
the top x - axis is the column density of h@xmath82 , @xmath89 .
the four panels ( a ) , ( b ) , ( c ) , and ( d ) correspond to the four cases of the excitation temperatures of the @xmath0co and c@xmath1o ( @xmath14=10 ) lines .
the blue and green dots denote the undepleted ambient and rest ( depletion factor @xmath141 1 and outflow regions ) components , respectively .
the black filled squares and their error - bars are the average values and the dispersion of @xmath4 in the 2.5mag @xmath7 bin .
the yellow contours indicate the surface density of ambient points at the 15@xmath24 , 45@xmath24 , and 75@xmath24 levels of the maximum surface density of the undepleted ambient points .
the horizontal dashed lines indicate the solar system value of 5.5 . ]
o , @xmath84 , and the abundance ratio , @xmath4 in l1551 and orion a. the blue and green dots denote the undepleted ambient and rest components in l1551 , respectively .
the gray dots denote the orion a data from shimajiri et al .
the yellow and black contours indicate the surface density of the blue and grey points at the 15@xmath24 , 45@xmath24 , and 75@xmath24 levels of their maximum surface density of the ambient component in l 1551 and orion a , respectively .
the horizontal dashed line indicates the solar system value of 5.5 . ] , and the integrated intensity of @xmath12co ( @xmath14=10 ) , @xmath107 .
the gray dots denotes all data points in the map .
the color - coded dots show the data points in the labeled regions , and the color represents the excitation temperature , @xmath59 .
the black contours indicate the surface density of the color - coded points at the 15@xmath24 , 45@xmath24 , and 75@xmath24 levels of the maximum surface density in each panel . ] , and the x - factor .
the gray dots denotes all data points in the map .
the color - coded dots show the data points in the labeled regions , and the color represents the excitation temperature , @xmath59 .
the blue contours indicate the surface density of the color - coded points at the 5@xmath24 , 15@xmath24 , 45@xmath24 , and 75@xmath24 levels of the maximum surface density of all data . ] , and the x - factor .
the gray dots denotes all data points in the map .
the color - coded dots show the data points in the labeled regions , and the color represents the visual extinction , @xmath7 .
the blue contours indicate the surface density of the color - coded points at the 5@xmath24 , 15@xmath24 , 45@xmath24 , and 75@xmath24 levels of the maximum surface density of all data . ]
co ( @xmath14=10 ) , @xmath126 , and the x - factor .
the gray dots denotes all data points in the map .
the color - coded dots show the data points in the labeled regions , and the color represents the excitation temperature , @xmath59 .
the blue contours indicate the surface density of the color - coded points at the 5@xmath24 , 15@xmath24 , 45@xmath24 , and 75@xmath24 levels of the maximum surface density of all data . ]
co ( @xmath14=10 ) in units of k. the numbers at the top - left corners denote the central lsr velocities of the individual channels in units of kms@xmath26 .
the effective beam sizes are 30@xmath250 .
the linear scale of 0.25pc is shown at the bottom - left corner of the first panel.[fig_ch_12co ] ] co ( @xmath14=10 ) , ( b ) @xmath0co ( @xmath14=10 ) , and ( c ) c@xmath1o ( @xmath14=10 ) in units of kms@xmath26 with effective beam sizes of 28@xmath253 , 28@xmath256 , and 28@xmath256 , respectively .
the gray and black dashed lines indicate in ( b ) the narrow filamentary structure and the redshifted streamer , respectively .
the velocity dispersion maps of ( d ) @xmath12co ( @xmath14=10 ) , ( e ) @xmath0co ( @xmath14=10 ) , and ( f ) c@xmath1o ( @xmath14=10 ) in units of kms@xmath26 with effective beam sizes of 28@xmath253 , 28@xmath256 , and 28@xmath256 , respectively .
the mean velocities and velocity dispersions of @xmath12co , @xmath0co , and c@xmath1o maps are calculated in @xmath57 = [ @xmath5815kms@xmath26 , 20kms@xmath26 ] , [ 3kms@xmath26 , 8kms@xmath26 ] , and [ 5.7kms@xmath26 , 7.5kms@xmath26 ] , respectively .
the rest symbols in the panels have the same meaning as in fig.1 . ]
co ( @xmath14=10 ) in units of k. the numbers at the top - left corners denote the central lsr velocities of the individual channels in units of kms@xmath26 .
the effective beam sizes are 30@xmath253 .
the linear scale of 0.25pc is shown at the bottom - left corner of the first panel.[fig_ch_13co ] ] o ( @xmath14=10 ) in units of k. the numbers at the top - left corners denote the central lsr velocities of the individual channels in units of kms@xmath26 .
the effective beam sizes are 30@xmath254 .
the linear scale of 0.25pc is shown at the bottom - left corner of the first panel.[fig_ch_c18o ] ] , and the integrated intensity of @xmath0co ( @xmath14=10 ) , @xmath133 .
the gray dots denotes all data points in the map .
the color - coded dots show the data points in the labeled regions , and the color represents the excitation temperature , @xmath59 .
the black contours indicate the surface density of the color - coded points at the 15@xmath24 , 45@xmath24 , and 75@xmath24 levels of the maximum surface density in each panel . ] , and the integrated intensity of c@xmath1o ( @xmath14=10 ) , @xmath134 .
the gray dots denotes all data points in the map .
the color - coded dots show the data points in the labeled regions , and the color represents the excitation temperature , @xmath59 .
the black contours indicate the surface density of the color - coded points at the 15@xmath24 , 45@xmath24 , and 75@xmath24 levels of the maximum surface density in each panel . ]
lccc molecular line & @xmath12co ( @xmath14=10 ) & @xmath0co ( @xmath14=10 ) & c@xmath1o ( @xmath14=10 ) + rest frequency [ ghz ] & 115.27120 & 110.20135 & 109.78218 + primary beam hpbw [ @xmath22 ] & @xmath215 & @xmath216 & @xmath216 + frequency resolution [ khz ] & 37.8 & 37.8 & 37.8 + main beam efficiency & 0.32 & 0.38 & 0.43 + observation & 2007 dec2008 jun & 2009 dec2010 feb & 2009 feb2009 may + scan mode & otf & otf & otf + mapping area [ arcmin@xmath15 ] & 44 @xmath3 44 & 42 @xmath3 43 & 30 @xmath3 30 + pixel grid size [ @xmath22 ] & 7.5 & 7.5 & 7.5 + effective beam size [ @xmath22 ] & 21.8 & 22.2 & 22.2 + velocity resolution [ kms@xmath26 ] & 0.1 & 0.1 & 0.1 + typical noise level in @xmath35 [ k ] & 1.23 & 0.94 & 0.67 + | we investigate the intrinsic abundance ratio of @xmath0co to c@xmath1o and the x - factor in l1551 using the nobeyama radio observatory ( nro ) 45 m telescope .
l1551 is chosen because it is relatively isolated in the taurus molecular cloud shielded from fuv photons , providing an ideal environment for studying the target properties .
our observations cover @xmath240@xmath340with resolution @xmath230 , which are the maps with highest spatial dynamical range to date .
we derive the @xmath4 value on the sub - parsec scales in the range of @xmath2327 with a mean value of [email protected] . comparing to the visual extinction map derived from the @xmath6 observations
, we found that the abundance ratio reaches its maximum at low @xmath7 ( i.e. , @xmath8 14mag ) , and decreases to the typical solar system value of 5.5 inside l1551 mc . the high @xmath4 value at the boundary of the cloud is most likely due to the selective fuv photodissociation of c@xmath1o .
this is in contrast with orion - a where its internal ob stars keep the abundance ratio at a high level greater than @xmath210 . in addition
, we explore the variation of the x - factor , because it is an uncertain but widely used quantity in extragalactic studies .
we found that x - factor @xmath9 @xmath10 which is consistent with previous simulations . excluding the high density region ,
the average x - factor is similar to the milky way average value . |
a number of experiments have returned microlensing results on the nature of the dark matter in the milky way .
the eros collaboration ( aubourg , 1993 ; lasserre 2000 ) monitored 17.5 million stars in the lmc for 2 years and 5.3 million stars for 6 years finding 3 microlensing candidates . using these events and an efficiency analysis they found an upper limit of 40% ( 95% c.l . ) on the halo mass fraction in machos in the mass range of @xmath0 to @xmath1 .
the macho collaboration ( alcock , 1993 ; 1997 ; 2000 ) monitored 11.9 million stars in the lmc over 5.7 years , finding between 13 and 17 microlensing candidates .
they did a careful efficiency analysis , including the effect of blending in their crowded fields and found a best fit macho halo fraction of 20% ( @xmath2 at 95% cl ) , which corresponds to an optical depth of @xmath3 .
it is important to realize that the macho estimate of dark matter halo fraction depended crucially on the use of standard models of the milky way thin disc , thick disc , spheroid , halo , and the lmc disc and halo .
in particular they estimated that between 2 and 4 events ( @xmath4 between @xmath5 and @xmath6 . should be from lensing of known stars , depending on galactic and lmc structure .
several groups ( e.g. wu 1994 ; sahu 1994 ; gates , 1996 ; evans , 1998 ; gould 1998 ; aubourg , 1999 ; alves & nelson 2000 ; weinberg 2000 ; evans & kerins 2000 ; distefano 2000 ; gyuk , 2000 ; mancini et al .
2004 ) have investigated whether additional stars in the lmc , or modification to the milky way structure could explain the extra microlensing events , with the augument being made both ways .
if the extra microlensing events are due to a new halo population of machos ( white dwarfs ? primordial black holes ? ) , they would represent roughly the mass of the milky way disc and would have a major effect on galactic chemical evolution , galaxy formation and cosmology .
thus , currently the most popular hypothesis is probably that the lmc has an additional , as yet undiscovered , extended population of stars , that is producing the extra microlensing .
the mystery of the extra lmc microlensing has thus been phrased as the question of whether there is lmc self - lensing or a new population of halo objects that make a small but significant contribution to the dark matter .
belokurov , evans , & le du ( 2003 , 2004 ) , and evans & belokurov ( 2004 ) ( bel / eb ) have suggested instead that the macho collaboration sample of microlensing events is severely contaminated with non - microlensing events , and that our optical depth is thus substantially over - estimated .
this is certainly a possibility as has been known from the beginning of the microlensing surveys , and is a topic worthy of investigation .
bel / eb reanalysed 22000 random macho light curves and 29 macho microlensing and sne candidates using a neural net and selected 7 microlensing events ( bel 2004 ) . in their first paper on the lmc , bel recognized that they had not analyzed all 11.9 million light curves , and did not claim to prove that the macho optical depth was greatly overestimated , but did strongly suggest that that was the case ( bel 2004 ) . in talks , and in the most recent paper ( evans & belokurov 2004 )
, they were less cautious ; in eb 2004 an efficiency calculation was performed and an optical depth was calculated .
strong claims are now being made that contamination in the macho data set is the solution to the mystery of lmc microlensing ( eb 2004 ) .
however , as we show below , there are severe errors in the bel / eb analysis , and so it is not yet useful in determining whether the macho optical depth is over - estimated .
fortunately there are several upcoming experiments that can address and hopefully solve the puzzle of the extra lmc microlensing .
the mega ( de jong 2004 ) and point - agape ( paulin - henriksson 2003 ) experiments are monitoring stars in m31 .
this is hopeful , since the large difference in optical depth between sources on m31 s near side and far side can distinguish between halo and disc microlensing .
preliminary results are still ambiguous with mega finding weak evidence in favour of halo microlensing with 4 events ( 3 on the far side and 1 on the near side ) , and point - agape finding weak evidence against halo microlensing with 3 events . in addition , the supermacho collaboration ( becker 2004 ) hopes to find enough new lmc events to clarify the situation .
the ultimate test may come from space missions , for example , the space interferometry mission microlensing key project ( unwin & turyshev 2002 ) which can measure the distance to the lenses ( gould & salim 1999 ) . even determining the distance to 3 or 4 lenses
will conclusively solve the problem of whether the extra microlensing is due to lmc or halo lenses .
in addition there is a proposed deep impact extended space mission ( dime ) ( cook , 2003 ) which is dedicated to microlensing parallax measurements from space and could answer the question if approved .
since these experiments are still underway , it is important that the community knows that the puzzle has not yet been solved and that the bel / eb claims should be treated with caution .
the optical depth to microlensing is just the chance that a given source star is undergoing microlensing .
since this probability is small , one monitors a large number of stars , @xmath7 , for a long period of time , @xmath8 , and estimates the optical depth by @xmath9 where the duration of the event , @xmath10 , is the einstein ring diameter crossing time and @xmath11 is the efficiency for that event ( probability of the combination of the experiment and selection criteria finding an event with duration @xmath10 in that field . ) in order to estimate @xmath4 one needs to search through all @xmath7 light curves and use some automated selection criteria to pick out the microlensing candidates .
this gives the set of microlensing events and their durations .
then one must apply exactly the same analysis and selection criteria to a large sample of simulated microlensing light curves and determine @xmath11 as the fraction of simulated events that the analysis plus selection criteria find .
every set of selection criteria will miss some real microlensing events , and also identify some non - microlensing events as microlensing , so developing selection criteria is a trade - off between losing real microlensing events and contaminating the sample with non - microlensing .
allowing false positives erroneously increases the optical depth , while cuts that are too restrictive eliminate actual events resulting in larger errors on the measured optical depth .
contamination is more serious than missing events since , if properly calculated , a decrease in efficiency will compensate for the reduction in number of events .
thus ideally , one uses the broadest set of cuts that appears to give negligible contamination by false positives . in the above framework ,
a neural network program is just a set of selection criteria , and so the above procedure applies .
neural network programs have found wide use recently for data analysis in particle physics , astrophysics , and many other arenas .
they have been found to be good at making efficient cuts in multi - dimensional parameter spaces , so we agree that the main idea of bel / eb , redoing the macho event selection with a neural net , is a good one . however , any set of selection criteria , even a neural net , is only as good as the data input to it and the way it is used to calculate the efficiencies and the final results .
so , first the raw light curve data must be converted into statistics that capture all the essential distinguishing properties of the microlensing events and the background .
then , after training the neural net on simulated events and some real and/or simulated background , all @xmath7 light curves should be fed to the net so it can produce a set of microlensing candidates
. then it must be run on a fair sample of simulated microlensing light curves to calculate its efficiency as a function of event duration .
bel / eb did not perform these steps correctly .
one problem that exists for every microlensing experiment is that whatever signal / noise ( s / n ) threshold one sets , most of the events will occur close to that threshold .
the s / n is closely related to the maximum magnification of the event , @xmath12 , which is roughly inversely related to the impact parameter , @xmath13 . since microlensing is uniformly probable in impact parameter
, more events occur with small @xmath12 than large @xmath12 . in the macho collaboration , we experimented with over 100 statistics derived from the light curves , investigating their usefulness using simulated microlensing and our experience with backgrounds .
we picked out around 20 of the most powerful statistics in making our selection criteria and in distinguishing microlensing from variable stars , supernovas , and noise .
we found that statistics involving fitting the photometric light curves with microlensing and supernova models , and those that used the photometric errors were especially useful in picking out the lower s / n events .
however , bel / eb did not apply their neural net program to our statistics , but instead created 5 of their own statistics , mostly based on cross and autocorrelation of the light curve data , as well as several additional statistics based on colour information for supernova discrimination .
none of their statistics make direct use of the photometric errors , and their main s / n statistic , @xmath14 , does not make use of the known microlensing light curve shape
. thus one would expect that they would have difficultly identifying low magnification microlensing events , as is evident from the macho events they throw out as non - microlensing . using less powerful statistics
makes it likely that their method would have lower efficiency at finding microlensing .
this does not have to be a problem however .
a less efficient set of selection criteria will give a lower number of microlensing candidates , but , since each event is weighted by the inverse of its detection efficiency , if the efficiencies are properly calculated , the optical depth estimate should be roughly the same . as we discuss below , bel / eb did not do a proper efficiency calculation . next , for the lmc , bel / eb ran their neural net on only 22000 out of 11.9 million light curves .
given the expected optical depth in the range of @xmath15 they should have expected to find less than one event .
they found 7 microlensing candidates because they also ran the net on the 29 events that macho selection criteria had picked out .
thus their selection criteria are actually the neural net ` anded with macho selection criteria .
since they were sequentially applying both the macho selection criteria and the neural net , it is clear that the efficiency of their method is necessarily lower than the macho selection criteria . in order to tell how much lower
, they would need to actually apply both the macho criteria and their net to a sample of simulated light curves .
they did not do this , and in fact , in figure 3 ( evans & belokurov 2004b ) they show a plot of efficiency versus @xmath10 that has a * higher * efficiency than our criteria b ' efficiencies ( which is our highest efficiency set ) .
they erroneously state that the efficiency of our criteria b is a lower limit to their efficiency when , in fact , our efficiencies are an upper limit to their efficiencies since they use our cuts before applying theirs .
we have extensive experience looking at light curves and examination of the light curves rejected by the neural net , shows they are mostly the low s / n events with low maximum magnification and shorter duration .
this is what we would expect given their less powerful statistics . considering only the less efficient and less contaminated macho selection criteria a for simplicity , bel /
eb keep only 3 of the 9 events with @xmath16 .
if this is a general rule then their real efficiency ( neural net plus selection criteria a ) should be lowered by roughly a factor of @xmath17 .
remarkably , if one corrects their efficiencies by this factor and uses the events they select as microlensing , one finds their estimate of optical depth increases to @xmath18 , consistent with ours ( with a larger error due to having fewer events ) .
of course , this correction is based upon speculation of what the bel / eb efficiencies are , as are the bel / eb results .
however , a point in favour of bel / eb efficiencies being lower , is that the microlensing interpretation of some of the events bel / eb claim are contaminants has been strengthened by follow - up data ( bennett et al .
2004 ) . as bel / eb emphasise
it is extremely important to be sure that there is not much contamination in the final microlensing sample .
one check we did was to compare our distributions of @xmath13 with the distribution expected from microlensing ( and not expected from variable stars or sne ) .
we performed a kolmogorov - smirnov test and found excellent agreement , especially after correcting for our efficiency as a function of @xmath13 . as we pointed out in alcock , ( 2000 ) , if one accepts the high magnification events as microlensing then there _ must _ be many more lower magnification events .
selection criteria ` a ' especially yield these events in just the right proportion .
another related and serious error is that if one took the bel / eb claim of having higher efficiencies than criteria b seriously , then in their search through the 11.9 million light curves , they should have found some additional real microlensing that we missed .
these events then should have been added into the optical depth estimation .
they claim that it is unlikely that they would find any additional microlensing candidates in the 11.9 million light curves because they did not find any in the 22000 macho light curve sample they investigated .
this claim does not make sense . if they had found even one new event in only 22000 light curves they then should expect to find more than 500 new events in the 11.9 million .
this would give a rather large value for @xmath4 .
however , there is a serious worry that the bel / eb net as currently configured would , in fact , find such a large number of additional microlensing candidates if it were run on 11.9 million light curves . in bel ( 2004 ) , and eb ( 2004 ) they say they found 2 false positives in 22000 light curves which were near the borderline between microlensing and noise events ( and we suppose were then removed by hand ? ) . to get an accurate estimate of efficiencies
, one must treat the real data in the same way as the simulated data , so one can not arbitrarily throw out such events .
if their false positive rate is 2 in 22000 for random macho light curves , they would be swamped with about 1000 non - microlensing contaminates .
this would require them to tighten their cuts and/or add more statistics , resulting in a lowering of their efficiencies . by relying on the macho selection criteria to remove false positives from the 11.9 million light curves , bel / eb were able to use less powerful statistics and looser cuts , thus fooling themselves that their net had a high efficiency .
the macho collaboration discovered the fact that sne occurring in background galaxies are a serious contaminant in microlensing surveys towards the lmc , and must be carefully identified and removed .
we use two criteria for identifying sne .
first , using hst or good seeing data if they exist , we check if there is a bright background galaxy overlapping the microlensing candidate .
this condition is very rare for a random star , and when it occurs on a microlensing event , basically ensures the event is a sn .
next we fit our candidates with a sn type ia light curve model and ask whether the sn fit is better than the microlensing fit .
luckily we found for our set of events that apart from one event ( event 26 which is not in set ` a ' ) these two tests always agreed .
thus we easily removed all supernova contamination , especially for set ` a ' .
bel / eb did not check for background galaxies , but used colour shift and peak symmetry information from the light curve to try to identify sne .
they correctly identified some of our sne , but their sn neural net ( bel 2004 ) did not correctly identify two of these sne ( events 10 and 24 ) .
they say they do not confirm the sn classification of the alcock , et al . "
but the clear presence of a background galaxy in the images shows that in fact these are sne and that the bel / eb sn net is not as reliable as our method .
we note that event lmc-10 is one of our best sn events with a near perfect fit to our type ia template .
bel / eb seem to reject it because it does not show typical type ia colours .
this illustrates the power of light curve fitting . in the end
, they leave these events out of their final set of microlensing candidates , but did not explain why ( bel 2004 ; eb 2004 ) . bel / eb make the case that their neural net is superior to our selection criteria in picking out microlensing and not producing false positives .
while we can not tell for sure until they apply it to all 11.9 million light curves and do a proper efficiency analysis , we find this quite unlikely .
as discussed above , bel / eb reduce the photometric data to a few statistics that throw away much of the information we have found most useful in picking out lower s / n microlensing events . for example their statistics can not distinguish between microlensing and any other achromactic , symmetric bump .
in addition , they clean " the photometric data by removing any point that deviates more than 3-sigma from its two neighbours .
thus they remove data from any fast rising short duration peak .
the macho collaboration also cleans the data by removing bad observations , but we find these by examining all stars in a given image and removing images that give many deviant photometric points .
this is illustrated by event macho-119.20738.3418 ( figure [ fig : lc ] ) , which is missed by bel / eb , possibly due to their cleaning method . in performing their cross- and autocorrelations
bel / eb also assume that the photometric data is evenly spaced in time , when in fact the intervals between observations vary widely
. therefore they may disregard information that we found useful and introduce unnecessary noise , especially during rapidly rising events .
we also suspect that the simulated microlensing events they used to train the net and to run their limited efficiency effort are quite unlike actual microlensing events in the macho data .
we found that the scatter in the data and the photometric errors vary greatly from observation to observation , depending strongly on seeing , local crowding , sky , etc . and are non - gaussian .
thus to create our simulated light curves we added microlensing flux to actual ( unlensed ) light curves being careful to scale the errors properly . if the simulated light curves are not very similar to real microlensing then the efficiency calculation will give incorrect results .
this is independent of whether a neural net or more traditional selection method is used . finally , much of the effort in the macho efficiency calculation went into understanding blending .
bel / eb claim that this effect cancels out to first order , and this is correct , but the 2nd order effect can be significant .
if one blindly uses our blending efficiencies with a different set of selection criteria ( the neural net ) one is just speculating on what the answer will be .
there is one variable star contaminant in the macho lmc data set that seems to be well established .
the eros collaboration ( glicenstein 2004 ) discovered that event 23 bumped again after 7 years and therefore is unlikely to be microlensing . since this event contributes 8% of the optical depth to set a " ( 6% to set
b " ) , all of our estimates of optical depth and macho halo fraction should be reduced by these amounts .
( 8% in set a and 6% for set b ) .
our estimated optical depth @xmath19 should be reduced to @xmath20 .
the bigger worry caused by this discovery is that there could be more such events in our sample .
we note that the bel / eb neural net incorrectly identified event 23 as microlensing and so is no help here . in alcock , 2000
, we were worried about such contamination which is why we employed two different sets of selection criteria .
the nature of microlensing is such that most of the events will be at low @xmath12 and therefore low s / n .
we attempted to study the magnitude of possible contamination systematic error by using a looser set of cuts ( set b " ) that we felt would be more seriously contaminated , and comparing with a tighter set of cuts ( set a " ) with less contamination ( and fewer real events ) .
we found that since the efficiency of the tighter cuts was lower , the final optical depth estimates were very similar .
we used the 17% difference between the two methods as an estimate of the size of the contamination error .
we note that the 8% reduction in @xmath4 implied by removing event 23 is within this reported error .
contamination will always increase the optical depth estimate , but there are other small systematic effects , such as the higher probability of missing binary events , that would shift the optical depth in the opposite direction .
we find that the macho collaboration estimates of optical should be reduced by around 8% due to the contamination of our set of events with a variable star ( event 23 ) .
while there may be other contaminants in our data sample , the distribution of @xmath13 and the agreement in optical depth from our more contaminated " and less contaminated " samples argue that this is unlikely to be substantial .
we find that the re - analysis of bel / eb is flawed and at this time does not make useful statements about contamination in the macho data set .
we are not saying that a neural net along the lines of bel / eb can not be useful in selecting microlensing events , but only that one has yet to be applied correctly .
neural nets can in many cases do better than more conventional cuts , and we applaud the effort at applying them to microlensing . table [ tab : compare ] gives a comparison of the macho and bel / eb analyses .
it is clear that machos can not make up the bulk of the milky way dark matter .
however , we find that the mystery of the lmc microlensing events , that is , whether they indicate a new halo population or lmc self lensing , is still experimentally unresolved , though lmc self lensing is the more popular explanation at this time for several reasons .
contamination by non - microlensing events such as lmc-23 is also still a possible explanation .
future experiments can answer this question definitively .
alcock , c. , et al . , 1993 , nature , 365 , 621 alcock , c. , et al . , 1997 ,
apj , 486 , 697 alcock , c. , et al . , 2000 , apj , 542 , 281 alves , d.r . ,
& nelson , c.a . ,
2000 , apj , 542 , 789 aubourg , e. et al . , 1993 ,
nature , 365 , 623 aubourg , e. , palanque - delabrouille , n. , salati , p. , spiro , m. , & taillet , r. , 1999 , a&a , 347 , 850 becker , a.c . ,
2004 , impact of gravitational lensing on cosmology , proceedings iau symposium no . 225 , 2004 belokurov , v. , evans , n.w . , & le du , y.l . , 2003 , mnras , 341 , 1373 belokurov , v. , evans , n.w .
, & le du , y.l . , 2004a , mnras , 352 , 233 bennett , d.p . ,
becker , a.c . , & tomaney , a.b .
, 2004 , astro - ph/0501101 cook , k. , et al .
, 2003 , baas , 20314302 dejong , j.t.a .
et al . , 2004 ,
a&a , 417 .
461 distefano , r. , 2000 , apj , 541 , 587 evans , n.w . & belokurov , v. , 2004b , astro - ph/0411222 evans , n.w . & kerins , e. , 2000 , apj , 529 , 917 evans , n.w . & gyuk , g. , turner , m.s . , & binney , j. , 1998 , apj , 501,45 gates , e.i . ,
gyuk , g. , & turner , m.s . , 1996 , phys rev d , 53 , 4138 glicenstein , j.f . , 2004 , talk at the hawaiian gravitational microlensing workshop gould , a. , 1998 , apj , 499 , 728 gould , a. & salim , s. , 1999 , apj , 524 , 794 gyuk , g. , dalal , n. , griest , k. , 2000 , apj , 535 , 90 lasserre , t. , et al . , 2000 ,
aa , 355 , l39 mancini , l. , calchi - novati , s. , jetzer , ph , & scarpetta , g. , 2004 , a&a , 427 , 61 minitti , d. , et al . , 2003 , sci , 301 , 1508 paulin - henriksson , s. , et al . , 2003 ,
a&a , 405 , 15 sahu , k.c . , 1994 ,
nature , 370 , 275 unwin , s. & turyshev , s. , 2002 , + @xmath22 wu , x .-
p . , 1994 ,
apj , 435 , 66 weinberg , m. , 2000 , apj , 532 , 922 | in a recent series of three papers , belokurov , evans , & le du ( 2002 , 2004 ) , and evans & belokurov ( 2004 ) , reanalysed the macho collaboration data and gave alternative sets of microlensing events and an alternative optical depth to microlensing toward the large magellanic cloud ( lmc ) . even though they examined less than 0.2% of the data they claimed that by using a neural net program they had reliably selected a better ( and smaller ) set of microlensing candidates . estimating the optical depth from this smaller set , they claim that the macho collaboration overestimated the optical depth by a significant factor and that the macho microlensing experiment is consistent with lensing by known stars in the milky way and lmc .
as we show below , the analysis by these authors contains several errors which render their conclusions meaningless .
their efficiency analysis is clearly in error , and since they did not search through the entire macho dataset , they do not know how many microlensing events their neural net would find in the data or what optical depth their method would give .
examination of their selected events suggests that their method misses low s / n events and thus would have lower efficiency than the macho selection criteria .
in addition , their method is likely to give many more false positives ( non - lensing events identified as lensing ) .
both effects would increase their estimated optical depth .
finally , we note that the eros discovery that lmc event-23 is a variable star reduces the macho collaboration estimates of optical depth and macho halo fraction by around 8% , and does open the question of additional contamination .
[ firstpage ] gravitational lensing
galaxy : halo dark matter
magellanic clouds |
a sub - wavelength object diffracts light into evanescent and propagating waves .
it is the evanescent part - the so - called forbidden light - that carries information on the sub - wavelength details of the object .
this evanescent contribution plays a key role in experiments targeted at imaging surface - plasmon polaritons ( spps ) , which are electron - photon hybrid states naturally confined at the boundary between a metal and an insulator . as such ,
spps are strongly modified by local changes of their environment at the nanoscale and it is therefore critical to find efficient methods to probe the interaction of spps with nanostructures .
this constitutes the central motivation of the present work . in this context
, imaging spps can be achieved using the ` forbidden light ' nsom arrangement [ 1 - 4 ] , where the forbidden light that couples into the sample glass substrate at angles larger than the critical incidence ( @xmath1 in fused silica of optical index @xmath2 ) is collected by an elliptic mirror .
an alternative method to collect the plasmonic forbidden light uses a fourier lens placed in the optical path of a lrm setup to selectively image in the fourier space spp beams propagating along specific directions @xcite .
+ in this paper , we consider a sub - wavelength slit patterned in a thin gold film that we image with a nsom using a point - like optical tip for illumination .
our first motivation is to extend previous works [ 1 - 5 ] by demonstrating that the use of fourier space spp signals for image reconstruction of this nanostructure provides optical images that are much better resolved ( about 4 times ) compared to images acquired in the real space directly .
our second motivation is related to the emerging field of quantum plasmonics , where one or several quantum emitters are coupled to plasmonic nanostructures to generate single spp quanta [ 6 - 12 ] .
this field is usually dealing with weak optical signals that can be polluted by a range of spurious incoherent light originating , e.g. , from gold fluorescence . cleaning
the useful quantum - related information from spurious incoherent signals could be crucial for future quantum plasmonics experiments : in addition to offering enhanced spatial resolution , our method offers a powerful way of making this cleaning step .
this is demonstrated here by using as point - like optical tip a scanning quantum source of light made of a single fluorescent nanodiamond glued at the apex of a fiber tip [ 13 - 18 ] .
a sketch of our setup is shown in fig . 1 . basically , this is a transmission nsom making use of a fluorescent nanodiamond - based optical tip @xcite for illumination of a thin ( 30 nm ) gold film deposited on a fused silica cover slip .
our motivations for using such a tip are multiple .
first , the small size ( @xmath3 25 nm ) of the nanodiamond mimics a point - like source of light , which has the potential for a better spatial resolution than standard metal - coated optical tips @xcite ( ultimately limited only by the scan height of the point source over the structure ) . this nanodiamond hosts three to four nitrogen - vacancy ( nv ) color - centers as revealed ; tl= tube lens ; @xmath4 , @xmath5 ( removable ) , @xmath6= achromatic lenses ; @xmath7= beam block ; @xmath8= multimode optical fiber ; apd= avalanche photodiode .
@xmath9 is the back - focal plane of @xmath4 .
@xmath10 is the objective back - focal plane and is located inside the objective itself @xcite .
the @xmath8-apd combination can be replaced by a camera ( not sketched ) aligned with the optical axis for imaging . in this setup , the tip is fixed and the sample is scanned in all three dimensions with nanometer accuracy .
the remaining excitation at 515 nm is removed by an optical filter ( not shown ) .
a limited number of light rays are indicated for clarity.,width=453 ] by time - intensity second - order correlation measurements [ 15 - 18 ] .
this means that the quantum source can not emit more than three to four photons at a time .
+ the diamond is illuminated with a @xmath11 nm laser light shone into the single mode optical fiber that is terminated by the tip . as we have shown previously
[ 16 - 18 ] , the red - orange near - field fluorescence of the nvs is able to launch spps into the gold film , whereas the green laser excitation can not do that because of strong interband absorptions in gold in this wavelength range : this is a second motivation for using a nanodiamond tip .
finally , our arrangement generates a range of undesired cross - excited nonplasmonic lights such as gold fluorescence in addition to the useful plasmonic signal : this will be advantageously used to demonstrate the ability of our method to discriminate among useful and spurious signals . in the quantum regime ,
where only a few photons couple to spps , it is crucial to eliminate this spurious light .
+ apart from the use of an active tip , our setup has additional special features .
first , a removable fourier lens @xmath5 is implemented . in the absence of @xmath5 , the light
is collected in the direct space in a standard way . in this case , the collection fiber ( @xmath8 in fig .
1 ) is optimally placed along the optical axis and the collected signal is sent to the apd . now adding @xmath5 so that the back - focal plane @xmath9 of the microscope is conjugated with the collection plane allows collecting light in the fourier space [ 5,16 - 18 ] ( note that the @xmath9 plane plays the same role as the back - focal plane of the objective @xmath10 that is located inside the objective itself and is difficult to access for optical imaging ) . in this case , the collection fiber @xmath8 can be laterally displaced to focus on the spp useful signal only ( see below ) , as shown schematically in fig .
in addition , an optional beam block ( @xmath7 in fig . 1 ) of adjusted lateral extension can be added in the fourier plane to remove all of the allowed light that couples into the silica substrate at smaller incidences not exceeding @xmath12 [ 18 ] .
finally , the collection fiber can be replaced by a wide - field camera to obtain either real - space or fourier - space instantaneous images of the spp propagation for a given tip position . used for image reconstruction .
( c ) cross sections of ( a ) and ( b ) along the blue dashed lines .
, width=453 ]
the principle of our method can be introduced by commenting on the fourier - space lrm images shown in figs .
2(a ) and 2(b ) , which are obtained with the tip positioned in the near field of a 30-nm thick gold film .
the fourier lens @xmath5 is set in and the image is recorded on the camera . in order to filter the light emitted in the allowed - light cone we introduce a beam block ( @xmath7 in fig .
1 ) centered in the fourier space ( compare figs . 2(a ) and 2(b ) ) .
a distinctive bright circle can be seen in both images .
it is typical for a spp field generated at the air - gold interface and leaking into the silica substrate [ 2,5,16 - 20 ] at leakage radiation angles @xmath13 , as a consequence of phase matching at the gold - silica interface . here
@xmath14 is spread over an angular @xmath15 range ( @xmath16 $ ] ) because of the phonon broadening in the nv emission used for spp excitation @xcite .
the spp circle corresponds to an effective index @xmath17 .
it therefore falls in the forbidden - light zone and , as such , can only be detected with a @xmath18 objective .
it represents a clear spp signature , essentially cleaned from any spurious light , which couples in silica at smaller incidences up to @xmath12 @xcite as already mentioned .
cross sections shown in fig .
2(c ) illustrate the importance of filtering the allowed light for clean spp imaging . from unsaturated images ( not shown )
we estimate the fraction of spurious light to @xmath19 of the total integrated signal in fourier space . therefore ,
in the following , we only focus our attention on those photons emitted in the forbidden - light region .
+ our method consists of mapping the intensity of the spp circle while the plasmonic system is raster scanning under the tip . for this purpose ,
the multimode fiber @xmath8 is positioned in the detection plane ( fig .
1 ) so as to coincide with this spp circle .
cross sections along the blue dashed lines in ( a ) and ( b ) [ respectively ( c ) and ( d ) ] . left scales stand for the optical signal levels , expressed in units of khz , right scales stand for the topography levels , expressed in nm.,width=453 ] the small circle in the upper - right part of the spp circle in fig .
2(b ) marks the imprint of @xmath8 used for image reconstruction .
note that only a small portion of the spp circle will be mapped out .
this does not affect the useful signal - to - background ratio since the spurious light is already filtered . to demonstrate our method , we have chosen to image a circular slit of @xmath3 120 nm rim thickness and @xmath3 1.5 @xmath20 m inner diameter patterned by focused - ion beam ( fib ) milling in a 30 nm thick gold film .
the results are shown in fig .
it is clear that the reconstructed image in fig .
3(d ) is much better resolved than the direct space image depicted in fig .
3(b ) . as a matter of fact
, the sharpness of the reconstructed image competes with that of the simultaneously acquired topographic image @xcite in fig .
this is confirmed by the cross sections shown in figs .
3(e ) and 3(f ) . in fig .
3(f ) , the full width at half maximum of the optical signal is around 130 nm , to compare to 100 nm @xcite in the corresponding topographic cross section , whereas in fig .
3(e ) , it is as large as 230 nm , compared to 120 nm in the topography .
therefore , reconstructed images from the fourier - filtered signals made only of high spatial frequencies , i.e. those due to spps leaking in the silica substrate , exhibit a four times enhanced spatial resolution of @xmath3 @xmath21 nm compared to @xmath3 @xmath22 nm obtained in the direct space .
this clearly shows the advantage of our method concerning resolution .
it is worthwhile to note that this 30 nm resolution fits well with the size of the nanodiamond and with the typical distance between tip and surface @xcite during scanning .
therefore , our imaging method is close to reaching the optimal spatial resolution achievable with a point - like optical tip @xcite .
+ we now wish to give some insight into the physics governing our observations .
we first point out that after fourier filtering the recorded image represents a fair measure of the the local amplitude of the excitation field at @xmath23 nm .
more precisely , below the saturation regime of the nvs where we are working , the fluorescence signal @xmath24 at the tip location @xmath25 is proportional to the excitation rate @xmath26 that depends on the local ( incident+ reflected ) field at the nv location ( @xmath27 is a transition dipole ) .
@xmath28 is strongly affected by the environment in the vicinity of the nanostructure @xcite . with our detection protocol
only those emitted photons coupling directly to spps _ or _ scattered by the structure at @xmath29 are recorded and contribute to the signal shown in fig .
the increased contrast between figs .
3(b ) and 3(d ) suggests that the coupling of the incident light at @xmath14 is enhanced only in the close vicinity of the aperture rim .
this is also in qualitative agreement with recent works [ 6 - 18 ] showing that plasmonic couplings depend critically on the distance between the quantum emitter and the nanostructure .
we remark that , while the light emission with the tip located far away from the rim is dominated by the leaky spp signal , the signal increase @xmath30 observed in the close vicinity of the structure results from a competition between plasmonic and non plasmonic light ( such as scattering and spurious fluorescence ) emitted at @xmath29 .
further analysis is required to identify the emission channel responsible for the increase in resolution reported in this work . +
it is worth pointing out that in refs .
@xcite , spp imaging was obtained with a classical nsom aperture - tip illuminated at a fixed laser wavelength . in particular ,
spp fringes were observed while scanning the tip in the vicinity of local protrusions @xcite or inside a square cavity made of grooves @xcite ( see also the theoretical analysis in refs .
@xcite ) . the absence of fringes in fig .
3(d ) is related to the fact that in our working regime we do not image the local photonic density - of - states at the emission wavelengths of the nv but the local excitation variation @xmath31 at @xmath32 nm . since no spp is excited at this wavelength no oscillation or fringe is expected to show up @xcite . + before summarizing , it is worth commenting on an additional feature of our method .
it was found recently that fourier filtering can lead to strong image distortions in the direct space in such a way that the intense spp signal expected at the center of the direct - space lrm images can be completely washed out @xcite .
this is due to interferences arising between propagating spp modes and the residual transmitted light ( here the nv fluorescence ) @xcite . in the present work
, we even found that it was not possible to image the circular rim in the direct space as shown in fig .
3(b ) , but with the fourier filter in place .
this complication is avoided by working in the fourier space .
a similar conclusion was reached recently in experiments aimed at probing the second - order coherence of spps launched by quantum emitters into a metallic film @xcite .
finally , it is worth recalling that high lateral resolution imaging was reported long time ago using the so called apertureless scanning plasmon near - field microscope ( spnm ) based on scanning tunnelling microscopy ( stm ) methods @xcite
. it would be interesting to compare precisely our method with this spnm approach , in particular in view of the recent development of stm methods coupled to lrm @xcite .
in summary , by probing leaky spps launched by a scanning quantum emitter , we have demonstrated a simple method to image a plasmonic device - a circular ring in a thin gold film - with a four times enhanced spatial resolution , i.e. a @xmath3 30 nm resolution .
the method implies working in the fourier space to reconstruct images in the direct space from spp signals that are cleaned from detrimental contributions with short spatial frequencies .
we believe that the approach presented here will be of interest to applications in the emerging field of quantum plasmonics in general , and for local studies of quantum emitters coupled to metal films in particular @xcite .
+ + acknowledgements : we thank g. dantelle and t. gacoin for providing us with nanodiamond samples and jean - franois motte for the optical tip manufacturing and fib milling .
this work was supported by agence nationale de la recherche ( anr ) , france , through the plastips and napho projects . | we demonstrate a simple scheme for high - resolution imaging of nanoplasmonic structures that basically removes most of the resolution limiting allowed light usually transmitted to the far field .
this is achieved by implementing a fourier lens in a near - field scanning optical microscope ( nsom ) operating in the leakage - radiation microscopy ( lrm ) mode .
the method consists of reconstructing optical images solely from the plasmonic ` forbidden ' light collected in the fourier space .
it is demonstrated by using a point - like nanodiamond - based tip that illuminates a thin gold film patterned with a sub - wavelength annular slit .
the reconstructed image of the slit shows a spatial resolution enhanced by a factor @xmath0 compared to nsom images acquired directly in the real space .
99 b. hecht , d. w. pohl , h. heinzelmann , and l. novotny , tunnel near - field optical microscopy : tnom-2 , " ultramicroscopy * 61 * , 99 - 104 ( 1995 ) .
b. hecht , h. bielefeldt , l. novotny , y. inouye , and d. w. pohl , local excitation , scattering , and interference of surface plasmons , " phys .
rev .
lett . * 77 * , 1889 - 1892 ( 1996 ) . c. chicane , t. david , r. quidant , j .- c .
weeber , y. lacroute , e. bourillot , a. dereux , g. colas des francs , and c. girard , imaging the local density of states of optical corrals , " phys .
rev .
lett . * 88 * , 097402 ( 2002 ) .
a. bouhelier , th .
huser , h. tamaru , h .- j .
gntherodt , and d. w. pohl,``plasmon optics of structured silver films , '' phys .
rev .
b * 63 * , 155404 ( 2001 ) .
a. drezet , a. hohenau , a. l. stepanov , h. ditlbacher , b. steinberger , n. galler , f. r. aussenegg , a. leitner , and j. r. krenn , how to erase surface plasmon fringes , " appl .
. lett . * 89 * , 091117 ( 2006 ) .
d. e. chang , a. s. srensen , p. r. hemmer , and m. d. lukin , quantum optics with surface plasmons , " phys .
rev .
lett . * 97 * , 053002 ( 2006 ) .
a. v. akimov , a. mukherjee , c. l. yu , d. e. chang , a. s. zibrov , p. r. hemmer , h. park , and m. d. lukin , generation of single optical plasmons in metallic nanowires coupled to quantum dots , " nature * 450 * , 402 - 406 ( 2007 ) .
s. schietinger , m. barth , t. aichele , and o. benson , plasmon - enhanced single photon emission from a nanoassembled metal - diamond hybrid structure at room temperature .
, " nano lett .
* 9 * , 1694 - 1698 ( 2009 ) .
y. fedutik , v.v .
temnov , o. schps , u. woggon , and m.v .
artemyev , exciton - plasmon - photon conversion in plasmonic nanostructures , " phys .
rev .
lett . * 99 * , 136802 ( 2007 ) .
a. w. schell , g. kewes , t. hanke , a. leitenstorfer , r. bratschitsch , o. benson , and t. aichele , single defect centers in diamond nanocrystals as quantum probes for plasmonic nanostructures , " opt .
express * 19 * , 7914 - 7920 ( 2011 ) .
r. kolesov , b. grotz , g. balasubramanian , r. j. sthr , a. a. l. nicolet , p. r. hemmer , f. jelezko , and j. wrachtrup , wave - particle duality of single surface plasmon polaritons , " nature phys . * 5 * , 470 - 474 ( 2009 ) .
a. huck , s. kumar , a. shakoor , and u. andersen , controlled coupling of a single nitrogen - vacancy center to a silver nanowire , " phys .
rev .
lett * 106 * , 096801 ( 2011 ) .
a. cuche , a. drezet , y. sonnefraud , o. faklaris , f. treussart , j .- f .
roch , and s. huant , near - field optical microscopy with a nanodiamond - based single - photon tip , " opt .
express * 17 * , 19969 - 19980 ( 2009 ) .
a. drezet , a. cuche , and s. huant , near - field microscopy with a single - photon point - like emitter : resolution versus the aperture tip ? ,
" opt .
commun .
* 284 * , 1444 - 1450 ( 2011 ) .
y. sonnefraud , a. cuche , o. faklaris , j .-
p .
boudou , t. sauvage , j .- f .
roch , f. treussart , and s. huant , diamond nanocrystals hosting single nitrogen - vacancy color centers sorted by photon - correlation near - field microscopy .
, " opt . lett . * 33 * , 611 - 613 ( 2008 ) .
a. cuche , o. mollet , a. drezet , and s. huant ,
` deterministic ' quantum plasmonics , " nano lett . *
10 * , 4566 - 4570 ( 2010 ) .
o. mollet , a. cuche , a. drezet , and s. huant , leakage radiation microscopy of surface plasmons launched by a nanodiamond - based tip , " diam .
relat . mater . *
20 * , 995 - 998 ( 2011 ) .
o. mollet , s. huant , g. dantelle , t. gacoin , and a. drezet , quantum plasmonics : second - order coherence of surface plasmons launched by quantum emitters into a metallic film , " phys .
rev .
b * 86 * , 045401 ( 2012 ) .
s. khn , c. hettich , c. schmitt , j .-
p .
poizat , and v. sandoghdar , `` diamond colour centres as a nanoscopic light source for scanning near - field optical microscopy , '' j. microsc . *
202 * , 2 - 6 ( 2001 ) .
r. marty , c. girard , a. arbouet , and g. colas des francs , near - field coupling of a point - like dipolar source with a thin metallic film : implication for stm plasmon excitations , " chem .
. lett . * 532 * , 100 - 105 ( 2012 ) .
a. gruber , a. drabenstedt , c. tietz , l. fleury , j. wrachtrup , and c. von borczyskowski , scanning confocal optical microscopy and magnetic resonance on single defect centers , " science * 276 * , 2012 - 2014 ( 1997 ) .
k. karrai and r. d. grober , piezoelectric tip - sample distance control for near field optical microscopes , " appl .
. lett .
* 60 * , 1842 - 1844 ( 1995 ) .
slight dispersions in the measured rim width are due to an imperfect patterning during the fib milling process . c. girard , o. j. f. martin , g. leveque , g. colas des francs , a. dereux , `` generalized bloch equations for optical interactions in confined geometries , '' chem .
. lett . * 404 * , 44 - 48 ( 2005 ) .
r. marty , a. arbouet , v. paillard , c. girard , and g. colas des francs , photon antibunching in the optical near - field , " phys .
rev .
b * 82 * , 081403 ( 2010 )
. f. i. baida , d. van labeke , a. bouhelier , t. huser , d. pohl , `` propagation and diffraction of locally excited surface plasmons , '' j. opt .
soc . am .
a * 18 * , 1552 - 1561 ( 2001 ) .
l. novotny , b. hecht and d. pohl,``interference of locally excited surface plasmons , '' j. appl .
phys . * 81 * , 1798 - 1806 ( 1997 ) .
a. hohenau , j. r. krenn , a. drezet , o. mollet , s. huant , c. genet , b. stein , and t. w. ebbesen , `` surface plasmon leakage radiation microscopy at the diffraction limit , '' opt .
express * 19 * , 25749 - 25762 ( 2011 ) .
m. specht , j. d. pedarnig , w. m. heckl , and t. w. hnsch , `` scanning plasmon near - field microscope , '' phys .
rev .
lett . *
68 * , 476 - 479 ( 1992 ) .
t. wang , e. boer - duchemin , y. zhang , g. comtet , and g. dujardin , `` excitation of propagating surface plasmons with a scanning tunneling microscope , '' nanotechnology * 22 * , 175201 ( 2011 ) .
p. bharadwaj , a. bouhelier , and l. novotny , `` electrical excitation of surface plasmons , '' phys .
rev .
lett .
* 106 * , 226802 ( 2011 ) . |
the @xmath0-@xmath1 model is one of prototypical examples for understanding strongly correlated electron systems .
the interplay between spin and charge degrees of freedom is a key ingredient of this model . in
spatially one - dimensional ( 1d ) case , this model ( or equivalently the large-@xmath6 hubbard model ) has been fully understood by the exact and numerical approaches @xcite .
therein , the electronic band dispersion splits into the spinon and holon branches , and this is called the spin - charge separation .
they are collective excitations , not the quasiparticles . in 2d case
, it is believed that this model captures anomalous features of hole - doped high-@xmath5 superconductivity in cuprates @xcite .
although it is difficult to solve it exactly , various types of numerical studies are proceeding .
the value of this model as the prototypical and minimal correlated system is still not lost , even though roughly 30 years have passed already after the discovery of high-@xmath5 cuprates . generally speaking ,
the spin and charge degrees of freedom are separated in 1d , while they are interacting with each other in 2d .
however , this statement is sometimes misleading , and for example in the context of the slave boson theory the phase string connects the spinon with the holon excitation @xcite .
the notion of the spin - charge separation and coupling may depend on what we are going to observe and on how to treat the spin and charge degrees of freedom . in this paper
, we would like to revisit the coupling of spin and charge degrees of freedom in 1d .
mathematically , the fundamental operator of the @xmath0-@xmath1 model corresponds to a fermionic particle with no double occupation at each site .
this operator does not obey the exact fermionic anti - commutation relation .
owing to this constraint , the equation of motion for this excitation necessarily induces more non - local spin excitations @xcite , and the equations of motion do not close .
a doped hole into the mott insulator becomes a composite particle dressed with the spin excitation .
this mathematical structure is common to both of 1d and 2d cases , and thus it is natural to consider that this non - locality is related to the phase string in 1d and the spin cloud in 2d .
the purpose of this paper is to examine the spectral properties of such composite excitation modes .
based on the comparison between 1d and 2d cases , we would like to also discuss about the anomalous band structure of high-@xmath5 cuprates . by this approach , it is possible to do more advanced spectroscopic calculation rather than the standard angle - resolved photoemission spectroscopy ( arpes ) . in terms of composite operators ,
the complicated band structures of correlated systems are composed of various types of composite particles .
different branches are represented by different composite operators , respectively . even if the single - particle spectrum is heavily damped by the strong self - energy correction at a particular energy - momentum region
, we may find the composite excitation that is well defined and has a long life time at the region .
in other words , we would like to find the best representation of the collective mode at given energy and momentum .
the organization of this paper is as follows . in the next section ,
we introduce the technical details .
in particular , we explain the physical meaning of composite excitations . in secs .
iii and iv , we present numerical results for 1d and 2d cases .
based on these results , we compare 1d with 2d results , and finally we summarize our paper .
we start with the @xmath0-@xmath1 model .
the hamiltonian is defined by @xmath7 where @xmath8 is the spinor representation of singly - occupied particle at site @xmath9 , @xmath10 with spin @xmath11 , @xmath12 , @xmath13 , and @xmath14 with the identity matrix @xmath15 and the pauli matrices @xmath16 .
this model includes electron hopping , @xmath17 , and the superexchange interaction @xmath1 . here ,
the hopping term between the nearest neighbor pair of sites @xmath9 and @xmath18 is represented as @xmath19 the fourier transform of @xmath20 is given by @xmath21 for the 1d chain , and is also given by @xmath22 for the 2d square lattice .
it is noted that the excitation @xmath23 is not a simple fermion , and satisfy the following algebrae @xmath24 = 2i\epsilon_{\mu\nu\lambda}n_{i,\lambda } , \end{aligned}\ ] ] where @xmath25 .
as we can easily imagine , the first equality necessarily produces spin and charge fluctuation .
as we go to higher order hierarchy of the heisenberg equation of motion for an excitation operator , the fundamental excitation is dressed with such quantum fluctuation .
we consider the heisenberg equation of motion for @xmath26 ( we take @xmath27 ) .
this equation of motion does not close , and induces spatially more extended composite excitations .
the equation for @xmath28 is given by @xmath29 where we define @xmath30 .
then , we find the new excitations @xmath31 and @xmath32 in which a doped hole combines with the nearest neighbor spin fluctuation . when we next consider the equation of motion for these composite states , we find @xmath33 and @xmath34 .
for instance , the important excitations up to @xmath35 for 2d @xmath36 lattice are @xmath37 these excitations have large spectral intensities that are defined as the diagonal parts of @xmath38 @xcite .
thus if we look at the spectra of these composite excitations , we may say the origin of each prominent band structure in the arpes spectrum .
we have confirmed that the excitation originating from the exchange term , @xmath39 , show similar behavior to @xmath40 , and hereafter we do not consider it . in this paper
, we calculate the diagonal part of the following propagator matrix @xmath41 where @xmath42 is the ground state , and @xmath43 is the fourier component of @xmath44 .
the spectra for the composite particles are given by the imaginary part of the retarded propagator matrix @xmath45 for instance , the usual single - particle spectrum is given by @xmath46 .
hereafter , we particulary focus on @xmath47 , @xmath48 , and @xmath49 . in 1d ( 2d ) case , we take @xmath50 ( @xmath51 ) lattice under the periodic boundary condition . in doped cases , we take @xmath52 holes ( @xmath53 electrons ) , and thus the hole doping rate is @xmath54 . the hamiltonian matrix is exactly diagonalized by the lanczos algorithm for the calculation of the propagators for composite excitations .
the propagators are also calculated by the kernel polynomial expansion , and we have confirmed the relevance between these methods . the broadening factor is taken to be @xmath55 . since the momentum resolution of each spatial direction is @xmath56 owing to severe finite - size effect , we naturally interpolate the numerical data that the mesh size becomes @xmath57 . we should note that this procedure does not produce new fine structures originating with lower energy physics .
we just introduce this method so that our data are visible .
) , width=321 ] , width=321 ] , width=321 ] , width=321 ] , width=321 ] , width=321 ] , width=321 ] , width=321 ] and ( b ) the electron removal process by @xmath58.,width=302 ] let us first examine the half - filled case .
we present the angle - resolved spectrum for each component of @xmath59 in figs .
[ fig1a]-[fig1h ] .
the figure [ fig1a ] for @xmath60 corresponds to the standard arpes result .
our numerical result of the arpes spectrum is consistent with the past analytical and numerical results , and we clearly observe the spinon and holon branches with different energy scales @xmath1 and @xmath0 , respectively . based on this consistency , let us particularly focus on the spectra for @xmath40 , @xmath61 , and @xmath58 in the following sentences .
the doped holes in these modes strongly couple with various sizes of the spin strings .
thus , they may not be simple single - particle - type excitations , although they are still fermionic excitations .
we expect that they would play crucial roles on the formation of collective excitations such as spinon and holon excitations . at first
, we examine the spectrum for @xmath62 in fig .
[ fig1b ] .
we find that the main band exactly traces the spinon branch .
this is quite unique , since such a property enables us to do more advanced spectroscopy in the sense that we can select one structure of complicated bands in correlated electron systems .
the reason for such separable spectroscopy is quite simple .
when we remove one electron from the af insulating ground state , the domain boundary ( spinon ) of the af state is created at around the position where the electron is removed .
since the electron removal by the operator @xmath40 counts the change in the near - neighbor spin structure as well as the created hole , the operator necessarily presents the spinon creation [ see fig .
[ fig1i](a ) ] .
therefore , we can regard @xmath40 as _the spinon operator_. note that this is different from the spinon operator @xmath63 in the context of the slave boson representation @xmath64 . on the other hand in figs .
[ fig1d ] and [ fig1h ] , the main bands for @xmath61 and @xmath58 trace the holon branch , and the spectral weight of the spinon branch is not strong .
thus , they are basically _the holon operators_. it is noted that the covering of the holon branch by these types of operators becomes much better as the spin string becomes longer . the readers may wonder why they behave like the holon excitation , although they are dressed with many spins .
the answer may originate in a fact that the spin string attached to @xmath61 and @xmath58 can regulate the spin structure after the photoemission so that the resulting final state has just one hole .
let us consider the situation in which they are operated to the af - like state , but the state contains one spinon bases owing to quantum fluctuation .
then , the operation of @xmath61 and @xmath58 to the state can remove the spinon as well as creation of the hole [ see fig . [ fig1i](b ) ] .
these results suggest that each branch in the arpes data originates from a particular composite excitation .
interestingly , these spectral intensities are much stronger than that of the single - particle spectrum , and there is only weak continuum part .
this weak continuum feature really represents that these composite excitations are more important than the simple @xmath65 state . in comparison with @xmath40 , @xmath61 , and @xmath58 , the spectra for @xmath66 , @xmath67 , @xmath68 , and @xmath69 as shown in figs .
[ fig1c ] , [ fig1e ] , [ fig1f ] , and [ fig1 g ] are more scattered , though a considerable amount of spectral intensity is still on the holon branch at @xmath70 . for these excitations ,
the multiple spin excitations are away from a created hole by the photoemission , thus they seem to be somehow different from @xmath40 , @xmath61 , and @xmath58 . ) , width=321 ] , width=321 ] , width=321 ] , width=321 ] , width=321 ] , width=321 ] , width=321 ] , width=321 ] next we examine the doped case in figs .
[ fig2a]-[fig2h ] . in this case
also , the important composite excitations near the fermi level are @xmath40 , @xmath61 , and @xmath58 .
we find that the spectral intensity concentrates on @xmath2 and @xmath71 as the doped hole is dressed with longer spin string . at this momentum ,
the spinon and holon branches merge together .
thus , the angle - resolved spectra for @xmath61 and @xmath58 have enough information for real electronic excitation as well as the phase string that connects spinon with holon . in the standard arpes spectrum ( fig . [ fig2a ] ) , we observe the metallic branch showing the tomonaga - luttinger liquid behavior .
the spectral weight above the fermi level originates in the holon branch at @xmath72 .
however , for the composite excitations @xmath40 , @xmath61 , and @xmath58 , the spectral intensity of this branch tends to be vanishing above the fermi level . in particular ,
the spectral intensity for @xmath61 and @xmath58 is strong only at the fermi level . on the other hand ,
the spectrum for @xmath66 has considerable weight above the fermi level .
therefore , the composite excitations behave quite differently from the fundamental excitation @xmath60 . although we observe single metallic band in the arpes spectrum , these results suggest that the character of the wave function changes at the fermi level .
this is because for instance @xmath40 corresponds to the self - energy correction to @xmath73 in the standard field - theoretical terminology , and this means that the damping mechanism of the fundamental excitation strongly depends on the momentum .
we will later discuss about this feature again in comparison with the presence of the pseudogap in 2d cases and high-@xmath5 cuprates .
the spectra for @xmath66 , @xmath67 , @xmath68 , and @xmath69 seem to contain different information . actually , we observe relatively strong spectral intensity at @xmath74 and @xmath75 .
this shadow band seems to naturally connect to the holon branch at @xmath70 , and intersects with the metallic band at around @xmath2 .
owing to this intersection , the band seems to split into two substructures .
later , we will also discuss more about this intersection in connection with the presence of the pseudogap in high-@xmath5 cuprates .
) at half - filling , width=321 ] at half - filling , width=321 ] at half - filling , width=321 ] at half - filling , width=321 ] ) for the two - hole state , width=321 ] for the two - hole state , width=321 ] for the two - hole state , width=321 ] for the two - hole state , width=321 ] let us move to discuss the band structure in 2d case .
we examine the hole doped case ( @xmath54 ) shown in figs .
[ fig3e]-[fig3h ] . for comparison
, we also show undoped case in figs .
[ fig3a]-[fig3d ]
. we should be careful for the fact that the numerical data may involve strong finite - size effects .
the excitations @xmath76 wrap the overall lattice sites , and their motions are highly restricted .
thus , we only show the spectral data for @xmath73 , @xmath40 , @xmath66 , and @xmath61 .
we examine the spectrum for @xmath60 .
the data are shown in fig .
[ fig3e ] . in the figure ,
the fermi level is located at the origin of the energy @xmath77 . comparing this with refs .
@xcite , we see that the result is reasonable .
the non - interacting band is highly renormalized , and the band width becomes roughly half . furthermore ,
when we look at the spectrum near the fermi level , we find that the weight at @xmath78 and @xmath79 is very strong .
when we look at the branch from @xmath3 to @xmath4 through @xmath79 , we find that the weight just above the fermi level is once reduced at around @xmath79 .
this tendency seems to be consistent with the presence of the fermi arc .
we have also confirmed that the gap clearly opens at this momentum by introducing long - range hopping @xmath80 and @xmath81 . according to ref .
@xcite , the pseudogap should be also seen just above the fermi level at around @xmath82 ( this gap is not seen experimentally , since the gap is located at the electron addition side ) .
the branch from @xmath3 to @xmath4 through @xmath82 suffers from the presence of the gap owing to our finite lattice system , and thus unfortunately we can not determine whether the gap exists .
it is also noted that there is a considerable amount of broad spectral weight at around @xmath83 and @xmath84 .
we also examine the spectra for @xmath40 , @xmath66 , and @xmath61 in figs .
[ fig3f ] , [ fig3 g ] , and [ fig3h ] , respectively .
we find that @xmath40 and @xmath61 have strong spectral intensity near the fermi level , while @xmath66 does not have the intensity near the fermi level .
thus , the strong coupling between the doped hole and the spin cloud is crucial for the intensity even in 2d cases .
furthermore , the intensity near the fermi level increases as the doped hole is dressed with larger and larger spin cloud .
all of the excitations do not have the remarkable spectral intensity above the fermi level .
this feature is also roughly consistent with 1d case , and strongly suggests that the character of the wave function changes at the fermi level .
it would be quite meaningful to compare the 1d results with the 2d results along @xmath3-@xmath4 direction through @xmath85 @xcite .
for instance , when we look at the spectra for @xmath40 in 1d and 2d cases , we see that they are quite similar with each other .
this observation suggests that the large spectral intensity at @xmath86 and @xmath87 in 2d originates in the spinon - like excitation observed in 1d .
originally , if we do not consider the spin fluctuation , there is no spectral weight at around @xmath83 and @xmath88 .
thus , this is quite reasonable .
however , if we look at @xmath61 and @xmath58 , the intensity at @xmath89 and @xmath88 in 2d is much enhanced than that in 1d .
this may be due to dimensionality dependence on the coupling strength between spin and charge degrees of freedom .
since the 1d metallic band originates in the holon branch before doping , in 2d case also the weight at around @xmath84 may be pushed above the fermi level by the hole doping .
this scenario is actually consistent with our conclusion that the character of the wave function changes at the fermi level .
if the dimensionality dependence is less pronounced on the spectra of composite particles , we may say that the presence of the pseudogap in the 2d @xmath0-@xmath1 model and high-@xmath5 cuprates can be understood by considering the vanishing spectral weights of the composite excitation above the fermi level in 1d cases .
i believe that this observation is a strong indication that the pseudogap is different from the superconducting gap .
however , this is still highly hypothesized conjecture , and thus we need to examine more about the vanishing spectral weight above the fermi level in 1d and 2d in more physical standpoints .
summarizing , we have examined spectral properties of composite excitations in the 1d and 2d @xmath0-@xmath1 models .
the most important goal of this numerical work is that these composite hole excitations well characterize prominent band structures in the angle - resolved photoemission spectrum .
therefore , it is possible to do more advanced spectroscopy rather than the standard arpes spectrum . in 1d case
, we have identified that the spinon and holon operators are represented as @xmath40 , @xmath61 , and @xmath58 . after doping to the mott insulator ,
the spectral weight of @xmath61 and @xmath58 concentrates on the fermi level , and they are physical excitations . we have also found that the 1d composite data are quite similar to those in 2d except for relatively large spectral intensity of 2d at @xmath89 and @xmath88 . in both of 1d and 2d cases ,
we have found that the band dispersions above and below the fermi level have different characters , respectively .
we have proposed that this feature would provide us quite important information on the formation of the pseudogap in high-@xmath5 cuprates .
hm acknowledges masaki fujita and his co - workers for valuable discussion .
99 m. ogata and h. shiba , phys .
b * 41 * , 2326 ( 1990 ) .
k. penc , k. hallberg , f. mila , and h. shiba , phys .
b * 55 * , 15475 ( 1997 ) ; phys . rev .
b * 55 * , r4859 ( 1997 ) ; phys . rev
. lett . * 77 * , 1390 ( 1996 ) ; k. penc , f. mila , and h. shiba , phys .
. lett . * 75 * , 894 ( 1995 ) .
s. maekawa , t. tohyama , s. e. barnes , s. ishihara , w. koshibae , and g. khaliullin , `` _ physics of transition metal oxides _ '' , springer ( 2004 ) .
h. suzuura and n. nagaosa , phys .
b * 56 * , 3548 ( 1997 ) .
h. matsumoto , m. sasaki , s. ishihara , and m. tachiki , phys .
b * 46 * , 3009 ( 1992 ) ; m. sasaki , h. matsumoto , and m. tachiki , phys . rev .
b * 46 * , 3022 ( 1992 ) .
h. matsueda and t. tohyama , in preparation .
t. tohyama , phys .
b * 70 * , 174517 ( 2004 ) .
m. kohno , phys .
105 * , 106402 ( 2010 ) ; phys .
lett . * 108 * , 076401 ( 2012 ) ; phys . rev .
b * 90 * , 035111 ( 2014 ) . | we examine spectral properties of doped holes dressed with surrounding spin cloud in the @xmath0-@xmath1 model .
these composite - hole excitations well characterize prominent band structures in the angle - resolved photoemission spectrum .
in one - dimensional ( 1d ) case at half - filling , we identify the composite operators that separately pick up the spinon and holon branches , respectively .
after hole doping , we find that the composite hole excitations with string - like spins tend to be localized at @xmath2 in the momentum space .
this means that such composite excitations should be actual electronic excitations , since the spinon and holon branches merge together at this momentum . in 2d case
, we find that the composite excitations with more non - local spin fluctuation have stronger intensity near the fermi level .
the composite band structure along diagonal @xmath3-@xmath4 direction in 2d has some similarity to that in 1d , and such non - local spin fluctuation plays an important role on the formation of the pseudogap in high-@xmath5 cuprates . |
Subsets and Splits
Data Mining Abstracts
This query searches for and retrieves 2000 documents containing the phrase 'data mining' in the abstract, providing limited but specific document samples.